Quantifying the Roman Economy: Methods and Problems [1 ed.] 0199562598, 9780199562596

Citation preview

OXFORD STUDIES ON THE ROMAN ECONOMY General Editors

Alan Bowman

Andrew Wilson

OXFORD STUDIES ON THE ROMAN ECONOMY This innovative monograph series reXects a vigorous revival of interest in the ancient economy, focusing on the Mediterranean world under Roman rule (c.100 bc to ad 350). Carefully quantiWed archaeological and documentary data will be integrated to help ancient historians, economic historians, and archaeologists think about economic behaviour collectively rather than from separate perspectives. The volumes will include a substantial comparative element and thus be of interest to historians of other periods and places.

Quantifying the Roman Economy Methods and Problems

Edited by

A L A N B OW M A N and A N D R EW W I L S O N

1

3

Great Clarendon Street, Oxford ox2 6dp Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide in Oxford New York Auckland Cape Town Dar es Salaam Hong Kong Karachi Kuala Lumpur Madrid Melbourne Mexico City Nairobi New Delhi Shanghai Taipei Toronto With oYces in Argentina Austria Brazil Chile Czech Republic France Greece Guatemala Hungary Italy Japan Poland Portugal Singapore South Korea Switzerland Thailand Turkey Ukraine Vietnam Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries Published in the United States by Oxford University Press Inc., New York # Oxford University Press 2009 The moral rights of the authors have been asserted Database right Oxford University Press (maker) First published 2009 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this book in any other binding or cover and you must impose the same condition on any acquirer British Library Cataloguing in Publication Data Data available Library of Congress Cataloging in Publication Data Data available Typeset by SPI Publisher Services, Pondicherry, India Printed in Great Britain on acid-free paper by the MPG Books Group in the UK ISBN 978–0–19–956259–6 1 3 5 7 9 10 8 6 4 2

Preface This volume is the Wrst of a series of monographs to be published under the general title Oxford Studies on the Roman Economy. It has its origin in a research programme directed by the Series Editors and funded by the Arts and Humanities Research Council for Wve years (2005–10). This project, entitled The Economy of the Roman Empire: Integration, Growth and Decline, starts from the realization that further debate on the Roman economy needs to be grounded in a better understanding of the evidence and that despite the lack of statistics in ancient written sources there is nevertheless a vast mass of archaeological and documentary (chieXy but not exclusively papyrological) data awaiting synthesis which could provide illuminating pointers. The project will attempt a detailed analysis of major areas of the economy where quantiWable bodies of archaeological and documentary data can be identiWed and compared. It will focus therefore on population and urbanization, agriculture, trade and commerce, and mining and coinage, examining for each a series of performance indicators which could track variation across space and time, and might enable a comparison of performance against the economies of other periods and cultures. Geographically the project will draw on material from across the entire empire, although Egypt (because of the papyrological evidence) will be a particular focus for the documentary record. Chronologically we will cover the period 100 bc to c. ad 350, aiming to test and quantify the general impression of economic growth to c. ad 200, and assess to what extent the third century might have been marked by stagnation, contraction, or decline; and how far there may have been some sort of recovery in the early fourth century. Besides assessing the rate and volume of economic growth and subsequent contraction we will also consider how far the Roman economy may have been integrated across the whole empire rather than acting as a set of loosely linked regional economies; and the interaction between state mechanisms of economic control and local ‘free market’ behaviour. Were the two part of a coherent economic

vi

Preface

‘system’? These general questions need to be approached through several underlying issues which include: the drivers and symptoms of ancient economic growth; levels of rationality in ancient economic behaviour; and the appropriateness of applying recent analysis of economic institutions and behaviour to the Roman economy. These will be examined through the analysis of a limited number of signiWcant bodies of published direct and proxy data, archaeological and documentary (particularly Greek papyri), within the four main diagnostic areas. SpeciWcally, we will consider the intensity of agricultural and manufacturing production, levels of supply and demand, the movement of goods over distance, the distribution of population and the eVects of urbanization, scale of the metal and money supply in relation to the behaviour of currency and prices, institutional control and free markets, and the application of developments in ancient technology. These topics will be discussed more exhaustively in future volumes which will be generated by a series of annual conferences, including four on the main diagnostic areas of analysis (population and urbanization; agriculture; trade and commerce; and mining and metal supply); this element is crucial because it allows us to involve a much wider range of perspectives and expertise, and provides opportunities for testing and feedback on our approaches. As the research progresses, we intend to make available on our website (http://oxrep.class.ox.ac. uk) a series of working papers, and eventually also the datasets which we and the project Research Assistants, Dr Myrto Malouta, Dr Annalisa Marzano, Dr Dario Nappo, and Dr Hannah Friedman, have collated, or which have been oVered by collaborators. As part of the project a doctoral student, Benjamin Russell, is writing his D.Phil. thesis on the economy of stone sculpture, focusing on quarrying, architectural production, sarcophagi, and honoriWc portrait statues. The Wnal objective of this programme is a volume in which we will attempt to summarize and synthesize the results of the research. As a preliminary, it is obviously essential to frame the key concepts and questions carefully and precisely, and to consider issues of approach and methodology. This is what the present volume sets out to do both at a general level and by illustrative analysis of material drawn from the four diagnostic areas. The contributions of the various authors originally took the form of presentations at a

Preface

vii

colloquium held at the Ashmolean Museum and the Classics Centre in Oxford in September 2006. The programme consisted of a number of main papers, to which brief responses were made. Following this, there were extended discussions on each of the key areas. This format explains the shape and content of the present volume, in which the main presentations (Lo Cascio, Fentress, Bowman, Wilson, Ponting, Rathbone) and most of the responses and the prefaces to the discussions (by Bagnall, Jongman, Mattingly, Fulford, Harris, Hitchner, Howgego, and Scheidel) are all represented. Some of the other respondents and discussants did not feel able to make a substantive written contribution to the volume but have kindly allowed us to integrate their observations or comments where appropriate. We hope that this strategy has given the volume a clear focus and made it more than a disparate conglomeration of conference papers. We are grateful to the AHRC for its Wnancial support of the project, to the staV at the Classics Centre and the Ashmolean Museum, Oxford, for their assistance with arrangements for the colloquium out of which this volume arises, and to the various contributors to the colloquium, both speakers and members of the audience who contributed points in discussion at the time or subsequently gave us feedback, especially Chris Wickham, Michael McCormick, A. J. Parker, and Ben Russell. Catherine Hardman of the Archaeology Data Service and Damian Robinson both gave valuable advice on IT and data archiving matters at the planning stage of the project, while Mark Austin has constructed the website, with initial input from Neil Leeder and Ben Chad. Julia Strauss kindly made available to us the data on Eastern Mediterranean shipwrecks she had collected in the course of her doctoral thesis at University College, London. Neither the organization of the conference nor the preparation of this volume could have been achieved without the hard work and dedication of the project’s Research Assistants, at the time Myrto Malouta and Annalisa Marzano. A.K.B. A.I.W. December 2007

Contents List of Contributors List of Figures List of Tables

xi xv xviii

INTRODUCTION: APPROACHES 1. Quantifying the Roman Economy: Integration, Growth, Decline? Alan Bowman and Andrew Wilson

3

I URBANIZATION 2. Urbanization as a Proxy of Demographic and Economic Growth Elio Lo Cascio 3. Response to Elio Lo Cascio Roger Bagnall

87 107

II FIELD SURVEY AND DEMOGRAPHY 4. Archaeology, Demography, and Roman Economic Growth Willem Jongman 5. Peopling the Countryside: Roman Demography in the Albegna Valley and Jerba Elizabeth Fentress 6. Peopling Ancient Landscapes: Potential and Problems David Mattingly

115

127 163

Contents

ix

III AGRICULTURE 7. Quantifying Egyptian Agriculture Alan Bowman 8. Response to Alan Bowman Roger Bagnall

177 205

IV TRA DE 9. Approaches to Quantifying Roman Trade Andrew Wilson 10. Approaches to Quantifying Roman Trade: Response Michael Fulford 11. A Comment on Andrew Wilson: ‘Approaches to Quantifying Roman Trade’ William Harris

213 250

259

V C OI NAGE 12. Roman Silver Coinage: Mints, Metallurgy, and Production Matthew Ponting 13. Coinage and Metal Supply Bruce Hitchner 14. Some Numismatic Approaches to Quantifying the Roman Economy Christopher Howgego

269 281

287

VI PRICES, EARNING S, AND STANDARDS OF LIVING 15. Earnings and Costs: Living Standards and the Roman Economy Dominic Rathbone

299

x

Contents

16. How Prosperous were the Romans? Evidence from Diodetian’s Price Edict (ad 301) Robert Allen 17. New Ways of Studying Incomes in the Roman Economy Walter Scheidel Index

327

346

353

List of Contributors Robert Allen, Professor of Economic History at Oxford University and Fellow of NuYeld College. He has written extensively on English agricultural history and is currently researching the global history of wages and prices, pre-industrial living standards around the world. The British Industrial Revolution in Global Perspective will be published in 2009. Roger Bagnall, Professor of Ancient History and Director of the Institute for the Study of the Ancient World, New York University. His research centres on the social, economic, and administrative history of the Hellenistic and Roman Near East, especially Egypt, and papyrology. Publications include Egypt in Late Antiquity (Princeton 1993) and (ed.), Egypt in the Byzantine World (Cambridge 2007). Alan Bowman, Camden Professor of Ancient History, University of Oxford. His areas of research are the social and economic history of the Roman empire, Ptolemaic and Roman Egypt, papyrology, and the Vindolanda writing-tablets. Publications include Agriculture in Egypt from Pharaonic to Modern Times (ed. with E. Rogan, London 1999), Life and Letters on the Roman Frontier: Vindolanda and its People (London, revised edition 2003). Elizabeth Fentress, independent scholar. Her research is on the archaeology of Italy and North Africa in the Punic, Roman, and early medieval periods. Publications include The Berbers (Oxford 1996, with Michael Brett), Cosa V, an Intermittent Town (Ann Arbor 2004), An Island through Time: Jerba Studies, vol. 1: The Punic and Roman Periods (JRA Suppl. 70, 2009, with A. Drine and R. Holod). Michael Fulford, Professor of Archaeology, University of Reading. His areas of research are Roman urbanism, economy, trade, and technology. Recent publications include Life and Labour in Late Roman Silchester: Excavations in Insula IX since 1997 (London

xii

List of Contributors

2006), Iron Age and Romano-British Settlements and Landscapes of Salisbury Plain (Salisbury 2006). William Harris, Shepherd Professor of History, Columbia University. He has written widely about the Roman economy and is the editor of The Monetary Systems of the Greeks and Romans (Oxford 2008). Bruce Hitchner, Professor of Classics and International Relations, Tufts University. His interests lie chieXy in the Roman economy, and the history and archaeology of Roman Africa and Gaul. Recent publications include: ‘Globalization avant la lettre: Globalization and the History of the Roman Empire’, New Global Studies 2 (2008), article 2 (http://www.bepress.com/ngs/vol2/iss2/art2); ‘The Mediterranean and the History of Antiquity’, in A. Erskine (ed.), Blackwell Companion to Ancient History (2008); ‘ ‘‘The fairest part of the earth’’: The evidence for intensive growth in the Roman Empire’, in J. Manning and I. Morris (eds.), The Ancient Economy: Evidence and Models (Stanford 2005), 207–22. Christopher Howgego, Keeper of the Heberden Coin Room in the Ashmolean Museum, Professor of Greek and Roman Numismatics in the University of Oxford, and a Fellow of Wolfson College. He directs the Roman Provincial Coinage in the Antonine Period Project (http:// rpc.ashmus.ox.ac.uk/), is author of Ancient History from Coins (London 1995) and Greek Imperial Countermarks: Studies in the Provincial Coinage of the Roman Empire (London 1985) and joint editor of Coinage and Identity in the Roman Provinces (Oxford 2005). Willem Jongman, Reader in Ancient History, University of Groningen. Areas of research include Roman urbanism and the measurement of Roman economic performance. Publications include The Economy and Society of Pompeii (Amsterdam 1988) and ‘The early Roman Empire: Consumption’, in W. Scheidel, I. Morris, and R. P. Saller (eds.), The Cambridge Economic History of the Greco-Roman World (Cambridge 2007), 592–618. Elio Lo Cascio, Professor of Roman History, Universita` di Roma ‘La Sapienza’. His published work primarily focuses on the administrative history of the Principate and of the Late Empire, the economic and social history of Rome, and Roman demography. Publications

List of Contributors

xiii

include Il princeps e il suo impero: Studi di storia amministrativa e Wnanziaria romana (Bari 2000); and the edited volumes Production and Public Powers in Antiquity (Cambridge 2000, with D. W. Rathbone) and Innovazione tecnica e progresso economico nel mondo romano (Bari 2006). David Mattingly, Professor of Roman Archaeology, University of Leicester. His main areas of research are rural settlement, farming technology and the economy, and urbanism in Roman Africa, olive cultivation in the Roman world, rural Weld survey in Italy, Libya and Jordan, and Roman imperialism. Major publications include Tripolitania (London 1995), The Archaeology of Fazzan: Volumes 1–2 (with C. M. Daniels, J. N. Dore, D. Edwards, and J. Hawthorne, London 2003 and 2007), An Imperial Possession: Britain in the Roman Empire (Penguin History of Britain Series, London 2006/7). Matthew Ponting, Lecturer in Science-Based Archaeology, School of Archaeology, Classics and Egyptology, University of Liverpool. His areas of research include the scientiWc investigation of Roman Imperial silver coinage, the origins of coinage in Palestine, aesthetic and technological choices in Roman military metalwork and the development of copper-alloy technology and usage in the Roman and Byzantine Near East. Recent publications include: K. Butcher and M. J. Ponting, ‘The Roman denarius under the Julio-Claudian emperors: Mints, metallurgy and technology’, Oxford Journal of Archaeology 24.2 (2005), 163–97, ‘The Egyptian billon tetradrachm under the Julio-Claudian emperors—Fiduciary or intrinsic?’, Schweizerische Numismatische Rundschau/Revue Suisse de Numismatique 84 (2005), 93–124. Dominic Rathbone, Professor of Ancient History at King’s College London. His main areas of research are in the economy of the Roman world especially Roman Egypt, the agrarian history of Roman Italy, and the Wnances of the Roman state. His main publications are Economic Rationalism and Rural Society in Third-Century ad Egypt: The Heroninos Archive and the Appianus Estate (Cambridge 1991) and Egypt from Alexander to the Copts: An Archaeological and Historical Guide (ed. with R. S. Bagnall, London 2004).

xiv

List of Contributors

Walter Scheidel, Professor of Classics and, by courtesy, History, Stanford University. His research focuses on ancient social and economic history, premodern historical demography, and comparative and transdisciplinary world history. He was the lead editor of The Cambridge Economic History of the Greco-Roman World (Cambridge 2007). Andrew Wilson, Professor of the Archaeology of the Roman Empire, University of Oxford. His research interests include the economy of the Roman empire; ancient technology; ancient water supply and usage; Roman North Africa, and archaeological Weld survey. Recent publications include: ‘Economy and trade’, in E. Bispham (ed.), The Short Oxford History of Europe, Vol. 2: Roman Europe (Oxford 2008), 170–202; chapters on ‘Hydraulic Engineering’, ‘Machines’, and ‘Large-scale manufacturing, standardization, and trade’, in J. P. Oleson (ed.), Handbook of Engineering and Technology in the Classical World (Oxford 2008); and ‘The Metal Supply of the Roman Empire’, in E. Papi (ed.), Supplying Rome and the Roman Empire (JRA Suppl. 69; Portsmouth, RI, 2007), 109–25.

List of Figures 1.1 Number of attestations of water-mills by century in archaeological Wnds, ancient representations, and literary sources, as known in 1984 (data in Wikander 1984)

35

1.2 Number of attestations of water-mills by century in archaeological Wnds, ancient representations, and literary sources, as known in 2006

35

1.3 Number of attestations of water-lifting devices by century in archaeological Wnds and literary sources, as known in 2006

37

1.4 Chronological distribution of surviving copies of literary texts on papyrus and parchment (Leuven Database of Ancient Books)

42

1.5 Roman slag heaps at lead and silver mines near Stojnic, Kosmaj (Serbia) (photo: A. Wilson)

66

5.1 Albegna Valley and Jerba Surveys, sampling designs

130

5.2 Location of the island of Jerba

132

5.3 Early Roman sites on Jerba

134

5.4 Reconstruction of an insula at Meninx

136

5.5 Distribution of population in Jerba, Early Roman period

140

5.6 The colony of Cosa c.150 bc: reconstruction

142

5.7 Cosa in the Augustan period

143

5.8 Albegna Valley: population minima and maxima

145

5.9 Centuria in the territory of Saturnia: internal divisions hypothetical

146

5.10 Albegna Valley: distribution of population, 50 bc–ad 100

147

5.11 Albegna Valley: the distribution of population over time

149

5.12 Vulci/Tarquinia coast (Augustan high 31/km2)

152

5.13 Caere (Augustan high 34/km2)

152

5.14 Jerba: distribution of population

153

5.15 ARS at Meninx and B urg u

154

5.16 B urg u and Meninx percentage Wneware/year

155

xvi

List of Figures

5.17 Jerba, villas and villages between the Early Punic and Late Antique periods

157

7.1 Egypt in the Early Roman period

180

9.1 Olive and wine presses in the suburbium of Rome (Corrente 1985, Wg. 85)

216

9.2 Mediterranean shipwrecks by century, using mid-points of date ranges (Parker 1992a, Wg. 3)

220

9.3 Mediterranean shipwrecks by century, using probability per annum (data from Parker 1992a)

223

9.4 Mediterranean shipwrecks by half-century, using probability per annum (data from Parker 1992a)

223

9.5 Mediterranean shipwrecks by quarter-century, using probability per annum (data from Parker 1992a)

225

9.6 Mediterranean shipwrecks by 20-year period, using probability per annum (data from Parker 1992a)

225

9.7 Percentage of Italian wine amphorae, provincial wine amphorae, and amphorae for other products, in selected contexts from Rome and Ostia, 50 bc to ad 600 (Panella and Tchernia 1994, Graphique 2)

230

9.8 Percentage of Italian wine amphorae from diVerent regions, and provincial wine amphorae, in selected contexts from Rome and Ostia, 50 bc to ad 600 (Panella and Tchernia 1994, Graphique 3)

231

9.9 Fish-salting vat capacities by half-century—all regions (36 factories at 23 diVerent sites) (Wilson 2006, Wg. 2)

235

9.10 Fish-salting vat capacities by half-century—southern Iberia and North Africa (28 factories at 14 diVerent sites) (Wilson 2004, Wg. 3)

236

9.11 Fish-salting vat capacities by half-century—Brittany (6 factories at 5 diVerent sites) (Wilson 2006, Wg. 4)

237

9.12 Graph of even frequencies for each ARS form compared to the mean derived from the dates of each form (Fentress et al. 2004, Wg. 11.2)

238

9.13 Comparison of ARS production and public building dedications in North Africa (Fentress and Perkins 1988, Wg. 4)

240

9.14 Jerba: percentage of ARS recovered over time (Fentress et al. 2004, Wg. 11.11)

241

List of Figures

xvii

9.15 UNESCO Libyan Valleys Survey: percentage of ARS recovered over time (Fentress et al. 2004, Wg. 11.13)

241

9.16 Distribution over time of Italian Terra Sigillata exported to the eastern Mediterranean (Poblome 2004, Wg. 6)

242

12.1 Scanning electron micrograph of a section through a post-reform denarius of Nero

271

16.1 Silver wages (Allen et al. 2007)

331

16.2 Welfare ratios with European respectability basket (Allen et al. 2007)

338

16.3 Welfare ratios with bare bones baskets (Allen et al. 2007)

342

17.1 Model of population regulation driven by exogenous change in the population growth rate (Lee 1986)

347

List of Tables 1.1 Estimates of remaining slag, and quantities of lead and silver extracted in the Roman period from the Kosmaj region in Upper Moesia

67

5.1 Population estimates for Early Roman Jerba

137

5.2 Population estimates for the Albegna Valley 50 bc–ad 100

144

9.1 Contents of the amphorae from the Grado wreck (2nd century ad) 233 15.1 Wheat prices

304

15.2 Pay in the Roman army

311

15.3 Living allowances and remuneration

315

15.4 Purchasing power of 1 lb gold

318

16.1 Northern European respectability basket

333

16.2 Mediterranean respectability basket

334

16.3 Bare bones subsistence baskets

341

17.1 Standardized index of real wages in Roman Egypt (2nd century ad ¼ 100)

348

17.2 Proxy variables for assessing the probability of real income growth in Roman Italy

348

17.3 Regional variation in real slave prices in the Graeco-Roman Mediterranean

350

17.4 Annual salaries in the Roman and Han empires

351

Introduction Approaches

This page intentionally left blank

1 Quantifying the Roman Economy: Integration, Growth, Decline? Alan Bowman and Andrew Wilson

I. INTRODUCTION The programme of research which this volume introduces is concerned with the fundamentals of the Roman imperial economy and will attempt detailed analysis of major economic activities (agriculture, trade, commerce, mining), utilizing quantiWable bodies of artefactual and documentary evidence and placing them in the broader structural context of regional variation, distribution, size and nature of markets, supply and demand. The chronological parameters are 100 bc–c. ad 350, covering the period of greatest imperial expansion and economic growth (to c. ad 200), followed by a century conventionally perceived as one of contraction or decline.1 Geographically, we will draw on material selected from all over the Mediterranean world: Egypt, North Africa, Spain, and Italy will be our most fruitful sources of data, which will be gleaned almost entirely from published archaeological and documentary sources. The project will provide a detailed basis for assessing the rate and volume of economic growth in the earlier empire and the extent of contraction after ad 200. We will also consider to what extent the Roman economy was integrated 1 We had originally speciWed the terminal date as c. ad 300, but we are persuaded that the quantity and quality of the evidence from the Wrst half of the fourth century makes it perverse to exclude this period arbitrarily. The dates in any case are only approximate.

4

Alan Bowman and Andrew Wilson

across the whole empire rather than a number of disparate but linked regional economies; and how far it was structurally integrated—that is, were state mechanisms of economic control and the behaviour of ‘free market’ local economies part of a coherent economic ‘system’? In addressing these general questions we will focus on several underlying issues which will, in combination, oVer a set of carefully framed deWnitions. These include: the constituents of economic growth; levels of sophistication and rationality in ancient economic strategies; inter-operation of regional markets; the extent to which recent analysis of economic institutions and behaviour is applicable to the Roman economy, in terms of intensity of production in agriculture and manufacture, markets, supply and demand, movement of goods over what distances, distribution of population and the eVects of urbanization, behaviour of currency and prices, institutional control and free markets, and the application of developments in ancient technology which we believe to be much more signiWcant than is generally recognized. Not long after we began to compile the material for the present volume, The Cambridge Economic History of the Greco-Roman World was published.2 Chapters 19–27 of this work cover the period on which we focus, from the late Republic to the later empire, with a combination of chronological, thematic, and regional treatments. As the work of many hands, this volume does not provide a single, uniWed vision of the economic history of the Mediterranean and north-west Europe under Roman rule and it does not conceal diVerences of emphasis or interpretation between the authors. But it does provide an excellent context in which to ground our own quantitative approach and it relieves us of the need to attempt or duplicate surveys of this sort. Furthermore, the Introduction (ch. 1) sketches the outlines of debate about the nature of the ancient economy and chapters 2–6, grouped under the rubric ‘Determinants of Economic Performance’ analyse many of the topics which we discuss in this volume (demography, urbanization, the agrarian regime, technology). Along the way, we are presented with a plethora of detail and with both old and new concepts and tools (such as ‘New Institutional Economics’) for analysing and interpreting it. In 2 Scheidel, Morris, and Saller 2007.

Quantifying the Roman Economy

5

summarizing what they believe they have achieved, the editors conclude their introduction thus (pp. 11–12): This emerging account of the Greco-Roman economy, we believe, is an advance over twentieth-century interpretations. It improves on substantivist approaches by providing crude statistics on economic performance, but it also goes beyond both sides in the old primitivist–modernist debate by developing general theoretical models of ancient economic behaviour and putting them in a global, comparative context. It recognizes that classical antiquity saw one of the strongest economic eZorescences in premodern history but keeps this in perspective, refusing to confuse the ancient economy with the modern. In short, it takes seriously Douglass North’s injunction to explain the structure and performance of economies through time.

This is not the place to examine the validity of this claim, which will undoubtedly be scrutinized in detail by reviewers, but a few comments are in order. The editors identify some long-term trends some of which will not provoke serious disagreement (pp. 9–11): that there was a signiWcant rise in per capita consumption over the period covered by the volume; that the population of the Mediterranean classical world greatly increased (perhaps doubled) over the period between c.800 bc and ad 200; that there was clearly signiWcant economic growth over this period. Underlying these general statements there are, however, some uncertainties which we intend to address in more detail in this and in subsequent volumes. The statistics they oVer are useful to us but in many ways, as they admit, crude. The population increase from 20 to 40 million over a millennium is posited with an assurance we do not share and conceals many major uncertainties. Per capita estimates for income and consumption are, we think, capable of further reWnement, at least for some regions and periods. We oVer some suggestions for ways in which this might be done. Following their lead more positively, we are certainly fully aware of the desirability of assessing the performance of the Roman economy in a comparative context and we certainly take seriously the need to address the structure and performance of economies over time (on all this more below). The work will be characterized throughout by commitment to integrate or correlate archaeological and documentary evidence

6

Alan Bowman and Andrew Wilson

where possible—the chosen topics are not all equally represented but there is suYcient complementarity to make this attempt worthwhile. Our chosen method is to take four key subject areas, widely recognized as diagnostic for the Roman imperial economy, for which we can compile, update, and exploit quantiWable data and rigorously analyse and deWne what each subject area can contribute to our key themes and questions. A crucial Wrst phase will be to select the most appropriate datasets and test their potential. The data compilation will comprise inter alia aggregation of speciWc features and assemblages from published archaeological reports, documentary evidence for agricultural yields, prices of commodities and labour, census data and units of habitation, and capacities of technological ‘devices’ (e.g. oil-presses, water-mills). Much of the work will consist of revising, updating, expanding, and synthesizing existing published datasets but we emphasize that none of these has been fully exploited by the potential of modern computing power. The potential of the data to generate, for example, comparative distribution patterns of ceramics or changes in the conWguration of land tenure and use can be tested. We have identiWed four key subject areas to focus on where there is useful evidence: demography and urbanization; agriculture; trade; metal supply and coinage. We are aware of the need to emphasize our recognition of the fact that such an undertaking can be neither deWnitive nor comprehensive. It is bound to omit some other important topics and sources of data such as osteoarchaeology which lack correlative documentary evidence,3 though we would certainly hope to take some account of results derived from such data. It is also impossible to attempt comprehensive treatment even of the selected topics across the whole period and the whole Mediterranean region. We will be satisWed if we can establish a framework and a methodology for suggesting how the evidence might address the crucial questions and issues, along with some detailed illustrations of how we might hope to make this work in the key subject areas for some periods and for some regions of the Roman empire. In sum, the subjects and the objectives we have identiWed will not yield clear-cut 3 Cf. Jongman 2007: 193–5 on femur length as reXection of stature and thus as an index of economic prosperity.

Quantifying the Roman Economy

7

and simple answers to questions about growth and decline over the whole empire in a period of nearly half a millennium. Our aim is rather to explore the complexities and suggest the trends. Explicit recognition that there is far more material than we can hope to exploit even in Wve years will, we hope, encourage further enhancement and analysis of economic data.

I I . D E F I N I T I O N S, M E THOD S, AND AP PROACH ES Inevitably, the subject raises a number of general issues which will recur throughout the programme.4 There now appears no need for us to revisit the old ‘primitivist–modernist debate’ in detail. As regards the comparative context, we will certainly keep in mind the features of what are generally characterized as ‘pre-capitalist’ economies which existed in areas that were eventually to fall under Roman rule and consider whether that eVected a fundamental transformation of their economic character. Whether or not we think that the Roman economy was a mixture of pre-capitalist and capitalist features (such as long-distance trade and production of goods for markets),5 the analysis of transformation and integration, such as might have occurred under the inXuence of Roman rule in large areas of the ancient Near East, is crucial and needs to take into account a more gradualist or developmental perspective, laying emphasis on the continuity rather than the watershed.6 This in turn will feed into analysis of the impact of institutional frameworks and structures which may be seen as central to economic growth when they were strong at the height of the empire and, conversely, crucial to economic decline if and when they weakened.7 Put another way, we need to consider whether we can meaningfully regard the Roman 4 Several are usefully discussed by Morley 2007. 5 Pleket 1993. 6 So Davies 2001: 13–14 noting that the Achaemenid empire shows ‘a stupefying range of economic and Wscal patterns and systems’, continuing into the Seleukid and Ptolemaic periods with the partial exception of monetization (on the latter see now von Reden 2007). 7 North 1990; see also Saller 2002.

8

Alan Bowman and Andrew Wilson

empire as a single unit or entity for any or all of the period encompassed by this study, which includes episodes of growth, contraction, and political fragmentation, as well as greater or lesser degrees of administrative control over time and space. An obvious focus of debate will be the question of whether it is useful or essential to study the empire region by region, which will inevitably reveal both diVerences and similarities over a huge range of geographical and institutional characteristics. A study of mining in Spain or agriculture in Egypt will tell us a lot about Spain or Egypt but little of more general importance about the empire as a whole, save perhaps what we already knew, that Spanish silver production underpinned the coinage, or that Egyptian grain fed a signiWcant proportion of the population of the city of Rome. In other words, a micro-regional approach shines only very narrow beams of light and no one would think it feasible (if only because of the very uneven spread of evidence and the huge gaps in our knowledge) to combine these narrow beams into one vast Xood of illumination. At the other end of the scale of generalizability, we might be urged to consider the usefulness of world-systems analysis, of which Wallerstein was a leading proponent.8 This was fashionably broad a decade ago and has indeed been discussed with reference to ancient Rome, but it has the crucial drawback that it tends to obscure or suppress the diVerences between regions which we must regard as vitally important characteristics of the various parts of the Mediterranean world and north-western Europe under scrutiny. To suppress them would from our point of view be very damaging, precisely because among the things we want to know about the Roman Mediterranean over time are the key similarities, diVerences, and connections from region to region, as part of a single economy or economic system, or simply as a number of small ‘economies’ linked to a greater or lesser degree. This debate cannot be concluded here but it is worth indicating brieXy: (1) that we think it is important to make economic sense of something which clearly was in many respects one political entity or ‘system’ which included Britain, North Africa, and Syria, without having to decide which was more or less ‘typical’; and (2) that the links and connections (or perhaps now ‘connectivities’) between the 8 Wallerstein 1974–80; see Woolf 1992.

Quantifying the Roman Economy

9

diVerent parts (conceived both geographically and structurally) of the Mediterranean will inform our understanding of the sense in which we can talk of unity, a single empire or diverse regions as part of a single political and economic ‘system’.9 Second, asking how far one part of the Roman empire might be compared with another naturally leads us to reXect on the beneWts of comparing the Roman empire with other political states, empires, and periods. We are in no doubt that in principle this must be potentially useful and failure to do it would rightly invite criticism. There are various ways in which we might do this. Broad comparisons between Rome and other examples are not exactly a new phenomenon, although most have focused on political structures, imperialism, colonialism, and the like rather than the economy.10 Economically, an obviously attractive comparison has been that between Rome and Mughal India, recently revived by Peter Bang;11 the evidence for Han China is less easily accessible to those who cannot read Chinese or Japanese. Egypt over the longue dure´e, and particularly in the nineteenth century, oVers some promise (though rarely unequivocal or universally accepted).12 We might anticipate that it would be illuminating to compare empires in respect of taxation policies and mechanisms, distribution of wealth, technological development, size of diVerent sectors (agriculture, trade, services, etc.), growth and decline. Implicit in such exercises is always the question of how exactly such comparisons help us. Simple juxtaposition for similarity or contrast is always intriguing but that hardly counts as analytical. We need to go further and determine how analysis of another imperial economy might Wll gaps in our knowledge of the Roman economy, or how it might suggest diVerent questions which can be put to the evidence we have. It seems to us that it is possible to gain some insight from other periods and areas 9 Cf. Horden and Purcell 2000. We do not think that the debate about the usefulness of the term ‘Romanization’ is relevant to this, see Mattingly 1997; 2004; Webster 2001. 10 For comparative studies see now Bang 2006, Meyer 2006; also Van der Spek 2006 on Babylonia. 11 See Bang 2002 and cf. Bang 2006, Bang and Bayly 2003. 12 This is implicit in the approach of Bowman and Rogan 1999. An example is cited by van Minnen 2000, n. 4.

10

Alan Bowman and Andrew Wilson

into the level and nature of economic performance in the Roman Mediterranean in the period under consideration: for example, to highlight what structural features of the Roman economy might be typical of large pre-industrial economies and, conversely, what aspects might be peculiar to the Roman world. A third general consideration concerns the dialogue between diVerent kinds of evidence. We have had suYcient warning of the dangers of using statistical evidence from ancient literary sources, where Wgures are notoriously unreliable and modern scholars often seem inclined to trust Wgures they can Wnd which suit their prejudices and reject those which do not.13 Such evidence may sometimes Wnd support, or the opposite, in documentary or archaeological sources. But our evidence is in general so lacunose and random that, more often than not, individual items are isolated and cannot be checked against a context. Once again, however, there are some exceptions and one of the themes which runs through the present programme is an attempt to create where possible a dialogue between archaeological and documentary evidence. This is, of course, not a new idea and there is a fairly recent scholarly literature which has reviewed progress and suggested ways forward.14 This has proved, perhaps surprisingly, diYcult to achieve very eVectively.15 Reconstruction of the physical conWguration of landholdings, an urban landscape, or a network of village settlements can be a frustrating exercise and it is striking how often such attempts result in virtual or relative topographies which cannot actually be pinned down to coordinates on a map, even when we have relevant documents;16 even straightforward lists of artefacts too often fail to harmonize with classiWcations based 13 e.g. Szaivert and Wolters 2005, see Scheidel 1996. Jones 1948: 10 considered Josephus’ Wgure for the population of Egypt (BJ 2.385) the only reliable statistic in a literary source but many have rejected this as far too high (e.g. Rathbone 1991, Bagnall and Frier 1994). 14 See Ørsted 1994, Hitchner 1995a, and Storey 1999. 15 For a useful survey see Storey 1999, noting (206) that ‘a balanced, dependable method for integrating textual and archaeological data is still lacking’. See also Hitchner 1995a. An example is MacKinnon 2004, who concludes (243) that ‘perhaps the greatest value of integrating zooarchaeological and textual data comes in terms of the analysis of animal breeds and the recognition of breeding improvement’; it is relevant that the textual evidence available to him is literary, not documentary. 16 Examples include an attempt to identify buildings under repair at Hermopolis in the 260s (van Minnen 2002), a map of villages in the Oxyrhynchite Nome

Quantifying the Roman Economy

11

on archaeological evidence.17 There are, however, also some data which allow us to make a direct rapprochement between documents which relate to domestic housing and the archaeological evidence for houses. In 1985 Hobson was able to examine the evidence of excavated structures in Fayum villages and to show that private houses in these communities were small, with about 60 m2 of Xoor space per storey.18 She proceeded to argue, on the basis of census documents and the like, that the density of occupation in these rural communities was ‘remarkably consistent with the Wgures given for Egypt in 1979, which show that in rural areas the average occupancy is 3.5 people per room, or 10.5 people per average 3-room house’.19 In general, we might expect it to be much easier to make these connections for urban structures or for villas than for agricultural land.20 A slightly diVerent and more promising approach is to treat, where possible, the documents themselves as artefacts in their archaeological context.21 Comparison of time series data for archaeological and documentary evidence for the same phenomenon can be instructive in illuminating possible biases in the data, as discussed below for two case-studies involving water-mills and water-lifting devices (pp. 33–8). The overarching aim here, then, is a series of studies which will suggest how we might identify major structural features, behaviour, and performance of the Mediterranean economy over 450 years of Roman domination, which can be compared with other periods and (Rowlandson 1996: xiv), the topography of Oxyrhynchus (Kru¨ger 1990), size and distribution of Fayum villages (Mu¨ller 2003a and 2003b, Mu¨ller and Lee, forthcoming). For a suggestion for Africa see Hitchner 1995a: 142. 17 For example a list of nails supplied for various types of footwear at Vindolanda in northern Britain undermines the classiWcation of nailed and un-nailed shoes established by archaeologists, see Bowman and Thomas 2003: 65. 18 Hobson 1985: 217. 19 Hobson 1985: 222. This can be supplemented by items of evidence such as the ground-plan of a house in P.Oxy. XXIV 2406, and measurements given in P.Lond. 50.7 (I, p. 49, house of 2113 cubits, courtyard 4  13 cubits). See further, pp. 57–8 below. 20 For a catalogue of Italian villas which might provide a starting-point, Marzano 2007. 21 Models for other periods may help, e.g. http://www.mnp.nl/bibliotheek/ rapporten/481508015.pdf, van Minnen 2000, n. 26.

12

Alan Bowman and Andrew Wilson

areas, by collecting and analysing quantiWable documentary and archaeological evidence for key areas and economic activities. The objectives are to compile datasets of evidence for each of the four diagnostic areas and to produce synthesizing studies for each area, plus an overview addressing the general and structural issues through the key areas. QuantiWcation studies on the Roman economy are not simply about counting and aggregating. Analysis of the data ought to enable us to measure change and variability—developments over time, variation across space, similarities with and diVerences from other cultures at diVerent times. Despite the slightly more optimistic view of the value and robustness of our data which we implied above, we cannot expect to derive a series of straightforward and unambiguous results from the data we have. At best, we may expect to identify a number of measurable indicators of trends that might tell us about diVerent aspects of the economy. We attempt in the remainder of this Introduction to point out what we see as some of the major issues and questions we face, and to oVer some examples of approaches to and methods of using archaeological and documentary data. The main general issues we need to address are the following: . The utility of estimates of GDP.22 We will need to consider whether we can estimate aggregate GDP (i.e. the sum of the value of all goods and services) for some places and some periods and what useful conclusions we might derive from that in relation to economic growth or contraction. We remain sceptical about estimating per capita GDP because of our uncertainty about population Wgures but we might suggest proxy data which might oVer some indications of per capita levels of income and expenditure. . We can attempt to identify drivers of growth including but not only those enumerated by Saller (trade, intensiWcation of capital investment, improved technology, education of the workforce, institutional attitudes and stimuli, increased division of labour, etc.).23 We also need to consider whether their absence (or opposite trends which might be postulated such as a decline in literacy,

22 See Temin 2006a.

23 Saller 2002 ¼ 2005.

Quantifying the Roman Economy

13

or a decrease in volume of shipping) is necessarily an indicator of decline. . We can also attempt to identify phenomena that are likely to be reXections of (rather than criteria for) growth, such as urbanization, import replacement, increased consumption, improved standards of living. The project relies on the collation of datasets which can then be analysed, and one of its key components is therefore the creation of a series of databases with appropriate query tools, ultimately to be made available on the internet for the use of other researchers at the end of the project, both in online searchable form and as static tables in a stable archive format to be archived with the Archaeology Data Service.24 This will involve a large amount of collation of material from published sources, but where possible, we would also want to make use of existing online database projects such as the epigraphic corpora of inscriptions from Heidelberg,25 the various papyrological databases brought together in the Trismegistos project and elsewhere,26 Roman Amphorae: A Digital Resource, etc.27 There is a growing research need, not currently met, for the development of data mining technologies to exploit the increasing mass of data—e.g. archaeological Wnds, inscriptions, or papyri—held in numerous diVerent databases generated by diVerent projects and hosted on diVerent servers. This would require, on the one hand, the development of middleware search layers and thesauri (to translate, for example, between diVerent classiWcation systems for the same type of amphora: Dressel 12 ¼ Augst 22 ¼ Beltra´n 3 ¼ Ostia 52 ¼ Peacock and Williams 14), and, on the other, greater interoperability between websites, so that one could link an epigraphic corpus to a mapping project for a visual distribution map of e.g. all fourth-century ad building inscriptions. The use of interoperable web-based database and GIS architecture could enable the daunting task of updating archaeological distribution maps (below, pp. 26–7) to be done as a collaborative venture between several projects, distributing the 24 25 26 27

http://ads.ahds.ac.uk/. http://www.uni-heidelberg.de/institute/sonst/adw/edh. http://www.trismegistos.org/. http://ads.ahds.ac.uk/catalogue/archive/amphora_ahrb_2005.

14

Alan Bowman and Andrew Wilson

workload and facilitating future updates by allowing individual excavators to report their new Wnds or to link their databases of Wnds from diVerent sites together.28 The comprehensive resolution of these issues is beyond the scope of the present project, but we hope to contribute to thinking and debate on the question and inXuence the development of such technologies. For the analysis of the data, we need to identify appropriate statistical techniques for quantiWcation. As regards speciWc methods and approaches, we rely essentially on a familiar repertoire of simple statistical tests, such as regression analysis which evaluates the likelihood that there is a relationship between one set of data (e.g. commodity prices) and another (e.g. distance from a major market or consumer centre); or the application of the Gini coeYcient (or Lorenz curve) to accumulated data to measure the degree of inequality of distribution expressed as a decimal Wgure from 0.1 (complete equality) to 1.0 (complete inequality).29 Studies of this sort inevitably rely to some extent on modelling the social and economic structures under scrutiny and looking for goodness-of-Wt in the evidence. We need to consider how we might best do this. The beneWts and the pitfalls of such model-based approaches are obvious. As existing examples of good practice, we can cite the use of demographic models and life-tables to describe the structural features of ancient populations which Wt the items of evidence available to us; that is, they are based on empirical data. Other examples of models which have been widely used include those which proceed from assumptions which are strictly unveriWable but judged likely to be in the correct range (e.g. that the population of the empire was 50 or 70 million) to conclusions which may be judged 28 This possibility is envisaged in the aims and the structure of the project ‘A Virtual Research Environment for the Study of Documents and Manuscripts’ (see http://www.bvreh.humanities.ox.ac.uk), supported by JISC as part of Phase 2 of its Virtual Research Environments programme. It will be realized through collaboration with the database structure developed in the Silchester Roman Town project, see http://www.ahessc.ac.uk/all-hands-2006. 29 These techniques and tests, e.g. Spearman’s rank correlation coeYcient, are described in standard handbooks such as Hudson (P.) 2000. Some of the simpler tests are available online (e.g. http://mercury.soas.ac.uk/users/sm97/teaching_intro_qm_ notes_gini_coeYcient.htm). For examples in the present subject area see Kessler and Temin forthcoming, Bowman 1985, and Bagnall 1992.

Quantifying the Roman Economy

15

more or less plausible, or more or less consistent with a range of ‘known’ phenomena (e.g. tax-rates or calorie requirements for human subsistence).30 Our approach to economic data more naturally leads us in the Wrst direction than the second, but it is inevitable that our inability to be certain about some of the crucial facts will sometimes force us to derive alternative models to be tested for goodness-of-Wt. We have used and continue throughout to use the terms ‘integration’, ‘growth’, and ‘decline’, and we now need to say what we mean by these in the context of the ancient Mediterranean, to identify the key indicators of these phenomena and the means of quantifying them.

III. AN INTEGRATED ECONOMY? The task we have outlined requires analysis of the nature and extent of the integration of the economy of the Mediterranean world under Roman rule. We need to say what we understand by the term ‘economic integration’, how we identify the key diagnostic features of an integrated economy and how we propose to measure or quantify the degree of integration (not a simple matter in the case of a process or a number of connected processes). First, to what extent was the Roman economy integrated across the whole empire rather than a number of disparate but linked regional economies (‘a system of interconnected sub-economies’31)? What is the extent of geographical integration, that is essentially but not only the progressive integration of the gradually acquired provinces? We also need to consider the well-attested economic contacts beyond the territories under direct control (a particularly important issue for Roman imperial rule in which a degree of suzerainty or control was often claimed for territories not under direct governance or 30 For examples relevant to the Egyptian agricultural economy, see Bagnall 1992, 2005, van Minnen 2000, and Bowman (this volume); for demography and social history, Bagnall and Frier 1994, Tacoma 2006. The model developed by Hopkins 1980 for taxes and trade is based on unveriWable assumptions. For further examples of models in ancient economic history see Manning and Morris 2005. 31 Harris 1993: 29, cf. Scheidel 2004.

16

Alan Bowman and Andrew Wilson

administration).32 These considerations will naturally demand that we keep in mind: (1) that the gradual transformation of the Hellenistic kingdoms, north Africa, and western Europe into a Roman empire inevitably involves greater economic integration than had existed earlier,33 and (2) that the processes of integration into an empire were not only economic processes; we have in mind, for example, the multifarious relations implied by the concept of ‘connectivity’ and ‘globalization’ (if applicable).34 There are both quantitative and qualitative issues. Does the economically integrated area shrink (pari passu with the reduction of the extent of territorial control)? Does the nature of integration change? Crudely put, is the Roman economy more integrated in the second century ad, less so in the third, and more so again in the fourth (following the rather simplistic but often cited cycle of growth to climax, decline, and recovery)? How do we describe regional diVerences and the links between regions? Are newly incorporated regions brought into the integrated structures to the same level and at the same pace (selfevidently not), or are some more integrated than others (we would naturally suppose so)? Second, how far was it structurally integrated—that is, are state mechanisms of economic control, and the behaviour of ‘free market’ local economies part of a coherent economic ‘system’? We can easily identify central features of both ‘sectors’ (on the one hand, regulation of currency, taxation levels, direct ownership of production, some areas of price control; on the other, local Xuctuation of supply and demand, of prices, entrepreneurial activity). But what exactly are the economic structures or frameworks that would eVectively weld these into a system (e.g. imperial policies, which have proved hard to identify convincingly, but perhaps include Diocletian’s Currency Reform and Price Edict, laws such as those aVecting landownership or navicularii, practices such as the mixture of state and private 32 For Rome’s eastern trade see the collection of evidence by Raschke 1978, with more recent archaeological work by Begley 1993; Will 2004; Peacock and Blue 2006; Peacock and Williams 2007. Evidence is now growing for Saharan trade, especially with the Garamantes of the Libyan Fazzan: Mattingly 2003. 33 Erdkamp 2005: 203 citing J. K. Davies to the eVect that ‘The Hellenistic Economy’ was not integrated. 34 Horden and Purcell 2000.

Quantifying the Roman Economy

17

activity in transport services, ‘tools’ such as taxation and currency)?35 On the one hand, we can attempt to identify speciWcs which have this integrative eVect; on the other, we can attempt to infer them and attribute to them a greater or lesser degree of deliberate manipulation, such as is implicit in the much-cited ‘taxes-and-trade’ model developed by Keith Hopkins.36 Third, we need to consider how to locate the processes of economic integration in relation to other processes and developments which are central to the debate about the character of the Roman empire. The two which most readily spring to mind are ‘Romanization’ and ‘globalization’. It is possible to indicate our approach to these issues without indulging in lengthy debate about the merits of the diVerent labels. Both of these portmanteau terms conceal the combination of a number of complex processes. The use of the term ‘Romanization’ has recently been subjected to severe criticism,37 but whatever term one uses to convey the facts, it is undeniable that the growth and development of the empire involved the spread of institutional, linguistic, cultural, economic, and religious phenomena which were in some important sense ‘Roman’ or promulgated by Rome (that wider deWnition will allow us to include in our purview Hellenistic and Graeco-Roman constituents too, as we certainly need to do).38 Archaeological assemblages in the provinces as well as Italy frequently reXect a kind of Roman consumption package—amphorae for wine, olive oil, and Wsh products, and red gloss table pottery; a qualitative perspective on the perceived integration of the Roman economy at least insofar as products from all over the empire were held to be widely available is provided by the verse inscription on the Mausoleum of the Flavii at Kasserine, which characterizes diVerent provinces by the products for which they are famous.39 Likewise, there are some deWnitions of ‘globalization’ that 35 See Paterson 2004. These issues are well analysed by Harris 2003. 36 Hopkins 1980; 2002. 37 Mattingly 1997; 2004; Webster 2001. 38 We do not seek to avoid the use of the term ‘Roman’, cf. Barrett 1997. Analogous issues have arisen over the use of the term ‘Graeco-Roman’, see Lewis (N.) 1968. 39 CIL 8.212 ll. 28–31: Graecia cum pueris, Hispania Pallados usu, venatu Libyae tellus, Orientis amomo; Aegyptos Phariis levitatibus artibus actis, Gallia semper ovans dives Campania vino. Cf. Groupe de Recherches sur l’Afrique antique 1993.

18

Alan Bowman and Andrew Wilson

are eVectively restricted to an economic perspective, pointing to the global interdependence of countries through increasing volume and variety of cross-border transactions in goods and services, free international capital Xows, rapid and widespread diVusion of technology, and so on. It is, however, more common now to regard globalization as a complex concept which certainly also encompasses major political and institutional aspects, intertwined with economic, cultural, and technological elements.40 Without adopting the extreme position that all history is economic history, we certainly regard economic integration as central to the complex of processes which hide behind these labels without wishing to suggest that economic integration is a suYcient deWnition of either, nor do we attempt to assert its necessary primacy, for such an assertion cannot rest on quantitative studies of the kind we propose. Naturally, we cannot at this stage attempt a full analysis of the consitutents or criteria of economic integration, but it is worth sketching out the areas which we think are important, with some indication of our approach to them.

III.1. Economic policy If we could demonstrate that Rome’s political leaders developed and implemented a coherent economic policy at any or all parts of the period under review, we would have a powerful argument for integration. Unsurprisingly, this is not easy to do, not least because we would hardly expect a consistency of economic policy over four centuries and ancient sources are in general notoriously lacking in 40 Morley 2007: 90–102; Jones (A.) 2006: 2: ‘the growing interconnectedness and interrelatedness of all aspects of society’, a process that ‘can be identiWed in almost every dimension of contemporary life’, ‘However, its proWle as a keyword springs largely from a more speciWc association with changes in the world economy in recent decades (and since the end of the Cold War in particular). For most non-academic commentators, globalization is Wrst and foremost about the development of a globalscale, capitalist free-market economy that is entwining activity across national borders in new and unprecedented ways. From an academic perspective, however, it has a plethora of further usages. It is applied to various diVerent dimensions of social life in the contemporary world, spanning cultural, technological, informational, environmental and political transformations—to name but a few of the more signiWcant spheres of its conceptual reach.’

Quantifying the Roman Economy

19

any sort of description or analysis of what the modern world (since Adam Smith) would regard as ‘economic policy’. It is much less hard to Wnd modern statements about the ‘economic policy’ of particular emperors, notably Augustus (encouraging trade), Vespasian (belttightening reorganization as in Suetonius), Trajan (building/trade), and Diocletian (economic revival) but these tend to be fairly narrow in focus.41 Roman economic policy is sometimes identiWed or implied at a more general level, e.g. in the Red Sea/eastern desert, where its major constituents include improvement of military control and communications, thus creating better conditions for trading. There can be no doubt at all that at least in the earlier empire the imperial house engaged directly in trade, agriculture, and manufacture over the whole empire in ways which must have had some direction from the top and invariably included interaction with private individuals.42 But the nature of the imperial acquisition of, for example, the major marble quarries, remains unclear. So too, although we can trace the imperial acquisition of major senatorial olive-oil producing estates in Istria in the Wrst century ad, or the acquisition in the course of the second century of the main brickyards in the Tiber Valley and the suburbium of Rome, it remains uncertain how far these processes reXect a deliberate and planned policy, and how far a more casual agglomeration of landholdings through marriage or conWscations.43 Nonetheless, ownership and direct management of land and other wealth-creating activities by the imperial authority, the level of which varies over time, must be an integral component of economic policy; by the same token, the decrease in direct ownership and management which is evident at least in Egypt after Diocletian must also be deliberate and strategic.44 There are, of course, a number of economically related measures taken by the emperor or the state which are recorded in the historical sources. These include the annona, the free distribution of quantities of grain to citizens at Rome, to which were added, progressively, pork (early second century), olive oil (under Severus), and wine 41 42 43 44

e.g. Paterson 2004. Harris 2003; Bowman 2008. Istria: Matijasic 1999. Rome and Tiber Valley: Seta¨la¨ 1977. Rowlandson 1996: 70–101.

20

Alan Bowman and Andrew Wilson

(Aurelian).45 The evidence of centralized olive amphora discard at Monte Testaccio and the control inscriptions on Dressel 20 amphorae from south-west Spain suggest state intervention in the olive oil supply of Rome long before the inclusion of oil in the annona, presumably with the aim of ensuring price stability through reliable supply by bulk state purchase and resale.46 By the third century ad some larger provincial towns had established their own annona schemes for the distribution of grain to a portion of the citizen body.47 But as far as our present evidence goes these measures all aVect particular regions only—Rome, certain other towns, and their main regions of supply (Egypt and North Africa in the case of the grain supply for Rome; Spain and North Africa in the case of olive oil supply). More promising perhaps are the second-century ad Lex Hadriana de rudibus agris encouraging development of marginal, uncultivated land in North Africa,48 or a statement in the Historia Augusta to the eVect that Probus’ investment in public building works brought about an economic return: ‘In Egypt there are works of his, which he had built by the soldiers, in many cities. But around the Nile he did so much that he singlehandedly augmented the grain tribute. He built bridges, temples, porticoes, and basilicae by military labour; he opened up the estuaries of many rivers, and drained several swamps, and established cornWelds and arable land in them.’49 Yet the Historia Augusta is a notoriously unreliable source, and both of these instances seem to be regional rather than universal measures. None of these examples is in itself invalid or useless, but we struggle to Wnd some overarching conceptual vision which encompasses the economy of the empire as a whole and is not so general as to be completely banal, nor can we Wnd explicit recognition of the need for budgeting on a macro-economic scale until the 45 Rickman 1980. 46 On Monte Testaccio see e.g. Rodrı´guez Almeida 1984; Bla´zquez Martı´nez and Remesal Rodrı´guez 1999; E´tienne and Mayet 2004, vol. 1: 23–39; Aldrete and Mattingly 1999 though none of these authors actually goes so far as to attempt to explain the state connection in the manner suggested here. 47 See P.Oxy. vol. XL. 48 See Kehoe 1988; Hitchner 1995a. 49 HA, Prob. 9.3–4.

Quantifying the Roman Economy

21

post-Diocletianic empire when, according to Jones, the praetorian prefects estimated the annual needs of the state and calculated the rates of levy required to meet them.50 Various possible explanations of this silence suggest themselves. We are asking the wrong question(s). Lack of explicit articulation does not mean that it was not somehow institutionally embedded. There was no overarching economic policy for the empire—just a series of measures to meet current needs, whatever they might be. We do not need a single explicit ‘policy’ or a consistent direction of policy to demonstrate integration, merely an array of integrated economic institutions or patterns of behaviour. This last is the direction in which we are heading.

III.2. Coinage and monetization This might appear to be the most straightforward of our diagnostics of integration but the appearance is somewhat deceptive. It would be good to be able to begin with the simple assertion that the monetary system, regulated through the imperial mints, achieved an extremely high degree of coherence and integration in a world which was to all intents and purposes highly monetized.51 Mines were perhaps not all in direct imperial ownership but this is certainly an area in which we should look for integration in the form of a high degree of state control complemented by private exploitation.52 In reality, the pattern of currency is much more complex.53 We must allow for signiWcant diVerences between west and east. In the former, although there are signs of residual non-Roman coins (except in Britain), Roman denominations quickly became universal. In the east, Rome took over areas with a multiplicity of existing local coinages which continued to circulate in a heavily monetized environment and the move to an almost wholly single Roman denomination (the radiate) was not achieved until the middle of the third century ad.54 The cases of Syria and Egypt are instructive. In the former, there is room for 50 52 53 54

Jones 1964: 449. 51 Howgego 1992. Hirt 2004, forthcoming as an OUP Classical Monograph. For a good summary see Burnett et al. 1992: 6–9. Butcher 2004: 253–6.

22

Alan Bowman and Andrew Wilson

doubt whether SC issues were freely used between cities or whether cities acquired them purely for local use. In local bronze coinage there was not a regional system: that is, coins found in one region originating in other regions in the province are no commoner or rarer than those originating in regions which were not part of the province.55 Egypt’s so-called ‘closed currency’, which continued until the Diocletianic reform of ad 296, may have coexisted with the circulation of gold aurei, although this is not certain. It has been shown that the Alexandrian mint made coins for Syria and cooperated with other mints, was part of an overall imperial scheme, all of which made it in eVect no more closed than other monetary areas.56 In order for these currencies to function in cooperation, which they clearly did, there must have been arrangements for exchange, small-scale examples of which we know where surtaxes were imposed or a premium was charged for commercial exchange; but we do not know the mechanisms which allowed this to work at provincial level (e.g. at Alexandria).57 We might reasonably postulate that a move to single currency would of course lower transaction costs, but one would guess that it would be diYcult to demonstrate this for the east in the period after ad 250. At all events, despite the complexity of the patterns, the eVective integration of the functioning of the currency and the monetary system seems beyond doubt and is fundamental to the phenomena described in the three following sub-sections.

III.3. Commercial institutions Coin was one of the vital instruments in moving goods and services across the whole empire but it is vital to recognize (1) that coin is not 55 Butcher 2004: 176. 56 Burnett 2005; Andreau 2005. Gara 1976 attempts to Wt the Egyptian evidence into a framework of Roman monetary domination. For the movement of Roman coinage to India via the Red Sea ports see Andreau 2005: 331–2, Rathbone 2001a; Bowman 2009. 57 Pergamum inscription, OGIS 484. As Andreau 2005: 330 says, at Alexandria the exchange must have been done through the public banks, which are attested though not in detail (Bogaert 1994, 1995). For a hypothesis as to the mechanism see van Minnen 2000: 208–10. The supposed equivalence of the denarius and the Alexandrian tetradrachma will have simpliWed matters but that too is not unproblematic (see Andreau 2005: 332–3).

Quantifying the Roman Economy

23

the only form of money and (2) that money is not the only mechanism for creating integration of movement. For money in noncoined form it will be suYcient to refer to the recent detailed article by William Harris, emphasizing the importance of credit, inter alia, as a form of money.58 Other vital institutions, with which this is linked, include banks and the various ways in which commercial transactions were documented and recognized as valid. A full appreciation of the complexity of this aspect of integration would require detailed discussion of the legalities of Roman contractual obligations and the increasing interpenetration of Roman and peregrine legal instruments (which is also closely connected to the important subject of social integration). This cannot be attempted here and we will be content with the proposition that even the superWcialities of commercial and Wnancial documents show how they underpin and validate the movement of goods and services across huge tracts of the empire.59 More speciWcally, however, we might note and explore the tendency under Roman rule towards greater uniformity of interest rates and the mechanisms of applying them. This phenomenon seems clear when one compares loan contracts in Ptolemaic and Roman Egypt and it is our impression that ‘standard’ Roman terms and rates became widespread elsewhere too.60 If this were sound, it would certainly be another diagnostic feature of economic integration.

III.4. Movement and trade The ability of the Roman imperial economy to move goods and services across huge distances in the empire and beyond its borders

58 Harris 2006. 59 e.g. TPSulp. 14 (Camodeca 1992), a vadimonium made in Rome between C. Sulpicius Cinnamus and an Alexandrian; Rhodian bank drafts, P.Oxy. L 3593–4. 60 Most obviously the explicit stipulation of interest charges which are at other times and places sometimes concealed in the capital loan and the rate of 12% per annum; proof that this was enforced as a maximum is lacking, see Andreau 1999: 90–9, esp. 92. For some interesting observations on the history and economic signiWcance of interest rates over the longue dure´e see Hudson (M.) 2000.

24

Alan Bowman and Andrew Wilson

is not in any doubt at all. Therefore, if we feel the need to demonstrate how that was done (see above) we do not feel the need to demonstrate that it was done. In this area of our inquiry, we will focus more on the range and scale of movement and the relationship between the constituent parts of the empire: movement of goods over the whole empire and beyond its borders, movement between regions, movement within regions (deWned as provinces or geographical ‘units’ such as Africa, Asia Minor, the Iberian peninsula), local movement between towns and villages. Naturally this also involves issues relating to the organization of production (both agricultural and ‘manufactured’) at the highest level (e.g. the imperial house) and at the lowest (e.g. the local farmer or artisan). Crucially, trade is the category of evidence which has been most heavily exploited in order to demonstrate growth and decline in the Roman economy. So one of the central features of the supposed economic ‘decline’ in the third century is thought to be the reduction in the volume, kind, and quality of goods moved over distance.61

III.5. Markets All of the topics discussed above are closely related to the major issue of the nature and structure of the market or markets in the empire. This is crucial to any deWnition of economic integration and is here singled out for discussion not least because it has been highlighted in recent studies many of which have come close to making it a suYcient condition of integration. Thus, the characterization of this aspect of the economy as a network of interconnected ‘free’ markets in which the levels and movement of prices are in economic terms rationally determined (e.g. obeying the laws of supply and demand, reXecting costs of transport over distance) has been powerfully propounded.62 It is argued that for the eastern Mediterranean there is no comparable level of market integration

61 Cf. Ward-Perkins 2005: 102–20. 62 Temin 2001, Rathbone 1997, but the quantity of evidence for price levels is rather less than one would ideally want. Market integration for the grain trade is well discussed by Erdkamp 2005: 143–205.

Quantifying the Roman Economy

25

in the economies of the Hellenistic period and the Roman domination will therefore mark a very important change in this respect. On this view we would expect to see increasing market integration over the period from c.100 bc to ad 200 and we will then have to test the hypothesis that this began to break down and become fragmented, as did the political unity of the empire, over the course of the next century. These broad brushstrokes conceal a plethora of detailed questions and issues. Foremost among the indicators of the behaviour and structure of the market(s) is the level and movement of prices over space and time. It is true that, as far as our data will take us, the general stability of price levels over the Roman Mediterranean seems to have endured for a quite remarkably long time in the Wrst two centuries ad, despite some incidences of coinage debasement which might naturally lead us to expect marked Xuctuation of prices. On the other hand, the amount of usable evidence at our disposal is not great or densely clustered and the calculations based on rather thin evidence for grain prices cry out for some further testing and focus closely on the relationship of prices in the capital and those places from which its grain supplies were derived (principally Egypt). It is nevertheless useful, we think, to exploit indicators of comparative value across time and place (such as the ratio of value/cost of wheat to barley).63 It is our impression that other scattered bodies of data present a much less tidy and ‘economically rational’ picture, in which there might be marked regional diVerences, or speciWc local conditions might cause fairly violent deviation from postulated ‘normal’ prices or market transactions. Possible factors which inXuence such distortions might be, for example, famine or crop failure, disruption of productive or supply processes by civil or military unrest. We also need to take account of systemic factors which might be broadly classiWed as governmental controls of market structure and functions. In a very general but nonetheless powerful way, this is one major impact of a taxation system of the sort which existed in the Roman Mediterranean. At a more speciWc level, the institutionalized

63 Bowman 2006: 85, a value/cost ratio of 3:1 for wheat:barley in Britain c. ad 100; according to Bagnall 1993: 25 in fourth-century Egypt the ratio was normally about 2:1.

26

Alan Bowman and Andrew Wilson

practice of control of supply and prices of goods can be seen in the imposition of requisitions for military or other governmental purchases which do not Wt the free market pattern. Third, and less systematically, price and market controls (e.g. strictures on grain hoarding or the requirement to declare stocks and prices) can be imposed to meet particular needs created by local shortages. This might suggest that we need to distinguish between a governmentregulated and a free market sector, to elucidate the relationship between them and to see whether that changed over time. Within that spectrum, it is also possible to discuss whether we are dealing with short-range (i.e. local) or medium-range (regional) or longrange (empire-wide) integration and whether there might be distinctions according to the commodities in the market.64 The question of market integration may be further complicated by the existence of ‘mass’ markets, which we take to mean the very few really large conurbations in the Roman Mediterranean,65 where the demands for basic commodities (grain, some other foodstuVs, slaves) might suggest a high degree of integration over long distances. This issue of local versus long-distance trade is taken up in Fulford’s response to Wilson’s paper on trade in this volume, stressing the potential of archaeology to illuminate the impact of transport networks. In particular, he contrasts the access that coastal cities had to imports with the relatively restricted distribution of certain types of imported cooking wares in inland zones, arguing that the relatively low costs of maritime transport created almost an extended ‘local’ zone for trading, but there was a sharp drop-oV in penetration of goods inland. If the distinction is as sharp as Fulford suggests, we ought to see marked diVerences in economic development between coastal zones and inland ones. Much more work is certainly needed systematically to analyse the now massive quantity of archaeological data for traded artefact distribution to explore this phenomenon; a major research desideratum would be the updating of distribution maps for some of the most common amphora forms and pottery types.66 This would be a truly massive task, but the use of 64 Cf. Hopkins 1983. 65 Cf. Erdkamp 2005: 177–81. 66 For early amphora distribution maps, see e.g. Riley 1979; Peacock and Williams 1986; for a more recent update, of Republican wine amphorae in Gaul, see Loughton 2003.

Quantifying the Roman Economy

27

interoperable web-based database and GIS architecture, as discussed above, could facilitate this.

III.6. Social integration We might also ask to what degree the Roman economy was integrated across society. Were peasants integrated into a market economy, or not?67 Were the state and the army actors, or separate sectors? What role did the elite play, not merely in consumption, but in production? There are signs of elite and probably senatorial involvement in the level of investment of productive infrastructure in, e.g. olive oil and wine production in North Africa and Istria.68 At a municipal level, we have evidence for local elites involved in the production of salted Wsh products and Wsh sauce, in the case of A. Umbricius Scaurus at Pompeii.69 What was the nature of imperial ownership and involvement?

III.7. Integration within industrial sectors Finally, it is worth exploring the evidence in particular sectors for a level of integration implied by supplier relationships between specialist stages of production, each occurring in separate, dedicated establishments. Such vertical specialization is one of the characteristics of an industry; the textile industry at Pompeii, for example, shows specialist wool-scourers, dyers, fullers, who must have been integrated with the spinners and weavers through a highly organized supply chain. Vertical specialization is also apparent within the Wshsalting industry, to the extent that Wsh-salting factories are found in relatively close regional association with evidence for salt production and amphora production, although usually the three activities occur in separate, specialist facilities which must have been vertically connected through supplier relationships.70 We do not, however, 67 Harris 1993: 20. 68 Wilson 2008a. 69 Curtis 1984; 1988–9. More generally, cf. D’Arms 1981. 70 Wilson 2006. On the Wsh-salting industry in the Iberian peninsula generally, and the various activities related to it, see E´tienne and Mayet 2002. Cf. E´tienne and

28

Alan Bowman and Andrew Wilson

frequently see the vertical integration of such diVerent stages of production within a single enterprise.

IV. ECONOMIC GROWTH

IV.1. Extent and nature When and how intensive growth has been generated is the central question of development economics.71

The extent and nature of economic growth achieved in antiquity continues to provoke debate. Hopkins argued that the period from 200 bc to ad 200 had seen some modest growth.72 Some historians have tended to downplay the signiWcance or extent of such growth as they may admit, emphasizing relative stagnation over the longue dure´e;73 others, and notably some archaeologists, are more optimistic about seeing growth in the material record.74 It is important to distinguish between two types of growth: aggregate or extensive growth, in which a simple increase in population leads to an increase in total GDP, and per capita or intensive growth, in which greater eYciency in production means that each worker is producing more. In reality, per capita growth and aggregate growth might occur together, and may be diYcult to disentangle, but per capita growth is the more signiWcant in demonstrating economic progress, and it is this issue which is most at stake in the debate. Few would deny the possibility, even the likelihood, of some aggregate growth in the Roman empire in the Wrst century bc and the Wrst century ad as a result of demographic growth, even though the extent and the regional distribution of any demographic growth are

Mayet 2004, vol. 1: 103–4, for ‘industrialized’ amphora production geared to the needs of olive oil producers in the Guadalcuivir valley. 71 Lal 1998: 2. 72 Hopkins 1978; 1980; 1995/6. 73 Millett 2000; Saller 2002 ¼ 2005. 74 Hitchner 1993; 2005; Wilson 2002a; Woolf 2000.

Quantifying the Roman Economy

29

debatable. As Bruce Frier puts it: ‘If we assume, as assume we must, very low levels of per capita income in the early Roman empire, it is nonetheless true that population growth could well have contributed far more than any other single source to any overall increase in the Roman empire’s economy during this period.’75 Frier has even raised the possibility that demographic growth may have led at times to overpopulation, leading to a drop in per capita income (from the diminishing returns of labour input into marginal lands) and reduced standards of living.76 Per capita or intensive economic growth may take one of two main forms: so-called ‘Smithian’ growth, in which growth in the extent of trade and the size of markets facilitates increased division of labour; and sustained intensive or ‘Promethean’ growth, based on fossil fuels and continuing technological improvement, which is characteristic of industrialized economies.77 Wrigley emphasizes that before the exploitation of fossil fuels, all economies were ‘organic’ in that they depended almost entirely on organic products not only for food and subsistence, but also for all energy, whether heat or mechanical, involved in the production of goods.78 Sources of energy (even wind- and water-power, through the sun’s determination of climate) ultimately depended on the sun; most importantly plants harnessed solar energy by photosynthesis, and these might be used directly as fuel, or eaten by animals or humans as nutrition to support the application of muscle power. Land-use choices were therefore a trade-oV not only between diVerent ways of feeding the human population, but also between diVerent modes of energy production (e.g. forestry for timber vs. pasture for draft animals); and these constraints set limits to the possibilities for growth, exacerbating the negative feedback constraints formulated by Malthus. The exploitation Wrst of coal and later of other fossil fuels enabled an escape from the Malthusian trap in which the inelasticity of land as a production factor determined a negative feedback between population growth and standards of living. Wrigley notes that the use of one ton of coal per year releases for other purposes about 1 hectare of 75 Frier 2001: 158. 76 Frier 2001. 77 Wrigley 2004; Lal 1998: 18–21. See also Scheidel 2004. 78 Wrigley 2004, esp. 29–35.

30

Alan Bowman and Andrew Wilson

forest otherwise needed to provide fuel.79 It is the new energy sources of fossil fuels (coal, and later oil/petroleum) which could release stored solar energy accumulated over millennia, that enable sustained, intensive, technologically based growth, sometimes called ‘Promethean’ growth because its basis is new means of producing Wre. By contrast, ‘Smithian’ growth is the only type of per capita growth we should expect to Wnd in pre-industrial, organic economies. That is not to deny the potential importance of technology in antiquity; merely to stress that sustained technological growth on the level of modern economies is a recent phenomenon. How, then, might one demonstrate, or even measure, per capita economic growth in the Roman world? There are two possible approaches. The Wrst is to identify the existence in antiquity of drivers of economic growth, to demonstrate that the conditions existed which would have made growth possible. Among those listed by Saller,80 one can identify archaeological evidence for: trade, intensiWcation of capital investment, improved technology, increased division of labour. The other factors—education and institutional attitudes and stimuli—are not so easily identiWable from the material remains, but may be approached to some extent through documentary and epigraphic evidence, especially in the case of education or rather literacy. The second approach is to identify symptoms that might suggest growth did indeed occur. Such indicators would include urbanization, import replacement, increased consumption, and higher standards of living.

IV.2. Trade and manufacture Increased trade contributes to economic growth through increasing the eVective size of markets reached by producers, enabling economies of scale and division of labour; and also by enabling distributed and more complex manufacturing, so that a wider range of goods may be produced at a given place using a mixture of local and imported materials or components. It also facilitates urbanization 79 Wrigley 2004: 39 n. 61.

80 Saller 2002: 261–2 ¼ 2005: 232.

Quantifying the Roman Economy

31

and the growth of cities beyond the capacity of their immediate hinterlands to supply them. Key issues to consider include the distances and distribution patterns over which goods were moved, and how these were aVected by the technologies and costs of maritime, riverine, and road transport; the extent to which the movement and exchange of goods represented market trade, or extraction as rents or taxes, or state-mandated supply of the court or the army. For many purposes it is likely to be more proWtable to break the somewhat vague concept of ‘trade’ down and view the phenomenon through the lenses of production, distribution, and consumption. Some of the methodological problems involved in assessing the scale and nature of ancient trade are discussed in the chapters by Wilson, Fulford, and Harris below.

IV.3. IntensiWcation of investment Investment in productive techniques or infrastructure is a critical driver of growth; but there is debate over whether it happened to any real extent in antiquity; and the question became one of the key issues in the debate about economic rationality. Literary and juridical sources tend to focus more on leasing and tenancy than on capital investment in the productivity of estates; documentary papyri from Egypt may, however, be read as reXecting capital investment.81 But the clearest evidence for large-scale investment in productive capacity is archaeological.82 It can be seen most obviously in the provision of batteries of olive or wine presses at large agricultural facilities in North Africa, Spain, and even Istria;83 in the large factories for Wsh-salting along the Iberian and North African coasts and in Brittany;84 in the adoption of water-mills and of new irrigation machinery;85 in the creation of irrigation schemes and channels;86 and in the creation of large-scale pottery production facilities (La 81 82 83 84 85 86

See Kehoe 1988; 1992; 1993; 1997; 2003; Rathbone 1991; 1994. Wilson 2008a, for a synthesis. Mattingly 1988a; 1988b; Wilson 2008a. Wilson 2006. Wilson 2002a; see also below, pp. 33–7. See also now the Hadrianic Lex Rivi Hiberiensis, Beltra´n Lloris 2006.

32

Alan Bowman and Andrew Wilson

Graufesenque, Le Rozier, Scoppieto), some of which seem to have been leased by landowners to potters.87 Investment in such largescale plant or in machinery improved productivity through economies of scale and more eYcient use of labour. But investment is present, and probably widespread, lower down the productive scale as well. The almost routine equipping of many villas in the western and central Mediterranean with one or more oil or wine presses is an indicator of this. Indeed, the very spread of the villa system, which underpinned the organization of much agricultural production at least in the western empire, represents an intensiWcation of investment in market-oriented agriculture.88 The analysis of some of these indicators of investment will be a core part of our work on the agricultural sector of the economy.

IV.4. Division of labour Little systematic research has yet been done on the extent of division of labour in the ancient world. However, plenty of evidence exists, and a study by Wilson focuses on the substantial archaeological evidence for division of labour in large-scale and mass production in the Roman period, notably in large urban bakeries and fulleries, in the large pottery production centres as at La Graufesenque, Le Rozier, and Scoppieto, and in the textile industry, where extreme division of labour (separate establishments at Pompeii for woolscouring, spinning, dyeing, weaving, and fulling) was coupled with vertical specialisation.89 Such division of labour should in Smithian terms reXect an increased size of market (and therefore increased trade), and should also achieve greater per capita labour eYciency in the sectors in which we can identify it. Attempts have also been made to quantify the evidence for urban craftsmen in Egyptian towns (and the comparison or contrast with rural villages is also relevant).90 The 87 Wilson 2008a. 88 Percival 1976; cf. for Italy, Purcell 1995 and Marzano 2007. 89 Wilson 2008a. 90 van Minnen 1987, using Weberian terminology and noting the estimate of 30–50% of the population of Byzantine Oxyrhynchus involved in urban crafts. It is important to bear in mind, however, that ‘urban craftsmen’ might also be involved to some extent in agricultural activity, if only as the small landholders who appear in land- and tax-lists and the like.

Quantifying the Roman Economy

33

results may tell us something about local supply and demand, measured against import over greater distances, and could help to reWne our notions of cities and towns as economic ‘producers’ or ‘consumers’. This is certainly an area where more research and systematic analysis could prove productive.

IV.5. Technology and development; archaeological and documentary sources The last twenty-Wve years have seen a radical overhaul of views on the level and importance of technological development achieved in the Roman world. From the 1960s to the 1980s the view prevailed that ancient technology in general, and Roman technology in particular, was stagnant and contributed little to the economy.91 Since then a number of studies have argued for a much higher level of ancient technological development, and a more rapid and widespread uptake of that technology.92 Ancient technology has re-entered the debate on the economy, and the task is now to assess what contribution technological developments might have made to economic growth. Our project aims to compare archaeological and documentary sources, and one of the things we might hope to get out of that is a better understanding of the nature and utility of those sources. This is illustrated in a striking way in a preliminary analysis of the uptake of certain mechanical technologies. Case study 1: The water-mill The water-mill is one of the earliest examples of human eVorts to harness natural forces to do mechanical work, and stands as the ancestor of a long line of machines. Water-milling greatly increased per capita productivity in the time-consuming and widely needed grinding of grain into Xour, and enabled greater specialization in labour, with the water-miller (molendinarius) emerging as a separate

91 Finley 1959; 1965, White 1962; 1980. 92 e.g. Wikander 1984; Greene 2000; Wilson 2002a; and the various papers in Lo Cascio 2006.

