Peasants, Citizens and Soldiers: Studies in the Demographic History of Roman Italy, 225 BC–AD 100 1107519128, 9781107519121

Recent years have witnessed an intense debate concerning the size of the population of Roman Italy. This book argues tha

241 43 2MB

English Pages 408 [407] Year 2015

Report DMCA / Copyright

DOWNLOAD PDF FILE

Table of contents :
Abstract
Title page
Contents
Tables
Maps
Note on abbreviations
Preface
1 Evidence, theories and models in Roman population history
2 Polybius’ manpower figures and the size of the Italian population on the eve of the Hannibalic War
3 Census procedures and the meaning of the republican and early-imperial census figures
4 Peasants, citizens and soldiers, 201 BC–28 BC
5 The Augustan census figures and Italy’s urban network
6 Survey archaeology and demographic developments in the Italian countryside
Epilogue
Appendix I: Cities and towns in early-imperial Cisalpina
Appendix II: Cities and towns in central and south Italy
Appendix III: Population figures for largest northern cities, AD 1600
Appendix IV: Some ingredients for a revised low-count model
Bibliography
Index
Recommend Papers

Peasants, Citizens and Soldiers: Studies in the Demographic History of Roman Italy, 225 BC–AD 100
 1107519128, 9781107519121

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

PEASANTS, CITIZENS AND SOLDIERS

Recent years have witnessed an intense debate concerning the size of the population of Roman Italy. This book argues that the combined literary, epigraphic and archaeological evidence supports the theory that early-imperial Italy had about 6 million inhabitants. At the same time the traditional view that the last century of the Republic witnessed a decline in the free Italian population is shown to be untenable. The main foci of its six chapters are: military participation rates; demographic recovery after the Second Punic War; the spread of slavery and the background to the Gracchan land reforms; the fast expansion of Italian towns after the Social War; emigration from Italy; and the fate of the Italian population during the first 150 years of the Principate. luuk de ligt is Professor of Ancient History at Leiden University. His book Fairs and Markets in the Roman Empire (1993) continues to be cited as a standard work on the rural economy of the Roman world. He has also published widely on Roman republican history and on the interplay between legal and economic developments. His edited volume, People, Land and Politics: Demographic Developments and the Transformation of Roman Italy, 300 bc –ad 14 (2008, with S. J. Northwood), has been greeted as a compulsory starting point for all those aspiring to understand the demographic, economic, social and political structures of Roman republican Italy.

PEASANTS, CITIZENS AND SOLDIERS Studies in the Demographic History of Roman Italy 225 bc –ad 100

LUUK DE LIGT

cambridge university press Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, São Paulo, Delhi, Mexico City Cambridge University Press The Edinburgh Building, Cambridge cb2 8ru, uk Published in the United States of America by Cambridge University Press, New York www.cambridge.org Information on this title: www.cambridge.org/9781107013186 © Luuk de Ligt 2012 This publication is in copyright. Subject to statutory exception and to the provisions of relevant collective licensing agreements, no reproduction of any part may take place without the written permission of Cambridge University Press. First published 2012 Printed in the United Kingdom at the University Press, Cambridge A catalogue record for this publication is available from the British Library Library of Congress Cataloging in Publication data Ligt, L. de. Peasants, citizens and soldiers: studies in the demographic history of Roman Italy 225 bc–ad 100 / Luuk de Ligt. p. cm. isbn 978-1-107-01318-6 1. Italy – Population – History. 2. Rome – History – Republic, 265–30 bc 3. Rome – History – Antonines, 96–192. I. Title. HB3599.L485 2012 304.60937ʹ09014–dc23 2011032182 isbn 978-1-107-01318-6 Hardback Cambridge University Press has no responsibility for the persistence or accuracy of URLs for external or third-party internet websites referred to in this publication, and does not guarantee that any content on such websites is, or will remain, accurate or appropriate.

Contents

page ix

List of tables List of maps Note on abbreviations Preface

xi xiii xv

Chapter 1 Evidence, theories and models in Roman population history 1 1.1 Introduction 1 1.2 Roman demography: low count versus high count 5 1.3 Traditional types of evidence: literary sources and rural survey data 1.4 New approaches 1.4.1 Evidence for land clearance 1.4.2 Evidence for the size of Italian towns 1.4.3 Osteological evidence 1.4.4 Estimates of carrying capacity 1.4.5 Comparative evidence 1.4.6 Climate change 1.4.7 Commercial grain imports 1.4.8 Imperial expansion and population growth 1.5 Old and new approaches to Roman demography: strengths and limitations

Chapter 2 Polybius’ manpower figures and the size of the Italian population on the eve of the Hannibalic War 2.1 Introduction 2.1.1 Low count 2.1.2 High count 2.1.3 Other interpretations 2.2 Some weaknesses of existing interpretations 2.2.1 Towards a new interpretation: the background to the tally of 225 bc 2.2.2 Age groups in the armies of the Republic 2.2.3 Some further implications

v

10 11 13 16 17 20 25 27 30 31 34 40 40 40 44 47 47 52 55 63

Contents

vi

2.4 The population of Italy in 225 bc 2.5 The Polybian manpower figures and Roman mobilization rates 2.6 Conclusions

71 72 77

Chapter 3 Census procedures and the meaning of the republican and early-imperial census figures 79 3.1 Introduction 79 3.1.1 A very short account of the republican census 80 3.2

3.3 3.4 3.5 3.6 3.7

3.8

3.1.2 The central problem: census procedures and registration rates The target population of the republican censuses: iuniores and seniores 3.2.1 Citizens sui iuris and citizens alieni iuris 3.2.2 Cives optimo iure and cives sine suffragio 3.2.3 Legionaries serving outside Italy 3.2.4 Assidui and proletarii Centralized and decentralized census procedures before the Social War The census figure for 86/85 bc Some other pieces of evidence Republican census procedures: some provisional conclusions A change in registration or reporting practices under Augustus? 3.7.1 Interpreting the Augustan census figures: philological and technical arguments 3.7.2 Comparative perspectives on the Augustan census figures Conclusions

Chapter 4 Peasants, citizens and soldiers, 201 bc–28 bc 4.1 Introduction 4.2 Developments between 201 bc and 163 bc: four questions 4.2.1 The impact of the Hannibalic War and the census figure for 203 bc 4.2.2 Demographic recovery after the Second Punic War 4.2.3 The rationale behind viritane distributions and colonization 4.2.4 The low count and the spread of agricultural slavery 4.3 Developments between 163 bc and 133 bc: the background to the Gracchan land reforms 4.3.1 The Gracchan land reforms in the literary tradition 4.3.2 Theories of demographic decline: poverty and its consequences 4.3.3 More slaves = fewer rural citizens? 4.3.4 Recruitment and losses on the battlefield 4.3.5 An alternative low-count reconstruction