34

Alan Bowman and Andrew Wilson

Wgure from the miller-baker (pistor) of the earlier Roman period. The spread and uptake of the water-mill is therefore important to questions of economic development and the relationship of capital investment in technology to economic growth.93 ¨ rjan Wikander published a brief but important landIn 1984 O mark study which overturned the commonly accepted idea that although the ancient world knew the water-mill, it was not until the early Middle Ages that its use became common.94 He showed that the large number of references to early medieval water-mills in surviving documents is a function of the nature of our sources. Much surviving Graeco-Roman literature is epic, didactic or elegiac poetry, narrative history, tragedy, or philosophy, none of which is likely to spend much time discussing water-mills. The bulk of the references that we do have come from scientiWc, technical, or encyclopaedic writings, which are comparatively rare. The massive increase in the rate of references to water-mills in written documents of the fourth and Wfth centuries ad is a product of the introduction of new genres of writing—legal codes, hagiography, and monastic charters, all of which are much more likely to refer to such relatively mundane devices. By contrast, the bulk of the archaeological material known in 1984 came from the second and third centuries ad, when no literary sources refer to water-mills at all—a reXection of the relatively small number of texts that have survived from this period (Fig. 1.1). Pictorial representations of water-mills are very rare— only two. In the 22 years following Wikander’s publication, our knowledge of ancient water-mills increased dramatically (Fig. 1.2). Some earlier written evidence dating back to the third century bc has now been identiWed, but while otherwise the number of known literary sources remains similar and the number of iconographic representations the same, the numerous archaeological discoveries have changed the overall picture. The number of archaeologically known sites has increased threefold. The second- and third-century ad peak is now much more pronounced, with appreciable numbers of water-mills 93 On the origin and spread of water-mills, see e.g. Wikander 1980; 1984; 2000; Lewis 1997; Brun and Borre´ani 1998; Wilson 2002a; Brun 2006. 94 Wikander 1984.

Quantifying the Roman Economy

35

25

20

15

10

5

0 1st c. BC 1st c. AD 2nd c. AD 3rd c. AD 4th c. AD 5th c. AD 6th c. AD 7th c. AD Archaeological finds

Ancient representations

Literary sources

Fig. 1.1. Number of attestations of water-mills by century in archaeological Wnds, ancient representations, and literary sources, as known in 1984 Source: Data in Wikander 1984.

25 20 15 10 5

AD

om an

c. th

10

h 9t

R

AD

AD

Ancient representations

c.

AD

8t

h

c.

AD

c. h 7t

AD

6t

h

c.

AD

c. h 5t

h

c.

AD

3r

d

c.

c. d

2n

Archaeological finds

4t

AD

D

C

.A

.B 1s

tc

BC

tc

1s

c. d

2n

3r

d

c.

BC

0

Literary sources

Fig. 1.2. Number of attestations of water-mills by century in archaeological Wnds, ancient representations, and literary sources, as known in 2006

36

Alan Bowman and Andrew Wilson

now apparent also in the Wrst and fourth centuries ad. The slight rise in the seventh century is principally due to a number of discoveries of wooden tidal mills in Ireland. The Roman-period peak would appear even sharper still if it were possible to distribute the six sites in the far right-hand column of Fig. 1.2, which are dated only as ‘Roman’ (in these cases, probably Wrst century ad to fourth century ad), among the centuries in which they really belong. Two conclusions follow from the diVerent pictures presented by the documentary and archaeological sources. First, the documentary sources do not by themselves provide a guide to the relative frequency over time of the use of water-power. The archaeological evidence oVers the best potential for assessing the comparative use of water-power over time, as it is the category least aVected by biases in the evidence. Secondly, the large number of documentary references to water-mills in the early Middle Ages, a time when the archaeological record is sparse, indicate that the archaeologically identiWed record must represent only a very small fraction of the original total. Case study 2: Water-lifting machines Interestingly, a similar picture is suggested by the literary and archaeological evidence for water-lifting devices (Fig. 1.3)—similar both in terms of the diVerence between the two categories of evidence, and in terms of the archaeological peak in the Wrst to third centuries. For this to be useful, we need to do more work on breaking the graph down by category of device—bilge pumps, irrigation devices, mine drainage, supply for urban baths, etc., and also to see how the literary and archaeological evidence compares with the papyrological evidence which is not graphed here. But this is work in progress; at the moment, we may note a correlation between uptake and use of—for the time— complex technology and the period most usually assumed to represent the height of the empire’s prosperity, in the Wrst to third centuries ad. What did such technological development mean in economic terms? Saller estimated that the adoption of animal- and waterpowered mills could at most have contributed to a growth at a rate of 0.025% per annum, and regards this as unimportant.95 However, 95 Saller 2002: 265 ¼ 2005: 235.

Quantifying the Roman Economy

37

70 60 50 40 30 20 10

Literary references

AD

h

c.

AD

7t

h 6t

c.

h

c.

AD

AD

5t

h

c.

AD

4t

c. 3r d

c. d 2n

tc

.A

AD

D

C

1s

.B tc 1s

c. d 2n

3r d

c.

BC

BC

0

Archaeological sites

Fig. 1.3. Number of attestations of water-lifting devices by century in archaeological Wnds and literary sources, as known in 2006

since he guessed that the growth postulated by Hopkins might amount ‘to perhaps as much as 25%’, which he considered to be spread over three centuries and ‘which would amount to less than 0.1% per year’ (the true Wgure would actually be 0.0747% compound per annum),96 the implication is that the adoption of new milling technologies alone could have contributed up to a third of the total annual growth of the empire. Both Wgures, 0.0747% and 0.025%, result from highly dubious guesses and could be contested on any number of counts; the point is simply that on Saller’s own Wgures the possible maximum contribution of milling technologies seems staggeringly important to the economy as a whole. Add to this technological improvements and 96 Saller 2002: 259–60 ¼ 2005: 231 confusingly spreads this 25% growth over three centuries in his text (thus equating to 0.0747% compound p.a.), but the graph (Saller 2002, Wg. 12.2 ¼ 2005, Wg. 11.2) which he purports to construct on the basis of Hopkins’s 1995/6 article (with just four data points!) actually shows this growth occurring over only the last two centuries bc, which would equate to 0.1125% p.a. The graph in any case has no value as evidence.

38

Alan Bowman and Andrew Wilson

innovations in mining, agriculture (notably irrigation devices), and the building industry (ubiquitous use of cranes, the development of labour-eYcient modes of construction using bricks and concrete), and it looks increasingly diYcult to deny the importance of technological developments to the achievement of per capita economic growth.

IV.6. Education of the workforce Investment in human capital—the skills and knowledge of the workforce—is today recognized as a major factor in intensive growth since the twentieth century. To what extent could this have been at all important in antiquity? Evidence for ancient education in general is poor, and tends to relate principally to the elite; direct evidence for education of the workforce is rare, and it is often assumed that ordinary workers were uneducated. However, occasional but unquantiWable hints suggest that the issue is at least worth considering. The Lex Metalli Vipascensis, regulating a mining community at Vipasca (Aljustrel, Portugal), contains a provision exempting schoolmasters from taxes levied by the procurator of the mine;97 they can hardly have taught anyone other than the children of miners, smelters, and the workforce engaged in ancillary services to the mining settlement. Overall, our best hope of approaching the issue in any quantiWable way is by looking at literacy. The positive role that literacy will have played in connection with the economy is well described in general terms by Keith Hopkins: The productivity of labour grew, as did total population and aggregate product. The surplus grew. There was a greater division of labour and more and bigger towns. The growth in literacy was both a consumption good—a way of integrating more people within a larger society—and a necessity. A larger-scale society needed (or operated better with) more writing.98

97 Lex Metalli Vipascensis 8; Domergue 1983: 56–7, 98–9. 98 Hopkins 1991: 136.

Quantifying the Roman Economy

39

Literacy rates in antiquity are notoriously diYcult to determine, though Harris doubts that levels reached even 5–10%.99 One might expect literacy rates among non-agricultural workers, however, to be higher than in the agricultural sector. Suggestive evidence is provided by the characteristically Roman habit of stamping certain products— bricks, transport amphorae, table pottery, lead pipes, even pastries or loaves of bread. Since brickstamps and stamps on amphorae and other pottery were applied before Wring, they must relate to the organization of production. In many cases it is likely that they may relate to locatio/conductio contracts between a landowner and a workshop manager who may be renting the production facilities. Similarly the charge-lists of La Graufesenque (bilingual in Latin and Gaulish), itemizing the numbers of vessels produced by individual potters which were loaded into massive shared kilns for Wring, are documents reXecting cooperation and division of labour between potters and kiln masters in a large-scale industry.100 Literacy here enables a more complex and sophisticated production system with greater specialization of labour. The overall frequency of stamping bricks reXects the general fortunes of the building industry as a whole in the area around Rome—epigraphic stamps start in the Wrst century ad, increase in quantity and frequency (and complexity) in the second century, and then stop abruptly after the Severans, with simple anepigraphic stamps only in the middle of the third century. Epigraphic stamps at Rome reappear with Diocletian’s reorganization of the brick industry, and continue in the west until Constantine, after which they stop. In the east, at Constantinople the practice of stamping bricks commenced in the late fourth century ad and continued until the early seventh century, again closely connected with the intensity of public building there.101 We should also consider two further aspects of literacy and documentation. The Wrst is the eVect of the bureaucratic practices

99 Cf. Harris 1989 and the various papers responding to him in Humphrey 1991; also Bowman and Woolf 1994. 100 Wilson 2008a; for the La Graufesenque documents, see Marichal 1988; Flobert 1992; Vernhet and Be´mont 1991; 1993. On the language of the documents, see Adams 2003: 687–724. 101 Bardill 2004.

40

Alan Bowman and Andrew Wilson

imposed by government through the mechanisms of record-oYces and the like. We might reasonably suppose that the documentation of economic activities enabled the government to reap Wscal beneWt in the form of higher taxation and that this would be a stimulus to higher productivity and economic growth. In some areas, such as Egypt, these practices were already widespread when the Romans annexed but they may well have intensiWed under Roman rule;102 in others they were no doubt introduced, as was the provincial census. The ceremonial burning of records of debt which took place under the eye and the presidency of the emperor Hadrian was heavily symbolic, but its economic implications were presumably the opposite of stimulating.103 The second point focuses more directly on quantiWcation. We can get a good sense of the great quantity of such archival material in a far smaller place, the Egyptian village of Tebtunis in the Arsinoite Nome for which we have abstracts of contracts deposited in the registry (grapheion) during speciWc years of the reign of the emperor Claudius (the last four months of ad 41/2, the whole of ad 45/6 and the Wrst four months of ad 46/7). These records show the care with which contracts were summarized and catalogued and yield, by a simple calculation, the mean volume of such contracts (most in Greek but some few in demotic Egyptian), which is 58 per month for the periods represented.104 These are not simply dead letters but are documents which were frequently consulted and played a large role in the regulation of the social and economic relations of the villagers. A further simple calculation reveals that over the course of a single year about 350 diVerent people, male and female, appear as parties to written contracts. At a rough estimate this is likely to constitute about 20% of the adult 102 See e.g. the edict of the provincial governor Mettius Rufus, issued in ad 89 (P.Oxy. II 237.27–34): ‘Claudius Arius the strategos of the Oxyrhynchite Nome has informed me that neither private nor public business is receiving proper treatment owing to the fact that for many years the abstracts in the property record-oYce have not been kept in the manner required, although the prefects before me have often ordered that they should undergo the necessary revision, which is not really practical unless copies are made from the beginning. Therefore I command all owners to register their property at the property record-oYce within six months, and all lenders the mortgages which they hold, and other persons the claims which they possess.’ 103 Boatwright 1987, pl. 41. 104 P.Mich. II 121, 123, P.Mich. V 238. For a detailed analysis see Toepel 1973, ch. 2.

Quantifying the Roman Economy

41

population of the village. The importance of this Wgure is not so much as a demonstration of rates of literacy, for we do not of course know how many of these individuals wrote or could have written their own contracts, but as an indication of the minimum proportion of people whose socio-economic relations in so-called ‘private’ aVairs were recorded and regulated by bureaucracy and written documentation. We cannot conclude that these were just people of higher social status for about one-third of the individuals have Greek names (generally an indication of higher social status) and about two-thirds Egyptian names (generally an indication of lower status). The direction in which this argument leads (which will need further detailed exposition) is that the individuals who document their economic activities either themselves or vicariously, are likely to be more economically productive. If this amounts to around one Wfth of the adult population of a Wrst-century Egyptian village, it renders the assertion of low literacy rates much less signiWcant. It is more diYcult to judge possible Xuctuations in levels of literacy over time.105 It is tempting to compare the Leuven Database of Ancient Books,106 which lists instances of surviving copies of literary texts in Egypt between the third century bc and the eighth century ad, and shows a marked peak in the second and third centuries ad in the sample of 14,246 texts on papyrus and parchment (Fig. 1.4). This of course represents something more than basic literacy, and may also be held in part to reXect variations in expenditure on literature. More seriously, however, it broadly corresponds to the survival rates of all papyri over time and may also be biased by Wnds at sites such as Oxyrhynchus where literary texts are for whatever reasons very heavily represented.107 A better sense of the relationship between literary texts and documentary material is oVered by the evidence from the Fayum villages and other sites but this also has to be assessed against the survival patterns of the Greek papyri as a whole.108

105 Harris’s view (1989) of a decline in late antiquity has not found general acceptance, see below, n. 140. 106 http://www.trismegistos.org/ldab, accessed on 12 Aug. 2007. 107 See Habermann 1998. 108 Clarysse 1983; van Minnen 1998.

42

Alan Bowman and Andrew Wilson

3000 2735

2500 2215

2000 1656

1500 1079

917

1000

810 774 713 616

481

AD

h 8t

c.

h 7t

c.

AD

AD

h

c.

AD

h

c.

AD

c.

5t

h

c.

AD

4t

AD

3r d

2n

d

c.

.A

D

C

tc

.B 1s

tc

Texts on papyrus

150

66

13

3

4

BC

d

c.

c.

BC

2

2n

3r d

292

293

1

0

643 461

6t

322

1s

500

Texts on parchment

Fig. 1.4. Chronological distribution of surviving copies of literary texts on papyrus and parchment (n¼14,246) Source: Leuven Database of Ancient Books.

IV.7. Institutional incentives and stimuli Institutional incentives to investment or to greater economic eYciency are likely to prove the most diYcult to quantify. The evidence for them is exclusively documentary in forms which are not very amenable to quantiWcation—chieXy in juridical writings, or anecdotally in literary sources. Nevertheless, we can at least demonstrate the existence of some incentives—the various state encouragements, through exemptions from munera, to merchants who put ships of a certain capacity at the service of the annona.109 Other examples are the Wrst-century ad Lex Manciana, governing share-cropping so long as estates were continuously cultivated, and the Lex Hadriana de rudibus agris, encouraging the development of marginal land on imperial estates in North Africa by exemptions from rent on land newly brought under cultivation until the olive or fruit trees planted had come to maturity.110 Field survey around 109 Claudius: Gaius, Inst. 1.32c. Second century ad: Dig. 50.5.3, Scaevola. See Wilson, Ch. 9 below, n. 29. 110 CIL 8.24953 (Aı¨n el-Djemala); cf. Kehoe 1984; 1988: 61–3. The Romans were not, of course, alone in such measures; there are institutional incentives for the

Quantifying the Roman Economy

43

Kasserine in Tunisia has demonstrated the intensive terracing and cultivation of marginal land between the Wrst and Wfth centuries ad which looks like a response to these sorts of legal incentives coupled with market opportunity.111 Also relevant is the decrease in direct ownership and management of land by the imperial authority in the later empire.112

IV.8. Symptoms of growth In addition to the drivers of growth, we can also seek indicators that may be a reXection of economic growth, or symptoms that it was occurring. These would include: increasing levels of urbanization, whose value as a proxy measure for growth is discussed by Lo Cascio in this volume; signs of import replacement as local provincial economies develop and begin producing goods that were formerly imported (see Wilson’s chapter on trade); increases in the consumption of material goods; and standards of living (discussed below in the chapters by Rathbone, Scheidel, and Allen). There is much current interest among economic historians of medieval and later periods in comparing living standards across diVerent pre-industrial societies,113 but ancient historians and archaeologists have only recently begun thinking about how to apply such analysis in any systematic way to the classical world. One approach has been suggested through analysis of house sizes and house contents, but this suVers from a number of limitations.114 Standards of living might be assessed through consumption patterns—reviews have recently been attempted by Jongman and Ward-Perkins, the latter contrasting the material culture of the Roman empire of the Wrst to third centuries ad with that of early medieval western Europe.115 Biological standards of living oVer excellent potential for comparison over long time periods, development of uncultivated land in the Achaemenid empire, where the construction of qanats in the desert zone at the foot of the Taurus mountains is said to have been fuelled by the grant of land for Wve generations to anyone who invested in the necessary irrigation technology (Polybius 10.28). 111 Hitchner 1988; 1990; 1993. 112 Rowlandson 1996: 70–101. 113 For a collection of recent work, see Allen, Bengtsson, and Dribe 2005. 114 e.g. Morris 2005 for the Archaic to Hellenistic Greek world, with discussion of methodological limitations. 115 Jongman 2007; Ward-Perkins 2005.

44

Alan Bowman and Andrew Wilson

as the systematic collection of osteological data from burials allows an assessment of variation in stature as a reXection of lifetime nutrition.116 In addition to large-scale anthropometrical projects attempting such comparisons over several thousand years of human history, a project directed by Jongman is analysing femur length over the Roman and late Roman period. Early results from this show a dramatic increase in average femur length for skeletons over the course of the Wrst century ad, with a signiWcant drop again in the second century, a slight recovery in the third and early fourth centuries, and a steep decline again in the later fourth and Wfth centuries.117 Further ways to compare standards of living involve comparisons of wages and prices, as the papers by Rathbone, Allen, and Scheidel in this volume discuss. Allen’s approach in particular lends itself to crosscultural comparisons, by establishing a ‘basket’ of prices which can be compared to wages. On his measure, the Roman labourer of the early fourth century ad earned a wage whose purchasing power compares with that of workers in eighteenth-century Europe and Asia, but was behind that of Wfteenth-century Europe (a period of high real wages).

IV.9. Contextualizing growth in the longue dure´e Our hope is that by the accumulation of various indices of the kinds discussed above, we might be able to produce a more robust idea of the shape of likely growth in the Roman world than, say, the highly impressionistic graph in Saller’s article on economic growth, which appears to be based on wholly spurious premisses.118 Saller’s source, Lucas, attempted to assess the impact of the Industrial Revolution, comparing it to the long history of pre-industrial societies that had gone before.119 As is not uncommon in works of economic history covering a long time span, the supposed data on ancient economic 116 Komlos 1994; Steckel and Rose 2002; cf. Koepke and Baten 2005; Kron 2005 with comments on methodology in Jongman 2007: 194. 117 Jongman 2007: 193–5. 118 Saller 2002, Wg. 12.1 ¼ 2005, Wg. 11.1. 119 Saller 2002 ¼ 2005 cites a working paper by Lucas as his source for the graph. We have not been able to trace the work referred to, but an internet publication with the same title is presumably much the same as the work cited by Saller: Lucas 2004 (http://minneapolisfed.org/pubs/region/04-05/essay.cfm#Wg2).

Quantifying the Roman Economy

45

performance are found to be shaky when interrogated. Lucas had no reliable Wgures for either GDP or population for the Wrst millennium ad, so it is unclear how he thought he was calculating GDP per capita. It appears that he assumed that all pre-industrial societies were pretty homogeneous and that per capita income could Xuctuate only within narrow bands. His paper makes a number of assumptions, among which are that all pre-industrial societies had a similar level of per capita income, and did not experience signiWcant growth: For poor societies—all societies before about 1800—we can reliably estimate income per capita using the idea that average living standards of most historical societies must have been very near the estimated per capita production Wgures of the poorest contemporary societies. Incomes in, say, ancient China cannot have been much lower than incomes in 1960 China and still sustained stable or growing populations. And if incomes in any part of the world in any time period had been much larger than the levels of the poor countries of today—a factor of two, say—we would have heard about it. If such enormous percentage diVerences had ever existed, they would have made some kind of appearance in the available accounts of the historically curious, from Herodotus to Marco Polo to Adam Smith.120

This latter point is optimistic to the point of naivety. Ancient authors writing between Herodotus and Marco Polo were notoriously uninterested in standards of living; and our sources are in any case scanty and disjointed. Their silence means very little. Moreover, a recent survey shows that incomes in pre-industrial societies could in fact vary from near-subsistence to about three times subsistence, thus allowing the potential for 100% more growth than Lucas would admit.121 The guess that growth over the period 200 bc to ad 100 might have amounted to some 25% is just that—a guess.122 120 Lucas 2004 (see above, n. 119). As it happens, Adam Smith does in fact stress the vast disparities possible between diVerent societies in standards of living at the bottom of the social scale: ‘and yet it may be true, perhaps, that the accommodation of an European prince does not always so much exceed that of an industrious and frugal peasant, as the accommodation of the latter exceeds that of many an African king, the absolute master of the lives and liberties of ten thousand naked savages’: Smith 1776, ch. 1 ¼ 3rd edn., 1811: 17. 121 Jongman 2007: 185, based on papers in Allen, Bengtsson, and Dribe 2005; several series show at least a threefold variation in real wages; cf. Allen in this volume. 122 Saller 2002: 265 ¼ 2005: 235.

46

Alan Bowman and Andrew Wilson

Moreover, the idea of a linear progression in aggregate (but not per capita) growth is apparent in the following passage, conceding some technological improvement: our increasing mastery of our environment is reXected in accelerating population growth over the centuries. Between year 0 [sic!] and year 1750, world population grew from around 160 million to perhaps 700 million (an increase of a factor of four in 1,750 years). In the assumed absence of growth in income per person, this means a factor of four increase in total production as well, which obviously could not have taken place without important technological changes. But in contrast to a modern society, a traditional agricultural society responds to technological change by increasing population, not living standards.123

We might question: (1) the lack of growth in per capita income in the Roman period; (2) the implied idea that population growth up to 1750 was steady and uniform—in the Mediterranean and north-west Europe there are, to the contrary, signs of growth in the Roman period followed by shrinking in the early medieval period (and again in the fourteenth century as a result of the Black Death), followed by growth again;124 (3) the notion that increased population and higher living standards are alternative responses to technological change in a traditional agricultural society. This model suggests sharper periodic growth rates sustained over two to three centuries, and also decline at other periods. We could not currently exclude the possibility that per capita income in the second century ad might have been double what it was in the sixth century ad in north-west Europe.

V. DECLINE To what extent can decline simply be measured by the absence or reversal of the indicators of growth discussed above?125 This would 123 Lucas 2004. 124 Russell 1958; 1972; Wickham 2005: 547–50. 125 For a later period, cf. Wilson 2006 on construction techniques and the collapse of the building trade in the late antique west. In general we are not addressing the issues presented and analysed by Ward-Perkins 2005 and Wickham 2005 which concern the period after ad 400.

Quantifying the Roman Economy

47

be a relatively straightforward approach. Without making any assumptions as to what the answers might be, we can at least identify what we would be looking for and this would certainly go some way towards answering the question. Whether it would provide a complete answer will need to be assessed in the light of considerations about ‘steady state’ and the balance of change in diVerent regions at diVerent times. Although most readers will naturally assume that in dealing with ‘decline’ we are looking at the period after c. ad 200, we should note, for instance, the implication of the view that the major period of growth was in the second to Wrst centuries bc, with only limited growth in the principate, the implication of which is that it may have ceased before ad 200.126

V.1. Trade and manufacture The following bleak statement by Amanda Claridge about the marble trade is representative of widely held views: The heyday was over by the middle of the third century. The market shrinks, the quarries wind down; highly experienced sculptors become fewer and fewer; the accumulated skills of thirty generations of unbroken tradition are less and less in demand. The marble was still there, stock-piled in Rome and abroad, and stock-piled in the quarries as well. But there were also armies of statues ready made to be re-used . . .127

We might be able to demonstrate a decrease in the volume and value of goods being made and traded; a decline in the quality of the manufactured objects (which might indicate a stagnating or declining technology and also a reduction in ‘standards of living’?); a reduction in the geographical range of trading, with shrinkage of long-distance trade, although the wide distribution in the east Mediterranean of African manufactured goods in the third and fourth centuries ad is well recognized. But there are complications, discussed further in Wilson’s chapter on trade in this volume. The disappearance of imports is not a straightforward indicator—import replacement by local production may be a sign of local economic 126 Morley 2007: 98.

127 Claridge 1988: 152.

48

Alan Bowman and Andrew Wilson

growth (see above, p. 43); but if imports are not replaced, and overall consumption levels drop, this would suggest decline.128 Equally, while an increase in the number of known shipwrecks between the second century bc and Wrst century ad probably does reXect an increase in trading activity in this period, a variety of factors including the increasing use of barrels as containers means that a stagnation and then decrease in these numbers thereafter does not necessarily translate simply into a parallel reduction of trading activity. Depending on what the evidence shows, we might deduce ‘decline’ at least as measured in terms of aggregate value of economic activity; measuring that in per capita terms is, of course, subject to the uncertainties in the demographic picture which we must constantly bear in mind.

V.2. Stagnation of capital investment Stagnation or decline in the level of investment and in productivity can be sought both in agriculture and in manufacture (see above). For the former, the interlinked factors which we might consider would include the amount of land under cultivation and its productivity: what, for example, would be implied by a shift in the balance of cereal crops and viticulture,129 how signiWcant is the phenomenon of declining rural communities, agri deserti and its distribution?130 Is there a widening gap between rich and poor which would aVect investment (making the wealthy wealthier stimulates investment?), forcing out the self-suYcient middle cohort, stimulating a move towards the ‘colonate’ with a reduction in the amount of slave labour (which might be either cause or eVect), a decrease in division of labour?131 And if these are real phenomena, in any way measurable, are they necessarily constituents of or proxies for decline?132

128 Import replacement, Woolf 1998: 193–202; Morley 2007: 98. 129 Bagnall 1993: 32; for a later period cf. Banaji 2002: 18–19. 130 Whittaker 1993, ch. 3; Bagnall 1985. 131 Bowman 1985; see Bowman in this volume. 132 Wickham 2005: 411–28 discusses the later evidence for Aphrodito and Jeme, noting the diVerent patterns in diVerent Egyptian villages.

Quantifying the Roman Economy

49

V.3. Institutional stimuli and attitudes Institutional stimuli and attitudes are key factors which have to be contextualized in our general notions of the political and economic character of the later empire. It seems obvious enough that if the creation of military security and the encouragement of urbanization in the earlier empire encouraged economic growth, the state could have failed to continue or could have reversed these processes by choice or necessity, largely the latter one supposes, in the case of military security. The power of Wscal controls, particularly changes in the methods and levels of taxation and the monetary policy, or better ‘strategy’, will have had profound eVects (higher taxes, more requisitions, changes in methods of collection, more in kind and less in cash, etc.) and the state could also have suppressed or reduced free market enterprise by other direct means (legislation aVecting trade associations etc.). Thus, set against the picture of the prosperity of the ‘high’ empire to c. ad 200, we have the supposed Diocletianic ‘recovery’, a more oppressive Wscal regime, a free population overburdened by tax demands largely generated by the need to maintain an overblown military and bureaucratic establishment and collected via the wealthy landowners.133 On the other hand, it may be that the burden of taxation was systematized rather than increased, that the state machinery became less directly involved in the major productive processes and that, at least in some regions, communal responsibility for ensuring key socio-economic activities was more marked (in contrast to the earlier culture of public service and euergetism), if not necessarily easy to put into eVect.134

V.4. Specialization of labour and education of the workforce One could imagine a measure of the decrease of diVerent types of productive labour in urban centres as demonstrated by ‘industrial’ sites, Wtting a picture of a return to reliance on local supply and a 133 This more or less standard view is summarized by Tate 1992: 1, but it can be found in many scholarly accounts. 134 Lepelley 1979, vol. 1: 59–73.

50

Alan Bowman and Andrew Wilson

shrinking communication network. The latter would have to include some estimate of a declining standard of literacy and education which, as already noted, is diYcult to achieve in quantiWed terms. Harris’s views of the constraints imposed by technology and of a declining literacy in the early Christian period have not proved very persuasive and are particularly hard to uphold for the eastern empire.135 However, if we look at stamps on Dressel 20 olive oil amphorae from south-west Spain during the Wrst to third centuries ad, we note in the third century a much higher incidence of stamps with the occasional retrograde letter in an otherwise left-to-right text. This would appear to reXect a decreasing competence in writing on the part of the potters or workshop managers producing the stamps.136 Then, after the cessation of production of the Dressel 20 form c. ad 267, its replacement, the Dressel 23, does not carry stamps, suggesting a simpler and smaller-scale organization of production. Although there may not be a direct causal connection, there does appear to be a broad correlation between a decrease in the volume of Spanish oil exports and the use of literacy in the bottling plants for this oil.

V.5. Urbanization, demography, and settlement patterns Here we might investigate the possibility of decreasing or shrinking urban centres, the counterpart of which might or might not be an increase in rural settlement which we would need to take into account in any estimate on an overall change in population levels. It is likely that such macro-estimates will prove elusive, but we might hope to gain some idea of the sizes and survival of settlements over time in selected areas where there have been reasonably systematic surveys. If those proved real, would we regard them as indicators of decline? The logic of taking urbanization as a proxy for economic growth would suggest that we ought to, but changes in settlement patterns and/or shrinkage of population are not necessarily incompatible with a steady state or 135 Harris 1989: 331 and ch. 8; Bowman 1991; Hopkins 1991. For urban artisanal activity in Egypt see van Minnen 1987 and cf. lists of artisans at Aphrodito cited by Wickham 2005: 413 with other material in Johnson and West 1949: 107–67. 136 E´tienne and Mayet 2004, vol. 1: 91.

Quantifying the Roman Economy

51

increased prosperity, which might be postulated for some areas in the third and fourth centuries, notably Syria.137 The building industry provides a perspective here: at Rome there are almost no large-scale building projects between the reigns of Alexander Severus and Diocletian, with the notable exception of the Aurelian Walls. In the provinces, public building activity (whether funded by imperial benefaction, municipal funds, or private euergetism) declines sharply after Alexander Severus for the rest of the third century.138 In view of this, it might seem somewhat paradoxical that at least in Egypt and Asia Minor there are clear signs of communal civic pride, matched by some expenditure, in the second half of the century.139 Even if these attempts at measurement were per impossibile to prove robust and decisive, they would still not encompass the full complexity and nuances of the processes under scrutiny in the period up to c. ad 350. It might be possible to agree with a straightforward assertion of decline, at least in the west, from c. ad 400 onwards, but in contrast to an earlier communis opinio of decline from c. ad 200, there is now much less of a consensus about ‘crisis’ and more of an inclination to emphasize changes and regional diVerences. In conformity with our general approach to the subject and the evidence, it seems reasonable to state the premiss that there was considerable regional diversity across the Roman Mediterranean, the nature and extent of which we would hope to illuminate by quantitative methods. Some positive results emerging from such investigations would enable us to test, at least for the earlier part of the period, current notions about the decline of the empire such as the cautiously expressed view of McCormick: The overall economic trend of the Roman world from c. 200 to 700 was downward. This is not to say that decline prevailed everywhere, all the time: far from it. But within these chronological limits, the overarching pattern is now clear, even if the details are sometimes sporadic and even contradictory.140

The notion of a drastic empire-wide economic decline in the third century (where, incidentally, the Egyptian evidence has played a 137 138 139 140

Tate 1992. Fentress and Perkins 1988, Wg. 4; cf. Wilson 2007b. Bowman 2000, contrast van Minnen 2002; Mitchell 1999. McCormick 2001: 30.

52

Alan Bowman and Andrew Wilson

central role in the collapsing economy view even for those who would insist on Egypt’s atypicality) and its antithesis might now seem like straw men, but the former is tenacious and has often been generalized on the basis of insuYcient detailed evidence. Central to it is the undeniable notion of collapse and debasement of the imperial currency and the more debatable one of rampant inXation.141 For Egypt, at least, we now have the beneWt of some up-to-date analysis of series of price data which shows convincingly that the ‘price-inXation’ of the later third century can to a signiWcant extent be explained by re-monetization.142 This removes our reliance on an oft-cited Oxyrhynchus papyrus of ad 260 which has been thought to show the government refusing to accept its own coinage as a sign of a collapsing currency.143 We also need to take into account the evidence, admittedly patchy and by no means totally unambiguous, of a certain vigour in civic aspirations and culture in some of the Egyptian metropoleis in the later third century which suggests that it is too simplistic to characterize the period as one of ‘economic collapse’.144 On the other hand, it is reasonably taken as a clear indication of economic trouble that in Africa and the western provinces new public building programmes decline dramatically after Alexander Severus until well into the fourth century (and even in Africa a very large number of the fourth-century building inscriptions probably record repairs in the wake of earthquakes in the 360s).145 These somewhat inconclusive indications suggest two further questions which we should address. We have a signiWcant quantity of dated documents and of other evidence, particularly numismatic, for the period c. ad 260–90 which reXect both communal (particularly at Oxyrhynchus and Hermopolis) and private economic activity in that period. They ought to reveal any signiWcant changes in patterns of economic behaviour on a communal or private level and thus suggest whether we are right to envisage a degree of abnormality. An answer to that question will form at least 141 See Rathbone 1996a. 142 Rathbone 1997, improving on Drexhage 1991. 143 P.Oxy. XII 1411. 144 Above, n. 139. 145 Fentress 1981: 199–201; Wilson 2007b: 293. But see Lepelley 1979 on the postDiocletianic prosperity and building in the fourth century (I, 59–120), although, curiously, he seems to ignore seismic activity in the 360s as a factor in the peak in repair inscriptions under Valentinian I (364–75).

Quantifying the Roman Economy

53

part of our characterization of the ‘third-century crisis’ by indicating whether or not people behaved in ways which suggest they believed that their prosperity or subsistence was in jeopardy because of systemic or structural collapse (rather than incidental or occasional diYculty). If we can do this, however incompletely or provisionally, for Egypt, we will still be aware of the need to show whether or how we can extrapolate to other areas. At the very least, greater scepticism about this kind of generalization is required. The second question is related, but looks ahead to the longer term. If the economy ‘recovered’ after c. ad 300 what was the basis of that recovery and was it equally eVective everywhere? One aspect of this will be the re-stabilization of the currency, on the basis of a reformed gold coinage, which was clearly eVective to a signiWcant degree and (amongst other important features) terminated the use in Egypt of the Alexandrian-minted tetradrachma. There were clearly hugely important changes in the ways in which Wscal administration, with stimuli and deterrents, operated after ad 300. Some studies of Egypt from the fourth century onwards have emphasized that this period laid new and diVerent foundations for agrarian vitality and prosperity at least in the east,146 others more negatively that it did not see the creation of conditions for a ‘Byzantine feudal system’.147 Re-analysis of shipwreck data also suggests a slight—and unsustained—rise in the number of known wrecks in the early fourth century, which may reXect some recovery in levels of trading activity (see Wilson’s chapter on trade in this volume). Our perspective here will not take us beyond the middle of the fourth century, but there is still a good deal of useful evidence available which may well make us feel uncomfortable about postulating universal decline in the period ad 300–50, whatever we may think ensued after the beginning of the Wfth century.

V I. A RE A S F O R A NA LYS IS The four main areas in which it seems to us most fruitful to attempt to quantify and compare both documentary and archaeological 146 Banaji 2002, see Kehoe 2003.

147 Bowman 1985, cf. Bagnall 1992.

54

Alan Bowman and Andrew Wilson

evidence for the Roman economy are demography (including settlement patterns and urbanization); agriculture; production and trade; and the relationship between mining, metal supply, and coinage. Within these areas we should be able to look for the indicators of integration, growth, and decline discussed above.

VI.1. Demography, settlement, and urbanization ‘The primordial element is population size, because in a pre-industrial society, the aggregate amount of wealth produced is directly connected with the numbers of people working.’148 Size, distribution, and movement of population are clearly crucial to economic performance, especially in relation to the ability of the land to support the population and the scale and concentration of ‘production’ within settlements. The question of economic growth and contraction is closely tied to questions of demographic change; and the question of per capita growth is closely connected with the proportion of the population engaged in non-agricultural activities, for which a proxy may be urbanization rates. We recognize the force of these observations but, as anyone familiar with the spate of recent studies on the demography of the Roman world will realize, we are in severe diYculty from the outset. We do not debate here the plausibility of the Beloch/Brunt estimates of a total population of 55/60 million for the whole empire in the Augustan period, which many still regard as the best guess-estimate, and their close relatives for the total size of the high empire population, but note only the continuing vitality of the debate about widely varying high and low counts for various parts of the empire at various times (notably Italy and Egypt, where no consensus has emerged) and the fact that what some scholars have proposed as ‘least worst’ hypotheses still produce major implausibilities and inconsistencies even where we do have some Wgures and (at least in the case of Egypt) one methodologically exemplary study for an area which has very rich (by ancient standards) evidence.149 For rural 148 McCormick 2001: 30. 149 Morley 2001; Scheidel 2001a; Lo Cascio and Malanima 2005; Kennedy 2006; for Egypt, Bagnall and Frier 1994. The Wgures given by Josephus for the total

Quantifying the Roman Economy

55

areas, large-scale reliable statistics of any kind are notoriously diYcult to obtain for anywhere except Egypt.150 Some well-informed discussion of these diYculties and possible approaches (hardly solutions) to them are included in this volume (Fentress, Lo Cascio), but we are not optimistic about the possibility of obtaining a signiWcant consensus or a Wrmer foundation than exists at present, nor are we enthusiastic about simply constructing alternative scenarios on more or less Ximsy premisses and testing them by the relative strength or weakness of the consequences which Xow from them. We can neither oVer a top–down approach of this sort, although it is diYcult to avoid being inXuenced to some extent by the Wgures which are routinely bandied about; nor could we possibly hope to reach a reliable conclusion by aggregating Wgures (even if we had them) for all the constituent parts of the empire over the whole period. What, then, can we do? The contributions of Fentress and Lo Cascio to this volume, and the associated discussions, indicate the importance we attach to estimating the levels of urbanization across the empire and attempting to relate survey data to population trends. Neither of these approaches is straightforward. Apart from the variability in the quantity and reliability of the evidence, there are fundamental issues to be resolved. On urbanization: what thresholds do we envisage (and are they the same for all places?—evidently not) in deciding whether a settlement is urban or not, in the grey areas populated by what might with equal plausibility be described as small cities or large villages? Criteria (none of them hard and fast) include size, legal status, institutional and physical infrastructure. It is interesting, and signiWcant not only in social, but also in economic terms, that a nucleated community of only 2,000 people might possess a forum, basilica, several temples, entertainment buildings, an aqueduct and street drainage system, together with a level of selfgovernment and administration. This is a considerable per capita ‘monumental overhead’ representing an expenditure of surplus that is comparatively larger than in some other pre-industrial societies.151 population in the Wrst century ad were regarded by Jones as reliable, but many have rejected them as too high (above, n. 13). 150 For a later period see Wickham 2005, ch. 8. 151 For comparison, Ragusa (Dubrovnik), whose population in the mid-Wfteenth century was perhaps 5,000–6,000, was unusual for a late medieval town in acquiring

56

Alan Bowman and Andrew Wilson

To regard that expenditure of surplus simply as ‘consumption’ is facile; both the physical and institutional infrastructures enabled the development of larger markets, oVering opportunities for increased division of labour, and lowered transaction costs. Second, the extrapolation of population trends from systematic surface survey is a relatively recent activity and the data derived therefore are subject to various important caveats.152 Not only does one have to make decisions about what population multipliers to apply to diVerent settlement categories (villages, villas, farms, etc.) in a particular region, but there are also wider issues of survey comparability and of possible diVerences in site detection rates caused by a variety of factors (Weldwalking intensity, modern land use and visibility conditions, post-antique alluviation, erosion, etc.).153 Nevertheless, by way of a brief sketch of the agenda for a future volume, the direction of recent work in demography suggests the potential value of quantitative data for some places and periods on the following topics: . Numbers and distribution of settlements classiWed by size and rank. We can, for example, compile and compare the evidence for settlements in particular areas, some of which will encourage us to bring documentary and archaeological evidence into relation. The increasing amount of archaeological evidence from survey and excavation in Egypt (the Delta and the Fayum in particular) is complemented by a wealth of documentary detail which may yield both actual and virtual maps of particular areas.154 Well-published archaeological evidence for Syria in the later Roman period, where there is clearly a high level of prosperity, can provide interesting material for comparative trends.155

an aqueduct-fed running water supply as early as 1438, which fed only two fountains. The city of Cambridge received a running water supply only in 1610 with the construction of ‘Hobson’s Conduit’ (or the ‘New River’) from Nine Wells near Shelford: Bushell 1938. 152 Sbonias 1999; Witcher 2005; Fentress in this volume. 153 Wilson 2008b. 154 Delta: http://www.ees.ac.uk/Weldwork/deltasurvey.htm; Fayum: Rathbone 1996b, Davoli 1998. Mapping: Kru¨ger 1990; Rowlandson 1996; Mu¨ller 2003a; 2003b. 155 Tate 1992; Gatier 1994.

Quantifying the Roman Economy

57

. Evidence for survival of settlements over time. For Egypt in particular we can use the documentary evidence to trace the survival of town and village names over time and to estimate the correlation between the survival of names in documents and the actual survival of settlements.156 . Physical dimensions of urban and rural settlements. Actual sizes of sites can be collated and the simple scaling of site plans and other (comparatively objective) physical data will contribute to the realization of valid comparisons across regions, which will take account of regional variation (meaning that we will not assume that a 10 ha site in Spain must have the same population as a 10 ha site in Syria).157 It is striking, for example, that in Egypt villages of 20–50 ha are not uncommon—comparable with cities like Thugga in Tunisia (25 ha), Italica in Spain (41.5 ha), or even Turin in Italy (47 ha). Clearly, estimates based on walled area are crude and may ignore suburbs; they also tend to produce a snapshot at a single, and often poorly dated, moment in time. Nevertheless, this approach does enable some very broad-brush comparisons across time or between diVerent cultures which may be illuminating: seventeenth-century Ragusa (Dubrovnik) is about two-thirds the size of Wrst-century Pompeii (and had substantially less ‘monumental overhead’). Size data could be used to produce basic rankings of settlements by area, which in turn could be crosschecked against a likely hierarchy of functions or services provided by these cities. . Numbers of living units and density of habitation. We can experiment with approaches that might allow some extrapolation from sizes of urban and non-urban settlement to populations. We prefer approaches based on population densities per hectare, and on numbers of dwellings per settlement, to approaches based on Xoor area of houses, which for our period tend to be vitiated by uncertainty over the existence, extent, and numbers of upper storeys. Nevertheless, the density- or dwelling-based approaches are not at all straightforward; comparative evidence from other 156 Based on toponyms collected for regions of Egypt, Drew-Bear 1979; Pruneti 1981; Falivene 1998. 157 For city sizes in Spain, see Carreras Monfort 1995–6.

58

Alan Bowman and Andrew Wilson societies suggests that a wide range of urban population densities per hectare are theoretically possible, while dwelling-based estimates are confused by uncertainty over how many slaves to allow for each household over and above the nuclear family unit that comparative anthropology frequently suggests might be in the range of 4.5–5 people per unit.158 Comparisons of census data with known plans suggest that medieval European cities commonly had densities of 100–120 people per hectare, with densely populated cities perhaps up to around 200/ha; but calculations for ancient Mesopotamian cities suggest 300–500/ha, and Zorn, using a series of converging indicators, makes a strong case for 470–590/ha in the Iron Age II phase of Tell en-Nasbeh in Israel.159 Comparisons with urban densities of other cultures or periods serve simply to show the range of the possible; they cannot be used as arguments for choosing a particular Wgure for Roman cities since urban population densities and living patterns may be culturally speciWc. However, in some extensively excavated cities (Ostia, Pompeii, and to some extent also for Timgad, Sabratha, and other cities) it is possible to estimate numbers of living units or households, and these may be able to provide a cross-check against density per hectare calculations. Figures of c.175/ha for Wrstcentury ad Pompeii, c.200/ha for second-century Sabratha, and c.360/ha for the Trajanic colony at Timgad look plausible on these grounds. We would expect Rome, Ostia, and some other large cities with evidence for multi-storey insula housing to reach substantially higher densities. The Egyptian evidence may prove particularly valuable in enabling comparisons of site sizes and documentary Wgures for population. Clearly there is going to be considerable variation over time, between regions, and even between towns in the same region (densities in second-century ad Ostia will diVer considerably from those in Wrst-century ad Pompeii). Nevertheless, assessing the extent of such diVerences is valuable; and we may be able at least to establish possible ranges

158 This approach was used by Beloch 1886, notably for Roman Italy. For more recent discussions of population estimates for ancient settlements, see e.g. Zorn 1994; Carreras Monfort 1995–6. 159 Zorn 1994.

Quantifying the Roman Economy

59

(not absolute Wgures) for a set of urban populations. These in turn could be fed into debates about the urban population as a proportion of the empire’s total population.160 . Public building activity, as recorded by, inter alia, documents and inscriptions recording or commemorating construction and repair. Collections of building inscriptions, though not comprehensive, are available and seem to us (as they have to others) to be testable and useful proxy evidence for growth or shrinkage.161 This is a particularly vital issue since the notion of a severe decline in the quantity and quality of public building, accompanying a decline in the incidence of euergetism, is still a tenacious element in notions of economic decline in the period after ad 200.162 . IdentiWcation of ‘units of production’ for speciWc artefacts and commodities. As Wilson has pointed out elsewhere, there is a signiWcant amount of published archaeological evidence for ‘manufacturing’ sites in towns and cities which has not yet been fully aggregated and quantiWed.163 There is a growing body of evidence which will allow us to analyse a relatively small number of individual communities in detail, and to estimate Xuctuations in size over comparatively long periods and large distances. Much of this relates to Egypt and the Near East and will be discussed in detail in a subsequent volume (we note in passing the opinion that it is no longer intellectually respectable to dismiss Egyptian evidence as having no signiWcance or value for demographic patterns in the ancient Mediterranean).164 The village of Philadelphia in the Egyptian Fayum may stand as an illustration. Its population in the Wrst century ad can be calculated with some precision since we know from tax documents that the number of adult males between the ages of 14 and 60 who paid poll-tax was between 900 and 950 in the period c. ad 50–100.165 To arrive at the total population we need to estimate and add the number of exempt 160 Wilson forthcoming. 161 Daguet-Gagey 1997; Horster 2001; JouVroy 1986; Fentress 1981; Fentress and Perkins 1988, Wg. 4; Wilson 1997; Wilson 2007b. 162 Contrast Lepelley 1979 on fourth-century Africa. 163 Wilson 2002b. 164 Bagnall and Frier 1994: 173. 165 P.Sijp. 26, 906.5 taxpayers, P.Coll. Youtie 20, cf. Rathbone 1990: 133.

60

Alan Bowman and Andrew Wilson

persons (mainly Roman citizens) and use the multiplier which represents the proportion of adult males in the population as a whole.166 These Wgures may be subject to some margin of error but there can be little doubt that the total population lay in the range of 3,000–3,500 in this period. From that relatively secure base, we might then adduce evidence which relates to: the physical size of the settlement; the size of the population in relation to the early Ptolemaic period (the Zenon archive) and the later Roman period; comparison with other villages in the Fayum; the sizes of houses and households in the Wrst century; the conWguration of landholding in the region of the village. It would be foolhardy to predict that the results of such research will produce a more reliable or robust version of the Brunt/Beloch estimates, or even a conWdent conclusion for any single province, city, or area that we do not already have (few enough). We might, however, hope to contribute something robust to the debate on levels of urbanization in the empire and in its diVerent regions, an issue of general signiWcance both for the fact that ‘urbanization’ is often claimed as one of the hallmarks of Roman imperial culture and for its potential value as an indicator of economic performance (Lo Cascio, this volume). This too involves problems in the overall demographic picture which will probably not admit simple solutions—for example, the probable incompatibility of a total population estimate of the order of 55/60 million and an urbanization rate of c.10%, preferred by some.167 If the population is of the order of 60 million then the corollary may be a rate of urbanization signiWcantly higher than 10%. We would hope, however, to be understood as sidestepping the insoluble problems and practising the art of the possible, in the present state of evidence. The gains should be signiWcant, if only in providing some basis for undermining the tendency to be too reliant on exemplary or anecdotal items of information. 166 There is no universally agreed Wgure. Those most commonly used fall in the range 2.9 to 3.5. Further discussion is postponed to the next volume. 167 Hopkins 1980; 1983: 85; Temin 2006a, b; but cf. Pleket 2006: 320: ‘It would not be irresponsibly audacious to posit a population of, say, 65 or 75 million’. See also Lo Cascio, this volume.

Quantifying the Roman Economy

61

VI.2. The agrarian economy It is common ground that agriculture was the dominant economic activity throughout the Mediterranean world in classical antiquity. We would not, at the outset, propose to quantify this in terms of the proportion of gross product which it represented and we deliberately eschew any approach based on any dichotomy between urban and rural economies or the notion that the urban centres parasitically absorbed the wealth produced by agriculture without reciprocity. Likewise, we are suspicious of overarching generalizations, such as those which propose the growth of Italian latifundia in the second century bc, the dominance of agricultural slave labour, or the development of a ‘feudal’ system in later antiquity. Equally, we avoid simply adopting the perspective of the literary sources (which is not the same as asserting that they have no value), whether the ‘manuals’ of Cato or Columella, or the anecdotal evidence of writers such as Pliny the Elder. Three general observations, to begin with. First, it would be obviously absurd to attempt a quantiWed study of the whole agrarian economy of the Roman Mediterranean in a project of the scale we envisage, even had we much more hard evidence than is the case. Apart from the enormous regional diversity, it involves an extremely complex interplay of diVerent activities and factors: not merely cultivation of the Mediterranean triad, cereal, vine, and olive (along with fruit, vegetables, and pulses), but also pastoralism, animal husbandry, Wsh-farming, wild plants, salt-mining, and so on.168 EVective exploitation of the natural ecology to feed the population and to produce surpluses which can be transformed into other forms of wealth for the state (tax) and the individual is underpinned by modes of deployment of labour (slave and free, waged and tenancy-based), use or absence of technology, methods of transport, and redistribution over local and regional networks, all of which would need somehow to be costed or given a value. We suspect that any general statement which tried to capture the sum of the dimensions or values of such elements and activities would be fragile to the point of uselessness. This is of a piece with our reservations about estimates of GDP (see above). However, it 168 Horden and Purcell 2000: 175–297.

62

Alan Bowman and Andrew Wilson

does not mean that we completely reject the utility of models such as the ‘taxes-and-trade’ model, but we cannot see how calculations of, for example, the income of Ptolemaic Egypt in the last century bc and the whole empire in the Wrst century ad can help us very much.169 Second, the one and only area where we have suYciently detailed evidence to attempt a quantiWed study with any conWdence at all is Egypt (see Bowman’s chapter, below, for possibilities and qualiWcations).170 Here we have the basis, however patchy the evidence may be, for some speculative models which might be applied to modes of activity (land distribution, crop-yields, food-supply, labour) or to particular ‘communities’ (family, village, local region). Whatever conclusions we derive from such attempts certainly cannot simply be mapped on to other regions. Cursory and random survey of recent research which has attempted analysis of the economics of agriculture in diVerent regions other than Egypt reveals that even the best examples tend to be micro-regional, far from comprehensive, and reach conclusions which are much more qualitative than quantitative. Two examples: Kehoe on African estates in the Bagradas Valley cannot produce estimates even of the areas of land under cultivation.171 Duncan-Jones, on investment in viticulture, uses evidence from literary sources which he had to adjust in order to reach conclusions which the author himself could regard as economically plausible.172 Third, what can we do, therefore? We believe that there is now a suYcient quantity of reliable, widespread, and detailed studies for particular regions or categories of activity, spread across the Roman Mediterranean chronologically and geographically, which will either serve as proxies for quantifying agricultural economics or will in the aggregate suggest similar trends or divergences. Examples include: olive oil production in Africa and elsewhere, irrigation technology, land distribution in Asia Minor, Weld survey evidence on rural settlement patterns and villa systems.173 The diYcult part 169 For the ‘taxes-and-trade’ model: Hopkins 1980, cf. Hopkins 2002. Income of Egypt, cf. Hopkins 2002: 196, illustrating how such estimates can produce wildly incompatible conclusions. 170 For a robust statement of this position see van Minnen 2000. 171 Kehoe 1988. 172 Duncan-Jones 1982: 31–59. 173 Mattingly 1985; 1988a; 1988b; 1993; Hitchner 1993; 1995b; Wilson 2002a; Thonemann 2007.

Quantifying the Roman Economy

63

will, once again, be to see what conclusions of general signiWcance we can coax out of these micro-examples, both in relation to agricultural activity in its narrower sense and in terms of the relationship of agriculture to the other economic aspects which we are considering. Let us suppose, however, that we might be able to compare patterns of land-ownership in fourth-century Egypt, Asia Minor, and rural Syria, or production capacities of processing ‘plants’ for Wsh, olives, or wine in relation to the physical size of the settlements in which they were located. It is our hope that a suYcient number of such comparisons or proxies (more than we could ourselves generate) will make a signiWcant contribution to our understanding of the broader economic infrastructures and trends, as might comparison with evidence from somewhat later periods.

VI.3. Production and trade Papyri and other documentary sources provide abundant evidence for commerce and trade. Archaeologically, there is an abundance of physical material that has clearly been produced in once place and consumed in another; problems and pitfalls of analysing this are explored in Wilson’s chapter below. We hope to examine trends in rising and falling market shares across the Mediterranean of diVerent ceramic classes and products shipped in amphorae, and we propose to pursue the suggestion by Fulford in his contribution to this volume of using artefact distribution to examine to what extent the diVerent costs of land, river, and sea transport in practice constrained distribution of diVerent categories of goods. We will focus on: . The production of certain kinds of goods where data lend themselves to quantiWcation of production infrastructure, e.g. the salted Wsh industry and related by-products of garum, liquamen, etc.174 . The movement of ceramics (including Wnewares and amphorae), plotting provenance against quantiWed distribution of Wnds. To what extent does the distribution of imports drop oV with distance from the sea and from major river corridors? This will involve the 174 For a Wrst attempt at this, see Wilson 2006; cf. Wilson, this volume.

64

Alan Bowman and Andrew Wilson

collation of datasets from various published Weld survey projects and excavations. . Comparison of ceramic trends against selected artefacts, notably sarcophagi, decorative marble, and millstones, for which existing petrographic provenancing studies can be synthesized. . The context in which goods moved, with the caveat that factors biasing wreck discovery make them more useful for indicating typical ship and cargo sizes, than for establishing patterns of trading routes. Reanalysis of the chronological distribution of wreck data, and augmentation of the dataset, may produce new insights (Wilson, this volume). Egyptian papyri and the Vindolanda and Murecine tablets are also promising for illuminating the context in which goods moved. . Social perspectives of consumption—on what types of sites are diVerent artefacts found, with what implications for access? Is there increased access to mass-produced goods by more of the population and does this reliably indicate ‘economic growth’? For local commerce and economic activity, we can take the grapheion records from the Fayum village of Tebtunis, with concentrated data for a very large number of contracts of various kinds deposited over a limited period early in the reign of Claudius (see above). From this one can extract not only the numbers and periodic distribution of the contracts but also (in a village whose size we can estimate with some degree of conWdence) the number of people involved in written economic transactions, the value of the individual transactions, and accumulated ‘value’ (even if not complete) over the period. Data of this kind might allow us to construct at least a partial picture of the economic activity in such a village which might suggest values of capital assets (property) and of transactions in diVerent ‘sectors’.

VI.4. Mining, metals, and metal supply The Roman empire made extensive use of coin, and bullion supply was crucial to the highly monetized Roman economy. Archaeological and documentary (principally epigraphic) evidence for mining will be collated with numismatic data, to test the hypotheses that the

Quantifying the Roman Economy

65

exploitation of new sources of metal was to a signiWcant degree responsible for economic growth from Augustus onwards, and that the abandonment or loss of these sources created or exacerbated Wscal problems in the late second and the third century. Key issues are: . Output of coinage metals (gold, silver, copper), but also other metals whose extraction was important for production of tools, weapons, and other artefacts. . Utilization of Greenland ice core data on atmospheric pollution as a proxy measure of diachronic production of copper, silver, and lead.175 . Survey of major mining sites, with dates of exploitation, geographical distribution, and archaeological evidence of workings and output. At some sites, the size of slag heaps (discussed below) can provide quantiWable data on the output of individual mines. . Estimates of the precious metal content of the coinage, utilizing existing work estimating proportional output of new and reminted coin. In this context, the new metallurgical analyses by Butcher and Ponting, presented by Ponting in this volume and discussed by Hitchner and Howgego in their contributions, are fundamental.176 . Estimates of modes and volume of loss of precious metal over the period. . Examination of archaeological indications for the opening up of new trans-Saharan routes giving access to new, sub-Saharan gold sources.177 Closely related, of course, to questions of currency and prices are questions of the relationship of money supply and metal supply. In an earlier paper on the metal supply of the Roman empire one of us tried to look at the correlations between, on the one hand, the exploitation of mining resources in the Roman empire and, on the other hand, the precious metal contents of the coinage, as we now

175 Hong et al. 1994; 1996; Rosman et al. 1997; Wilson 2002a: 25–7. 176 Butcher and Ponting 1995; 1997; 1998; 2005. For earlier work, see Cope et al. 1997 and, for gold coinage, Morrisson 1985. 177 Garrard 1982; Wilson 2007a.

66

Alan Bowman and Andrew Wilson

Fig. 1.5. Roman slag heaps at lead and silver mines near Stojnic, Kosmaj (Serbia) (Photo: A. Wilson)

understand it principally from the work of Butcher and Ponting (cf. Ponting, this volume).178 The cessation of most large-scale activity at the major Spanish mines before the end of the second century squeezed the supply of precious metals just as demands for increased military pay rose. An obvious question is: were the mining resources providing the new metal which was still needed even for the much debased and recycled coinage of the third century? Case study 3: Mining in Upper Moesia In this regard, Dragana Mladenovic´ has recently drawn attention to some interesting quantitative evidence from the Kosmaj region of Serbia (Upper Moesia), where vast slag heaps of the lead and silver mines near Stojnic cover an area of some 6 km2 (Fig. 1.5).179 She has collected estimates of the amount of slag remaining in the 1960s, totalling over 2 million tonnes, and estimates of the amounts of lead 178 Wilson 2007a. 179 The evidence is treated at greater length in her doctoral thesis currently in progress, ‘The Romanization of Upper Moesia’. We thank Dragana Mladenovic´ both for this information and permission to summarize it here, and for the opportunity to explore the Stojnic slag heaps.

Quantifying the Roman Economy

67

Table 1.1. Estimates of remaining slag, and quantities of lead and silver extracted in the Roman period from the Kosmaj region in Upper Moesia Slag (tonnes) Removed 1880–1905 Removed in WW I Removed in WW II and 1956–66 Remaining in 1966–7 Original total Annual average over 300 years

336 227,686 ? ? 2,002,750 2,230,772 þ

Pb extracted (tonnes, þ/ 10%)

588,710 655,737 þ 2,186 þ

Ag extracted (tonnes, þ/ 10%)

4,754 5,295 þ 17.65 þ

Source: D. Mladenovic´.

and silver that were likely to have been extracted from that slag using Roman-period technology (Table 1.1). When one adds to the amount remaining in 1966–7 the various amounts of slag known to have been taken away for reprocessing in modern times, especially by the AustroHungarians during the First World War, one arrives at over 2.2 million tonnes of slag, from which over 650,000 tonnes of lead, and 5,295 tonnes of silver, had been extracted. Over a period of 300 years of operation, this averages out at at least 2,186 tonnes of lead and 17.65 tonnes of silver per year; real totals will have been higher, perhaps by 10% or more, since the Nazi forces also removed an unknown quantity of slag in the Second World War, which for a variety of reasons is likely to have been more than the Austro-Hungarians did. For comparison, this Wgure of over 17 tonnes of silver per year is closely comparable with the 20 tonnes per year produced by the Laurion mines at their peak of operation in the Wfth century bc. But the Kosmaj region is not Serbia’s main silver mining resource; it does not even Wgure on a 1990 map of the Upper Moesian mining regions, which concentrates on the better-known regions of Kosovo and the Upper Timok valley;180 nor does it appear on a modern map of silver deposits,181 perhaps because the Romans worked the deposits out to the water table. The much larger silver mines of Kosovo and the Upper Timok valley should have been even more important, and suggest that we need to think further about the relative 180 Parovic´-Pesˇikan 1987/90. 181 Monthel et al. 2002, Wg. 20: http://www.mineralinfo.org/Cartes/Serbie/pot_ minier.pdf.

68

Alan Bowman and Andrew Wilson

importance of the Balkan mines in the changing conditions of metal supply over the imperial period. This is borne out by the fact that, together with Bessa in northern Italy and Dolaucothi in Wales, the Balkans are the only region outside Spain where techniques of hydraulic mining were used to achieve gold extraction on a scale not paralleled again until the nineteenth century—Roman workings of alluvial gold deposits in the valley of the Vrbas river in Bosnia and the valley of the River Zsil in Dacia; and in stockwerk (layered vein) deposits in Dacia at Piatra Seaca and at Braza.182 The scale of mining achieved in the Roman world is related not only to the demand for metals for coinage, but also determined by the technology and capital investment available. Recent work has stressed the probable economic impact of the obvious and often spectacular advances in mining technology in the Roman period—the uptake of water-lifting devices for drainage, allowing mining of enriched deposits below the natural water table; and the often massive opencast gold mines worked by hydraulic erosion.183 However, work remains to be done on the less spectacular advances made in the technology of underground mining, including shoring and enlarged gallery size for ease of ore transport and removal, enabling the use of animals and even wagons in some mines;184 these advances are important because the majority of mining involved underground tunnelling rather than hydraulic opencast workings, and so advances made in this area could be more widely applied. Clearly, therefore, there is useful research that can be done to relate the archaeology and chronology of diVerent mining areas to the metal production of diVerent provinces at diVerent times, and to developments in coinage and the economy at large.185 One question, in the light of the apparent reduction in the number and scale of gold-mining sites by the fourth century ad, is whether the increased purchasing power of gold by that time noted by Rathbone (this volume) might in fact be due to an increase in its relative scarcity.186 182 Davies (O.) 1935: 186, 201, and 203. 183 e.g. Wilson 2002a; Domergue and Bordes 2006. 184 Rosumek 1982 for a preliminary collection of the evidence; cf. also Lewis (M.) 2001 for wagons and even an early mining ‘railway’. 185 Cf. Wilson 2007a. 186 Rathbone appears to discount this, but cf. Wilson 2007a.

Quantifying the Roman Economy

69

V I I. C ON C LU S I ON A project like this inevitably needs to concentrate on speciWcs; in order to quantify, we need particular case studies and areas. We cannot look at the whole economy, nor are we trying to. What we are trying to do is identify a series of indicators that can suggest trends; and these may be diVerent in diVerent regions. We need to break the problem down into constituent parts and analyse in some detail, and then re-aggregate the trends identiWed. By so doing we hope to obtain a clearer idea of how the Roman economy compares with periods before and after, and with other cultures and periods; and how diVerent provinces or regions performed over time within the Roman period. Most importantly, we hope to move discussion of the Roman economy in new directions, to stimulate debate, and to provide a means of allowing analyses of the Roman economy to be worked into economic histories of the very long run, which until recently have tended to start around ad 1000 for lack of published time series data before this. Bibliography Adams, J. N. (2003). Bilingualism and the Latin Language. Cambridge. Aldrete, G., and Mattingly, D. J. (1999). ‘Feeding the city: The organization, operation, and scale of the supply system for Rome’, in D. J. Mattingly and D. S. Potter (eds.), Life, Death and Entertainment in the Roman Empire. Ann Arbor, 171–204. Allen, R. C., Bengtsson, T., and Dribe, M. (eds. 2005). Living Standards in the Past: New Perspectives on Well-being in Asia and Europe. Oxford. Andreau, J. (1999). Banking and Business in the Roman World. Cambridge. —— (2005). ‘Le syste`me mone´taire partiellement ‘‘ferme´’’ de l’E´gypte romaine’, in T. F. Duyrat and O. Picard (eds.), L’Exception e´gyptienne (E´tudes Alexandrines 10 IFAO). Cairo, 329–38. Aubert, J.-J. (1994). Business Managers in Ancient Rome: A Social and Economic Study of Institores, 200 b.c.–a.d. 250. Leiden. Bagnall, R. S. (1985). ‘Agricultural productivity and taxation in later Roman Egypt’, TAPA 115: 289–308. —— (1992). ‘Landholding in late Roman Egypt: The distribution of wealth’, JRS 82: 128–49.

70

Alan Bowman and Andrew Wilson

Bagnall, R. S. (1993). Egypt in Late Antiquity. Princeton. —— (2005). ‘Evidence and models for the economy of Roman Egypt’, in J. G. Manning and I. Morris (eds.), The Ancient Economy: Evidence and Models. Stanford, 187–204 (repr. in Hellenistic and Roman Egypt (A30), chapter III). —— and Frier, B. W. (1994). The Demography of Roman Egypt. Cambridge. Banaji, J. (2002). Agrarian Change in Late Antiquity: Gold, Labour, and Aristocratic Dominance. Oxford. Bang, P. F. (2002). ‘Romans and Mughals: Integration in a tributary empire’, in L. de Blois and J. Rich (eds.), The Transformation of Economic Life under the Roman Empire. Amsterdam, 1–27. —— (2006). ‘Imperial Bazaar: Towards a comparative understanding of markets in the Roman Empire’, in P. F. Bang et al. (eds.), Ancient Economies, Modern Methodologies: Archaeology, Comparative History, Models and Institutions. Bari, 51–88. —— and Bayly, C. A. (eds. 2003). ‘Tributary empires in history: Comparative perspectives from antiquity to the late medieval’, Medieval History Journal 6.2 (special issue). —— Ikeguchi, M., and Ziche, H. G. (eds. 2006). Ancient Economies, Modern Methodologies. Archaeology, Comparative History, Models and Institutions. Bari. Bardill, J. (2004). Brickstamps of Constantinople (Oxford Monographs on Classical Archaeology) 2 vols. Oxford. Barrett, J. C. (1997). ‘Romanization: A critical comment’, in D. J. Mattingly (ed.), Dialogues in Roman Imperialism: Power, Discourse and Discrepant Experience in the Roman Empire (JRA Suppl. 23). Portsmouth, RI, 51–64. Begley, V. (1993). ‘New investigations at the port of Arikamedu’, JRA 6: 93–108. Beloch, K. J. (1886). Die Bevo¨lkerung der griechisch-ro¨mischen Welt. Leipzig. BeltrA´ n Lloris, F. (2006). ‘An irrigation decree from Roman Spain: The Lex Rivi Hiberiensis’, JRS 96: 147–97. BlA´ zquez Martnez, J. M., and Remesal Rodrguez, J. (eds. 1999). Estudios sobre el Monte Testaccio (Roma), 1. Barcelona. Boatwright, M. T. (1987). Hadrian and the City of Rome. Princeton. Bogaert, R. (1994). Trapezitica Aegyptiaca, recueil de recherches sur la banque en E´gypte gre´co-romaine (Pap. Flor. XXV). Florence. —— (1995). ‘Liste ge´ographique des banques et des banquiers de l’E´gypte romaine, 30a–284’, ZPE 109: 133–73. Bowman, A. K. (1985). ‘Landholding in the Hermopolite Nome in the fourth century ad’, JRS 75: 137–63.

Quantifying the Roman Economy

71

—— (1991). ‘Literacy in the Roman Empire: Mass and mode’, in J. Humphrey (ed.), Literacy in the Roman World (JRA Suppl. 3). Ann Arbor, 119–31. —— (2000). ‘Urbanisation in Roman Egypt’, in E. Fentress (ed.), Romanization and the City: Creation, Transformations and Failures (JRA Suppl. 38). Portsmouth, RI, 173–87. —— (2006). ‘Outposts of empire: Vindolanda, Egypt and the empire of Rome’, JRA 19: 75–93. —— (forthcoming). ‘Trade and the Xag’, in D. Robinson and A. I. Wilson (eds.), The Trade and Topography of Egypt’s North-West Delta (Oxford Centre for Maritime Archaeology). Oxford. —— and Rogan, E. (eds. 1999). Agriculture in Egypt from Pharaonic to Modern Times. London. —— and Thomas, J. D. (2003). The Vindolanda Writing Tablets: Tabulae Vindolandenses III. London. —— and Woolf, G. D. (eds. 1994). Literacy and Power in the Ancient World. Cambridge. Brun, J.-P. (2006). ‘L’E´nergie hydraulique durant l’empire romain: quel impact sur l’e´conomie agricole?’, in E. Lo Cascio (ed.), Innovazione tecnica e progresso economico nel mondo romano (Pragmateiai). Bari, 101–30. —— and BorrE´ani, M. (1998). ‘Deux moulins hydrauliques du Haut Empire romain en Narbonnaise: Villae des Mesclans a` La Crau et de Saint-Pierre/Les Laurons aux Arcs (Var)’, Gallia 55: 279–326. Brunt, P. (1971). Italian Manpower, 225 b.c.–a.d. 14. Oxford. —— (1987). Italian Manpower, 225 b.c.–a.d. 14, rev. edn. Oxford. Burnett, A. (2005). ‘The imperial coinage of Egypt in the Wrst century ad’, in T. F. Duyrat and O. Picard (eds.), L’Exception e´gyptienne (E´tudes Alexandrines 10 IFAO). Cairo, 261–77. —— Amandry, M., and RipollŁs, P. P. (eds. 1992). Roman Provincial Coinage, vol. I: From the Death of Caesar to the Death of Vitellius (44 bc–ad 69). London. Bushell, W. D. (1938). Hobson’s Conduit: The New River at Cambridge commonly called Hobson’s River. Cambridge. Butcher, K. (2004). Coinage in Roman Syria: Northern Syria, 64 bc–ad 253. London. —— and Ponting, M. J. (1995). ‘Rome and the East. Production of Roman provincial silver coinage for Caesarea in Cappadocia under Vespasian, ad 69–79’, OJA 14.1: 63–77. —— —— (1997). ‘A study of the chemical composition of Roman silver coinage, a.d. 196–197’, AJN 9: 17–36.

72

Alan Bowman and Andrew Wilson

Butcher, K. and Ponting, M. J. (1998). ‘Atomic absorption spectrometry and Roman silver coins’, in W. A. Oddy and M. R. Cowell (eds.), Metallurgy in Numismatics 4. London, 308–34. —— —— (2005). ‘The Roman denarius under the Julio-Claudian emperors: Mints, metallurgy and technology’, OJA 24.2, May: 163–97. Camodeca, G. (1992–). Tabulae Pompeianae Sulpiciorum (TPSulp.). Edizione critica dell’archivio puteolano dei Sulpicii. Naples. Carlsen, J., Ørsted, P., Skydsgaard, J. E. (eds. 1994). Land Use in the Roman Empire (ARID Suppl. 22). Rome. Carreras Monfort, C. (1995–6). ‘A new perspective for the demographic study of Roman Spain’, Revista de Historia da Arte e Arqueologia 2: 59–82. Claridge, A. (1988). ‘Roman statuary and the supply of statuary marble’, in J. C. Fant (ed.), Ancient Marble Quarrying and Trade (BAR International Series 453). Oxford, 139–52. Clarysse, W. (1983). ‘Literary papyri in documentary ‘‘archives’’ ’, in E. Van’t Dack, P. Van Dessel, and W. Van Gucht (eds.), Egypt and the Hellenistic World (Studia Hellenistica 27). Leuven, 43–61. Cope, L. H., King, C. E., Northover, J. P., and Clay, T. (1997). Metal Analyses of Roman Coins Minted under the Empire (British Museum Occasional Paper 120). London. Curtis, R. I. (1984). ‘A personalized Xoor mosaic from Pompeii’, AJA 88.4: 557–66. —— (1988–9). ‘A. Umbricius Scaurus of Pompeii’, in R. I. Curtis (ed.), Studia Pompeiana et Classica in Honor of W. F. Jashemski 1. New Rochelle, NY, 19–49. D’Arms, J. H. (1981). Commerce and Social Standing in Ancient Rome. Cambridge, Mass. Daguet-Gagey, A. (1997). Les opera publica a` Rome (180–305 ap. J.-C.). Construction et administration. Paris. Davies, J. K. (2001). ‘Hellenistic economies in the post-Finley era’, in Z. H. Archibald et al. (eds.), Hellenistic Economies. London, 11–62. Davies, O. (1935). Roman Mines in Europe. Oxford. Davoli, P. (1998). L’archeologia urbana nel Fayyum di eta` ellenistica e romana. Naples. de Blois, L., and Rich, J. (eds. 2002). The Transformation of Economic Life under the Roman Empire. Amsterdam. Domergue, C. (1983). ‘La mine antique d’Aljustrel (Portugal) et les tables de bronze de Vipasca’, Conimbriga 22: 5–193. —— and Bordes, J.-L. (2006). ‘Quelques nouveaute´s techniques dans les mines et la me´tallurgie a` l’e´poque romaine: leur eYcacite´ et leurs eVets sur

Quantifying the Roman Economy

73

la production’, in E. Lo Cascio (ed.), Innovazione tecnica e progresso economico nel mondo romano (Pragmateiai). Bari, 197–223. Drew-Bear, M. (1979). Le Nome Hermopolite, toponymes et sites. Missoula. Drexhage, H.-J. (1991). Preise, Mieten/Pachten, Kosten und Lo¨hne im ¨ gypten bis zum Regierungsantritt Diokletians. St Katherinen. ro¨mischen A —— Konen, H., and Ruffing, K. (2002). Die Wirtschaft des ro¨mischen Reiches. Berlin. Duncan-Jones, R. P. (1982). The Economy of the Roman Empire: Quantitative Studies, 2nd edn. Cambridge. Erdkamp, P. (2005). The Grain Market in the Roman Empire. Cambridge. E´tienne, R., and Mayet, F. (2002). Salaisons et sauces de poisson hispaniques (Trois cle´s de l’e´conomie de l’Hispanie romaine 2). Paris. —— —— (2004). L’Huile hispanique (Trois cle´s de l’e´conomie de l’Hispanie romaine 3) 2 vols. Paris. Falivene, M. R. (1998). The Heracleopolite Nome, a Catalogue of the Toponyms. Atlanta. Fentress, E. (1981). ‘African building. Money, politics and crisis in Auzia’, in A. C. King and M. Henig (eds.), The Roman West in the Third Century: Contributions from Archaeology and History (BAR International Series 109). Oxford, 199–210. —— and Perkins, P. (1988). ‘Counting African red slip ware’, in A. Mastino (ed.), L’Africa romana. Atti del V Convegno di studio. Sassari, 205–14. Finley, M. I. (1959). ‘Technology in the ancient world’, EHR 2nd ser. 12: 120–5. —— (1965). ‘Technical innovation and economic progress in the ancient world’, EHR 2nd ser. 18: 29–45. —— (1973). The Ancient Economy (Sather Classical Lectures 43). Berkeley. —— (1985). The Ancient Economy, 2nd edn. Berkeley and Los Angeles. Flobert, P. (1992). ‘Les graYtes de La Graufesenque: un te´moinage sur le gallo-latin sous Ne´ron’, in M. Iliescu and W. Marxgut (eds.), Latin vulgaire—latin tardif III. Tu¨bingen, 103–14. Frank, T. (1933–40). An Economic Survey of Ancient Rome, 5 vols. repr. 1975. Baltimore. Frederiksen, M. W. (1975). ‘Theory, evidence and the ancient economy’, JRS 65: 164–71. Frier, B. (2001). ‘More is worse: Some observations on the population of the Roman Empire’, in W. Scheidel (ed.), Debating Roman Demography. Leiden, 139–59. Gara, A. (1976). Prosdiagraphomena e circolazione monetaria: Aspetti dell’organizzazione Wscale in rapporto alla politica monetaria dell’Egitto romano. Milan.