81 82 83 87 95 98 106 112 116 120 120 121 128 134 135 135 137 138 142 150 154 157 158 159 162 165 167

Contents Competing low-count readings of the census figures for the period 163 bc–124 bc 4.3.7 The high count and the Gracchan land reforms 4.3.8 Survey archaeology and the Gracchan ‘crisis’ 4.4 Developments between 133 bc and 28 bc 4.4.1 Marius and the proletarianization of the legions 4.4.2 Emigration from Italy 4.4.3 Emigration, urbanization and the decline of the free population 4.5 Conclusions

vii

4.3.6

Chapter 5 The Augustan census figures and Italy’s urban network 5.1 Introduction 5.1.1 The geographical distribution of the population 5.1.2 Italy’s urban network at the time of Augustus 5.1.3 Definitional problems 5.1.4 Some methodological considerations 5.2 The urban network of Cisalpine Gaul in 28 bc 5.2.1 Some ingredients for a more detailed analysis 5.2.2 Expected urbanization rates 5.2.3 Expected urban population densities 5.2.4 A low-count model for Cisalpine Gaul 5.2.5 A high-count model for Cisalpine Gaul? 5.3 The population of central and southern Italy 5.3.1 Expected urbanization rates in central and southern Italy 5.3.2 Expected urban population densities in central and southern Italy 5.4 The urban network of central and southern Italy in 28 bc: a brief outline 5.4.1 Some low-count models for central and southern Italy 5.4.2 Some high-count models for central and southern Italy 5.5 Some general conclusions

171 177 179 182 183 184 187 191 193 193 194 196 199 202 205 210 211 213 224 227 228 230 233 235 238 241 242

Chapter 6 Survey archaeology and demographic developments in the Italian countryside 247 6.1 Introduction 247 6.2 Methodological problems 249 6.3 Diachronic patterns and their interpretation 254 6.4 Trends in site numbers, 300 bc–ad 100 257 6.5 Rural survey data and demographic models 265 Site densities in the suburbium: towards a low-count explanation 6.7 Land and labour in central-western Italy, 200 bc–ad 100 6.8 Conclusions 6.6

271 279 283

viii

Contents

Epilogue

284

Appendices i Cities and towns in early-imperial Cisalpina ii Cities and towns in central and south Italy iii Population figures for largest northern cities, ad 1600 iv Some ingredients for a revised low-count model Bibliography Index

289 304 337 340 345 382

Tables

1.1 Hopkins’ version of the low count 2.1 Roman and allied manpower figures in 225 bc according to Polybius 2.2 Allied manpower in 225 bc 4.1 Roman census figures (204 bc–114 bc) 4.2 Age structure of a population of 340,000 adult males 4.3 Alternative low-count model for the late Republic 5.1 Eight low-count models of urbanization in early Augustan Cisalpina 5.2 Eight low-count models of urbanization in early Augustan Transpadana 5.3 Four high-count models of urbanization in early Augustan Cisalpina 5.4 Eight low-count models of urbanization in central and southern Italy 5.5 Four high-count models of urbanization in central and southern Italy

ix

page 7 41 69 136 144 190 225 226 227 239 241

Maps

5.1. The large and medium-sized cities of northern Italy, c. 28 bc 5.2. The large and medium-sized cities of central and southern Italy, c. 28 bc

xi

page 209 237

Note on abbreviations

Names and texts of Greek authors are abbreviated in accordance with Liddell and Scott, Greek-English Lexicon, those of Latin authors in accordance with the Oxford Classical Dictionary, with only a handful of minor adaptations. For abbreviations of periodicals, the reader is referred to L’Année Philologique.

xiii

Preface

The origin of this book may be traced back to my inaugural lecture, in which I argued that the quantitative importance of rural slavery during the second century bc had been greatly exaggerated and that the widely held theory that the free rural population of central-western Italy began to decline in the late 160s bc rested on dubious premises. Even at that time it was clear that, if these basic ideas were correct, it would be necessary to reexamine not only the demographic history of Roman Italy during the last two centuries of the Republic, but a whole range of related issues, such as the economic and social background to the Gracchan land reforms, the gradual proletarianization of the legions, the growth of towns and the multitude of problems posed by the large quantities of archaeological data collected during the field-walking campaigns of the past 50 years. It was clear from the start that proper treatment of these topics would require a collaborative effort by a team of senior and junior researchers. I therefore owe an enormous debt to the Netherlands Organization for Scientific Research (NWO), whose generous grant made it possible for me to launch an extensive research project involving three PhD students and two post-doctoral researchers. During the past 5 years, I have incurred great personal and intellectual debts to those who have kept me on my toes by pulling me in completely different directions. From my friends and colleagues outside the Leiden Institute for History, I should like to single out Peter Garnsey, who volunteered to read the entire manuscript. Only he knows how different some sections of this book would have looked without his perceptive comments. I am also indebted to Bas van Bavel, Professor of Medieval History at Utrecht University, for valuable discussions on historical demography, and to John Bintliff, of the Leiden Faculty of Archaeology, for his acute observations on the methodological difficulties raised by the interpretation of the archaeological evidence. xv

xvi

Preface

In preparing the final manuscript I have also benefited from the critical observations made by the anonymous readers of the Cambridge University Press. Of the members of our research group, I should like to thank Saskia Hin for encouraging me not to give up entirely on the fragmentary literary sources. Without this stimulus, the second chapter of this book would never have been written. I also learned a great deal from Saskia Roselaar, whose firm grasp of the legal evidence can only be called frightening, and from Jeremia Pelgrom, who kept reminding me of the many ways in which pre-conceived conceptions of Roman and Italian society continue to shape our readings of the material record. Finally, I cherish happy memories of conversations with Simon Northwood and Paul Erdkamp, who turned out to be inexhaustible sources of information on Roman census procedures, on military matters and on the topic of migration. On the linguistic front, thanks are due to my long-time friend Rosamund Annetts, who displayed great ingenuity in bringing my English closer to native standards. As always, my greatest debts are to Olga, for her unwavering and loving support, and to Thijs and Lydian, for constantly drawing me into the fun of family life, travel and sports.