74

Alan Bowman and Andrew Wilson

Garrard, T. F. (1982). ‘Myth and metrology: The early trans-Saharan gold trade’, Journal of African History 23.4: 443–61. Gatier, P.-L. (1994). ‘Villages du Proche-Orient protobyzantin (4e`me– 7e`me s.)’, in G. D. R. King and A. Cameron (eds.), The Byzantine and Early Islamic Near East II. Princeton, 17–48. Greene, K. (2000). ‘Technological innovation and economic progress in the ancient world: M. I. Finley re-considered’, EHR 53: 29–59. Groupe de Recherches sur l’Afrique antique (1993). Les Flavii de Cillium: E´tude architecturale, e´pigraphique, historique et litte´raire du mausole´e de Kasserine (CIL VIII, 211–216), (Collection de l’E´cole franc¸aise de Rome 169). Rome. Habermann, W. (1998). ‘Zur chronologischen Verteilung der papyrologischen Zeugnisse’, ZPE 122: 144–60. Harris, W. V. (1989). Ancient Literacy. Princeton. —— (1993). ‘Between archaic and modern: Some current problems in the history of the Roman economy’, in W. V. Harris (ed.), The Inscribed Economy: Production and Distribution in the Roman Empire in the Light of instrumentum domesticum (JRA Suppl. 6). Ann Arbor, 11–29. —— (2003). ‘Roman governments and commerce, 300 bc–ad 300’, in C. Zaccagnini (ed.), Mercanti e politica nel mondo antico. Rome, 275–305. —— (2006). ‘A revisionist view of Roman money’, JRS 96: 1–24. Hirt, A. M. (2004). ‘Mines and Quarries in the Roman Empire: Organizational Aspects 27 bc–ad 235’. University of Oxford, D.Phil. thesis. Hitchner, R. B. (1988). ‘The Kasserine archaeological survey, 1982–1986’, Ant. Afr. 24: 7–41. —— (1990). ‘The Kasserine archaeological survey—1987’, Ant. Afr. 26: 231–59. With contributions by Simon Ellis, Alan Graham, David Mattingly, and Lucinda Neuru. —— (1993). ‘Olive production and the Roman economy: The case for intensive growth in the Roman empire’, in M.-C. Amouretti and J.-P. Brun (eds.), La production du vin et de l’huile en Me´diterrane´e (Bulletin de correspondance helle´nique, Suppl. 26). Athens, 499–508. —— (1995a). ‘Historical text and archaeological context in Roman North Africa: The Albertini Tablets and the Kasserine Survey’, in D. B. Small (ed.), Methods in the Mediterranean: Historical and Archaeological Views on Texts and Archaeology. Leiden, 124–42. —— (1995b). ‘Irrigation, terraces, dams and aqueducts in the region of Cillium (mod. Kasserine): The role of water works in the development of a Roman-African town and its countryside’, in P. Trousset (ed.), Productions et exportations africaines: Actualite´s arche´ologiques en Afrique du Nord antique et medie´vale. VIe colloque international sur l’histoire et

Quantifying the Roman Economy

75

l’arche´ologie de l’Afrique du Nord (PAU, octobre 1993—118e congre`s). Paris, 143–57. —— (2005). ‘ ‘‘The advantages of wealth and luxury’’: The case for economic growth in the Roman empire’, in J. G. Manning and I. Morris (eds.), The Ancient Economy: Evidence and Models. Stanford, 207–22. Hobson, D. W. (1985). ‘House and household in Roman Egypt’, YClS 28: 211–29. Hong, S., Candelone, J.-P. and Boutron, C. (1996). ‘Changes in natural lead, copper, zinc and cadmium concentrations in central Greenland ice from 8250 to 149,100 years ago: their association with climatic changes and resultant variations of dominant source contributions’, Earth Planet. Sci. Lett. 143: 233–44. —— —— , Patterson, C. C., and Boutron, C. (1994). ‘Greenland evidence of hemispheric pollution for lead two millennia ago by Greek and Roman civilizations’, Science 265: 1841–3. Hopkins, K. (1972). Review of Brunt (1971), JRS 62: 192–3. —— (1978). ‘Economic growth and towns in classical antiquity’, in P. Abrams and E. A. Wrigley (eds.), Towns in Societies: Essays in Economic History and Historical Sociology. Cambridge, 35–77. —— (1980). ‘Taxes and trade in the Roman Empire (200 bc–ad 400)’, JRS 70: 101–25. —— (1983). ‘Models, ships and staples’, in P. Garnsey and C. R. Whittaker (eds.), Trade and Famine in Classical Antiquity (PCPS Suppl. 8). Cambridge, 84–109. —— (1991). ‘Conquest by book’, in J. Humphrey (ed.), Literacy in the Roman World (JRA Suppl. 3). Ann Arbor, 133–58. —— (1995/6). ‘Rome, taxes, rents and trade’, Kodai 5/6: 41–75. —— (2002). ‘Rome, taxes, rents and trade’, in W. Scheidel and S. von Reden (eds.), The Ancient Economy. Edinburgh, 190–230. Horden, P., and Purcell, N. (2000). The Corrupting Sea: A Study of Mediterranean History. Oxford. Horster, M. (2001). Bauinschriften ro¨mischer Kaiser: Untersuchungen zu Inschriftenpraxis und Bauta¨tigkeit in Sta¨dten des westlichen Imperium Romanum in der Zeit des Prinzipa¨ts. Stuttgart. Howgego, C. (1992). ‘The supply and use of money in the Roman world 300 b.c. to a.d 300’, JRS 82: 1–31. —— (1994). ‘Coin circulation and the integration of the Roman economy’, JRA 7: 5–21. Hudson, M. (2000). ‘How interest rates were set, 2500 bc–1000 ad’, Journal of the Economic and Social History of the Orient 43: 132–61.

76

Alan Bowman and Andrew Wilson

Hudson, P. (2000). History by Numbers: An Introduction to Quantitative Approaches. London. Humphrey, J. (ed. 1991). Literacy in the Roman World (JRA Suppl. 3). Ann Arbor. Johnson, A. C., and West, L. C. (1949). Byzantine Egypt: Economic Studies. Princeton. Jones, A. (2006). Dictionary of Globalization. Cambridge. Jones, A. H. M. (1948). Ancient Economic History (Inaugural lecture). London. —— (1964). The Later Roman Empire, 3 vols. Oxford. —— (1974). The Roman Economy: Studies in Ancient Economic and Administrative History, ed. P. A. Brunt. Oxford. Jongman, W. (2007). ‘Gibbon was right: The decline and fall of the Roman economy’, in O. Hekster, G. de Kleijn, and D. Slootjes (eds.), Crises and the Roman Empire. Proceedings of the Seventh Workshop of the International Network Impact of Empire (Nijmegen, June 20–24, 2006). Leiden, 183–99. Jouffroy, H. (1986). La construction publique en Italie et dans l’Afrique (E´tudes et Travaux 2). Strasbourg. Kehoe, D. P. (1984). ‘Private and imperial management of Roman estates in North Africa’, Law and History Review 2: 241–63. —— (1988). The Economics of Agriculture on Imperial Estates in Roman North Africa (Hypomnemata 89). Go¨ttingen. —— (1992). Management and Investment on Estates in Roman Egypt during the Early Empire (Papyrologische Texte und Abhandlungen 40). Bonn. —— (1993). ‘Economic rationalism in Roman agriculture’, JRA 6: 476–84. —— (1997). Investment, ProWt and Tenancy: The Jurists and the Roman Agrarian Economy. Ann Arbor. —— (2003). ‘Aristocratic dominance in the Late Roman agrarian economy and the question of economic growth’, JRA 16: 711–21. Kennedy, D. L. (2006). ‘Demography, the population of Syria and the census of Q. Aemilius Secundus’, Levant 38: 109–24. Kessler D., and Temin, P. (2008). ‘Money and prices in the early Roman empire’, in W. V. Harris (ed.), The Monetary System of the Greeks and Romans. Oxford, 174–207. Koepke, N., and Baten, J. (2005). ‘The biological standard of living in Europe during the last two millennia’, European Review of Economic History 9.1: 61–95. Komlos, J. (ed. 1994). Stature, Living Standards and Economic Development: Essays in Anthropometric History. Chicago. Kron, G. (2005). ‘Anthropometry, physical anthropology and the reconstruction of ancient health, nutrition and living standards’, Historia 54.1: 68–83.

Quantifying the Roman Economy

77

Krger, J. (1990). Oxyrhynchos in der Kaiserzeit: Studien zur Topographie und Literaturrezeption. Frankfurt. Lal, D. (1998). Unintended Consequences: The Impact of Factor Endowments, Culture and Politics on Long-Run Economic Performance. Cambridge, Mass., and London. Lepelley, C. (1979–81). Les cite´s de l’Afrique romaine au Bas-Empire: E´tude d’histoire municipale, 2 vols. Paris. Lewis, M. J. T. (1997). Millstone and Hammer: The Origins of Water Power. Hull. —— (2001). ‘Railways in the Greek and Roman world’, in A. Guy and J. Rees (eds.), Early Railways: A Selection of Papers from the First International Early Railways Conference. London, 8–19. Lewis, N. (1968). ‘Graeco-Roman Egypt: Fact or Wction?’, in Proceedings of the Twelfth International Congress of Papyrology, Ann Arbor 1968 (ASP 7). Toronto, 3–14. Lo Cascio, E. (1994). ‘The size of the Roman population: Beloch and the meaning of the Augustan census Wgures’, JRS 84: 23–40. —— (1999). ‘La popolazione dell’Egitto romano’, in M. Bellancourt-Valdher and J.-N. Corvisier (eds.), La de´mographie historique antique (Arras), 153–69. —— (ed. 2006). Innovazione tecnica e progresso economico nel mondo romano (Pragmateiai). Bari. —— and Malanima, P. (2005). ‘Cycles and stability: Italian population before the demographic transition (225 bc–ad 1900)’, Rivista di Storia Economica 21: 5–40. Loane, H. J. (1938). Industry and Commerce of the City of Rome (50 bc–200 ad) (Johns Hopkins University Studies in Historical and Political Science, ser. 56. 2). Baltimore. Loughton, M. E. (2003). ‘The distribution of Republican amphorae in France’, OJA 22.2: 177–207. McCormick, M. (2001). Origins of the European Economy. Cambridge. MacKinnon, M. (2004). Production and Consumption of Animals in Roman Italy: Integrating the Zooarchaeological and Textual Evidence (JRA Suppl. 54). Portsmouth, RI. Manning, J. G., and Morris, I. (eds. 2005). The Ancient Economy: Evidence and Models. Stanford. Marichal, R. (1988). Les graYtes de La Graufesenque (Gallia, Suppl. 47). Paris. Marzano, A. (2007). Roman Villas in Central Italy: A Social and Economic History (Columbia Studies in the Classical Tradition 30). Leiden.

78

Alan Bowman and Andrew Wilson

Matijasic, R. (1999). ‘La presenza imperiale nell’economia dell’Istria romana e nel contesto Adriatico’, Histria Antiqua 4: 15–22. Mattingly, D. J. (1985). ‘Olive oil production in Roman Tripolitania’, in D. J. Buck and D. J. Mattingly (eds.), Town and Country in Roman Tripolitania: Papers in honour of Olwen Hackett (BAR International Series 274). Oxford, 27–46. —— (1988a). ‘Oil for export? A comparison of Libyan, Spanish and Tunisian olive oil production in the Roman empire’, JRA 1: 33–56. —— (1988b). ‘The olive boom: Oil surpluses, wealth and power in Roman Tripolitania’, Lib. Stud. 19: 21–41. —— (1993). ‘Maximum Wgures and maximizing strategies of oil production? Further thoughts on the processing capacity of Roman olive presses’, in M.-C. Amouretti and J.-P. Brun (eds.) La production du vin et de l’huile en Me´diterrane´e (Bulletin de correspondance helle´nique, Suppl. 26). Athens and Paris, 483–98. —— (ed. 1997). Dialogues in Roman Imperialism: Power, Discourse and Discrepant Experience in the Roman Empire (JRA Suppl. 23). Portsmouth, RI. —— (ed. 2003). The Archaeology of Fazzan, vol. 1 Synthesis. London. —— (2004). ‘Being Roman: Expressing identity in a provincial setting’, JRA 17: 5–25. Meyer, J. C. (2006). ‘Trade in Bronze Age and Iron Age Empires, a comparison’, in P. F. Bang et al. (eds.), Ancient Economies, Modern Methodologies: Archaeology, Comparative History, Models and Institutions. Bari, 89–106. Millett, P. (2000). ‘Productive to some purpose? The problem of ancient economic growth’, in D. J. Mattingly and J. Salmon (eds.), Economies beyond Agriculture in the Classical World (Leicester–Nottingham Studies in Ancient Society). London, 17–48. Mitchell, S. (1999). ‘Greek epigraphy and social change: A study of Romanisation in south-west Asia Minor in the third century ad’, in XI Congresso Internazionale di EpigraWa Greca e Latina, Atti II. Rome, 419–33. Monthel, J., Vadala, P., Leistel, J. M., and Cottard, F., with the collaboration of Ilic, M., Strumberger, A., Tosovic, R., Stepanovic, A. (2002). Mineral Deposits and Mining Districts of Serbia: Compilation Map and GIS Databases. BRGM/RC-51448-FR. http://www.mineralinfo. org/Cartes/Serbie/pot_minier.pdf. Morley, N. (2001). ‘The transformation of Italy 225–28 bc’, JRS 91: 50–62. —— (2007). Trade in Classical Antiquity. Cambridge.

Quantifying the Roman Economy

79

Morris, I. (2005), ‘Archaeology, standards of living, and Greek economic history’, in J. G. Manning and I. Morris (eds.), The Ancient Economy: Evidence and Models. Stanford, 91–126. Morrisson, C. (1985). L’Or monnaye´ (Cahiers Ernest Babelon 2, 1 PuriWcation et alterations de Rome a` Byzance). Paris. Mller, K. (2003a). ‘Places and spaces in the Themistou Meris (Fayum/ Graeco-Roman Egypt): Locating settlements by Multidimensional Scaling of papyri’, Ancient Society 33: 103–125. —— (2003b). ‘Mastering matrices and clusters: Locating Graeco-Roman settlements in the Meris of Herakleides (Fayum/Egypt) by MonteCarlo-Simulation’, Archiv fu¨r Papyrusforschung 49: 218–52. —— and Lee, W. (forthcoming). ‘From mess to matrix and beyond: Estimating the size of settlements in the Ptolemaic Fayum/Egypt’, Journal of Archaeological Science. North, D. C. (1990). Institutions, Institutional Change and Economic Performance. Cambridge. Ørsted, P. (1994). ‘From Henchir Mettich to the Albertini Tablets: A study in the economic and social signiWcance of the Roman lease system (locatio-conductio)’, in J. Carlsen, P. Ørsted, and J. E. Skydsgaard (eds.), Land Use in the Roman Empire (ARID Suppl. 22). Rome, 115–25. ParoviC´ -PeSˇikan, M. (1987/90). ‘Kompleks metalursˇkih pec´i na Ulpijani’, Zbornik radova Muzeja rudarstva i metalurgije 5–6: 33–65. Paterson, J. (2004). ‘Autocracy and political economy’, Mediterraneo Antico VII.2: 571–89. Peacock, D. P. S., and Blue, L. (2006). Myos Hormos—Quseir al-Qadim: Roman and Islamic Ports on the Red Sea. Oxford. —— and Williams, D. F. (1986). Amphorae and the Roman Economy: An Introductory Guide (Longman Archaeology Series). London and New York. —— —— (eds. 2007). Food for the Gods: New Light on the Ancient Incense Trade. Oxford. Percival, J. (1976). The Roman Villa: An Historical Introduction. London. Pleket, H. W. (1993), ‘Agriculture in the Roman empire in comparative perspective’, in H. Sancisi-Weerdenburg et al. (eds.), De Agricultura: In Memoriam P. W. de Neeve. Amsterdam, 317–42. Ponting, M. J., Evans, J. A., and Pashley, V. (2003). ‘Fingerprinting of Roman mints using laser-ablation MC-ICP-MS lead isotope analysis’, Archaeometry 45.4: 591–7. Pruneti, P. (1981). I centri abitati dell’Ossirinchite: Repertorio toponomastico. Florence.

80

Alan Bowman and Andrew Wilson

Purcell, N. (1995). ‘The Roman villa and the landscape of production’, in T. J. Cornell and K. Lomas (eds.), Urban Society in Roman Italy. London, 151–79. Raschke, M. G. (1978). ‘New studies in Roman commerce with the East’, in H. Temporini and W. Haase (eds.), Aufstieg und Niedergang der ro¨mischen Welt II.9.2. Berlin, 604–1378. Rathbone, D. W. (1990). ‘Villages, land and population in Graeco-Roman Egypt’, PCPS 36: 103–42. —— (1991). Economic Rationalism and Rural Society in Third-Century ad Egypt: The Heroninos Archive and the Appianus Estate. Cambridge. —— (1994). ‘More (or less?) economic rationalism in Roman agriculture’, JRA 7: 432–6. —— (1996a). ‘Monetisation, not price inXation, in third-century Egypt’, in C. E. King and D. G. Wigg (eds.), Coin Finds and Coin Use in the Roman World. The Thirteenth Oxford Symposium on Coinage and Monetary History (Studien zu Fundmunzen der Antike 10). Berlin, 321–39. —— (1996b). ‘Towards a historical topography of the Fayum’, in D. M. Bailey (ed.), Archaeological Research in Roman Egypt (JRA Suppl. 19). Ann Arbor, 50–6. —— (1997). ‘Prices and price formation in Roman Egypt’, E´conomie Antique, Prix et formation des prix dans les e´conomies antiques (Entretiens d’arche´ologie et d’histoire St-Bertrand-de-Comminges). Paris, 183–244. —— (2001a). ‘The ‘‘Muziris’’ papyrus (SB XVIII 13167): Financing Roman trade with India’, in Alexandrian Studies II in Honor of Mostafa el-Abbadi (BSAA 4–6). Alexandria, 39–50. —— (2001b). ‘PSI XI 1183: Record of a Roman census declaration of ad 47/8’, in T. Gagos and R. S. Bagnall (eds.), Essays and Texts in Honor of J. David Thomas (American Studies in Papyrology 42). Oakville, Conn., 99–114. Rickman, G. (1980). The Corn Supply of Ancient Rome. Oxford. Riley, J. A. (1979). ‘The coarse pottery from Berenice’, in J. A. Lloyd (ed.), Excavations at Sidi Khrebish Benghazi (Berenice) (Supplements to Libya Antiqua 5.2). Tripoli, 91–467. Rodrguez Almeida, E. (1984). Il Monte Testaccio: Ambiente, storia, materiali. Rome. Rosman, K. J. R., Chisholm, W., Hong, S., Candelone, J.-P., and Boutron, C. F. (1997). ‘Lead from Carthaginian and Roman Spanish mines isotopically identiWed in Greenland ice dated from 600 b.c. to 300 a.d.’, Environment, Science and Technology 31: 3413–16. Rosumek, P. (1982). Technischer Fortschritt und Rationalisierung im antiken Bergbau (Habelts Dissertationsdrucke, Reihe alte Geschichte 15). Bonn.

Quantifying the Roman Economy

81

Rowlandson, J. L. (1996). Landowners and Tenants in Roman Egypt: The Social Relations of Agriculture in the Oxyrhynchite Nome. Oxford. Russell, J. C. (1958). ‘Late ancient and medieval population’, TAPS 48.3: 1–152. —— (1972). ‘Population in Europe 500–1500’, in C. M. Cipolla (ed.), The Fontana Economic History of Europe 1. Glasgow, 25–71. Saller, R. P. (2002). ‘Framing the debate over growth in the ancient economy’, in W. Scheidel and S. von Reden (eds.), The Ancient Economy. Edinburgh, 251–69. —— (2005). ‘Framing the debate over growth in the ancient economy’, in J. G. Manning and I. Morris (eds.), The Ancient Economy: Evidence and Models. Stanford, 223–38. Sbonias, K. (1999). ‘Introduction to issues in demography and survey’, in J. BintliV and K. Sbonias (eds.), Reconstructing Past Population Trends in Mediterranean Europe. Oxford, 1–20. Scheidel, W. (1996). ‘Finances, Wgures and Wction’, CQ 46: 222–38. —— (ed. 2001a). Debating Roman Demography (Mnemosyne Suppl. 211). Leiden. —— (2001b). Death on the Nile: Disease and the Demography of Roman Egypt (Mnemosyne Suppl. 228). Leiden. —— (2004). ‘Demographic and economic development in the ancient Mediterranean world’, Journal of Institutional and Theoretical Economics 160: 743–57. —— (2007). ‘A model of real income growth in Roman Italy’, Historia 56: 322–46. —— and von Reden, S. (eds. 2002). The Ancient Economy. Edinburgh. —— Morris, I., and Saller, R. (eds. 2007). The Cambridge Economic History of the Greco-Roman World. Cambridge. Setl, P. (1977). Private Domini in Roman Brick Stamps of the Empire: A Historical and Prosopographical Study of Landowners in the District of Rome (Acta Instituti Romani Finlandiae 9.2). Helsinki. Smith, A. (1776). An Inquiry into the Nature and Causes of the Wealth of Nations, 4 vols. Edinburgh. Steckel, R. H., and Rose, J. C. (eds. 2002). The Backbone of History: Health and Nutrition in the Western Hemisphere. Cambridge and New York. Storey, G. R. (1999). ‘Archaeology and Roman society: Integrating textual and archaeological data’, Journal of Archaeological Research 7.3: 203–48. Szaivert, W., and Wolters, R. (2005). Lo¨hne, Preise, Werte: Quellen zur ro¨mischen Geldwirtschaft. Darmstadt. Tacoma, L. E. (2006). Fragile Hierarchies: The Urban Elites of Third-Century Roman Egypt. Leiden.

82

Alan Bowman and Andrew Wilson

Tate, G. (1992). Les campagnes de la Syrie du Nord du IIe au VIe sie`cle: Une exemple d’expansion de´mographique et ´economique a` la Wn d’antiquite´, Institut franc¸ais d’arche´ologie du proche-orient (Bibliothe`que historique et arche´ologique 133, 1). Paris. Temin, P. (2001). ‘A market economy in the early Roman Empire’, JRS 91: 169–81. —— (2006a). ‘Estimating the GDP of the Early Roman Empire’, in E. Lo Cascio (ed.), Innovazione tecnica e progresso economico nel mondo romano (Pragmateiai). Bari, 31–54. —— (2006b). ‘The economy of the early Roman empire’, Journal of Economic Perspectives 20: 133–51. Thonemann, P. (2007). ‘Estates and the land in Late Roman Asia Minor’, Chiron 37: 435–78. Toepel, L. R. (1973). ‘Studies in the Administrative and Economic History of Tebtunis in the First Century ad’. Diss. Duke University. Van Der Spek, R. J. (2006). ‘How to measure prosperity? The case of Hellenistic Babylonia’, in R. Descat (ed.), Approches de l’e´conomie helle´nistique (Entretiens d’arche´ologie et d’histoire St-Bertrand-de-Comminges 7). Paris, 287–310. van Minnen, P. (1987). ‘Urban craftsmen in Roman Egypt’, MBAH 6.1: 31–88. —— (1998). ‘Boorish or bookish? Literature in Egyptian villages in the Fayum in the Graeco-Roman period’, Journal of Juristic Papyrology 28: 99–184. —— (2000). ‘Agriculture and the ‘‘Taxes-and-Trade’’ Model in Roman Egypt’, ZPE 133: 205–20. —— (2002). ‘Hermopolis in the crisis of the Roman empire’, in W. Jongman and M. Kleiwegt (eds.), After the Past: Essays in Ancient History in Honour of H. W. Pleket. Leiden, 285–304. Vernhet, A., and BE´ mont, C. (1991). ‘Un nouveau compte de potiers de La Graufesenque portant mention des Xamines’, Annales de Pegasus 1: 12–14. —— —— (1993). ‘Le graYte des nones d’octobre’, Annales de Pegasus 2: 19–21. von Reden, S. (2007). Money in Ptolemaic Egypt, from the Macedonian Conquest to the End of the Third Century bc. Cambridge. Wallace, S. L. (1938). Taxation in Egypt from Augustus to Diocletian. Princeton. Wallerstein, I. (1974–80). The Modern World-System. London and New York. Ward-Perkins, B. (2005). The Fall of Rome and the End of Civilization. Oxford.

Quantifying the Roman Economy

83

Webster, J. (2001). ‘Creolizing Roman Britain’, AJA 105.2: 209–25. White, L. T. (1962). Medieval Technology and Social Change. Oxford. —— (1980). ‘Technological development in the transition from antiquity to the Middle Ages’, in Tecnologia, economia e societa` nel mondo romano: Atti del convegno di Como, 27/28/29 settembre 1979. Como, 235–51. Whittaker, C. R. (1993). Land, City and Trade in the Roman Empire (Variorum Collected Studies 408). Aldershot. Wickham, C. (2005). Framing the Early Middle Ages: Europe and the Mediterranean, 400–800. Oxford. ¨ . (1980). Vattenmo¨llor och mo¨llare i det romerska riket. Lund. Wikander, O —— (1984). Exploitation of Water-Power or Technological Stagnation? A Reappraisal of the Productive Forces in the Roman Empire (Studier utgivna av Kungl. Humanistiska Vetenskapssamfundet i Lund, 1983–1984. 3). Lund. ¨ . Wikander (ed.), Handbook of Ancient —— (2000). ‘The water-mill’, in O Water Technology (Technology and Change in History 2). Leiden, 371–400. Will, E. L. (2004). ‘Mediterranean amphoras in India’, in J. Eiring and J. Lund (eds.), Transport Amphorae and Trade in the Eastern Mediterranean (Monographs of the Danish Institute at Athens 5). Athens, 433–40. Wilson, A. I. (1997). ‘Water Management and Usage in Roman North Africa: A Social and Technological Study’. D.Phil. thesis, University of Oxford. —— (2002a). ‘Machines, power and the ancient economy’, JRS 92: 1–32. —— (2002b). ‘Urban production in the Roman world: The view from North Africa’, PBSR 70: 231–73. —— (2006). ‘The economic impact of technological advances in the Roman construction industry’, in E. Lo Cascio (ed.), Innovazione tecnica e progresso economico nel mondo romano (Pragmateiai). Bari, 225–36. —— (2007a). ‘The metal supply of the Roman Empire’, in E. Papi and B. Scardigli (eds.), Supplying Rome (JRA Suppl. 69), 109–25. —— (2007b). ‘Urban development in the Severan Empire’, in S. C. R. Swain., S. J. Harrison, and J. Elsner (eds.), Severan Culture. Cambridge, 290–326. —— (2008a). ‘Large-scale manufacturing, standardization, and trade’, in J. P. Oleson (ed.), Handbook of Engineering and Technology in the Classical World. Oxford, 393–417. —— (2008b). ‘Site recovery rates and the ancient population of the Biferno Valley’, in G. Lock and A. Faustoferri (eds.), Archaeology and Landscape in Central Italy: Papers in Memory of John A. Lloyd (Oxford University School of Archaeology Monographs 69). Oxford, 233–53.

84

Alan Bowman and Andrew Wilson

Wilson, A. I. (forthcoming). ‘City sizes and urbanization in the Roman Empire’, in A. K. Bowman and A. I. Wilson (eds.), Population, Settlement, and the Roman Economy (Oxford Studies on the Roman Economy 2). Oxford. Witcher, R. (2005). ‘The extended metropolis: Urbs, suburbium and population’, JRA 18: 120–38. Woolf, G. (1992). ‘Imperialism, empire and the integration of the Roman economy’, WorldA 23.3: 283–93. —— (1998). Becoming Roman: The Origins of Provincial Civilisation in Gaul. Cambridge. —— (2000). ‘Regional productions in early Roman Gaul’, in D. J. Mattingly and J. Salmon (eds.), Economies beyond Agriculture in the Classical World (Leicester–Nottingham Studies in Ancient Society). London, 49–65. Wrigley, E. A. (2004). Poverty, Progress and Population. Cambridge. Zorn, J. R. (1994). ‘Estimating the population size of ancient settlements: Methods, problems, solutions, and a case study’, BASO 295: 31–48.

Part I Urbanization

This page intentionally left blank

2 Urbanization as a Proxy of Demographic and Economic Growth Elio Lo Cascio

It is common wisdom, presumably an uncontroversial notion, that the Greek World, from the eighth century bc, witnessed a set of phenomena to a certain extent connected: population growth, the resulting growth of total production, widespread urbanization. More controversial is the extent to which, if any, this ‘extensive growth’ was accompanied, mainly through the enlargement of the cultivated area and a process of intensiWcation, by ‘intensive growth’, that is, a growth in per capita production. If we look at the Italian peninsula uniWed by Rome in the period from the third to the Wrst century bc, it seems to be equally uncontroversial that there was an increase in urbanization, both in terms of the number of urban centres and of their dimension, and it is also certain that there was an increase in production, although it is highly disputed whether this growth was matched by a demographic growth, or even whether there was any increase at all in population. To put it in another way: the character of the demographic trend is hotly debated. If we look at the empire as a whole, and in particular at its western provinces, it seems to me that the general consensus is that population growth, growth of GDP and of per capita income, and urbanization went hand in hand. In what follows I would like to advance some observations on the relationship between the three variables (population, production, and rate of urbanization) in the Roman world resorting to comparative material in addition. I will ask, in particular, with reference to Italy

88

Elio Lo Cascio

and to other regions of the empire, whether the increase in the rate of urbanization can be thought to be a proxy of population increase— and therefore of extensive growth—and a proxy of intensive growth, and speciWcally of an increase in agricultural productivity. In order to clear the Weld of possible misunderstandings I should like to clarify at the outset the sense in which I use the term ‘urbanization’ and the expression ‘rate of urbanization’. It is hardly necessary to say that I do not refer to the ancient meaning of the words polis or res publica or civitas, nor do I dwell on the famous passage of Pausanias, which is always quoted in this context, for example by Moses Finley in the opening sentences of his seminal paper on the ancient city: the Periegetes dismisses the claim of Panopeus, a little town in central Greece, to city status, by observing that the people who live there have ‘no government oYces, no gymnasium, no theatre, no market-place, no water descending to a fountain, but live in bare shelters just like mountain cabins, right on a ravine’ (10.4.1).1 It ought to be obvious that I am simply referring to the numerical and dimensional increase of nucleated settlements above some threshold of population. I use ‘rate of urbanization’ to indicate the percentage of the inhabitants of an area who live in the nucleated settlements above that threshold. Such numerical and dimensional expansion of the nucleated settlements might not imply, in theory, any increase in population, but simply reXect a change in the mode of settlement. It is undeniable, however, that historically the clustering of population in nucleated settlements goes hand in hand with population increase. Having deWned ‘urbanization’ and ‘rate of urbanization’ in this way, one may then ask whether, beyond the social and cultural factors that promote the rise of nucleated settlements, it is legitimate to say, following Boserup, that a given density of population of an area is a sine qua non for the emergence of an urban scenario—a necessary condition, of course, not a suYcient condition. With regard to the early urbanization in the ancient world, usually explained as the eVect of the greater agricultural productivity of the areas in which it occurs, and/or of a rise in the technological level of agriculture, Boserup observes: ‘The two explanations of early urbanization miss the crucial point because they focus on agricultural surplus per 1 Finley 1981: 3.

Urbanization as a Proxy of Growth

89

worker rather than on total surplus . . . It was the number of foodproducing families on which the size of the surplus for urban consumption primarily depended . . . whether urban centers with large nonagricultural populations could appear depended primarily upon demographic factors, i.e., size and density of population in that area. Fertility of land and agricultural technology were secondary factors; the prime condition for early urbanization was a suYciently large and dense population.’2 In other terms, there is a correlation between density of population and urbanization and one can say that below a certain threshold of density, the emergence of urban centres with a proportion of population not engaged in agricultural production is impossible. In this sense, urbanization can be considered a proxy of population density. Boserup then points to the impact that technology in the transport of food has in aVecting the size of the settlements and the distance between one another. Nucleated settlements of small dimension can be, of course, just the result of a speciWc mode of settlement of an agricultural population, which does not live in scattered farms and villages. This is certainly the case for a substantial part of classical Greece, but also the case for southern Italy, and above all for Sicily, in modern times. It is hardly necessary to stress the diVerence between an ‘urban population’ and a population engaged (or engaged primarily) in non-agricultural activities; therefore, the rural/urban split does not correspond to the split between the population engaged in agriculture and the population engaged in non-agricultural activities. In general the speciWc phenomenon of the so-called ‘agro-town’ can explain how and why it was possible to have a very high rate of urbanization in classical Greece, even within the limits of a traditional agricultural economy. As I observed a few years ago referring to the population of Boeotia, ‘it is only by supposing that Boeotian towns were in fact predominantly agricultural that it is possible to accept the whole reconstruction . . . given by BintliV’,3 which implies that in the fourth century ‘some 70% of the whole population were town-dwellers, 30% were farm and village-dwellers’.4 This is a common feature of some other Mediterranean areas in medieval and

2 Boserup 1981: 65.

3 Lo Cascio 1999b: 165.

4 BintliV 1997; 1999: 22.

90

Elio Lo Cascio

modern times, for example southern Italy and Sicily as opposed to central and northern Italy. As Paolo Malanima has put it, by looking at the development of urbanization and the Italian economy during the last millennium, ‘in the pre-modern world, agrarian families, although almost invariably present in both small and large towns, were usually only a minority. This was the case in medieval cities of central and northern Italy, but not in the south, where the majority of the inhabitants of both large and small centres were often peasants. This was especially true of Sicily, where sparse settlements did not exist at all, and the population lived in large urban villages. If we consider these villages as towns, Sicily at the beginning of the nineteenth century would rank as the most urbanised region in Europe, and perhaps in the world, with an ‘‘urban’’ population of 66%; that is to say, the same urban percentage as Italy in the late twentieth century.’5 Of course, this apparently paradoxical conclusion points to the inherent ambiguities of a notion like the ‘urbanization rate’. It is, however, proper to ask whether this state of aVairs, which admittedly was more or less prevalent in the world of the Greek poleis, can be legitimately generalized to Roman Italy and the Roman empire. I must register the opinion put forward by Peter Garnsey many years ago, that the phenomenon of the ‘agro-town’ was not actually widespread in Italy.6 Only the smallest urban centres, in the regions of Roman Italy where they were more dense, could have been ‘agrotowns’. But, just to oVer an example, it seems clear from what Cicero says, that even in Ameria, which was, according to Cicero’s deWnition, one among the municipia rusticana (pro Roscio Amerino 43), a deWnition which can be thought to recall precisely the idea of the ‘agro-towns’,7 evidently considered by him typical of certain areas of central Italy, there was in any case a divide between the residents of the urban centre and the people living permanently in the villas and farms of the countryside. A scenario of ‘agro-towns’ like the one still dominant in the interior parts of Sicily in the twentieth century, presupposes a more or less empty countryside, and we do not Wnd an empty countryside in the regions of Roman Italy, where towns were 5 Malanima 2005: 98–9. 6 Garnsey 1979, repr. in Garnsey 1998 with addendum by Scheidel. 7 Lo Cascio 2006.

Urbanization as a Proxy of Growth

91

smaller and denser. Once the case of the ‘agro-towns’ is set aside, one may consider the rate of urbanization as representing more or less the percentage of the population not engaged in agricultural activities and, consequently, an indicator of agricultural productivity. I say, obviously, ‘more or less’, because, apart from the proportion of the agricultural population living in towns, a substantial, even if not quantiWable, proportion of the rural population was not engaged in agricultural activities or was engaged in them part-time. There is reason to believe that this proportion must have varied widely in the Roman world. Comparative evidence for other better known periods, however, would suggest that this was low, perhaps lower than 10% (the Wgure suggested for Wfteenth-century Tuscany by Herlihy and Klapisch-Zuber is 6%, which, according to Federico and Malanima, seems too low).8 But this percentage was much higher in Italy in 1881, when the economy of the country was still predominantly a traditional agricultural one: whereas the population of Italy not engaged in agricultural activities was 42% of the total population, the urban population (in centres above 5,000 inhabitants) was around 20%. Elsewhere I have used a very conservative estimate given by Bairoch: in pre-industrial societies ‘the percentage of nonagricultural activities exceeded the proportional weighting of the urban population by 4 to 5 percentage points’.9 To what extent, then, is it legitimate to consider an increasing rate of urbanization as a proxy of intensive growth? And to what extent is it legitimate to consider it as such in the Roman world? By themselves, neither the numerical and dimensional increase of urban centres, nor the increase in the rate of urbanization must necessarily reXect intensive growth, if the productivity of the factors of production—capital, and therefore land and labour—does not change over time. But historically things seem to have been diVerent. It is indisputable that even in the traditional agricultural economies of the past productivity changed over time and therefore the relationship between population growth and intensive growth changed as well. DiVerent models for singling out the relationship between demographic change and change in the agricultural production in traditional economies have 8 Herlihy and Klapisch-Zuber 1985; Federico and Malanima 2004: 439 n. 8. 9 Bairoch 1989: 266.

92

Elio Lo Cascio

been devised, starting from the crude and oversimplifying Malthusian model, to the Boserupian model of the impact of population pressure on the adoption of technological innovations in agriculture which can induce intensive growth. I will refer to the model adopted in a recent reassessment of the demographic development of Italy over two thousand years before the demographic transition,10 which is in essence a neo-classical growth model, adapted to pre-modern economies by replacing the focus on capital with the focus on what is the dynamic and unstable factor of production in those economies, that is, labour. The model highlights the eVects of a population increase on the productivity of both land and labour, assuming that capital (chieXy land) is stable, and considering technology as an exogenous factor. In this case, with the increase of population, the product per unit of capital increases as well, but labour productivity diminishes as a consequence of the lower and lower availability of capital (chieXy land) per worker, to the point when total production is barely able to support the current population. But, if we relax the conditions of a stable capital (agricultural land) and consider the eVect of technological advance on the ratio between product and capital, it is possible that not only the product per unit of capital, but also the product per unit of worker increases as well. It is possible, therefore, that population and per capita income both increase jointly, especially when new lands are put under cultivation, when better climatic conditions allow the enlargement of the cultivated areas, and when specialization, intensiWcation, or the adoption of new technical devices are able to increase the output. But there is a ceiling to per capita growth in the traditional economies before the industrial revolution and the innovations in the use of new sources of energy, since in the end the lower and lower availability of agricultural land per worker will determine a decrease in labour productivity. In some historical situations, however, these limits will be reached only after a long period of growth, and after a visible rise in the standard of living, as was the case with the world of the Greek poleis.11 10 Lo Cascio and Malanima 2005. 11 Morris 2004.

Urbanization as a Proxy of Growth

93

The increase in output per unit of worker can be measured, in speciWc historical situations, through the increase in the number of the people who are not engaged in the production of food. In this sense a change in the urbanization rate can be considered as a proxy of intensive growth, provided that we consider the large majority of the population of the so-called urban centres (that is, with a population above a given threshold) as not engaged in agricultural activities and provided that we can consider the rural population not engaged in agricultural activities as not substantial. I refer here to the model devised by Tony Wrigley (and adopted by Federico and Malanima in their study of the development of agricultural production in medieval and modern Italy), in order to estimate agricultural production when direct data on it are missing or unreliable, and when it would be impossible to derive them from the demand side since the series of prices and wages are equally missing.12 In this case ‘the urbanization rate may be used in order to estimate output per worker, albeit crudely’, if we assume that the whole agricultural production is consumed, that agricultural per capita consumption is constant (that is, it does not change following the changes in prices or income), that the ratio of the total workforce to population is constant, and that the time allocation between agricultural and nonagricultural work for all workers is constant. In this case, aggregate agricultural output equals per caput consumption of agricultural goods multiplied by population (P), and agricultural employment equals the whole population minus the urban population and rural non-agricultural population . . . Thus output per worker (y) can be calculated as:

y¼

P 1 ¼ P  P(Ur þ Rna ) 1  (Ur þ Rna )

where Ur is the urbanization rate and Rna is the proportion of the rural nonagricultural workers in the total population.13

In this way Federico and Malanima were able to calculate that output per worker grew in northern and central Italy by about one-Wfth between ad 1000 and 1300. 12 Wrigley 1986; Federico and Malanima 2004. 13 Federico and Malanima 2004: 438–9.

94

Elio Lo Cascio

Is it possible to attempt a similar exercise for the Roman empire or for some of its regions? There is no need to emphasize the absence of data from which to estimate agricultural production directly, or the lack of series of prices and wages which would allow us to estimate it indirectly from the demand side. Do we have the possibility of estimating the urbanization rate above a certain threshold? Certainly not; but we can propose, in limited cases at least, plausible orders of magnitude. This exercise allows us conversely to estimate, by comparison with other better known situations and the use of comparative evidence, the maximum possible level of the urbanization rate permitted by the carrying capacity of a territory, given the technological limits imposed on a traditional agricultural economy. In this sense, the exercise allows us to single out ‘the limits of the possible’, or ‘the parameters of the possible’, as labelled by Tacoma, with reference to population and to production.14 Put schematically and roughly: the more eYcient we consider Roman agriculture, the higher will be the urbanization rate, the lower will be the population engaged in agricultural production necessary to sustain the urban population (or, better, the population not engaged in agricultural activities). One can say that in this way a new dimension is added to the debate on the nature of the typical ancient city, whether consumer or producer. In theory, the urban population (or the population not engaged in agricultural activities) could consist just of people who are sustained thanks to entitlements of various kinds to beneWt from agricultural output. But it would be unrealistic to suppose that these people can constitute the majority of the urban population. Therefore an estimate of the ratio of urban population to the total population can give us an idea also of the extent of the secondary and tertiary sectors. In this connection, two qualifications are in order. First, I do not consider the direct evidence of the scope and quantitative importance of the secondary and tertiary sectors in the diVerent urban scenarios of Italy and the Roman provinces oVered chieXy by the papyrological and archaeological evidence, and the many studies that have challenged the ‘minimalist’ view of urban production and manufacturing activities in the towns of the Roman empire. I would like to make a reference in passing to the exemplary paper that Andrew Wilson 14 Tacoma 2006.

Urbanization as a Proxy of Growth

95

devoted a few years ago to the study of Timgad, Sabratha, and other African towns, in which an attempt was made to estimate the proportion of town-dwellers engaged in manufacturing activities.15 The second qualification is that I do not consider either the decisive role that a rising level of per capita agricultural productivity can have had in providing not only food for an expanding population, but also the raw materials and energy needed for the expansion of manufacture and transport. I refer here to the recent, illuminating paper on the development of English agriculture between 1300 and 1800 by Tony Wrigley—on what he calls ‘the transition to an advanced organic economy’.16 By looking at the end of the process, Wrigley observes that changes in agricultural technology ‘were perhaps less widespread and far-reaching than was commonly supposed’ even if ‘the introduction of new crops and new crop rotations was certainly highly important’, but he adds: ‘Yet changes in farm size, in market opportunity and structures, and in organizational forms, which were, of course, closely interrelated, probably played the decisive role in securing the gains that were achieved’.17 I wonder whether it is legitimate to apply this kind of argument to the development of Roman agriculture in Italy Wrst, and then in the provinces. Reverting to the relationship between agricultural productivity and rate of urbanization, it is of course impossible to give a quantitative estimate of the population of the perhaps 2,000 urban centres of the empire in ad 200,18 like the estimates, for instance, advanced by Jan de Vries for European urbanization in the modern age.19 And it is of course also impossible to create a database like the one built by Malanima, on the data collected by Ginatempo and Sandri and other scholars, for medieval and modern Italy.20 We can try to glimpse, through diVerent kinds of evidence, the population of some urban centres that we can consider more or less representative. The data that can be derived from the literary sources cannot be simply 15 Wilson 2002. 16 Wrigley 2006. 17 Wrigley 2006: 438–9. 18 For this Wgure Scheidel 2004b: 747. 19 De Vries 1984. 20 L’urbanizzazione in Italia, http://www.issm.cnr.it/asp/cv/malanima/dati/urban. pdf; see also Malanima 1998 and 2005; Ginatempo and Sandri 1990.

96

Elio Lo Cascio

dismissed as valueless (especially the data we have for Rome and for other so-called Me´gapoles of the empire).21 Sometimes they allow us to establish a hierarchy of urban centres in a speciWc region. I think not only of the famous Ordo nobilium urbium of Ausonius, but also of the much more interesting data that can be inferred, for example for northern Italy, from Strabo, data that have perhaps not been exploited as fully as they deserve, since it is clear that Strabo is interested in the size of the urban centres he describes.22 Better data can be drawn, as Richard Duncan-Jones showed many years ago, from the epigraphic sources documenting large-scale gifts for feasts or distributions. Even better data, perhaps the best we possess, can be drawn for some of the Egyptian metropoleis, especially valuable when they can be compared with the extension of the built-up area; certainly more debatable in their value are the data that can be elicited from the archaeological record of the extension of the towns, especially as deWned by their walled areas. There is, however, a feature of urbanization in the Roman empire in comparison with European urbanization that needs to be stressed as particularly signiWcant: the sheer number of urban centres with several hundred thousands of inhabitants, the so-called Me´gapoles, was higher in the Roman empire than in Europe till at least the Wfteenth or the sixteenth century, and modern Europe did not have a city the size of Rome till the early nineteenth century. Many social, political, and cultural reasons can be invoked to explain this diVerence, but are they enough? Greg Woolf has recently studied the development of the ‘urban networks’ of the Roman empire and especially of the Roman East, through the perspectives of the rank– size analysis and has compared therefore the rank–size distributions of European cities in diVerent times since the fourteenth century, as established by de Vries, with the tentative rank–size distributions which can be established—admittedly on the basis of very poor empirical data—for several regions of the empire.23 One of his main suggestions is that, whereas the number of primary centres in the Roman empire was certainly comparable with medieval and modern Europe or was even greater, secondary centres would have 21 Lo Cascio 2000.

22 Contra Morley 1996: 181.

23 Woolf 1997.

Urbanization as a Proxy of Growth

97

been much less numerous. He bases his argument on the analysis of diVerent areas of the empire: Egypt, Achaia, and especially Asia Minor. A similar argument has been advanced with reference to Egypt by Rens Tacoma.24 Now, this conclusion has been challenged, with good arguments, by Pleket, speciWcally with reference to Asia Minor.25 I would like to make a more general point: if the idea that in the Roman empire the rank–size distribution was much steeper than in modern Europe is not acceptable, then the total urban population of the empire must have been bigger than normally thought; and, if so, either the rate of urbanization must have been higher than 10% for the empire at large, the Wgure often proposed, also recently by Peter Temin in his estimate of the Roman GDP (where he notes that this is the ‘level of Britain, France and Germany in 1600’), or the population of the empire must have been larger than 55 million, the Wgure advocated by Temin.26 Perhaps, the most plausible solution is to think that both values must be higher than estimated. In my view this is the conclusion that in any case is suggested by what we know about the two regions for which we have the best set of data, Egypt and Italy, which apparently exhibit an astonishing rate of urbanization—they would have been incredibly important ‘poles of growth’, if we accept the lower estimates of their population. Apart from Alexandria, we can evaluate with a high degree of conWdence the population of several Egyptian metropoleis,27 starting from Hermopolis, whose population, over an extension of 120 hectares, ‘might reasonably fall between 25,000 and 50,000, probably nearer the upper than lower Wgure’.28 The estimate given by Bagnall and Frier in their book is in fact 37,000, with a density of c.300 inhabitants per hectare, a Wgure that they compare with that of eighteenthcentury Aleppo.29 Allowing 25,000 inhabitants on average to more or less 50 capitals of the nomes in Roman times (urban centres which certainly were not ‘agro-towns’), they reckon the total population of Egyptian metropoleis at 1.25 million, and assessing at 500,000 the population of Alexandria, they calculate an urban population of 1.75 24 27 28 29

Tacoma 2006. 25 Pleket 2003. 26 Temin 2006. Rathbone 1990; Bagnall 1993; Bagnall and Frier 1994; Tacoma 2006. Bagnall 1993: 53. Bagnall and Frier 1994.

98

Elio Lo Cascio

million, which is 37% of the whole population of Egypt estimated at 4.75 million. I have criticized this last estimate elsewhere and I do not want to dwell on this here.30 I just want to stress that the strongest argument against so low an evaluation of the total Egyptian population is that it implies a very high rate of urbanization, which, if we take into account the substantial export of grain from the country, would entail a very high labour productivity and therefore a very eYcient, perhaps too eYcient, agriculture.31 I advanced a similar argument, with reference to Italy.32 The sheer number of the urban centres in Italy—430 in the Augustan age—is so high that it would make the ‘low count’ of the total Italian population advocated by Beloch and Brunt implausible, since, even allowing for the fact that Rome was fed by the provinces, the low count would entail a ratio of non-agrarian population to the total population impossibly high for a traditional agricultural economy.33 This argument still fails to convince Walter Scheidel, who has recently produced a new thorough discussion of the topic and reiterated his rebuttal of the ‘high count’ and of the ‘revisionist’ perspective, as he has called it.34 We have now a new set of estimates of the population of Italy in the early empire by Morley and Scheidel, changing in detail those given by Hopkins in his book and I will refer to them.35 Scheidel takes the estimate of 1.3 million for the nonagrarian population of Italy outside Rome and considers it ‘about a quarter of the total population of Italy outside Rome’. He observes: ‘It is unclear if the non-farming population exceeded the urban population: if it did, the non-metropolitan urban total need not be put at much more than one million, or 20 to 25% of Italy’s population outside Rome’ and adds ‘In the Wrst half of the sixteenth century, about 15% of the population of peninsular Italy lived in cities of 5,000 or more’: a threshold which he thinks in any case too high for many or perhaps most of the 430 cities of Italy, a substantial part of 30 Lo Cascio 1999a. 31 For criticism along this line see also Scheidel 2001: 183–4; 247–8, and generally ch. 3; Tacoma 2006. 32 Lo Cascio 1994; 1999b. 33 Beloch 1886; Brunt 1971. 34 Scheidel 2004a. 35 Hopkins 1978: 68–9; Morley 1996: 181–3; Scheidel 2004a: 2–9.

Urbanization as a Proxy of Growth

99

which would be in any case ‘agro-towns’.36 Morley attempted to establish an urban hierarchy in Italy, singling out Wve (or rather four) levels: apart from the Wrst, Rome itself, the second level is that of the major ports, Ostia and Puteoli, credited with 30,000 each; the third level is the one of the regional centres, Mediolanum, Patavium, and Capua (but not Aquileia, since ‘it is highly unlikely that it should be included as part of an Italian urban system centred on Rome’), from 25,000 to 40,000, averaging 30,000, the fourth level is that of 25 major cities, from 5,000 to 25,000, averaging 15,000, the Wfth level is the one constituted by 400 minor cities, from 1,000 to 5,000, averaging 2,000. The total is 2,325,000 urban inhabitants in Roman Italy. Morley concludes, rightly in my opinion, that ‘at Wrst sight, the total Wgure for the urban population is exceptionally high; apparently nearly 40% of Italians lived in cities (25% if the city of Rome is excluded), a remarkable Wgure by the standards of pre-industrial societies’. Morley then goes on to advance a round estimate of 1,000 for the number of farming people in each town and a round estimate of 500 ‘involved in crafts and services’ in the ‘average small town’. In this way he is able to produce lower, and therefore apparently more credible, Wgures for the rate of non-farming population.37 Even if I think that it is a completely acceptable strategy to base the study of urbanism in Roman Italy on the hierarchy suggested by Strabo, I just want to observe that it would be enough to raise the average of the 25 major cities from 15,000 to 20,000 and the average of the 400 minor cities from 2,000 to 3,000 to boost the total to 2.9 million, which is dangerously close to 50% of the total population of Italy as estimated by Morley! There is, however, something that seems to me quite remarkable, and that is the comparison that we can make with Egypt and its metropoleis. This is a comparison that has been made by Scheidel and by Tacoma. Commenting on the hierarchy of Italian towns established by Morley, Scheidel observes: ‘Apart from a few dozen major centres, most towns appear to have been rather small. This is well brought out by a comparison with Roman Egypt: endowed with a similarly sized population, that region contained no more than Wfty ‘‘cities’’ alongside numerous 36 Scheidel 2004a: 6.

37 Morley 1996: 182–3.

100

Elio Lo Cascio

‘‘villages’’, the largest of which could house several thousands. The main diVerence between the urban systems of Italy and Egypt is one of deWnition: in Italy settlements that were as large as the most substantial Egyptian villages regularly enjoyed urban status.’38 I agree with this; but then it is necessary to explain why the urban centres which averaged 25,000 were more or less 50 in Egypt, and the centres which would have averaged 30,000 in Italy would have been just Wve, whereas there would have been just 25 other centres which averaged only 15,000! The complex and elaborate attempt by Rens Tacoma to establish a rank–size distribution of the urban centres of Roman Egypt on the basis of the extension of the nome territories highlights the same problem: how to explain that so many towns in Egypt averaged 14,000–22,000 and so few did in Italy?39 Whatever ‘numbers game’ we decide to play, it remains hardly credible that urban population outside Rome, in the core region and the most urbanized country of the Roman world would have been so low in absolute terms. I pointed elsewhere to the conservative data inferred by DuncanJones concerning ten secondary towns mostly of central Italy, some of which had free populations of the order of several tens of thousands (like Pisaurum, Comum, and Spoletium), and it is certainly true, as was objected by Scheidel, that only in the case of some of these Wgures can we be absolutely sure that they refer just to the plebs urbana.40 But at least they seem to suggest that the medium-sized towns must have been more numerous and widespread, and that even the smallest ones must have been not so small after all. Other estimates of the population of many Italian towns have been inferred since Beloch from the extension of the walled area. These estimates are certainly debatable, since they deduce rather mechanically the number of inhabitants from the extension of the walled area. My impression is that more often than not the built-up area must have been substantially larger because of the consistent presence of suburbia, even if the opposite could occur as well, namely that the built-up area was less extended than the walled area. In any case, the walls in the Roman towns, at least in the early empire, did not have the same function of 38 Scheidel 2004a: 15. 39 Tacoma 2006. 40 Lo Cascio 1999b: 165; Duncan-Jones 19822: 262–77; Scheidel 2004a: 15.

Urbanization as a Proxy of Growth

101

marking the limit of the town as in medieval cities, as Ginatempo has observed.41 We certainly need a thorough collection and analysis of these data on the built-up area of the Italian towns, which new techniques can now make more easily feasible. But even taking the walled area as a proxy of the population we have to consider that the number of centres with a walled area around 100 hectares, especially but not only in northern Italy, was substantial. Therefore, 1 million for the population of all the urban centres (apparently above the threshold of 1,000), the Wgure which appears to be advocated by Scheidel, for what was by far the most urbanized region of the empire seems too low.42 Let us, however, assume that the Wgures for the total population of Italy and the non-agrarian part of it advanced by Scheidel are more or less in the right order of magnitude: what do they imply? From the fourteenth century on, when data on Italian urbanization are better known, until 1861, the urbanization rate remained between 16 and 21% of the whole population, if we consider under ‘cities’ the centres with more than 5,000 inhabitants43 (the percentage is substantially higher, if we take the centres with more than 2,000 or 1,000). These values refer again to what was still the most urbanized area in Europe in the late Middle Ages and among the most urbanized regions of the continent in early modern times. Only the small Dutch region exceeded Italy from the sixteenth century on. England reached the Italian urbanization rate in 1700. Now, as we have seen, the values put forward by Morley and Scheidel imply an urbanization rate of at least 39%, Rome included, or, if we are justiWed in excluding Rome since it would have been fed by the provinces, of at least 20–25%. What level of agricultural productivity is implied by a rate of urbanization for the whole of Italy of 39 or even 25%? We have seen that the level of urbanization is limited, in pre-modern agrarian economies, by the productivity of labour in agriculture, which set a ceiling to the possibilities of urban expansion. This is the reason why, before 1800, urbanization seldom exceeded the level of 20% (taking again as the threshold the centres with more than 5,000). In 41 Ginatempo in Ginatempo and Sandri 1990. 42 Scheidel 2004a: 15. 43 Malanima 1998; 2005.

102

Elio Lo Cascio

order to estimate the agricultural labour productivity implied by the Wgures proposed by Scheidel, we can adapt to the speciWc case of Augustan Italy the equation already introduced. What we have to do is to use Scheidel’s Wgures for the total population, free and slave (5.8 million, that is, 4.6 million in ad 14, the mid-point of his estimate of 4.4 to 4.8 million,44 plus 1.2 million slaves45), for the inhabitants of Rome not depending on Italian agricultural production (700,000), and for the proportion of the population employed in agriculture. In this way we can estimate the number of people supported by a peasant with his work. The equation we have to solve is the following: y¼

P  P(Pfr ) 1  Pfr ¼ P  P(U þ Rna ) þ P(Ua ) 1  (U þ Rna ) þ Ua

where the symbols represent: y

output per worker (computed as the people supported by a worker in agriculture); P population; Pfr the proportion of the inhabitants of Rome fed with external cereals; U the urban coeYcient (Rome included); Rna the proportion of rural population employed in non-agricultural tasks; Uathe proportion of urban inhabitants employed in agriculture. According to the low estimate of the Italian population, if we assume a low percentage (0.10) of rural non-agricultural population Rna and low values (0.10) for the agricultural population living in the cities Ua (since it is diYcult to suppose that there were many peasants among the inhabitants of Rome), we reach high, perhaps too high, results: assuming that the total population was 5.8 million, that the total urban population above the threshold of 1,000 was 2 million (1 million in Rome, 1 million in the other urban centres),46 and 44 Scheidel 2004a: 9. 45 Scheidel 2005: 67, 71. 46 The assumption, which is by no means correct, is that all the centres with more than 1,000 inhabitants would be ‘towns’ in Roman terms, that is communities with an independent status.

Urbanization as a Proxy of Growth

103

therefore more or less 35% of the total population, that the inhabitants of Rome fed with external cereals were 700,000 (and therefore that P(Pfr) is 0.119), that the rural population employed in nonagricultural tasks was 580,000 and that the urban inhabitants employed in agriculture were 580,000, the value of y is 1.34, that is, 100 agricultural workers produced food for 34 other persons. In order to get more plausible results we have to assume that nonagrarian people in rural settings were negligible in number and, on the contrary, that the proportion of urban inhabitants employed in agricultural activities was much more than 20%, in the urban centres outside of Rome, both hypotheses which do not seem to be in line with the general scenario of the economy of Italy at the height of its expansion. For the sake of completeness I would add that if we accept the Wgure for the total population advocated by the so-called ‘high-counters’, and allow for a more realistic, even though still conservative value for the urban population outside Rome, we will reach a more realistic value also for the labour productivity of Italian agriculture, more in line with the productivity of the Italian agricultural economy in medieval and modern times. Assuming that the total population was 14 million, that the total urban population was 2.9 million and therefore 21% of the total population, that the inhabitants of Rome fed with external cereals were 700,000, that the rural population employed in non-agricultural tasks was 1.4 million and that the urban inhabitants employed in agriculture were 1.4 million, y would be equal to 1.20, that is, 100 agricultural workers would have produced food for 20 other persons. Even this value would put the economic performance of the Italian peninsula well above most of the other Mediterranean regions at that time and for many centuries afterwards. Bibliography Bagnall, R. S. (1993). Egypt in Late Antiquity. Princeton. —— and Frier, B. W. (1994). The Demography of Roman Egypt. Cambridge. Bairoch, P. (1989). ‘Urbanization and the Wndings of two decades of research’, Journal of European Economic History 18: 239–90. Beloch, K. J. (1886). Die Bevo¨lkerung der griechisch-ro¨mischen Welt. Leipzig.

104

Elio Lo Cascio

Bintliff, J. L. (1997). ‘Further considerations on the population of ancient Boeotia’, in J. L. BintliV (ed.), Recent Research on the History and Archaeology of Central Greece. Oxford, 231–52. —— (1999). ‘Regional Weld surveys and population cycles’, in J. L. BintliV and K. Sbonias (eds.), Reconstructing Past Population Trends in Mediterranean Europe (3000 bc–ad 1800). Oxford, 21–33. Boserup, E. (1981). Population and Technology. Oxford. Brunt, P. (1971). Italian Manpower 225 b.c.–a.d. 14. Oxford. de Vries, J. (1984). European Urbanization 1500–1800. London and Cambridge, Mass. Duncan-Jones, R. (19822). The Economy of the Roman Empire: Quantitative Studies. Cambridge. Federico, G., and Malanima, P. (2004). ‘Progress, decline, growth: Product and productivity in Italian agriculture’, 57.3: 437–64. Finley, M. I. (1981). ‘The ancient city from Fustel de Coulanges to Max Weber and beyond’, in M. I. Finley, Economy and Society in Ancient Greece (ed. with an introd. by B. D. Shaw and R. P. Saller). London and New York, 3–23 (1st pub. in Comparative Studies in Society and History 19 (1977), 305–27). Garnsey, P. (1979). ‘Where did Italian peasants live?’, PCPS 25: 1–25 (repr. in P. Garnsey (1998). Cities, Peasants and Food in Classical Antiquity: Essays in Social and Economic History, ed. with addenda by W. Scheidel. Cambridge, 107–31). Ginatempo, M., and Sandri, L. (1990). L’Italia delle citta`. Il popolamento urbano tra Medioevo e Rinascimento (secoli XIII–XVI). Florence. Herlihy, D., and Klapisch-Zuber, Ch. (1985). Tuscans and their Families: A Study of the Florentine Catasto of 1427 (Engl. trans.). New Haven. Hopkins, K. (1978). Conquerors and Slaves. Cambridge. Lo Cascio, E. (1994). ‘The size of the Roman population: Beloch and the meaning of the Augustan census Wgures’, JRS 84: 23–40. —— (1999a). ‘La popolazione dell’Egitto romano’, in M. BellancourtValdher and J.-N. Corvisier (eds.), La de´mographie historique antique. Actes du premier colloque international de de´mographie historique antique, Arras, 22–23 novembre 1996. Arras, 153–69 (repr. in Studi Storici 40 (1999), 425–47). —— (1999b) ‘The population of Roman Italy in town and country’, in J. BintliV and K. Sbonias (eds.), Reconstructing Past Population Trends in Mediterranean Europe (3000 bc–ad 1800). Oxford, 161–71. —— (2000). ‘Registrazioni di tipo censuale e stime della popolazione di Roma nell’antichita`’, in C. Nicolet, R. Ilbert, and J.-Ch. Depaule (eds.), Me´gapoles Me´diterrane´ennes: Ge´ographie urbaine retrospective. Actes du

Urbanization as a Proxy of Growth

105

colloque organise´ par l’Ecole franc¸aise de Rome et la Maison me´diterrane´enne des sciences de l’homme, Rome, 8–11 mai 1996. Paris, 628–59. —— (2006). ‘Realta` e rappresentazione: la caratterizzazione degli homines ex municipiis rusticanis nella pro Roscio Amerino’, in G. Petrone and A. Casamento (eds.), Lo spettacolo della giustizia: le orazioni di Cicerone. Palermo, 49–62. —— and Malanima, P. (2005). ‘Cycles and stability: Italian population before the demographic transition (225 b.c.–a.d. 1900)’, Rivista di Storia Economica 21.3: 197–232. Malanima, P. ‘L’urbanizzazione in Italia’, http://www.issm.cnr.it/asp/cv/ malanima/dati/urban.pdf (accessed 11 June 2008). —— (1998). ‘Italian cities 1300–1800: A quantitative approach’, Rivista di Storia Economica 14: 91–126. —— (2005). ‘Urbanisation and the Italian economy during the last millennium’, European Review of Economic History 9: 97–122. Morley, N. (1996). Metropolis and Hinterland: The City of Rome and the Italian Economy 200 b.c.–a.d. 200. Cambridge. Morris, I. (2004). ‘Economic growth in ancient Greece’, Journal of Institutional and Theoretical Economics, 180.4 (Dec.), 709–42. Pleket, H. W. (2003). ‘Economy and urbanization: Was there an impact of empire in Asia Minor?’, in E. Schwertheim and E. Winter (eds.), Stadt und Stadtentwicklung in Kleinasien. Bonn, 85–95. Rathbone, D. (1990). ‘Villages, land and population in Graeco-Roman Egypt’, PCPS 36: 103–42. Scheidel, W. (2001). Death on the Nile: Disease and the Demography of Roman Egypt. Leiden, Boston, and Cologne. —— (2004a). ‘Human mobility in Roman Italy, I: The free population’, JRS 94: 1–26. —— (2004b) ‘Demographic and economic development in the ancient Mediterranean world’, Journal of Institutional and Theoretical Economics 180.4 (Dec.), 743–57. —— (2005). ‘Human mobility in Roman Italy, II: The slave population’, JRS 95: 64–79. Tacoma, L. (2006). Fragile Hierarchies: The Urban Elites of Third Century Roman Egypt. Leiden and Boston. Temin, P. (2006). ‘Estimating GDP in the early Roman empire’, in E. Lo Cascio (ed.), Innovazione tecnica e progresso economico nel mondo romano. Bari, 31–54. Wilson, A. I. (2002). ‘Urban production in the Roman World: The view from North Africa’, PBSR 70: 231–73.

106

Elio Lo Cascio

Woolf, G. (1997). ‘The Roman urbanization of the East’, in S. E. Alcock (ed.), The Early Roman Empire in the East. Oxford, 1–14. Wrigley, E. A. (1986). ‘Urban growth and agricultural change: England and the Continent in the early modern period’, in R. I. Rotberg and Th. K. Rabb (eds.), Population and Economy: Population and History from the Traditional to the Modern World. Cambridge, 123–68. —— (2006). ‘The transition to an advanced organic economy: Half a millennium of English agriculture’, EHR 59: 435–80.

3 Response to Elio Lo Cascio Roger Bagnall

What do we mean by urbanization? If I have a number of qualms about the resourceful and wide-ranging assessment that Elio Lo Cascio oVers of the degree of urbanization of the Roman imperial world, in considerable part they stem from discomfort with his answers, explicit and implicit, to the question of what counts as urban in our calculations of the percentage of population living in urban centres. I do not see any sign of consensus on this subject in the fraction of the relevant literature that I have read. A lack of clarity on this point is also at the root of the inconsistency with which Italy and Egypt are treated in this chapter. Lo Cascio’s opening remarks treat his position as self-evident, not needing justiWcation. That seems to me unwise. Lo Cascio’s view is that size matters, and little else does. In particular, he is not attached to juridical deWnitions of the status of settlements as a discriminant. One can understand why this seems a reasonable approach. Just because a settlement of 2,000 people is called a city does not mean that it belongs in the same category with Ephesos. And for cross-period comparisons a size cut-oV seems plausible. But in fact oYcial status has an impact on the nature of a settlement and its relationship to its surroundings. In particular, cities provided or were expected to provide a certain range of services, and these helped to structure that relationship. Small or large, the presence of major cult centres, organs of justice, and markets (to name only the three most obvious) makes a great diVerence in the degree to which a settlement employs its population in non-agricultural occupations

108

Roger Bagnall

and can draw on the surplus of others’ agricultural activity. If we are asking about urbanization not in the abstract but as a pointer to (e.g.) the percentage of the population not employed in agriculture, then oYcial status matters. For this reason, I do not see an alternative to the painstaking counting-up of cities. We would agree, however, that ancient Roman cities were not in the main (or should not be deWned as) what Lo Cascio calls agrotowns—or, at least, that in speaking of urbanization we should exclude these from consideration. For Egypt, on which I shall concentrate my remarks, the non-agricultural character of the nome capitals is put beyond doubt by the Hermopolite land registers of the fourth century, where on almost any reasonable estimate of the populations of the city and the nome one can hardly argue that the landowners (with their families) amounted to more than about 15% of the population; and those with enough land to live solely on rents from it can hardly be more than half that. It is true that the register omits the pagus nearest to the city, which will have contained intensively cultivated garden and orchard land. But much of that no doubt belonged to the same wealthy people who owned the largest share of urban holdings in the rest of the Hermopolite, and one pagus out of seventeen cannot alone change the picture greatly. It is also true that some urban residents undoubtedly worked as hired cultivators of one sort or another in the surrounding countryside, on land owned by the wealthy or by women. Practically speaking, however, considerations of distance make it unlikely that this was more than a modest percentage of the population. On any reckoning, then, a large majority of Hermopolis’s population must have been supported by something other than landowning or agricultural labour. That gives us a fourth-century marker. It does not tell us of itself anything about change during the previous several centuries, and in fact nothing in Lo Cascio’s chapter really gets at the dimension of growth. This is hardly surprising. Even for Egypt, our data about city populations, however imperfect, are also scattered across the centuries. If recent attempts to deduce a structure to the distribution of metropolitan populations have had diYculty with the limited quantity and quality of data,1 still less have they been able to trace change, 1 Tacoma 2006, with earlier bibliography cited by him.

Response to Elio Lo Cascio

109

as we tend to have at best one data point for any given city. Probably the best we can do is to compare the metropolis of the Arsinoite nome in the third century bc, under Ptolemaic rule, and at somewhere in the third to fourth century. The census records from the second half of the third century bc suggest a population for Crocodilopolis of only about 4,000; estimates for the imperial peak range from 27,000 to 46,000,2 with Tacoma recently estimating ‘44,000 or less’.3 Although we should not underestimate the urban character of even the small city of the Ptolemaic period,4 most scholars of GraecoRoman Egypt would agree, rightly or wrongly, that most of the growth came in the Roman period, with the advent of the class of small notables living on agricultural rents. Since Clarysse and Thompson have estimated the population of the Arsinoite nome at 85,000 to 90,000 at the same period at which Arsinoe itself was in the vicinity of 4,000,5 the early Ptolemaic period would apparently have had an urbanization rate of only 5% or so in this district. It is perhaps worth another look at the ratio of urban to rural population in Roman Egypt, taking into account the work published since The Demography of Roman Egypt.6 I remain convinced that the concept of carrying capacity will get us nowhere. Egypt in Hellenistic and Roman times always produced a substantial surplus for export, whether through the sale abroad of a royal tax surplus, private shipments, imperial tax wheat sent to Rome for free distribution, or some combination of these at various points. We have Wgures for the tax wheat sent abroad, but we have no idea how much was sold at any given point; there is thus no way of estimating what was consumed in Egypt. I would begin by the best Wgure we have, namely that for the number of Hermopolite households, which probably amounted to about 7,000. What multiplier to use for individuals is debatable, but it seems unlikely that a number per house is relevant; the number per household is more useful, and that for urban households we estimated at 5.3. That is where a population Wgure of 37,000 came from (a Wgure per hectare is only derivative, not the source of the 2 Clarysse and Thompson 2006, vol. 2: 100. 3 Tacoma 2006. 4 Bingen 2007: 114–21. 5 Clarysse and Thompson 2006, vol. 2: 95. 6 Bagnall and Frier 1994, esp. 53–7.

110

Roger Bagnall

calculation). It may of course be too high or too low, for all sorts of reasons we cannot know. The Hermopolite nome amounted to about 4% of the arable land of Egypt; the percentage might be slightly higher, but probably not lower, as some recent writers have tended to more conservative rather than more expansive estimates of the surface under cultivation (see Chapter 7 for further discussion). If the number is approximately correct, it was about double the size of the average nome, on the assumption that there were about Wfty nomes. We do not know if the relationship between the size of a nome and the population of its metropolis was relatively constant. Tacoma’s analysis tends to support a fairly stable relationship, but it has to be said that our data are scanty enough that it would be hard to exclude absolutely the possibility of something closer to a power law, in which the metropolitan population relative to rural population was greater in the largest nomes. The villages do seem to exhibit a power law distribution, to judge particularly from the vestis militaris codex P.Col. IX 247 (also Hermopolite). The Hermopolite probably had something like 15 to 20 sizeable villages at the head of a number four to Wve times as large of smaller settlements. But there is no obvious reason to believe that the distribution of metropolis populations followed a similar pattern. If we leave that wrinkle aside for the moment, the average metropolis would on Tacoma’s reckoning have had a population smaller than the 25,000 average that Frier and I used in our analysis, perhaps more like half the estimated Hermopolite population, or 18,500. Fifty such cities would yield an urban population (not counting Alexandria) of 925,000; add a half-million for the capital, and the total is 1,425,000. It is unlikely that we can do much better than that with present data. But even that is better than we can do for the rural population, where diVerent positions are staked out by various scholars without really much basis. If the cities represented a quarter of the non-Alexandrian population, the chora would have had a population of 3.7 million; if a Wfth, 4.6 million. Of course, these are static numbers. Both may have been true at one point or another, if (as we suppose) there was slow population growth over the Wrst two centuries of Roman rule, down to the Antonine plague. Either seems more plausible than the possibility of an urbanization rate lower than 15%, which a total population Wgure for Egypt of 8 million would imply.

Response to Elio Lo Cascio

111

On almost any reckoning, however, we remain faced with the conundrum that Lo Cascio poses: Why did Egypt have more centres with populations greater than (say) 20,000 than Italy, with also a high urbanization rate and a population probably on the same order of magnitude? One might add the question why Egypt also probably had more settlements in the 2,000 to 5,000 range than Italy did, as an extrapolation from the Hermopolite vestis register would seem to me to suggest. One could juxtapose two possible answers. One is that Egypt’s administrative structure itself, with its nomes and toparchies, later pagi, provided a means of concentrating agricultural surpluses via rents, taxes, and provision of services. It did not leave entirely to the chances of historical development the attribution of land to cities. The other is that Egypt’s agriculture was simply that much more productive than Italy’s. I suspect that both of these are true, but it would take us very far aWeld to try to quantify either. In concluding it is worth returning to the problem of change, which is largely lost in Lo Cascio’s attempt to arrive at general levels of urbanization—not surprisingly, given how diYcult it is to answer the question for any time, let alone for several. It is only the very long-term data for the Arsinoite nome that may give us some sense of the scale of change. Even on a fairly conservative assessment, urbanization in that region in the middle to late Roman period is unlikely to have been less than 15%, and if the 5% estimate for the third century bc is anywhere near the mark, then a tripling of the concentration of population in the city over a half-millennium seems plausible. If one believes, as I do, that the urban population was more nearly a quarter of the total, then a quintupling is likely. How much of that took place under the Ptolemies and how much under the Romans, it seems impossible to say. At all events, this is the sort of direction in which I would look for attempting to quantify change. Bibliography Bagnall, R. S., and Frier, B. W. (1994, 20062). The Demography of Roman Egypt. Cambridge. Bingen, J. (2007). Hellenistic Egypt: Monarchy, Society, Economy, Culture. Edinburgh.

112

Roger Bagnall

Clarysse, W., and Thompson, D. J. (2006). Counting the People in Hellenistic Egypt, 2 vols. Cambridge. Tacoma, L. (2006). Fragile Hierarchies: The Urban Elites of Third Century Roman Egypt. Leiden and Boston.

Part II Field Survey and Demography

This page intentionally left blank

4 Archaeology, Demography, and Roman Economic Growth Willem Jongman

A M ILLENNIAL VIEW We do not need to go back in time any further than our own youth to realize that we live in a far more prosperous society than our parents ever did. Even corrected for inXation, incomes have roughly doubled over a generation. We can spend these higher incomes on more goods and services, including many that did not even exist a generation ago. We are also healthier than our parents’ generation: we are taller, we are Wtter in our mature age, and we die later. Moreover, since there are also vastly more of us, our aggregate wealth has increased even more. I am not sure the old story is still true that there are now more people alive than have ever died in all of human history, but even if it is not, nineteenth and particularly twentieth-century population growth has been staggering. Environmental Armageddon may well be the only—Wnal—frontier. The modern growth experience has been overwhelming, and has imprinted on us the indelible belief that history moves in only one direction: upwards. And yet there was a time when it was commonly believed that life in the past had been better, and that the present was but a bleak shadow of a past golden age. This may seem like hopeless Romanticism, but it is in fact not. There is good reason to believe that the move from hunting and gathering to sedentary agriculture was one of dire necessity from population pressure, and carried the price

116

Willem Jongman

of a declining standard of living and reduced life expectancy.1 Thus, hunter gatherers were ‘the original aZuent society’ because they had an abundant, diverse, and stable food supply that could be obtained with fewer working hours than ever after.2 The subsequent Neolithic Revolution allowed the planet to carry many more people, but the price for that was hard work, deteriorating diets, and worsening health.3 Life had become brutish and short. The most graphic expression I know of such a downward trend is that of two thousand years of Dutch body length: from the Roman period the trend was forever down, and was only changed sometime late in the nineteenth century, with the beginning of the modern trend that made the Dutch into what they are now: the tallest people in the world, and the tallest people in history.4

DEMOGRAPHIC STRUCTURE Demographic structure is a traditional and revealing indicator of economic performance. Modern societies are characterized by low birth rates and low death rates, and high life expectancies. This, however, has only been a very recent thing. Before sometime in the nineteenth century most Western societies had desperately high death rates, and needed high birth rates to Wll the empty spaces. Life expectancy at birth tended to be in the range 20–30 years, i.e. signiWcantly less than in the unhealthiest of modern Third World countries. Here, the argument is that of itself a demographic regime with high mortality, high natality, and a low life expectancy betrays an economy that was unsuccessful in at least some respects. Keith Hopkins’s brilliant seminal essay on the use of model life tables for Roman demographic history has spurred what I believe to have been the most important new view of Roman life of the last few decades: high and unpredictable mortality not only brought great personal sadness to many, but also created a social structure of great fragility, and invited high levels of social mobility.5 For me the highlights of 1 Lee and Devore 1968. 2 Sahlins 1968. 3 Cohen 1977. 4 Maat 2003: 57–88; see p. 63. 5 Hopkins 1966.

Archaeology, Demography, and Growth

117

this research have been Hopkins’s own work on the Roman Senate, and Saller’s work on the Roman family.6 My view of precarious life in disease-infected ancient Rome will never be the same again.7 Whatever Rome’s economic successes, it failed to provide that ultimate scarce good we have come to take almost for granted: a long and healthy life for many. No doubt this was a big failure common to all pre-industrial economies. And yet the explanation is more complex than just low standard of living. Life expectancy was usually lower in the cities, even if urban per capita incomes were often higher.8 Life expectancy of the rich, moreover, was usually almost as low as that of the poor.9 Death was democratic, for the simple reason that infectious diseases were the big killers, and they infected the rich as much as the poor. In fact, in so far as the rich preferred to live in the cities, they unwisely subjected themselves to a particularly murderous disease regime. Low life expectancy reXects the poverty of society rather than the poverty of individuals. Hopkins’s brilliant choice to invoke model life tables has taught us much about Roman population structure, but we would obviously like to have real ancient data as well, if only to study diVerences in real mortality patterns between, for example, town and country, but also between one period and another.10 Was pre-industrial mortality lower during periods of higher prosperity, or perhaps precisely the opposite (i.e. was mortality a price that had to be paid for prosperity)? By its nature Hopkins’s methodology for all its logical simplicity cannot answer such pressing questions. One often mentioned possibility to obtain real data would be to study ages at death from the skeletons of real Romans. If tombstones are misleading because they reXect commemoration rather than demography, perhaps the skeletons themselves can tell us more about reality. The problem is that it is hard to tell age of death from an adult skeleton, even if some physical anthropologists will 6 Hopkins 1983; Saller 1994. 7 Cf. Scheidel 2003: 158–76; Jongman 2003: 100–22. 8 de Vries 1984; Jongman 2003. 9 For ancient Rome: Scheidel 1999. 10 So far, the best real data are census data from Roman Egypt: Bagnall and Frier 1994.