chapter 1

Evidence, theories and models in Roman population history

1.1 introduction The central importance of demographic trends to understanding economic and social developments in pre-modern and modern societies has long been recognized. As early as the eighteenth century, Malthus argued that if humans failed to control their sexual passions the population would inevitably grow faster than its agricultural output. The discrepancy between population and food resources resulting from this would then be resolved primarily by ‘Malthusian’ checks such as famines and large-scale epidemics.1 During the twentieth century, the study of population dynamics was taken up and developed by various medieval and early-modern historians working in the Annales tradition. Drawing their inspiration from Malthus, they tended to regard demographic growth as an autonomous factor. In studies of this type, demographic developments are seen as influencing a wide range of economic parameters such as levels of production, land productivity, labour productivity, wages, commodity prices and rents.2 An obvious weakness of a strictly Malthusian approach is that history provides quite a few examples of drastic demographic downturns that had little or nothing to do with any imbalance between population and resources. One famous illustration is the decimation of the native population of Peru, Mexico and North America by new diseases imported from Europe.3 Another classic example is the dramatic fall in European population levels brought about by the Black Death. It is true that Malthus listed ‘the whole train of common diseases and epidemics, wars, pestilence, plague and famine’ among his positive checks on population growth.4 That, however, does not alter the fundamental fact that the arrival and persistence of the bubonic plague had far less to do with any decline in living standards 1 2 4

Malthus (1798/1992). For a lucid discussion see Livi-Bacci (2001), 64–72. The classic study in this field is Le Roy Ladurie (1974). 3 E.g. Livi-Bacci (2001), 42–8. Malthus (1798/1992), 23.

1

2

Evidence, theories and models

resulting from excessive population growth than with the independent behaviour of micro-organisms, fleas and rats.5 Another interesting development is the recent upsurge of interest in climatic history. It has been pointed out, for instance, that from the late thirteenth century onwards large parts of northwest Europe seem to have suffered from increasingly unstable weather conditions. These caused the low-lying coastal areas of England, Holland, Flanders and Scandinavia to suffer from flooding, while the upper altitudinal margins of the rural economy in the interior districts were reduced. According to some specialists, these developments help to explain why some parts of medieval Europe began to experience serious economic problems long before the Black Death.6 Like those publications that emphasize the major demographic impact of the arrival and spread of lethal pathogens, the recent literature on climatic change questions the validity of the Malthusian assumption that population change operated as an independent variable. Reading through some of the key publications on the demographic history of medieval Europe, it is also possible to detect an increasing emphasis on social and political structures as major determinants of the concrete economic and social effects of changes in population levels. One illustration of this is the period following the Black Death, which witnessed the gradual abolition of traditional serfdom in England, but also saw an intensification of rural repression in eastern Europe. It has been convincingly argued that these completely different outcomes can be understood only by paying close attention to the distribution of political, social and economic power among kings, nobles, towns and peasants in the countries and areas concerned.7 In another wave of studies that began in the mid-1960s, the focus of attention shifted to the impact of economic, social and cultural factors as determinants of rates of population growth. Up to a point this involved a departure from the traditional Malthusian approach, because the 5

6

7

E.g. Hatcher and Bailey (2001), 57. The literary sources for the period that will be studied in this book report various outbreaks of epidemic disease; see e.g. Livy 27.23 (208 bc), Livy 28.46 and 29.10 (205 bc), Obseq. 6, 10, 13 and 22 (181 bc, 175 bc, 165 bc, 142 bc). Although some of these epidemics clearly had a considerable impact on a local or regional level, there is nothing to suggest that they had any discernible effect upon population trends in Italy as a whole. For this reason, epidemic diseases will play no part in this book. Recent literature on the Black Death is, of course, of fundamental importance to those interested in the demographic, economic and social impact of the pestis Antoniniana and the plague of the mid-sixth century ad. E.g. Bailey (1991). For a brief examination of the impact of climatic change on population change in Roman Italy, see section 1.4.6. E.g. Aston and Philpin (1985).

Introduction

3

underlying assumption was that population trends in the past were partly determined by the economic and social structures of the societies concerned. It has been suggested, for instance, that the relatively slow pace of population growth in northwest Europe in the early-modern period may have been connected to a pattern of late marriage for women.8 In at least some areas, this pattern, in its turn, seems to have been linked to opportunities to acquire farmland, determined by the distribution of property rights and by inheritance customs.9 Finally and perhaps most importantly, many specialists on demographic history have been influenced by the theories of Esther Boserup, who argued that both in the distant and in the more recent past population growth has often stimulated people to intensify production by employing more effective technologies, whether new or pre-existing. One of the insights underlying this theory is that it makes little sense to develop or use more intensive methods of agricultural production when demographic pressure is low because, at least in pre-industrial economies, the adoption of such methods often results in lower levels of productivity per hour worked. Boserup and her followers suggest that, precisely because pre-industrial societies offer plenty of scope for agricultural innovation, many pre-modern populations were able to avoid the Malthusian trap of diminishing returns by increasing per capita and overall output. This approach, in other words, overturns the Malthusian model by taking population growth as one of the main factors determining economic development, rather than as an inherently uncontrollable process that pushes a given agrarian economy to its productive limits.10 Although this brief survey of some important trends in historical demography is of necessity superficial, it suffices to show that, even if we abandon the Malthusian thesis of demographic development as an independent force that determines the fate of pre-modern populations, it is still possible to see population change as a major determinant of economic and social change. In fact, the extensive literature on demographic developments in pre-modern Europe makes it clear that we cannot hope to understand the 8 9

10

This approach was pioneered by Hajnal (1965). Cf. also Crone (1989), 152–4. E.g. Kertzer and Bretell (1987); Van Bavel (2002). For a brief discussion of Roman patterns of inheritance and their impact on ages of marriage, see Chapter 4, at notes 41–2. Boserup (1965), Livi-Bacci (2001), 72–5. For a critical assessment of Boserup’s theory from an archaeological perspective, see Morrison (1994). As Morrison observes, processes of intensification are often driven by factors other than population pressure (for instance, by market-driven, prestigedriven or politically driven demand). In Morrison (2007), she emphasizes the importance of power relations in a given society. Brookfield (2001), 199, followed by Fisher (2007), 93, holds Boserup responsible for imposing ‘a deterministic template’ on a diverse set of actual histories.