118

Willem Jongman

gladly pronounce on it.11 If you can be wrong by a couple of decades, the range of uncertainty is far greater than what is still historically meaningful. As with tombstones, the chance to be buried was sometimes also age-speciWc, and the resulting age pyramid may thus reXect commemoration rather than demography. If that is so, we can indeed only judge the plausibility of such reconstructions of ancient age structure by the extent to which they conform to the model life tables we are trying to improve upon. New techniques may evolve, but with current technology analysis of human skeletons adds little to our understanding of the demographic structure of the ancient world. Age at death, however, is not the only thing a skeleton can tell us. Far more promising than the unreliable determination of age at death are studies of stature, (mal)nutrition, disease history, and workload. Here, the advances in physical anthropology and forensic science are opening up views of the human past that were unimaginable only a few years ago.12 Taken together, arguments from demographic structure remind us that antiquity was Wrmly located in the pre-industrial world, with the fragility and brevity of life that went with it. The hard evidence we have from Roman Egypt seems to locate it at the lower end of the pre-industrial range. The explanation may have been in high Roman population densities. If Rome was prosperous, that prosperity did not, it seems, translate into a long life.

DEMOGRAPHY A ND ECONOMIC GROWTH In the very long run of world history aggregate population and aggregate production both show trends that moved up for long periods of time, even if there were sometimes interruptions.13 What 11 Bouquet-Appel and Masset 1982; Mays 1998: 50; Cox 2000: 75. Cf. Molleson 1995. I am very grateful to Geertje Klein Goldwijk who made me aware of these publications. 12 Garnsey 1999 for examples from classical antiquity of the potential of what can be learned. A fascinating recent example is Prowse et al. 2007. 13 MacEvedy and Jones 1978.

Archaeology, Demography, and Growth

119

is not so clear, however, is the extent to which per capita incomes and prosperity also moved up, or perhaps moved down. Was population growth the product of rising prosperity, and was population growth in and of itself a further stimulus to the economy, or was it at the expense of standard of living (there were more people to share the same pie/land)? Thus any long-term economic history needs to address not only changes in aggregate population and aggregate production, but also changes in per capita incomes. Yet, and perhaps surprisingly, Roman historians have barely addressed the economy in those terms, even though the Roman economy poses a fascinating problem: population density was high by comparative historical standards (even if controversy remains about how high exactly), and still there now seem to be good reasons to believe that popular standard of living was equally high.14 The bleak neoclassical model predicts that that is not really possible. The millennial trend in population growth implied a change in the proportions of land and labour used in production. Thus, there was movement along what is called the production function, but not of the production function itself. A production function is a mathematical description of the relation between changes in the input of factors of production (land, labour, capital) and output.15 Movement along the production function describes how much more output we have if we increase the input of one factor of production. In that sense, it describes the consequences of a given level of technology. Unfortunately movements along the production function incur decreasing marginal returns. Under demographic pressure, more labour is used on the same land. Output increases, but not by as much as labour input: labour suVers from declining marginal labour productivity. Because wages are theoretically equal to marginal labour revenue productivity (i.e. everyone’s wages are equal to the extra revenue produced by the last worker), this is a grim tale: population pressure depresses wages, and the more quickly the more quickly marginal labour productivity declines.16 It is a logic that operated both in the very long run of the millennial trend, but also in the shorter run of the secular trend as 14 Jongman 1988: 65–154 for an early and more pessimistic eVort. 15 Every economics textbook will have a good explanation. I have learnt much from Stonier and Hague 1980. 16 Jongman 1988.

120

Willem Jongman

before and after the Black Death of the fourteenth century. It is a bleak scenario, because the underlying assumption is that any shortterm prosperity growth will lead to more births (humans are just rabbits), and thus to a subsequent reversal in the standard of living. As Ester Boserup and Jan de Vries have argued, however, there is an alternative scenario.17 After a period of population increases, and consequent diminishing marginal returns to labour, new agricultural technologies could be applied to use the additional labour more eVectively. When technology changes, given quantities of factors produce more output: the production function itself shifts. IntensiWcation is not only a process of gradually getting more in each other’s way, but also allows qualitative transformations in the organization of agricultural production, and changes in fallowing in particular. Hunting and gathering were succeeded by slash-and-burn agriculture, and they in turn by long fallow, the two-Weld system, the threeWeld system, annual cropping, and multicropping and irrigation agriculture. Each of these successive systems was indeed more labour-intensive, but made sensible use of that additional labour, rather than waste it. To this succession of Weld systems Jan de Vries has added the possibility of rural-urban specialization, as yet one more way to make productive use of the extra labour. The great success of the early-modern Dutch economy was not only due to its colonial brutality and its maritime trade, but equally to its specialized commercial agriculture. Historically, the question is simple, therefore: did per capita incomes decline as a result of population pressure, or not? And if not, why not?

T H E A RC H A E O LO G Y O F P O P U L AT I O N What neither model life tables nor physical anthropology can do, of course, is to tell us anything about (changes in) aggregate population levels. And yet, those are at least as interesting for our understanding of Roman economic performance. Was population density low, and thus standard of living probably high? Or was population density 17 Boserup 1965; de Vries 1974.

Archaeology, Demography, and Growth

121

high, with a low standard of living as a Malthusian consequence? Or, Wnally, was population density high because the economy was so successful? Thus far, much of the debate has been concerned with the ever more intricate interpretation of Roman census Wgures, in what I think is an increasingly futile because irresolvable debate between believers of the low count and disciples of the high count. Can we think ourselves out of the box? One alternative is to look once more at what was possible, and under what conditions. Part of the traditional argument against the high count was the view that Italy could not have sustained such a high population. My own earlier calculations are a case in point.18 Starting from the belief that standard of living was low and that diet was unlikely to have been much better than largely cereal-based, I proceeded to calculate the caloriWc output of Italian agriculture on the simplifying assumption that all of it was cereals. The grand total of this carrying capacity was more than was needed for the population of Italy of the low count, but signiWcantly less than what was needed if Italy had been as densely populated as the high count presupposes. Without simpliWcation, understanding is impossible, but oversimpliWcation is misleading. This analysis is indeed conWned to movements along the production function and ignores the potential for a shift of the production function itself. Some crops other than cereals produce far more calories per hectare: in the case of wine and olive oil some Wve times more calories per hectare than wheat.19 To do that, however, they need a much bigger labour input. Thus, wine and oil were expensive calories because they were labour-intensive. However, unlike cereal agriculture they could make good productive use of abundant labour. Paradoxically, if Romans were prosperous enough to consume the more expensive calories in wine and olives, there could also be far more Romans. In recent years, archaeologists have claimed that they may have the hard data to set the debate on a much surer footing. And indeed, in principle this could be true, even if in practice margins of error may still be large. Ester van Joolen’s work on archaeological land evaluation shows what is methodologically possible, and unpublished 18 Jongman 1988.

19 Jongman 2007a.

122

Willem Jongman

work by Leontien Spoelstra demonstrates that if conditions were right high population densities could indeed easily be sustained by the growth of other crops than just cereals.20 A commercial agriculture with crops such as wine, oil, or horticulture yielding large quantities of calories could have supported many rural people, and fed large urban populations. So what actually happened? Archaeologists have indeed tried actually to count numbers of people in the archaeological record. Archaeological Weld surveys have shown a dense pattern of surface scatters that are interpreted as traces of sites, both large and small. The rural landscape of the classical period was a landscape full of such sites, but the almost insurmountable diYculty is to translate these patterns of scatters into numbers of people.21 The Wrst diYculty is that no Weld survey ever succeeds in recovering all sites. Even with intensive survey methods substantial numbers of sites escape recovery, and it is often impossible to know what proportion they are. Sometimes, when the distribution pattern is very obviously regular, we can estimate what we are failing to recover. At other times this is impossible. Yet this does not mean that all is lost—it just depends on the argument we try to make. If we try to use survey data to argue for a low population density, we are obviously using a dangerous argumentum e silentio. If, on the other hand, we are using the recovered pattern as a prop for a high count, this is not so: if we had a higher recovery rate, our argument would only be stronger. This is important, because I do believe that even with low recovery rates, many areas already show patterns of pretty high density. The second problem is that it is not quite so easy unequivocally to translate sites into people. How small a site must we assume to be represented by a small scatter, and how many inhabitants does that imply? Similarly, even if large sites are quite few in number, their aggregate population may have been large since they were inhabited by far more people. But by how many more? As Robin Osborne has shown, the choices we make have major implications.22 The problem is aggravated even more if we appreciate that an argument about 20 van Joolen 2003; Spoelstra 2007. 21 Robin Osborne’s (2004) thoughtfully sceptical review of the value of Greek Weld surveys for demographic reconstruction is just as relevant for the Roman world. 22 Osborne 2004.

Archaeology, Demography, and Growth

123

absolute population density often has to serve a comparative historical purpose. We want to argue, for example, that ancient population densities were higher than those in medieval or early modern times. If methodologies to construct population estimates diVer for the diVerent periods, they have to be all the more accurate. Thus, the archaeological reconstruction of absolute population numbers is fraught with diYculty and uncertainty. My personal instinct is to view the aggregate meaning of a few decades of Weld surveying as a powerful argument for a rather fuller landscape than we imagined before, but that is obviously weak methodology. However, the survey data also allow a diVerent manipulation. If absolute numbers are hard to reconstruct, the same does not necessarily apply to relative numbers. Thus far, Weld surveying has been under the spell of the Braudellian paradigm of a largely unchanging pre-industrial economy constrained near subsistence by static technology and, particularly, ecology. If ecology is the most important variable, change becomes hard to understand, of course, let alone economic growth. And yet, time series analysis of survey results can and does show important changes over time. Moreover, this analysis does not suVer nearly as much from data problems as does the reconstruction of absolute population numbers. In eVect, it simply treats what we have as a sample, and the only problem it therefore needs to deal with is how representative the sample is: are there systematic biases? Finally, relative population change is what we needed anyway, to compare it to the relative change in output, and relative change in standard of living.

RO ME: PROSPERITY FOR MANY? Roman historians have largely failed to discuss the implications of prevailing high population densities. Discussion was concerned with the hard choice between the high estimate for the population of republican Roman Italy, and the low one. Since until recently the high estimate had little street credibility, no one seemed to have realized that even at the low estimate, population densities were

124

Willem Jongman

high by comparative historical standards.23 Now that the high count is gaining respectability, the need to think about the economic implications of population pressure is all the more urgent. The historical question is a simple one: did high and increasing population density erode the standard of living of the mass of the population, or not? It was the question I tried to answer twenty years ago (even if few critics appreciated the economic logic of what I tried to do).24 My answer at the time was that high population density did indeed depress popular standard of living, and increased social inequality. What I failed to do is document that mass poverty, perhaps because I could not possibly imagine how one could ever test this hypothesis. In recent years, however, such validation is coming within reach. For population trends the archaeological Weld survey work by Lisa Fentress and others has shown that for a while during the middle and late republic population in Italy probably did indeed increase, to reach a peak in the Wrst century bc or the Wrst century ad, and to decline again in later antiquity (the turning point would seem to have been in the second half of the second century ad). At the peak of this trend population density may well have been as high as the so-called high count interpretation of the census records suggests. What is problematic about these estimates is that they are based on small samples of mostly Italian territory, that they are based on divergent methodologies, and that they eschew systematic econometric metaanalysis. It is also fair to say that much archaeological evidence now points to increases in standard of living during precisely this period.25 Shipwrecks and metal pollution in the Greenland ice core had already showed dramatic rises in economic activity far exceeding the likely magnitude of population growth, and thus argued for per capita increases in shipping and metal extraction. The same now applies to a range of other activities, such as construction, meat consumption, Wsh sauce production, or human body length. For a while, ordinary Romans seem to have been much better oV than they had been for a long time, and people would be for a long time to come. The potential for detailed time-series analysis of such archaeological 23 Jongman 1988.

24 Jongman 1988.

25 Jongman 2007b: 183–99.

Archaeology, Demography, and Growth

125

data has only just begun to be utilized. There are amazing data out there, waiting to be given a voice. Bibliography Bagnall, R. S., and Frier, B. W. (1994). The Demography of Roman Egypt. Cambridge. Boserup, E. (1965). The Conditions of Agricultural Growth. Chicago. Bouquet-Appel, J. P., and Masset, C. (1982). ‘Farewell to palaeodemography’, Journal of Human Evolution, 11: 321–33. Cohen, M. N. (1977). The Food Crisis in Prehistory: Overpopulation and the Origins of Agriculture. New Haven. Cox, M. (2000). ‘Ageing adults from the skeleton’, in M. Cox and S. Mays (eds.), Human Osteology in Archaeology and Forensic Science. London, 61–81. de Vries, J. (1974). The Dutch Rural Economy in the Golden Age 1500–1700. New Haven and London. —— (1984). European Urbanization 1500–1800. London. Edwards, C., and Woolf, G. (eds. 2003). Rome the Cosmopolis. Cambridge. Garnsey, P. (1999). Food and Society in Classical Antiquity. Cambridge. Hekster, O., de Kleijn G., and Slootjes, D. (eds. 2007). Crises and the Roman Empire. Proceedings of the Seventh Workshop of the International Network Impact of Empire (Nijmegen, June 20–24, 2006). Leiden. Hopkins, K. (1966). ‘On the probable age structure of the Roman population’, Population Studies 20: 245–64. —— (1983). Death and Renewal. Cambridge. Jongman, W. M. (1988). The Economy and Society of Pompeii. Amsterdam. —— (2003). ‘Slavery and the growth of Rome: The transformation of Italy in the Wrst and second century bce’, in C. Edwards and G. Woolf (eds.), Rome the Cosmopolis. Cambridge, 100–22. —— (2007a). ‘The early Roman Empire: Consumption’, in W. Scheidel, I. Morris, and R. P. Saller (eds.), The Cambridge Economic History of the Greco-Roman World. Cambridge, 592–618. —— (2007b). ‘Gibbon was right: The decline and fall of the Roman economy’, in O. Hekster, G. de Kleijn, and D. Slootjes (eds.), Crises and the Roman Empire. Proceedings of the Seventh Workshop of the International Network Impact of Empire (Nijmegen, June 20–24, 2006). Leiden, 183–99. Lee, R. B., and Devore, I. (eds. 1968). Man the Hunter. New York. Maat, G. J. R. (2003). ‘Male stature: A parameter of health and wealth in the low countries, 50–1997 ad’, in W. H. Metz (ed.), Wealth, Health and Human Remains in Archaeology. Amsterdam, 57–88.

126

Willem Jongman

MacEvedy, C., and Jones, R. (1978). Atlas of World Population History. Harmondsworth. Mays, S. (1998). The Archaeology of Human Bones. London. Metz, W. H. (ed. 2003). Wealth, Health and Human Remains in Archaeology. Amsterdam. Molleson, T. (1995). ‘Rates of ageing in the eighteenth century’, in S. R. Saunders and A. Herring (eds.), Grave ReXections: Portraying the Past through Cemetery Studies. Toronto, 199–222. Osborne, R. (2004). ‘Demography and survey’, in S. E. Alcock and J. F. Cherry (eds.), Side-by-side Survey: Comparative Regional Studies in the Mediterranean World. Oxford, 163–72. Prowse, T. L., Schwarcz, H. R., Garnsey, P., Knyf, M., Macchiarelli, R., and Bondioli, L. (2007). ‘Isotopic evidence for age-related immigration to imperial Rome’, American Journal of Physical Anthropology 132.4: 510–19. Sahlins, M. (1968). ‘Notes on the original aZuent society’, in R. B. Lee and I. Devore (eds.), Man the Hunter. New York, 85–9. Saller, R. P. (1994). Patriarchy, Property and Death in the Roman Family. Cambridge. Scheidel, W. (1999). ‘Emperors, aristocrats, and the Grim Reaper: Towards a demographic proWle of the Roman elite’, CQ ns 49.1: 254–81. —— (2003). ‘Germs for Rome’, in C. Edwards and G. Woolf, Rome the Cosmopolis. Cambridge, 158–76. Spoelstra, L. (2007). ‘High Count or Low Count? Een studie naar bestaansmogelijkheid, gebaseerd op landbouwgewassen in de Pontijnse regio’. Unpublished BA thesis, Groningen University. Stonier, A. W., and Hague, D. C. (1980). A Textbook of Economic Theory. London. van Joolen, E. (2003). Archaeological Land Evaluation: A Reconstruction of the Suitability of Ancient Landscapes for Various Land Uses in Italy Focussed on the First Millennium bc. Ph. D. diss. Groningen University.

5 Peopling the Countryside: Roman Demography in the Albegna Valley and Jerba Elizabeth Fentress

The use of the comparative method involves quite enormous though I shall not say insuperable problems . . . Unfortunately positivism, which assumes that the evidence is available on the ground and can be divided into 100 m. square grids for comparison by computer or other allied techniques does not actually solve the problem either. (E. Hobsbawm 1979: 248)

This chapter builds on the Wne work recently carried out by Robert Witcher on the area around Rome, which he loosely calls the suburbium, as well as two projects carried out in the last twenty years.1 The Wrst, the Albegna Valley, was published without the relevant chapter, while the second, the Jerba Project, is in press.2 However, Phil Perkins has published good, although still preliminary, calculations for the Albegna Valley for the period up to 50 bc.3 In this chapter I want to discuss some of the issues relating to the calculation of population from Weld survey, and then give the calculations for our two areas for the Augustan period, a high point in each case. Perhaps more interesting, however, is the further information that comes from Xow data, recording the change over time in the population and its habitat. Some aspects of this will be sketched here. 1 Witcher 2005: 120–38. Carandini et al. 2002; See now Fentress et al. 2008. A preliminary report in Fentress 2001. See also Attolini et al. 1991. 2 For the Albegna Valley, Attolini et al. 1991; Carandini et al. 2002; for Jerba, preliminary report in Fentress 2001 and see now Fentress, Drine, and Holod 2009. 3 Perkins 1999.

128

Elizabeth Fentress

Although the calculations from both projects inevitably contain a substantial margin of error, the comparisons between the two, and between any one period and the next, strengthen and support the conclusions from just one period or one project, giving the ‘positivist’ approach to data which is, after all, generated from a large number of individual hectares, a comparative support. Every area has its own particularities—we would expect higher concentrations in an oasis or in the suburbs of a major city—and has to be taken in its context. But neither Jerba nor the Albegna oVer unique environments, and it seems plausible to base some generalizations on them. The Wrst step in making demographic calculations is, of course, to estimate the total number and type of sites in a given period in a given area. Ideally, we would survey the area under plough and count the sites, although as anyone used to dealing with survey data knows, the large number of techniques that have been adopted over the past Wfty years have changed our estimates as to the reliability of such counts.4 Since the 1970s, more or less intensive approaches have quite justly prevailed, but even then approaches to—or even the acceptance of—sampling diVer widely, ranging from Forma Italiae’s standard coverage of an IGM map sheet of 100 km2, which can be assumed to represent itself alone, to our own systematic transect sampling with added purposive samples, and various compromises in between. Witcher selects six surveys, all of them isolated blocks, with a total of 935 sites from 323 km2, and uses these data to estimate a total number of Roman sites of 18,100 for the area within a 50-km radius of Rome, an area of roughly 5,415 km2, once allowances are made for sea, marsh, the city of Rome, and land over 650 m in elevation. Now while each of these surveys is reasonably intensive, it is worth asking whether they actually constitute a viable sample of the territory of Rome, insofar as the two largest are from the Alban hills (Tusculum and Tibur), with a notoriously high density of sites. Archaeologists rarely go willingly into areas that do not promise to yield sites, and the uniformly high densities represented here may not Wnd echo in areas with forest or rough ground. However, this is to some extent a quibble, and Witcher’s estimate that each of these square kilometres had a villa and two farms, while one in a hundred 4 For recent work on demography and Weld survey Sbonias 1999, Tre´ment 1999.

Peopling the Countryside

129

has a village, and one in two hundred a road station, may be taken, with caution, as acceptable, and gives a density of 60 people per km2 within 50 km of Rome, and, using another set of surveys, 42 between 50 and 100 km. If we want to be more precise, however, we have to look at projects with systematic samples, like the unpublished Tuscania survey5 and the Albegna Valley survey (Fig. 5.1), in which transects (in both cases) or randomly selected samples (Tuscania) were used to give a sample of an entire area rather than its particularly rich bits. As the Jerba Valley survey design shows, I have stuck stubbornly with the original transect scheme, which is intuitively satisfying in that it samples all land forms from watershed to valley bottom in the proportion that they occur.6 I am here fortiWed by a remark by Albert Ammerman that regular patterns will make better sense to future generations and are a lot less diYcult to work with.7 This approach is coupled with another that is not particularly followed in the surveys cited by Witcher. That is that a sampling design should represent a clearly deWned geographic entity rather than something like an IGM map sheet which is not so much systematic as arbitrary. A correlate to this is that the sample should be large enough to mean something, thus the 3 km from Cures used by Witcher mean exactly that— 3 km.8 Although we will always be in the dark about the population we are sampling it is no good trying to say something about an unknown population using thirteen sites—even though, admittedly, the Cures survey (from which that number is derived) was designed to test a previous, less intensive approach. It is simply not a big 5 Preliminary reports in Barker and Rasmussen 1988; 1991. However, the results are published in such a way as to make them unusable for our purposes. 6 Objections to sampling include that voiced by Mattingly at the conference, that we might miss a substantial urban site in the interval between the samples. The reply to this was given by Flannery (1976: 159) 30 years ago in response to the same objection, roughly, that ‘you couldn’t miss Teotihuacan if you tried’ (his classic book, particularly its chapter on sampling, should still be required reading for students). This is valid for the Etruscan and Roman towns in the valley as well: in other words, extensive survey and local knowledge will tend to give a reasonable representation of major sites outside the area sampled: archaeologists are not artiWcially blinkered. In the Albegna Valley all known villa sites were recorded, regardless of the sample, but they have not been counted for the purpose of this study. 7 Pers. comm. c.1995. 8 Di Giuseppe et al. 2002.

130

Jerba

Albegna Valley Probabilistic sample

Survey transects (1996–2000) Purposive sample

Hu-mit al-su-q

Sidi jmu-r

EL

SA

N

A

Aghi-r

AL

BEG

Ajiim

N Developed areas/not walked Probabalistic sample (random hectares) Purposive sample (non-random hectares)

0

5km

Fig. 5.1. Albegna Valley and Jerba Surveys, sampling designs

0 1 2 4 6 8 10 Km

Elizabeth Fentress

Midu-n Sidi Garu-s

Peopling the Countryside

131

enough sample to be signiWcant or meaningful and should not be used.9 Further, the intensity of the survey must be matched by proper conditions for visibility in at least some of the area. People undertake surveys in woodlands, or in areas where nothing whatsoever is ploughed.10 This is a problem. Peter van Dommelen and others have recently tried to counteract the worrying tendency of many areas of the Mediterranean to fall out of cultivation by hoe-clearing little patches of earth at regular intervals,11 which is laudable, and perhaps the best one can do, but the very real problem in much of Italy is that Ward-Perkins was right in saying, in 1955: ‘If this material is to be recorded, the record must be made at once’.12 A combination of deep ploughing and abandonment of marginal land has left the possibilities for fruitful survey in very much worse shape than they were Wfty years ago. We may, indeed, have to rely on surveys such as the South Etruria Survey, now being prepared for publication in a third-millenium format, rather than attempting a new one, because in many cases that chance has been missed. This brings us to a Wnal problem that survey aYcionados are bored with, but which some have not, perhaps, yet heard of: background noise or, as it is more correctly known, oV-site material, has in some way to be taken into account. But how? I have argued elsewhere that oV-site material most likely represents sites which have been so destroyed as to not leave a clear signal in the ploughsoil.13 This is, however, unquantiWable. At best it can serve as a visual correlate to a map of sites and a statement of their numbers, conWrming evidence of absence or warning us of the likelihood that we are not seeing the full picture. But this should not put us oV from recording it, simply restrain us from trying to quantify it in any meaningful way. I now turn to the Jerba survey, which, in a sense, presents far fewer problems than the Albegna Valley survey. The survey was carried out 9 For a longer analysis on the importance of absolute sample size in Weld survey Fentress 2000a; for proportionate sample size Fentress 2002b, with a simulation of the results of diVerently sized samples on the same material. 10 On visibility and site recovery see Terrenato and Ammerman 1996. 11 van Dommelen and Sharpe 2004. 12 Ward-Perkins 1955: 55. 13 Fentress 2001.

132

Elizabeth Fentress Ostia Puteoli

Caralis

Lilybaeum Carthage

Syracusa Cossyra

Hadrumetum

Melita

Lepti Minus Cercina Tacape Gigthis

Jerba

Girba Meninx Sabratha

250 km

Oea

Lepcis Magna

Fig. 5.2. Location of the island of Jerba

under the direction of Renata Holod, Ali Drine, and myself between 1995 and 2000. The island of Jerba lies just oV the coast of southern Tunisia, between the Greater and Lesser Syrtis (Fig. 5.2). It covers an area of just under 600 km2. The landscape variability is slight, although perhaps half of the land is unsuitable for irrigated agriculture because of the lack of decent available groundwater: there is minimal rainfall and there are no springs. The visibility was uniformly good

Peopling the Countryside

133

because of the almost total absence of ground cover, although without those startling moments you get in central Italy when it is clear that a very large plough has just taken out a major site, and you have the impression that you are looking at an entire villa which has been put through a blender. We walked blocks of one hectare at a time, recording sites in the usual fashion, but also the presence or absence of pottery from pre-Roman, Roman, early and late Medieval, and early Modern periods This gives us a certain control on the areas where sites were probably present. In looking at the demography I start with the Early Roman period (Fig. 5.3), although the same type of calculation will be used for all other periods. The Early Roman period gives us a maximum, and allows us to compare our data with that of Witcher. It is also a useful period to begin with as it is that in which we are most certain as to whether a site was occupied or not. Early Arretine sigillata, with a date range of perhaps Wfty years, is ubiquitous, and under reasonable conditions absurdly easy to Wnd. If a site does not have it, it is reasonable to assume it was not occupied. While the date range we give for this period is wider, from 50 bc to ad 100, in fact practically all the sites for both surveys are dated by this pottery, and were thus occupied at the same time, roughly the Augustan period. For these estimates we have used only the data from the background sample. This represents only half of our sites, and just 13.4% of the area of the island. However, as the purposive sample was all drawn from the most heavily occupied part of the island, it would have provided a signiWcant sample error. As in Witcher’s case, we are using some arbitrary assessments of population per site. However, as he points out, these provide answers inherently more plausible than those derived from size alone; estimates of this sort record population Wgures derived from a calculation of, for example, a pottery scatter covering 60 m2 with a density of 1 person per 10 m2, yielding a population of 6.14 The procedure, like that of Witcher, was simply to count the number of sites in each of the rural categories—farms, villages, and villas—and to multiply them by standard ‘household’ sizes: 5–10 in the case of the farms, representing a low estimate which would be a single nuclear family, and a high estimate which adds on 14 For examples of this sort of evidence see Hassan 1981: 63–6.

134

Elizabeth Fentress B002

B001 Hu-mit al-Su-q

A001

Ghi-zin

G002

G024 E004 Sidi Jmu-r E007

G089 G049

Bu- rgu-

F022 G015

G056

F042 G084G090

F054

G087

Sidi Garu- s K002 H004 K006 K182 L006 K021 K048L007 K019 K165 K107 K056 Aghi-r K027 K032 K037 K023 K135 K049 K050K046K047 K144K176 K008

J020 J023 J026

I001

Haribus Aji-m

Suq al-Qibli-

K150 K149 Ta-la K202 O003 O010

Meninx

O011 O013

Zitha

Early Roman Sites

N

Urban centres

kiln

village/hamlet

Roads? (hypothetical)

farm

villa

Developed areas/not walked

necropolis

Ceramic scatter (all Roman)

mausoleum

0 1 2

6

4

8

10

km

Fig. 5.3. Early Roman sites on Jerba

dependants. Now, 5 people is the commonest standard estimate of household size. Witcher suggests it is low, saying that it does not take into account high infant mortality, conscription, and so on.15 On the other hand, the Wgure would accurately reXect the size of a family living in a house, after infant mortality and conscription had taken 15 Witcher 2005: n. 52.

Peopling the Countryside

135

their toll, rather than the actual number of children born to the family—in other words there is a diVerence between household size and family size.16 However, I am here using the average of a range of between 5 and 10, as we did not attempt to class our farm sites into large and small ones, and there is every likelihood that a larger site might have supported a number of dependants. In the case of villages, I have assumed 8 to 16 households, giving a Wgure of 40–80 (compare Witcher’s 50–200 with an informed guess of 100). For villas we estimate a population between 30 and 50 (compare Witcher’s 15–50 with an informed guess of 25). The estimates, then, are not the same as Witcher’s but they are comparable—slightly higher in the case of a villa, lower in the case of a village, and almost identical in the case of a farm. The minimum and maximum population totals were then multiplied by 7.47 to restore the sample to 100% and calculate the probable number of sites in the areas not surveyed. The procedure was carried out twice, Wrst on the number of certain sites, and then counting in the possible sites as well. We cannot, for the usual reasons, add anything for the ‘missing’ sites, but counting the ‘possibles’ should go some way towards compensating for this. This iteration gave us a range between a minimum (certain sites, minimum population estimate) and maximum (all sites, maximum population estimate). We averaged these two, which may give a more reliable Wgure than the extremes of the range. However, no consideration has been given to the ‘missing’ sites that are presumably disguised by the oVsite scatters shown in pale grey, as there is no way of estimating their numbers. This will to some extent aVect the number of farms recorded, although probably not those of villas. Thus the crude average (rounded, for these purposes, to the nearest 500) is probably itself a conservative estimate. In urban contexts—the two cities, B urg u and Meninx, and the two ports, Haribus and Ghı¯zen—we have used the Wgure of 120 people/ ha. This is derived from the evidence from the excavations illustrated here (Fig. 5.4): we have reconstructed the insula examined as covering around 38  50 m. If allowance is made for four roads, this gives 16 For example, Scheidel’s estimate that 1 in 2 men would have had to enlist at 17 years of age (1996: 94) would have removed any serious chance of keeping an extended family.

136

Elizabeth Fentress

Fig. 5.4. Reconstruction of an insula at Meninx

roughly a quarter of a hectare. If, as reconstructed, there were six houses of varying sizes on the insula, and each was occupied by a family of 5,17 the population per hectare would have been 120. Again, this is at the low end of the scale, and Wgures using 200 people/ha are 17 Again, this number is apparently low, but it takes account of the varying sizes of the houses. Further, as Perkins 1999: 109 points out, it Wts a population with life expectancy at birth of 25–35.

Peopling the Countryside

137

Table 5.1. Population estimates for Early Roman Jerba (rounded to signiWcant Wgures) Certain Rural Sites Farms x 5 Farms x 10 Villages x 40 Villages x 80 Villas x 30 Villas x 50

59 59 11 11 23 23

Total rural

Pop. (*7.47)

295 590 440 880 690 1,150

2,200 4,410 3,290 6,580 5,150 8,590

Total

83 83 17 17 25 25

Pop. (*7.47) Min.

Max. Average

415 3,100 2,200 6,200 830 6,200 680 5,080 3,290 10,200 1,360 10,200 750 5,600 5,150 9,340 1,250 9,340

Total urban

6,750 7,250

10,640 25,740 18,200

Urban sites Min. (ha) (*120) (*200) Max. (ha) (*120) (200) Meninx 26 3,120 5,200 60 7,200 12,000 3,120 12,000 B urg u 21 2,520 4,200 21 2,520 4,200 2,520 4,200 Ghı¯zin 5 600 1,000 5 600 1,000 600 1,000 Haribus 8 960 1,600 8 960 1,600 960 1,600 Total population

4,200

7,560 3,360 800 1,280

7,200 18,800 13,000 17,840 44,540 31,200

also given.18 Again, where there is doubt about the size of the site in a given period minimum and maximum estimates are given, the average calculated and rounded. The results for the Early Roman period are shown in Tables 5.1 and 5.2. The Wrst thing to notice about these Wgures is the relative unimportance of the isolated farms, whose population constitutes only 13% of the total. Now, as it is the farms which we are most likely to miss, this suggests that our ‘missing’ sites would not radically change the total Wgures. Far more important is the Wgure per hectare assigned to the urban sites, which changes our Wgures radically. Indeed, I have adopted a somewhat small calculation for Meninx, insofar as it is impossible to calculate how much of the town was actually occupied in this period, as we cannot distinguish between massive ‘slag heaps’ of shells, which cover almost everywhere, and the areas which were actually inhabited at any given moment. The 26 ha 18 This would assume a population of 8 per house. Wilkinson 1999: 46 gives a bewildering number of Wgures used to calculate urban populations per hectare, ranging from 100 to 1,200. However, the general range tends to be around 100–200. See also Hassan 1981: 66. Wallace-Hadrill 1995: 95–6 puts the population of Pompeii at 120–300/ha, using roughly similar methods to ours. See most recently Chamberlain et al. 2006: 50–5.

138

Elizabeth Fentress

cited here as a minimum Wgure represents, in a positivist fashion, that area of the town on which Wrst-century pottery was actually found by our survey—in fact, the map of these shows that very likely the whole of the town was occupied at this point. However, no deductions have been made for public buildings and open areas, so that the best guess of 43 ha (the average of the minimum and maximum), will have to do: for comparison, this area is just under 1/8 of the size of Carthage. Finally, I should note that we may well be missing much of the settlement at Haribus, which appears to be largely under the sea, and thus possibly signiWcantly bigger than the 6-ha site visible today.19 The ‘best guess’ population of the island in the Early Roman Period is 31,000, with roughly 50% of them living in urban contexts. This gives a population density of 52 people per km2, which is close to Witcher’s 60 for the inner suburbia. However, if we take away the urban population the density is reduced to 30, which, given the geographical diVerences between a distant island and the hinterland of the megalopolis, seems a more reasonable Wgure. Compare, for example, the estimate of 70 per km2 from the far richer Fayum cited by Rathbone for the third century bc.20 The danger, of course, is that we are willing to accept or reject these Wgures precisely on the grounds of their apparent plausibility, which opens the door to accusations of trying to make it come out ‘right’. Common sense is a trap that is always open. My instinct would be to raise the estimate of the population of Meninx even higher; my method does not allow this, but others may, of course, wish to adopt the higher Wgures. The growth rate from the previous period, where the same sort of calculation provides a population of 25,500, can be accommodated within a ‘normal’ growth rate of around 0.16% per year over 200 years.21 We thus can see no compelling evidence for colonization on the island, or for any signiWcant number of immigrants. However, as 19 The consonants of its name are suspiciously like those of the town of Girba known from later sources, which gave its name to the island. For an extended (but inconclusive) discussion of this possibility, see Fentress et al. 2009: 174–6. 20 Rathbone 1997. 21 Although much higher rates are possible over short periods, this is close to the Wgure of 0.1% per year which Hassan 1981: 221 uses for Palaeolithic populations. For comparison, growth in Europe after the Black Death was around 0.3%.

Peopling the Countryside

139

all of the growth occurred in urban contexts, which grew at a rate of 0.38% compared to a complete lack of growth in the countryside, we may imagine that it was stimulated to some degree by the movement to the cities of any surplus rural population, and probably drew from the mainland as well. Further, some of the rural labour force could have been replaced by slaves imported from sub-Saharan Africa, for Meninx seems to have formed an alternative terminus to Lepcis Magna for the slave trade through the Fezzan.22 We are thus probably seeing a diminution in the free population of the countryside. Whether or not the absolute Wgures are accurate, what is interesting here is the distribution of the population (Fig. 5.5). Small farms are relatively few. The relative importance of the agglomerations— villas and villages—is striking; 45% of the population is found in them. Their dominance in rural settlement reXects a high degree of labour concentration in agricultural production. This labour was probably managed at least in part on private estates, where the workforce could cooperate during periods of intensive activity, such as pruning and harvesting, or the heavy maintenance involved in irrigated agriculture. This pattern of cooperation in intensive agriculture would result in a more signiWcant surplus value from the crops. As the urban population grew, the demand for agricultural surplus would have expanded, creating new stimulus for the rural economy, particularly in the market garden sector in the vicinity of Meninx. However, there is little doubt that the murex purple trade remained the principal source of income at Meninx itself. Further, the high proportion of urban to rural population makes it likely that grain was being imported in considerable quantities. It is generally assumed that 75%–80% of the active labour force is needed to support a non-farming urban population.23 That could not be the case here, with 42% of the population in the cities, and a large, if unknown proportion of the rest involved in the production of an export crop, which I have argued elsewhere was passum wine,24 that would have contributed relatively little to the diet. Indeed it seems unlikely that the island could have produced enough grain for a population this size even if the workforce existed. Michael Frachetti 22 Fentress et al. 2009: 207. 23 Hassan 1981. 24 Fentress 2001.

140

Elizabeth Fentress Distribution of population in Jerba, Early Roman Period

Fig. 5.5. Distribution of population in Jerba, Early Roman period

Farms 11%

Towns 49%

Villages 20%

Villas 20%

has calculated that Jerba’s richest soils could have yielded approximately 1,550 metric tons of wheat using pre-modern techniques.25 The potential grain yields would be suYcient for a population of 6,000–8,000 assuming that each individual consumed roughly 200 kg of grain per year, and that all of the arable land was devoted to its production. There is thus no doubt that Meninx relied on imported grain. Rather than farming grain, the rural labour force was probably employed in more intensive irrigated market gardening, as well as wine and possibly oil production. If we turn to the more familiar circumstances of the Albegna Valley the diVerence is striking. Here we are dealing with a much older survey, carried out by myself and Maria Grazia Celuzza with a large team between 1978 and 1984. Again, we used a transect survey, and walked intensively. The relationship between the sample transects and the actual geography of the valley has been well studied by Perkins, and seems to provide an accurate reXection of the variables with the valley.26 The transect sample amounts to 249 km2 out of a total area of 1,428, or 17.5% of the valley.27 Conditions of visibility were more variable than at Jerba, as, while much of the land was deep ploughed, 25 Frachetti in Fentress et al. 2009: 64. 26 Perkins 1999. 27 We have used here the transects of the Albegna Valley without the purposive additions, although the purposive samples of the Valle d’Oro and Capalbio were maintained. This would tend to slightly overestimate the population.

Peopling the Countryside

141

some was under vines and already some of the higher ground was falling out of cultivation. As a rule of thumb, we estimated that some 75% of each transect was actually visited under good conditions. This is hardly the Wrst attempt to calculate population totals in the Albegna Valley: I made one calculation for the early Empire in the late 1980s, published by Carandini and Cambi in 2002,28 Perkins made a series for the Etruscan and Republican periods in 1999,29 and Cambi calculated for the third century and the early Republic in the same year.30 The upshot is that there is little relationship between these Wgures—or rather, there is some relationship between mine and Perkins’s, and no relationship whatsoever between those and Cambi’s. Some of the diVerences can be explained, others at present are not understandable (but are being subjected to a Wnal recheck of the data from the original sheets). What I have done in this iteration is to take the data and treat them almost exactly as the Jerba data, with some exceptions. Our population estimates are a little diVerent from those of Jerba, in that we distinguished three diVerent types of farm, designated as: the curiously named ‘Farm/tomb’, which simply expressed radical uncertainty as to the nature of small scatters; Farm 1; and Farm 2. I have thus avoided maximum and minimum estimates, assuming that half of the Farm/ tombs were actually of Farm 1 type, attributing a single nuclear family to a Farm 1, and assuming a family plus Wve dependants in a Farm 2. The villa and village estimates remain the same. Thanks to the excavations at the colony of Cosa (Fig. 5.6) we have a fairly good idea of the density of the urban population over the Roman period. The housing at Cosa looks very diVerent from our excavations at Meninx, with ground set aside for a garden in each house plot.31 At its (re)foundation in the early second century, I have calculated that there were 224 small houses and twenty-four larger ones: counting Wve per household for the small ones and ten for the larger ones, this gives us 1,360 people, with a density per hectare of roughly 100—smaller than the minimum Wgure calculated at Meninx, although plausible when one considers the extensive amount of land devoted to kitchen gardens in the Republican colony.32 Other 28 Fentress 2002a: 191. 31 Fentress 2004, Wg. 12.

29 Perkins 1999: 168. 30 Cambi 1999. 32 Fentress 2004: 25 and Wg. 10.

142

Elizabeth Fentress

Fig. 5.6. The colony of Cosa c.150 bc: reconstruction

Republican colonies, Saturnia and Heba, are given the same densities, as is the slightly mysterious site of Orbetello, which might just have been a second site for the location of the Cosan colonists.33 However, an increasingly small area of land at Cosa was inhabited in subsequent periods. The Augustan re-foundation covered just over 5 ha, and could hardly have held much more than 500 people (Fig. 5.7).34 In the second century much of this was abandoned. A small resettlement in the third century covered only 3 ha, giving a maximum of 300, while in subsequent centuries the site was all but abandoned.35 Heba also occupies 14 ha, and follows a similar trajectory, while at Saturnia the epigraphic evidence for the early third 33 We give here 672 families divided between Cosa and Orbetello, which would account for about 17% of the perhaps 4,000 colonists deducted to Cosa in 273 bc (see Celuzza 2002: 106–7). There is no certainty that Orbetello was occupied by colonists: the walls, however, are identical to those of Cosa, and some stratigraphy, as well as various tombs, dates to this period. See Ciampoltrini 1995. 34 Fentress 2004: 32–4 and Wg. 14. 35 Fentress 2004: Wgs. 25 and 29.

Peopling the Countryside

143

Fig. 5.7. Cosa in the Augustan period

century contrasts with the complete lack of Wnds for any period after the Wrst. A well-known testamentary inscription seemed to allow Duncan-Jones to calculate that the free urban population of Saturnia in the Severan period was between 945 and 1,720, although Jacques subsequently reduced this estimate to 500–700. This might tempt us to trump the archaeological evidence with text, although the lack of evidence of any pottery at all does leave me somewhat suspicious.36 The calculations for the period between 50 bc and the end of the Wrst century ad come out at 31,000 (Table 5.2). This gives a population density of 21 per km2, or 19 if only the rural population is taken into account, a density less than 40% of that of Jerba. For comparison, the ISTAT data for 1936 gives a density of 22 per km2,37 in a time when the Maremma was still in part a malarial swamp. Also 36 CIL 11. 2650; Duncan-Jones 19822: 272–3; Jacques 1993. 37 ISTAT data. See, however, Lo Cascio’s strictures against the use of these comparisons (1999a: 169).

144

Elizabeth Fentress

Table 5.2. Population estimates for the Albegna Valley 50 bc–ad 100 (rounded to significant figures) Certain Pop. (*5.70) Rural Sites Farm/Tomb x 5 Farm 1 Farm 2 Villages x 40 Villages x 80 Villas x 30 Villas x 50

48 115 86 9 9 65 65

120 575 860 360 720 1,950 3,250

684 3,280 4,900 2,050 4,100 11,100 18,500

Total sites Pop. (*5.70) Min.

52 116 86 9 9 65 65

130 580 860 360 720 1,950 3,250

Total rural Urban sites Cosa Saturnia Heba Total urban

741 3,300 4,900 2,050 4,100 11,100 18,500

Max.

684 741 3,280 3,300 4,900 4,900 2,050 4,100

Average

713 3,290 4,900 3,080

11,100 18,500 14,800 22,000 31,500 26,800

Size (ha) 5 24 14

(*100) 500 2,400 1,400 4,300

Total

4,300 4,300

4,300

26,300 35,800 31,100

in contrast to Jerba, only 14% of the population appear to live in urban contexts, while 47% are found on villas employing intensive agricultural techniques and, as we know from the work of Carandini, exporting their surplus. Although, as Perkins points out, we have no idea what the carrying capacity of the valley actually is, the fact that it appears to have been an exporting region does tend to suggest that the population had not reached that capacity. The calculated density is, however, only half that suggested by Witcher for the area around Rome. If both Wgures are reliable it goes some way to demonstrate the very special nature of the suburban population, and allows us to consider the case of the Albegna Valley as a more normal density for a producing region. However, we should bear in mind that it is considerably less than Lo Cascio’s estimate for the average density of 28 per km2 of the lands of the Etruscans, Romans, and Latins in the third century bc38—a time when, be it noted, our own estimates give a density between 5 and 7. Like Witcher’s results, this seems to support the lower range of estimates, those of Beloch, Brunt, and 38 Lo Cascio 1999a: 169.

Peopling the Countryside

−400

−200

0 Min.

200 Max.

400

145

600

800

Average

Fig. 5.8. Albegna Valley: population minima and maxima

Hopkins,39 for the population of Italy at the beginning of the empire—although even here Witcher calculates that Beloch’s estimate would yield 48 per km2. So how reliable are these Wgures? And is it possible to wriggle out of this by dismissing them? Our recovery of 839 Roman sites gives 3.3 per km2, not dissimilar to the area around Rome, although the typology is diVerent. Is there any way of estimating the ‘real’ density? First, a look at the range between maxima and minima shows just how shaky the average Wgures are (Fig. 5.8). Only in the early empire does the range shrink to under a third of the total. But things get worse. Cambi has pointed out that the third-century Wgures for the territory of Cosa, where we may assume a deduction of 4,000 colonists and thus at least 12,000 people, are woefully short.40 A similar situation is clear in the territory of Saturnia. There we reconstructed a hypothetical distribution of twenty lots, arranged, pace Wilson, per strigas (Fig. 5.9).41 Seven small farms were found on them in a particularly attentive search. This would suggest that we are failing to see fully 2/ 3 of our sites. Now, this would be worrying indeed if it were true, but 39 Beloch 1886; Brunt 1971; Hopkins 1978. 40 Cambi 1999. 41 Fentress and Jacques 2002, Wg. 43; Wilson 2004.

146

Elizabeth Fentress

Fig. 5.9. Centuria in the territory of Saturnia: internal divisions hypothetical

there are a number of mitigating factors. First, the assumption that we are missing sites, while highly probable, does not apply to all sites, only to those in the size range between 25 and 200 m2, or, precisely, the small farms. Villas, and even large farms, are very hard to miss. Second, the century in question is right next to Saturnia, and it is logical to assume that at least some of the allotments were cultivated by people living in the city. Cambi’s criticism for the territory of Cosa, however, is certainly correct up to a point, although I have some diYculty with the entity and duration of the third-century colony of Cosa, for which neither I, nor anyone else, has ever found any houses.42 In both cases, however, these may be very ephemeral settlements indeed, which survived long enough to build the walls but not much else, knocked out by a new wave of recruitment at the time of the Second Punic war, or some other unrecorded cataclysm. 42 Ciampoltrini 1995.

Peopling the Countryside

Urban 14%

Farm1 13% Farm2 16%

Villa 47%

147

Fig. 5.10. Albegna Valley: distribution of population, 50 bc–ad 100

Village 10%

This might mean that the third-century rural settlement, too, was far more ephemeral than we imagine. But the main point is, as BintliV has pointed out,43 that even if we assume that we are underestimating the small farms by as much as 200% the change is not enormous. If we use the Wgures for the late Republic and early Empire, the added 8,000 people we would gain by tripling the small farms would only amount to 21% of the total, bringing up the population to 39,000, and the density to 27 per km2. This might suggest that the maxima are more reliable than the average Wgures, but it makes it very, very diYcult to bulk up the Wgures in this relatively favoured region of Italy until they resemble the density suggested by Lo Cascio’s calculations. Indeed, if we accept Beloch’s estimate that an area of roughly 100,000 km2 was occupied44 we can see that a far higher density in the suburban area and Campania would be required in order to oVset the low density represented by our highest Wgures, if we extend it down the coast of the Maremma. Of the population how many were slaves? The easiest way to go about the calculation is to assume that everyone living in the villas was a slave, which is clearly an exaggeration but perhaps not too much of one,45 and that everyone living in the towns, farms, and villages was free, which will be equally untrue, although in the other direction. This leads us to conclude that 47% of the population of the valley during this period were slaves (Fig. 5.10). This is higher than 43 BintliV 1999; see also Sbonias 1999. 44 Beloch’s estimate that 40% of the territory was occupied gives a rough and ready Wgure of 100,000 km2 (Beloch 1886: 417; see also Jongman 1988: 65–7). 45 Although see Marzano 2007: 125–48.

148

Elizabeth Fentress

Beloch’s estimate of roughly a third of a population of 6 million,46 but probably Wts if one considers the very low level of urbanization in the valley: elsewhere, particularly closer to the Apennines, the proportion of free-born would have been higher. Some years ago Rathbone suggested that we were seriously underestimating the free population of the countryside, as no villa would have maintained yearlong the number of slaves necessary for the harvest.47 In fact, Cato talks of at least Wfty extra components in the labour force, which seems to presuppose a reserve of free labour, or a free population almost as large as the servile one. Now, although our Wgures for the valley as a whole seem to suggest a similar situation, if we look at the distribution of small farms and villages they are regularly far from the areas of the villas: in the heartland of the Cosan territory there are almost none at all. The harvest problem in terms of the territory of Cosa can, I think, be partially solved in other ways—perhaps, as Ulrike Roth has recently suggested,48 by female slaves from the villa itself, who would otherwise have been involved in textile production, or by slaves who worked in turn for diVerent proprietors who collaborated with each other for the harvest—this could be the sense of de Agricultura 4: opera facilius locabis, operaios facilius conduces. Finally, some labour may have come from farther away. In the Maremma before the land reform of the 1950s a large number of braccianti immigrated to the area for the harvest: in 1911, for example, the immigrant workforce amounted to 21% of the population. But these came from as far away as the Casentino.49 In our case, the upper valley may have served as a labour reserve. A free component of the population was certainly present, if we take the territory as a whole. Now, if we examine the distribution of this population in the valley during this period it is immediately obvious that they are not in any way evenly spread across the landscape.50 Settlement 46 Beloch 1886: 416–18. See also Brunt 1971: 121–30 who puts the Wgure at 3 million. Scheidel estimates that no more than 1/5 of the population of Italy lived north of the Rubicon (1996: 168), and that 40% of the Italian population were slaves. 47 Rathbone 1981; see also Lo Cascio 1999b: 222–1. 48 Roth 2007. 49 Bagianti 1983. 50 See the figure in Carandini et al. 2002: 187.

Peopling the Countryside

149

35000 Dispersed

30000

Nucleated

25000

Urban

20000 15000 10000 5000

550–

450–500 AD

350–

250–

150–

50–100 AD

50–1 BC

150–

250–

0

Fig. 5.11. Albegna Valley: the distribution of population over time

is denser around the towns, thinning out in a fairly regular way as one moves away from them. However, there is a large area which does not appear to have been settled at all, which we have interpreted elsewhere as the ager publicus of Cosa, not aVected by the colonization and used for extensive activities like pasture and wood production. While the Valle d’Oro and the lower Albegna Valley are occupied almost exclusively by villas, the upper valley, particularly the area around Saturnia, is largely occupied by farms. This suggests that there was more mixed farming in that area, and a greater survival of a free population. All of these points have been made before and at greater length;51 I am reiterating them here to stress that only projects on this sort of scale, with adequate sampling technique, can tell us about the regional distribution of population and sites, rather than giving us glimpses of particular cases. I now turn to the Xow data, or the change of population over time. Fig. 5.11 shows the change in population in the Albegna Valley, broken down into its various components. It is immediately obvious that they tend to support the contention of Lo Cascio and others that the population of the countryside was rising until somewhere in the 51 e.g. Attolini et al. 1991; Celuzza 2002; Fentress 2002a.

150

Elizabeth Fentress

early Empire, although the uncertainty of our Wgures for the Republican period leaves the exact nature of the rise in doubt—I suspect that when we have Wnished checking the pottery52 they will resemble far more closely those of the early Empire. If the rise between the third century and the second, which amounts to 10 per thousand seems to have been based to a large extent on in-migration—the importation of slaves and the second colonization of Cosa in 197 bc—the rise into the Augustan period could to some extent have depended on normal population growth—the Wgure of 1.5 per thousand compares with that of 1.7 per thousand suggested by Lo Cascio’s estimate of the population growth from the census data of 28 bc and ad 14. There is evidence in the large number of new villas for a large inXux of slaves from elsewhere in the early part of this period (SetteWnestre is an example), and we know of a certain degree of triumviral and Augustan colonization, so I would certainly include a substantial in-migration here.53 However, the population does not continue to grow: already the second century ad shows it is declining steeply by a rate around 4 per thousand since the Augustan period, a decline which steepens slightly into the third century to 5 per thousand. This contrasts sharply with Lo Cascio’s perception of a continued growth into that period, although he is of course generalizing for Italy as a whole.54 The steepening of the decline between the Wrst and the second to that between the second and the third centuries might be due to the eVects of the Antonine Plague, although, given the highly rural nature of the population this is uncertain, and we cannot pin the date closely enough. Nor, in the generally deserted Maremma of the sixth century, can we see any intensiWcation caused by the Justinianic plague. However, we might well ask whether another epidemiological factor was at work, malaria. Sallares has demonstrated convincingly that Malaria Falcipara was hyperendemic on the coast from classical times, which goes very badly with our relatively high populations in the late Republic and early Empire, 52 Figures are missing for the Republican amphorae in the central part of the territory: these are particularly important as they would date the sites on which black glaze is not present. 53 Cf. Fentress 2002a: 182. 54 Lo Cascio 1994.

Peopling the Countryside

151

and their distribution over low-lying areas.55 However, it should be noted that the fall-oV over time in the area of the coast, the ager Cosanus and that around Talamone, was far greater than that in the upper valley, where swampy and wet conditions would have been less prevalent. We have always attributed this coastal collapse to the greater intensity of the villa system there, and its economic weakness after the early empire, but it does seem from Sallares’s work that we cannot rule out environmental factors. Indeed, it is not impossible to interpret much of our evidence in this light: the high susceptibility of adults with no acquired immunity might have occasioned the almost immediate collapse of the third-century colony. Cosa, re-founded and repopulated in the beginning of the second century bc, was again deserted in the 70s—apparently due to a pirate attack, but it was not repopulated until Augustus’ time. In its hinterland the small farms of the second and early Wrst century bc disappeared rapidly as well.56 Repeated attempts to resettle it, particularly the Augustan colonization of Cosa, were, as I have shown elsewhere, failures.57 In order to accept malaria as the principal determining factor in the disappearance of the free population of the coast we would have, however, to explain the relative immunity of the slave-run villas, although Sallares suggests that their high mortality would have occasioned only Wnancial worries for their proprietors as they were replaced by new slaves: further, the ever-larger estates would have been farmed extensively rather than with intensive wine production on the SetteWnestre model.58 The diminution in the rate of decline from the third century onward would suggest that the population, perhaps slaves bred in the villas, had acquired immunity to some degree. I propose to test the hypothesis that malaria inXuenced the demographic Xow by contrasting the low-lying areas with higher ground, but have not yet been able to do this. However, even the rather cruder Wgures published for the Maremman coast at Caere and Vulci show a similar pattern of rapid decline (Figs. 5.12 and 5.13), and there is every likelihood that it can be generalized for the whole of the Maremma.59 55 Sallares 1999; 2002. Pliny refers to the gravis et pestilens ora Tuscorum, quae per litus extenditur (Ep. 5.6.2). 56 For the greater susceptibility of children and immigrants see Sallares 1999: 167. 57 Fentress 2004: 138–41. 58 Sallares 2002: 254–5. 59 Enei 2001; Corsi 1998.

152

Elizabeth Fentress

1800 1600 1400 1200 1000 800 600 400 200 0

Farms

0 55

0 45

0 35

0 25

0 15

50

0 −5

50 −1

−2

50

Villas

Fig. 5.12. Vulci/Tarquinia coast (Augustan high 31/km2)

The population densities in those areas, however, are somewhat higher than those of the Albegna Valley, increasing with proximity to Rome. The population of the Albegna Valley is overwhelmingly rural in all periods except the third century bc. Even in the early empire a rather high estimate of the urban population gives only 14% of the total population, while in the subsequent period, in which Cosa has already been abandoned and we have little evidence for settlement in

3500 3000

Farms

2500

Villas

2000 1500 1000 500

Fig. 5.13. Caere (Augustan high 34/km2)

0 55

0 45

0 35

0 25

0 15

50

−5 0

50 −1

−2

50

0

Peopling the Countryside

153

35000 30000 25000 20000 15000 10000 5000 0 −300 −200 −100

1 Farms

100

200

300

Nucleated

400

500

600

Urban

Fig. 5.14. Jerba: distribution of population

either Heba or Saturnia, the percentage is likely to have been far smaller. By the end of the third century I would suggest that the area was entirely de-urbanized. This low urban population suggests that the valley continued its vocation as a net agricultural exporter: its continuing linkage to the networks of trade and distribution of the outside world and, in particular, with Rome is, of course, suggested by the quantity of African Red Slip found in all areas of the valley up until the sixth century. The quantity per site, as we have noted elsewhere, is substantially higher than that of Jerba.60 If we return to Jerba the picture is very diVerent, although anomalous in terms of the African surveys which tend to show a far more dynamic settlement pattern well into the late Empire. It is, as we have seen, much more densely populated than the Etruscan coast, in spite of its apparent inability to produce enough food to feed itself: clearly the high industrial productivity of Meninx created an economic spin-oV in the form of textile production, Wsheries, and transhumant stock raising that supported a growing population up through the Wrst century bc (Fig. 5.14). The population stabilizes in the early Roman period and remains high until 400, when it begins to drop oV steeply. This is rather diVerent from the pattern shown by other surveys, which show the number of sites rising into the fourth or even Wfth centuries. The urban pattern is of course far more constant 60 Fentress et al. 2004.

154

Elizabeth Fentress

Fig. 5.15. ARS at Meninx and B urg u

and important than it is in the Albegna Valley. Meninx, by the third century, concentrates over 50% of the population of the island. We might expect in this highly urbanized environment that the eVects of the Antonine and Justinianic plagues would be visible here. However, the steep fall in the presence of well-known Red Slip forms at the town of B urg u seems to date to the beginning of the third century. Thereafter the town is reduced to the size of a village, and the remaining population seems to have concentrated at Meninx, where, however, we can see no major change in this phase (Figs. 5.15 and 5.16). Its demise, however, is equally abrupt, coinciding with the middle of the

Peopling the Countryside

155

% Fineware / year

3

2

1

0 −30 20

70 120 170 220 270 320 370 420 470 520 570 620 670 Years Meninx

AD

- Burgu

Fig. 5.16. B urg u and Meninx percentage Wneware/year

sixth century. Its status as a major port city would have made it especially vulnerable to the Justinianic plague, probably the Wrst instance of the Bubonic plague, carried by black rats from port to port.61 However, the relatively steady population throughout the empire does seem to rule out any endemic eVects of malaria, which has never to my knowledge been recorded on the island—a complete lack of surface water will have protected its population. Finally, it is worth considering the fragility of the dispersed population in both surveys. As a rule of thumb, the villas and the villages survive, and the small farms melt away, although at diVerent rates at diVerent times. In both cases the villas predominate over villages, although, of course, we do not know the exact status of sites which began as villas in the Wrst century bc by the time they have been 61 For the eVects of this plague McCormick 2001: 109–11; Little 2006.

156

Elizabeth Fentress

around for five hundred years. My excavations of a villa near Marsala in Sicily showed that it burned under Hadrian, but continued to be occupied by small structures well into the sixth century.62 But the fact remains that someone owned these sites, and the people who lived there can probably not be counted as free peasants, although whether they were actually slaves depends on a series of incognita. The fragility of the small farms, however, remains obvious. Now, while the Albegna Valley in the middle and late Empire is notable for its total inertia as far as new sites are concerned, Jerba is far less so. There seems to be evidence for continuous investment in the countryside: sixteen of the farms were new in the late empire, while fully seven new villages emerged, and the village at B urg u seems to have consolidated its settlement. The new villages were larger than those of previous periods. There is, in fact, an interesting relationship between the villages and the villas. The earliest villages were few but substantial. They declined in size to a low in the early Roman period, when the largest villas were constructed. From that point on their average size increased. The villas follow a diVerent pattern, with the large sites disappearing more quickly than the smaller ones (Fig. 5.17). While new villas ceased to be built, the villages remained active. This shift in emphasis might suggest that individual landlords had become much less important in the organization of production, and the fundi less of a focus of agricultural settlement.63 Most of the original villas continued in occupation, but there is no sign of investment in their buildings in the later period, and they may have come to resemble the villages. This has necessarily been a very rapid look at the light these two surveys shed on the population of the empire and its distribution. Many of the conclusions, particularly as to their total population, are still softer than one would wish. But they are, for all their uncertainties, hard to equal in terms of sheer quantity of data: we eagerly await Tuscania and the South Etruria Survey for comparison. And because of the similarity of the data collection we seem safe in making comparative statements about them: there is no doubt, for instance, that Jerba is far more ‘developed’ than the Albegna Valley, in spite of its 62 Fentress 1999. 63 On the shift from villas to villages in the late Empire, see Wickham 2005: 473–88.

Peopling the Countryside

157

3.5 3

Hectares

2.5 2 1.5 1 0.5

0 EP

P

ER

MR

LA1

LA2

Period

Villas

Villages

Fig. 5.17. Jerba, villas and villages between the Early Punic and Late Antique periods

greater distance from Rome. Its intensive, almost industrialized economy is in no way peripheral. The two very diVerent trajectories of these regions have in common only the almost total lack of evidence for the seventh and eighth centuries. The comparison between the two shows, yet again, the importance, not only of studying regional samples, but of comparing them. Sallares stresses that ‘all valid studies must commence at the local level’,64 to which one could add that only by the comparison of a number of local situations can we understand any individual one. The software for the publication of this sort of data on the web in GIS format is all there and I hope to have shown how we are taking advantage of this for Jerba. Albegna languishes for want of Wnancial 64 Sallares 2002: 118.

158

Elizabeth Fentress

resources, however. For future work on demography from this perspective, it would be immensely valuable to assemble the data from this, Tuscania, and other projects in standard format and group them on one site. In this way we might signiWcantly move the argument beyond individual case studies, and allow a larger number of scholars and datasets to contribute to the debate. Bibliography Attolini, I., et al. (1991). ‘Political geography and productive geography between the valleys of the Albegna and the Fiora in southern Etruria’, in G. Barker and J. Lloyd (eds.), Roman Landscapes: Archaeological Survey in the Mediterranean Region. London, 142–52. Bagianti, I. (1983). ‘Migrazione dalle Montagne Toscane alla Maremma nel ‘‘900’’ ’, in Campagne maremmane fra 800 e 900. Florence. Barker, G., and Rasmussen, T. (1988). ‘The archaeology of an Etruscan polis: A preliminary report on the Tuscania Project (1986 and 1987 seasons)’, PBSR 56: 25–42. —— —— (1991). ‘The Tuscania survey’, in G. Barker and J. Lloyd (eds.), Roman Landscapes: Archaeological Survey in the Mediterranean Region. London, 104–14. Beloch, K. J. (1886). Die Bevo¨lkerung der griechisch-ro¨mischen Welt. Leipzig. Bintliff, J. L. (1999). ‘Further considerations on the population of Ancient Boeotia’, in J. L. BintliV (ed.), Recent Developments in the History and Archaeology of Central Greece. Proceedings of the 6th International Boeotian Conference (BAR Int. Series 666). Oxford, 231–52. Brunt, P. A. (1971). Italian Manpower 225 b.c.–a.d. 14. Oxford. Cambi, F. (1999). ‘Demography and Romanization in central Italy’, in J. L. BintliV and K. Sbonias (eds.), Reconstructing Past Population Trends in Mediterranean Europe (3000 b.c.–a.d. 1800). The Archaeology of Mediterranean Landscapes 1. Oxford, 115–28. Carandini, A., Cambi, F., Celuzza, M.-G., and Fentress, E. (eds. 2002). Paesaggi d’Etruria: Valle dell’Albegna, Valle d’Oro, Valle del Chiarone, Valle del Tafone. Progetto di ricerca italo-britannico seguito allo scavo di SetteWnestre. Rome. Celuzza, M.-G. (2002). ‘La romanizzazione’, in A. Carandini, F. Cambi, M.-G. Celuzza, and E. Fentress (eds.), Paesaggi d’Etruria: Valle dell’Albegna, Valle d’Oro, Valle del Chiarone, Valle del Tafone. Progetto di ricerca italo-britannico seguito allo scavo di SetteWnestre. Rome, 103–12. Chamberlain, A., Barker, G., and Slater, E. (2006). Demography in Archaeology. Cambridge.

Peopling the Countryside

159

Ciampoltrini, G. (1995). ‘Un pocolom e le mura di Orbetello’, AC 47: 289–302. Corsi, C. (1998). ‘Atlante Fondario Romano. L’insediamento rurale di eta` romana e tardoantica del territorio tra Tarquinia e Vulci’, Journal of Ancient Topography 8: 223–55. Di Giuseppe, H., Sansoni, M., Williams, J., and Witcher, R., et al. (2002). ‘The Sabinensis Ager revisited: A Weld survey in the Sabina Tiberina’, PBSR 70: 99–150. Duncan-Jones, R. P. (198263). The Economy of the Roman Empire: Quantitative Studies. Oxford. Enei, F. (2001). Progetto Ager Caeretanus. Il litorale di Alsium. Ricognizioni archeologiche nel territorio dei comuni di Ladispoli, Cerveteri e Fiumicino. Ladispoli. Fentress, E. (1999). ‘The house of the (Sicilian) Greeks: Excavations at Contrada Mirabile’, in A. Frazer (ed.), The Roman Villa: Villa urbana. First Williams Symposium on Classical Architecture held at the University of Pennsylvania, Philadelphia, April 21–22, 1990. Philadelphia. —— (2000). ‘What are we counting for?’, in R. Francovich and H. Patterson (eds.), Extracting Meaning from Ploughsoil Assemblages. The Archaeology of Mediterranean Landscapes 5. Oxford, 44–52. —— (2001). ‘Villas, wine and kilns: The landscape of late Hellenistic Jerba’, JRA 14: 249–68. —— (2002a). ‘La tarda eta` repubblicana e la prima eta` imperiale’, in A. Carandini, F. Cambi, M.-G. Celuzza, and E. Fentress (eds.), Paesaggi d’Etruria: Valle dell’Albegna, Valle d’Oro, Valle del Chiarone, Valle del Tafone. Progetto di ricerca italo-britannico seguito allo scavo di SetteWnestre. Rome, 181–95. —— (2002b). ‘La campionatura: strategia del progetto’, in A. Carandini, F. Cambi, M.-G. Celuzza, and E. Fentress (eds.), Paesaggi d’Etruria: Valle dell’Albegna, Valle d’Oro, Valle del Chiarone, Valle del Tafone. Progetto di ricerca italo-britannico seguito allo scavo di SetteWnestre. Rome, 39–43. —— (2004). Cosa V: An Intermittent Town: Excavations 1990–1997 (MAAR Suppl.). Ann Arbor. —— Drine, A., and Holod, R. (eds. 2009). An Island through Time: Jerba Studies, vol. I: The Punic and Roman Periods (JRA Suppl. 70). Portsmouth, RI. —— Fontana, S., Hitchner, R. B., and Perkins, P. (2004). ‘African Red Slip and its market. Fineware and sites in Sicily and Africa’, in S. Alcock and J. Cherry (eds.), Side by Side Survey. Oxford, 147–62. —— and Jacques, F. (2002). ‘Saturnia: la centuriazione’, in A. Carandini, F. Cambi, M.-G. Celuzza, and E. Fentress (eds.), Paesaggi d’Etruria: Valle

160

Elizabeth Fentress

dell’Albegna, Valle d’Oro, Valle del Chiarone, Valle del Tafone. Progetto di ricerca italo-britannico seguito allo scavo di SetteWnestre. Rome, 124–6. Flannery, K. (1976). The Early Mesoamerican Village. New York. Hassan, F. (1981). Demographic Archaeology. New York. Hobsbawm, E. (1979). ‘An Historian’s Comments’, in B. Burnham and J. Kingsbury (eds.), Space, Hierarchy and Society: Interdisciplinary Studies in Social Area Analysis (BAR Int. Series 59). Oxford, 247–52. Hopkins, K. (1978). Conquerors and Slaves. Cambridge. Jacques, F. (1993). ‘CIL XI 2650 et le population de Saturnia’, ZPE 99: 217–19. Jongman, W. (1988). The Economy and Society of Pompeii. Amsterdam. Little, L. (ed. 2006). Plague and the End of Antiquity: The Pandemic of 541–750. Cambridge. Lo Cascio, E. (1994). ‘The size of the Roman population: Beloch and the meaning of the Augustan census Wgures’, JRS 84: 23–40. —— (1999a). ‘The population of Roman Italy in town and country’, in J. L. BintliV and K. Sbonias (eds.), Reconstructing Past Population Trends in Mediterranean Europe (3000 b.c.–a.d. 1800). The Archaeology of Mediterranean Landscapes 1. Oxford, 161–71. —— (1999b). ‘Popolazione e risorse agricole nell’Italia del II secolo a.C.’, in D. Vera (ed.), DemograWa, Sistemi Agrari, Regimi Alimentari nel Mondo Antico. Bari, 217–45. McCormick, M. (2001). The Origins of the European Economy: Communications and Commerce ad 300–800. Cambridge. Marzano, A. (2007). Roman Villas in Central Italy: A Social and Economic History. Leiden and Boston. Perkins, P. (1999). Settlement, Society and Material Culture in Central Coastal Etruria (BAR Int. Series 788). Oxford. Rathbone, D. W. (1981). ‘The development of agriculture in the ager Cosanus during the Roman Republic’, JRS 71: 10–23. —— (1997). ‘Surface survey and the settlement history of the ancient Fayum’, in Archeologia e papiri nel Fayyum: Storia della ricerca, problemi e prospettive. Atti del Convegno Internazionale, Siracusa 1996: 7–20. Roth, U. (2007). Thinking Tools: Agricultural Slavery between Evidence and Models. London. Saggin, A. (2003). ‘Orbetello: i ritrovamenti archeologici recenti’, in Annali della fondazione per il museo Claudio Faina. Rome, 355–62. Sallares, R. (1991). The Ecology of the Ancient Greek World. London. —— (1999). ‘Malattie e demograWa nel Lazio e in Toscana nell’Antichita`’, in D. Vera (ed.), DemograWa, Sistemi Agrari, Regimi Alimentari nel Mondo Antico. Bari, 131–88. —— (2002). Malaria and Rome. Oxford.

Peopling the Countryside

161

Sbonias, K. (1999). ‘Introduction to issues in demography and survey’, in J. L. BintliV and K. Sbonias (eds.), Reconstructing Past Population Trends in Mediterranean Europe (3000 b.c.–a.d. 1800). The Archaeology of Mediterranean Landscapes 1. Oxford, 1–20. Scheidel, W. (1996). Measuring Sex, Age and Death in the Roman Empire: Explorations in Ancient Demography (JRA Suppl. 21). Portsmouth, RI. —— (1997). ‘Quantifying the sources of slaves in the early Roman empire’, JRS 87: 156–69. —— (2001). Debating Roman Demography (Mnemosyne Suppl. 211). Leiden. Terrenato, N., and Ammerman, A. (1996). ‘Visibility and site recovery in the Cecina Valley Survey, Italy’, Journal of Field Archaeology 23: 91–109. Trement, F. (1999). ‘Prospection arche´ologique et de´mographie en France’, in J. L. BintliV and K. Sbonias (eds.), Reconstructing Past Population Trends in Mediterranean Europe (3000 b.c.–a.d. 1800). The Archaeology of Mediterranean Landscapes 1. Oxford, 93–114. van Dommelen, P., and Sharpe, L. (2004). ‘Surveying Punic rural settlement: The Terralba Project’, Antiquity 78, no. 299 (Mar.), http://antiquity. ac.uk/ProjGall/vandommelen/index.html (accessed Oct. 2007). Wallace-Hadrill, A. (1995). Houses and Society in Pompeii and Herculanium. Princeton. Ward-Perkins, J. B. (1955). ‘Notes on Southern Etruria and the Ager Veientanus’, PBSR 23: 44–72. Wickham, C. J. (2005). Framing the Early Middle Ages: Europe and the Mediterranean, 400–800. Oxford. Wilkinson, T. (1999). ‘Demographic trends from archaeological survey’, in J. L. BintliV and K. Sbonias (eds.), Reconstructing Past Population Trends in Mediterranean Europe (3000 b.c.–a.d. 1800). The Archaeology of Mediterranean Landscapes 1. Oxford, 45–64. Wilson, A. I. (2004). ‘Tuscan landscapes: Surveying the Albegna valley’, JRA 17: 569–76. Witcher, R. (2005). ‘The extended metropolis: Urbs, suburbium and population’, JRA 18: 120–38. Zuckerman, L. (1999). The Potato. New York.

This page intentionally left blank

6 Peopling Ancient Landscapes: Potential and Problems David Mattingly

The chapter by Lisa Fentress is a characteristically elegant and robust contribution to the debate about how best to mobilize Weld survey results in the context of broader debates about settlement trends and ancient demography.1 There is much to be gleaned from this richly detailed and positivistic account and I fear that in the space available to me I cannot address fully all the issues that she raises.2 The explosion of Weld survey projects in the late twentieth century has contributed to a general sense that rural settlement expanded considerably in many areas of the Roman empire, but has not yet had much impact on overall demographic models and estimates for Roman population of provinces or the empire as a whole. For this reason alone, I applaud the attempt by Fentress (along with recent papers by Rob Witcher) to generate sophisticated argument about ancient demography from raw survey data.3 In laying down some caveats about the methodology and conclusions that Fentress has developed, I am (with all the beneWt of hindsight) reXecting on whether I would have approached the survey of the Albegna Valley or Jerba in the same way.4 1 This theme of the OXREP reprises one of the main themes of the POPULUS project, see BintliV and Sbonias 1999; Francovich et al. 2000; Gillings et al. 1999; Leveau et al. 1999; Pasquinucci and Tre´ment 2000. 2 Furthermore, any comments on the Jerba data must be preliminary until the full publication of the survey. 3 See also Witcher 2005; 2006a. 4 Previous publications will make clear some of the philosophical and methodological diVerences between us (see especially Fentress 2000; Mattingly 2000), though

164

David Mattingly

Although there is an understandable desire to relate numbers of sites recorded by Weld surveys to ancient population numbers, as Fentress acknowledges it is not straightforward to achieve with the typically fuzzy and lacunose data provided by Weld survey. There are other issues that these data may be just as successfully applied to5 and it is worth pondering whether, in emphasizing the demographic aspects, we are striving for an unachievable level of numerical precision or imposing an unrealistic weight of expectation on our results. There are many issues raised explicitly or implicitly by this chapter: . How to maximize the interpretational power of new archaeological data on regional settlement patterns? . Can ‘sites’ be translated into ‘people’ in a straightforward way? Do periods of increasing numbers of sites correlate with growth in population or economic growth? Do periods when site numbers decline automatically indicate population shrinkage? . Urban and larger nucleated sites can comprise a surprisingly large percentage of the overall population (though the two surveys considered here show opposite characteristics in this regard). . What are the advantages of random and standardized sampling procedures in Weld survey, such as were used in the Albegna Valley and Jerba surveys? . How successfully can we draw out inter-regional similarities and diVerences—as here between Albegna and Jerba? . What is the true relationship between the sites that are recorded in Weld survey and those that once existed in the landscape? . Is survey better at tracking trends over time (Fentress’s Xow data) than at providing snapshot population Wgures? I cannot hope to cover all of these, and intend to concentrate on areas where my own views diVer somewhat. It needs stressing at the outset that there is much in this chapter with which I concur and which enshrines best practice. I would argue that both of us have produced valid and signiWcant survey data through our diVerent approaches. 5 Witcher 2006b, for a perceptive discussion of what we should be doing with survey data.

Peopling Ancient Landscapes

165

RU RAL SET TLEMENT NUMBERS It is a truism that the rural landscape is now seen to be a great deal busier than was the case in earlier scholarly generations, prior to the upsurge in landscape archaeology. The scale of incremental knowledge gain can be huge. By way of example, consider the Roman period rural settlements known in England in 1955 when Hoskins wrote his pioneering study The Making of the English Landscape.6 He estimated that there were c.100 towns and small towns and only about 2,500 rural sites of all types. This led him to conclude that much of Britain was still forest or uncultivated wasteland and that the total population was under 1 million. Using 1914 Wgures for land under crops or pasturage for England and Wales of 27 million acres (c.11 million ha), he estimated that only 2–3% of this was already under cultivation in Roman times. Fifty years on, the picture is dramatically diVerent. Roman period settlements in England in 2005 are estimated at c.150 towns, small towns, and villages with at least 100,000 rural settlements.7 Figures for small towns and villas have grown far less than those of lesser rural sites, but the overall increase is nothing short of spectacular. With the new data, and despite a tendency to reduce estimates of urban populations, the overall population Wgures for Britain under Rome have tended to rise.8

UNKNOWN UNKNOW NS Reports that say that something hasn’t happened are always interesting to me, because as we know, there are known knowns; there are things we know we know. We also know there are known unknowns; that is to say we know there are some things

6 Hoskins 1955: 33–44. 7 Taylor 2007; Mattingly and Witcher 2004: 175–6. 8 Millett 1990: 181–6.

166

David Mattingly we do not know. But there are also unknown unknowns—the ones we don’t know we don’t know.9

In the context of the disastrous US policy in Iraq, Rumsfeld’s ‘unknown unknowns’ might be translated as ‘the unpleasant truths that the White House was in denial about from the outset’ or perhaps ‘unacknowledged unknowns’. How worried should we be about Rumsfeldian ‘unknown unknowns’ or blindspots in Weld survey? To translate Rumsfeld’s categories into archaeological ones, we have: 1. Known knowns—the artefacts and sites that a survey collects. Most archaeologists inevitably focus on these in building their accounts. 2. Known unknowns—the fact that some sites are known to have been destroyed or were inaccessible and the consequences on recovery rates of decisions taken about collection method, survey intensity, and sampling, etc. Again, many surveys make a claim to be able to compensate for deWciencies in visibility, logistical resources, etc. 3. Unknown unknowns—the exact relationship between sites recovered and sites that once existed, or between pottery sherds and people. These problems are not really ‘unknown’ of course, but they tend to represent things that we would prefer to ignore or downplay. This is potentially a big problem and limits our ability to form a clear sense of what proportion of the evidence has been destroyed or is otherwise not visible. Recent very large-scale and expensively funded French work in the Rhoˆne Valley accompanying the TGV line has transformed understanding of the degree to which sites have been buried beneath alluviation in that most active of river valleys. Sites recorded by traditional surface survey represented c.5% of total ancient sites recorded through regular excavations along the line of the railway (though a higher percentage for the Roman period). In relation to the case studies presented by Fentress, geomorphological considerations are probably more of an issue for the Albegna estimates than for Jerba. 9 Donald Rumsfeld, the former US Secretary of Defense famously coined the expression ‘unknown unknowns’ at the Defense Department BrieWng (12 Feb. 2002): http://en.wikipedia.org/wiki/Unknown_unknown.

Peopling Ancient Landscapes

167

Our ability to use archaeological survey data to contribute meaningfully to the high/low count debate is potentially compromised by such unknown unknowns. Both high and low counts could in theory be supported by interpreting the survey data according to diVerent sets of assumptions. In the Wnal analysis, the survey data are probably better suited for generating Xow data on the broad upward/ downward trends over time. For the same sorts of reasons, it is extremely hazardous to tie the archaeological longue dure´e to speciWc historical events or demographic crises occasioned by malaria and plague. Fentress is at her best when describing patterns, rather than in the wilder Xights of interpretation she applies to those patterns.

R A N DO M S U RV EYS A N D SI T E M O D E L L I N G More than twenty years ago, Colin Haselgrove illustrated the key problem of levels of inference in Weld survey.10 The issue is that the sample of data recovered by Weldwalking is separated from the original site universe by multiple depositional and post-depositional events, by sampling and methodological biases, and by several levels of inference commonly used to Wll in the gaps. We are still none the wiser about the numerical relationship between what we Wnd and what once existed. The case has been convincingly made by Fentress and many others for the necessity of random sampling methods if sites discovered in a given sample area are to be modelled out to a wider territorial space.11 However, while I agree with this starting position, at the 2006 OXREP conference I commented on some potential shortcomings of the systematic sample sets collected by the two projects and the way Fentress uses them. Random sampling in itself is not the complete answer and there are other decisions relating to the design of the sample and the Weld methodology that can materially aVect the reliability of the results. 10 Haselgrove 1985.

11 Binford 1964; Fentress 2000; Plog 1976.

168

David Mattingly

Problem 1. Numbers and how they are generated The fact that three separate attempts to model the Albegna Valley population data by three diVerent members of the original team have yielded signiWcantly diVerent results should be a cause for greater concern (and analysis). There is also a methodological issue as I see it in the way the Jerba survey sample has been used to generate the overall site universe and population Wgures.12 I cannot correlate the population estimates with the random sample transect data.

Problem 2. Random sampling Both the Albegna and Jerba surveys used systematic patterns of transects to create a ‘random’ sample of the landscape. They represent one of the most rigorous attempts to impose such sampling procedures in Mediterranean Weld survey.13 However, although the generation of regular linear transects across the island ought to produce a representative sample of the overall site distribution and thus allow the settlement pattern to be reconstructed with conWdence, this depends on a number of assumptions.14 One of the ways to improve the sophistication of such random samples would be to make them proportional to diVerent land blocks in the surveyed area. The Wrst assumption of the Jerba survey is that all parts of the island were broadly the same and all equally attractive to human settlement. As a consequence the random sample transects were all laid out as narrow north–south corridors across the width of the island. This provided a 13.4% sample of the landmass, but is not in 12 Fentress properly states (p. 133) that she will only use the site data from the systematic transects (covering c.13.4% of Jerba’s surface area) as a basis for generating Wgures for the island as a whole—thus ignoring the additional judgemental surveyed areas, especially around Meninx and the south-east coast of the island. However, comparison between the Jerba site/population tables and Fig. 5.3 suggests that this process was not followed in practice—the Wgures for small farms, villas, and villages contain all sites known on the island and thus the multiplication of these numbers by 7.47 ‘to restore the sample to 100%’ is surely invalid. 13 See Given and Knapp 2003 for another. 14 There is also the issue of whether the slight tinkering with oVset placement of some transect lines (for instance to take them across Meninx and B urg u) undermines the claim that this is a truly random sample!

Peopling Ancient Landscapes

169

fact an equal sample of all the diVerent categories of landform on the island and most signiWcantly the coastal/shoreline zone appears to be under-represented—especially as the north or south ends of several transects run into areas of modern development. I remain unconvinced that the project achieved 13.4% coverage of the coastlands of the island, yet the additional judgemental/purposive sample area in the south-east of the island suggests that the coastal environs may have been particularly heavily occupied—a point borne out by the fact that of all Roman sites known on the island, 9/11 villages and 3/4 known urban centres and 15/23 villas are located within 2 km of the shoreline. Now, as it happens, the random transect sample of the coastline area is somewhat unrepresentative of the density of sites in this sector of the landscape.15 If we blank out all sites on the map of Jerba apart from those within the random north–south transect lines, how easy is it to model up from just these data to produce an accurate facsimile of the pattern we know more about from the purposive sample? The problem here is not that the principle of random survey is wrong, but rather that there are some potential distortions produced in the dataset by the use of long narrow transects.16 My own preference would have been to Weldwalk within blocks of say 6  6 km that were randomly located in proportion to the areas of discrete units of the landscape (coastal plain, interior farmland, salt Xats), with additional areas being allocated where necessary to make up for areas obliterated by modern development.

Problem 3. The recoverability of rare upper-echelon sites in random samples Because there were relatively (surprisingly) few small rural sites recorded in the interior of the island, the demographic signiWcance of the upper-echelon settlements on the island is thus fairly high, 15 One factor is that in addition to several north and south extremities of transects running into areas of modern development (and thus zero archaeological visibility), the transect that runs down the west coast of the island crosses a large salt Xat that is unrepresentative of other parts of the coastal plain. 16 There are issues with edge eVects of long linear transects, which need to be taken into account, see Robinson and Zubrow 1999.

170

David Mattingly

especially the towns and villages. Fentress has responded to a point I raised in my verbal response (see her n. 5) that a systematic sampling strategy will always be better at recording the most abundant categories of site, than the larger and rarer upper-echelon sites. She rightly dismisses the possibility of the survey having missed the equivalent of a Teotihuacan on Jerba, but that was not the implication of my original comment.17 Sites that belong to rare types are quite likely to be under-represented within a random survey, and for this reason their overall numbers are diYcult to model from a 10 or 20% sample. For a number of reasons the estimated numbers of towns and villages on Jerba may be inaccurate. Having worked on the antiquarian record for Jerba in compiling the relevant sheet of the Barrington Atlas,18 I am aware how poor knowledge of the island’s archaeology was prior to the Jerba survey. The large-scale hotel development in the north-east of the island and the expansion of the modern towns of Humit al-Suq and Ajı¯m have eVectively destroyed without record a large swathe of the coastal zone. There is a still unresolved debate about how many towns existed on the island in antiquity, with the possibility of an additional centre beyond the four recorded by Fentress. Humit al-Suq or Ajı¯m are the most likely locations for a missing urban site (which could add an additional 10% to the population Wgures on its own). The existence of harbours at both locations constitutes a strong argument in favour of them at the very least having been ancient villages. Furthermore, as villages clustered on or close to the coast—perhaps reXecting the dual exploitation of the sea and the land19—there are implications for the overall modelling of population if there has been an undercount of their real numbers consequent on the placement of the random 17 Actually the discovery of the large and previously unrecognized Etruscan town of Doganella by the Albegna Valley survey is a salutary lesson in the potential for random survey to miss even exceptional and very large sites (Perkins and Walker 1990; Walker 1985). The site was discovered by chance when a 1 km wide transect of the rural survey happened to run across the area of the Etruscan town (whose size exceeded 200 ha)—it could easily have been missed. The survey corridor was extended a further km in width to encompass all the ancient city. 18 Talbert 2000, map 35 and gazetteer. 19 The assumptions about the carrying capacity of the island appear to take no account of the contribution of Wsh protein, both to diet and as an exchange commodity for importing cereals.

Peopling Ancient Landscapes

171

survey units (with modern coastal development having destroyed an important proportion of this landscape zone). My point is that we cannot be conWdent that the survey located an accurate 13.4% of urban and village sites and the consequences of any error in these categories will have a disproportionate impact on the demographic models produced.

Problem 4. The possibility of missing farms Site continuity and visibility is somewhat unacknowledged in this account. In general surface visibility is good within the typical olive groves of the island, though there were clearly scatters of Wnds encountered that hinted at the presence of sites, without fully convincing the Weld teams. As the island lacks natural springs, settlement has relied heavily on building and maintaining cisterns for storing water and we might expect a high degree of continuity of occupation or reoccupation of locations where such catchments existed. In consequence, I suspect that there is a serious visibility problem in the centre of the island—focused on the locations of farms still in occupation. This is where the collection of oV-site data in large-scale surveys is important, because they may give clues as to the numbers and location of ‘lost’ or degraded sites, such as those that lie beneath modern farms. Haloes of Roman sherds around such locations could point to the existence of Roman farms, as Fentress acknowledges, even where structural evidence and artefact densities are below the ‘site threshold’ imposed by the survey. Fentress may well be right that the overall numbers of small farms were less signiWcant on Jerba to the total population than the towns, villages, and villas, but the extent of shortfall is an ‘unknown unknown’. If only half or a quarter of the Roman farms in the transects have been securely identiWed, the population deWcit could be as high as 10–20% of the overall total.

Problem 5. Site interpretation in the ploughzone There is a potential diYculty in attaching interpretative labels to the sites found in ploughzone Weld survey, where the surface representation

172

David Mattingly

may be little more than heavily abraded sherds and sparse building materials. Criteria can be set, of course, to discriminate between farm and villa, villa and village, village and town (or indeed between diVerent sizes of farm).20 But any serious Weld survey practitioner will know that things are not always clear-cut and that our categorizations are often ‘best Wt’ guesses, selected from a relatively narrow range of options. Geophysical survey and/or excavation can occasionally shed greater light, but can only conWrm interpretation for the individual site, not for an entire group of sites. In many areas of the ancient world there are so few well-excavated rural sites that the sort of round numbers for population levels at diVerent categories of site remain very crude instruments (Fentress’s minimum and maximum Wgures).

Problem 6. Evidence of absence and absence of evidence The dating of phases of human activity is not straightforward—even within the Roman period, as Martin Millett has ably demonstrated.21 Some regions were at certain times much more closely integrated with Mediterranean trade, with the result that there are periods with a larger abundance of diagnostic Wnds. By contrast, a preceding or subsequent phase with less diagnostic material may appear one of drastically reduced activity. While the association between demographic downturn and these less dynamic economic phases is plausible, some caution needs to be exercised in giving this numerical precision—as the Xow data calculations appear to.

CONCLUSIONS This is an important study that takes numerical analysis of rural demography to new levels. But I am not convinced that it is the last word on the high/low population question or even on the Albegna and Jerba surveys. The ‘unknown unknowns’ are suYciently large for 20 And multi-period sites often grow or shrink over time, making estimations of status based on surface spreads particularly tricky. 21 Millett 1991; 2000.

Peopling Ancient Landscapes

173

the debate to remain open. I suspect our main point of disagreement relates to the relative completeness of survey site discovery rates— Fentress seems to assume that the vast majority of sites within the transects will have been located. I am more sceptical and suspect that what is recorded represents a smaller percentage of the original site universe, due to long-term processes of destruction, burial and erosion, ploughing damage, modern construction as well as methodological and sampling issues of modern Weldwork. What we can all agree upon, however, is that the full publication of survey data will allow this sort of quantiWcation exercise to be played out for a range of Weld surveys. Fentress is to be thanked for having the courage to oVer her own models for criticism. She will not be surprised that, in the interests of promoting further debate, I have not accepted unequivocally all her calculations. Bibliography Alcock, S. E., and Cherry, J. F. (eds. 2004). Side-by-Side Survey: Comparative Regional Studies in the Mediterranean World. Oxford. Barker, G. W., and Lloyd, J. (eds. 1991). Roman Landscapes. Archaeological Survey in the Mediterranean Region. London. Binford, L. (1964). ‘A consideration of archaeological research design’, American Antiquity 29: 425–41. Bintliff, J., and Sbonias, K. (eds. 1999). Reconstructing Past Population Trends in Mediterranean Europe (3000 b.c.–a.d. 1800). The Archaeology of Mediterranean Landscapes 1. Oxford. Fentress, E. (2000). ‘What are we counting for?’, in Francovich et al. 2000: 44–52. Flannery, K. V. (ed. 1976). The Early Mesoamerican Village. New York. Francovich, R., Patterson, H., and Barker, G. (eds. 2000). Extracting Meaning from Ploughsoil Assemblages. The Archaeology of Mediterranean Landscapes 5. Oxford. Gillings, M., Mattingly, D. J., and van Dalen, J. (eds. 1999). Geographical Information Systems and Landscape Archaeology. The Archaeology of Mediterranean Landscapes 3. Oxford. Given, M., and Knapp, A. B. (2003). The Sydney Cyprus Project. Social Approaches to Regional Archaeological Survey. Los Angeles. Haselgrove, C. (1985). ‘Inference from ploughsoil artefact samples’, in C. Haselgrove et al. 1985: 7–29.

174

David Mattingly

Haselgrove, C., Millet, M., and Smith, A. (eds. 1985). Archaeology from the Ploughsoil. Studies in the Collection and Interpretation of Field Survey Data. SheYeld. Hoskins, W. G. (1955). The Making of the English Landscape. London. Leveau, P., Trement, F., Walsh, K., and Barker, G. (eds. 1999). Environmental Reconstruction in Mediterranean Landscape Archaeology. The Archaeology of Mediterranean Landscapes 2. Oxford. Mattingly, D. J. (2000). ‘Methods of collection, recording and quantiWcation’, in Francovich et al. 2000: 5–15. —— and Witcher, R. (2004). ‘Mapping the Roman world: The contribution of Weld survey data’, in Alcock and Cherry 2004: 173–86. Millett, M. (1990). The Romanization of Britain. Cambridge. —— (1991). ‘Pottery: Population or supply pattern’, in Barker and Lloyd 1991: 18–26. —— (2000). ‘Dating, quantifying and utilizing pottery assemblages from surface survey’, in Francovich et al. 2000: 53–9. Pasquinucci, M., and Trement, F. (eds. 2000). Non-Destructive Techniques Applied to Landscape Archaeology. The Archaeology of Mediterranean Landscapes 4. Oxford. Perkins, P., and Walker, L. (1990). ‘Survey of an Etruscan city at Doganella, in the Albegna valley’, PBSR 58: 1–150. Plog, S. (1976). ‘Relative eYciencies of sampling techniques for archaeological surveys’, in Flannery 1976: 136–58. Robinson, J., and Zubrow, E. (1999). ‘Between spaces: Interpolation in archaeology’, in Gillings et al. 1999: 65–83. Talbert, R. J. A. (ed. 2000). Barrington Atlas of the Greek and Roman World. Princeton. Taylor, J. (2007). An Atlas of Roman Rural Settlement in England. Council for British Archaeology. York. Walker, L. (1985). ‘Survey of a settlement: A strategy for the Etruscan site of Doganella in the Albegna valley’, in Haselgrove et al. 1985: 87–94. Witcher, R. (2005). ‘The extended metropolis: Urbs, suburbium and population’, JRA 18: 120–38. —— (2006a). ‘Settlement and society in Early Imperial Etruria’, JRS 96: 88–123. —— (2006b). ‘Broken pots and meaningless dots? Surveying the rural landscapes of Roman Italy’, PBSR 74: 39–72.

Part III Agriculture

This page intentionally left blank

7 Quantifying Egyptian Agriculture Alan Bowman

INTRODU C TION There is a formidable amount of evidence for the agricultural economy of Ptolemaic, Roman, and Byzantine Egypt, mainly in the papyri but also signiWcantly (and increasingly) in the accumulating archaeological data.1 In contrast to other areas where historians frequently bemoan the lack of quantiWable evidence, it may be that the quantity and complexity of the documentary evidence has deterred many specialists, both Egyptian and non-Egyptian, from attempting studies which involve quantiWcation. For this reason, among others, the subject is a useful testing ground for an analysis of methods and approaches in economic quantiWcation. We can at present beneWt from a substantial number of published studies on diVerent aspects of the Egyptian agrarian economy and there is plenty of scope for adding to these and for gathering more primary source material. In what follows I concentrate on describing, with some signiWcant examples, the kind of direct and indirect evidence we have, and on considering what kind of questions it might be made to answer. This is just one of the building-blocks which might form part of a better understanding of the economy of the Roman empire. Considering it here in isolation from other aspects is an unfortunate 1 Abbreviations used for editions of papyri may be found in the Checklist of editions of Greek and Latin papyri, ostraca, and tablets: http://scriptorium.lib.duke. edu/papyrus/texts/clist.html. For archaeological evidence see Bagnall 2001.

178

Alan Bowman

but unavoidable tactical necessity. In any attempt to quantify the economy of Egypt we need to take the agricultural economy as part of a larger picture which also involves other major topics addressed in a general way in this volume: demography and settlement patterns; taxation; trade and exchange. We also need to relate it to the wider canvas of the Roman Mediterranean and to address the matter of generalization, or at least applicability to that wider world. Egyptian conditions are unique in some important respects but that should not prevent Egypt from making a contribution to our understanding of the Roman economy, especially methodologically. For the framework, there are two obvious problems, common to all such studies of the ancient history of the Mediterranean. First, comparatively abundant though the evidence for Egypt is, it is still heavily skewed chronologically and geographically, almost completely lacking representation in the Delta and providing less evidence for the late Ptolemaic and early Roman periods than for the early Ptolemaic and later Roman periods.2 Any attempt at correlating the chronological and spatial axes in a graph of the distribution of evidence is not a simple matter.3 There is thus the perennial problem of the general applicability of conclusions drawn from analysis of speciWc datasets, many of which are derived from groups of documents which are not random. The second problem is one that is common to many aspects of the study of the ancient Mediterranean, a general lack of robustness in many of the estimates, particularly population numbers, which are crucial to the macroeconomic picture.4 It is clear that with the possible margins of error implied here in estimates of magnitude which are essentially unknowable, deductions about or models of important economic factors (GDP, population density, rate of urbanization, size of agricultural sector, tax yield and surpluses) will be fragile and speculative at best, selfcontradictory or impossible at worst. On the other hand, as will be shown below, on the regional or microeconomic scale (in Egypt and 2 For a summary of the distribution, see Habermann 1998. 3 For a good analysis of the diYculties involved see Bagnall 1995, ch. 4. 4 Cf. Introduction. For a summary of the state of opinion on Rome and Italy see Morley 2001, with some bibliography in n. 6, Scheidel 2001a, and Lo Cascio in this volume. The Wgure of 55–60 million originates with Beloch (1886) and has subsequently been used by Brunt (1971) and many others. For Roman Egypt see Bagnall and Frier 1994: 51–7.

Quantifying Egyptian Agriculture

179

elsewhere) there are data which inspire considerably more conWdence and seem subject to smaller margins of error. The challenge, of course, is to move securely from the latter to the former without begging the question and while it is clearly impossible to accumulate all relevant data for the ancient Mediterranean, it ought at least to be possible to show how one could use speciWc datasets within a plausible framework on our timescale to achieve a better understanding of economic structures and long-term developments. It should be stressed, however, that the suggestions made below are at this stage of a speculative nature and need rigorous testing against aggregated data. What follows here is Wrst of all a brief mise-en-sce`ne which attempts to indicate how we might hope to achieve some results in quantifying agricultural activity, with illustrations and examples under four headings: . . . .

Quantity and distribution of land Crop production and consumption Animal power and transport Wages, costs, and prices.

In this preliminary attempt to suggest how we might use the available data, I avoid discussing many complex technical details which will be obvious to those who know the papyrological evidence intimately. I conclude with some remarks on methods of analysing the data and identifying the crucial questions which can be addressed by developing models which are consistent with the data.

Q UA N T I T Y A N D D I S T R I B U T I O N OF LA N D I begin with the question of scale and perspective, and some brief remarks on the evidence for Egypt as a whole, for its regional divisions and for local areas (towns and villages). As indicated above, the most diYcult thing is to produce any macroeconomic estimates in whose precision one can have any conWdence at all. The total surface area of the ancient kingdom or province is plausibly estimated by Baines and Malek at 34,000 km2,

Fig. 7.1. Egypt in the Early Roman period

Quantifying Egyptian Agriculture

181

with twice as much land available in the Delta as in the Valley.5 Within that it has been variously calculated that the amount of land under agricultural cultivation was probably between 20,000 and 25,000 km2, and be it noted that parameters of 20–25% will signiWcantly aVect any resulting calculations.6 The overall Wgure itself is probably mainly attributable to Karl Butzer, who oVered a breakdown of 16,000 km2 for the Delta, 10,000 km2 for the Valley, and 1,300 km2 for the Fayum; this Wgure was calculated for the midPtolemaic period, c.150 bc, and was therefore presumably intended to include the Ptolemaic expansion of cultivation in the Fayum.7 It seems to me a Xawed method to break down these estimates by nome, town, and village given the inherent uncertainty in the grand total (and the total lack of detail for the Delta). But it should be noted that the grand total is not entirely lacking in evidential context since the Edfu Donations Text of the Ptolemaic period gives a total of 9 million arouras of cultivated land, close to 25,000 km2.8 Since we are just as badly oV as regards the related question of total population Wgures, with estimates ranging between 4.5 and 8 million, the procedure of using such macroeconomic guess-estimates as a basis seems doubly treacherous.9 The obvious danger is that of piling one uncertain estimate on top of another, compounding the probable distortions as we do so. For agriculture and the agricultural population, one such potential distortion lurks in estimates and multipliers for density of habitation and for the ratio of urban to rural population.10 Another lies in the value and level of productivity, the tax yields, and the level of exports to Rome and later to 5 Baines and Malek 1980: 16–17. 6 For the lower Wgure, Scheidel 2001b: 220. Van Minnen (1997: 29) estimated 20,000–23,000 km2 (7.25–8.35 million arouras); Bowman (1996: 13), a slightly higher Wgure but that too is admittedly impressionistic. 7 Butzer 1976: 83 (and cf. Bagnall 1993: 110–14). 8 Rathbone 1989: 174. 9 This subject is reserved for separate, detailed treatment. For representative samples in the range see Bowman 1996: 17, Bagnall and Frier 1994: 54, van Minnen 1997, Scheidel 2001a and 2001b. Many have rejected Josephus’ Wgure (BJ 2.385) of 7.5 million excluding Alexandria in the later Wrst century ad. Butzer’s Wgure (1976: 83) for c.150 bc is 4.9 million. Clarysse and Thompson (II 2006: 102) estimate c.1.5 million for 250 bc, much lower than most other suggested totals. 10 Butzer’s overall Wgure is 240/km2. These issues are also discussed by Lo Cascio in this volume.

182

Alan Bowman

Constantinople. Several scholars, including myself, have been tempted to comparisons with nineteenth-century Egypt, for which there are some quantitative data and some census Wgures, although the margin of uncertainty is far from negligible there also.11 In short, it is diYcult to make any signiWcant inferences with conWdence from these guess-estimates, either upwards to the wider empire (especially given Egypt’s wholly exceptional agricultural productivity) or downwards to the regional and local scale. One could, of course, multiply a crude gross Wgure for land under cultivation by a notional average yield in cereal equivalent,12 taking into account non-arable land, but the usefulness of such a Wgure is not clear to me given all the other uncertainties in our calculations. By the same token, it is diYcult to see that we can make any sense of a GDP-based approach without putting considerable eVort into generating more robust Wgures; this applies particularly to the derivation of useful per capita estimates but there may be more proWtable ways of looking at this (see the Introduction to this volume, above). In fact, the only valid conclusion to be drawn from this representative but by no means exhaustive survey is that the whole enterprise is, in its current state, a mess. One might, nevertheless, see potential in attempting some overall quantiWcation of the magnitude of the agricultural sector, or parts of it.13 To this end, there seems to me more to gain by working from the bottom up, with the evidence we have, which is in fact quite substantial, and on the whole much better than elsewhere, if not without its own diYculties. The problems are (i) that it is patchy and (ii) that it is diYcult to know how to make it address ‘the bigger picture’. It is possible to derive some reasonably precise Wgures for sizes of a few particular districts (nomes), of which there are between 40 and 50 altogether in the Roman period. Since sizes were obviously widely variable, we need to avoid simple multipliers and crude extrapolations from a very few. The Oxyrhynchite, the Hermopolite, and the Arsinoite, which were undoubtedly among the larger nomes, are the best examples: for the Hermopolite 1,140 km2 ¼ 413,820 11 27,659 km2 according to the census of 1882, Butzer 1976: 83. Bowman 1996: 17–18, Bowman and Rogan 1999: 1–7, Bagnall 1992: 141. 12 It is generally acknowledged that average yields in Egypt were exceptionally high but here again modern estimates vary quite widely, see Rowlandson 1996: 251. 13 For a recent robust attempt see van Minnen 2000, discussed further below.

Quantifying Egyptian Agriculture

183

arouras; the Oxyrhynchite 780 km2 ¼ 283,140 arouras; the Arsinoite (Fayum) c.900 km2 ¼ 326,700 arouras.14 And within those, there are some further data for the amount of land under cultivation; for the Oxyrhynchite we have a Wrm Wgure in a document of 202,534 arouras (¼ c.557 km2) of land taxable in grain in the early fourth century ad, which would represent c.72% of total area; a speculative model developed by Bagnall for the Hermopolite Nome derives a total arable area of c.298,000 arouras (¼ c.820 km2) simply by calculating 72% of the total area by analogy with the Oxyrhynchite.15 At some point, we will naturally want to relate this to settlement and habitation patterns. The demographic issues cannot be dealt with in detail here, but we do already have some reasonable estimates (which are capable of further reWnement) of the size of some of the urban settlements (metropoleis) and also of some villages, particularly those in the Fayum.16 For the agricultural economy, the size and number of villages in some of the better known nomes and the area of land around them is of more crucial signiWcance. It is worth drawing attention to the fact that we do have some data from the documents on village names and the periods over which they are attested in the nomes mentioned above, the Oxyrhynchite, Hermopolite, and Arsinoite, and also in the Herakleopolite. For the present, I oVer only a very crude count suggesting orders of magnitude, but it ought to be possible to reWne this and to achieve a more nuanced picture of changes in the number of settlements over time, relative sizes, and perhaps spatial relationships.17 Archaeological

14 For the estimates see Bagnall 1993: 333–5; note that these are based to a signiWcant extent on Butzer’s overall calculation. 15 Bagnall and Worp 1980, Bagnall 1992: 137. 16 Bowman 2000; the basis for some of the calculations in Alston 2002: 332–4 is not secure. See also Tacoma 2006. 17 Oxyrhynchite: 86 villages with their ‘own administration’, Rathbone 1990; c.120 village names in the Oxyrhynchite, Pruneti 1981; Hermopolite: P.Col. IX 247 (incomplete), has 72 toponyms in the early fourth century ad, from which Bagnall (1993: 110), derives a total of c.100 villages; there are c.170 village names in M. DrewBear 1979, over the period from the Wrst century bc to the fourth century ad; Arsinoite: 165 villages, P.Teb. II, App. II (see also http://fayum.arts.kuleuven.ac.be/ general.html); Herakleopolite: c.130 names in Falivene 1998. Bagnall (loc. cit.) also derives a total Wgure of 2,300 villages in the whole of Egypt but there may be a large margin of error here.

184

Alan Bowman

surveys, of which a signiWcant number have been initiated in recent years, will add more useful evidence here.18 For individual towns and villages we also have a reasonable amount of quite precise data (though not always complete) for the size of territories and agricultural land dependent on them, for example at Kerkeosiris in the Ptolemaic period and at several other villages in the Roman period. There is a noticeable bias towards the Fayum here, but the evidence for relative sizes of villages elsewhere might give us some proxies for possible quantities of agricultural land attached. In some cases it is diYcult to be sure whether the Wgures for village territories are complete or partial but these data are actually quite rich and there is much more than I am able to show here.19 It reveals, as one would expect, a great deal of variation in scale and size. . Kerkeosiris (Fayum), second century bc: c.4,700 arouras (c.1,300 ha) of which the village itself and waste land account for 240 arouras (66 ha).20 . Karanis, ad 179/80 or 211/12: 11,500 arouras (3,160 ha) of grainland (3,832 ¼ c.1,050 ha private land); early fourth century ad, 4,271 arouras (c.1,174 ha).21 . Fayum villages, ad 167: Hiera Nesos 4,061 arouras (c.1,116 ha), Drumos of Hiera Nesos 1,753 arouras (c.482 ha), Drumos of Kerkeesis 556 arouras (152 ha), Ptolemais Nea 3,924 arouras (1,079 ha), epoikion of Perkeesis 2,161 arouras (594 ha), total c.3,423 ha.22 . Philadelphia, early third century ad: 4,583 arouras (1,260 ha) of private land.23 . Bacchias, ad 219/20: survey of unXooded land includes 70,600 arouras (c.19,415 ha) of crown land.24

18 Rathbone 1996, http://www.ees.ac.uk/deltasurvey/dsintro.html. http://www. dur.ac.uk/penelope.wilson/Delta/Survey.html. 19 For some further detail see Bagnall 1993: 110–14. 20 Crawford 1971: 44. 21 P.Mich. VI 372, Bagnall 1992. 22 P.Bouriant 42. 23 P.Yale III 137. 24 P.Gron. 2. This Wgure is of an entirely diVerent order of magnitude from the others and requires closer scrutiny or explanation.

Quantifying Egyptian Agriculture

185

. Tebtunis, ad 222/3: 3,588 arouras (986 ha) of unXooded land.25 . Naboo (Apollonopolite Heptacomias, middle Egypt), early second century ad: 7,075 arouras (1,945 ha).26 Detailed analysis is reserved for a future volume, but it can be indicated here that with reasonably secure estimates for the population of some of these villages at some periods (for example Kerkeosiris in the second century bc, Philadelphia, Karanis, and Tebtunis, all probably in the range of 3,000–4,000 in the period c. ad 50–150), we might hope to be able to achieve some notion of the relationship between population, land, and agricultural product. Moving down from the communal picture to the anatomy of individuals’ landholdings, there is some good evidence for distribution and tenure of land, which oVers some points of departure. The documents recording land-tenure in the Hermopolite Nome in the mid-fourth century oVer evidence for a high degree of inequality of distribution among urban landholders and allow some comparison of that level of inequality over time. They were also particularly useful in an earlier study because the two almost contemporaneous registers allowed us to track changes within the cohort of landholders. For comparison I used evidence from Fayum villages, which showed a more equal distribution.27 That work was ampliWed and extended by Bagnall in a very useful study of landholding in the rural villages of the nome.28 This enabled him to construct a model for the nome as a whole, an area estimated at around 1,140 km2, which would constitute about 4% of Egypt’s cultivable land. The model population comprises 7,400 rural landowners with an average holding of 24.3 arouras, 952 residents of the city of Hermopolis with an average holding of 39 arouras, and makes suggestions about the proportional holdings of urban and rural residents and the patterns of distribution. Bagnall has also analysed a late second-century tax register from Panopolis from which he deduces, albeit speculatively, some Wgures for the residents of the metropolis holding vineyards and orchard land in the nome: the proportions of such land to cereal land seem to be roughly in line with what can be deduced for other places in Egypt, women seem to hold a higher percentage of such land than their grainland-holding counterparts at Hermopolis (150 25 P.Giss.Univ. VI 52. 27 Bowman 1985.

26 P.Giss. 60. 28 Bagnall 1992.

186

Alan Bowman

years later), and the Gini coeYcient of inequality is very close to that for metropolitan landholders at Philadelphia in the early third century.29 In summary, the Gini coeYcients are as follows: . Kerkeosiris, late second century bc: 0.374 (all holders of all categories of village land). . Philadelphia, early third century ad, private land only: 0.532 (entries with complete data), 0.516 (all entries). . Karanis, early fourth century ad: 0.638 (landholders residing in the metropolis of the nome), 0.431 (landholders residing in the village). . Hermopolite Nome, mid-fourth century ad: two registers, G(iessen) and F(lorence), list landholders resident in Herm(opolis) and Ant(inoopolis). G Herm. ¼ 0.794, G.Ant. ¼ 0.637, combined ¼ 0.715. F Herm. ¼ 0.815, F.Ant. ¼ 0.616, combined ¼ 0.758. . Panopolis, early second century ad, vineyard and orchard land: 0.702. . Krokodilopolis, mid-Wrst century ad: 0.553.30 In the cases where the Wgures themselves are extrapolated from those actually given in the texts of the registers they are obviously subject to some uncertainty, but the claim would be that the quantiWcation is likely to be broadly accurate and that it allows some important general conclusions about the character of Egyptian agriculture. The evidence from the Fayum village of Kerkeosiris in the Hellenistic period (second century bc) shows a signiWcantly greater degree of equality than the later Wgures. But, broadly speaking, the evidence for landholdings of village residents in the Fayum from the late second century bc to the early fourth century ad shows signiWcantly more equality of distribution than those of urban landholders at Hermopolis in the mid-fourth century ad, with more equality in holdings of public (demosia) than private (idiotike) land. By this later date the complex mixture of royal, imperial, public, and private ownership had disappeared in favour of almost universal private ownership.31 It is not clear whether that in itself tended 29 Bagnall 2002. 30 See Sharp 1998: 37. 31 Bowman 1985. The distinction marked by the survival in the mid-fourth century of the terms demosia ge and idiotike ge is simply the rate of taxation, the former being higher than the latter.

Quantifying Egyptian Agriculture

187

to favour the concentration of more land in the hands of the urban elites, but even among them large landholdings tended to be fragmented and there is no clear evidence for the build-up of large estates which one could characterize as consolidated and ‘feudal’. There are many other aspects to the evidence for land tenure (not least the matters of fragmentation of holdings of individuals and ‘privatization’), of which I think two are worth considering brieXy here. One is the question of ‘average’ holdings which we can calculate from lists such as those for Hermopolis, Philadelphia, and Krokodilopolis; are these more valuable than clusters around a mean?32 It is a simple but important point that, taken on their own, they do, of course, have the disadvantage of concealing inequality of distribution and may correspond to no reality at all. The other is that the comparison of the landholdings of individuals in the Hermopolite registers at diVerent times allows some inferences about the amount (and proportion) of land changing hands in this community.33 A diVerent approach to the question of quantifying the mobility of land might be oVered by the evidence of contracts for sale and lease of land at the Fayum village of Tebtunis over a very short period in the ad 40s. The registers of contracts from the village record-oYce are understood to present a full record of contracts deposited and this enables us to calculate how much land (though not as a percentage of a total since we do not have such a Wgure) was changing hands and how many people were involved in such transactions over an 18-month period in a village with (probably) around 3,000 inhabitants.34 Analysis of such data might suggest whether the market in land became more or less active over time.

CROP PRODUCTION AND CONSUMPTION For agricultural production, estimates of Egypt’s total yield in wheat equivalent are as ill-founded as any of the macroeconomic Wgures 32 For Krokodilopolis, see Sharp 1998: 37. 33 Bowman 1985: 154. 34 The grapheion registers are analysed from a bureaucratic rather than an economic perspective by Toepel (1973). See also the Introduction to this volume, pp. 40–1. I hope to present some further statistical analysis on the website (http://www.oxrep.class.ox.ac.uk).

188

Alan Bowman

discussed above. In this case, they tend to be derived from individual examples of Wgures from documents or other sources which are used to infer some sort of an average, then multiplied by the Wgure which represents the total amount of land under cultivation. We might be able to do better than this, but not just by plucking Wgures out of thin air. There are, after all, some larger scale Wgures, such as that in Justinian, Edict XII, providing a basis for a calculation which allowed Rathbone to conclude that the total assessment of taxes in kind remained broadly stable from the Roman annexation in 30 bc (perhaps even from the Ptolemaic period), until after the Arab conquest.35 If that is correct, it would naturally raise doubts about postulating any signiWcant overall reduction in the amount of land under cultivation in the late Roman and Byzantine periods, despite clear evidence of shrinkage in some Fayum villages. I leave that aside for the moment to concentrate on the ways in which we might use some sample data. The documents oVer evidence of the range of crops grown and this can be supplemented by recent, admittedly patchy, archaeological data which are valuable for indications of the presence of commodities outside the mainstream (wheat, barley, fodder, wine, oil).36 There are some documents which allow estimates of the proportions of land devoted to diVerent crops, principally the Wrst four just mentioned. These are speciWc to particular localities and cannot simply be extrapolated or multiplied to give regional or provincial totals. One might intuitively expect to Wnd a very high proportion of wheat cultivation overall but that is too imprecise to be of any real use in the present context. There are, at least, some plausible hypotheses which one might bear in mind and continue to test: for example, that the proportion of orchard and garden land (including vineyards) under cultivation was in the region of 8% of the total land under cultivation. Did this increase as the popularity of wine increased to 35 Rathbone 1989: 175. See also Lo Cascio’s discussion in this volume of the relationship between urbanization rate and agricultural output. 36 For example, plant and faunal remains: Karanis (Boak 1933: 87–93); Quseir (Whitcomb and Johnson 1982: ch. 16: Wattenmaker, 17: Wetterstrom); Coptos (Herbert and Berlin 2003: 206–20: Redding); Berenike (Sidebotham and Wendrich 1996: chs. 19–20, 1998: chs. 15, 17, 1999: chs. 16, 18, 2000: ch. 13); Bacchias (Pernigotti et al. 2000: 95–110: Lentini, 2001: 99–119: Lentini). There is also some evidence in the Murecine tablets (e.g. TPSulp. 45–7).

Quantifying Egyptian Agriculture

189

the detriment of the traditional Egyptian preference for beer and can it conWrm or undermine the likelihood of a move towards an ‘intense boom which stimulated investments in wine-growing throughout the Byzantine period’?37 The vagaries of such data are illustrated by the fact that a list of private landholdings from Philadelphia shows around 80% of grainland in the early third century, whereas another document from Theadelphia shows only 40% of its land under wheat in ad 164/5, which seems intuitively likely to be well below the norm.38 Another approach is oVered by a substantial number of documents which give totals of tax or rent for speciWc areas from which we might hope to deduce the total yields. For the Herakleides division of the Arsinoite Nome (Fayum), for example, we have a very large total of wheat revenues in ad 184/5, and cash taxes from garden land in the same division in a document of the second century.39 There are many others which might allow us to compare revenues over time and to test the proposition that Egypt was able to provide revenues of the same order of magnitude in the sixth century as in the Wrst.40 The diYculty with these Wgures lies in trying to calculate what proportion of the total yields is represented by the tax/rent, since there are evidently no uniform rates in either category (at least before ad 300) and much depends on the irrigation status of the land in question. Similarly, and perhaps more reliably, the rents speciWed in land-leases can be related to the yields from the land under lease.41 Any approach to the construction of a total yield for the agricultural economy will also have to recognize the importance of livestock, for which the data are much more sporadic and incomplete, consisting largely of simple registrations of Xocks of sheep and goats.42 Large commodity surpluses are a major feature of the Egyptian agricultural economy and it will be important to attempt to quantify the amounts or the relative proportions accounted for by internal 37 The quotation is from Banaji (2001: 18). 38 Range and proportion of crops sown: P.Petr. III 75, Arsinoite mid-third century bc; P.Berl.Leihg. II 32, Theadelphia ad 164/5, 40% wheat (cf. Sharp 1998: 53–4); orchard/vineyard c.7.5%, P.Yale III 137, Philadelphia early third century ad, cf. Panopolis (Bagnall 2002, Rowlandson 1996: 238). 39 Much evidence for yields of tax/rent in kind and in cash on diVerent categories of land, e.g. P.Bouriant 42, Fayum villages, ad 167; P.Oxy. LXVI 4527, Herakleides division of Arsinoite ad 184/5 (see Bagnall 2000); vineyards, P.Teb. II 503. 40 Rathbone 1989. 41 Rowlandson 1996: 251. 42 Habermann 2001.

190

Alan Bowman

consumption and by export. Evidently Egypt was able to supply Rome and later Constantinople with large amounts of grain (much of it raised as tax) and the landowners and owners of other means of production, operating in a market economy in and outside Egypt, must have turned their surpluses into ‘wealth’. There are a number of constituents of such a calculation for which we do have a reasonably Wrm basis. We have evidence for rations for agricultural workers (even if we are not sure whether they constitute the whole of their wages), we know the standard amounts distributed in the grain-dole, and there is reasonably good, if spotty, evidence for the number of loaves produced from an artab of wheat, the capacities of bakeries, and the amount of animal power required for milling.43 These can be related to the parameters of standard subsistence calorie requirements for individuals and we can probably also make some estimates of rations required for animals as well as the numbers of animals required for transport and processing (see below). It is therefore possible to make reasonably plausible estimates for a range of communities of speciWc (even if hypothetical) sizes and to attempt some calculation of the food supplies which they required.44

ANIMAL POWER AND TRANSPORT As is implied above, animal power is clearly a factor of central importance in the economics of production and consumption and it requires some attempt at quantiWcation.45 A great deal of the evidence relates to donkeys, which were the major mode of land transport for agricultural produce on the local roads.46 Camels are also well attested but tend to 43 Oxyrhynchite dole (P.Oxy. XL), 4,000 recipients @1 artab per month requires 48,000 art. wheat per annum. P.Oxy. XII 1454, bakers produce 30 loaves from 1 artab of (milled) wheat. PSI VII 908 ¼ WChr. 426, milling power of animals produces enough Xour for 3,600 people per day. 44 For Oxyrhynchus see Sharp 1998 and 2007. 45 QuantiWed data in Adams 2007, Habermann 1989 and 1990, Sharp 1998, Jo¨rdens 1995. 46 I note in passing that donkey bones are conspicuous by their absence from available reports of archaeological assemblages, but that is presumably because they tended to die ‘on the road’ and the carcasses were left to rot where they fell; also perhaps because they were not eaten and therefore do not appear in food refuse deposits.

Quantifying Egyptian Agriculture

191

be used more for long-distance desert routes, especially in the eastern desert. Animal (donkey) power is also essential for driving some kinds of irrigation machinery and for milling. We have much less information for oxen, although these were undoubtedly crucial for ploughing. Numbers of donkeys and camels, loads carried, rations required are all attested in the documents; if they do not produce global totals, they at least give reliable indications of capability (we have much less information for horses and oxen). A standard load of wheat for a donkey was 3 artabs, for a camel 6 artabs (or 4 for a foal), a wagon load 5/6 artabs. An adult camel could carry 4/4.5 metretae of oil, or 4 keramia of wine.47 A donkey will consume 3–5 artabs of barley per month and a camel 360 bundles of hay per month. One would naturally proceed from the assumption that local and regional transport is signiWcantly landbased, with grain destined for Alexandria and for export largely transported by boat. We have some evidence for the capacities of various types of boats too.48 It is obvious that the transport business is an integral element in the realization of the agricultural wealth of the country, even in the local context, and it is therefore crucial to address the questions of scale and organization, which are important both from the point of view of the state and the private individual. The pervasive importance and the complexity of this element of the agricultural economy have perhaps not been given the analysis they deserve because of the tendency to see an oversimpliWed pattern of, on the one hand, local ‘subsistence’ agriculture (requiring relatively little movement of commodities) and, on the other, long-distance transport by river to Alexandria and Rome. For the organization of transport we have a few points of entry. The archive of Nikanor is our best guide to the organization of a transportation business, and it is clear that contracts were undertaken both for the government and for private individuals (surely a ubiquitous pattern). The documentation of the Appianus estate in the third-century Fayum oVers some important evidence for the mechanics of transportation within a single ownership, but involving activity in dispersed locations, some of it connected with the sale of 47 P.Oxy. XVII 2128 records 411 donkeys carrying 1,734 loads in 7 weeks, cf. Sharp 1998: 196 for calculations of donkey and camel loads passing through customs. 48 Bagnall 1993: 35.

192

Alan Bowman

the estate’s produce.49 We can also make some attempt to estimate the charges made for transportation by land and by river and relate that to the distances travelled and the capital costs, as well as the upkeep, of the animals and the input of human labour.

WAGES, COSTS, AND PRICES These are clearly fundamental to an attempt to quantify the agricultural economy. There is a very great deal of scattered information in the papyri, but it is diYcult to contextualize and rationalize, let alone to relate to prices and costs outside Egypt. We can make use of the collection of evidence by Drexhage, although it is now somewhat out-of-date and much of the detail needs correction or reWnement.50 Comparisons may be made with evidence collected from other areas of the Roman empire (principally Italy and Africa, by Duncan-Jones)51 and for other areas and periods in the past.52 Evidence for levels of costs and prices touch on various diVerent elements in the agricultural economy. There are prices for sales of land (as well as for rents), which are of course notoriously variable according to quality and other factors.53 This makes calculation of averages or means a hazardous undertaking and perhaps a relatively unhelpful one. There may, however, be other ways in which aggregation of the amounts and the prices of land changing hands can be suggestive. Labour costs and wages associated with agricultural production are clearly crucial but the accounts in which they appear do not always allow us to feel conWdent that the amounts of cash or rations represent the full wage and they can be diYcult to contextualize.54 We also have to take account of the value of labour input which is not directly waged, such 49 The family business of Nikanor owns 30 camels (Wrst century ad), RuYng 1993. For the Appianus estate, Rathbone 1991. 50 Drexhage 1991. 51 Duncan-Jones 1982: chs. 3–4. 52 For the data from Hellenistic Babylon (the Astronomical Diaries) compiled by R. J. van der Spek see http://www.iisg.nl/hpw/data.php#babylon and cf. van der Spek 2000, 2005, Temin 2002. 53 Rowlandson 1996, Drexhage 1991. 54 e.g. Drexhage 1991, Rathbone 1991: 156–7.

Quantifying Egyptian Agriculture

193

as the dike-corve´e, and slave labour, admittedly less common in Egypt than elsewhere but not completely absent, and requiring the cost of buying and maintaining slaves to be factored in.55 Various scholars have attempted to track the prices of staple commodities, most recently Dominic Rathbone for wheat, wine, and donkeys, which do allow us to observe interesting general trends over the three centuries of the empire.56 There is also evidence which allows us to calculate transportation costs, as well as some quantiWed data in pottery leases showing the costs of producing vessels for transport of agricultural produce.57 If we can produce relatively good data for some of these elements or activities, how can they help us? On the microeconomic level we might certainly hope to aggregate estimates of costs in agricultural enterprises of diVerent types and sizes and of the relative cost of diVerent elements within such enterprises (land, seed, labour transport, irrigation, tax, and so on). I hope that it will also be possible to bring amounts of money spent on land into relation with other costs and prices in the agricultural economy. And we might also be able to speculate as to whether these changed signiWcantly over time; were they, for example, aVected by the privatization of land which had occurred by the fourth century? It should be possible and useful, too, to distinguish between free-market prices and state-regulated prices or values for agricultural produce (for example for military supplies) and to test the hypothesis that prices (for example for grain, but not only that) can be ‘rationalized’ in terms of an integrated market economy, as Temin and Kessler have attempted to do.58 For all the costs and individual prices, we need to devise some indexing system which will enable us to relate them and to some extent track the value of land, labour, and commodities.59 Finally, we might hope to compute approximately the value of the agricultural sector and the relative proportions of the product needed for subsistence and tax and, consequently, the magnitude of the surpluses. 55 For prices of slaves see Biezunska-Małowist 1977. Much of the documentary evidence from Egypt for slave sales relates to use in the domestic rather than the agricultural sector, where slaves commonly supply labour in the absence (for whatever reason) of a male head-of-household (Bagnall and Frier 1994: 70–1). 56 Rathbone 1997. 57 Cockle 1981 ¼ P.Oxy. L 3595–7. 58 Kessler and Temin 2008. 59 For an attempt at indexation for fourth-century prices see Bagnall 1985.

194

Alan Bowman METHODS, QUESTIONS, CONTEXTS

Extended discussion of quantitative methods in detail is avoided here. This conclusion considers in outline the potential of the type of evidence described above. It bears repeating that any estimate at quantiWcation on the larger provincial scale (Egypt as a whole) from the top down appears very fragile in the present state of knowledge. Exercises of this kind do have a fairly long history. The most accessible is that of A. C. Johnson in 1936 who attempted to compute the revenue which Rome drew from Egypt and the expenditures which it supported, referring back to an article by Tenney Frank in 1933 which was itself in part a critique of Wilcken and calculated an annual revenue of about 75 million HS.60 Although there is still a good deal of value in much of the detail collected by Johnson in that volume, it is evident that the macroeconomic estimates rely largely on well-known and hardly robust pieces of evidence from literary sources: Velleius on the revenues under Augustus, the annual 20 million modii of grain for Rome (‘Victor’), Strabo’s note that Ptolemy Auletes drew 12,500 talents in annual revenue according to Cicero, and so on.61 These calculations of Johnson were, of course, essentially estimates of tax revenue, not total product, leaving the thorny question of what proportion of the total product these revenues (which are of course not drawn solely from the agricultural sector) might represent. Assessments of the usefulness or viability of a GDP-based approach have diVered. Temin has produced an estimate of the GDP of the Roman empire of HS 10–12 billion (where 1 billion ¼ 1,000 million) and he derives a Wgure of about HS 190 for the per capita GDP but this, of course, depends on a total population Wgure in the 55/60 million range, derived from the Wgures Wrst proposed by Beloch and subsequently used by Brunt and others; these are no more than best (or least worst) guesses and may actually be signiWcantly wide of the mark, as they certainly are for the ‘high-counters’ such as Lo 60 Johnson 1936: 481–90, Frank 1933: 147–8. 61 For a very sceptical view of the value of evidence of this kind see Scheidel 1996.

Quantifying Egyptian Agriculture

195

Cascio.62 Debates about the size of the empire’s population reveal a degree of fragility similar to those about the population of Egypt. Furthermore, as Scheidel notes, such Wgures ‘would tell us little about actual quality of life, or even how the Roman economy performed within the constraints imposed by pre-industrial levels of productivity’.63 Rathbone has adopted another approach which points to the ‘relative absence of poverty in Roman Egypt’, suggesting that the economic data point to a low-pressure demographic regime where the population reproduced itself up to a total ceiling well below the carrying capacity of the productive resources available.64 If that were correct, it would be consistent with the generation of high surpluses which would be available as revenue to the imperial power. But it can hardly claim to be established, even if we think that estimates for gross product can be more plausibly constructed than in other provinces and might be able to suggest something about levels and distribution of wealth or poverty. Nevertheless, even if there is no secure way forward with per capita calculations, it does not follow that we cannot derive something useful from estimates of volume and value of production and the like; this possibility remains to be explored in detail. The content of the present volume and the whole research project which it represents is clearly underpinned by the notion that there are quantiWable data which will yield useful results and I here suggest the way in which progress might be achieved in thinking about the agrarian sector. First, we might hope to achieve some working hypothesis for yields from leases and levels of rent and tax, and to compare those over time, as also with land distribution, transport costs, and so on. As suggested above, I think we can also derive some useful conclusions about the size and density of urban and rural settlement and the quantities of land attached to settlements, which should be considered in relation to the available archaeological evidence. We can also take standard estimates for subsistence levels and calculate them in relation to estimates of yield and tax (though uniquely in the case of Egypt we have to allow for quite signiWcant variation from

62 Temin 2006.

63 Scheidel 2006: 54–5.

64 Rathbone 2006.

196

Alan Bowman

year to year caused by the level of the inundation, but we can estimate averages or means). We ought also to be able to calculate quantities of commodities which needed to be moved, the motor power required to do it, and the cost (in commodities like fodder and cash value). Second, even if we cannot identify a large number of individual components which can be straightforwardly aggregated to produce a ‘big picture’, we might be tempted to think that we have something with which we can work towards a reasonable estimate of the size and value of the agricultural sector in Egypt as a whole, identifying the components in a number of diVerent ways. Thus, one possibility is to envisage it in terms of intensity and value of production using a number of diVerent tools of measurement. In fact, given some reasonably plausible basis for estimating the monetary value of goods and services (such as wheat or agricultural labour), we might then be able to calculate: . Value of agricultural production in a locality or region, related to volume of money. . Measure of value against other kinds of economic products and services, relating it to trade and artisanal activity, that is relative costs which might yield some sort of index. . Income levels for communities and individuals, with some idea of range and income distribution, and therefore some notion of standard of living. This might lead to a clearer idea of scale of wealth of elites, amount and value of surpluses, level of subsistence at the lower end of the scale. . Movement (literal and virtual) of goods and value—by this I mean the attempt to detect shifts in control and exploitation of wealth, testing for example the notion of the accumulation of large landholdings in the hands of a decreasing number of people. These observations naturally demand to be placed in the context of a model-based approach to Egyptian agriculture, which is what they actually imply. Here it is useful to take van Minnen’s recent article as a point of departure which may illustrate the advantages and the

Quantifying Egyptian Agriculture

197

diYculties inherent in such an approach.65 One of his investigations derives the size of the agricultural production from the yields implied by rental level in leases over the period from the Wrst to the seventh century. What the graph (212) actually shows is a dip in the level of attested rents from the fourth century onwards. In order to deduce a drop in yields one must suppose that the rent as a proportion of the yield remained relatively constant, given which one can further infer heavier taxation, decreasing population, and so on. None of the conclusions would follow if it were the case, as it might be, that for various possible reasons (for example privatization of land, lower management and production costs, etc.) there was a tendency towards decrease in rent as a proportion of the yield. What is needed is quantiWcation of other kinds of data which may support one or other explanation for the decrease in rents. In the following section of the article, dealing with what he calls the ‘nitty-gritty’ of Egyptian agriculture, van Minnen focuses more on the evidence provided by particular communities or villages, which is where in my view the real strength of the evidence lies. With the help of metrics of this sort discussed above, I suggest that we might be able to construct a limited number of economic models which are derived from available data for ‘units’ of diVerent kinds and sizes, with some chronological and geographical distribution in which we could measure and evaluate the agricultural sector and place it in the context of the economic anatomy of the unit as a whole. Thus, we might envisage reconstructing a region (the Fayum, the Oxyrhynchite Nome), a major urban community (the most promising candidates are likely to be the two metropoleis of Oxyrhynchus and Hermopolis), a small number of villages, for which the Fayum will provide some evidence (Karanis, Philadelphia, Theadelphia,66 and Tebtunis), an estate (Appianus, Sarapion), an individual landholder. These will inevitably oVer diVerent perspectives and emphases, partly as a result of the nature of the evidence which has happened to survive and partly as a result of genuine diVerences from unit to unit or in diVerent regions. 65 ‘We must have recourse to a historical model which has to be dynamic enough to accommodate long-term changes and variable enough to adapt to local circumstances’ (van Minnen 2000: 207). 66 Analysed in some detail by van Minnen (2000: 214), following Sharp (1999) and the unpublished Leuven dissertation of J. France; see also Sharp 1998.

198

Alan Bowman

In urban centres, for example, we will be likely to focus on landowners and the mode and scale of property and management. Such scenarios will not satisfy those who insist on demonstrating typicality or generalizability, but they ought to indicate possible diVerences over time and space which can be compared to data for other kinds of economic activity. How might all or any of this contribute to the wider picture of the Roman economy? We have several hopes. One is that it will suggest a suYcient number of useful proxies for economic trends. A second is that it will illustrate and demonstrate approaches which are methodologically sound, will evade the diYculties which have so far been evident in top-down estimates and a number of smaller but more robustly constructed elements. These can only be understood as exemplars or indicators of particular kinds against which evidence for other places and periods (almost all of it of a diVerent kind from that in the papyri) can be measured. Ancient historians have been used to thinking that our evidence is inadequate, compared to other pre-modern periods in European and Mediterranean history. But we are gradually being persuaded by historians of those periods that this is simply not so. Egypt is not ‘typical’, but the attempt to deWne and describe ‘typicality’ across this huge and diverse geographical area is more likely to lead us astray than to help. Egypt’s agricultural economy is, for various reasons, very idiosyncratic. It would be a reductio ad absurdum of the most obvious kind to take the value of Egypt’s agricultural product (if we could compute it), derive a per capita or even a per hectare Wgure, and then use these as multipliers for other provinces or regions to derive an estimate for the empire as a whole. However, if we can estimate the gross economic value of agricultural and other kinds of activity, with some notion of subsistence requirements, tax revenue, and surpluses, that is a positive result which, in turn, helps to characterize the social and economic structures (for example the role of the wealthy elite in the wider eastern Mediterranean canvas). If we can model the structure and functions of individual communities or units in the agrarian economy we will have made some signiWcant progress. We might then have some indication of whether that gross product was larger or smaller in ad 1 than ad 350, what the Xuctuations were over that period, and whether it was diVerently distributed. In general, we need to place

Quantifying Egyptian Agriculture

199

this in the context of an eastern empire in which some have claimed to identify clear signs of a relatively prosperous and in some areas growing population which does not sit well with the dogma of ‘economic decline’ in the later empire. There is still a great deal to be done in these areas as the recent work of McCormick and Wickham shows,67 not least in emphasizing that no simple statement about ‘decline’ and ‘later empire’ is likely to be either useful or right.

NOTE ON METROLOGY Aroura, unit of land measurement: 0.68 acre, 0.275 ha Artab, dry measure: 38.78 litres ¼ 4.5 modii Italici ¼ 30.2 kg of milled wheat Modius Italicus: 8.4 litres Medimnus, Attic measure: 52.5 litres ¼ 1.32 artabs ¼ 6 modii Drachma ¼ 6 or 7 obols ¼ 1 sestertius Tetradrachmon ¼ 1 denarius Talent ¼ 6,000 drachmas Bibliography Adams, C. E. P. (2007). Land Transport in Roman Egypt. Oxford. Alston, R. (2002). The City in Roman and Byzantine Egypt. London. —— and Adams, C. E. P. (1997). ‘Urbanism and the urban community in Roman Egypt’, JEA 83: 199–216. Alston, R. (2002). The City in Roman and Byzantine Egypt. London. Bagnall, R. S. (1985). Currency and InXation in Fourth Century Egypt (BASP Suppl. 5). Chico, Calif. —— (1992). ‘Landholding in late Roman Egypt: The distribution of wealth’, JRS 82: 128–49. —— (1993). Egypt in Late Antiquity. Princeton. —— (1995). Reading Papyri, Writing Ancient History. London. —— (2000). ‘P.Oxy. 4527 and the Antonine plague in Egypt: Death or Xight?’, JRA 13: 288–92. —— (2001). ‘Archaeological work on Hellenistic and Roman Egypt 1995–2000’, AJA 105.2: 227–43. 67 McCormick 2001, Wickham 2005.

200

Alan Bowman

Bagnall, R. S. (2002). ‘Public administration and the documentation of Roman Panopolis’, in A. Egberts et al. (eds.), Perspectives on Panopolis, an Egyptian Town from Alexander the Great to the Arab Conquest. Leiden, 1–12. —— and Frier, B. W. (1994). The Demography of Roman Egypt. Cambridge. —— and Worp, K. A. (1980). ‘Grain land in the Oxyrhynchite Nome’, ZPE 37: 263–4. Baines, J. R., and Malek, J. (1980). Atlas of Ancient Egypt. Oxford. Banaji, J. (2001). Agrarian Change in Late Antiquity. Oxford. Barois, J. (1911). Les irrigations en Egypte. Paris. Beloch, K.-J. (1886). Die Bevo¨lkerung der griechischen-ro¨mischen Welt. Leipzig. Biezunska-Ma£owist, I. (1977). L’esclavage dans l’Egypte romaine. Wrocław. Boak, A. E. R. (ed. 1933). Karanis: The Temples, Coin Hoards, Botanical and Zoological Reports; Seasons 1924–31. Ann Arbor. Bowman, A. K. (1985). ‘Landholding in the Hermopolite Nome in the fourth century ad’, JRS 75: 137–63. —— (1996). Egypt after the Pharaohs. London and Berkeley. —— (2000). ‘Urbanisation in Roman Egypt’, in E. Fentress (ed.), Romanization and the City: Creation, Transformations and Failures (JRA Suppl. 38). Portsmouth, RI, 173–87. —— and Rogan, E. (eds. 1999). Agriculture in Egypt from Pharaonic to Modern Times. London. —— and Thomas, J. D. (2003). The Vindolanda Writing-Tablets (Tabulae Vindolandenses III). London. Brunt, P. A. (1971). Italian Manpower 225 bc–ad 14. Oxford. Butzer, K. W. (1976). Early Hydraulic Civilisation in Egypt: a Study in Cultural Ecology. Chicago. Clarysse, W. (2004). ‘The Fayum’, http://fayum.arts.kuleuven.ac.be/ general.html (accessed 8 Oct. 2007). —— and Thompson, D. J. (2006). Counting the People in Hellenistic Egypt, vols. I–II. Cambridge. Cockle, H. W. (1981). ‘Pottery manufacture in Roman Egypt: A new papyrus’, JRS 71: 87–97. Crawford, D. J. (1971). Kerkeosiris. Cambridge. Drew-Bear, M. (1979). Le Nome Hermopolite, toponymes et sites. Missoula, Mont. Drexhage, H.-J. (1991). Preise, Mieten/Pachten, Kosten und Lo¨hne im ¨ gypten bis zum Regierungsantritt Diokletians. St Katharinen. ro¨mischen A Duncan-Jones, R. P. (1982). The Economy of the Roman Empire: Quantitative Studies. Cambridge.

Quantifying Egyptian Agriculture

201

Egypt Exploration Society. ‘Delta Survey’, http://www.ees.ac.uk/deltasurvey/ dsintro.html (accessed 8 Oct. 2007). Erdkamp, P. (2005). The Grain Market in the Roman Empire. Cambridge. Falivene, M. R. (1998). The Heracleopolite Nome, a Catalogue of the Toponyms. Atlanta. France, J. (1999). ‘Theadelpheia and Euhemereia: Village History in Graeco-Roman Egypt’. Unpublished diss. Leuven. Frank, T. (1933). ‘On Augustus and the aerarium’, JRS 23: 143–8. Habermann, W. (1989–90). ‘Datenanalyse an den Zolldokumenten im ro¨mischer Zeit I’, in H.-J. Drexhage and J. Su¨nskes (eds.), Migratio et commutatio: Studien zur alten Geschichte und deren Nachleben. St Katharinen, 157–75; ‘II’, MBAH 9: 50–94. —— (1998). ‘Zur chronologischen Verteilung der papyrologischen Zeugnisse’, ZPE 122: 144–60. —— (2001). ‘Die Deklarationen von Kleinvieh (Schafe und Ziegen) im ¨ gypten: Quantitative Aspekte’, in P. Herz and G. Waldherr ro¨mischen A (eds.), Landwirtschaft im Imperium Romanum (Pharos XIV). St Katharinen, 77–100. Herbert, S. C., and Berlin, A. (eds. 2003). Excavations at Coptos (Qift) in Upper Egypt 1987–1992 (JRA Suppl. 53). Portsmouth, RI. Hitchner, R. B. (1995). ‘Historical text and archaeological context in Roman North Africa: The Albertini tablets and the Kasserine survey’, in D. B. Small (ed.), Methods in the Mediterranean: Historical and Archaeological Views on Texts and Archaeology. Leiden, 124–42. Hudson, P. (2000). History by Numbers: An Introduction to Quantitative Approaches. London. Johnson, A. C. (1936). Roman Egypt to the Reign of Diocletian (¼ T. Frank (ed.), Economic Survey of Ancient Rome II). Baltimore. Jçrdens, A. (1995). ‘Sozialstrukturen im Arbeitstierhandel des kaiserzeitli¨ gypten’, Tyche 10: 37–100. chen A Kessler, D., and Temin, P. (2008). ‘Money and prices in the early Roman empire’, in W. V. Harris (ed.), The Monetary System of the Greeks and Romans. Oxford, 174–207. Krger, J. (1990). Oxyrhynchos in der Kaiserzeit: Studien zur Topographie und Literaturrezeption. Frankfurt. Lo Cascio, E. (1994). ‘The size of the Roman population: Beloch and the meaning of the Augustan census Wgures’, JRS 84: 23–40. —— (1999). ‘La popolazione dell’Egitto romano’, in M. Bellancourt-Valdher and J.-N. Corvisier (eds.), La de´mographie historique antique. Arras, 153–69. McCormick, M. (2001). Origins of the European Economy. Cambridge.

202

Alan Bowman

Morley, N. (2001). ‘The transformation of Italy 225–28 bc’, JRS 91: 50–62. Mller, K. (2003a). ‘Places and spaces in the Themistou Meris (Fayum/ Graeco-Roman Egypt): Locating settlements by Multidimensional Scaling of papyri’, Ancient Society 33: 103–25. —— (2003b). ‘Mastering matrices and clusters. Locating Graeco-Roman settlements in the Meris of Herakleides (Fayum/Egypt) by MonteCarlo-Simulation’, Archiv fu¨r Papyrusforschung 49: 218–52. —— and Lee, W. (forthcoming). ‘From mess to matrix and beyond. Estimating the size of settlements in the Ptolemaic Fayum/Egypt’, Journal of Archaeological Science. Oates, J. F., et al. (2005). ‘Checklist of editions of Greek, Latin, Demotic, and Coptic papyri, ostraca, and tablets’, http://scriptorium.lib.duke.edu/ papyrus/texts/clist.html (accessed 8 Oct. 2007). Pernigotti, S., et al. (eds. 2000–1). Bakchias VII, rapporto preliminare della campagna di scavo 2000; Bakchias VIII, rapporto preliminare della campagna di scavo 2001. Bologna. Pruneti, P. (1981). I centri abitati dell’Ossirinchite: repertorio toponomastico. Florence. Rathbone, D. W. (1989). ‘The ancient economy and Graeco-Roman Egypt’, in L. Criscuolo and G. Geraci (eds.), Egitto e storia antica dall’ellenismo all’eta` araba. Atti del colloquio internazionale, Bologna, 31 Agosto– 2 Settembre, 1987. Bologna, 159–76. —— (1990). ‘Villages, land and population in Graeco-Roman Egypt’, PCPS 36: 103–42. —— (1991). Economic Rationalism and Rural Society in Third-Century ad Egypt: The Heroninus Archive and the Appianus Estate. Cambridge. —— (1993). ‘Egypt, Augustus and Roman taxation’, CCG 4: 81–112. —— (1996). ‘Towards a historical topography of the Fayum’, in D. M. Bailey (ed.), Archaeological Research in Roman Egypt (JRA Suppl. 19) Ann Arbor, 50–6. —— (1997). ‘Prices and price formation in Roman Egypt’, Economie antique: Prix et formation des prix dans les e´conomies antiques (Entretiens d’arche´ologie et d’histoire 3). St-Bertrand-de-Comminges, 183–244. —— (2006). ‘Poverty and population in Roman Egypt’, in M. Atkins and R. G. Osborne (eds.), Poverty in the Roman World. Cambridge, 100–14. Rowlandson, J. L. (1996). Landowners and Tenants in Roman Egypt: The Social Relations of Agriculture in the Oxyrhynchite Nome. Oxford. Ruffing, K. (1993). ‘Das Nikanor-Archiv und der ro¨mische Su¨d- und Osthandel’, MBAH 12: 1–26. Scheidel, W. (1996). ‘Finances, Wgures and Wction’, CQ 46: 222–38.

Quantifying Egyptian Agriculture

203

—— (ed. 2001a). Debating Roman Demography (Mnemosyne Suppl. 211). Leiden. —— (2001b). Death on the Nile: Disease and the Demography of Roman Egypt (Mnemosyne Suppl. 228). Leiden. —— (2006). ‘StratiWcation, deprivation and the quality of life’, in M. Atkins and R. G. Osborne (eds.), Poverty in the Roman World. Cambridge, 40–59. Sharp, M. (1998). ‘The Food Supply in Roman Egypt’. D.Phil. thesis, Oxford. —— (1999). ‘The village of Theadelphia in the Fayyum: Land and population in the second century’, in A. K. Bowman and E. Rogan (eds.), Agriculture in Egypt from Pharaonic to Modern Times. London, 159–72. —— (2007). ‘The food supply’, in A. K. Bowman et al. (eds.), Oxyrhynchus, a City and its Texts. London. Sidebotham, S. E., and Wendrich, W. (eds. 1996). Berenike 1995: Preliminary Report of the 1995 Excavations at Berenike (Egyptian Red Sea Coast) and the Survey of the Eastern Desert. Leiden. —— —— (eds. 1998). Berenike 1996: Preliminary Report of the 1996 Excavations at Berenike (Egyptian Red Sea Coast) and the Survey of the Eastern Desert. Leiden. —— —— (eds. 1999). Berenike 1997: Preliminary Report of the 1997 Excavations at Berenike (Egyptian Red Sea Coast) and the Survey of the Eastern Desert, including Excavations at Shenshef. Leiden. —— —— (eds. 2000). Berenike 1998: Preliminary Report of the 1997 Excavations at Berenike (Egyptian Red Sea Coast) and the Survey of the Eastern Desert, including Excavations at Wadi Kalat. Leiden. Storey, D. B. (1999). ‘Archaeology and Roman society: Interpreting textual and archaeological data’, Journal of Archaeological Research 7.3: 203–48. Tacoma, L. (2006). Fragile Hierarchies: The Urban Elites of Third-Century Roman Egypt. Leiden and Boston. Temin, P. (2002). ‘Price behaviour in ancient Babylon’, Explorations in Economic History 39: 46–60. —— (2006). ‘Estimating the GDP of the Early Roman Empire’, in E. Lo Cascio (ed.), Innovazione tecnica e progresso economico nel mondo romano. Bari, 31–54. Toepel, L. R. (1973). Studies in the Administrative and Economic History of Tebtunis in the First Century ad. Diss., Duke University. van der Spek, R. J. (2000). ‘The eVect of war on the prices of barley and agricultural land in Hellenistic Babylonia’, in J. Andreau, P. Briant, and R. Descat (eds.), La guerre dans les ´economies antiques (Entretiens d’arche´ologie et d’histoire 5). St-Bertrand-de-Comminges, 293–313.

204

Alan Bowman

van der Spek, R. J. (2005). ‘How to measure prosperity? The case of Hellenistic Babylon’, in R. Descat (ed.), Approches de l’e´conomie helle´nistique (Entretiens d’arche´ologie et d’histoire 7). St-Bertrand-de-Comminges, 287–310. —— (2007). ‘Commodity prices in Babylon 385–61 bc’, http://www.iisg.nl/ hpw/data.php#babylon, http://www.iisg.nl/hpw/babylon.php (accessed 8 Oct. 2007). van Minnen, P. (1994). ‘House-to-house enquiries: An interdisciplinary approach to Roman Karanis’, ZPE 100: 227–51. —— (1997). Roman Hermopolis: A Study of the Social and Economic History of a Roman Town in the First Four Centuries ad. Doctoral thesis, Katholieke Universiteit Leuven. —— (2000). ‘Agriculture and the taxes-and-trade model in Roman Egypt’, ZPE 133: 205–20. —— (2002). ‘Hermopolis in the crisis of the Roman empire’, in W. Jongman and M. Kleiwegt (eds.), After the Past: Essays in Ancient History in Honour of H. W. Pleket. Leiden, 285–304. Whitcomb, D. S., and Johnson, J. H. (1982). Quseir al-Qadim 1980: Preliminary Report. Malibu. Wickham, C. (2005). Framing the Early Middle Ages: Europe and the Mediterranean, 400–800. Oxford. Wilson, P. (2007). ‘The Western Delta Regional Survey’, http://www.dur.ac. uk/penelope.wilson/Delta/Survey.html (accessed 8 Oct. 2007).

8 Response to Alan Bowman Roger Bagnall

It is hard to be sure whether this survey of the bases of our quantitative knowledge of the agricultural economy of Roman Egypt should be regarded as optimistic or pessimistic. Bowman sides with those who, looking from outside, think that actually we have quite a lot of information; but his analysis of the available data for large-scale parameters like population and land on the whole emphasizes more the gaps in our knowledge and the squishiness of some of the numbers we frequently use than it stresses their utility. Those two points of view can of course be reconciled by falling back on micro-level analyses and case studies, as the latter part of the chapter seems to suggest. We might, for example, understand in Wne detail the balance sheet and income statement of an individual without having any means of escaping microhistory in favour of the quantiWcation of a village, nome, province, or empire. But it is hard to see that this would constitute a positive outcome for a project aiming at quantiWcation. In my view we do not need to surrender to a negative assessment of our prospects. The principal reason is that I do not see the goal as a set of numbers on paper but as an interactive model in which every user is free to choose diVerent values for particular variables, and which is created with as plausible a set of values as we can deduce. This model then allows us to examine the consequences of changing the value of a particular variable. If we adopt a particular percentage of urbanization, and let that drive the calculations, do we arrive at unacceptable results somewhere else? Some values for some variables will deWnitely produce results capable of falsiWcation, and at that

206

Roger Bagnall

point the model will begin to become bounded in some dimensions. Relationships between variables are likely to be more controllable than some of the macroeconomic values, but that does not mean that the latter can be invented freely. An important element in making such a model cogent is the process of inventorying our ignorance as assiduously as our sources. Production requires inputs. Bowman recognizes this in several instances, most importantly perhaps transportation. We can take this process into areas for which the papyri are lamentably lacking in information, of which the most salient is probably energy deriving from sources other than human and animal power. These gaps cannot be Wlled with evidence, at least not at present, but they can be modelled. How much energy did it take to produce an amphora or a cooking pot? How much labour was required to make a sack? How many new items of this sort did an average person need per year? We will not arrive at precise knowledge, but we can probably build an approximation that is much better than nothing. And the process will be strongly recursive: that fuel needed animals to get it to the kiln, and the animals had to eat, and their food had to be grown by someone. Each major element in the economy needs to be traced through in this fashion. A big gap of a diVerent sort from the unreliability of macroeconomic indicators is the absence of information from the Delta, as Bowman points out. This also is not to be despaired of. There is in fact a lot of information about agriculture in the carbonized papyri from Thmouis, which have been subjected to detailed tabulation and analysis in the recent dissertation of Katherine Blouin.1 We can recover critical information like the percentage of land allocated to diVerent crops and other uses, thus helping to provide a set of values for the range of land use to help us contextualize the Wgures we have from other nomes. They are not perfect, of course, but the amount of information is likely to be surprising to anyone raised on papyrological pessimism about this region. We need to try to sketch rough borders for the Delta nomes and calculate their sizes, a project with plenty of its own hazards but again capable of being carried out within certain margins of error. 1 See Blouin 2007a; 2007b; forthcoming.

Response to Alan Bowman

207

I would suggest that it is at the nome level that we have the best chance of building a Wrst-round model, which might then be extrapolated to the entire province. Extrapolation will carry with it a signiWcant decline in the level of conWdence, naturally, but that does not mean that it will not be worthwhile. (If one accepts the estimate that the Hermopolite amounted to about 4% of all of Egypt, straight-line extrapolation is particularly easy in that case, cf. my response to Lo Cascio; no doubt wrong, too, but by how much it remains to be seen.) I tried in the article of 1992 cited by Bowman to oVer the merest sketch of what such a model for the Hermopolite nome might look like. It could be deepened signiWcantly with other information, most notably the vestis militaris codex (P. Col. IX 247) to which Bowman refers. With this list of villages for two-thirds of the nome together with tax assessments (in garments, but capable of being turned into cash with only modest approximateness) we can in eVect build a model of the entire list of villages for the nome, with their relative assessments. That is already a signiWcant step beyond the aggregate numbers I used. It can then be linked to the sixthcentury tax codex from the Hermopolite village of Temseu Skordon, admittedly outside the boundaries of this project. That codex (P.Lond.Copt. 1075), which is in preparation for publication by James Keenan, Leslie MacCoull, and me, allows an estimate of the number of taxpaying landowners in this village. One could then extrapolate the number of landowners in each of the other villages in the fourth-century codex and the model derived from it. Temseu Skordon was responsible for 3.37% of the total nome taxes in that codex; since the codex preserves about ¯ of the total for the nome, we may extrapolate from this that it was liable for about 2.25% of the total taxes of the nome at that date. We do not know if it may have changed in relative or absolute size between the fourth and sixth centuries, but in any event it should be an example of a relatively large village. By my reckoning there are 359 individual taxpayers in the Temseu Skordon codex. The last link in building the Wrst stage of the model would then be the number of persons in the population per landowner. The number of landowners in the 1992 model was derived by dividing the arable land supposed to have been owned by villagers by an average holding extrapolated from the Karanis data. That Wgure could be given much

208

Roger Bagnall

more work from other sources; it might be too high. If Temseu Skordon had 2.25% of the Hermopolite cultivable land owned by villagers, using the estimate that Bowman quotes, it would have had about 4,050 arouras. That comes to about 11.3 arouras per landowner, or somewhat less than half of the Wgure that I used in the Hermopolite model. The Wgure would be much larger if one calculated based on total arouras rather than those owned by villagers. My point is not to push for a particular Wgure here, only to say that we have materials from which to build a model. The larger point here is that attempting to quantify the economy of Roman Egypt is not best thought of as a process with an end point consisting of a set of answers. Rather, it is better conceived as an ongoing, iterative exercise in the building of an open model that can be used by many scholars experimenting with diVerent values and incorporating new evidence as it becomes available. It will have various components, including an input-output table and a set of sub-tables for diVerent activities linked to it. With all of its gaps, our evidence is certainly suYcient to allow the building of enough components of this model for the whole to be capable of excluding some answers and helping us deWne ranges for others. Agriculture is the largest part of this model, of course, but we will get wrong answers for many values if we do not build it into a model that asks what the range of the possible is for non-agricultural activities. We have far less data for most of those domains, of course, but much can be obtained from comparative evidence. One Wnal point about modern Egypt as a comparand may be worthwhile. The landscape of Egypt has changed greatly as a result of the two Aswan dams, barrages elsewhere, the introduction of perennial irrigation, and the reconWguration of the canal system. Survey work today can never hope of itself to recover the landscape of antiquity. But maps and accounts from the nineteenth century and before, including the account of the Fayum by an-Nabulsi which James Keenan has done so much to bring to our attention, can help Wll the gap. Particularly worthwhile, I think, would be attempts to do just what Bowman suggests, build up a picture of the area and nature of some of the nomes we know relatively well from documentation, starting at the toparchy level. The results can be compared with our scattered information for these districts in the documents. I would

Response to Alan Bowman

209

guess that at least for the Hermopolite, Oxyrhynchite, Arsinoite, and Herakleopolite nomes some real progress might be made in this fashion, providing a test of the kind of model-building that I have suggested. Bibliography Blouin, K. (2007a). Homme et milieu dans le nome mende´sien a` l’e´poque romaine (1er au 6e s.). Doctoral thesis, Que´bec Nice. —— (2007b). ‘De Mende`s a` Thmouis (delta du Nil, E´gypte): Hydrologie mobile, socie´te´ mobile?’, in E. Hermon (ed.), L’eau comme patrimoine—de la Me´diterrane´e a` l’Ame´rique du Nord. Que´bec, 107–28. —— (2007). ‘Environnement et Wsc dans le nome mende´sien a` l’e´poque romaine: Re´alite´s et enjeux de la diversiWcation’, BASP 44: 135–66.