4

Evidence, theories and models

economic and social history of this or any other part of the world if we do not take population trends into account. In my view, there can be no doubt that this conclusion applies to the economic and social history of late-republican and early-imperial Italy. Here, too, competing reconstructions of population levels and of demographic developments are highly significant to our understanding not only of economic structures and social conditions, but also of at least some episodes of Roman political history and of the burden imposed by Roman imperialism.11 For this reason alone, it is not a realistic option to ignore Italian demographic history on the grounds that the surviving evidence is too scanty or too ambiguous. At the very least, we should try to clarify the economic, social, and military implications of the proposed reconstructions. It is equally important that we should assess the strengths and weaknesses of these reconstructions, using every method available, and examine the feasibility or otherwise of resolving whatever weaknesses may come to light. This book will focus principally upon demographic change in mainland Italy between the late third century bc and the first century of the Principate. One reason for narrowing down the chronological scope of our investigations is that our written sources concerning Roman or Italian population levels during the first 250 years of the Republic are few and of dubious quality.12 For the early-imperial period we have four census figures, but since three of these are Augustan it is difficult to say much about later developments. Since even for the period 225 bc–ad 14 the written record is fragmentary, we cannot afford to ignore the rich body of archaeological data brought to light by the extensive field-walking campaigns carried out in the Italian countryside over the past 50 years. For the specific purposes of my investigations, the most important archaeological periods are those during which black gloss pottery and early-imperial Italian sigillata were the most popular finewares in Roman Italy. These two periods span the last three centuries bc and the first century ad.13 Geographically, my enquiries will be limited to mainland Italy. This focus partly reflects the fact that most of the textual evidence relating to 11 12

13

Cf. below, at notes 36–7. According to Beloch (1886), 339–40 and Brunt (1971/1987), 26–7, all census figures relating to the period before the Latin War of 340–338 bc are unreliable. Cf. also the brilliant discussion by Alföldi (1977), 125–7, who argues that even some of the figures for the early third century bc are likely to be fictitious. For a more optimistic view see Coarelli (1988); Ward (1990); Lo Cascio (2001b), 566. Production of mid-republican black gloss may actually have begun in around 350 bc (Witcher 2008a, 276). According to some recent publications (e.g. Bousquet, Felici and Zampini 2008), the transition from early-imperial to mid-imperial Italian sigillata took place at around ad 80.

Roman demography: low count versus high count

5

Roman population history between the end of the First Punic War and the death of Augustus refers to developments in this area. Literary and archaeological material relating to the demographic history of the entire region is, moreover, so abundant (and so problematic) that no single monograph could do justice to it.14 It is not, of course, difficult to identify various forms of demographic and economic interaction between Italy and the provinces that directly affected population levels and urbanization rates in mainland Italy. Of these, emigration from Italy and the importation of large amounts of tax grain are perhaps the most obvious examples. Although it is extremely difficult to gauge the quantitative dimensions of these phenomena, they cannot be ignored. I shall not, however, attempt to contribute to the debate about population levels in the provinces, partly because my discussion of developments in Italy is already overloaded and partly because the surviving evidence seems to offer no basis for reliable conclusions.15 1.2 roman demography: low count versus high count In the light of the intensity and theoretical sophistication of scholarly debate concerning the causes and effects of demographic developments in medieval and early-modern Europe, it is in a way surprising that it has taken so long for ancient historians to accept certain of the challenges posed by historical demography. In fact, as Frier and Scheidel observed in 2001, demography played no part in the intense debates ignited by the publication of Finley’s The Ancient Economy and continued to languish as a marginal field of research until the early 1990s.16 This does not, of course, mean that ancient historians had no interest in the sizes of Greek and Roman populations prior to the publication of Parkin’s book on Roman demography and that of Bagnall’s and Frier’s monograph on the demography of Roman Egypt.17 In fact, when a growing number of specialists began to apply approaches and concepts originating in the field of historical demography to the ancient world, some of them did so 14

15 17

As noted in the preface, this book is just one product of a much larger research project investigating the demographic history of Roman Italy. For an in-depth discussion of the impact of demographic developments on patterns on landholding and on the evolution of property rights, see Roselaar (2010). Hin (2009) offers a systematic discussion of Roman patterns of fertility, mortality and migration. Pelgrom (forthcoming) looks at the evolution of settlement patterns as revealed by rural surveys and discusses attempts to establish a meaningful relationship between known or estimated population levels and changes in the archaeological record. Cf. below, note 35. 16 Frier (2001), 158; Scheidel (2001c), 3–10, 72. Parkin (1992); Bagnall and Frier (1994).

6

Evidence, theories and models

with the aim of rekindling an old debate about the size of the Roman population that began in the late nineteenth century. One of the most important landmarks in this controversy was the appearance of Julius Beloch’s Die Bevölkerung der griechisch-römischen Welt (1886), in which it was argued that during the early years of the Principate the free population of mainland Italy stood at around 4 million. Eighty-five years after the publication of Beloch’s monograph, his overall reconstruction was endorsed in Peter Brunt’s massive Italian Manpower (1971). In recent years the most prominent defender of this view has been Walter Scheidel, who has tried to buttress it with a plethora of new arguments, many based on comparisons with other pre-modern societies.18 In scholarly literature, the term ‘low count’ is now often used as convenient shorthand for the basic interpretation defended by this group of scholars. In the final analysis, the low count espoused by Beloch and Brunt rests on the interpretation of two pieces of numerical evidence found in the written sources. One of these is the famous survey of Roman and Italian manpower resources contained in the second book of Polybius’ Histories.19 These figures relate to the military strength of Rome on the eve of the Gallic invasion of 225 bc. According to Beloch and Brunt, the Polybian figures suggest that central and southern Italy had a free population of between 2.7 and 3 million.20 Beloch assigned between 0.5 million and 1 million inhabitants to Cisalpine Gaul, while Brunt put the free population of northern Italy at between 1 and 1.4 million.21 Using this last figure, Brunt ended by suggesting that mainland Italy might have had some 4.5 million free inhabitants in 225 bc. The other building block of the low count is a certain interpretation of the results of the census of 28 bc, when 4,063,000 capita civium were registered.22 Both Beloch and Brunt assumed that, unlike the republican census figures, this figure should be interpreted as including all men, women and children of citizen status. Assuming that an additional 1 million citizens remained unregistered, Brunt eventually arrived at an estimate of 4 million people of citizen status in Italy and a further 1 million in the provinces.23 Building on the figures for 225 bc and 28 bc, it is possible to draw certain conclusions about the trajectory followed by the free population of Italy. 18