This page intentionally left blank

Part IV Trade

This page intentionally left blank

9 Approaches to Quantifying Roman Trade Andrew Wilson

INTRODU C TION What is the point of trying to quantify trading activity in the Roman world? An understanding of the scale and the nature of Roman trade is critical for understanding Roman urbanization, economies of scale, and investment opportunities—was it the intensity and ubiquity of long-distance trade that made it possible to support the urban centres of the Roman empire, feeding them from outside their immediate hinterlands? Did long-distance trade operate on a scale suYcient to increase the overall size of markets in certain goods or commodities, enabling specialization and division of labour, and thus Smithian growth? What were the relationships between local, regional, and long-distance trade subsystems? Who proWted from trade, and to what levels of society was trade an important part of their incomes? What was the state’s role in trade, both in promoting and protecting trading activity, and in beneWting from it, whether from more stable prices, or taxes on transactions? But quantifying trade in what? As soon as one starts to probe the question of trade, one has to break it down into trade in speciWc products, to and from speciWc areas; and it rapidly becomes apparent that we can only track very small and selected parts of the total. We have to piece together a few speciWc areas of the whole, and then use those to suggest a larger picture, while remaining conscious that most of the jigsaw is missing.

214

Andrew Wilson

Archaeological data recovered and synthesized over the last thirty years or so have demonstrated that long-distance trade in the Roman world was normal and common, and involved not merely luxury goods (silks, spices, marble) but a range of staple foods and other relatively low-cost commodities—wine, oil, salted Wsh, grain, table pottery, glass, millstones, etc. The sheer quantity of material and the nature of the contexts in which much of it is found usually suggest that these goods were moving as trade rather than personal equipment, gift exchange, or even imperial shipments.1 But it is remarkably diYcult to move from this impressionistic view of vibrant, largescale, long-distance trade in even humdrum goods and artefacts, to any quantiWed analysis. Given that it is archaeological evidence which has largely brought about the shift in perception of Roman trade, we might expect that the archaeological material would lend itself to quantiWcation of trade data, either to demonstrate chronological developments or regional variation. After all, there are hundreds of published excavations of Roman sites with hundreds of thousands of pottery fragments, and many other artefacts, and much of that pottery was not made at the sites where it was found. Surely we must be able to use these data to quantify trends? But this is not as easy as it might seem, for several reasons, some to do with biases in preservation, excavation, and publication, and others connected with the problems of synthesizing the published material. What I aim to do in this chapter is to try to work through some of the main biases inherent in the archaeological record, discuss ways of dealing with them, and suggest some ways in which we can quantify trends in the trade of particular goods over time. The Wrst, and obvious, point is that we can only trace things that survive. Thus trade in people, grain, lentils, beans and pulses, fruit, Xowers, hides, furs, textiles, spices, papyrus, basketry, etc. will remain 1 Cf. Whittaker 1985, 1987, and 1989 for a somewhat diVerent view. While the army on the frontiers was clearly a major market, the Vindolanda Tablets (Bowman 20032) show that civilian merchants were involved in their supply; and despite Whittaker’s arguments about elites transferring produce between their diVerent estates, outside the market, such redistribution can only have accounted for a fraction of the distribution attested by archaeological Wnds—especially those from urban and lower-status sites—nor can it account for amphorae containing Wsh products.

Approaches to Quantifying Roman Trade

215

largely invisible archaeologically. A particular value of documentary evidence in antiquity is that it reminds us of the large-scale trade in such archaeologically invisible goods, such as the 7,000 modii of Alexandrian wheat and 4,000 modii of chickpeas, beans, and lentils given as security against a loan of 10,000 sestertii made at Puteoli in ad 37 (TPSulp. 51 and 52); or civilians trading in hides between Catterick and Vindolanda; or the trade and sale of slaves.2 Archaeologically, we are forced to concentrate on durable artefacts, and in particular those for which we can determine a provenance. In practice, this largely means objects of pottery or stone—Wne table pottery, cooking and common wares, transport amphorae, stone sculpture, architectural stone (usually marble), and millstones (usually in volcanic lavas). Occasionally provenance can be suggested for glasswares and some metal artefacts, though this usually requires costly scientiWc analysis and the number of artefacts for which this has been done therefore remains limited. Long-distance trade is easier to track than short-distance trade. The archaeological evidence for trade frequently relies on petrology to pinpoint the geological regions of origin of ceramic or stone goods. Clearly, one can only discriminate between local and longer distance trade if the products move outside a region of geological distinctiveness (which may be quite extensive). Of course, in terms of economic transactions, short-distance trade and exchange may be very signiWcant—for example, exchanges between a town and its territory—and may well be more important than long-distance trade. An example of regional trade is wine production in Roman Egypt. We know from papyrus contracts for the lease of potteries making wine amphorae, from vast mounds of amphora wasters at kiln sites, and from the Wnds of numerous wine-making installations, that wine production in Egypt was organized on a massive scale in the Roman and late Roman periods.3 Yet the export of Egyptian wine amphorae was on nothing like such a scale, and some types are rarely found outside Egypt. The implication is that most Egyptian wine was 2 Murecine tablets: Camodeca 1999. Vindolanda tablets: Bowman 20032: 136–8. Documentation of traYc in slaves: Tomlin 2003. 3 Cockle 1981.

216

Andrew Wilson

Fig. 9.1. Olive and wine presses in the suburbium of Rome (Corrente 1985, Wg. 85)

consumed within the province—at Alexandria, the metropoleis (nome capitals), or in smaller towns and villages. Similarly, in the region around Rome, the distribution of oil and wine presses in an area with apparently no local amphora production strongly suggests the production of wine and oil for the urban and suburban markets of Rome, to which they were shipped in skins (Fig. 9.1).4 In this case, one could only measure the trends in production infrastructure—the construction and use of oil and wine presses. A similar pattern obtains for Cyrenaica, where in the late Roman period some sites have numerous olive presses (c.50 at the village of Lamluda, for instance, east of Cyrene), but there is no

4 Corrente (1985 [1986]), 113 Wg. 85, cf. Bellini and Rea (1985 [1986]); Panella and Tchernia 1994: 147–8.

Approaches to Quantifying Roman Trade

217

corresponding amphora production or export, implying local or regional consumption within Cyrenaica.5 Why is it important to distinguish between long-distance trade and regional trade? One answer is the impact on imperial Wnances. Where we can demonstrate long-distance trade across the imperial frontiers, or between one province and another, we are investigating activity which contributed to state revenues, via the customs tax of 25% across the empire’s frontiers (at least on the eastern frontiers) and the 5% portoria between provinces. State cargoes were exempt from the portoria, and this exemption will have included grain cargoes and perhaps also subsidized olive oil import to Rome even before oil’s inclusion in the annona. Nevertheless, the trade in other goods, as attested for example by the widespread and abundant distribution of wine and Wsh product amphorae, suggests that inter-provincial trade may have made a considerable contribution to state revenues. External trade also seems to have provided a not negligible source of revenue. The Muziris papyrus refers to a cargo of spices and luxury goods from India, valued before tax at over 9 million sestertii, on which a customs duty of 25% was payable.6 Strabo (Geography 4.5.3), writing in the reign of Augustus, said that customs duties on trade with pre-conquest Britain exceeded the likely revenues of imposing a tribute and the costs of collecting it. What proportion of the traded total do the goods we have found represent? There is simply no way of extrapolating from the recovered assemblage to the original total. This means that we cannot hope to recover full Wgures; we cannot produce a series of data points representing absolute quantities of a commodity imported to a region over time. We can only hope to compare proportions between diVerent assemblages. But, again, proportions of what? We can, for example, express the (recovered) imported African Red Slip ware 5 Personal observation. 6 Rathbone 2001; cargo valuation after payment of 25% duty equivalent to HS 6,926,852 (p. 48), thus HS 9,235,803 before tax. The state’s take was therefore c. HS 2,308,951 on this one cargo. More generally, for exports of wine to India and import of textiles, cf. Will 2004, Williams 2004, Wild and Wild 2005. But the issue of eastern trade is more complex, involving also interrelations with regional trade networks involving incense from Arabia and Ethiopia—cf. Peacock and Williams 2007.

218

Andrew Wilson

(ARS) at a site as a proportion of the total recovered tablewares at the site. But to do this we need an assemblage where we can distinguish the tablewares from the cooking wares in the sherd counts of the publication, as comparing ARS with cookwares is not comparing like with like. I will return to this problem later.

MA RBL E The trade in stone, especially marble, oVers potential for analysis: architectural marble, sculptures, and sarcophagi oVer a wealth of data which can be provenanced, tracked, and quantiWed. This will be explored within the context of our project by Benjamin Russell, in his doctoral thesis on ‘Sculpted stone in the Roman economy, 100 bc—ad 300’. Here I note merely that besides quantities, we can also detect qualitative and organizational trends—in the second century, the increasing standardization of certain sarcophagus types, and even portrait statues, shipped part-Wnished from the quarries with Wnal detailing to be added at the receiving centre.7 This matches a trend in architectural marble where the quarries moved from a quarry-to-order system to a quarry-to-stock system in the second century, as overall demand increased. The other main category of stone objects that could be provenanced and used for analysis of trade is millstones. There is plentiful evidence of long-distance and regional trade in lava millstones— querns in Niedermendig lava from Mayen in Germany are widely distributed on military sites in Britain; lava querns and animal mills are common imports on Mediterranean urban sites distant from volcanic regions. Masses of published data on hand-querns and millstones in archaeological reports await synthesis and analysis, although this will require a truly enormous amount of work: almost every museum in western Europe must have its collection of Roman period hand-querns, and occasionally powered millstones. David Peacock has done a preliminary sketch of the Mediterranean trade in lava millstones while Ruth ShaVrey has studied the regional 7 Wilson 2008: 402–5.

Approaches to Quantifying Roman Trade

219

distribution of Lodsworth greensand, Old Red Sandstone, and Mayen lava querns within Britain.8 But to achieve this level of analysis for wider areas is beyond the scope of our project: I mention it here chieXy to Xag up an opportunity for future research.

SHIP WRE CKS Shipwrecks are an immensely important source of data, but need to be used with great caution and an awareness of the numerous biases which have created the present record. A. J. Parker’s pioneering synthesis of ancient shipwrecks has been used by a number of historians to illustrate the growth and shrinkage in long-distance trade during antiquity. Parker himself was careful to point out some of the limits and biases inherent in the dataset, which he explored in a series of articles, and was properly cautious about the conclusions he drew, but the data have often been employed rather too uncritically by others.9 Parker’s graph of shipwrecks by century (Fig. 9.2) shows an impressive peak in the last two centuries bc and the Wrst two centuries ad, after which third-century wrecks fall to half the number of second-century wrecks. Nevertheless, the implications for trading activity are not as clear-cut as the graph of chronological distribution of wrecks may at Wrst sight suggest. To a certain extent, this is partly a graph of amphora usage.10 Wrecks are spotted under water usually by the mound formed on the seabed by their cargo; this protects the parts of the hull underneath from breaking up. The shipwrecks we know about therefore are mostly amphorae cargoes and loads of architectural marble, as these survive on the seabed as a protective 8 Peacock 1980; ShaVrey 1998; 2003. 9 Parker 1990a; 1990b; 1992a; 1992b; 1996. Use of Parker’s graph: Hopkins 1980: 105–6 (using an early version); Gibbins 2001, Wg. 10; de Callatay¨ 2005; Davies 2006: 84–5; Jongman 2007: 188; Morley 2007a: 572–3; 2007b: 98 (with caveats about underrepresentation of African trade). Cf. Gibbins 2001: 273–83 on the strengths and weaknesses of wrecks as evidence for trade. Whittaker 1989 pointed out many of the problems in using these shipwreck data to chart changes in trade over time. 10 Parker 1990a: 342; 1992a: 8; Horden and Purcell 2000: 371–5.

220

Andrew Wilson

200 180 160 140 120 100 80 60 40 20

2500

2000

1500

1000

500

0

500

1000

1500

BC AD

Fig. 9.2. Mediterranean shipwrecks by century, using mid-points of date ranges (after Parker 1992a, Wg. 3)

mound in the way that other cargoes, including barrels, do not. Warships, grain freighters, and ships carrying perishable cargoes are rarely found. By the early medieval period the amphora had been largely replaced, at least in the west, by the barrel.11 Does the decline in the number of known wrecks from the second to seventh centuries ad therefore reXect an absolute decline in the number of trading voyages, or is it at least partly explicable by a move from amphorae to wooden barrels as the preferred container for trading liquids?12 11 Sealey 1985: 125–6 (barrels used for wine trade from late Wrst century bc); Panella and Tchernia 1994; Tchernia 2006 on increasing use of barrels for wine in Roman Italy from the third century ad onwards, with the shift from amphorae to barrels largely complete in the western Mediterranean by the Wfth or sixth centuries ad. On the archaeology of the barrel generally, Marlie`re 2002; also E´tienne and Mayet 2000: 21–58 for Lusitania. 12 Against this interpretation it might be objected that the number of barrels known from excavations on land peaks in the Wrst century ad and declines thereafter

Approaches to Quantifying Roman Trade

221

The barrel hypothesis might receive some support from Parker’s regional breakdown,13 which shows a greater number of shipwrecks continuing through the late antique and early medieval period in the eastern Mediterranean, where amphorae remained in use longer, than in the central and western Mediterranean. Nevertheless, this regional breakdown could equally, or better, be interpreted as suggesting longer-lived and more robust trade in the Byzantine world than in the early medieval West. While the rise in numbers of wrecks in the Hellenistic/early Roman periods, before the usage of the barrel became widespread, must reXect a genuine increase in trading activity, there remains considerable uncertainty about how one interprets the decline in wreck numbers in the mid- to late Roman period. The regional analysis also highlights some factors underlying quirks of the main graph. It has always seemed a little odd that the overall distribution of Parker’s Mediterranean shipwrecks peaks in the Wrst century bc rather than, say, the Wrst or second century ad, although the numbers are not vastly diVerent. But Parker’s graph for wrecks oV the coast of southern France shows that this area accounts for quite a number of the second- and Wrst-century bc wrecks;14 many of them carried cargoes of Italian wine in Dressel 1 amphorae. The wine trade from Italy to southern Gaul in the late Republic accounts for an extraordinary number of shipwrecks. The lower numbers of wrecks carrying Dressel 2–4 wine amphorae of the Wrst century ad from Italy to France therefore show the eVects of import replacement, as the growth of viticulture in southern Gaul under the Empire enabled much of the Gallic market to be supplied from more local sources. This is a reminder that an apparent decline in certain indicators of trade—here, shipwrecks—may nevertheless be consistent with a model of economic growth. Regional variations are also complicated by the fact that wrecks in areas where underwater archaeology and leisure diving are well developed (e.g. Spain, southern France, Italy, Croatia) will be over-represented by comparison until at least the Wfth century. Yet this comparison itself may be biased because most of these barrels are preserved as linings for wells; and the drop in numbers in the early medieval period may be due to the cessation of this practice, rather than to the number of barrels in use: Marlie`re 2002: 174–5; cf. Desbat 1991. Once again, secondary factors may skew attempts at quantiWcation. 13 Parker 1992a, Wg. 7. 14 Parker 1992a, Wg. 7.

222

Andrew Wilson

with Greece or the entire southern littoral of the Mediterranean, where very little underwater archaeology has been carried out or published. Regional analyses of sub-sets of the main database, for what Gibbins terms ‘key sample areas’ where the seabed has been relatively intensively explored,15 are very useful in highlighting possible explanations for some of the features of the overall graph, and the availability of GIS tools now greatly facilitates such analyses. Importantly, the shape of the graph is a function of the method used to deal with long date ranges. Parker’s graph of Mediterranean shipwrecks over time uses the mid-points of the date range for each shipwreck, but this does not deal well with wrecks whose chronological bracket spans several centuries.16 (His Wg. 5, which graphs just the well-dated wrecks in the western Mediterranean, is not routinely cited by ancient historians.) The numerous wrecks dated simply as ‘Roman’ (considered as 150 bc to ad 400, with a mid point of ad 125) constitute many of the wrecks in the second-century ad column of the histogram. Such long-dated wrecks must instead be distributed across the histogram, by converting their date range to a probability that the wreck sank in a particular year, and then accumulating that annual probability for the number of years within the date range which fall within each chronological period on the graph.17 Thus, for a wreck dated ad 125–75, there is a probability of 0.02 that it sank in any individual year within that range, a probability of 0.5 that it sank between ad 125 and ad 150, and a probability of 1 that it sank in the second century ad. Figures 9.3–9.6 display the data from Parker (1992a) using this method, but graphing the wrecks by century, by half-century, by 25-year period, and by 20-year period. It is important to stress that exactly the same data and probability calculations are used for each graph; what changes is the length of the chronological bracket for each bar of the histogram. Several diVerences from Parker’s mid-point graph are apparent in the probability graphs. First, in the graph of wrecks by century (Fig. 9.3), the absolute peak now comes in the Wrst century ad, not 15 Gibbins 2001: 278–81. 16 Martin-Kilcher 1998, Wg. 10; 2004: 266 n. 1. On the problems of dating wrecks, see Tchernia 1990. 17 Comparable methods of graphing data are used in the analysis of ARS by Fentress and Perkins 1988, and Fentress et al. 2004, discussed below.

Approaches to Quantifying Roman Trade

223

220 200

Wrecks per century

180 160 140 120 100 80 60 40 20 −1500 −1400 −1300 −1200 −1100 −1000 −900 −800 −700 −600 −500 −400 −300 −200 −100 0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500

0

120 110 100 90 80 70 60 50 40 30 20 10 0

−1500 −1400 −1300 −1200 −1100 −1000 −900 −800 −700 −600 −500 −400 −300 −200 −100 0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500

Wrecks per 50-year period

Fig. 9.3. Mediterranean shipwrecks by century, using probability per annum (data from Parker 1992a)

Fig. 9.4. Mediterranean shipwrecks by half-century, using probability per annum (data from Parker 1992a)

224

Andrew Wilson

the Wrst century bc. The second-century ad column is now much shorter—less than half the number of Wrst-century ad wrecks, and about the same number as the third-century wrecks, largely as a result of smoothing out the generically ‘Roman’ wrecks. But analysis of the other graphs, with progressively Wner resolution, throws up some other patterns and also shows that some elements of the picture that emerges may be dependent on the categorization by time period chosen. The half-century graph (Fig. 9.4) shows high numbers of wrecks in the later second century bc and the early Wrst century bc, constituting a secondary peak; the Wrst-century ad peak is now revealed to be weighted more towards the Wrst half of that century. A very slight, and not sustained, recovery in wreck numbers is perhaps discernible in the early fourth century ad. Figures 9.5–9.6, 25-year and 20-year periods, show Wner resolution and display diVerences dependent on the length of period selected: the 25-year graph now shows an absolute peak in the Wrst quarter of the Wrst century bc, slightly exceeding the Wrst two quarters of the Wrst century ad; in the 20-year graph this peak is still present but only slightly exceeds that of the Wrst 20 years of the Wrst century ad. The peak in the Wrst century bc is preceded also by a steep increase in wrecks at the end of the second century bc, and is very clearly the result of the late Republican wine trade between Italy and Gaul. Both these graphs show a marked dip in the early Augustan period (25–1 bc or 20–1 bc), which is perhaps unexpected,18 and they also suggest a slight recovery in Wgures in the early fourth century ad after the low numbers of the late second and third century. The marked drop between the Wrst and second centuries ad appears surprising, and is not easily explained. The numbers are fairly constant throughout the second century, and the drop occurs too early to be a result of the Antonine plague. More promising is the idea that this drop may reXect the eVects of economic growth in the provinces, leading to import replacement, but there are diYculties also with this; the provinces were exporting to Italy, and to each other, at this time. Are we in fact seeing the impact of the widespread use of the barrel in the Mediterranean already by the second century ad?19 18 Although it appears also in Parker’s graph of Roman wrecks in the western Mediterranean (Parker 1992a, Wg. 5), which is based on well-dated wrecks. 19 It was certainly common in Gaul and north-west Europe from the Wrst century ad: Desbat 1991; Marlie`re 2002.

−1500 −1420 −1340 −1260 −1180 −1100 −1020 −940 −860 −780 −700 −620 −540 −460 −380 −300 −220 −140 −60 20 100 180 260 340 420 500 580 660 740 820 900 980 1060 1140 1220 1300 1380 1460

Wrecks per 20- year period

−1500 −1400 −1300 −1200 −1100 −1000 −900 −800 −700 −600 −500 −400 −300 −200 −100 0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500

Wrecks 25-year period

Approaches to Quantifying Roman Trade 225

60

55

50

45

40

35

30

25

20

15

10

5

0

Fig. 9.5. Mediterranean shipwrecks by quarter-century, using probability per annum (data from Parker 1992a)

55

50

45

40

35

30

25

20

15

10

5

0

Fig. 9.6. Mediterranean shipwrecks by 20-year period, using probability per annum (data from Parker 1992a)

226

Andrew Wilson

If some of the Wrst-century ad peak is accounted for by wrecks carrying Spanish wine in Pascual 1 and Dressel 2–4 amphorae, does the disappearance of these containers signal the end of large-scale Spanish wine exports, or a shift to other containers? The amphorae from Les Alle´es de Tourny at Bordeaux show a large import of wine from Italy and Spain in the Wrst century bc, and a lesser import from Spain in the Wrst half of the Wrst century ad.20 Thereafter, there are few wine amphorae of any type, local or imported, although occupation lasted until the third century ad; rather than assume that the inhabitants of Bordeaux stopped drinking wine, we must either assume provision from local sources in skins, or import in barrels. In Lusitania, the number of cupa-type funerary monuments, which depict barrels, peaks in the second century, and implies a local production of wine sold in barrels, though not necessarily its export.21 But even explanations appealing to barrels and import replacement do not cope fully with the high levels of trade in amphora-borne olive oil and Wsh products from Spain and Africa, as attested from excavations on land sites. We know that oil and grain exports to Rome were maintained, and the widespread distribution of ARS suggests high levels of trade in the second and third centuries ad. Are African wrecks under-represented because of the dearth of underwater archaeology along the North African littoral, so that an increase in African exports after the Wrst century ad may not be fully reXected in the wreck evidence? However one looks at it, there is a mismatch between the wreck data and the impression gained from assemblages from land sites, which invites further research. While the left-hand half of the graphs, showing an increase in wreck numbers from the Hellenistic period to the Wrst century ad, may reasonably be taken as showing an increase in maritime trading activity, interpretation of the right-hand half of the graphs is not so straightforward. In the Wfteen years since Parker’s synthesis was published, several hundred more ancient wrecks have been reported, and even if the rate of increase in information may now have slowed somewhat, the database could be substantially expanded.22 Progress in this direction 20 E´tienne and Mayet 2000: 232. 21 E´tienne and Mayet 2000: 53–4. 22 Gibbins 2001 estimates that some 1,500 Mediterranean shipwrecks dated before ad 1500 were known by the year 2000 (compared with 1,189 in Parker’s 1992 catalogue).

Approaches to Quantifying Roman Trade

227

has been made by Julia Strauss for her Ph.D. thesis on trade in the eastern Mediterranean, and I am indebted to her for the information she has shared.23 A preliminary update of Parker’s catalogue, based on Strauss’s work with further additions, suggests that for 1,356 Mediterranean and Black Sea wrecks dated before ad 1500 the centre of gravity shifts even further towards the Wrst century ad. If one disaggregates these into marble and amphora wrecks the peak in marble wrecks occurs in the second century ad, as one would expect from the widespread provincial use of marble evident in the architectural record. An analysis of cargo size over time suggests that in the Roman period the largest ships were much larger than in previous or later centuries. There is a marked distinction between the period 150 bc–ad 400, in which there are several wrecks of between 100 and 500 tons, and the periods before and after, which lack wrecks over 100 tons.24 Within the Roman period, no signiWcant trend in cargo sizes is discernible; there are both small and large wrecks throughout.25 Some of the very largest known wrecks are associated with the massive trade in wine between Italy and southern Gaul in the late Republic. Perhaps the best known are the wrecks from the Madrague de Giens, a 290–390 tonne ship which sank between 60 and 50 bc while carrying some 5,800–7,800 Dressel 1 wine amphorae, and assorted table pottery; and the Albenga wreck, a ship which sank around 100–90 bc. The Albenga ship, estimated at 450–500 tonnes, carried some 10,000 Dressel 1 wine amphorae; since each amphora carried c.26 litres of wine, the cargo was some 260,000 litres of wine. To put this into some sort of perspective, that is some 347,000 modern wine bottles, and with a cheap £2.99 bottle the cargo value would exceed a million pounds in today’s prices; that doubtless greatly understates the relative value in antiquity as the characteristics of the Republican wine market suggest that Gaulish consumers were willing to pay relatively high prices for Italian wine. Large merchant vessels were not only a phenomenon of the late Republican wine trade. Indeed, although we have evidence from shipwrecks for large ships of between 300 and 500 tonnes in the Wrst 23 Strauss 2006. 24 Parker 1990a: 340–2, Wg. 7; 1992b, Wg. 1. 25 Parker 1990a: 341; 1992b: 89–90.

228

Andrew Wilson

century bc and Wrst century ad, carrying amphorae, and again in the second century ad carrying cargoes of marble, these were not the largest ancient ships.26 The grain freighters running between Rome and Alexandria, and Rome and Carthage, are likely to have been larger—all the more so after Rome’s harbour facilities were improved under Claudius and then Trajan by the construction of the harbour at Portus. This is actually a case where the institutional framework should have encouraged the construction of larger ships—both by making the necessary harbour infrastructure available, and by the various tax incentives oVered by certain emperors to those who placed ships at the disposal of the annona.27 An estimate of 1,000–1,300 tonnes for large grain freighters is accepted by some.28 Yet the perishable nature of grain, carried in sacks and thus creating no durable mound to protect the hull, means that no wreck of a large grain freighter has been found.29 In other words, really large ships only become visible to us under particular circumstances at the peak of trading certain durable cargoes—the late Republican wine trade, or the imperial marble trade of the second century ad—and this may not reXect overall trends in the sizes of ships, or intensity of trade as a whole. Other variables may be at work aVecting recovery here. The tendency is to assume that ships had an equal chance of sinking at all periods throughout history, and that variations in the graphs therefore reXect variations in the number of sailings of which wrecks should 26 Pomey and Tchernia 1978. The large obelisks moved from Egypt to Rome in the imperial period imply very large ships—the two obelisks moved in 10 bc (now in Piazza del Popolo and Piazza Montecitorio) weigh 234 and 239 tonnes; the Vatican obelisk, moved by Caligula in ad 37, weighs 336 tonnes, and the Lateran obelisk, shipped in ad 357, weighs 462 tonnes. Pliny says that the ship transporting the Vatican obelisk also carried 120,000 modii of lentils as ballast (NH 16.76.201); its total tonnage has been estimated at 1,100 tonnes. The other loads will also have needed stabilizing ballast and imply ships of substantially greater tonnages than the weights of the obelisks themselves. 27 Gaius, Inst. 1.32c: Claudius granted privileges to those who built grain ships of 10,000 modii capacity (69 tonnes). Digest 50.5.3 (relating to late second century ad): exemption from public services for shipowners who make available for the annona a ship of 50,000 modii (¼ 345 tonnes) or several vessels of 10,000 modii. Casson 1971: 171–2 n. 23. 28 Casson 1971: 172–3; Gould 2000: 145. 29 A cargo of corn (Triticum sp.) dated c. ad 600–625 was found in the Saint Gervais B wreck, a smallish ship of 41–50 tonnes. Some corn was found in Les Laurons B (ad 175–200), though this also carried amphorae. Parker 1992a, nos. 1001 and 578.

Approaches to Quantifying Roman Trade

229

represent a constant proportion. But this ignores the possibility of technological development and the better sailing abilities of larger ships; one would expect larger ships to cope better with heavy seas than smaller ships, so they may have sunk less frequently. Alternatively, larger ships may more regularly have sailed the open water routes, and if they sank, did so in deep waters beyond the reach of scuba divers. In other words, sailing activity by larger ships may be under-represented in the record of identiWed wrecks, and if there was a greater use of larger ships under the empire, as suggested by the sources for the grain trade, the evidence of marble wrecks, and the infrastructural improvements to harbours, this could seriously distort the overall picture. For comparison, the largest merchant ships known from medieval Europe and from Song dynasty China in the thirteenth century ad had a displacement of some 375 tonnes.30 The troop transports commissioned from the shipyards of Genoa, Venice, and Marseilles by Louis IX of France for his Crusades in 1248–51 and 1270, which had a displacement of up to 800 tonnes,31 seem to have been exceptional vessels for the period, made possible only by the royal underwriting of their construction. The regular construction of merchant vessels exceeding 500 tonnes probably did not occur again until the fourteenth or Wfteenth centuries. While smaller ships of 25 to 100 tonnes were common at all periods, at the upper end of the range the economies of scale in maritime transport oVered by the large merchant ships of the Roman world between 200 bc and ad 300 were not matched again for another 1,000 years.

A MP HO RA E Amphorae, as the transport containers for a variety of goods—wine, oil, Wsh products, but also nuts, pitch, alum, preserved fruits, sometimes grain etc.—ought to be another primary resource for the quantiWcation of ancient trade. Nevertheless, they have their own problems which need to be understood; they are another form of proxy, since we are usually more interested in what they carried than 30 Gould 2000: 193.

31 Gould 2000: 190.

230

Andrew Wilson

80

% of total amphorae

70 60

Italian wine amphorae Provincial wine amphorae Amphorae with other contents

50 40 30 20 10 0 −50

0

50

100 150 200 250 300 350 400 450 500 550 600 Year

Fig. 9.7. Percentage of Italian wine amphorae, provincial wine amphorae, and amphorae for other products, in selected contexts from Rome and Ostia, 50 bc to ad 600 (Panella and Tchernia 1994, Graphique 2)

in the containers themselves. We need to know the average capacity of a form, and its typical contents.32 A tiny rim sherd of a Dressel 1 represents 27 litres of Italian wine; so does a complete amphora. But a Late Roman 3 amphora held only between 3 and 6 litres of wine, and a Tripolitanian amphora some 60 litres of olive oil. Clementina Panella and Andre´ Tchernia have presented graphs of imports to Ostia and Rome, showing the proportions of amphorae carrying Italian wines, provincial wines, and other products as a percentage of the total amphora assemblage (Figs. 9.7–9.8).33 These graphs show that from the Wrst century ad onwards the proportion of non-Italian wine imports rises. Both Italian and non-Italian wine as a proportion of total amphora imports drop sharply after ad 250. Panella and Tchernia are of course well aware of the shortcomings of graphing numbers of amphorae, and use the graphs as a starting point for further discussion of other factors aVecting the total picture. They note that the decline in amphora-borne wine imports 32 Cf. e.g. Sealey 1985, passim; Ejstrud 2005. 33 Panella and Tchernia 1994.

Approaches to Quantifying Roman Trade

231

50 45 % of total amphorae

40 35 30 25 20 15 10 5 0 −50

0

50

100 150 200 250 300 350 400 450 500 550 600 Year Tyrrhenian

Adriatic

Bruttium

Provincial wines

Central Italy

Fig. 9.8. Percentage of Italian wine amphorae from different regions, and provincial wine amphorae, in selected contexts from Rome and Ostia, 50 bc to ad 600 (Panella and Tchernia 1994, Graphique 3)

is doubtless even steeper than the graph (Fig. 9.8) shows, for two reasons. One is that the later Roman wine amphorae were much smaller than their early imperial counterparts, so that the amount of wine represented by the later part of the wine lines is actually much smaller than the percentage of amphora counts suggests. The slight rise again in the Wfth century needs to be balanced by the realization that Rome’s population declined drastically after 410, and so the total of demand, and therefore presumably supply, must have dropped massively. If one is trying to imagine quantities rather than the proportions which the graph actually shows, one has mentally to compress the vertical axis in the later part of the graph. Importantly, Panella and Tchernia see in the sharp decline in amphora-borne wine imports after 250 evidence for the increasing use of barrels. Count-based graphs such as these are therefore useful for suggesting trends but their appreciation has to be nuanced by awareness of the shortcomings of using data solely on amphora counts.

232

Andrew Wilson

Re-analysis of the data using average volumes for diVerent types of amphora might be useful, throwing the barrel question into sharper relief. We also need to know what the amphorae contained. For many types this is well established—Dressel 1 carried wine; Dressel 20 carried olive oil; Beltra´n II carried Wsh products, etc. But for other types the identiWcation of contents is not secure: several Tripolitanian II amphorae, a form until recently thought to have carried olive oil like many other Tripolitanian amphorae, have been shown to have been pitched, and therefore probably carried wine or Wsh products.34 Haltern 70, which tituli picti show carried defrutum or sapa, or olives preserved in defrutum, or sometimes muria (a Wsh sauce), has long been the subject of debate, between those who saw it as a wine amphora, on the basis that defrutum was a ‘vin cuit’, and those who interpreted defrutum as a preservative syrup. The most recent study establishes that defrutum was boiled must, used as a preservative or sweetener, or as an additive in the winemaking process.35 Then there is the problem of reuse. Occasional reuse is to be expected, and is unlikely to skew statistics signiWcantly.36 But occasionally entire cargoes were composed of reused amphorae. The Grado wreck (mid-second century ad) carried a cargo of 23–5 tonnes of salted Wsh or garum all in some 600 reused amphorae of various types, whose principal contents were normally olive oil or wine (Table 9.1).37 Had it not been for the Wsh remains found packed 34 Bonifay 2004: 89 and 92; 2007. 35 Carreras and Aguilera 2004. For earlier interpretations, see e.g. Parker and Price 1981, Sealey 1985, both arguing for defrutum as a preservative syrup; Liou 1988, E´tienne and Mayet 2000, interpreting defrutum as a ‘vin cuit’. 36 Cf. Liou 1988. All 8 Dressel 1C amphorae from the La Cavalie`re wreck contained olives, possibly preserved in defrutum (Sealey 1985: 25); 3 Dressel 1B amphorae from the Albenga wreck (60–50 bc) contained hazelnuts, and several contained grain, although the vast majority (thousands) of Dressel 1B from this wreck contained wine (Parker 1992a: 50); Dressel 2–4 is primarily a wine amphora but tituli picti indicate that it could carry related products such as vinegar or defrutum; some Dressel 2–4 amphorae from the Dramont D and La Tradelie`re wrecks carried dates (Sealey 1985: 46–7). The Wfty or so Dressel 1 amphorae from the 1970s excavations at Colchester Sheepen were produced 20 years or more before the occupation of the site and are thought to have arrived at the site as reused vessels (Sealey 1985: 11–13, 101–8). 37 Auriemma 2000.

Approaches to Quantifying Roman Trade

233

Table 9.1. Contents of the amphorae from the Grado wreck (2nd century ad) Quantity 200 20 150 20

Type

Normal contents

Here used for

Africana I (piccolo) Tripolitana I Knossos A/53 Grado I

Olive oil Olive oil Wine Garum

Sardina pilchardus 10–11 tonnes Fish (14 with mackerel) 2 tonnes Sardina pilchardus Garum

Tonnage

Note: There were also a smaller number of other, more diverse amphorae (also reused?); the only one whose contents could be examined contained Wsh. Source: Auriemma 2000.

inside the amphorae, it would undoubtedly have been concluded that the ship had carried mainly oil and wine. Even if we can Wnd assemblages in which we have a suYciently high proportion of amphorae types whose contents and capacity are well known, it is diYcult to achieve a meaningful analysis of relative proportions of e.g. wine, oil, and Wsh products imported to a site since we do not know how much was carried in other containers, such as skins or barrels. A valiant attempt to deal with this problem has been made by Bo Ejstrud.38 This uses quantitative data for amphorae from Augst, Nijmegen, Avenches, Saint-Romain-en-Gal, and Rome, combined with capacity information for particular amphora types, to calculate the relative proportions of the volumes of wine, oil, and garum imports. The relationship between oil and garum remains fairly stable across the sites, but the proportion of wine varies considerably—over half the assemblage in Saint-Romainen-Gal and Rome, but around a quarter or less in the three sites north of the Alps. Ejstrud assumes that the discrepancy is to be explained by the greater use of barrels for transporting wine in northern Europe, and attempts to calculate how much wine might have been imported in barrels to the northern sites to give overall proportions of the three products which would match those from the Mediterranean sites. This is ingenious, but could be questioned on several counts. First, the assumption that consumption patterns were uniform is contentious (meat consumption, for example, varied regionally).39 38 Ejstrud 2005.

39 King 1999.

234

Andrew Wilson

Moreover, the argument relies on wine being the only product carried in barrels in northern Europe. Several Wsh-salting sites with their characteristic salting vats are known from Armorica (northwest Brittany), including one at Plomarc’h which is the second largest known salting installation in the Roman world.40 As no amphora production is known from this area, the Breton salted Wsh must have been exported in barrels, thus giving two archaeologically invisible products and vitiating the possibility of isolating wine as a single variable factor in Ejstrud’s analysis. One would imagine that major markets for Breton salted Wsh would have included the urban and military sites of Britain and the lower Rhine frontier. Augst, on the upper Rhine, may be close enough to the Mediterranean to have been supplied principally with amphoraborne salted Wsh from Mediterranean sources, but Nijmegen, one of the other sites compared, may have received part of the Breton export in barrels. While the peak in Breton salted Wsh production is essentially a second-century phenomenon, too late to aVect the contexts of ad 30–70 analysed by Ejstrud, one cannot wholly exclude the shipment of salted Wsh in barrels at this date. If amphorae are problematic to use for quantifying trade, we might get somewhere by using a diVerent kind of proxy—not the container for the goods, but the infrastructure used to produce them, provided that we can be sure that the product was being made for export or long-distance trade. One instance where this can be tried is in the production of salted Wsh and related by-products, such as the sauces garum, liquamen, and allec. Although Wsh were salted and sauces produced on a fairly small scale in pithoi or dolia from at least the Wfth century bc, and continued to be right into the early medieval period, from the Wrst century bc until the Wfth century ad we see the appearance of large concrete vats, providing a permanent built infrastructure for large-scale salting. Such factories spread around the western Mediterranean, especially along the coast of North Africa and around the Straits of Gibraltar; there are also clusters in the Black Sea, and on the coast of Armorica. They concentrate in regions where there are large seasonal migrations of Wsh; the resulting massive catches at 40 Sanquer and Galliou 1972; Curtis 1991: 74–8, with references.

Approaches to Quantifying Roman Trade

235

3000

Aggregate capacity (m3)

2500

2000

1500

1000

500

45

0 55

0

1– 50

AD

1–

50

0

0

45

40

1– AD

0

40 1–

35 AD

AD

0

35

1– 30

AD

25

1–

30

0

0

25

20

1– AD

0

20 AD

AD

15

1–

15

00

1–

–1

10 AD

50

51 AD

1– AD

10

0–

50 50 BC –1

BC

0

Fig. 9.9. Fish-salting vat capacities by half-century—all regions (36 factories at 23 different sites) (Wilson 2006, Wg. 2)

certain times of the year could not be consumed locally and thus had to be salted to allow preservation and export to more distant markets. The Wve or six centuries of the large Wsh-salting factories thus represent a phenomenon where the cultural taste for salt Wsh and garum, combined with access to a pan-Mediterranean market under the Roman empire, oVered the potential for large-scale exports, and this in turn attracted very considerable investment in the productive infrastructure.41 Within this period of six centuries, where the production capacity and the market must have been of a diVerent order of magnitude than that implied by the dolium-based production before and afterwards, we can track some trends by comparing the aggregate vat capacity of certain factories over time. We are currently limited to some 34 factories for which minimum estimates of vat capacities are possible, and where the chronology is suYciently clearly published to 41 Wilson 2006.

236

Andrew Wilson

Aggregate vat capacity (m3)

1800 1600 1400 1200 1000 800 600 400 200

10 0 10 1– AD 15 15 0 1– AD 2 20 00 1– AD 2 25 50 1– AD 3 30 00 1– AD 3 35 50 1– AD 4 40 00 1– AD 4 45 50 1– AD 5 50 00 1– 55 0

50

AD

51 –

AD

1– AD

10 0–

50 50 BC –1

BC

0

Fig. 9.10. Fish-salting vat capacities by half-century—southern Iberia and North Africa (28 factories at 14 different sites) (Wilson 2004, Wg. 3)

relate these capacities and the periods for which they were in use at least to 50-year periods. Figure 9.9 shows the total capacities aggregated by half-century; it is very much a view of minimum capacities for these sites, parts of which have been lost to erosion, or may not have been completely excavated. The sample is not large, but the main outline of the trend is clear. There is a rise in production capacity to a peak in the later second century, with a sharp drop in the third and a steady decline from the fourth century onwards.42 Disaggregating the graph into two of its main component regions, we see that around the Straits of Gibraltar the peak is earlier and more sustained (Fig. 9.10), from the late Wrst through to the end of the second century, with a sharp drop in the third, rising again slightly to a low peak in the early fourth century, then a gentle decline again followed by a steep drop-oV in the early Wfth century. By contrast in Brittany (Fig. 9.11) the investment in Wsh-salting plants

42 Wilson 2006.

237

1200 1000 800 600 400 200

0

0 50

1–

55

0

50

45

1–

AD

0

45 AD

40

1–

40

0 35

1– AD

0

35

30

1– AD

0

30

1– 25

AD

0

25 1–

20 AD

AD

0

20 1–

15 AD

AD

10

1–

15

00

50

–1 51

AD

1– AD

10

0–

BC

0

50 50 BC –1

Aggregate vat capacity (m3)

Approaches to Quantifying Roman Trade

Fig. 9.11. Fish-salting vat capacities by half-century—Brittany (6 factories at 5 different sites) (Wilson 2006, Wg. 4)

begins in the late Wrst century (both epigraphy and the technology of the vats show this is a clear instance of investment by Mediterranean Romans), with capacity rising to a massive peak in the second, much of this accounted for by the vast factory at Plomarc’h in the Bay of Douarnenez, with a capacity of 500 m3; production declines only slightly until the late third century, but drops sharply in the fourth and ceases by the Wfth. The Armorican salted products, we should remember, were not transported in amphorae, and it is study of the production sites that begins to give us some idea of the quantity of produce that must have been transported in barrels to the markets in northern Europe, doubtless including the armies in Britain and on the Rhine frontier.

P OT TE RY Table pottery and cooking pottery, being traded in its own right rather than as a container for other goods, at least lacks two dimensions of uncertainty (contents and volume) that amphorae may have. The best way to track trends over time in production and distribution is that pioneered by Fentress and Perkins in 1988 and later

238

Andrew Wilson Even % ARS / year

2 Even Mean 1.5

1

0.5

0 −30

25

80

135 190 245 300 355 410 465 520 575 630 685

Fig. 9.12. Graph of even frequencies for each ARS form compared to the mean derived from the dates of each form (Fentress et al. 2004, Wg. 11.2)

developed with the help of Bruce Hitchner and Sergio Fontana.43 Taking data from several Weld surveys, where the material was unstratiWed but unlikely to be biased towards a particular period or context, they plotted quantities of African Red Slip ware over time. They assumed that a sherd of a form produced over a 100-year period had an equal probability of being made in any year within that range; ‘200 sherds [of that form] would thus result in a probable distribution of two pots per year’. The resulting aggregate totals are thus to some extent inXuenced by the date ranges assigned to each form, but the eVect of this can be assessed by generating a dummy curve with one sherd of each form (Fig. 9.12), and comparing the resulting curve (dark line) with the mean from actual data (grey line).

43 Fentress and Perkins 1988; Fentress et al. 2004. The following discussion summarizes the main arguments of these two articles.

Approaches to Quantifying Roman Trade

239

The simulated data suggest that the dating ranges used create the following eVects: 1. the sharp rise to ad 100, as 63 new forms Wrst appear in the later Wrst century ad; 2. the steep drop at 245/250—45 forms, including many common ones, have a latest date of ad 250; 3. the slow rise from 245 to 300; 4. the peak at 500–20, caused by overlap between a set of 22 forms with a latest date of 500 and 25 forms with a start date of 500–10; dating one set of forms ten years earlier or the other ten years later would largely remove this peak. In a way, the graph with its third-century low is basically a composite of two normal distribution curves for the earlier ‘A’ production and the later ‘D’ production, with a subsidiary normal distribution curve in between them representing ‘C’ production. The growth of C wares, produced in Byzacena, parallels an increase in contemporary settlement there, and looks like an example of import substitution, with Byzacena even competing in the export markets with northern Tunisia where the A and D productions were made. This leaves as real features of the production the increase from ad 100 hitting a peak around 155 and then falling sharply to 175. Suggested reasons for the fall include the Antonine plague, or the creation of the Classis Africana for the annona in the 150s; if the navicularii were prevented from carrying non-state cargoes Africa might have lost an advantage in transport if Wne wares had previously travelled with annona cargoes on private ships. Although the date ranges of the forms aVect the graph, the number of forms produced at any one period—another key inXuence on the graph’s shape—might be taken as a crude reXection of economic vitality. Encouragement for this interpretation is given by Fentress’s comparison of the ARS graph with a plot of African building inscriptions per year (Fig. 9.13); the very similar shapes suggest that, as it is unlikely that one is directly dependent on the other, both reXect overall patterns in the economic health of the African provinces.44 44 Fentress and Perkins 1988: 210–11.

240

Andrew Wilson

4 1

ARS % per annum

2

Building inscriptions per annum

3

2

1

200

250

300

350

400

450

Fig. 9.13. Comparison of ARS production and public building dedications in North Africa (Fentress and Perkins 1988, Wg. 4)

If the overall curve for the mean of the various diVerent survey assemblages analysed should reXect trends in overall production, the diVerences between this mean and the curve for an individual survey should reXect peculiarities in the supply to that region—in other words, local characteristics of the trade between it and the production centres. On Jerba, for example, the ARS imports follow the general shape of the mean but at a much lower level until the midfourth century, when they leap above it and remain above it until the seventh century (Fig. 9.14). Interestingly, this relative rise in imports in the later period is not reXected in a greater vitality of settlement— rather the opposite. In the Libyan valleys survey the picture is reversed (Fig. 9.15)—greater relative imports until ad 245, and then less than the mean as import substitution from Tripolitanian Red Slip ware kicks in—but again, the import curve is the opposite of the settlement trend. While the African sites receive less than the mean, sites on Sicily receive more ARS than the average of the various surveys, with perhaps a closer correlation between import activity and settlement vitality. A possible explanation for the comparative

2 Jerba Mean 1.5

1

0.5

0 −30

25

80

135 190 245 300 355 410 465 520 575 630 685

Fig. 9.14. Jerba: percentage of ARS recovered over time (Fentress et al. 2004, Wg. 11.11) 3 Libyan Valleys Mean 2.5

2

1.5

1

0.5

0 −30

25

80

135 190 245

300 355 410 465 520 575 630 685

Fig. 9.15. UNESCO Libyan Valleys Survey: percentage of ARS recovered over time (Fentress et al. 2004, Wg. 11.13)

242

Andrew Wilson

Greece Palestine

Asia Minor North Africa

Cyprus Africa Proc.

AD

15

0

AD

13

5–

13

5

AD

0– 12

5– 10

90

–1

12

05

0

AD

AD

0

AD

5

–9 75

AD

0 –6

–7 60

AD

5 45

–4 30

0

AD

AD

–3 15

15 1–

–1 15

30

–1

5

BC

BC

40 35 30 25 20 15 10 5 0

Syria average

Fig. 9.16. Distribution over time of Italian Terra Sigillata exported to the eastern Mediterranean (Poblome 2004, Wg. 6)

scarcity of ARS on African sites may be that it was primarily an export ware, and Sicily had privileged access, lying close to Africa and en route to major markets at Rome. There is considerable potential for applying this type of analysis to other major tableware productions to which it has not yet been systematically applied—Italian Terra Sigillata, Gaulish Samian wares, Eastern Sigillata A. Jeroen Poblome has used the technique to chart the export of Italian sigillata to the eastern Mediterranean (Fig. 9.16); a study of ARS supply to the eastern Mediterranean, primarily a late phenomenon, would be very interesting. The most recent study of Samian in Britain also uses a similar approach, graphing the diVerent regional Samian productions against each other as percentages of the total Samian supply.45 A more systematic comparison of ARS with these other pottery types would allow some comparison of regional trade in the diVerent wares, and further elucidate the changing distributions and market share of the various regional productions. One could also apply this approach to some cooking ware productions, as there is now increasing recognition of the importance of trade in these from the Hellenistic period as

45 Willis 2005.

Approaches to Quantifying Roman Trade

243

represented at Euesperides, right through until late antiquity with Pantellerian ware and other productions.46

C E RA MI C A S SE MB L AG E S Several studies have used the quantiWcation of ceramic assemblages from diVerent sites to explore the degree to which Roman cities were interconnected and the relative importance of their diVerent sources of imports, notably Fulford’s comparisons of assemblages from Ostia, Carthage, and Berenice.47 However, the extent to which pottery quantiWcation has been attempted for excavations varies considerably across the Roman world. In Britain and northern Europe quantiWcation has been regarded as a normal component of excavation pottery reports since the 1970s or 1980s; its application is still haphazard for excavations in the Mediterranean (though is now quite common in Italian, French, and recent Spanish excavations), and rarer still for sites in the Eastern Roman empire. Moreover, pottery reports of the 1970s did not always distinguish between coarse wares and amphorae in their quantiWcation Wgures, reXecting the organization of the excavation Wnds processing operation or the post-excavation programme, yet for economic analysis these categories of material need to be treated separately since they represent totally diVerent categories of goods. If one wants to compare imports to diVerent regions across time, one needs sites where quantiWed ceramic assemblages of a statistically relevant size are available from several diVerent phases to provide a diachronic picture for the site. (This has been done for the amphorae from Ostia and several other sites,48 but not for other pottery categories.) The assemblages need to have been quantiWed in a comparable way (sherd count, sherd weight, Minimum Vessel Count, Estimated Vessel Equivalents, etc.). The combination of these factors means there is still a dearth of sites where all are present, rendering a widespread 46 Euesperides: Wilson 2005; Swift 2006. 47 Fulford 1987; 1989. 48 Panella and Tchernia 1994.

244

Andrew Wilson

comparison of import patterns complicated. A synthesis of available assemblages meeting these criteria is a key research priority.

FUTURE PERSPECTIVES Despite the huge amount of published archaeological data, there remain problems of synthesizing material from published sources. Frequently this is to do with comparability—in terms of the manner of excavation and recording, of classiWcation of Wnds, and of their presentation in publication. But it can be done for many assemblages and projects; and the publication of some datasets on the Internet should in theory make this easier. Data mining of such datasets needs to be a major high-level research goal of the future. Here again we run into problems of compatibility of publications and formats, but increasingly tools are being developed for cross-searching of databases, including intermediate interface layers such as thesauri which allow a search engine to equate, for example, diVerent names for the same amphora type on diVerent sites. But this is going to require further work from IT- and web-literate archaeologists, both in developing search tools and in designing websites for future publications, before we can search a variety of Weld survey databases to get a picture of e.g. ARS or Samian supply calculated from their data in the manner pioneered by Fentress et al. How do we compare the Roman picture with that for later or earlier times? Most evidence that has shaped our ideas about the vitality of medieval trade comes from documentary sources such as the Geniza Archive, or the letters and accounts of Francesco Datini, the fourteenth-century ‘Merchant of Prato’. On the one hand, the documents recently emerging from Vindolanda, and the Egyptian documentary papyri, begin to suggest that such documents were present in antiquity also, but have suVered from poor survival rates. But properly to compare trade in the Roman and medieval periods we need to compare like with like—which predominantly means comparing the archaeological record for diVerent periods. Much useful work for the early medieval period has been done in this respect by Wickham, in Framing the Early Middle Ages; this has

Approaches to Quantifying Roman Trade

245

served to point up the considerable diVerence between levels of Roman exchange and those in the second half of the Wrst millennium ad.49 What we really need now is a study covering the archaeology of, say, 1200 to 1800, to facilitate comparison with the Roman world and see where in this period diVerent facets of the traded economy might return to Roman levels. Finally, it needs to be stressed that there is little prospect of quantifying vague notions of ‘trade’ in the abstract; the question must be broken down into smaller parts—trade in what goods, between what regions, at what times? Each time we have looked at a large dataset, e.g. shipwrecks, we have seen the possibility of certain biases in the data, which need to be tested by isolating a subset of the data, e.g. the wine amphora wrecks oV the southern French coast which account for much of the Republican peak. This process may simultaneously concretize our understanding of one part of the picture—the wine trade between Italy and Gaul, followed by import replacement—while reducing our conWdence in a simplistic reading of the larger shape of the graph, i.e. the uncertainty over the extent to which the use of barrels aVects the recovery of later wrecks. Nevertheless, this is the way we must proceed: to accumulate large general datasets, analyse them, test them for biases, isolate the biases by probing more detailed case studies, and then see how those studies aVect our understanding of the total dataset again. Bibliography Auriemma, R. (2000). ‘Le anfore del relitto di Grado e il loro contenuto’, MEFRA 112: 27–51. Bellini, G. R., and Rea, R. (1985 [1986]). ‘Note sugli impianti di produzione vinicolo-olearia nel suburbio di Roma’, in Misurare la terra: centuriazione e coloni nel mondo romano. Citta`, agricoltura e commercio: materiali da Roma e dal suburbio. Modena, 119–31. Bonifay, M. (2004). E´tudes sur la ce´ramique romaine tardive d’Afrique. Oxford. —— (2007). ‘Que transportaient done des amphores africaines?’, in E. Papi (ed.), Supplying Rome and the Roman Empire (JRA Suppl. 69). Portsmouth, RI, 8–25. 49 Wickham 2005.

246

Andrew Wilson

Bowman, A. K. (20032). Life and Letters on the Roman Frontier: Vindolanda and its People. London. Camodeca, G. (1999). Tabulae Pompeianae Sulpiciorum (TPSulp.). Edizione critica dell’archivio puteolano dei Sulpicii, 2 vols. Rome. Carreras, C., and Aguilera, A. (2004). ‘El Producte envasat’, in Culip VIII i les a`mfores Haltern 70. Girona, 117–32. Casson, L. (1971). Ships and Seamanship in the Ancient World. Princeton. Cockle, H. (1981). ‘Pottery manufacture in Roman Egypt: A new papyrus’, JRS 71: 87–95. Corrente, M. (1985 [1986]). ‘Alcuni esempi di forme economiche nel settore Est del suburbio romano’, in Misurare la terra: centuriazione e coloni nel mondo romano. Citta`, agricoltura e commercio: materiali da Roma e dal suburbio. Modena, 112–18. Curtis, R. I. (1991). Garum and Salsamenta: Production and Commerce in Materia Medica. Leiden. Davies, J. K. (2006). ‘Hellenistic economies’, in G. R. Bugh (ed.), The Cambridge Companion to the Hellenistic World. Cambridge, 73–91. de Callata, F. (2005). ‘The Graeco-Roman economy in the super longrun: Lead, copper and shipwrecks’, JRA 18: 361–72. Desbat, A. (1991). ‘Un bouchon de bois du Ier s. apre`s J.-C. recueilli dans la Saoˆne a` Lyon et la question du tonneau a` l’e´poque romaine’, Gallia 48: 319–36. Ejstrud, B. (2005). ‘Size matters: Estimating trade of wine, oil and Wsh-sauce from amphorae in the Wrst century ad’, in T. Bekker-Nielsen (ed.), Ancient Fishing and Fish-Processing in the Black Sea Region. Aarhus, 171–81. E´tienne, R., and Mayet, F. (2000). Le vin hispanique. Paris. Fentress, E., Fontana, S., Hitchner, R. B., and Perkins, P. (2004). ‘Accounting for ARS: Fineware and sites in Sicily and Africa’, in S. E. Alcock and J. F. Cherry (eds.), Side-by-Side Survey. Oxford, 147–62. —— and Perkins, P. (1988). ‘Counting African red slip ware’, in A. Mastino (ed.), L’Africa romana. Atti del V Convegno di studio. Sassari, 205–14. Fulford, M. G. (1987). ‘Economic interdependence among urban communities of the Roman Mediterranean’, World Arch. 19.1: 58–75. —— (1989). ‘To East and West: The Mediterranean trade of Cyrenaica and Tripolitania in antiquity’, in D. J. Mattingly and J. A. Lloyd (eds.), Libya: Research in Archaeology, Environment, History and Society 1969–1989. Lib. St. 20, Special issue, 169–91. Gibbins, D. (2001). ‘Shipwrecks and Hellenistic trade’, in Z. H. Archibald, J. Davies, V. Gabrielsen, and G. J. Oliver (eds.), Hellenistic Economies. London, 273–312.

Approaches to Quantifying Roman Trade

247

Gould, R. A. (2000). Archaeology and the Social History of Ships. Cambridge. Hopkins, K. (1980). ‘Taxes and trade in the Roman Empire (200 bc–ad 400)’, JRS 70: 101–25. Horden, P., and Purcell, N. (2000). The Corrupting Sea: A Study of Mediterranean History. Oxford. Jongman, W. M. (2007). ‘Gibbon was right: The decline and fall of the Roman economy’, in O. Hekster, G. de Kleijn, and D. Slootjes (eds.), Crises and the Roman Empire. Proceedings of the Seventh Workshop of the International Network Impact of Empire (Nijmegen, June 20–24, 2006). Leiden, 183–99. King, A. (1999). ‘Diet in the Roman world: A regional inter-site comparison of the mammal bones’, JRA 12: 168–202. Liou, B. (1988). ‘Le contenu des amphores, typologie, et e´pigraphie: quelques cas aberrants ou embarrassants’, in Socie´te´ Franc¸aise d’E´tude de la Ce´ramique Antique en Gaule—Actes du Congre`s d’Orange. Marseille, 171–7. MarliŁre, E. (2002). L’outre et le tonneau dans l’Occident romain. Montagnac. Martin-Kilcher, S. (1998). ‘Le vin dans la Colonia Iulia Karthago’, in M. Comas and P. Pa´dros (eds.), El vi a l’antiguitat: Economia, produccio´ i comerc¸ al Mediterrani occidental. Badalona, 511–29. —— (2004). ‘Amphorae in the Roman West: Discussion and research since 1989’, in J. Eiring and J. Lund (eds.), Transport Amphorae and Trade in the Eastern Mediterranean (Monographs of the Danish Institute at Athens 5). Athens, 263–72. Morley, N. (2007a). ‘The Early Roman Empire: Distribution’, in W. Scheidel, I. Morris, and R. Saller (eds.), The Cambridge Economic History of the Greco-Roman World. Cambridge, 570–91. —— (2007b). Trade in Classical Antiquity. Cambridge. Panella, C., and Tchernia, A. (1994). ‘Produits agricoles transporte´s en amphores: L’huile et surtout le vin’, in L’Italie d’Auguste a` Diocle´tien. Rome, 145–65. Parker, A. J. (1990a). ‘Classical antiquity: The maritime dimension’, Antiquity 64.243 (June): 335–46. —— (1990b). ‘The wines of Roman Italy’, JRA 3: 325–31. —— (1992a). Ancient Shipwrecks of the Mediterranean and the Roman Provinces. Oxford. —— (1992b). ‘Cargoes, containers and stowage: The Ancient Mediterranean’, IJNA 21: 89–100. —— (1996). ‘Sea transport and trade in the ancient Mediterranean’, in E. E. Rice (ed.), The Sea and History. Stroud, 97–109.

248

Andrew Wilson

Parker, A. J. and Price, J. (1981). ‘Spanish exports of the Claudian period: The signiWcance of the Port Vendres II wreck reconsidered’, IJNA 10.3: 221–8. Peacock, D. P. S. (1980). ‘The Roman millstone trade: A petrological sketch’, World Arch. 12.1: 43–53. —— and Williams, D. (eds. 2007). Food for the Gods: New Light on the Ancient Incense Trade. Oxford. Poblome, J. (2004). ‘Italian sigillata in the Eastern Mediterranean’, in J. Poblome, P. Talloen, R. Brulet, and M. Waelkens (eds.), Early Italian Sigillata: The Chronological Framework and Trade Patterns. Proceedings of the First International ROCT-Congress, Leuven, May 7 and 8, 1999. Leuven, 17–30. Pomey, P., and Tchernia, A. (1978). ‘Le tonnage maximum des navires de commerce romains’, Archaeonautica 2: 233–51. Rathbone, D. (2001). ‘The ‘‘Muziris’’ papyrus (SB XVIII 13167): Financing Roman trade with India’, in M. Abd-el-Ghani, S. Z. Bassiouni, and W. A. Farag (eds.), Alexandrian Studies II in Honor of Mostafa el Abbadi. Alexandria, 39–50. Sanquer, R., and Galliou, P. (1972). Garum, sel et salaisons en Armorique gallo-romaine’, Gallia 30: 199–223. Sealey, P. R. (1985). Amphoras from the 1970s Excavations at Colchester Sheepen (BAR British Series 142). Oxford. Shaffrey, R. (1998). ‘The Use of Old Red Sandstone in Roman Britain’. Ph. D. thesis, University of Reading. —— (2003). ‘The rotary querns from the Society of Antiquaries’ excavations at Silchester, 1890–1909’, Britannia 34: 143–74. Strauss, E. J. (2006). ‘Roman Trade in the Eastern Mediterranean’, Ph.D. thesis, University College London. Swift, K. (2006). ‘Classical and Hellenistic Coarse Pottery from Euesperides (Benghazi, Libya): Archaeological and Petrological Approaches to Production and Inter-Regional Distribution’. D.Phil. thesis, University of Oxford. Tchernia, A. (1990). ‘Contre les e´paves’, in A. Duval, J.-P. Morel, and Y. Roman (eds.), Gaule interne et Gaule me´diterrane´enne aux IIe et Ier sie`cles avant J.-C.: Confrontations chronologiques. Actes de la table ronde de Valbonne, 11–13 novembre 1986. Paris, 291–301. —— (2006). ‘La crise de l’Italie impe´riale et la concurrence des provinces’, Cahiers du Centre de Recherches historiques 37: 137–56. Tomlin, R. S. O. (2003). ‘ ‘‘The Girl in Question’’: A new text from Roman London’, Britannia 34: 41–51.

Approaches to Quantifying Roman Trade

249

Whittaker, C. R. (1985 [1987]). ‘Trade and the aristocracy in the Roman empire’, Opus 4: 49–75. —— (1989). ‘Amphorae and trade’, in Amphores romaines et histoire e´conomique: Dix ans de recherche (CEFR 114). Rome, 537–9. Wickham, C. (2005). Framing the Early Middle Ages: Europe and the Mediterranean, 400–800. Oxford. Wild, J. P., and Wild, F. C. (2005). ‘Rome and India: Early Indian cotton textiles from Berenike, Red Sea coast of Egypt’, in R. Barnes (ed.), Textiles in Indian Ocean Societies. London and New York, 11–16. Will, E. L. (2004). ‘Mediterranean amphoras in India’, in J. Eiring and J. Lund (eds.), Transport Amphorae and Trade in the Eastern Mediterranean (Monographs of the Danish Institute at Athens 5). Athens, 433–40. Williams, D. F. (2004). ‘The eruption of Vesuvius and its implications for the early Roman amphora trade with India’, in J. Eiring and J. Lund (eds.), Transport Amphorae and Trade in the Eastern Mediterranean (Monographs of the Danish Institute at Athens 5). Athens, 441–50. Willis, S. (2005). ‘Samian Pottery, a resource for the study of Roman Britain and beyond: The results of the English Heritage funded Samian Project. An e-monograph’, Internet Archaeology 17, http://intarch.ac.uk/ journal/issue17/willistoc.html. Wilson, A. I. (2005). ‘Une cite´ grecque de Libye: Fouilles d’Euesperides (Benghazi)’, CRAI 2003: 1648–75. —— (2006). ‘Fishy business: Roman exploitation of marine resources’, JRA 19: 525–37. —— (2008). ‘Large-scale manufacturing, standardization, and trade’, in J. P. Oleson (ed.), Handbook of Engineering and Technology in the Classical World. Oxford, 393–417.

10 Approaches to Quantifying Roman Trade: Response Michael Fulford

After a brief consideration of ‘what is the point of trying to quantify trading activity in the Roman world’, Andrew Wilson’s chapter is concerned primarily with the ‘what’ of trade. His focus is primarily on the actual and potential contribution that the study of commodities can make: the trade in marble and other stone, such as querns and millstones; the study of amphorae—the ceramic containers for perishables like wine, olive oil, Wsh products, preserved fruits, etc; and of pottery, particularly the widely distributed Wne wares, such as the North African-manufactured African Red Slip Wares. To these examples he reXects a little on what the study of the infrastructure of production can add—in this case the study of the factories producing salted Wsh and related by-products. On a diVerent but related topic concerning infrastructure, and using the chosen category of evidence as a measure of the changing volume and direction of (Mediterranean) trade, he has also considered the study of ancient shipwrecks. With each of his case studies, Wilson is careful to probe their strengths and weaknesses. In the case of categories of artefact or commodity the prime concern is the availability of suitably quantiWed data that can be tested with rigorous comparative analysis, but there is also the need to set the respective categories in their material context. Barrels or skins, rather than amphorae, may have been the preferred medium of transport of wine, olive oil, and other commodities in certain regions at certain times, but these survive

Response to Andrew Wilson

251

exceptionally in the archaeological record; equally, the incidence of an artefact category such as ceramic tableware or a quern has to be seen in the context of its Wnd circumstance and the extent of Weldwork, and in relation to similar products—another material type or make of tableware or quern, for example. Then there is the question of how the incidence relates to diVering intensities of Weldwork; do large quantities of amphorae on military sites in the north-western provinces simply reXect the huge amount of excavation that has been undertaken on them in preference to towns and rural sites? Can pottery tablewares be considered in isolation from glassware and metalwork (which tend to be recycled when discarded)? To what extent is our understanding of shipwreck data biased by the relative visibility on the seabed of cargoes of amphorae and the lack of rigorous and methodical, regional survey to identify the ships which carried less archaeologically visible cargoes? Probably the greatest expected incidence of wrecks in the western Mediterranean between the mid-Wrst century ad and the mid-Wfth century ad might be expected in the seaways between Ostia/Portus and the sources of grain and olive oil emanating from North African ports, particularly Africa Proconsularis. There is no hint of this in the shipwreck record. One explanation for this might be that even the volume of traYc generated by the corn supply was less than that carrying other commodities in other regions; another would be that the relevant coasts and deep sea have not been systematically sampled by rigorous survey. If Wilson’s focus is on the ‘what’ of trade, should we not also consider the collective evidence of all categories of available evidence, but this time from the perspective of the consumer? Gradually we are acquiring more and more datasets of Wnds from excavated sites throughout the settlement hierarchy in several provinces and regions of the Roman empire. If these datasets have sound chronological control we have the possibility, Wrst, of exploring change through time at each settlement under consideration. Second, by comparative analysis, we can observe variation in patterns of consumption among diVerent types of settlement in the region or province; then region to region, province to province. This approach could consider Wrst diversity in material and biological remains, the latter including the direct evidence of animals and plant types consumed as well as the indirect as indicated by transport amphorae, etc. Second, there

252

Michael Fulford

is the extent of the contribution to the life of the community concerned from outside its immediate territory. What derives from the region, outside the immediate territory, but within a radius of, say, 100 km; what derives from further aWeld? What is the proportion of long-distance traded foodstuVs and material goods? How do these ratios change with time? This approach assumes that the sources of all material goods and foodstuVs are known, but of course this is seldom, if ever, the case. However, it is possible to develop probabilities in determining sources for the ‘unknown’ categories, using provenancing techniques in the case of some materials, such as ceramics and stone, or making assessments of workshop traditions for rarer items. Just as the study of individual categories of material in Wilson’s study is limited to the available sample of recorded examples, so, too, is the study of a settlement’s consumption limited by the size of the available dataset which seriously inhibits the answering of the question ‘how much?’. Nevertheless the correlation of diversity against relative abundance could form the basis for the development of an index, period by period, of a settlement’s relative wealth or complexity and thereby form the basis for inter-provincial or inter-regional comparisons. It will also address from the point of view of the consumer the impact of the rise and fall of particular industries and food producers. How, for example, did the decline of Baetican olive oil export and the consequent rise of African oil aVect the overall consumption of olive oil among settlements around the shores of the west and central Mediterranean? At the moment we do not have the basis for making such comparisons between settlements other than on the basis of size, usually taken as the walled area of a settlement, and on the number and diversity of public buildings, which, in the absence of written evidence, are often taken as an indicator of the settlement’s legal status. The emphasis on the walled area privileges the settlement’s development trajectory at one particular moment, as does the provision of public buildings, usually taken from their estimated date of construction. While geophysical survey and surface collection of datable ceramics—particularly valuable when the techniques are combined— are beginning to provide alternative approaches to calculating size of the built environment and change over time, the post-construction

Response to Andrew Wilson

253

histories of public buildings do not provide a safe insight into the settlement’s changing fortunes over time, though their initial construction does provide evidence of scale of investment at that particular moment. Approaching Roman-period economic behaviour from the point of view of studying both the ‘what’ of trade and the variable patterns of material and biological consumption through the settlement hierarchy on a regional or provincial basis is clearly one, interconnected way forward for examining change over time and regional variation. We need to deWne the artefacts and their provenances; and we also need to study their consumption in context. In addition, however, it is important to begin to think about the relative and changing value of goods and foodstuVs over time and distance and determine what is local, as opposed to regional or luxury (long-distance) trade. Let us take the Mediterranean, for example, where, as many have commented before, access to the sea undoubtedly enhanced the ease with which goods were transported around its shores and determined new approaches to the technology of transportation. It is generally accepted that the development of the shape of the amphora as a vessel for transporting liquids was to facilitate its tight packing into the holds of ships. But, in a Mediterranean context, we are immediately confronted with the question of what constitutes local for an amphora-carried wine which is produced near to and exported from a particular port? Setting aside the capital costs of the ship, what was the cost diVerential between transporting an amphora 15 miles, the land-based equivalent of local, or 500 miles, the land-based equivalent of long-distance or luxury? In the terrestrial context, the pattern of distribution inland from the point of production, constrained by the costs of transport, is likely to be clearly deWned, with a steep fall-oV of volume with distance from the origin. In the maritime situation, however, the fall-oV of volume with distance is much more gradual, making the task of distinguishing between local, regional, and long-distance traYc much harder. Consider the cost implications of distributing wine to any location around, say, the west basin of the Mediterranean from a port also located on the shore of the same basin. Can wine exported from Cosa in the later Republic in Dressel 1 amphorae, for example, be regarded as local a wine to consumers in Carthage, Marseilles, Naples,

254

Michael Fulford

Tarragona, etc, as it was to consumers in Cosa and its hinterland? Is the Mediterranean distribution of Dressel 1 diVerent in any way from that of earlier wine amphorae? Is it a question of the volume of the trade, or rather the pattern of distribution away from the Mediterranean coast into the hinterland of Iberia or Gaul, and as far as southern Britain which is distinctive and distinguishes Dressel 1 from its predecessors?1 That some, if not all, of the coastal consumption of wine carried in Dressel 1 amphorae might be regarded as local is borne out by the pattern of behaviour of ordinary cooking wares produced close to the sea and with correspondingly extensive maritime distributions, such as the hand-made pottery produced on the island of Pantelleria, midway between Sicily and North Africa, in late Republican and imperial times.2 This was as ‘local’ in ports along the North African and west Italian coasts as it was in Pantelleria, but there is no evidence for signiWcant inland distribution away from the ports. In a terrestrial context where distribution is entirely based on landbased transport, the circulation of cooking wares is generally concentrated within a radius of about 30 miles of the production centre. In a maritime situation 30 land-based miles translates potentially into hundreds of marine miles. If we regarded cooking wares of any particular period as essentially an indicator of the local we might regard the normal distance coastally produced wares travelled by sea as determining the ‘local’ in a maritime context. Allowing that local, regional, and long-distance can carry diVerent meanings in a maritime as opposed to a terrestrial context, it is then necessary to determine whether a distribution has exceptional characteristics. In the case of wine carried in Dressel 1 amphorae it has been suggested above that signiWcant inland distribution in Gaul and Iberia might distinguish it from wines with an essentially coastal distribution. Similar observations can be made about the distribution of olive oil-carrying Dressel 20 amphorae produced along the Guadalquivir and exported in signiWcant quantity via Cadiz, situated at the mouth of that river.3 This oil, too, has a very pronounced inland distribution with signiWcant quantities of Wnds on the

1 Tyers 1996: 89–90.

2 Peacock 1982: 78–80.

3 Tyers 1996: 87–9.

Response to Andrew Wilson

255

northern frontier, particularly along the Rhine and in Britain, and probably transported via the Rhoˆne and the Rhine. At the same time there appears to be a marked bias in the Mediterranean distribution towards Rome where very large quantities have been recorded from Monte Testaccio. Indeed, on the evidence that we have this mound of discarded amphorae seems to be very largely composed of Dressel 20 amphorae. The distribution of this type of transport vessel appears therefore to distinguish itself from others with a Mediterranean circulation in two respects, extreme variability in the maritime context and a maintained and focused terrestrial distribution, evident in the lack of decline in quantity with distance from the production areas to the frontiers. Such a diVerence in the nature of the distribution points to contributing factors other than those of a straightforward commercial character. The signiWcant presence at Rome and on the northern frontier suggests that Baetican olive oil was bought and distributed for political reasons, supplying garrisons and the populace in Rome. While these subjective observations about Dressel 1 and Dressel 20 amphorae can in due course be tested statistically with quantiWed data, the behaviour of the latter type at least reminds us that any consideration of patterns of consumption may need to take account of a political context. Potentially, both a competitive, entrepreneurial mode of distribution and consumption can run alongside a system driven and paid for by the state, in this case to contribute to Rome’s annona and the supply of the army. Clear distinctions in the pattern of ceramic supplies to military as opposed to civilian sites have been observed in early Roman Britain. South Gaulish sigillata (tableware) and olive oil-carrying amphorae (predominantly Baetican Dressel 20s) dominate military assemblages, while North Gaulish, Gallo-Belgic pottery (tableware) and wine amphorae dominate the others.4 Implications of these studies are not only that the value and signiWcance of these goods and commodities are diVerent according to the context of their consumption, but also that there were parallel, but not necessarily mutually exclusive, systems of transport or trade, one in the domain of the state, the other in the domain of the private merchant. 4 Willis 1996; Pitts and Perring 2006.

256

Michael Fulford

In relation to value, the evidence of diVerentiated distribution systems raises more general questions about luxury and long-distance trade. In terms of the energies expended to deliver amphorae from the Mediterranean to Britain, no distinction can be drawn between those which were consumed by the army and those which reached the native elites. For the former, the goods delivered were part of their expectation—wine, olive oil, sigillata tableware, and many other goods were part of what was perceived to be necessary to support an appropriate and accustomed lifestyle among the frontier armies and garrisons. The Roman pottery and other Wnds from the Neronian fortress at Usk in South Wales together provide a typical example of the range and quantities of imported goods to be found in a Wrstcentury Roman military context.5 Barely half a century later the documents recovered from the fort at Vindolanda (ad 90–120) on the northern frontier of Britain provide a complementary picture to that of the archaeology.6 In the case of foodstuVs, the items likely to have been contained by amphorae are supplemented by a list which includes further exotica, such as spices and pepper, from south or south-east Asia, but also a range of food, more than two-thirds of the total, such as cereals, eggs, pork fat, venison, etc, and drink, notably beer, which could have been provided from the region, if not very locally.7 The totality of more than forty diVerent types of food and drink, and the ways in which they would have been prepared and served to table, rather than the individual item, are striking. It is not clear what would, or might have been, considered a luxury in this context, but there is little doubt that the maintenance of a distinct, ‘Roman’ or ‘Roman military’ way of life, was an essential ingredient in maintaining morale on the frontiers. Cash certainly changed hands for some of the items listed, but we do not know how costs impacted on the standard of living of the individual oYcer or ordinary soldier. The contrast in the archaeological record with contemporary native sites in northern Britain for which we do not have any written sources is pronounced.8 In terms of the ceramic assemblages alone, only small quantities of imported amphorae, sigillata, and other wares are 5 6 7 8

Greene 1979; Manning 1993; Manning, Price, and Webster 1995. Bowman 1994; Bowman and Thomas 1983; 1994; 2003. Bowman 1994: 69–70; Bowman and Thomas 2003: 15–16. Willis 1996.

Response to Andrew Wilson

257

recorded. As with Vindolanda we cannot easily gauge the impact acquiring these luxuries had on the native elites and their retinues. Even if the exchange rate was no longer one slave for one amphora of wine as Diodorus Siculus described for the Gauls in the late Republic (V. 26.3), the price was likely to have been signiWcant. In the case of the southern British native elites, who had been in receipt of similar goods from the Mediterranean region for several generations before annexation, the wine represented a luxury, a commodity which revealed and reinforced their status and was a prominent feature of their burial assemblages. Luxury to a native chief was but one element in the complex, standardized way of life of a Roman oYcer. In order to develop further analyses, such as the above consideration of luxury, and to conclude, it is essential that we not only continue to develop rigorous approaches and consistent methodologies of quantiWcation in the way that Wilson advocates, but that we also do so in a comprehensive fashion at the various points of consumption through the settlement hierarchy. Approaches which take account not only of the character and variability in consumption in the settlement context, but also of the individual categories of evidence which constitute our evidence of trade, will also encourage the development of more robust and relevant models of interpretation beyond reciprocity, redistribution, and marketing. And to address the question of ‘what is the point?’ with just one example: without rigorous classiWcation and characterization of material and biological evidence linked with systematic quantitative approaches, we would not begin to understand the extent of the eVort and energies involved in supplying the Roman army in the north-western provinces between the late Wrst century bc and the mid-third century ad. Bibliography Bowman, A. K. (1994). Life and Letters on the Roman Frontier. London. —— and Thomas, J. D. (1983). Vindolanda: The Latin Writing Tablets (Britannia Monogr. 4). London. —— —— (1994). The Vindolanda Writing Tablets (Tabulae Vindolandenses II). London. —— —— (2003). The Vindolanda Writing Tablets (Tabulae Vindolandenses III). London.

258

Michael Fulford

Greene, K. T. (1979). Report on the Excavations at Usk, 1965–76: The PreFlavian Fine Wares. CardiV. Manning, W. H. (1993). Report on the Excavations at Usk, 1965–76: The Roman Pottery. CardiV. —— Price, J., and Webster, J. (1995). Report on the Excavations at Usk, 1965– 76: The Roman Small Finds. CardiV. Peacock, D. P. S. (1982). Pottery in the Roman World: An Ethnoarchaeological Approach. London. Pitts, M., and Perring, D. (2006). ‘The making of Britain’s Wrst urban landscapes: The case of late Iron Age and Roman Essex’, Britannia 37: 189–212. Tyers, P. A. (1996). Roman Pottery in Britain. London. Willis, S. (1996). ‘The Romanization of pottery assemblages in the east and north-east of England during the Wrst century ad: A comparative analysis’, Britannia 27: 179–221.

11 A Comment on Andrew Wilson: ‘Approaches to Quantifying Roman Trade’ William Harris

1. SHIPW REC KS A major contribution of Andrew Wilson’s excellent survey is his analysis of the statistics derived by A. J. Parker from shipwreck archaeology. Parker’s achievement remains remarkable, and he himself uttered plenty of warnings about how to interpret this kind of evidence. But Wilson’s fresh analysis still leaves us with a considerable historical puzzle, as he well appreciates. Neither the textual nor the material evidence from sites on land would lead us to suspect that there was a decline in maritime trade in the western Mediterranean (measured by the number of vessels employed) after ad 50 and especially after 100, as the shipwreck evidence suggests. The supposed replacement of amphorae by barrels does not seem to be an adequate explanation, all the less so if we factor in the strong possibility that many regions of the Roman empire were by this date suVering from serious deforestation.1 Nor do the rise and the fall of the various areas that produced amphora-borne commodities seem to be a suYcient answer. There is no reason either to suppose that the average size of amphora-carrying ships increased between the Wrst century bc and the second century ad. It has seldom if ever been noted that shipwrecks in which metal ingots are the main reported cargo—which might be thought to be a 1 Cf. Harris 2005: 35–6.

260

William Harris

better index of overall trade and of overall economic activity than the amphorae—also appear to decline somewhat from the Wrst century bc to the Wrst and second centuries ad. That makes our problem even more severe. Since we seem to be faced with a development speciWcally aVecting maritime commerce, I suggest that we examine further the possibility that it was maritime safety that changed. There can be little doubt that in the Wrst century bc the introduction of ships’ pumps decreased the relative number of shipwrecked merchant vessels.2 Scholars more knowledgeable than I am about ancient ships and sailing practices might tell us whether in high imperial times all progress in these areas was lacking. Casson’s classic account is not to any great extent developmental. The well-known volume on navigation edited by Patrice Pomey strongly suggests that improvements were quite common in the Roman period, without making a detailed case.3 One notices from time to time new features that presumably made merchant Xeets more seaworthy, for example deadeyes in the rigging of ships of the mid-second century ad and later.4 In Parker’s brief account of this matter, what is most intriguing is the apparent continuation of improvements into late antiquity.5

2. AIMS AND RESULTS What is the point of attempting to quantify Roman trade? Wilson oVers various suggestions, three of which are especially clear and suggestive: quantiWcation might help us to tell (a) whether the ‘urban centres’ of the Roman empire depended on long-distance trade, and (b) how extensively trade oVered investment opportunities. And it might help us to judge (c) whether there was sustained growth in GDP per capita in any period.6 The Wrst question perhaps answers 2 Parker 1992: 28. 3 Pomey 1997; see esp. 85–6. 4 Whitewright 2007. 5 Parker 1992: 25. 6 I will leave on one side here the intriguing questions whether quantifying Roman trade might help us to estimate how much the state invested in infrastructure or exacted in tax revenue.

Comment on Andrew Wilson

261

itself, once we take into account the full range of long-distance imports that the more prosperous kind of Roman town consumed in the late Republic and high Empire (and once we take into account the exports that in most cases paid for them). Tchernia’s well-known demonstration of the area from which Rome drew its wine will serve as an illustration of the eVects of the few genuine cities (in the modern sense), while the huge range of metal artefacts from excavated Pompeii shows beyond doubt that the quality of Pompeian life was deeply dependent on imports.7 All this is elementary, but it leads to two further questions which I shall take up brieXy later on—what was the full range of important traded commodities? And how typical were the consumption patterns of the Vesuvian cities? The question of investment opportunities undoubtedly deserves further investigation. It is to be linked both to the question of the money supply and to social inhibitions about types of investment. But Wilson’s report does not show how quantiWcation in itself will help us to get a handle on the question whether Romans with capital could normally Wnd ways to invest it, or the question whether they normally wanted to ‘invest’ in the modern sense.8 The question of growth, obviously central to the economic history of the Roman empire, requires us to use all the evidence both textual and material.9 It also requires us to pay attention to the whole population of the Roman empire—or explain why we are not doing so. The material evidence certainly gives some very vivid impressions about long-term change, say between 100 bc, ad 150, and ad 400. Can we do better than that?10 We might, for example, create two (or more) competing models, avoiding extreme versions that have no genuine chance of being right. We might weigh a choice between (a) a Greene–Hitchner model, with relatively extensive 7 Ciarallo and De Carolis 1999 gives an incomparably rich impression of this material. 8 On Roman attitudes to investing see Kehoe forthcoming. 9 For divergent views as to how to approach this question, with respect to the late republican period, see Scheidel 2007 and Harris 2007. 10 Not according to de Blois and Rich 2002, who write (xiv–xv) that ‘the relation between the ‘‘command-economy’’ of tax goods [sic] and requisitional commodities [sic], and the ‘‘market-economy’’ constitutes an almost insurmountable obstacle for those who would like to gauge the role of inter-regional trade and thereby the extent to which ‘‘Smithian’’ growth took place’. But these are not the real problems.

262

William Harris

diVusion of productive technological innovation and large spurts of growth;11 and (b) a more modest, essentially Malthusian, model in which there is indeed some diVusion of productive technology, but no breakthrough. Important evidence continues to accumulate: take the case of the wooden force pumps studied by Richard Stein.12 But how much eVect did they have on production, and do they really count as evidence in favour of Model (a)? The material evidence is, in my opinion, most likely to help us to make progress with the growth problem if we concentrate on the causes of economic growth in known historical societies. The role of fossil fuels in the Industrial Revolution will always save us from excessive RostovtzeYan modernizations of the Roman economy. But the way archaeology can best contribute, as Wilson in eVect showed in a landmark article some Wve years ago,13 is by indicating how far productive technological changes spread at various dates. The more we can extend and reWne our knowledge of the spread of water mills and other forms of hydraulic technology, the closer we shall get to a satisfying answer.

3 . S E E I N G W H AT I S ( H A L F- ) H I D D E N ‘Archaeologically’, Wilson writes, ‘we are forced to concentrate on durable artefacts, and in particular those for which we can attribute a provenance. In practice, this largely means objects of pottery or stone . . .’. But if we concentrate on objects of pottery or stone, we are never I think going to establish in a deWnitive fashion the superiority of either of the two models sketched above. There are precious few lentils left in the archaeological record, that is true. But there are, to start with, enormous quantities of metal artefacts. How to make use of them, quantitatively? This question could be answered in several diVerent ways. It should be possible to hypothesize 11 See among other publications Greene 2000 and Hitchner 2005. I do not mean to imply that these two scholars agree with each other in every respect. 12 Stein 2004. They were an innovation of the late Wrst or early second century ad, Stein 2004: 246. 13 Wilson 2002.

Comment on Andrew Wilson

263

the annual volume of ironware purchased in Flavian Pompeii, and even Flavian Italy, within certain broad limits; and it would be possible to compare this with what happened in economies that are considered by many to be broadly comparable in nature, if not in size, to the economies of the more advanced regions of the Roman empire, say those of western Europe in the sixteenth and seventeenth centuries. And we need to ask what happens to productivity when an agrarian economy begins to make much greater use of metal tools. More generally we need to speculate in as informed a way as possible about the volume of major commodities that leave little or no archaeological record, without forgetting the vast range of other commodities from lapis lazuli to lions.14 It was a useful exercise to consider the slave trade from this point of view. There is much more to say about timber: once again, we need some informed speculation about regional and overall consumption. We need Roman archaeologists who know the consumption patterns of both Italians and provincials—such as Wilson—to take fairly speciWc periods in Roman history and engage in some informed and sensible guesswork about the volume of trade in metals, timber, and a number of other commodities, keeping constantly in mind the comparanda from Renaissance and early modern Europe. How much perfume did it take to make Rome smell sweet, how much paint did it take to cover the walls of Pompeii and hundreds of other Roman towns? Then there is the matter of textiles.

4. ‘URBAN CENTRES’ The revolution in Roman archaeology in the last generation, so at least it could be argued, was survey archaeology. But it is not to be forgotten that from Silchester to Aphrodisias we have also learned a great deal more about Roman urban settlements. Is it not time that a team of archaeologists attempted a more ambitious typology of the towns and villages of the Roman empire, considering their territories 14 It must be added that the dichotomy between luxuries and staple commodities (Wilson, at the beginning) is almost meaningless, as I have argued elsewhere, and it has done nothing but harm to the economic history of Rome. See now Morley 2007: 39–43. For a long catalogue of commodities, though not the full range, see Harris 2000.

264

William Harris

as well as their nuclei, a typology aimed at telling us in the end how to rate the typicality of the consumption patterns of the Vesuvian towns and of Egyptian villages? The documents concerning the thirdcentury Fayum estate of Appianus rarely mention objects made of iron (which hardly means, however, that the estate almost entirely lacked iron tools); contrast the wide diversity of metal-working craftsmen attested at Tebtunis.15 To Wnd out what is generalizable in this and other respects, we need to work more on rural and urban consumption, in all provinces, over long periods. This work should be highly quantitative. A task for OXREP? Bibliography Ciarallo, A., and De Carolis, E. (1999). Homo faber: natura, scienza e tecnica nell’antica Pompei. Milan. de Blois, L., and Rich, J. (2002). ‘Introduction’, in L. de Blois, and J. Rich (eds.), The Transformation of Economic Life under the Roman Empire. Amsterdam, ix–xx. Greene, K. (2000). ‘Technological innovation and economic progress in the ancient world: M. I. Finley re-considered’, EHR 53: 29–59. Harris, W. V. (2000). ‘Trade’, in A. K. Bowman, P. Garnsey, and D. Rathbone (eds.), Cambridge Ancient History XI (2nd edn.). Cambridge, 710–40. —— (2005). ‘The Mediterranean and ancient history’, in W. V. Harris (ed.), Rethinking the Mediterranean. Oxford, 1–42. —— (2007). ‘The Roman economy in the Late Republic, 133–31 bc’, in W. Scheidel, I. Morris, and R. Saller (eds.), The Cambridge Economic History of the Greco-Roman World. Cambridge. Hitchner, R. B. (2005). ‘ ‘‘The advantages of wealth and luxury’’: The case for economic growth in the Roman Empire’, in J. G. Manning and I. Morris (eds.), The Ancient Economy: Evidence and Models. Stanford, 207–22. Johnson, A. C. (1936). Roman Egypt (An Economic Survey of Ancient Rome II). Baltimore. Kehoe, D. (forthcoming). ‘The Early Roman Empire: Production’, in W. Scheidel, I. Morris, and R. Saller (eds.), The Cambridge Economic History of the Greco-Roman World. Cambridge. Morley, N. (2007). Trade in Classical Antiquity. Cambridge.

15 Johnson 1936: 343.

Comment on Andrew Wilson

265

Parker, A. J. (1992). Ancient Shipwrecks of the Mediterranean and the Roman Provinces (BAR-IS 580). Oxford. Pomey, P. (ed. 1997). La Navigation dans l’antiquite´. Aix-en-Provence. Scheidel, W. (2007). ‘A model of real income growth in Roman Italy’, Historia 56: 322–46. Stein, R. (2004). ‘Roman wooden force pumps: A case-study in innovation’, JRA 17: 221–47. Whitewright, J. (2007). ‘Roman rigging material from the Red Sea Port of Myos Hormos’, International Journal of Nautical Archaeology 36: 282–92. Wilson, A. (2002). ‘Machines, power and the ancient economy’, JRS 92: 1–32.

This page intentionally left blank

Part V Coinage

This page intentionally left blank

12 Roman Silver Coinage: Mints, Metallurgy, and Production Matthew Ponting

Scholars have been interested in the composition of Roman silver coins since at least 1834 when Ernst von Bibra published assays of Wfteen early imperial denarii. The reason for this interest is simple— the amount of silver in the coins provides important evidence for the study of the Roman economy. Ostensibly a straightforward aim, one might think, but one which has proved more complex the more knowledge has been gained. The Wrst analyses were conducted using classic wet-chemical techniques requiring large samples cut from coins. The results were usually very reliable, but the techniques were time-consuming and the fact that the coin analysed was largely destroyed made obtaining statistically valid sample numbers impossible, especially for the period before the later third century ad. The introduction of instrumental analysis based on x-rays in the 1970s meant that coins could be analysed non-destructively and these techniques were seized upon by the archaeological and numismatic community as the long-awaited panacea. Curators and archaeologists were more inclined to permit analysis of large numbers of coins by a method that could be used without damage and this resulted in a number of analytical projects. The most important of these from our point of view was David Walker’s Metrology of the Roman Silver Coinage that used x-ray Xuorescence (XRF) to chart the variation in the silver content of Roman coinage from Augustus to

270

Matthew Ponting

the mid-third century.1 Walker measured the silver and copper contents quantitatively and also noted the presence of gold and lead where they were detected, but the main aim of the project was to determine the changes in the silver content. The technique was not sensitive enough to quantify minor constituents fully. Walker was clearly aware of the problem of surface enrichment caused by burial that is usually encountered when trying to measure the silver content of ancient silver-alloys. He was careful to abrade the area selected for analysis before analysis took place and to take repeated measurements, further abrading the area between each measurement, until a series of three consistent measurements was obtained. It was then assumed that the enriched surface layer had been pierced and that the average of the consistent measurements was representative of the original alloy from which the coin was made. Walker’s metrology has remained the main source of information about the composition of Roman silver coinage until quite recently, and indeed is still the only source for many issues. However, in 1989 Mike Cowell of the British Museum Department of ScientiWc Research demonstrated that it was possible that Walker’s method did not adequately penetrate the enriched surface layer, thereby casting doubt on the reliability of Walker’s published data.2 In 1994 I started a project analysing Flavian silver coins from Caesarea in Cappadocia using atomic absorption spectroscopy (AAS), a technique that requires a sample to be removed from the coin and dissolved in acid. To take the sample I adopted the tried and tested technique of drilling into the cylindrical edge of the coin and using the turnings from the drill as the sample. To avoid including metal that was silver-enriched I discarded the Wrst millimetre or two of material and only collected the heart-metal. Using this method I reanalysed several coins in the British Museum collection that had been previously analysed by Walker and obtained diVerent results that conWrmed Cowell’s suspicions. Further work, obviously necessary to explain why Walker’s data are unreliable, has shown that the surfaces of debased Roman silver coins were usually artiWcially 1 Walker 1976–8.

2 Schmitt-Korte and Cowell 1989.

Roman Silver Coinage

271

Fig. 12.1. Scanning electron micrograph of a section through a post-reform denarius of Nero. The boxes marked correspond to two areas that were analysed (see text for details).

enriched in silver, a process that produced a much thicker enrichment layer than would have been produced purely by natural processes. Consequently, the abrade-measure-abrade technique used by Walker results in measuring the silver content within the artiWcially produced enriched zone of the coin. This can be demonstrated on the SEM micrograph of a section through a denarius of Nero shown in Fig. 12.1; at the surface, area 1 has a silver content of 91.5%, whereas the heart-metal (area 2) has a silver content of 80.8%. The abrasion of a few microns will only result in the analysis of the enriched metal. Comparing the results of Walker’s analysis with the results of the AAS analysis shows a signiWcant diVerence in the silver contents and also in the consistency of the results; indeed Walker was surprised at the variability of some of his results, which he put down to poor mint practice. However, the new data clearly show that the standards were usually rigorously maintained; it was the variability in the depth of the enriched zone that caused the variability noted by Walker.

272

Matthew Ponting

The artiWcially enriched zone was produced by oxidizing the copper in the alloy and then stripping the oxidized copper out of the alloy by pickling in dilute acid. This was achieved by leaving the coin blanks in a furnace at a red heat for a prolonged period of time and then soaking them in acid before striking. It is likely that the method developed out of the necessity of cleaning the black copper oxide Wlm oV freshly cast blanks of debased silver-alloy. It must have been observed that blanks that were left exposed in the air whilst still hot from the mould would have had a heavier black coating and that, once this was removed, the blank appeared brighter and more silvery. An almost identical process is still used today for the production of modern commemorative coins, although the scientiWc principles are now so well understood that the thickness of the enriched zone can be precisely controlled; as Walker’s results show, the ancient mint was unable to do this. It should be stressed that the technique used by Walker is perfectly reliable when used on coins with silver contents greater than 90%. This is the point when a mixture of silver and copper becomes homogeneous, below which the alloy exists as two distinct phases (as seen in the denarius of Nero) and takes on an increasingly pinkish tinge. However, the artiWcial enrichment process, or ‘blanching’ as it is sometimes called, not only disguised the pinkish tinge that would become increasingly noticeable to the Roman public as the amount of copper added to the alloy was increased, but also deceived the modern analyst into believing that natural corrosion was the only problem to be surmounted. Believing a coin to be made of a highsilver alloy and therefore suitable for surface analysis on the basis of appearance is clearly dangerous; the only reliable way to determine this is to take a sample from the heart-metal of the coin. Another non-destructive technique that has achieved increasing acceptability is neutron activation, especially the fast instrumental version developed in France by Guerra and Barrandon.3 This technique provides a highly sensitive analysis of an entire coin as it is today—and as such is a very reliable non-destructive technique. The problem is the very fact that it does provide an accurate chemical analysis of the entire coin as it is today. It essentially homogenizes 3 Guerra and Barrandon 1998.

Roman Silver Coinage

273

any variability in composition caused by either corrosion processes or by artiWcial means to provide an average bulk composition. The problem with this approach was demonstrated way back in 1972 by two other French scientists, Condamin and Picon, who compared the ‘present total composition’ of silver denarii with the composition of the remaining heart-metal of the coin and found that in certain cases between one-third and one-half of the original total copper content had been lost from the coins.4 This translates into a total present silver content of 56% where the original composition as measured in the heart-metal is only 47%. Such measurements may be an accurate determination of the total composition of an individual coin as it is today, but is not an accurate determination of the original alloy from which the coin was made. As an archaeologist, my aim is to establish the original composition of the alloy mixed in the mint and used to make the coins in order to understand how the Roman state manipulated its coinage and controlled its bullion supplies, not merely to measure the total composition of individual coins as they come down to us in the archaeological record. In order to do this, a sample of the remaining heart-metal within the coin needs to be analysed, and the least damaging way to do this is to use a very thin drill. Almost any elemental analytical technique can be used to determine the composition of this sample, although the sample size is small (10 mg) and the fragmentary and irregular nature of the drillings makes it most suitable for techniques that are highly sensitive and require the sample to be dissolved in dilute acid. Such techniques include AAS and inductively-coupled plasma atomic emission spectrometry (ICP-AES). ICP-AES is much quicker and more sensitive for most elements of interest, which is an advantage when the sample weight is so small, and this is why it is used in the current project. Furthermore, using such a technique not only enables accurate measurement of the copper and silver in samples that are reliably representative of the original alloy, but also allows the accurate quantiWcation of the lead and gold that Walker was unable to achieve. This is particularly useful in establishing the standards to which the ancient moneyer believed that he or she was adhering. Modern instrumental analysis measures the elemental abundances of 4 Condamin and Picon 1972.

274

Matthew Ponting

silver, a pure chemical element; the silver metal extracted from various ores using traditional technology was not chemically pure; it contained contaminants from both the original ores and from the reWning processes. These contaminants are principally gold, lead, and bismuth; so the silver bullion measured out at the mint should be calculated as the sum of the silver, gold, lead, and bismuth rather than just the elemental silver. Recent work undertaken using this approach has begun to change our understanding of the debasement of Roman silver coinage. According to Walker it was in Nero’s reign that debasement started, coinciding with the general reform of ad 64. Before this the denarius was made of pure silver bullion. Our recent analyses of the JulioClaudian series5 agree with Walker that it was in Nero’s reign that the debasement started, but it now appears that the process was more complex than his data suggest. The Wrst intentional additions of copper appear in the corona civica issues of ad 60/1, and may have started earlier, although this question has yet to be addressed by further analyses. This debasement was minimal, consisting of the addition of between 0.5% and 1.5% of copper to the silver bullion, and seems to have been limited to the one issue; the Virtus and Roma issues of ad 63/4 were struck on the old pure Julio-Claudian standard. However, in agreement with Walker’s data, the reform of ad 64 is the main event, although Walker’s Wgure of a minimal debasement down to an average of 93.5% silver is incorrect. The heart-metal of thirty denarii so far analysed shows that the silver bullion content was reduced to 80% and that this was disguised by blanching. Curiously, at the end of Nero’s reign the last issue (MacDowell type 3b)6 was struck with a bullion content of 90%, an apparent attempt to improve the purity of the coinage. This standard continues to be used under Otho, but with the Civil Wars the situation becomes more complex. Over one hundred denarii from this period have been analysed so far, but the data have yet to be studied in detail. First impressions suggest that the anonymous issues were predominantly pure bullion, but also included many plated coins. The silver content of the issues of the four contenders vary somewhat; Otho started with the late Neronian standard of 90%, as already 5 Butcher and Ponting 2005a.

6 MacDowell 1979.

Roman Silver Coinage

275

mentioned, but his last issue was down to 80%. Galba continues the 90% at Rome, but struck pure bullion at his Gallic mints and in Spain. Vitellius also struck at two standards; again pure bullion at Lyon, but back to the Neronian 80% at Rome. Vespasian appears to have largely stuck to the 80% standard, although there are certain issues that clearly buck the trend. These are very provisional results and the conclusions are inevitably going to be reWned and changed as more data are acquired and more closely scrutinized. However, one overall conclusion that can be made is that the blanching of the debased issues has served to disguise the severity of the reform of ad 64 and obscured the complexities and Xuctuations of the subsequent issues. Clearly the Roman state manipulated the silver bullion content of the denarius coinage with considerable frequency especially when under pressure, and was highly successful in disguising these acts. The current AHRC-funded project will involve the analysis of some 1,500 coins covering the period from the death of Nero to the Wrst issues of Septimius Severus and is set to Wnish in early 2009. Analysis by ICP-AES can also measure several other useful trace elements as well as the gold, lead, and bismuth already mentioned. These can provide additional data that may relate to production technology or even allow speculation on the origin of the silver or copper. This is an area that has received relatively little systematic scientiWc study. Although the trace elements that are now measurable in silver alloys can show interesting patterning that allows the chemical characterization of speciWc issues and linking diVerent issues together to suggest a common mint or workshop, they cannot be used to trace the original geographical source of the silver. However, certain types of silver ore have chemical characteristics that can be used to suggest which ore types may have been used. In particular gold and bismuth can be useful in this way; the oxidized silver ores (cerussite and anglesite) and jarosite ores generally contain high levels of gold (1.5% or more), whereas silver from galena ore will contain very little (0.01% to 0.1%). Jarosite ores will also usually contain signiWcant levels of bismuth, unlike the oxidized ores, although silver from galena can also contain up to 1% of bismuth. Furthermore, similar ore types occur at many diVerent places, although there are certain regions that are known for speciWc ore types, such as jarosite ores in the Huelva region of Spain or galena

276

Matthew Ponting

from Lavrion in Greece. It is therefore signiWcant that Julio-Claudian silver denarii show a gradual decline in their gold content with Augustan denarii generally having the higher gold contents and also high bismuth. Tiberius’ denarii are lower in gold, but have a similar range of bismuth contents as do the two denarii of Caligula and the six denarii of Claudius so far analysed. It therefore appears that high-gold high-bismuth silver sources were being used under Augustus, possibly Spanish jarosite ores, which were gradually diluted by silver from ores containing lower levels of gold under subsequent rulers. The overall impression is one of gradual change caused by the mixing of new metal with the recycling of older coin. However, Nero’s denarii behave quite diVerently, with the two prereform issues analysed dividing neatly between the corona civica issue having high gold and low bismuth and the Virtus and Roma issues conforming to the low gold and moderate bismuth portion of the previous reigns. The majority of Nero’s post-reform denarii form a well-deWned cluster marked by relatively high gold and low bismuth; similar to the corona civica issue. The relationship of the corona civica issue, as the Wrst issue to be adulterated with copper, with the postreform issues surely cannot be circumstantial and seems to mark an abrupt change in the silver bullion source(s) being used. It is purely conjectural, but high-gold and low-bismuth contents are features of the so-called ‘dry-ores’ (chlorargyrite and argentite); these are rich ores that, it is commonly accepted, would have been all but worked out by the Wrst century ad and could therefore suggest that these coins were produced from the recycling of older silver bullion objects. A problem that besets the interpretation of silver coin analyses is that precious metal resources are, by their very nature, limited and therefore available metal was carefully controlled. Recycling must therefore have played a large part in silver coin production and would have led to the homogenization of any regionally speciWc chemical signature quite quickly. Indeed, this is precisely what was noticed in the compositions of the coinage of Caesarea in Cappadocia. Here, starting in the Flavian period, numismatists had noticed that the coins were produced in two distinct styles and that one of these styles was identical to that of the contemporary denarii struck in Rome. The suggestion was that the Rome-style Caesarean

Roman Silver Coinage

277

coins were struck from dies that were cut in Rome or that the coins themselves were struck in Rome and shipped out to the province. It was hoped that analysis would resolve this question; if the Romestyle coins were made of the same metal as the denarii known to have been struck in Rome, and if both of these were chemically diVerent from the Caesarea coins of local style, it would indicate that the Rome-style coins were struck in Rome and shipped to Caesarea. This was indeed found to be the case, with the compositions of the two style groups being signiWcantly diVerent and the ‘Rome-style’ coins having the same gold and lead levels as the contemporary denarii.7 However, by the reign of Trajan the diVerences between the Rome-style and local-style coins are far less distinct, although the match between the denarii and the Rome-style coins remains. This sort of loss of resolution is common in Roman silver analyses from the end of the Wrst century onwards and is best explained by increasing recycling of silver bullion; a diVerent source of chemical structure therefore needs to be found for later coinages. Luckily, as the silver coinage becomes more debased, the trace contaminants coming into the alloy with the copper become increasingly measurable as the concentration of copper rises. These elements (arsenic, nickel, cobalt, chromium, iron, manganese, antimony, zinc, and tin) can provide the basis for compositional characterization that is independent of the silver. For example, by the reign of Severus the bullion content of the denarius had been reduced to 50% and the silver is highly homogenized. However, the copper alloyed with the silver bullion is not yet homogenized and appears to have been locally or at least consistently procured. This enables the stylistically identiWed products of Severus’ eastern mints to be distinguished from the stylistically identiWed products of the mint of Rome by the amount of nickel they contain, and also permits coins of ambiguous style to be correctly attributed.8 The copper alloyed with the silver bullion appears to have been pure reWned copper directly brought from the smeltery and not recycled copper from a general pool of previously alloyed metal. Prior to the third century there is no evidence of copper being used in alloying with silver that had been recycled in the same way as appears to have been the case with the 7 Butcher and Ponting 1995.

8 Gitler and Ponting 2003.

278

Matthew Ponting

silver bullion. This is conWrmed by the fact that tin and zinc remain at the levels expected in traditionally reWned copper and it is only in the latest issues of Severus and his family that traces of these metals start to appear at levels suggesting that recycled bronze was Wnding its way into the copper reserved for coinage. This supports the contention that despite the increased recycling of silver bullion obscuring any characteristic chemical signature associated with it, the trace elements accompanying the copper will enable meaningful characterization to be attempted. This is particularly interesting when the chemistry of some of the provincial coinages is examined. Some issues of Alexandria in Egypt during the Julio-Claudian period were analysed recently and it was found that there is a signiWcant relationship between issue date and the concentrations of arsenic and nickel.9 The suggestion here is that the Alexandrian mints obtained copper on a yearly basis from at least three sources and that sources were returned to in subsequent years although more than one source could supply metal for an issue in a single year. Similar patterning has been found in the coinage of Antioch and seems to be a feature of the provincial issues related to the way metal was procured. To investigate further the sources of the metal used for the production of Roman silver coins, lead isotope analysis is also being used. Lead isotope analysis is a technique that measures the abundances of the four isotopic forms of lead. Because lead occurs at low levels in silver and the ratios of these abundances are often speciWc to ores from certain locations, lead isotope analysis could provide a means of provenancing Roman silver coins. However, lead would also be added to concentrate the silver in the extraction of silver from jarosite ores and would have been used in the extraction of silver from recycled coin. This added lead would have its own isotopic signature that would swamp any signature from the lead originally in the silver. Consequently, the lead isotope signature of Roman silver coins is likely to be that of the lead added either during extraction or recycling. However, this is potentially useful, especially as it means that some of the conventional arguments surrounding the use of lead isotope analysis can be sidestepped. Because lead was an abundant 9 Butcher and Ponting 2005b.

Roman Silver Coinage

279

and low-value metal, it is unlikely to have been heavily recycled, meaning that the isotopic signature of such metal may still relate to its origin and allow discussion of the procurement of lead used in silver production. With these points in mind, a small number of isotope determinations were made on Julio-Claudian denarii to see if such data could be useful.10 The results were informative and even surprising; Lavrion can clearly be ruled out as a source for JulioClaudian silver, as can known Italian deposits. An Augustan denarius struck in Lyon has a close match with lead and silver metal excavated at Rio Tinto in Spain, and therefore supports the idea of Spanish silver being used under Augustus. The surprise was that a close match was found between denarii of Augustus, Tiberius, and Nero and lead from Britain! Whilst the database of comparative lead isotope data is not exhaustive, and data continue to be added, a British origin for lead used to smelt silver cannot yet be ruled out, even before the conquest. It is known that lead from Cartagena in Spain was shipped the 550 km to Rio Tinto to be used in extracting silver from jarosite ore, so the idea that lead was being shipped from Britain to Spain, perhaps as a cheap return cargo for all that Spanish wine and garum that was being imported by the British elite, is perhaps not as ridiculous as all that. Many more data need to be collected before this suggestion can be substantiated or ignored and a much larger programme of lead isotope analyses has just commenced.

AC KN OWLE DG EME NT S This project is a collaboration between myself and the numismatist Kevin Butcher at the American University of Beirut (now at the University of Warwick). Any project of this nature and scale cannot possibly succeed without the help, support, and cooperation of a large number of individuals and institutions. I would therefore like to acknowledge the following people, in no particular order and apologize to those I miss out: Richard Reece, Ian Carradice (St Andrews University), Chris SomerWeld (Scheme, University of Nottingham), Jane Evans and Vanessa Pashley (NIGL, British Geological 10 Butcher and Ponting 2005a: 191.

280

Matthew Ponting

Survey, Keyworth), Jonathan Williams and Richard Abdy (British Museum), Mike Cowell (British Museum, retired), Jyrki Muona (Helsinki), Justine Bayley and David Dungworth (English Heritage), John Davies (Norwich Castle Museum), Peter Robinson (Doncaster Museum), Jenny Hall (Museum of London), Steven Minnitt (Somerset Museums), Chris Howgego (Ashmolean Museum), Philip Wise (Colchester Castle Museum), Gale Boyle (Bristol City Museum), Ticca Ogilvie (National Museum of Scotland), Richard Greaves and Benedikt Zaech (Mu¨nzkabinett Winterthur), Michel Prieur and Keith Sugden (Manchester Museum), and The Garstang Museum. Bibliography Butcher, K., and Ponting, M. (1995). ‘Rome and the East: Production of Roman provincial silver coinage for Caesarea in Cappadocia under Vespasian, ad 69–79’, Oxford Journal of Archaeology 14.1: 63–77. —— —— (2005a). ‘The Roman denarius under the Julio-Claudian emperors: Mints, metallurgy and technology’, Oxford Journal of Archaeology 24.2: 163–97. —— —— (2005b). ‘The Egyptian billon tetradrachm under the JulioClaudian emperors—Wduciary or intrinsic?’, Schweizerische Numismatische Rundschau 84: 93–124. Condamin, J., and Picon, M. (1972). ‘Changes suVered by coins in the course of time and the inXuence of these on the results of diVerent methods of analysis’, in E. T. Hall and D. M. Metcalf (eds.), Methods of Chemical and Metallurgical Investigation of Ancient Coinage (Royal Numismatic Society Special Publication No. 8). London, 49–66. Gitler, H., and Ponting, M. (2003). The Silver Coinage of Septimius Severus and his Family (ad 193–211). Milan. Guerra, M. F., and Barrandon, J.-N. (1998). ‘Ion beam activation analysis with a cyclotron’, in W. A. Oddy and M. R. Cowell (eds.), Metallurgy in Numismatics, vol. 4 (Royal Numismatic Society Special Publication No. 30). London, 15–34. MacDowell, D. W. (1979). The Western Coinages of Nero. New York. Schmitt-Korte, K., and Cowell, M. (1989). ‘Nabatean coinage—Part I: The silver content measured by x-ray Xuorescence analysis’, The Numismatic Chronicle 149: 33–58. Walker, D. R. (1976–8). The Metrology of the Roman Silver Coinage, Part I: From Augustus to Domitian (BAR Suppl. 5); Part II: From Nerva to Commodus (BAR Suppl. 22); Part III: From Pertinax to Uranius Antoninus (BAR Suppl. 40). Oxford.

13 Coinage and Metal Supply Bruce Hitchner

Matthew Ponting’s chapter above clariWes what we already knew: that the Roman state manipulated the silver bullion content of the denarius coinage under pressure. What it exposes quite neatly is the care and eVort that went into disguising this. But this also raises a number of interesting questions not only about the nature of Roman manipulation of the money supply, but about its desired impact on the latter and more generally on the Roman economy. SpeciWcally, when we speak of monetary manipulation, what activities should we consider to fall under this category? If we understand manipulation, broadly speaking, to be about the state’s desire to regulate, however crudely, the supply of money available in the general economy and indirectly to ensure that its revenue requirements are met, then I would add to the list, in addition to metallurgical manipulation, such activities as regulating the volume of emissions, debt cancellations and tax remissions, and even state loans.1 Let us now consider the purposes and impact of each. First, metallurgical and emission manipulation. Since the Roman state depended for money on hard metallic currency and did not borrow from the private sector to meet any deWcits, it was compelled to manipulate physically the coinage, in particular the denarius and aureus, as a means of meeting its public obligations, collecting taxes, and preventing deWcits, and there were only two ways of doing this: by increasing or reducing the precious metal content or annually 1 I would also add changes in the messaging on coinage in certain circumstances.

282

Bruce Hitchner

increasing or reducing the volume of emissions. As long as there was relative political and economic stability, and expansionist warfare was kept to a workable/proWtable minimum (booty ¼ costs of making war), the Roman state could produce suYcient silver and gold at a high standard through the emission of coinage that combined new metal and recycled older coin. However, political instability and major warfare, whatever the reason, placed immediate pressure on state revenues and Wscality. Since borrowing from the private sector was not seen as a viable political mechanism for addressing this problem (only the emperor could provide such relief, as other lenders, either individually or in combination, could be seen as a political threat to the emperor), the only option was to manipulate the coinage. Once this was done, as in the case of Nero’s manipulation, it became a legitimate (indeed the only legitimate) mechanism for oVsetting deWcits. However, in order to maintain monetary stability the Roman state was careful to ensure that denarius debasements were gradual. Silver content fell slightly from 94 to 88% between Nero and Antoninus Pius. However, the decline in silver content was much more signiWcant after that, falling to 78% under Marcus Aurelius, then to 74% by late in Commodus’ reign. The drop in silver content continued thereafter falling to 55% in the reign of Septimius, 51% under Caracalla, and about 45% under Elagabalus and Severus Alexander, but this decline was oVset from the reign of Severus onward by a substantial increase in the available volume of denarii (achieved through the introduction of a retariVed denarius and a pay increase for the army), which per year was higher than under other emperors from Nero onward except Vespasian. Hence, stability and public conWdence were, to all indications, maintained by a combination of small, carefully modulated debasements or manipulations until the crisis of the late second century forced the state to oVset more substantial reductions in silver content of the denarius with the emission of more denarii. At no time during this period did the state see Wt to radically reform or replace the currency, as it was forced to do in the later third century. All this being said, did the Roman state manipulate the coinage only in response to political or economic problems, or to problems of metal supply, or was it in any way in reaction to problems associated

Coinage and Metal Supply

283

with growth in a hard money economy? The Roman empire was a hard currency economy in which the supply of money varied, and was not linked in any way to positive population and economic growth. It is likely therefore to have suVered inXation from time to time, but also perhaps no less often deXation, i.e. a decrease in the general price level, over a period of time. DeXation will have meant a decline in the available amount of hard currency per person, in eVect making money scarcer, with a consequent increase in the purchasing power of each unit of currency. Debasement of the coinage would therefore be an eVective tool for spreading a given amount of specie around and spreading the costs to everyone. In other words, debasement expanded the money supply to keep up with transactions, for given velocity and prices (MV ¼ PT). Conversely, returns to higher silver or gold contents could be seen as eVorts to make money demand stay equal to money supply, and to ensure that the state beneWted from the seignorage. Productive eYciency driven by proWt motives could have promoted deXation in the empire by lowering the overall price of goods and increasing purchasing power in the Wrst and second centuries. Such increases in purchasing power might have stimulated diYculties in Roman society as many people’s net worth was held in illiquid assets such as homes, land, and other forms of private property. It could also have enhanced overall debt as the payments made in servicing debts would have represented a larger amount of purchasing power than they did when the debt was Wrst incurred. Consequently, deXation can be thought of as a phantom ampliWcation of the interest rate on a loan. To return then to the main question, we may speculate that manipulations of the coinage, in particular metallurgical manipulation and changes in the volume of coin emission until the mid-third century, were as much a consequence of overall economic growth as of individual political and economic events. From the standpoint of metallurgical manipulation, such manipulations further suggest that increases or decreases in metal extraction for minting purposes may have been closely coordinated with perceptions of inXationary or deXationary forces, the former requiring increased production of copper, the latter leading to increased demand for new silver and gold.

284

Bruce Hitchner

Let us turn now to non-physical monetary manipulation such as state loans, debt cancellations, and tax remissions. The Roman state injected money on an irregular basis into provincial economies by all three means in combination. This was possible because the government itself did not borrow money and built up monetary reserves against future expenditures. These revenues were stored both at Rome in the treasuries (aeraria and/or Wscus) and in provincial municipalities where taxes were collected. The most famous instance of a socalled state loan is Tiberius’ injection of one hundred million sesterces through banks and loans free of interest in ad 33 to avert a Wnancial crisis and restore liquidity to the Wnancial and real estate market at Rome.2 In fact, this looks much like a large private aristocratic injection of cash, not unlike earlier personal contributions made to the state by Augustus. More signiWcant for our purposes, and something to which Temin has recently drawn attention, are loans of state reserves arranged by Pliny the Younger on advice from Trajan.3 SpeciWcally, while governor of Bithynia, Pliny asked Trajan whether it might be possible to loan unused public monies, from the tax revenues collected locally at a rate below or at the standard 12% interest, or to parcel it out among local town oYcials (decuriones) even if they were not willing to receive it, to abate the capital shortage.4 Trajan responded by agreeing to the lower interest rate. However, he opposed compelling oYcials to borrow unwillingly as it was not consistent, in his view, with the ‘justice of our time’. Two things stand out here. First, Pliny’s reference to not leaving the money unemployed suggests that it may have been a regular practice for governors to make state tax revenues, collected locally, available for loans. Second, Trajan’s willingness to oVer the money at a lower interest rate is based, as Temin has pointed out, on market, not administrative 2 Tacitus, Annals 6.16–17; Dio 58.21.1–5; Suetonius, Tiberius 48.1; Rodewald 1976: 1–17; Duncan-Jones 1994: 23–5 and bibliography cited therein. 3 Temin 2004: 726–7; Pliny, Ep. 10.54. 4 Pecuniae publicae, domine, providentia tua et ministerio nostro et iam exactae sunt et exiguntur: quae uereor ne otiosae iaceant. On interest rates see Sherwin-White 1966: 635–6; Temin 2004. Rates varied, suggesting a ‘demand curve for loans’. In Africa and Asia 12% is the highest rate for charitable and municipal endowments. Columella 7.18.4 implies a rate of 6% for loans. Digest 50.10.5 sets a 6% rate for certain municipal bequests under Antoninus Pius. Non-redeemable loans for Trajan’s alimenta were at 5% at Veleia and lower in other places.

Coinage and Metal Supply

285

realities, which, in the end, allowed the state to beneWt Wnancially from the loans while also providing a break on interest rates for those who borrowed from the state. If this was indeed a regular practice, it represents an important state mechanism for adjusting the amount of money available in the local economy. Finally, tax remissions and debt cancellations could serve a similar purpose. Grants of tax immunity, remission, or abatement of tax arrears made to communities or individuals by Roman oYcials in the late Republic and emperors under the Principate frequently come under criticism by modern historians for being politically driven and thus lacking, from a modern perspective, a sense of both fairness and ‘economic rationality’.5 But there was a long and venerable history of rulers granting tax exemptions in antiquity and, to be frank, it is a practice that still Wnds its way into the tax and Wscal policies of modern governments in both the developed and developing world. But if we set aside issues of ethics and rationality in taxation for the moment, and apply a strictly Wnancial perspective to the practice, tax immunities and especially remissions may be seen as devices which increased liquidity, in certain circumstances, and thus the availability of capital in the towns and provinces of the empire. Indeed, there are two good reasons for arguing that tax immunities or remissions did not have an adverse eVect on state Wnances. First, there is some indication that imperial advisers assessed the potential impact of such grants on the budget.6 While this did not preclude emperors from ignoring their advisers’ advice and being over-generous (Nero’s grant of tax freedom being the most notorious example, Suet. Vesp. 8.4), it points to the existence of an informal, internal auditing process designed to limit the potential for making excessive tax remissions—hardly surprising in a government that worked on a non-borrowing cash basis. Second, and perhaps more importantly, the Roman government probably exceeded its anticipated oYcial tax 5 Duncan-Jones 1994: 59–63 for a discussion and partial list of tax reviews and remissions. Note in particular the lists in Digest 50.15.1; 50.15.8. Duncan-Jones (1994: 4) also charges the Roman government with ignoring economic rationalism ‘in its willingness to assign tax burdens in accordance with political privilege rather than capacity to pay’ in exempting some communities from taxation. 6 Suetonius (Titus 8), for example, reports that Titus promised more to his petitioners for tax remissions than his advisers said he could aVord.

286

Bruce Hitchner

targets on a regular basis, making any tax remissions a rather minimal cost to the state of doing political business in the Roman world.7 That imperial tax remissions may have had a positive eVect on the Roman economy can be seen in the impact of the Wscal policies of the emperor Hadrian. Hadrian remitted taxes totalling possibly 900 million sesterces.8 He also drastically reduced the taxes on imperial properties and eventually postponed money taxes on them, and did the same for mine operators starting new concessions in government-owned mines. Hadrian likewise encouraged the agricultural development of unused or virgin land in Africa and perhaps Spain. As a result, over the course of both his reign and those of his immediate successors, these measures seem to have resulted in both agricultural expansion and increased building activity across the empire, particularly in the west. Bibliography Duncan-Jones, R. P. (1994). Money and Government in the Roman Empire. Cambridge. MacMullen, R. (1987). ‘Tax pressure in the Roman Empire’, Latomus 46: 737–54. Rodewald, C. A. (1976). Money in the Age of Tiberius. Manchester. Sherwin-White, A. N. (1966). The Letters of Pliny: A Historical and Social Commentary. Oxford. Temin, P. (2004). ‘Financial intermediation in the Early Roman Empire’, Journal of Economy History 64: 705–33. 7 Drawing on his examination of Roman taxes in Egypt, Duncan-Jones concluded that ‘the amount extracted from the taxpayers [in the provinces] probably exceeded the oYcial target, by a substantial margin in many cases’. Duncan-Jones 1994: 59–63, esp. 63, also table 4.8 showing the Wfteen-year tax cycle and tax events, ad 41–180. 8 See MacMullen 1987: 741 and n. 19 drawing on ILS 309 ( ¼ CIL 6.967); DuncanJones (1994: 59 n. 63) considers the Wgure of 900 million sesterces to be too high.

14 Some Numismatic Approaches to Quantifying the Roman Economy Christopher Howgego

The programme of metal analysis being undertaken by Matt Ponting with Kevin Butcher is arguably the most important numismatic contribution currently being made to the quantiWcation of the Roman economy. A few general comments may go some way to explaining why this is so. In respect of the production of coinage, no mint records survive so that quantiWcation must be based on the study of the number of dies used to strike the coinage. One should not be too optimistic about progress in this area. Until the fourth century ad the Roman monetary system was complex, with regional coinages playing a part alongside the imperial.1 So any view of the monetary economy must engage with both. Even if this complexity is set aside there is not much to go on. There are hardly any die studies of major issues of coins in the imperial period.2 The sheer scale of the coinage makes such studies incredibly onerous. Large-scale die studies of imperial coinage may well have to await the development of an automated process, probably reliant on 3D imaging.3 In the meantime quantiWcation in the Roman imperial period relies on huge extrapolation from tiny samples, with margins for error which render conclusions of an economic nature very tentative 1 RPC; Howgego et al. 2005. 2 Howgego 2005a: 110. 3 Howgego 2005a.

288

Christopher Howgego

indeed.4 An assessment of the approximate scale of individual issues is within our reach, but this is a long way from engagement with the Quantity Theory of Money. The Fischer equation allows us to relate the quantity of money to prices and the scale of monetized transactions. The equation is a truism. The supply of money (M) times how hard it works (V for Velocity) equals the level of Prices (P) times the level of monetary Transactions (T). That is, MV ¼ PT. What does this mean in terms of the Roman economy? It is important to remember that the Fischer equation properly deals only with the monetized economy. It is therefore a potential source of confusion to substitute estimates of gross domestic product (GDP) for PT, if non-monetized production is included in GDP. Moreover, it is entirely likely that levels of monetization changed over time, so that an increase in the quantity of money might well have been taken up by an increase in the proportion of transactions which were monetized. And what should count as money? A case can be made for the existence of credit-money in addition to coined money in the Roman world.5 (The case rather depends on how one chooses to deWne money.) Credit was certainly available in many forms, records of debts might be transferred within certain circles, and transactions might take place without the physical movement of coin. Thus the use of credit must be factored into the MV side of the equation somehow, either as an increase in Money or as increasing the eVective Velocity of circulation.6 A signiWcant further problem is that in all periods Velocity is an unknown variable within the equation, and may only be calculated when the other terms may be taken as given (Money, Prices, and Transactions).7 So inevitably there are doubts about the calculation of Money from coin alone. Such questions about how to operate the Fischer equation in a Roman context may be illuminated only partially by recourse to comparative evidence from the English Middle Ages. For what it is 4 Note, for example, the slender evidential base available to Duncan-Jones 1994: 150–62. 5 Harris 2006. 6 Harris 2006; Howgego 1992: 12–16, 27–8. 7 Howgego 1995: 122–5; 1992: 12–16. For ways of considering Velocity in a medieval context, see Mayhew 1995a and 1995b.

Numismatic Approaches to QuantiWcation

289

worth the medieval evidence suggests that the use of credit tended to move in step with the money supply (rather than, for example, acting to make up for a shortfall of money), and that an increase in levels of monetization correlated with an enlarged supply of money and a decrease in velocity.8 For the Roman economy the Fischer equation is certainly a good tool to think with, but it is very uncertain what numbers should be plugged into it. Numismatic studies oVer other ways to make progress in quantifying the Roman economy, which might ultimately prove more fruitful. Two areas in particular look hopeful. One is the analysis and interpretation of already well-deWned patterns of circulation: to what extent did coins move about the empire (and indeed outside it) and how long did such movements take? The fourth century ad is the easiest to investigate in this regard, as there was a single style of imperial coinage produced at identiWed decentralized mints, but earlier periods provide useful evidence too.9 This type of material may be brought to bear not only on models such as that of Hopkins, but also, and perhaps more naturally, on the concept of connectivity as outlined by Horden and Purcell.10 It might address questions such as whether the Mediterranean is a natural unit of analysis, and of how connected, and how diVerent or similar, temperate Europe looks, as seen through the lens of coin circulation. It might allow us to investigate the proposition that the Roman world was exceptional in its degree of connectivity between the Mediterranean and temperate Europe, and to consider whether this was a factor in the exceptional performance of the Roman economy. If circulation is one fruitful area, metal analysis is another. That is why the work being undertaken by Ponting and Butcher is so important. Ponting and Butcher’s work is substantially a revision of David Walker’s three-volume Metrology of the Roman Silver Coinage.11 Walker’s work was magisterial, well-constructed, and innovative. It would be sad if the revision of his results caused his volumes to be less used. They contain a great deal of continuing value. Although his absolute results have been shown to be inaccurate, the trends he 8 Mayhew 1995a and 1995b. 9 Howgego 1994. 10 Hopkins 1980; Horden and Purcell 2000. 11 Walker 1976–8.

290

Christopher Howgego

charted are in general being vindicated. Much of his interpretation remains valid, both in terms of detail and in general observations. The few pages he penned on the moral dimension to Roman monetary policy deserve to be read by all.12 In a world which viewed itself in terms of golden ages which might decline to ages of iron there were inXuences on decision-making which went beyond what we might term hard economics. The attempt to improve the quality of the coinage at the end of Nero’s reign is not so surprising in the light of Roman morality. And it certainly falls into a pattern. All earlier debasements had been rapidly reversed. What is new is that this one could not be reversed for long. Improvements were later attempted by Domitian, Pertinax, Macrinus, and Gordian I–II/Balbinus and Pupienus. But from Nero onwards, none of the improvements lasted long. There is one key aspect of Ponting and Butcher’s approach which perhaps deserves further development so that the results may usefully be applied to economic history. Ponting states clearly that his aim as an archaeologist is to establish the original composition of the alloy mixed in the mint and used to make coins. There is every reason to believe that he achieves his aim. But to the economic historian, as opposed to the historian of metallurgy, the key aim is surely to establish the amount of silver in the coins, or perhaps rather the amount of silver which the Romans thought was in their coins. There are two determinants of silver content: Wneness and weight standard. In order to calculate the Wneness of the Wnished coin, the percentage we should be using is arguably not that of the core metal, but the weighted average of the coin as a whole (that is including the enriched layer). SuperWcially neutron activation is more suitable for this purpose, in that in provides an analysis of the entire coin. The qualiWcation ‘superWcially’ is necessary, because this technique analyses coins as they are now, so that it cannot diVerentiate between surface enrichment in the mint and subsequent chemical change (for example the leaching of coins by agro-chemicals while they are in the ground or by subsequent cleaning). It would be useful to have these theoretical concerns addressed directly. How signiWcant are they in practice? What is the best way to deal with them?

12 Walker 1976–8: vol. III, 106–10.

Numismatic Approaches to QuantiWcation

291

The second determinant of silver content is weight standard. Again, for the economic historian the actual standard used in the mint is perhaps more important than any notional standard. The weight of individual coins may be greatly aVected subsequent to minting by wear, damage, and leaching. The best approach is probably to weigh as many coins in as good condition as possible. It is also important to select coins from a range of hoards to exclude both selectivity of heavier coins by any given hoarder and chemically induced weight loss during burial and/or subsequent cleaning. Weights are a key component of what we want to know. Richard Duncan-Jones has done important work in this area.13 From Wneness alone we get a picture of irregular and sometimes dramatic drops in Wneness, say to 80% under Nero, and to 50% under Septimius Severus. This gives a somewhat misleading impression in respect of monetary standards. If one factors in declining weights, one gets a much smoother picture of a gradual decrease over time in the actual amount of silver in the denarius.14 Weights and Wneness were managed together. Turning to what Ponting has to say about metal sources, this is all new and interesting. It also relates to his other topic. For example, the high standard of Galba’s issues might well have something to do with where he got his metal from. The bullion might have come direct from the mines, or as bullion and plate from sympathizers.15 In order to further our understanding of metal sources we should in particular welcome the enlargement of Ponting’s programme of lead isotope analyses. The mixing of metal from diVerent sources skews results. It is only when we have repeat readings from a number of coins within a narrow isotopic range that we may believe that the signature is indicative of a particular source, rather than randomly generated by the mixing of metals from diVerent sources. If we know which sources were being used we can say more about inter-regional transport. It is fascinating to consider that British lead might have been used in reWning Spanish silver. Further work may allow us to make inferences from the cessation of some sources, and the opening up of new. It was surely a contribution to our understanding of the fourth century that a change in the platinum content of the gold coinage 13 Duncan-Jones 1994: 219–23. 14 Duncan-Jones 1994: 229, Wg. 15.7. 15 Plutarch, Galba 20, 2.

292

Christopher Howgego

demonstrated that a new source of gold became available from the 340s.16 Ponting and Butcher’s work has a bearing on a whole range of topics of interest to the economic historian. Four are listed below by way of example: First, the debasement of the coinage is arguably as good an indicator as we have of the Wscal inadequacy of the Roman empire. Under the Republic crisis-driven debasements, whether in the Second Punic War, the Social War, or in the run-up to Actium, were always reversed as soon as possible. Something changed. As mentioned earlier, the attempt to improve standards under Nero was short-lived, as were all subsequent attempts. Fiscality cannot defy gravity for long. The third-century crisis began early. Secondly, we need to know to what extent price change was driven by monetary change, and to what extent by other causes.17 In particular we may look forward to Ponting and Butcher’s forthcoming work on the second century. Rathbone has characterized Egyptian prices as static at least from ad 45 until they doubled between 160 and 190. Prices then remained level until 274/5.18 It will be of considerable interest to have greater deWnition on the second-century debasement in Egypt, which was broadly contemporary with the doubling of prices. We can then think about this in the context of the virtual hiatus in the production of debased silver coinage in Egypt from 170 to 180,19 in relation to what was happening with the imperial currency, and indeed against the background of the dark years of the 170s, the revolt of Avidius Cassius (if that is what it was), and the plague. On this topic Ponting and Butcher’s work can be joined up with the Roman Provincial Coinage Project currently being undertaken at the Ashmolean Museum.20 Thirdly, in thinking about the unity or diversity of the economy of the Roman empire, it will be useful to chart with greater accuracy the extent to which monetary change proceeded in parallel across the Roman world, and to what extent regions went their own way. Under 16 18 19 20

Morrisson et al. 1985: 92–5. 17 Howgego 1995: 121–35. Rathbone 1996. Broken only by issues for both Marcus and Commodus in year 17 (ad 177). See Roman Provincial Coinage Online http://rpc.ashmus.ox.ac.uk/.

Numismatic Approaches to QuantiWcation

293

Nero there was some coordination between the reform of the imperial coinage and reforms in Crete, Cappadocia, Syria, and Egypt, but the reforms were not identical.21 Was this typical of coinage reform in the Roman empire? Lastly, Romanists often take for granted the Wxed equation of one gold aureus to 25 silver denarii. To monetary historians of other periods the maintenance of a Wxed bimetallic system for two and a half centuries despite variable market rates for gold and silver is almost unbelievable. This stability and the surprise it evokes might be a clue to something special about the Roman economy and provide a reason why economic historians might want to take an interest in it. Part of the answer may be that, broadly speaking, the Roman economy was a monetized economy not surrounded by other monetary economies. Such a perspective raises questions of wider interest. We might, for example, consider to what extent observed monetary behaviour is driven internally, within an economy, and to what extent it depends upon economic pressures from outside. In order to investigate this topic, we should Wrst consider the extent, if any, to which manipulation of the coinage played a role in achieving stability. Here too Ponting and Butcher’s work has the potential to make a signiWcant contribution to our understanding of the nature of the Roman monetary economy. These four key topics stand to be illuminated by Ponting and Butcher’s work. There are many others, but this list should be suYcient to convey the importance of what is being done and to incite interest in further results. Bibliography Bowman, A. K., and Brady, M. (eds. 2005). Images and Artefacts of the Ancient World. Oxford. Britnell, R. H., and Campbell, B. M. S. (1995). A Commercialising Economy: England 1086 to c. 1300. Manchester. Burnett, A. (2005). ‘The imperial coinage of Egypt in the Wrst century ad’, in F. Duyrat and O. Picard (eds.), L’exception ´egyptienne? Production et

21 RPC vol. 1, ch. 6; Burnett 2005: 274.

294

Christopher Howgego

e´changes mone´taires en E´gypte helle´nistique et romaine. Actes du colloque d’Alexandrie, 13–15 avril 2002. Cairo, 261–77. Duncan-Jones, R. (1994). Money and Government in the Roman Empire. Cambridge. Duyrat, F., and Picard, O. (eds. 2005). L’exception e´gyptienne? Production et e´changes mone´taires en E´gypte helle´nistique et romaine. Actes du colloque d’Alexandrie, 13–15 avril 2002. Cairo. Harris, W. V. (2006). ‘A revisionist view of Roman money’, JRS 96: 1–24. Hopkins, K. (1980). ‘Taxes and trade in the Roman Empire (200 bc–ad 400)’, JRS 70: 101–25. Horden, P., and Purcell, N. (2000). The Corrupting Sea: A Study of Mediterranean History. Oxford. Howgego, C. (1992). ‘The supply and use of money in the Roman world 200 bc to ad 300’, JRS 82: 1–31. —— (1994). ‘Coin circulation and the integration of the Roman economy’, JRA 7: 5–21. —— (1995). Ancient History from Coins. London. —— (2005a). ‘The potential for image analysis in numismatics’, in A. K. Bowman and M. Brady (eds.), Images and Artefacts of the Ancient World. Oxford, 109–113. —— (2005b). ‘Roman provincial coinage online’, http://rpc.ashmus.ox.ac. uk/ (accessed 9 Oct. 2007). —— Heuchert, V., and Burnett, A. (eds. 2005). Coinage and Identity in the Roman Provinces. Oxford. King, C. E., and Wigg, D. G. (eds. 1996). Coin Finds and Coin Use in the Roman World. Berlin. Mayhew, N. J. (1995a). ‘Population, money supply, and the velocity of circulation in England, 1300–1700’, EHR 48: 238–57. —— (1995b). ‘Modelling medieval monetization’, in R. H. Britnell and B. M. S. Campbell (eds.), A Commercialising Economy: England 1086 to c. 1300. Manchester, 55–77. Morrisson, C., Brenot, C., Callu, J.-P., Barrandon, J.-N., Poirier, J., and Halleux, R. (1985). L’or monnaye´ I: PuriWcation et alte´rations de Rome a` Byzance. Paris. Rathbone, D. (1996). ‘Monetization, not price-inXation, in third-century A.D. Egypt?’, in C. E. King and D. G. Wigg (eds.), Coin Finds and Coin Use in the Roman World. Berlin, 321–39. Walker, D. R. (1976–8). The Metrology of the Roman Silver Coinage, Part I: From Augustus to Domitian (BAR Suppl. 5); Part II: From Nerva to Commodus (BAR Suppl. 22); Part III: From Pertinax to Uranius Antoninus (BAR Suppl. 40). Oxford.

Numismatic Approaches to QuantiWcation

295

Abbreviations RPC

A. Burnett and M. Amandry (eds. 1992–). Roman Provincial Coinage. London and Paris. Vol. I (1992): From the Death of Caesar to the Death of Vitellius (44 bc–ad 69) by A. Burnett, M. Amandry, and P. P. Ripolle`s. Supplement I (1998) by A. Burnett, M. Amandry, and P. P. Ripolle`s. Vol. II: From Vespasian to Domitian (ad 69–96) by A. Burnett, M. Amandry, and I. Carradice. Vol. VII: De Gordien Ier a` Gordien III (238–244 apre`s J.C.). 1. Province d’Asie, by M. Spoerri Butcher.

This page intentionally left blank

Part VI Prices, Earnings, and Standards of Living

This page intentionally left blank

15 Earnings and Costs: Living Standards and the Roman Economy (First to Third Centuries ad) Dominic Rathbone

1 . INTROD U C TION If we had plenty of data for costs and earnings from the Roman empire of the Principate (Wrst to third centuries ad) we could organize them in sets to show: 1. The cost- and wage-levels in diVerent areas of the empire in diVerent periods. 2. The diVerentials in each area and period between the living standards, that is costs over earnings, of diVerent groups, such as urban and rural workers, skilled and unskilled, state and private employees, and so on. Through analysis of these datasets we could try to answer questions such as: 1. How far was there an integrated market for goods and labour in the Roman empire? Or were there distinct economic zones at regional or even local level? 2. What were the living standards of diVerent groups, and how great were the diVerentials between them? 3. Are any signiWcant changes across time visible in market integration or living standards?

300

Dominic Rathbone

Our answers would contribute to a general understanding of the structure and workings of the Roman economy, perhaps including whether and when there was signiWcant economic growth, an issue to which I will return. Unfortunately we have very few such data for most of the Roman empire in the Wrst to third centuries, and even for Egypt nothing like the Delos temple accounts or the Babylonian astronomical diaries of the Hellenistic period. That is why Heichelheim’s statistical study Wirtschaftliche Schwankungen der Zeit von Alexander bis Augustus (1930) stopped with Augustus, while his 1954/5 article, with its extended overview into the seventh century, focused on the eastern Mediterranean. There have been collections of data for particular provinces and periods, notably the empire-wide survey edited by Frank (1933–40); Duncan-Jones (1974/1982) mainly on Italy and Africa; Drexhage (1991) on Roman Egypt; Sperber (1974/1991) on later Roman Palestine; plus now the collection by Szaivert and Wolters (2005) of all Wgures in literary sources. But the only studies of long-term change are Mrozek (1975) and Corbier (1985), who compare Italian prices of the Wrst century ad (Pliny, NH; Pompeii) with prices in Diocletian’s Edict on Maximum Prices of December 301. Their main interest was monetary inXation. So too I reexamined Egyptian prices to argue for price stability rather than inXation up to 274, while also making some other points.1 Heichelheim remains the only model for a long-term pan-Mediterranean analysis of ancient costs and earnings. He has long been unfashionable because of his default modernizing, but nowadays most of us are happy not to make an either-or choice between economic and sociological approaches. An ambitious project to collate and interpret price and wage data for the ancient world from Persepolis to Hadrian’s Wall is now being masterminded by van der Spek (Vrije Universiteit Amsterdam). This chapter draws on some of my ongoing research for that project to illustrate how the limited data we have can be used and what sort of conclusions we can and cannot draw. I discuss all the ‘normal’ wheat prices I know of the Wrst to third centuries along with some groups of data for living allowances and

1 Rathbone 1996, 1997.

Earnings and Costs

301

remuneration, and also a few fourth-century and later data which pose interesting questions of comparability.

2 . M E T RO LO G Y: THE A RT OF GETTING IT RIG H T It is tedious but essential to try to get the value of ancient units of measurement and ancient coinages right. When the data are few, small errors can matter. All previous lists are to a greater or lesser extent unreliable because of metrological inaccuracies and errors, an assertion which declares open season for the hunting down of mistakes in this chapter. We have no up-to-date handbook which is more generally useful and reliable than Hultsch’s Metrologie of 1862, revised 1882, but there have been many advances of detail since, not least in the works of Duncan-Jones. In my experience metrological excursuses provoke yawns, so I just state the equivalences which I think correct and use here: . 1 Roman libra (pound weight) ¼ 323 g. . 1 standard modius (mod.; dry measure) ¼ 8.62 l, holding 6.8 kg of Italian wheat. . 1 modius castrensis ¼ 1.5 standard modii. . 1 standard Roman-period artaba (art.; Egyptian dry measure) ¼ 4.5 modii ¼ 38.8 l, holding 30 kg of Egyptian wheat. . There were 16 asses ¼ 4 sesterces (HS) to the denarius (den.), and 25 denarii to the aureus. . In the Greek East the drachma (dr.) and in Egypt the Alexandrian tetradrachm (i.e. 4 Egyptian dr.) were equivalent to the denarius. . Roman silver and gold coins were minted at an oYcial target weight of so many to the libra:  84 silver denarii (¼ 3.85 g) to ad 64, thereafter 96 (¼ 3.36 g).  40 gold aurei until 64 (¼ 8.08 g), thereafter normally around 45 (¼ 7.18 g).  72 late imperial gold solidi (¼ 4.49 g). Diocletian’s Edict priced a pound of silver at 6,000 denarii and a pound of gold at 72,000 denarii.

302

Dominic Rathbone

The basic methodological problem is how to compare earnings and prices from diVerent systems. Heichelheim created baselines for indexing prices from diVerent systems by selecting pairs of more or less contemporary prices as equivalents, for instance the later secondcentury bc wheat prices of 6˜ asses per modius at Rome and 2` drachmas per artaba in Egypt.2 This procedure is rather arbitrary and unsafe when data are so scarce. Mrozek converted prices of the Wrst century ad and 301 into the weight of gold coin which they represented. Corbier used prices in denarii, and calculated their percentage rise or fall by converting the Wrst-century prices Wrst on the 1:72 ratio of the Augustan denarius to the Edict’s denarius in terms of face value per pound of gold, and then on the 1:64 ratio of the Neronian denarius. Corbier’s method is just a more complex way of doing what Mrozek did, and only allows measurement of aggregate change between two points with no allowance for intermediate variations. I use the same procedure as Mrozek with some reWnements. First, where necessary I use a large standard commodity unit, for example one tonne of wheat, so that the bullion values are in nice clear numbers. Secondly, I use silver as my standard of value from the mid-third century bc to mid-third century ad, and gold from the earlier fourth century ad onwards, with some earlier Wgures also calculated in gold to make a link. This reXects the historical fact that the monetary system of the Republic was based on the denarius (no gold coins), that of the Principate on the denarius with the aureus (linked gold and silver coins), and that of late antiquity on the solidus (no silver coins). It is fairly straightforward to compare values in gold and silver because from Augustus to Diocletian there was a standard oYcial 1:12 weight ratio between the two. Thirdly, I allow for the reforms of the denarius by using its diVerent successive target weights. However, I make no allowance for debasement, unlike Lo Cascio who adjusted third-century prices from Egypt in line with the progressive reduction in weight of the aureus, on the assumption, for which there is no evidence, that these reductions corresponded accurately to increases in the monetary value, or purchasing power, of gold bullion.3 I prefer the much simpler explanation that the weight 2 Heichelheim 1930: 74; cf. 1954/5: 501 n. 1. 3 Lo Cascio 1997.

Earnings and Costs

303

reductions were the result of shortage of bullion for new minting, and also represented a proWt for the state in that the new aurei, like the debased denarii, retained their oYcial face value. This is because, once the denarius was tied by Augustus to a gold coin-standard in a ratio of 1:25 which was not altered as long as that coinage system lasted, it became a token coin (as did the occasional silver fractions), and increasingly so as it was debased; the same is true of the third-century silver antoninianus and the underweight aurei. The Egyptian price data show that the Alexandrian tetradrachm retained its face value, which was oYcially equivalent to the denarius (and hence in the same relationship to the aureus), until Aurelian’s reforms because, in my view, the government accepted its coinage back from taxpayers at its face value and did not raise basic tax-rates. In short, Roman coins were pegged to Roman taxation. Of course there were variations, sometimes sharp if brief, in the free market values of silver and gold bullion, but the oYcial ratio, backed by the reality of taxation, seems in practice to have dominated normal market exchange. This was possible because the Roman government, through imperial ownership of mines, was the monopoly producer of gold and silver bullion as well as coins, and because there were no competitor coinages of any signiWcance in neighbouring states. Jongman is scathing about this procedure for indexing: ‘converting prices into weights of precious metal . . . is theoretical nonsense’, because bullion values themselves are variable; he prefers calculation of earning power in wheat equivalent.4 I may be missing something, but this seems to face the same objection: wheat prices vary too. We just have to be precise about what we think our calculations show. Ideally we would check variances over time between a number of diVerent costs, including bullion and earnings.

3. WHEAT PRICES In Table 15.1 below I list the data I can Wnd for ‘normal’ wheat prices in the Roman empire. The problem of paucity of data is immediately 4 Jongman 2007.

304

Dominic Rathbone

Table 15.1 Wheat prices —gold/silver bullion values per tonne (1,000 kg) 1. 123 – 58 bc Rome Livy, Epit. 60; Schol. Bob. 2.135 St.; Cic. pro Sestio 55 Subsidized price of Gracchan frumentatio was 61=3 asses per mod., presumably from 5 mod. (monthly ration) for 2 den., i.e. HS 1.6 per mod. ¼ 226 g silver 2a. 74 bc Sicily Cicero, in Verrem 2.3.213–16 Wheat cheap in 76, but expensive in 75 because taken (Cicero was quaestor) to meet shortage at Rome. In 74 (shortage at Rome again) pre-harvest price reached 5 den. (¼ HS 20) per mod. ¼ 2,827 g silver commutation of frumentum in cellam at 3 den. per mod. was fair pre-harvest, but unfair after, i.e. price had dropped below HS 12. ¼ 1,696 g silver 2b. 73 – 71 bc Sicily Cicero, in Verrem 2.3.163, 173–4; 188–97, 201–2 73 bc Lex Terentia Cassia revived frumentatio at Rome: Verres was instructed to purchase a second tithe at HS 3 per mod. and frumentum emptum at HS 3.5; at one stage the average price was HS 2 to 2.5 per mod.; later(?), the senate authorized Verres to buy frumentum in cellam at HS 4; the normal market price across all Sicily was then(?) HS 2 to 3 per mod. (HS 2 to 3) ¼ 283 to 424 g silver 3. 2nd – 1st c. bc (?) Rome Dionys. Hal. 12.1.12 (wrote late 1st c. bc) Shortage of 439 bc: when price had reached 12 dr. (¼ HS 48) per mod., Maelius sold grain at 2 dr. (HS 8) per mod. (I take these prices to be inserted from contemporary Rome by Dionysius or a second-/Wrst-century annalistic account.) (HS 8) ¼ 1,130 g silver 4. ad 19 Rome Tacitus, Ann. 2.87 Shortage: Tiberius set maximum price and paid merchants compensation of 2 nummi (HS) per mod. 5. ad 64 Rome Tacitus, Ann. 15.39 Fire caused fear of shortage: Nero reduced the price to 3 nummi per unit, i.e. HS 3 per mod. (¼ 36 g gold) ¼ 424 g silver 6. ad 70s (?) Pompeii Duncan-Jones, 1974/1982: 146 (corrected) Prices equivalent to HS 3 and, probably, HS 7.5 per mod. ¼ 370 to 926? g silver 7. c. ad 100 – 150 (?) Forum Sempronii CIL 11. 6117 ( þ p. 1397) Shortage: benefactor provided wheat at 1 den. (HS 4) per mod. (¼ 42 g gold) ¼ 495 g silver 8. ad 93/4 Antioch, Pisidia AE 1925.126b ¼ Docs. Flav. 464.ii Winter shortage: Roman governor of province ordered surplus stocks of wheat to be sold at not more than 1 den. per mod. Before the shortage the price had been 8 or 9 asses (HS 2 to 2.25) per mod. ¼ 245 to 278 g silver 9. late 1st – early 3rd c. ad Judaea Sperber 1974/1991: 102 (rabbinic texts) Wheat cheap at 1 den. per se’ah, expensive at 4 den. per se’ah. If se’ah ¼ mod. (so Sperber), range is HS 4 to 16 ¼ 495 to 1,979 g (?) silver If se’ah ¼ mod. castr. (cf. Heichelheim, in Frank 1938: 181), range per standard mod. is HS 2.67 to 10.67 ¼ 330 to 1,319 (?) silver 10. ad 80s – 160s middle Egypt Rathbone 1997 ‘Farm-gate’ prices, range of 6 to 12 dr. per art. ¼ 165 to 330 g silver (¼ 21 g gold) ¼ 247 g silver median 9 dr. per art. ¼ HS 2 per mod.

Earnings and Costs 11. ad 190s – 250s middle Egypt Rathbone 1997 ‘Farm-gate’ prices, range 12 to 20 dr. per art. median 16 dr. per art. ¼ HS 3.56 per mod.

305 ¼ 330 to 550 g silver (¼ 38 g gold) ¼ 440 g silver

12. ad 301, December eastern provinces (?) Diocletian, MPE 1.1a G Maximum 100 denarii per mod. castr. (and 60 den. for barley). ¼ 44 g gold ¼ 528 g silver 13. 4th – 6th c. ad Egypt Carrie 1997 range from 12 to 8 art. per solidus ¼ 12 to 18 g gold median 10 art. ¼ 45 mod. per solidus ¼ 15 g gold 14. c. ad 493 – 526 Italy Anon. Valesianus 12.73 Prosperity under Theoderic: wheat normally 60 mod. (Italici?) per solidus.¼ 11 g gold

apparent. Most of the wheat prices recorded outside Egypt are the result of special circumstances. My normal prices include state subsidies and Diocletian’s maximum price; what I exclude are the sometimes fantastic prices cited when severe shortages were feared or happened. If I had included them we would have seen that when people had little or no hope of extra supplies, the sky was the limit, while price rises during local shortages when outside supplies could be acquired were much less sharp, and that Roman levies for special needs often provoked regional shortages and price rises, as attested, for example, in Sicily in the 70s bc (item 2a) and at Antioch when the emperor Julian wintered there with his army in ad 362/3.5 None of these points is new or surprising, and they can be illustrated just as well by the many cases for which no prices, whether real or exaggerated, are cited. Our best sets of data come from Egypt (items 10–11, which simply present the results of an earlier study by me). We like to deal with an average or normal price. These data show that the norm is seasonal and inter-annual Xuctuation between certain parameters: in the 80s to 160s between 6 and 12 drachmas per artaba. The median value of 9 dr./art. is useful, but should be read with þ /  warning signs. The Egyptian data also show the danger of calculating unilinear long-term trends, whether as Corbier did or by using linear regression.6 At some stage between the ad 160s and 190s there was a brusque doubling of the prices of wheat and all other reasonably attested commodities and also of wages, from 5 Jones 1964: I. 446.