19 21

22

E.g. Scheidel (1996a), 167–8. In some publications, for instance Scheidel (2008), Scheidel seems to adopt a more evasive position, but all of his recent publications on the sources of slaves, human mobility, political participation rates, real income growth and the size of the economy take the low count as their starting point. See e.g. Scheidel (1997), (2004), (2005a), (2006), (2007a) and (2009c). Plb. 2.24. 20 Beloch (1886), 413, 418, 435: 2.7 million; Brunt (1971/1987), 59, 120: 3 million. Brunt (1971/1987), 189. Although Beloch was reluctant to offer any specific estimate for Cisalpine Gaul (1886, 428), he did suggest 3.5 million as a maximum figure for the free and non-free population of peninsular Italy in 225 bc and 4 or 4.5 million as a maximum figure for Italy as a whole (1886, 435). Aug. RG 8.2. 23 Beloch (1886), 372–8; Brunt (1971/1987), 116.

Roman demography: low count versus high count

7

Table 1.1 Hopkins’ version of the low count A – Population changes, 225–28 bc Men, women and Adult males children (aged 17+ years) 225 bc Free Slave Total

4,500,000 500,000 5,000,000

28 bc

225 bc

28 bc

4,000,000 2,000,000 6,000,000

1,350,000 150,000 1,500,000

1,220,000 600,000 1,820,000

B – Rural/urban split Rural free Rural slaves Urban slaves Italian towns free City of Rome free Total

4,100,000 500,000 in all 250,000 150,000 5,000,000

C – Migration from Italy overseas Adult males (aged 17+ years) Before 69 bc 69–49 bc 49–28 bc Sub-total: Killed/counted twice over: Net migration

125,000 25,000 165,000 315,000 50,000 – 265,000

2,900,000 1,200,000 800,000 500,000 600,000 6,000,000

1,230,000 150,000 in all 75,000 45,000 1,500,000

870,000 360,000 240,000 150,000 200,000 1,820,000

D – Decline of free rural population Adult males (aged 17+ years) Migrants 225–28 bc to provinces: to Italian towns: Total loss:

265,000 100,000 365,000

Brunt argued that this trend was negative. He believed that there were about 500,000 fewer free Italians in the early years of the Principate than just prior to the Gallic invasion. At the same time, he argued that the slave population of Italy increased from c. 500,000 in 225 bc to c. 3 million in 28 bc. This figure would give early-imperial Italy a population of c. 7 million.24 Some other scholars who favour a low-count interpretation consider Brunt’s estimate of slave numbers too high. One such scholar is Hopkins, who puts the number of urban and rural slaves at c. 2 million and the total population of Italy at 6 million (Table 1.1).25 In his view too, however, the 24 25

Brunt (1971/1987), 124. Hopkins (1978), 68, following Beloch (1886), 415–16. Lower estimates in De Ligt (2004), 745–6; Scheidel (2005a), 70–1.

8

Evidence, theories and models

period 225–28 bc witnessed a significant decline in the free Italian population, a marked increase in the number of slaves and a robust expansion of the combined free and non-free population of Italy as a whole. Until recently, the only alternative theory was that by 28 bc there were about three times as many free Italians as Beloch and his followers believed. This view too dates from the late nineteenth century. It is, however, especially associated with the American ancient historian Tenney Frank, who put the number of free adult male Italians during the time of Augustus at 3.5 million and the total Italian population (including slaves and foreigners) at 14 million.26 During the second half of the twentieth century, this theory was taken up first by Wiseman and then by Lo Cascio. Their reconstruction has become known as the ‘high count’.27 Polybius’ manpower figures had only a marginal role in early expositions of the ‘high count’ theory. In a series of recent publications, however, Lo Cascio has used these data to estimate the size of the Italian population on the eve of the Second Punic War, arguing that they point to a free population of c. 3.4 million for central and southern Italy (excluding Bruttium and the territories of the Greek cities). Extending the average population density implied by this figure to the areas not mentioned by Polybius (including Cisalpine Gaul), he arrives at an estimate of between 6 and 8 million people for pre-Hannibalic Italy as a whole.28 Lo Cascio’s interpretation of the Augustan census figures is more or less identical to that of Frank, who took them to refer to adult male citizens only. Assuming that males aged 17 years or over made up c. 30 per cent of the population, we arrive at a total citizen population of c. 13.4 million at the time of the census of 28 bc. Since it is probable that at least 10 per cent of the population went unregistered, this figure must be increased to c. 14.75 million.29 Of these hypothetical 14.75 million citizens, roughly 1.25 million can be assigned to the provinces,30 leaving us with a figure of 13.5 million for the free population of Italy at the start of the Principate. If we repeat these calculations for ad 14, when 4,937,000 civium capita were registered, we end up with a free Italian population of about 16 million, if we assume that the number of citizens living abroad had risen to 1.9 million by this date.31

26 28 29

30

Frank (1924). 27 Wiseman (1969); Lo Cascio (1994). Lo Cascio and Malanima (2005), 201–2. Scheidel (1996a), 167. I do not understand Lo Cascio’s failure to address this important point in his most recent publications. Lo Cascio (1999a), 164. 31 (4,937,000 × 3.3 × 1.1) − 1,900,000 = 16,021,310.

Roman demography: low count versus high count

9

In various publications, Lo Cascio has also argued that slaves made up between 10 and 20 per cent of the population of early-imperial Italy.32 This would imply that there were between 1.5 million and 3.5 million slaves.33 Adding these to the free population of Italy suggests that in 28 bc the total population was between 15 and 17 million. According to this argument, the minimum figure for ad 14 would be 17.5 million. In the end, Lo Cascio assigns mainland Italy 14 million inhabitants in 28 bc and 15 to 16 million inhabitants in ad 14.34 All these estimates are on the low side. They could, however, be defended by assigning an additional 2 or 3 million citizens to the provinces.35 As we have seen, the most recent version of the high count proposes that the population of 225 bc was between 50 and 100 per cent larger than the low count would suggest. For the early Augustan period, the difference climbs to at least 150 per cent. If the high count could be proved to be correct, ancient historians would be forced to abandon many cherished conceptions about Roman society between, say, 250 bc and ad 150. For instance, if mainland Italy really had some 13.5 million free inhabitants in 28 bc, its population during the middle and late Republic must also have been much larger than previously thought. This would mean not only that military participation rates were far lower than has previously been believed (which would require a revision of republican Italy’s ‘militaristic’ image), but also that rates of political participation were lower (which would undermine any theory that we should take Polybius’ identification of a democratic element in the Roman constitution seriously).36 We would also have to accept that the demographic impact of the Second Punic War was quite limited and that Tiberius Gracchus’ claim that the expansion of slavery was pushing the free Italian peasantry into decline was completely false. It would, moreover, follow that the emigration 32 33

34 35

36

Lo Cascio and Malanima (2005), 204. In another article (2002, 62), Lo Cascio argues that slaves cannot have made up more than 15 or 20 per cent of the Italian population. The former figure would imply 2.5 million slaves. Lo Cascio and Malanima (2005), 203, 208, Table 2. The view that more citizens may properly be assigned to the provinces has been defended by Crawford (2008), 640–1. Cf. also Launaro (2008), 183–4. Valerius Maximus’ claim that Mithridates’ supporters killed 80,000 citizens in Asia in 88 bc (Val. Max. 9.2, ext. 3) is, however, generally rejected as utterly unreliable, and Cicero’s statement that ‘Gaul was packed with businessmen, full of Roman citizens’ (Font. 11) has no clear quantitative implications. Note that there is nothing to support Crawford’s suggestion that Sertorius, Pompey and Caesar might have bestowed Roman citizenship upon several million provincials. Interestingly, some experts on the history of particular provinces have argued that Brunt’s estimates for the number of citizens living in the provinces are too high rather than too low. Cf. Appendix iv. For contrasting views regarding the Roman constitution’s democratic element, see Millar (1984) and (1998), North (1990), Mouritsen (2001). On political participation rates, see Scheidel (2006).

10

Evidence, theories and models

of large numbers of veterans to colonies in the provinces between 49 bc and 28 bc had little to do with any boom in slave-staffed estates, but was a natural consequence of rapid population growth throughout Italy. Finally, and perhaps most importantly, we would have to posit the existence of a sophisticated economy capable of feeding 13.5 million free persons and between 1.5 million and 3 million slaves. As some high counters have pointed out, this ought to imply high yields in Italian agriculture, far larger imports of provincial tax grain than the existing literature would suggest, or a combination of both.37 This would make complete nonsense of the widely shared view that Italy’s economy was less sophisticated during the late Republic and early Empire than it was to become during the High Middle Ages and in earlymodern times. In fact, we would have to face the possibility that by 28 bc or ad 14 the Italian population had grown to a size that would not be surpassed until the late eighteenth or early nineteenth century. The proposition for which I shall argue in this book is that the high-count theory runs up against insuperable difficulties. I shall also, however, try to demonstrate that none of the various versions of the low count formulated by previous scholars is entirely satisfactory. In particular, I shall argue against the traditional low-count assertion that the Roman citizen body began to decline or to stagnate from the mid-second century bc onwards and also against the widely held view that accepting the low count necessarily means accepting that a dramatic decline in the number of free people living in rural Italy took place during the last two centuries of the Republic. It is my contention that, if these misconceptions are abandoned, the low count provides us with an interpretative framework superior to that offered by the high count. 1.3 traditional types of evidence: literary sources and rural survey data As we have seen, most early reconstructions of the demographic history of Roman Italy were based on competing interpretations of a handful of figures found in the written sources. In recent years, a number of scholars have expressed their dissatisfaction with the ways in which this type of evidence has been used by participants in the debate about population numbers.38 The reasons for this unease are easily detected. The ingenuity 37 38

Cf. below, sections 1.4.4 and 1.4.7. E.g. Scheidel (2004). Cf. also Jongman (2003) for a complete rejection of all literary references to rural depopulation and of the entire literary tradition concerning the background to the Gracchan land reforms. For a salutary warning against the dangers posed by this hypercritical approach, see Santangelo (2007b), 473.

New approaches

11

of those interpreting the figures preserved in the literary sources seems endless, with the predictable result that exactly the same data can be used to buttress diametrically opposed theories. A fortiori, the same observation applies to various non-quantitative statements made by ancient authors concerning the expansion of slavery and the decline in the free rural population. The moral overtones of these statements are evident and their frustratingly vague terminology is open to many interpretations. Against this background, one can only sympathize with those who have tried to develop new perspectives upon the demographic and agrarian history of Italy by drawing on the large body of data that has been collected – and is still being collected – by archaeologists in the Italian countryside. This approach was much favoured in the 1970s, when some specialists claimed that the archaeological record revealed the literary tradition concerning the social and demographic background to the Gracchan land reforms to be full of rhetorical distortions.39 As we shall see in a later chapter, however, the supposedly ‘hard’ data recovered by archaeological surveys are now acknowledged to pose almost as many problems of interpretation as the literary sources.40 It has been observed, for instance, that some of these campaigns resulted only in the identification of an impossibly small number of sites. This can be explained by assuming that many peasant dwellings were of flimsy construction and that their inhabitants did not use enough fineware to leave behind detectable plough scatters of diagnostic material.41 Alternatively, it can be argued that during some phases of Roman republican history a large part of the rural population lived in small, nucleated settlements that have escaped detection because they were situated on hilltops that have never been properly investigated.42 From a purely methodological viewpoint, this loss of innocence in the field of archaeology can only be applauded. Unfortunately, the other side of the coin is that archaeologists and ancient historians are now faced with the uncomfortable conclusion that every type of ‘evidence’ available to them is ‘soft’ in the sense of being compatible with a variety of interpretations.

1.4 new approaches In view of the enormous interpretational difficulties posed by literary evidence and rural survey data alike, it is perhaps not surprising that ancient historians have begun to look for alternative methods of shedding new light 39

Frederiksen (1970–1971).

40

Chapter 6.

41

Rathbone (1993b).

42

Pelgrom (2008).