6 Corbier 1985.

306

Dominic Rathbone

which a new sixty-year period of price and wage stability emerged. I have argued that this was a result of economic dislocation caused by the Antonine plague, and also that something similar probably happened throughout the rest of the empire.7 Not everyone is convinced. The lack of third-century prices from other provinces makes the question of typicality unanswerable. That brings us to regional price levels. The median price of wheat in Sicily of the 70s bc (345 g; item 2b) was 42% higher than the median in Egypt of the ad 80s to 160s (item 10), with some overlap in their normal ranges (almost a Wfth of their combined ranges), which does not seem implausible granted the relative size and fertility of the two provinces. In comparison it is striking that the normal price at Pisidian Antioch in ad 93 (item 8) falls squarely within the Egyptian price range, and the apparent price range in Roman Judaea (item 9) is also comparable, at least in the later second to third centuries (also assuming that the se’ah was equivalent to the modius castrensis). Pisidian Antioch had a productive hinterland, and Judaea could perhaps draw on Syrian supplies, but the correspondence hints at the emergence in the early empire of a common band of normal prices for wheat in the main coastal areas of the eastern Mediterranean, that is the nucleus of an integrated regional market. Diocletian’s Price Edict presupposes general market integration, at least as regards wages and wheat and some other goods in the coastal areas of the eastern provinces (Section 7). Diocletian reminds us that imperial and civic intervention was always an important factor. The imperial state collected wheat as tax and rent and by compulsory purchase to supply its armed forces, for the regular distributions at Rome (frumentationes), and for ad hoc subventions in Rome and other cities (altogether the annona); civic authorities also tried to maintain local supplies at reasonable prices. The wheat distributed to all male citizens in Rome from 123 to 58 bc was sold at an amazing subsidy (item 1), even lower than the local price in Sicily (item 2b), and from 57 bc it was distributed free. In 2 bc Augustus capped the number of recipients at 200,000,8 somewhat over a half of the then adult male population, in the interest, so Suetonius tells us,9 of farmers and merchants (and the treasury). So 7 Rathbone 1996.

8 Res Gestae 15.4; Dio 55.10.1.

9 Aug. 42.

Earnings and Costs

307

too in ad 19 Tiberius was careful to compensate merchants; his compensation perhaps implies that his subsidized price was HS 4/ modius (item 4). Diocletian’s Edict on Maximum Prices is another example of this mentality, albeit an extreme one and designed, as his preface makes clear, to beneWt the soldiery rather than the urban population. The eVects of state intervention may not always have been positive, but in the long term it was conducive to the belief in and reality of an integrated market because it reinforced trade links and common expectations. An indication of these common expectations is iconic prices, that is prices widely used to typify urban well-being. In the Republic the iconic price of wheat for the urban consumer in happy times was one as per modius. Around ad 100 Martial10 put a bitter rural twist on the trope, perhaps criticizing the pandering to urban populations: ‘When an amphora sells for twenty asses and a modius for one as, the farmer is drunk, stuVed and penniless’. The data suggest that Romans of the Wrst and early second centuries ad thought a fair urban price was HS 4 per modius, that is an iconic one modius for one denarius. This was the price set by the benefactor at Forum Sempronii (item 7), the Roman governor at Antioch (item 8), and perhaps Tiberius in ad 19 (item 4), while Nero, typically, had to go one better in ad 64 (item 5). Rabbinic writers use the same Wgure, although perhaps with the modius castrensis (item 9). The price of one denarius per modius is implicit in the deduction of 240 denarii for rations from soldiers’ pay to ad 84, assuming that they received 5 modii a month (Table 15.2), also in Seneca’s allowance for a slave at Rome, assuming a half-andhalf package, and might explain why the alimenta allowances are almost all multiples of HS 4 (both Table 15.3b). The tradition persisted: the maximum price for wheat set by Diocletian in 301 of 100 denarii per modius castrensis, whether or not it was unrealistically low (Section 7), was clearly an iconic price which was probably used as the template for Wxing other maxima like the daily wages of 25 and 50 denarii. What was the normal range of market prices in Rome and Italy of the Principate? Pompeii and Herculaneum have been a big disappointment in terms of price data: just a couple of graYti of uncertain 10 12.76.

308

Dominic Rathbone

interpretation (item 6). So scholars have turned to the state prices. Frank guessed that the Gracchan price was half the market price in Rome,11 and Duncan-Jones suggests the same for Nero’s price in 64,12 also arguing from milling costs for a price in imperial Rome of nearer HS 6 than 8 per modius. Somewhat inconsistently DuncanJones takes the benefactor’s price from Forum Sempronii of HS 4 per modius (item 7) to represent the norm for Italy rather than half the normal price.13 I add a possible reXection of Wrst-century bc experience at Rome from Dionysius of Halicarnassus’ elaboration of a story about early Rome (item 3), that wheat when short might reach HS 48 per modius while HS 8 per modius was a reasonable price. My hunch is that prices in central peninsular Italy followed prices in Rome, with a normal price range of say HS 6 to 10 or 12 per modius, comparable in variability to the normal price ranges attested in Roman Egypt (items 10, 11). That is as low as is plausible, in free market terms, in comparison with the Sicilian and Egyptian prices. The conventional estimates of wheat yields, both probably too low, are 1:4 for Roman Italy, 1:10 for Egypt, a threefold net diVerence. We must also allow for transport costs. But how far was there a free market in grain? Kessler and Temin have argued from regression analysis of six pairs of wheat prices at Rome and diVerent provincial locations that the Roman empire had a single integrated market for wheat driven by prices at Rome;14 prices elsewhere simply related to distance from Rome. So too Hopkins envisaged that ‘Rome was at the peak of a pyramid of rising prices’.15 Conversely Erdkamp, following the Persson thesis that the high variability of grain prices in individual localities in early modern Europe indicates a limited market in grain incapable of equalizing local surpluses and shortages, has argued that there was little private trade in grain in the Roman empire, and hence imperial and civic authorities were obliged to intervene to deal with local shortages.16 However, these two contrasting views both work from an unreasonably abstract model of market eYciency. I have argued that the variation in Egyptian wheat prices does not indicate an inadequate market but was determined principally by Xuctuations in yields due to the Nile inundation and was 11 Frank 1933: 192. 12 Duncan-Jones 1974/1982: 345–6. 13 Duncan-Jones 1974/1982: 50. 14 Kessler and Temin 2005. 15 Hopkins 1995/6: 220. 16 Erdkamp 2005.

Earnings and Costs

309

limited by Egypt’s excellent storage and transport facilities.17 Erdkamp’s assertion that most grain was moved by the state rather than by private trade is simply implausible. Kessler and Temin, apart from the problem that their data are selective and Xawed, ignore diVering local and regional patterns of production and demand and also transport circumstances. Hopkins’s model is more historical: the provinces exported goods to earn back what they paid Rome in taxation. But both sides probably overestimate the vampiric suction of Rome. Did high yields in Egypt, for instance, mean cheap wheat for Rome rather than more prosperous Egyptians? The issue of what may be termed ‘entitlement’18 needs more thought. We have seen hints of the existence of regional price levels: one in the eastern Mediterranean and another in Rome and Italy, with a fourfold diVerential between them which roughly accords with relative productivity and transport costs. If we had more data we could test and nuance this, and might well end up with a picture of several interlocking price zones. We have also seen that the Romans believed that their empire had an overarching common market, and that imperial and civic intervention helped to make this belief a fact by acting as a damper on the upper range of urban price levels. Lastly, a look towards late antiquity. It is sometimes said that in the long term in the later Roman empire, indeed from the early Principate, the price of wheat remained stable relative to the value of gold (and hence silver). So Mrozek and Corbier, for example, though they stopped in 301 and compared Diocletian’s maximum price with a low price of HS 4.19 This might imply that there was not much of a free market in grain in the Roman world. Already it is clear that the prices from Egypt (items 10, 11) do not square with this: wheat was almost twice as expensive in terms of silver and gold after the Antonine plague than it had been before it. But even more extraordinary is the drastic collapse in wheat prices in late antiquity, or, to put it the other way round, the massive increase in the purchasing power of gold (items 13, 14). For Egypt, for everywhere (I do not list all cases here), the levels sink well below what they had been before the Antonine plague. Despite the reasonable distribution of the evidence, do we Wnd this suspicious? Could the beneWts of 17 Rathbone 1997.

18 Sen 1981.

19 Cf. Sperber 1974/1991.

310

Dominic Rathbone

Theoderic’s rule have made Italy more fertile than Egypt? To identify and assess possible answers we would need to carry out a much broader comparison of prices and earnings between the Principate and late antiquity.

4. MILITARY SALARIES The Roman army was the single largest employer in the Roman world: with say 350,000 men in a total population of 50 million, it employed 2% of adult males. It also sets a handy benchmark because it had a single basic rate of pay for legionary and auxiliary infantry, which can be reconstructed from the literary sources and military accounts from Egypt,20 and because soldiers were stationed throughout the empire. Some general points are worth making. First, soldiers were paid the same sum throughout their career. There were no increments and very little promotion: only about 3% became principales on 1.5 or two times basic pay. OYcers from centurion up were mostly recruited directly from the municipal gentry and were paid enormously more than soldiers. Second, soldiers were paid the same in all provinces, irrespective of local price levels, and supplies were deducted at the same standard rates. That is another sign that the Romans thought that in broad terms their empire had an integrated economy. It also facilitated transfers of men and units between provinces. In reality soldiers were probably better oV in real terms in less developed frontier provinces. The urban cohorts in Rome, on 1.5 times basic pay, can hardly have been better oV in real terms than an ordinary legionary, and it is clear why Augustus set praetorian pay at three times the basic rate, double the Republican norm. Third, pay rates were stable for long periods, which implies that before the Antonine plague there was little observable price inXation. Domitian’s pay-rise in ad 85 of 33% implies, if we assume that none of it represented a rise in real terms, a compound rate of inXation of 0.2% per annum over the 130 years since Julius Caesar. Of course it may 20 More detailed argument in Rathbone 2007a.

Earnings and Costs

311

Table 15.2. Pay in the Roman army 2a. Legionary/auxiliary infantry pay Augustus to Domitian (ad 84)

gross HS 900 p.a. wheat equivalent at HS 8/mod. ¼ 765 kg deductions 240 rations 36 boots 30 hay net 594 (66%)

ad 85 to Septimius Severus

gross HS 1,200 p.a. wheat equivalent at HS 10/mod. ¼ 816 kg deductions 328 rations 48 boots 40 hay net 784 (65%)

214(?) – ?

gross HS 3,600 wheat equivalent at HS 20/mod. ¼ 1,224 kg

ad 301 (P.Panop. Beatty 2) gross 1,200 den. stipendium þ 600 den. ‘cost of annona’ p.a. also donatives of 10,000 den., perhaps 20,000 den.wheat equivalent at 100 den./mod. castr. ¼ 185 þ 1,020 or 2,040 kg 2b. DiVerentials 1.5  basic: principales (few; some 2  ), cavalrymen, urban cohorts 3  basic: praetorians—plus free wheat from ad 65 15  basic: ordinary centurions

have been a reaction to a recent sharper rise in prices, perhaps after Nero’s coinage reforms of ad 65 and the civil war of 68–9, but even a price increase of 33% over Wfteen years would imply a compound rate of under 2% a year and a century of price stability up to ad 65. The basic rate of pay from Julius Caesar to Domitian was not generous compared to civic oYcials at Urso or private employees (Sections 5 and 6), but seems to lie in the middle range of civilian earnings. Even if we had adequate data, it would be diYcult to convert soldiers’ earnings into wheat equivalent because their pay was uniform whatever the local range of wheat prices. In Table 15.2 I give guesstimate prices for Italy which serve only as a possible index of change in purchasing power. Domitian’s increase in ad 85 may

312

Dominic Rathbone

have upped the real value of military pay slightly, but he also increased charges for food and clothing by a third. If Septimius Severus doubled pay for all soldiers, he did no more than compensate them for the doubling of prices sparked oV by the Antonine plague. The Wrst real improvement to the salaries of Roman soldiers was Caracalla’s rise of (probably) 50%, but even that only put them on par with the remuneration of urban craftsmen. It is a mystery how soldiers made the wealth they could apparently accumulate to judge from their tombstones. Booty was rare and so, before the later third century, were donatives. Soldiers did receive a discharge bonus of HS 12,000, at least in theory, which up to ad 84 represented a third of their total remuneration, and nearer a quarter thereafter, but I also suspect that many or most were involved in private business ventures alongside their military service. Lastly, it is patent that despite Diocletian’s extraordinary eVorts to cap prices, the stipendium had become worthless by the fourth century, and it is clear that soldiers were utterly reliant for their remuneration on regular and generous donatives.

5 . S TATE CIVIL IAN E M P LOY E ES The state paid its civilian employees rather better. In Urso, a colony founded by Caesar in Spain in the 40s bc, the town charter set the annual salary for clerks (scribae) of the senior magistrates at HS 1,200 p.a. and for Xunkies (lictors) at HS 600.21 Other staV were to be paid between HS 300 and 700 per annum. Most, perhaps all, of these posts were part-time. The charter was re-inscribed in the late Wrst century ad, but these are probably the original salaries because the common Flavian charter for Spanish cities leaves salaries to the discretion of each town council;22 again a lack of noticeable inXation is implied. At a price of HS 6 per modius (a pure guess for Wrst-century ad Spain), the salaries for senior clerks and lictors were worth 1,360 kg and 680 kg of wheat respectively. In the mid-second century ad at the imperial quarry of Mons Claudianus in the eastern desert of Egypt 21 Roman Statutes I 25, ch. 62.

22 Ch. 73.

Earnings and Costs

313

the quarrymen, masons, and smiths hired from outside were paid 47 dr. monthly (¼ 564 dr. p.a.), plus a free artaba of wheat (worth around 8 dr.) and free wine.23 In eVect their gross remuneration was 720 dr. a year, that is 2 dr. a day, the common rate at that time for private craftsmen in the Nile valley, which at a price of 9 dr. per artaba would have purchased 2,400 kg wheat. For the top imperial oYcials, I present again the Wgures Wrst compared by Jones24 as amended in Garnsey and Humfress.25 In the Principate the proconsuls of Africa and Asia, in oYce for a single year (most other governors averaged three years), received an annual allowance of 250,000 denarii, equivalent then to 250 lbs gold; at a price of HS 8 per modius, this would have purchased 850,000 kg wheat. Procurators in the Principate, typically in any post for about three years, received annual salaries of from 15,000 to 75,000 denarii p.a., equivalent then to 15 to 75 lbs gold. In the fourth century ad the annual allowance (rations) of the governor of Lower Libya was commuted to 400 solidi, equivalent to 5.6 lbs gold, with which he could have purchased, at the presumably low oYcial commutation rate of 40 modii per solidus, 110,000 kg wheat. Under Justinian the Praetorian Prefect was paid 100 lbs gold per annum, which at 40 modii per solidus could have purchased almost 2,000,000 kg wheat. Fourth-century governors did not earn as much, in terms of gold, as the most junior procurator of the Principate. Hence, say Jones and Garnsey and Humfress, their enthusiasm for bribes. I would tend to reverse the proposition: oYcials of the Principate were paid huge salaries, which is why the elite was so keen to hold oYce, but typically they served for short terms, so were keen on bribes as well. Perhaps the fourth-century ‘bureaucracy’ in fact gave more government for less cost. When in the mid-sixth century Justinian increased senior administrative salaries, even the Praetorian Prefect received far less gold than the old proconsuls, but the increased spending power of gold, even if we used an arguably more realistic price of 20 modii per solidus, made him at least their equal in real remuneration.

23 Cuvigny 1996. 24 Jones 1964: 396–401, 644, 677. 25 Garnsey and Humfress 2001: 40–1, 45–6.

314

Dominic Rathbone 6. PRIVAT E EMPLOYEES

A survey of the better known cases of monthly wheat rations and living allowances gives us an idea of what the Romans thought was a ‘living wage’ against which we can evaluate attested cash wages. In Rome, and hence in its empire, the frumentatio of 5 modii a month (408 kg p.a.) seems to have set a norm which phased out the mid-Republican tradition of 4 modii. A ration of 5 modii per month for all soldiers is perhaps implied by the sum deducted for food from their salary up to ad 84, and is a better Wt than 4 modii in accounts like AE 1998.838 (Carlisle, 80s ad). Egypt retained its traditional 1 artaba a month, now made equivalent to 4.5 modii, which was the standard ration for all Roman soldiers stationed there. Maybe this is a hint of a rise in living standards over the second to Wrst centuries bc. The alimenta are not much help; at best they may indicate what Romans thought the average food costs of pre-adults were. I guess that in Italy of the Wrst to second centuries an adult could not subsist on less, in monetary terms, than HS 480 p.a., that is HS 1.33 a day, which is the modal amount in the subsistence allowances cited in the legal sources (item 3b.4), and at HS 6 per modius would have purchased 544 kg of wheat p.a. The diYcult question is what the average wage was. As we might have expected, urban and skilled workers are better paid than rural workers. The perhaps surprising twofold diVerential of Diocletian’s Edict is supported by the two stray Wgures we have for late Republican Italy (items 3c.1–2) and the Egyptian data (items 3c.4b–c) which point to twofold to threefold diVerentials. It looks as though investment in training produced a signiWcant economic reward, presumably because trained craftsmen had more clout in negotiating their pay. Our one Wgure for wages at Rome (item 3c.2), at roughly twice minimum subsistence, was relatively poor. If it is in fact typical, it implies that free baths (you had to pay in Egyptian villages), food handouts, and the like were an economic necessity for the unskilled inhabitants of Rome. The net remuneration of rural labour in Egypt, although low in absolute terms, was not only, in terms of purchasing power, much better than at Rome, but also seems to have improved from the second to third century, even

Earnings and Costs

315

Table 15.3. Living allowances and remuneration [square brackets denote implied rates] 3a. Monthly wheat rations 5 modii 123 bc – 3rd c. ad frumentationes ad 60s slave allowance [Principate all soldiers(?) [ boys’ alimenta

¼ 408 kg p.a. Rome, Oxyrhynchus Rome (Seneca, Ep. Mor. 80.7) general] Terracina]

4.5 modii (¼ 1 artaba in Egypt) earlier 2nd c. bc summer Weldworkers Italy (Cato, de Agric. 156) Principate infantry, quarrymen, farmhands, etc. Egypt 4 modii earlier 2nd c. bc mid-2nd c. bc [Principate

¼ 326 kg p.a. winter Weldworkers Italy (Cato) infantry general (Polybius 6.39.13) boys’/girls’ alimenta Veleia/Terracina]

3 modii earlier 2nd c. bc [Principate

light rural workers Italy (Cato) girls’ alimenta Veleia]

3b. Allowances 1. ad 60s Rome (Seneca, Ep. Mor. 80.7) allowance for slave: 5 modii wheat (at HS 4?) and 5 denarii (HS 20) per month 2. c. ad 100 alimenta in Italy (Duncan-Jones 1974/1982: 172) Terracina (private): boys HS 20, girls HS 16 per month Veleia (imperial): boys HS 16, girls and illegitimate boys HS 12, illegitimate girls HS 10 3. c.175–180 alimenta in Africa (Duncan-Jones 1974/1982: 102) Sicca (private): boys HS 10, girls HS 8 per month 4. Principate 13 subsistence sums in legal cases (Frier 1993) median/modal value HS 480 p.a. 3c. Private wages 1. mid-2nd c. bc, Italy: workman assembling olive mill, HS 8 per day (Cato, de Agric. 21.5); can buy (at HS 4) 4,900 kg wheat p.a. 2. 70s bc, Rome: casual labour HS 3 per day (Cic., pro Rosc. Com. 28); can buy (at HS 6) 1,220 kg wheat p.a. 3. ad 163 and 164, Dacia: miners’ contracts (cf. Cuvigny 1996) no. 9: HS 360 for 388 days, i.e. HS 1 per day? at HS 3/modius ¼ 820 kg wheat p.a. no. 10: HS 280 for 179 days, and HS 40 for children; together HS 2 per day? (Continued)

316

Dominic Rathbone

Table 15.3. (Continued) 4a. mid-2nd c. ad, Fayum (Egypt): rural labour, sample modal average 1.2 dr. per day at median 9 dr./art. can buy 1,440 kg wheat p.a. or, high 2 dr. per day at high 12 dr./art. can buy 1,800 kg wheat p.a. 4b. mid-3rd c. ad, Fayum: rural labour, sample modal average 2.3 dr. per day at median 16 dr./art. can buy 1,550 kg wheat p.a. or, high 4 dr. per day at high 20 dr./art. can buy 2,160 kg wheat p.a. 4c. mid-3rd c. ad, Fayum: rural craftsmen typically earn 4 dr. per day at median 16 dr./art. can buy 2,700 kg wheat p.a. shipwrights in city earn 7 dr. per day (sawyers 8 dr.), can buy 4,730 kg 5. ad 301, Diocletian, MPE 7.1a–13 G: rural worker with rations, maximum 25 den. per day, can buy (unadjusted) including rations (920) 1,290 kg wheat p.a. craftsman with rations, typical maximum 50 denarii per day, can buy (unadjusted) including rations (1,840) 2,200 kg wheat p.a. shipwright with rations, maximum 60 denarii per day, can buy (unadjusted) including rations (2,200) 2,570 kg wheat p.a.

before we take into account the decline in real value of taxation in cash caused by monetary inXation. This is what Scheidel argued,26 but the data need more careful checking, and still there is the question of what the implications really are. It is also worth noting that Egyptian workers were in real terms better oV if both wages and wheat prices were high than if they were both low. Dacian miners, in contrast, seem to have been very ill-paid, even if we assume a low local price of wheat. There seems to have been more of a free market and more regional variation in wages than in wheat prices, which implies a considerable regional variation in living standards. But the data outside Egypt are pitiably thin. The Diocletianic maximum wages are not directly comparable with earlier data because they are all for employment with rations as well as a cash wage. We can adjust them approximately by assuming standard rations of 3 modii castrenses a month worth a maximum of 300 denarii, that is by adding 10 denarii a day. The adjusted maximum wages for rural labour still remain relatively low, but perhaps not impossibly so, whereas the maximum wages for 26 Scheidel 2002.

Earnings and Costs

317

craftsmen look unreasonably low. I return to this problem in the next section.

7. DIOCLETIAN’S PRICES Diocletian’s Edict on Maximum Prices, promulgated at Nicomedia in December ad 301, provides our only extensive set of prices and earnings data from the Roman empire. A big question is how realistic its maximum prices are, and how useful they are for economic historians. This has been little tested. Hopkins noted the broad correspondence of maritime freight charges to actual nautical distances, with the proviso that the drafters followed the geographical tradition which made the western Mediterranean smaller than it actually is.27 Mrozek and Corbier tried to use Diocletian’s prices to calculate inXation between the mid-Wrst century and ad 301, and Corbier concluded that, in terms of gold equivalent, the prices of many spices and perfumes had fallen considerably, on a rough average around 60%.28 Allen (this volume) suggests a diVerent approach to the exploitation of Diocletian’s maximum prices. In Table 15.4 I index a small selection of Diocletian’s prices against earlier known data, converting both into what a Roman pound of gold would have bought at the oYcial valuation of the silver coinage. Most of the prices used have been discussed above. I take the higher possible total for military pay attested in P.Panop. Beatty 2 (ad 301) of 21,800 denarii (the Price Edict does not list soldiers’ pay because it was Wxed). I add Diocletian’s maximum prices of 800 den./lb for pepper, 400 den./lb for trog(l)odytic myrrh and 2[00 þ ?] den./lb for stacte myrrh,29 with which I compare the top prices of 15 den./lb, 16.5 den./lb, and 50 den./lb cited as current in Rome, probably around ad 70, by the elder Pliny.30 I also add the maximum price for ivory of 150 den./lb,31 with which I compare the sale prices of 9 den./lb in TPSulp. 101 and of 70 and 100 dr./mna in SB XVIII 13167.32

27 Hopkins 1983: 102–4. 28 Mrozek 1975; Corbier 1985. 29 34.68 G; Roueche 1989: 303. 30 NH 12.28, 70. 31 16.10a G. 32 With Rathbone 2000 (mna = 525g).

318

Dominic Rathbone

Table 15.4 Purchasing power of 1 lb gold

oYcial silver weight of denarii—to ad 64 —from ad 65 wheat at HS 4/mod. iconic price, 1st – 2nd c. ad wheat at 12 dr./art. high price, Egypt, 80s–160s wheat at 20 dr./art. high price, Egypt, 190s–250s rural wage of 2 dr./day high, Egypt, mid-2nd c. ad rural wage of 4 dr./day high, Egypt, mid-3rd c. ad

1st – 3rd c. ad (A)

Diocletian’s Edict (B)

3,845 g

3,876 g

3,785 g 1,000 mod.

100 B/A 101 102

1,080 mod. Ital.

108

1,687 mod.

64

1,013 mod.

107

2,250 days

2,057 days

1,125 days

91 183

shipwrights/sawyers, say norm of 7.5 dr./day, Egypt, mid-3rd c. ad

600 days

1,028 days

166

soldier’s pay at HS 1,200 p.a., 2nd c. ad at HS 3,600 p.a., late 3rd c. ad

1,350 days 450 days

1,189 days

88 264

best pepper, Rome, c. ad 70 best trogodytic myrrh best stacte myrrh

75 lbs 68.2 lbs 22.5 þ lbs

90 lbs 180 lbs 360 lbs

120 264 1,600

ivory (pieces?), Puteoli, ad 48 ivory pieces, and whole tusks, Alexandria, mid-2nd c.

111 lbs 105 lbs 73 lbs

480 lbs

432 457 658

The index Wgures in the last column of Table 15.4 express the relationship of each Diocletianic price to the corresponding earlier price reckoned as 100. An index Wgure of 100 thus indicates that one Roman pound of gold could buy the same by Diocletian’s price as by the earlier price; an index Wgure below 100, that gold now bought less, and one above 100, that gold now bought more. The almost exact parity for silver coins illustrates the consistent oYcial tariYng from Augustus to Diocletian of silver to gold coins on a 12:1 weight ratio. Diocletian’s maximum price for wheat was extraordinarily close to the general iconic price of the early Principate, which may be a Xuke or may possibly indicate that the Roman government also had some traditional idea that wheat, like silver, should have a Wxed ratio to gold. Probably more signiWcant is the correspondence of the

Earnings and Costs

319

maximum price for wheat to the high end of the normal range in Egypt before the big price rises of 274 and the later 290s. In contrast, the maximum wages set for rural workers and craftsmen (even adjusted as in Table 15.3 to include rations) represented a considerable reduction in terms of gold and also wheat, and soldiers’ pay was cut even more severely from its third-century peak. While the maximum price for pepper was close to its top Wrst-century price, according to Pliny, the prices for myrrh and ivory look wildly low in historical terms. It is not clear how these discrepant variations are to be best explained. One hypothesis, following the argument of Lo Cascio, might be that Diocletian’s Edict recognized and tried to cap a cumulative rise in the purchasing power of gold.33 However, that is not likely. According to the Edict’s preface, Diocletian’s principal aim was to protect the purchasing power of military pay; also, though not advertised, all state expenditure was to beneWt. So why cap the purchasing power of gold when the government was the monopoly producer and single biggest holder of gold bullion? And even if this is what happened, why then overvalue wheat in relation to wages when it was one of the most basic purchases made by soldiers and state? There is also the question of which of the widely variant index Wgures should be taken to reXect the supposed actual rise in the purchasing power of gold. The contrary supposition seems to work better. As the Wgures for wheat and wages in the Wrst column of numerals in Table 15.4 illustrate, the doubling of prices which followed the Antonine plague had in eVect halved the purchasing power of gold and hence silver too. The even sharper price rises of the late third century in response to the coinage reforms of Aurelian in 274 and Diocletian in 294–6 had presumably had a similar eVect, whose scale will have depended on the precise nature of those ill-understood reforms. If so, the related Prices and Coinage Edicts of 301 may represent an attempt to restore the purchasing power of gold, that is of the imperial coinage, to what it had been, at least in the eastern provinces, in the most recent era of price stability a generation earlier. (No one in 301 can have known that the purchasing power of gold had been 33 Lo Cascio 1997; cf. section 2 above.

320

Dominic Rathbone

twice as high in the early Principate.) But it was not an economically and socially neutral operation. As we have seen, maximum civilian wages were set considerably lower in relation to the price of wheat, presumably a deliberate attempt to make labour cheaper. Then there are the puzzlingly diVerent but enormous reductions in the prices for myrrh and ivory (and other spices and perfumes).34 The earlier ivory prices, incidentally, look odd because we would expect the price to have been signiWcantly higher in Italy than in Alexandria,35 but ivory prices doubtless experienced big ad hoc Xuctuations, and also may have risen at Rome from the Wrst to second centuries simply through supply and demand. Corbier suggested that the low prices of pepper (on one of her calculations), other spices, and perfumes implied a ‘banalisation’ or ‘democratisation’ of consumption of these goods;36 in fact, price reductions on this scale would have required a massive growth in supply which is unlikely. One solution would be to conclude that prices for luxury goods were so variable anyway, depending on their quality and stage of processing and on haphazard and unconnected sharp peaks of supply and demand, that it was very diYcult to pick a typical or realistic price, and that some or all of the earlier prices and the Edict’s prices are not comparable or unrepresentative. This must be sometimes or often the case, but the common trend of the Edict to downprice these goods severely suggests that Diocletian may have had a moral aim, that is to restrain the import trade in ‘exotic’ goods by slashing its potential proWtability in the tradition of Roman censors of old who had set price limits on elite expenditure. Pepper perhaps escaped this fate because its use was now so common (‘democratisation’) that it no longer counted as a luxury. Whatever the correct explanation for the vast but varied reductions, in historical terms, of the prices for spices, perfumes, and ivory, they call into question the realism of the price structure of the Edict. But the problem goes deeper. I have of necessity used Egyptian wheat prices and wages for comparison, but if they had been, in coinage terms, consistently slightly lower than those in other eastern provinces, the Edict’s maximum prices and wages will have been more generally reductive than appears here. Furthermore, the prices of the 34 See Corbier 1985. 35 Cf. Pliny, NH 8.31. 36 Corbier 1985: 95, 101–2.

Earnings and Costs

321

Edict will have been infeasibly low for Italy, assuming that the normal range of wheat prices there remained up to four times higher than in Egypt. And we know almost nothing of historical price levels in most of the European provinces or the inland areas of the east. Table 15.2 suggests that soldiers in 301 were as well or better oV than in the third century in terms of wheat equivalent at the Edict’s maximum price only because for the third-century calculation I used a high guesstimate price for Italy rather than the low Egyptian price. Diocletian’s soldiers will only have been smiling if they could hold sellers, even in Italy, to the maximum prices of his Edict. It may be signiWcant that no fragments of the Edict have been found in the west. In short, Diocletian’s Edict on Maximum Prices has potential and pitfalls for the economic historian. Some of its prices (e.g. wheat, pepper) may reXect, at a low level, the actual price range particular to the eastern coastal zone around ad 301; some (for example, wages) may be more reductive but still in some relation to other prices; others (for example, myrrh, ivory) are grossly and variably reductive. As a whole, therefore, the Edict’s prices do not reXect any external historical reality (except insofar as they may have been obeyed for a time), nor do they form an internally consistent structure. For progress in assessing the prices of the Edict as a whole there are various desiderata: more data from earlier periods, especially prices for textiles, the single biggest category of goods in the Edict; a revised text of the Edict with all the new Aphrodisias fragments, including those of the Coinage Edict; an updated study of the fourth-century prices from Egypt,37 our only worthwhile evidence for price behaviour after the Edict.

8. CONCLUSIONS It seems that the Roman government and elite had some belief that there was an overarching integrated market in the empire, and acted in various ways to make this belief reality, for example by keeping the oYcial silver to gold weight ratio of the coinage at 12:1, intervening 37 Bagnall 1985.

322

Dominic Rathbone

in shortages to provide wheat at a ‘fair’ price of HS 4 per modius, and maintaining a unitary rate of pay for soldiers. There were regional price zones for wheat in the Roman empire: it seems that Rome and Italy formed one and the coastal areas of the eastern Mediterranean another. If we had more data, we would probably discern other zones, such as the north-western provinces. Wheat probably cost up to four times more in Italy than in Egypt, reXecting relative agricultural productivity and transport costs. However, this was probably the extreme diVerence for wheat prices in the empire because it coupled the most productive producer with the most intense consumer. The regional price zones were interlocked at stable levels by private trade spurred by the ideology and practice of state intervention. There appears to have been much more regional variation in wages, which would imply variation in labour supply if wheat and other basics were relatively similarly priced. In developed areas, such as the urban cultures of the east, rural craftsmen were paid double the modal agricultural wage, and much more in cities. I suspect that manual labourers in cities were poorer than agricultural workers, and that you had to be a craftsman to beneWt fully from the city. In Rome the unskilled poor probably depended on state and private handouts to survive. There is no evidence for Wnancial reward of long service or experience; indeed there was virtually no promotion for Roman soldiers. Independent craftsmen may have been able to raise their rates, especially for piece work, if their reputation grew. In contrast the elite rewarded themselves handsomely out of state funds for running the empire as oYcers and civilian oYcials, and with higher salaries for higher status posts. To assess the economy and prosperity of the Roman empire it would be interesting to compare earnings in terms of bullion and wheat equivalent with comparable urbanized empires such as Han China or Georgian Britain. Many of the attested Roman earnings, including those of soldiers, apparently fall below a wheat equivalent of 1,500 kg p.a., roughly three times subsistence, which is often taken as the benchmark of prosperity in pre-industrial economies. However we need to be alert to possible deWciencies in our assumptions and calculations: soldiers, for instance, also had a hefty discharge bonus and had some private business aVairs, while agricultural

Earnings and Costs

323

labourers often came from families with a diverse portfolio of part-time activities. There is also the problem of telling which is the typical situation. An average which included elite earnings would be much higher. For the nature of an economy it is as important how earnings are distributed as what their brute level was. We could ask how moderately or greedily, in comparative historical terms, the Roman elite paid themselves, how well or badly the acquisition of craft skills was rewarded, and so on. Our data for Roman prices and earnings, thin though they be, permit some conclusions about changes across time. The Egyptian data, not discussed in full here, show two long periods of stability from the ad 80s to 160s and from the 190s to 274. Although we lack the price data to prove that this was an empire-wide phenomenon, it Wts with military pay which remained the same from the later Wrst century bc to ad 84/5, when it was raised by a third, and then again stayed Wxed to the start of the third century. Of course prices of individual goods such as ivory may have risen or fallen against the general pattern because of particular factors of supply and demand. There are two slight indications of rises in living standards: the apparent increase in the early Principate of the standard Roman wheat ration from four to Wve modii per month, that is 326 to 408 kg p.a. (Table 15.3a), and the enhanced earnings of rural labour in third-century Egypt. Also, the doubling of prices and wages which followed the Antonine plague brought everyone a great saving in tax payments in that Xat cash taxes, such as the poll tax, were not increased. Unfortunately none of these data can be used to demonstrate economic growth in the sense of increased productivity, for improved earnings may just indicate that the entitlement of workers had grown. Of course the two could happen together. Elsewhere I have argued that the large estates which developed in Egypt after the Antonine plague had to pay more for labour because it was relatively short, but that they also increased agricultural productivity.38 Because the coinage system of the Roman empire was oYcially pegged to a bullion weight standard, the price rises of the late second and late third centuries in eVect reduced the purchasing power of gold and silver. I have suggested that the aim of Diocletian’s Edict on 38 Rathbone 2007b.

324

Dominic Rathbone

Maximum Prices and Coinage Edict of ad 301 was to restore the purchasing power of gold, that is of the imperial coinage, to its pre274 level. Whatever short-term success Diocletian had, by the later fourth century the purchasing power of gold had rocketed to two and a half times its pre-274 level, where it remained for the next two centuries. Patently the cause of this was not economic growth, nor a twofold drop in the empire’s stock of circulating gold. Perhaps again it was to do with entitlement, and was the result of a revolution which, in the context of declining urbanism, expanded the socioeconomic diVerentiation between a gold-rich elite and a cash-poor peasantry. Bibliography Andreau, J., Briant, P., and Descat, R. (eds. 1997). E´conomie antique: Prix et formation des prix dans les e conomies antiques. St-Bertrand-deComminges. Bagnall, R. (1985). Currency and InXation in Fourth Century Egypt (BASP Suppl. 5). Chico, Calif. Carrie, J.-M. (1997). ‘L’arithmetique sociale de l’economie agraire: prix de la terre, rente foncie`re et prix des cereales dans l’E´gypte romano-byzantine’, in J. Andreau, P. Briant, and R. Descat (eds.), E´conomie antique: Prix et formation des prix dans les e conomies antiques. St-Bertrand-de-Comminges, 121–46. Corbier, M. (1985). ‘Devaluations et evolution des prix (Ier–IIIe sie`cles)’, RN 27: 69–106. Cuvigny, H. (1996). ‘The amount of wages paid to the quarry-workers at Mons Claudianus’, JRS 86: 139–45. Drexhage, H.-J. (1991). Preise, Mieten/Pachten, Kosten und Lo¨hne im ¨ gypten bis zum Regierungsantritt Diokletians. St Katharinen. ro¨mischen A Duncan-Jones, R. P. (1974/1982). The Economy of the Roman Empire: Quantitative Studies, 1st and 2nd edn. Cambridge. Erdkamp, P. (2005). The Grain Market in the Roman Empire. Cambridge. Frank, T. (ed. 1933–40). Economic Survey of Ancient Rome, 6 vols. Baltimore. Frier, B. W. (1993). ‘Subsistence annuities and per capita income in the early Roman empire’, CP 88: 222–30. Garnsey, P., and Humfress, C. (2001). The Evolution of the Late Antique World. Cambridge.

Earnings and Costs

325

Heichelheim, F. (1930). Wirtschaftliche Schwankungen der Zeit von Alexander bis Augustus. Jena (repr. New York 1979). —— (1954/5). ‘On ancient price trends from the early Wrst millennium b.c. to Heraclius I’, Finanzarchiv 15: 498–511. Hopkins, K. (1983). ‘Models, ships and staples’, in P. Garnsey and C. R. Whittaker (eds.), Trade and Famine in Classical Antiquity. Cambridge, 84–109. —— (1995/6). ‘Rome, taxes, rents and trade’, Kodai 6/7: 41–75; repr. (2002) in W. Scheidel and S. von Reden (eds.), The Ancient Economy. Edinburgh, 190–230. Hultsch, F. (1862/1882). Griechische und ro¨mische Metrologie, 1st and 2nd edn. Berlin (repr. Graz 1971). Jones, A. H. M. (1964). The Later Roman Empire 284–602, 2 vols. Oxford. Jongman, W. M. (2007). ‘The early Roman Empire: Consumption’, in I. Morris, R. Saller, and W. Scheidel (eds.), The Cambridge Economic History of the Greco-Roman World. Cambridge, 592–618. Kessler, D., and Temin, P. (2005). ‘Money and prices in the early Roman Empire’, MIT Dept of Economics, Working Papers 05-11, http://ssrn.com/ abstract¼704724 (accessed 19 Oct. 2007). Lo Cascio, E. (1997). ‘Prezzi in oro e prezzi in unita` di conto tra il III e il IV sec. d.C.’, in J. Andreau, P. Briant, and R. Descat (eds.), E´conomie antique: Prix et formation des prix dans les e conomies antiques. St-Bertrand-deComminges, 161–82. Morris, I., Saller, R., and Scheidel, W. (eds. 2007). The Cambridge Economic History of the Greco-Roman World. Cambridge. Mrozek, S. (1975). Prix et re muneration dans l’Occident romain (31 av. n.e`. – 250 de n.e`.). Gdansk. Rathbone, D. W. (1996). ‘Monetisation, not price-inXation, in thirdcentury A.D. Egypt?’, in C. E. King and D. G. Wigg (eds.), Coin Finds and Coin Use in the Roman World. Berlin, 321–39. —— (1997). ‘Prices and price-formation in Roman Egypt’, in J. Andreau, P. Briant, and R. Descat (eds.), E´conomie antique: Prix et formation des prix dans les e conomies antiques. St-Bertrand-de-Comminges, 183–244. —— (2000). ‘The ‘‘Muziris’’ papyrus (SB XVIII 13167): Financing Roman trade with India’, in Alexandrian Studies II in Honour of Mostafa el Abbadi, Bulletin de la Socie te d’Arche ologie d’Alexandrie 46: 39–50. —— (2007a). ‘Military Wnance and supply’, in P. Sabin, H. van Wees, and M. Whitby (eds.), The Cambridge History of Greek and Roman Warfare II. Cambridge, 158–76. —— (2007b). ‘Roman Egypt’, in I. Morris, R. Saller, and W. Scheidel (eds.), The Cambridge Economic History of the Greco-Roman World. Cambridge, 698–719.

326

Dominic Rathbone

Roueche, C. (1989). Aphrodisias in Late Antiquity: The Late Roman and Byzantine Inscriptions. London. Scheidel, W. (2002). ‘A model of demographic and economic change in Egypt after the Antonine plague’, JRA 15: 97–114. Sen, A. (1981). Poverty and Famines: An Essay in Entitlement and Deprivation. Oxford. Sperber, D. (1974/1991). Roman Palestine, 200–400: Money and Prices, 1st and 2nd edn. Ramat-Gan. Szaivert, W., and Wolters, R. (2005). Lo¨hne, Preise, Werte: Quellen zur ro¨mischen Geldwirtschaft. Darmstadt.

16 How Prosperous were the Romans?: Evidence from Diocletian’s Price Edict (ad 301) Robert C. Allen

How prosperous were the Romans?1 Their individual experiences ranged from wretched poverty to fabulous wealth, and that variety makes generalizations diYcult. Many kinds of evidence can be used to address this question.2 Three approaches to the problem are particularly direct and encompassing. The Wrst approach is to calculate the average income. This equals Gross Domestic Product divided by the population since GDP equals both the value of total production, and the sum of everyone’s income. The GDP approach is appealing since it ties income into the production structure and makes explicit the connection between the standard of living and the eYciency of agriculture and manufacturing.3 However, there is a corresponding drawback: the GDP approach requires either (1) a great deal of economic information that is either unavailable or not known with much accuracy, or (2) very strong equilibrium assumptions so that the small amount of information we do have can be used as proxy data for what we do not know about the economy. In addition, the population must also be known in order to 1 I thank Alan Bowman and Andrew Wilson for their comments on an earlier draft of this chapter. Remaining errors are my own. 2 Atkins and Osborne 2006; Brown 2001; Jongman 2007; Ward-Perkins 2005. 3 Hopkins 1980, 2002; Goldsmith 1984; Maddison 2001; Temin 2006.

328

Robert C. Allen

calculate average income, and population estimates are also controversial. Consequently, while GDP calculations help organize what we know about the Roman economy, any calculation of per capita GDP is bound to be problematic. A second approach to ascertaining the average standard of living uses skeletal evidence.4 The idea is that if people were better nourished during their youth, they would have been taller as adults. If large, random samples of skeletons were excavated, average adult height could be determined and the standard of living established. Promising research has begun in this area, but it is already creating controversy rather than consensus. There turn out to be serious problems in estimating average height—the prevalence of cremation, the size and representativeness of the samples, inferring height from the length of the femur, and so forth—and there is the further complexity that the correlation between height and income is not exact. So even if we could establish the ‘biological standard of living’, we would still not be certain about the economic standard of living. In view of these diYculties, this chapter proposes a third approach to measuring Roman living standards. Instead of estimating the average income (or height) of Romans, I study the income of an ‘average’ Roman, a somewhat diVerent concept. In particular, I study the purchasing power of an unskilled, free male labourer.5 How much could such a person buy with his earnings? There are, of course, diYculties with this method: how representative was a labourer? What if he did not work full time? What about other people? Nevertheless, this provides us with a new approach to the problem that incorporates information that other approaches do not use. It provides a useful complement to them. The purchasing power of wages (the real wage) is intrinsically comparative. In this case, we concentrate on the question: could workers in the Roman empire buy more or less than their counterparts in other times and places? We can establish comparisons with medieval and early modern Europe as well as with other great empires like Qing China and India under Moghuls and the Raj. 4 Koepke and Baten 2005; Jongman 2007. 5 For some earlier treatments of wages, costs, and purchasing power see Reece 1973, Duncan-Jones 1978, and Fre´zouls 1977, 1978.

How Prosperous were the Romans?

329

The reason that these comparisons are possible is because many researchers are studying real wages around the world in a systematic way.6 One of the grand questions in modern economic history is why Europe pulled ahead of the rest of the world, and that question cannot be answered until we know when the ‘great divergence’ happened.7 Real wages are one indicator of that separation. By using standard methodologies developed to chart the great divergence, Roman historians can avail themselves of the emerging picture of post-medieval living standards and use it to judge Roman performance. At the same time, the Roman empire will be integrated into this broader world-view. This chapter is a Wrst stab at that problem. To measure the purchasing power of wages, we need information about wages and the prices of consumer goods. The data used in this chapter are all derived from Diocletian’s Edict on Maximum Prices issued in ad 301.8 There are questions about how long the edict was enforced and how the maximum prices it speciWed relate to actual wages and prices in diVerent parts of the empire;9 indeed, one way to improve on the results here is to rework the calculations with wages and prices speciWc to a variety of times and places. However, since the Price Edict is such a well-known source and includes most of the necessary information it is worth seeing what it has to say. The approach I use is straightforward. First, I calculate from daily wages what a labourer would earn in a year if he worked full time. Second, I calculate the cost of supporting a family. This involves specifying the items—and the quantities—they would have consumed in a year. This list is called their ‘consumption basket’. Multiplying the quantities by their prices deWnes the cost of living. Third, I compare income to living costs. Could the labourer purchase the speciWed basket? Doing the same calculations for the Roman ¨ zmucur and Pamuk 2002; Bassino and Ma 2005; 6 van Zanden 1999; Allen 2001; O Allen et al. 2007. 7 Pomeranz 2000. 8 For the prices in the Edict I have used the edition of LauVer (1971). The most important modern publications are Reynolds 1971, Erim et al. 1970 and 1973, Giacchero 1974. See also Ermatinger 1990. The fullest English translation of the text, as it was then known, is that of Graser (1940). 9 See Ermatinger 1990.

330

Robert C. Allen

empire and for other scenarios shows us whether the Roman labourer fared better or worse than his counterparts in other periods and places. The remainder of the chapter discusses these issues in more detail and presents the results.

WAG E R AT E S In the Roman empire, some workers were paid by the day and some according to the work they did. Here I only consider the day workers, although it would be worthwhile to extend the analysis to the others. The Price Edict speciWes the wage for a variety of unskilled workers (for example farm labourers, camel and mule drivers, water carriers, and sewer cleaners: VII, 1a, 17, 19, 31, 32) as 25 denarii per day plus maintenance. A cash wage plus a food allowance has been a common payment scheme in many poor countries. According to Duncan-Jones, food allowances were about 5 Italic modii of wheat per month.10 Since the price of wheat was set by the Edict at 100 denarii per castrensis modius (i.e. 66.67 denarii per Italic modius, which was two-thirds the size of the castrensis modius), the food allowance was worth 333 denarii per month. I treat this as an increase in income rather than as a reduction in the quantity of food purchased. One could debate how to combine the money wage and the food allowance since the Wrst is per day and the second per month. I assume that the food allowance was worth 11.1 denarii (¼ 333 denarii/30 days). On this reading, unskilled labour earned 36.1 denarii per day. It is interesting to compare the wage rate in the Roman empire to wages elsewhere. To do that, we must establish an exchange rate between denarii and other currencies. This is an ever-present problem in price history. Since silver was so often the medium of 10 Duncan-Jones 1982: 146. The Wve Italic modii used for food allowances is the amount of the Roman ‘corn-dole’ of the earlier empire, the use of which can be justiWed on the view (though not universally accepted) that the dole was not welfare for the destitute but an allowance for the more or less self-suYcient (see P.Oxy. XL 2892–940 general introduction: 8–15).

How Prosperous were the Romans?

331

20 London

Amsterdam

Vienna

Florence

Delhi

Beijing

Grams of silver per day

15

10

5

0 1375

1475

1575

1675

1775

Roman empire AD 301

Fig. 16.1. Silver wages Source: Allen et al. (2007) and text.

exchange, currencies are usually converted according to their silver equivalents. How much silver was in a denarius at the time of the Price Edict? I take it to have been 0.032 grams of pure silver.11 In that case, the daily wage of a labourer was 1.16 grams of silver. 11 According to King (1993: 11), the argenteus, which was the coin worth one hundred denarii, had a theoretical weight of 3.38 grams of silver. In practice, however, the actual weight was a bit lower. It should be noted that the welfare ratios computed in this chapter are independent of the silver content assumed for the denarius. The reason is that the welfare ratios are ratios. The silver exchange rate appears in the numerator where it converts the wage from denarii to grams of silver and in the denominator where it does the same for the prices. Thus, the silver exchange rate cancels out.

332

Robert C. Allen

Figure 16.1 puts that wage on a graph of daily wages for labourers in leading cities in Europe and Asia from the late Middle Ages through to the eighteenth century.12 There was little divergence in wages among European cities in the Wfteenth century, but afterwards wages rose dramatically in north-western Europe where the economy was booming due to the expansion in global trade.13 Wages in Asian cities, for which data only become available in the seventeenth century, were low—on a par with those in central and southern Europe where the economy was stagnant. The daily wage in the Roman empire is shown with a star on the graph, although it does not, of course, apply to the year for which it is plotted. The wage rate in Rome was lower than in early modern Europe or Asia. Despite Diocletian’s concern with inXation, wages in his empire were not high by later standards.

THE EUROPEAN RESPECTABILITY BASKET Figure 16.1 shows that a worker in eighteenth-century London earned much more silver per day than a worker in the Roman empire, but does that mean that the English worker had a higher standard of living? The answer depends on the cost of living. How should it be measured? The simplest approach is to calculate how much grain the worker could buy with his wage. Sometimes ‘grain wages’ give the right answer, but they are problematic since neither London nor many Roman workers ate grain. Instead, they ate bread, which, as a processed food, was a more expensive source of calories. Furthermore, virtually all people consume things besides the basic carbohydrate. Ideally, one should specify a complete ‘consumption bundle’ and compute its cost. Many bundles are possible. Table 16.1 is the ‘northern European respectability basket’ that I Wrst used to compute the cost of living in early modern Europe. The basket was inspired by English and Dutch studies of working-class 12 The European and Asian wages and prices used in this chapter are available on my website at http://www.nuV.ox.ac.uk/sociology. 13 Allen 2003a; Acemoglu, Johnson, and Robinson 2005.

How Prosperous were the Romans?

333

Table 16.1. Northern European respectability basket Nutrients/day Quantity per Price g. silver Spending Nutrients/day grams of person per year per unit share (%) calories protein Bread Beans/peas Meat Butter Cheese Eggs Beer Soap Linen Candles Lamp oil Fuel Total

182 kg 52 l 26 kg 5.2 kg 5.2 kg 52 each 182 l 2.6 kg 5m 2.6 kg 2.6 l 5.0 M BTU

0.693 0.477 2.213 3.470 2.843 0.010 0.470 2.880 4.369 4.980 7.545 4.164

30.4 6.0 13.9 4.3 3.6 1.3 20.6 1.8 5.3 3.1 4.7 5.0

1,223 160 178 104 53 11 212

50 10 14 0 3 1 2

414.899

100.0

1,941

80

budgets and suggests the spending pattern of ‘respectable’ workers. The budget is medieval, however, in that it does not include foods like sugar and potatoes that only entered mass consumption after the European discovery of America. The budget shown in Table 16.1 represents the spending of an adult male. It should be noted that wheat bread was consumed in some regions, while rye bread was consumed in others. In some places, veal was eaten, in others, it was beef or pork. I have followed the maxim ‘when in Rome, do as the Romans’ and have calculated the cost of the basket using the price of the most widely consumed type of bread or meat. The budget’s daily calorie intake (1,940) would put the person in the third decile from the bottom of the English distribution in the late eighteenth century and in the fourth decile of the French.14 Protein availability was high. This was ‘respectability’ rather than destitution. Frequently mentioned non-food commodities like cloth, candles, and fuel are also included. Textiles presented a problem. The only type of cloth shown is linen, although much wool was also worn. The available price data for wool are hard to interpret, however. One needs the price of the same quality of woollen cloth in the places compared, but wool varied enormously in terms of quality, and 14 Fogel 1991: 45.

334

Robert C. Allen

quality diVerences seem to have driven the diVerences in the prices we have. Wool cloth is, therefore, not useful for the present exercise. Linen does not appear to suVer from the same problem. I have included enough linen in the budget, so that the share of linen in spending is what one would expect for all textiles. Table 16.1 lists average prices in grams of silver for 1745–54 prevailing in Strasbourg. When the cost of the basket is calculated with these prices, it amounts to 414.899 grams of silver per year. Prices were diVerent in other cities and in other times, but this gives an indication of the price level in eighteenth-century Europe. Table 16.1 also omits the cost of housing. Budget studies suggest it amounted to 5–10% of spending. I have not been able to Wnd the rent of comparable housing in the cities studied, so I have dealt with this problem by increasing the cost of everything except the housing by 5%. While Table 16.1 describes spending in northern Europe, it does not apply to the Mediterranean. In southern Europe, wine was consumed instead of beer, and olive oil was used instead of butter. In reckoning the cost of living in Mediterranean cities, I substituted olive oil for butter and wine for beer. In the latter case, the quantity was reduced to maintain the same annual consumption of alcohol. Table 16.2 shows the resulting Mediterranean budget. Its cost in

Table 16.2. Mediterranean respectability basket Quantity per person per year Bread Beans/peas Meat Olive oil Cheese Eggs Wine Soap Linen Candles Lamp oil Fuel Total

182 kg 52 l 26 kg 5.2 l 5.2 kg 52 each 68.25 l 2.6 kg 5m 2.6 kg 2.6 l 5.0 M BTU

Strasbourg price g. silver per unit

Naples price g. silver per unit

0.693 0.477 2.213 7.545 2.843 0.010 0.965 2.880 4.369 4.980 7.545 4.164

0.790 0.479 2.571 2.505 2.571 0.127 0.300 2.029 4.854 1.405 2.505 5.452

416.3583

355.9249

Diocletian price g. silver per unit 0.394 0.408 1.290 1.160 1.290 0.053 0.774 1.160 4.031 1.160 1.160 1.586 163.921

How Prosperous were the Romans?

335

Strasbourg prices was virtually identical to the cost of the northern basket. Neapolitan prices are also shown, and the Mediterranean basket cost less in Naples than it did in Strasbourg. While the baskets in Tables 16.1 and 2 would support an adult male, the question we want to address is whether a man’s earnings were suYcient to support a family, so the consumption bundle in Table 16.1 or 2 must be scaled up accordingly. The cost of the family depended on the ages and conditions of the wife and children. I have not made a close calculation of this question, but have simply multiplied the cost of the basket in Tables 16.1 and 2 by three. This allows for a man, a woman, and some children given that equivalence scales represent a women as eating less than a man and young children as requiring very much less. In sum, the cost of maintaining a family at the standard of the northern European respectability basket equals the cost of the basket in Table 16.1 or 2, multiplied by 1.05 to include rent, multiplied by 3 to scale it up to a family.

THE COST OF THE BASKET IN DIOCLETIAN’S REIGN To establish a comparison with the Roman empire, we must compute the cost of the basket with prices from Diocletian’s Price Edict. Some of these present few problems of interpretation, while others present diYculties. These prices have been converted to grams of silver per metric unit. The particular prices were as follows (parenthetical references are to chapters and lines in the Price Edict): Beans/peas: 100 denarii per castrensis modius as for crushed beans, lentils, chickpeas, crushed peas, dried kidney beans (I, 9, 11, 13, 15, 21); Meat: 8 denarii per Roman pound as for beef, goat, or mutton (IV, 2, 3); Cheese: 8 denarii per Roman pound as for fresh cheese (VI, 96); Eggs: 1 denarius each as for eggs (VI, 43); Oil: 12 denarii per Roman pint as in oil, ordinary (III, 3);

336

Robert C. Allen

Wine: 8 denarii per pint as for ordinary wine (II, 10). Cloth: cover to be used as a tent, 16 feet in length and breadth, dyed (XIX, 4). The cloth is 256 square Roman feet in area and priced at 2,500 denarii or 9.77 denarii per square Roman foot. A Roman foot was 296.2 mm, so the cloth cost 111.3089 denarii or 5.877 grams of silver per square metre. The early modern prices are for pieces one metre long and typically about one ell (27 inches or 0.6858 metres) wide. A piece this size would have cost 4.03 grams of silver in ad 301. Fuel: ‘wagon load of wood, 1200 pounds’ worth 150 denarii (XIV, 8) or 0.125 denarii per Roman pound (0.32745 kg) or 0.012216 grams of silver per kilogram (0.032  0.125/0.32745). I assume that one kilogram of wood yields 7,701 BTUs.15 The price of energy in grams of silver per million BTUs was, therefore, 1.586 ¼ 1,000  0.012216/7.701. Bread: bread does not appear in the Edict. It does give a maximum price for wheat (100 denarii per castrensis modius of 12.936 litres: I, 1) and the wage for a skilled worker, in particular a baker at 50 denarii per day plus maintenance (VII, 12). I estimated the price of bread from these data using the ‘bread equation’ I estimated previously.16 This is a statistical equation estimated from an early modern dataset of bread prices, grain prices, and wage rates. The equation Wts the data well and provides a means of estimating the price of bread from the cost of its two main inputs. The equation used here is: price of bread ¼ 0.063 þ 1.226  price of wheat þ 0.014  daily wage of skilled labour The prices and wages are all in grams of silver, the price of bread is grams of silver per kilogram, and the price of wheat is grams of silver per litre. Table 16.2 shows the prices used in computing the cost of living in Diocletian’s day. When the prices are multiplied by the quantities shown, the cost of the basket can be calculated. It works out at 163.921 grams of silver per year. This was a much lower cost of living than in Naples or Strasbourg. 15 Allen (2003b) tabulates the energy content of many fuels measured by weight and volume. This paper is available on my website at http://www.nuV.ox.ac.uk/sociology. 16 Allen 2001.

How Prosperous were the Romans?

337

THE STANDARD OF LIVING OF A RO MA N L AB O U RE R We can bring together the wage and price information to work out the standard of living of the Roman labourer. We proceed in three steps: annual income, annual consumption cost, and the ratio of income to expenditure. Annual income: Including food allowance, the daily wage of the Roman labourer was 1.16 grams of silver. How many days per year did he work? I assume that he worked 250 days, which is what was also assumed for early modern workers. This looks a plausible number in view of the large number of festivals in the late Roman empire.17 In that case, annual earnings were 289 grams of silver. Annual consumption cost: The basket in Table 16.2 cost 163.921 grams of silver. To get the annual cost of a family, rent at 5% must be added and the total scaled up by a factor of three. The annual cost of supporting a family was therefore: 163.921 grams silver  1.05 for rent  3 ¼ 516.352 grams of silver. Welfare ratio: The worker earned 56% of the cost of the basket (289/516.352). This ratio is called his welfare ratio. Evidently, the worker did not earn enough to buy the basket. How he coped will be considered shortly. For the moment, however, we can compare his welfare ratio to welfare ratios of workers in other times and places. In all cases, we are comparing the purchasing power of their earnings. Fig. 16.2 plots welfare ratios for the same cities shown in Fig. 16.1. Workers in London and Amsterdam did best. There was Xuctuation in their earnings. In the worst periods they earned barely enough to buy the respectability basket, but in most years they had some excess. These workers had the highest silver wages (Fig. 16.1). That advantage was reduced but not eliminated when account is taken of the prices of consumer goods they bought. 17 At the mid-point of the high and low parameters oVered by Duncan-Jones (1978: 160).

338

Robert C. Allen

2 London

Amsterdam

Vienna

Florence

Delhi

Beijing

1.5

1

0.5

0 1375

1475

1575

1675

1775

1875

Roman empire AD 301

Fig. 16.2. Welfare ratios with European respectability basket Source: Allen et al. (2007) and text.

Workers elsewhere in Europe and Asia experienced real wage slides across the early modern period. There was a wage peak in the Wfteenth century following the population drop caused by the Black Death in 1348/9 and subsequent outbreaks. As population grew from the sixteenth century onward, real wages fell. By the eighteenth century, most workers in Eurasia had earnings that only equalled half of the cost of the basket. Workers in the Dutch Republic and England were the only ones to avoid this fate as the economies of these countries grew rapidly due to their success in the global economy.

How Prosperous were the Romans?

339

BARE B O NES SUBSISTENCE BASKETS What happened to workers in southern Europe, east Asia, and the Roman empire who were too poor to buy the baskets—meagre as they are—shown in Tables 16.1 and 2? One strategy was to increase the time worked. Men could work more days and longer hours. No allowance has been made thus far for the earnings of women and children. They could work many hours, although their wage rates were usually so low that their employment could not decisively change the situation. China was an exception to this generalization, for women there contributed a signiWcant proportion of family income.18 Cutting expenditure was the other strategy. That was possible because the food shown was highly processed either by animals (meat) or by people (bread and beer). Indeed, the way to cut costs was sharply to curtail most of the non-food items, eliminate the alcohol and most of the animal protein, and eat boiled grains rather than bread. Scottish highlanders ate oatmeal, Italians shifted from bread to polenta, and many Asians ate mostly rice. In the Yangzi Delta, poor peasants even ate boiled wheat in the summer.19 Early modern travellers’ accounts often depict many people subsisting on diets like this. ‘It appears from contemporary accounts that the articles in the diet of the common people in most parts of India consisted chieXy of rice, millets and pulses.’20 Francisco Palsaert, a Dutch East India Company captain who visited the subcontinent in the early seventeenth century, called the Indian diet ‘monotonous’. In the Delhi-Agra region, the people ‘have nothing but a little kitchery [kedgeree] made of green pulse mixed with rice . . . eaten with butter in the evening, in the day time they munch a little parched pulse or other grain’. The workmen ‘know little of the taste of meat’. Indeed, pigs, cattle, chickens, and eggs were all taboo. Where available, Wsh was the only source of animal protein. It was a similar story in western India. Wheat was not eaten by the labouring population, 18 Li 1998: 150-5; Pomeranz 2000: 290, 319–20. 19 Li 1998: 207 n. 25. 20 Raychaudhuri and Habib 1982, I: 164.

340

Robert C. Allen

whose main source of carbohydrates was millet. This was ground into a course Xour and fried up as chapatis that were eaten with pulses and vegetables. Charles Lockyer, who toured Asia in the early eighteenth century on the English East India Company ship Streatham, observed of the Arab sailors in the Indian Ocean: ‘They serve for small Wages, and are Victual’d at a much cheaper Rate than our Ship’s Companys: Salt-Wsh, Rice, Gee, and Doll, with a few Fowls, being all the Provisions they care for. Doll is a small Grain, less than Fetches, contains a Substance like our white Peas, and being boil’d with Rice makes Kutcheree.’21 I speciWed such quasi-vegetarian ‘bare bones’ spending patterns for each of the places being compared. Oatmeal was the basis of the diet in north-western Europe, polenta in northern Italy, and millet in western India. Table 16.3 shows the bare bones basket for northwestern Europe and for the Romans. The Roman diet was dominated by wheat and beans, which supply most of the calories and protein. The calorie level has been kept at 1940 calories per day, as before. While animal protein has been greatly reduced, the wheat and beans provide a diet that is comparably rich in protein. The great thing about the bare bones diet is that it cost much less than the respectability diet—in the case of Rome, 83.163 grams of silver versus 163.921. Hence, the cost of supporting a family was cut to 261.963. This was something that the Roman labourer could aVord. Indeed, the ratio of income to the consumption expenditure jumped to 1.10, so the labourer had a bit of income to spend on ‘luxuries’. All workers see a rise in their welfare ratios, when bare bones spending patterns are the basis for comparison, as Fig. 16.3 shows. The Figure shows the same relative values as Fig. 16.2, but the poor workers of the eighteenth century all had welfare ratios of one or a bit above. Labourers in London and Amsterdam earned three or four times the cost of the bare bones diet. They did not multiply their oatmeal consumption by four, however. Instead, they increased their purchases of non-food items and bought the more expensive foods of the respectability diet. The essence of the matter is captured by 21 Lockyer 1711: 25.

Table 16.3. Bare bones subsistence baskets European oats Quantity per person per year Wheat Oats Beans/peas Meat Butter Olive oil Soap Linen Candles Lamp oil Fuel Total

155 kg 20 kg 5 kg 3 kg

Diocletian wheat

Nutrients/day calories

Nutrients/day grams of protein

1657 187 34 60

72 14 3 0

1.3 kg 3m 1.3 kg 1.3 l 2.0 M BTU 1938

89

Quantity per person per year

Nutrients/day calories

Nutrients/day grams of protein

172 kg

1607

50

20 kg 5 kg

187 34

14 3

5l 1.3 kg 3m 1.3 kg 1.3 l 2.0 M BTU

112

0

1940

67

342

Robert C. Allen

1 = bare bones subsistence income

6

5

London

Amsterdam

Vienna

Florence

Delhi

Beijing

4

3

2

1

0 1375

1475

1575

1675

1775

1875

Roman empire AD 301

Fig. 16.3. Welfare ratios with bare bones baskets Source: Allen et al. (2007) and text.

Doctor Johnson’s deWnition of oats as ‘a grain, which in England is given to horses, but in Scotland supports the people’.

C ON C LU S I ON The real wage evidence supports a guardedly optimistic view of Roman living standards. Certainly, the Roman worker in Diocletian’s time was doing about as well as most workers in eighteenth-century Europe or Asia. One lesson that price history teaches, however, is that real wages Xuctuated in the past. Indeed, Roman performance looks unimpressive in comparison with Wfteenth-century Europe when real wages were so high. In addition, generalizations are confounded by

How Prosperous were the Romans?

343

variation across space with workers in booming parts of the continent earning much more than their counterparts in more placid provinces. One suspects that the same was true of the Roman empire. The next step in the research agenda should, therefore, be to use actual wages and prices (rather than legal maxima) to measure real wages in the Roman period and to do this for diVerent regions of the empire and diVerent time periods. That investigation might reveal as much variation as we observe in the early modern world and lead to a more complex assessment of Roman living standards. Bibliography Acemoglu, D., Johnson, S., and Robinson, J. (2005). ‘The rise of Europe: Atlantic trade, institutional change and economic growth’, American Economic Review 95: 546–79. Allen, R. C. (2001). ‘The great divergence in European wages and prices from the Middle Ages to the First World War’, Explorations in Economic History 38: 411–47. —— (2003a). ‘Poverty and progress in early modern Europe’, EHR 56: 403–43. —— (2003b). ‘Was there a timber crisis in early modern Europe?’, Economia e energia secc. xiii–xviii, Serie II—Atti delle ‘Settimane di Studi’ e altri Convegni, Instituto Internazionale di Storia Economica ‘F. Datini’, Prato, 34: 469–82. —— Bassino, J.-P., Ma, D., Moll-Murata, C., and van Zanden, J. L. (2007). ‘Wages, prices, and living standards in China, 1739–1925: In comparison with Europe, Japan, and India’. Oxford University Department of Economics, Working Paper 316. Atkins, M., and Osborne, R. G. (eds. 2006). Poverty in the Roman World. Cambridge. Bassino, J.-P., and Ma, D. (2005). ‘Japanese unskilled wages in international perspective, 1741–1913’, Research in Economic History 23: 229–48. Brown, P. R. L. (2001). Poverty and Leadership in the Later Roman Empire. Hanover and London. Corcoran, S. (2000). The Empire of the Tetrarchs: Imperial Pronouncements and Government ad 284–324. Oxford. Duncan-Jones, R. (1978). ‘Two possible indices of the purchasing power of money in Greek and Roman antiquity’, in Les De´valuations a` Rome, E´poque re´publicaine et impe´riale (Coll. de l’E´cole franc¸aise de Rome 37) I. Rome, 159–68.

344

Robert C. Allen

Duncan-Jones, R. (19822). The Economy of the Roman Empire. Cambridge. Erim, K. T., and Reynolds, J. (1973). ‘The Aphrodisias copy of Diocletian’s edict on maximum prices’, JRS 63: 99–110. —— —— Wild, J. P., and Ballance, M. H. (1970). ‘The Copy of Diocletian’s Edict on Maximum Prices from Aphrodisias in Caria’, JRS 60: 120–41. Ermatinger, J. W. (1990). ‘Diocletian’s economic revolution’, Mu¨nstersche Beitra¨ge zur antiken Handelsgeschichte 9.1: 45–9. Figueira, T. J. (1999). Review of J. Andreau, P. Briant, and R. Descat (eds. 1997), The Ancient Economy: Prices and the Establishment of Prices in the Ancient Economies, St-Bertrand-de-Comminges, in The Classical World 92.6: 575–7. Fogel, R. W. (1991). ‘The conquest of high mortality and hunger in Europe and America: Timing and mechanisms’, in P. Higonnet, D. S. Landes, and H. Rosovsky (eds.), Favorites of Fortune: Technology, Growth, and Economic Development since the Industrial Revolution. Cambridge, Mass. 33–71. FrE´zouls, E. (1977). ‘Prix, salaires et niveaux de vie: quelques enseignements de l’Edit du Maximum’, Ktema 2: 253–68. —— (1978). ‘Prix, salaires et niveaux de vie: quelques enseignements de l’Edit du Maximum (II)’, Ktema 3: 289–300. Giacchero, M. (ed. 1974). Edictum Diocletiani et collegarum de pretiis rerum venalium: in integrum fere restitutum e latinis graecisque fragmentis, 2 vols. Genova. Goldsmith, R. W. (1984). ‘An estimate of the size and structure of the National Product of the early Roman Empire’, Review of Income and Wealth 30: 263–88. Graser, E. R. (1940). ‘The Edict of Diocletian on Maximum Prices’, in T. Frank (ed.), An Economic Survey of Ancient Rome, vol. V: Rome and Italy of the Empire. Baltimore, 305–420. Hopkins, K. (1980). ‘Taxes and trade in the Roman Empire (200 b.c.–a.d. 400)’, JRS 70: 101–25. —— (2002). ‘Rome, taxes, rents and trade’, in W. Scheidel and S. von Reden (eds.), The Ancient Economy. Edinburgh, 190–230. Jongman, W. M. (2007). ‘The early Roman Empire: Consumption’, in W. Scheidel (ed.) The Cambridge Economic History of the Greco-Roman World. Cambridge, 592–618. King, C. E. (1993). ‘The fourth century coinage’, in L. Camilli and S. Sorda (eds.), L’‘inXazione’ nel quarto secolo d.c. Atti dell’Incontro di Studio 1988. Rome, 1–87. Koepke, N., and Baten, J. (2005). ‘The biological standard of living in Europe during the last two millennia’, European Review of Economic History 9: 61–95.

How Prosperous were the Romans?

345

Lauffer, S. (19712). Diokletians Preisedikt (Texte und Kommentare 5). Berlin. Li, B. (1998). Agricultural Development in Jiangnan, 1620–1850. Basingstoke. Lockyer, C. (1711). An Account of the Trade in India. London. Maddison, A. (2001). The World Economy: A Millennial Perspective. Paris. Michell, H. (1947). ‘The edict of Diocletian: A study of price Wxing in the Roman Empire’, Canadian Journal of Economics and Political Science 13.1: 1–12. ¨ zmucur, S., and Pamuk, S. (2002). ‘Real wages and standards of living O in the Ottoman Empire, 1489–1914’, Journal of Economic History 62: 293–321. Pomeranz, K. (2000). The Great Divergence: China, Europe, and the Making of the Modern World. Princeton. Raychaudhuri, T., and Habib, I. (1982). The Cambridge Economic History of India, vol. I, c.1200–c.1750. Cambridge. Reece, R. (1973). ‘Wages and prices’, in D. E. Strong (ed.), Archaeological Theory and Practice. London, 240–5. Reynolds, J. M. (1971) ‘Libya and Diocletian’s Edict on Maximum Prices’, Libya antiqua 8: 33–42. Temin, P. (2006). ‘Estimating the GDP of the Early Roman Empire’, in E. Lo Cascio (ed.), Innovazione tecnica e progresso economico nel mondo romano. Bari, 31–54. van Zanden, J. L. (1999). ‘Wages and the standards of living in Europe, 1500–1800’, European Review of Economic History 3: 175–98. Ward-Perkins, B. (2005). The Fall of Rome and the End of Civilization. Oxford.

17 New Ways of Studying Incomes in the Roman Economy Walter Scheidel

In his chapter Dominic Rathbone highlights many of the problems that undermine modern attempts to reconstruct price and income levels in the Roman world. In my comments, I touch very brieXy on three ways of expanding the scope of our studies beyond the limits of the existing and mostly inadequate data on costs and wages, by considering the determinants of real incomes; the use of proxy data for real incomes; and the potential of cross-cultural comparison.

D E T E R MINANTS OF R E AL INC OMES Real wages are strongly inXuenced by demographic conditions, given that they tend to be fairly sensitive to deviations from an economic system’s equilibrium level of population size. A classic exposition of this principle is provided by Lee’s analysis of the correlation of real wages and demographic growth rates in pre-industrial England (Fig. 17.1).1 Papyrological evidence from Roman Egypt oVers an opportunity to study this relationship in an ancient economy. However, my earlier 1 Lee 1986. For similar as well as diVerent correlations elsewhere, cf. Reher and Ortega Osona 2000.

New Ways of Studying Roman Incomes Population growth rate

Real wage

Deviation from mean value

Population size

347

Time

Fig. 17.1. Model of population regulation driven by exogenous change in the population growth rate Source: Lee 1986.

attempt to explain an apparent increase in real wages (and concurrent drop in land rents) in third-century ad Egypt with reference to population losses incurred in the ‘Antonine plague’ of the late second century ad (Table 17.1) has met with some scepticism due to the fact that even in that region, the available data may be insuYcient to test this assumption in a rigorous fashion.2 These uncertainties, in turn, eVectively rule out comparable empirical studies of other, less well documented parts of the Roman empire. But all is not lost. Even in the absence of reliable or representative wage and price data, it may nevertheless be possible to identify factors that are likely to aVect real incomes, and to surmise their probable consequences. In a Wrst step, I have set up a matrix of variables which—judging from comparative evidence—can reasonably be expected to have mediated real income levels in Roman Italy (Table 17.2).3 This approach, necessarily fraught with uncertainties about the precise properties and relative weight of each variable, works best in those cases where consistent trends emerge from the record, that is, when all or most factors are either putatively favourable or unfavourable to real income growth. As I have argued elsewhere, 2 Scheidel 2002; Bagnall 2002. 3 Scheidel 2007.

348

Walter Scheidel

Table 17.1. Standardized index of real wages in Roman Egypt (2nd century ad ¼100) Ratio

Period

Mean

Median

Daily wage/wheat

2nd c. 3rd c. 2nd c. 3rd c. 2nd c. 3rd c. 2nd c. 3rd c.

100 121 (111) 100 115 100 147 (134) 100 134

100 102 100 134 100 134 100 156

Monthly wage/wheat Daily wage/wine & oil Monthly wage/wine & oil

Source: Scheidel 2002: the Wgures in parentheses are based on all daily wages from the 3rd century, excluding the data from Memphis.

Table 17.2. Proxy variables for assessing the probability of real income growth in Roman Italy (favourable conditions in italics)

Demographic attrition Geographical mobility Net free population growth Net expansion of slavery Redistribution Moderate/low Commercial development Tributary inXows Preconditions for, and indices of, real income growth

300–200 bc 200–100 bc 100–30 bc

30 bc–

High Moderate Stagnant Increasing

Moderate Moderate Moderate Strong

High High ?Stagnant/Moderate Strong

Moderate Moderate Moderate ?Stagnant

Moderate Moderate Low Mixed

Moderate Strong Growing Mixed

Strong Strong Growing Good

Slowing Shrinking Poor

Source: Scheidel 2007.

this method predicts real income growth at the end of the Republican period and a reversal thereafter. It is true that this model critically depends on speciWc assumptions, most notably about underlying trends in population size, and that signiWcantly diVerent assumptions would occasion substantial revisions. Even so, inasmuch as modern observers will Wnd it less taxing or hopeless to reconstruct general contextual conditions (such as the ones listed in Table 17.2) than to measure wages and costs per se, this approach, however conjectural, opens up a new avenue in the study of economic development that allows us to overcome the impasse created by the pervasive scarcity of income data.

New Ways of Studying Roman Incomes

349

PROXY DATA F OR REAL INCOMES Information on the cost of unfree labour may also shed some light on real incomes in the economy as a whole. Thus, we might conjecture that high slave prices are indicative of high wages overall, or vice versa. Jongman advances this argument in his chapter on consumption in the early Roman empire for the Cambridge Economic History of the Greco-Roman World, arguing that the substantial manumission premiums recorded at Delphi point to fairly high nominal as well as real wages for non-slave workers.4 We must bear in mind that much of this argument hinges on the representative nature of these freed slaves: if they were atypical—that is more highly skilled or otherwise more privileged than the average slave—their fees need not tell us a great deal about basic income levels. Moreover, my own survey of slave prices and free wages reveals further complications: broadly speaking, it appears that slaves in classical Athens were rather cheap (relative to wage levels), whereas in Roman Egypt (and probably in other parts of the empire as well) they were relatively more expensive (Table 17.3).5 This suggests that the relationship between slave prices and wages was more complex and mediated by secondary factors such as turnover risk, i.e. the stability of labour markets.6 Nonetheless, at least in principle, this indirect approach once again permits us to extend our studies beyond the narrow conWnes of existing datasets on wages.

C O M PAR AT I V E EV I DE NC E In his chapter, Rathbone raises the question of how the salaries of Roman government oYcials compared to those of Han China. A simple comparison shows dramatically higher levels of income as well as inequality among Roman imperial functionaries (Table 17.4). 4 Jongman 2007. 5 Scheidel 2008, based on data discussed in Scheidel 2005. 6 Discussed in Scheidel 2008.

350

Walter Scheidel

Table 17.3. Regional variation in real slave prices in the Graeco-Roman Mediterranean Wheat equivalent (in tonnes) Context (Delphi) (2nd c. bc (1st c. bc Italy (1st/2nd c. ad) Roman legal (2nd/3rd c. ad) Egypt (2nd/3rd c. ad) (Levant, 2nd c. ad (Dacia, 2nd c. ad (Dura, 3rd c. ad Price Edict (ad 301)

Unskilled rural labour (in years)

Range

Rough mean

?