12

Evidence, theories and models

upon Roman demographic history. Recent literature reveals at least four such approaches. The first of these approaches seeks to drive the debate over Italian population numbers forward by looking for new types of proxy evidence offering some insight into long-term population trends, changes in the balance between population and resources available for consumption, or even population levels during certain periods. While some archaeologists would still argue that rural survey data can be used for this purpose, recent studies have begun to explore the potential of other data sets, such as evidence for land reclamation, osteological data and archaeological and epigraphic evidence concerning the size of towns. A second wave of recent studies focuses on levels of productivity in Roman agriculture in an attempt to calculate how many people the farming population of late-republican and early-imperial Italy would have been capable of sustaining. Although some of the figures fed into calculations of this type are drawn from ancient or early-modern sources, these are essentially mathematical exercises that explore the implications of a given set of starting assumptions. A third tactic is to examine population levels in medieval and earlymodern Italy. Unlike the previous method, this comparative approach focuses upon agricultural systems and demographic regimes whose historical existence is undisputed, inviting us to ask specific questions about the similarities and differences between agricultural practices during different periods of Italian history. Finally, the authors of some recent publications have tried to sidestep the difficult questions raised by the surviving literary and archaeological evidence by investigating various factors likely to have favoured population growth. Following the example of some specialists in the field of later European history, certain ancient historians have begun to argue that a rise in average temperatures and a spectacular increase in public and private income from the Empire may have allowed the population of laterepublican Italy to reach levels higher than those recorded in medieval and early-modern times. In what follows I shall briefly explore the capacity of these new approaches to shed light on the issue of population size. Some of my findings will be negative. It will be argued, for instance, that modern reconstructions of climate change during the late Republic and early Empire and even the relatively abundant evidence for land reclamation in northern and central-western Italy are compatible with all existing models of population change in Roman Italy.

New approaches

13

For clarity’s sake I want to emphasize that, precisely because my starting question is narrowly quantitative, my negative conclusions do not imply that an in-depth investigation of topics such as land reclamation, carrying capacity and climate change cannot be expected to produce any interesting results. My point is merely that the various new types of evidence brought into play by some recent studies reveal little or nothing about the strengths and weaknesses of competing demographic models. For this reason, there seems to be little point in examining the results of these studies in fine detail, or in trying to integrate them into an analysis that draws upon completely different categories of evidence. I shall also try to make positive points concerning the four approaches that we shall be examining. One of these will be that comparative study of Italian population levels and of underlying economic structures is more useful than most other indirect or deductive approaches because it helps us to identify what may be called ‘the limit of the plausible’ within a much wider range of theoretical possibilities. My second positive point will be that if we accept the debatable premise that, at least in the long run, human populations tend to increase as more consumable resources become available to them, there is every reason to suspect that the expansion of the Italian population did not come to an end during the Augustan period, but continued until the end of the first century ad and quite possibly up until the arrival of the Antonine Plague. The ultimate aim of my survey of recent developments in the field of quantitative Roman demography is to establish which types of evidence are best suited to putting the explicit and implicit claims of competing demographic models to the test. My answer will be that there is no real alternative to the use of various categories of textual evidence and of certain types of archaeological data, with the crucial proviso that it would be naïve to use any type of evidence as a basis for straightforward inductive inferences. This is the conclusion underlying the specific research agenda that will be pursued in the five chapters forming the core of this book. 1.4.1 Evidence for land clearance In recent years some participants in the debate between low counters and high counters have tried to buttress their views by drawing attention to the demographic implications of written and archaeological evidence for the opening up of previously marginal areas for arable cultivation. Although no attempt has been made to put together a synthetic picture of Italy as a whole, various publications dealing with specific localities or regions offer

14

Evidence, theories and models

valuable information. It seems clear, for instance, that in the territory of the southern Etrurian town of Sutrium large tracts of land were cleared of forest as early as the third century bc, while in the Albegna Valley in centralwestern Etruria tracts of woodland seem to have been opened up for arable cultivation during the first decades of the second century bc.43 Recent work by Quilici and Quilici Gigli has revealed evidence of an expansion of arable cultivation in various parts of Latium, most of which seems to date from the first century bc, and evidence for deforestation has also been detected around Monte Massico in northern Campania.44 It seems reasonable to interpret these indications as pointing to population growth in central-western Italy. Unfortunately, this does not get us much further. The obvious reason for this is that all existing reconstructions of Italian population history, including the most pessimistic versions of the low count, assume that the expansion of the city of Rome stimulated agricultural production in South Etruria, Latium and Campania.45 The ploughing up of previously marginal land in these parts of Italy therefore cannot be used to back up any specific reconstruction of demographic development during the Republic. Further methodological difficulties are highlighted by developments in North Italy. It has often been noted that the Roman conquest of Aemilia and other northern areas was followed by large-scale reclamations that left clear traces in the form of centuriation grids. It seems reasonable to interpret these projects as evidence of demographic expansion.46 Any attempt to put even a very approximate figure on the populations of the areas concerned is, however, impeded by the fact that very little is known about patterns of land use and their evolution over time. In the case of Aquileia, for instance, it has been suggested that a large proportion of the land assigned to the colonial population was initially used for stock-raising rather than for arable farming or wine-growing.47 One clue in favour of this hypothesis is the fact that during the first century or so of its existence Aquileia appears to have imported wine from other parts of Italy; evidence of amphora production in its territory suggests that this town did not become a major exporter of wine until the last decades of the Republic. 43 45 46

47

Morselli (1980), 15; Cambi (2002), 142. 44 Quilici and Quilici Gigli (2009); Arthur (1991), 41. Morley (1996) remains the best account of this topic. Kron (2005a), 478–82. He correctly observes that the Romans managed to drain various low-lying parts of the Po delta (such as the Valli di Comacchio) that were subsequently abandoned, but fails to take into account the fact that land levels were often higher during republican and early-imperial times and that some areas were flooded as a result of later subsidence. Cf. e.g. Grove and Rackham (2001), 340–1. Bandelli (2009).

New approaches

15

This example suggests that, at least in this particular part of northern Italy, population growth stimulated the introduction of new crops requiring a higher input of labour per unit area. It also supports the idea that this development commenced during the last 150 years of the Republic. Our knowledge of the original population of Aquileia is, however, confined to the fact that it took in 3,000 adult male colonists (implying c. 10,000 free inhabitants of Latin status) in 181 bc and had to be reinforced in 169 bc.48 These very general observations do not, therefore, allow us to go beyond the conclusion that Aquileia and its rural territory are likely to have had substantially more than 10,000 inhabitants during the early years of the Principate. Although this information is not without interest, it is compatible with all existing reconstructions of the demographic history of republican Italy. In the specific case of northern Italy, we also have to face the fact that extensive alluvial deposits have obliterated most traces of Roman occupation in the lower-lying areas of the Po delta, making it almost impossible to follow any expansion of centuriation grids that may have taken place after the end of the Republic.49 This problem is all the more frustrating because at least some of the centuriation grids that have been detected in the north are now assigned to the early Empire.50 How, then, can we be certain that the amount of land under cultivation and the population of northern Italy peaked at precisely the time of the first emperor, as the high count seems to imply? In assessing the demographic implications of the reclamation schemes carried out in central-western Italy and in various parts of Cisalpine Gaul, we must also consider the possibility that the spread of more intensive forms of land use in these areas may have been partly offset by diminishing agricultural production in other areas. Some ancient historians have argued that large parts of Apulia, Magna Graecia, Lucania and Bruttium experienced a serious demographic downturn during the last two centuries of the Republic.51 As we shall see in a later chapter, not only is the significance of this trend disputed, but its chronology and the forces lying behind it remain under debate.52 Nevertheless, even those whose view of developments in the 48 49

50

51

Livy 40.34.2; 43.17.1. See e.g. Barbieri and Manzelli (2006), 196. Kron (2005a), 481–2, notes this problem with respect to the Republic, but ignores its relevance to the Empire. As Gabba (1989/1994), 200, points out, most of Transpadana’s centuriations are post-Caesarian. According to Capuis (1994), 73, the large grid to the south of Padova belongs to the reign of either Claudius or Nero. In central Italy, the centuriation grid of Alba Fucens has been dated to ad 149 (e.g. Iaculli 2008, 25). E.g. Toynbee (1965), followed by Cornell (1996). 52 Chapter 6.

Evidence, theories and models

16

deep south is particularly optimistic seem to agree that between the midthird century bc and the age of Augustus many towns in this area probably lost a significant proportion of their populations.53 This suggests that less intensive forms of agricultural production must have become increasingly widespread, except in those areas from which commercial crops could easily be exported to markets in other areas. At a more general level, Boserup’s observations regarding the relationship between demographic developments and agricultural intensification or its opposite cast serious doubt on any attempt to derive population estimates from the extent of the areas cultivated. In fact, as we shall presently see, completely different estimates of the size of the Italian population can be defended even if we assume (without good evidence) that all Italian land suitable for arable cultivation had been deforested, drained and ploughed up by the end of the Augustan period. I conclude that our evidence regarding land clearance in Roman Italy is compatible with all existing reconstructions of demographic developments. At least in theory, it should be possible to achieve further advances by collecting evidence from a larger number of regions, by refining chronologies and by combining the results with archaeological evidence concerning settlement densities and patterns of land use. For the moment, however, archaeological evidence of deforestation, drainage and other forms of land clearance cannot be expected to shed much new light on the problem of Italian population numbers. 1.4.2 Evidence for the size of Italian towns It is generally accepted that the last two centuries of the Republic witnessed a spectacular increase in the populations of Italian towns. According to Hopkins, the cities and towns of mainland Italy had about 1.9 million inhabitants at the time of the census of 28 bc, as against only about 650,000 in 225 bc.54 These estimates imply a trebling of the urban population during the last two centuries of the Republic and an urbanization rate of about 32 per cent for early-Augustan Italy. Since about 430 settlements of urban status existed during this period, it would also appear that the vast majority of Italian towns had tiny populations. In various publications, Lo Cascio has tried to undermine this reconstruction by arguing that early-Augustan Italy contained a fairly large number of substantial towns and that the total urban population was 53

Cf. Chapter 6, at note 73.

54

Hopkins (1978), 68. Cf. above, Table 1.

New approaches

17

much larger than Hopkins would suggest.55 In similar vein, Kron has argued that northern Italy contained more substantial towns in the age of Augustus than during the first decades of the fourteenth century or during earlymodern times.56 If these contentions could be proved to be correct, they would cast considerable doubt on the viability of the low count, since, even if we take into account the fact that Rome was the capital of a huge empire, an urbanization rate in excess of 32 per cent would be anomalous from a comparative point of view. There can be no doubt that this debate about the size of towns provides us with a promising new angle on the demographic history of Roman Italy. At the same time, even a superficial reading of the literature is enough to reveal that every reconstruction so far suggested rests on very weak foundations. The reason for this is that no one has ever attempted to determine how many large towns there were in late-republican and early-imperial Italy. The challenge posed by this unsatisfactory state of affairs will be taken up in Chapter 5, where it will be argued that the archaeological evidence for the physical size of the towns of Roman Italy provides strong support for the low count. 1.4.3 Osteological evidence Drawing on a relatively small body of skeletal evidence from central and southern Italy, Kron has tried to demonstrate that the mean height of adult male Italians between 500 bc and ad 500 was 168.3 cm. This figure contrasts strongly with the average height of Italian conscripts in 1854, which was only 162.6 cm.57 Arguing that height is determined largely by levels of protein consumption, Kron arrives at the conclusion that the inhabitants of Roman Italy must have enjoyed a higher economic standard of living than the Italians of any later period before the twentieth century.58 In recent years Klein-Goldewijk has assembled a much larger collection of osteological evidence as part of an ambitious PhD project. Although her database is still expanding and has not yet been made available in published form, Jongman has already used the material that was available in 2006 as a basis for certain provisional conclusions.59 His main finding is that the 55 57 58

59

Lo Cascio (1999a), 165, and especially (2009). 56 Kron (2005a), 474–5. Kron (2005a) and (2008), 81. If correct, this finding might be interpreted as supporting Kron’s optimistic view regarding levels of meat consumption in Roman Italy. Cf. below, at note 77. Jongman (2007b), 193–5.

18

Evidence, theories and models

advent of the Roman empire coincided with a spectacular increase in height in southern and northwestern Europe. In his view, the inhabitants of the Roman empire were not only taller than the Italians of the nineteenth century, but taller than any other pre-industrial Europeans. Like Kron, he interprets the Roman data as pointing to a significant increase in per capita income during the late Republic and early Empire. Can this evidence be used to shed new light on Italian population levels during the Republic and early Empire? In a recent article, Scheidel has argued that it can. His main argument is that the high count logically implies high population pressure in Italy and that this would have resulted in low levels of material well-being for the majority of the population. Since the bone evidence collected by Kron and the unpublished data used by Jongman seem to point in the opposite direction, it seems difficult to avoid the conclusion that the high count cannot be correct.60 Although Scheidel’s inference is certainly ingenious, it stands or falls with the reliability of Kron’s and Jongman’s findings. As noted above, Jongman’s conclusion is based on bone evidence from an area much larger than Italy. Most of the underlying data, in fact, come from Britain.61 As long as KleinGoldewijk’s much larger database remains unpublished, we cannot be sure that the patterns that appear to be emerging are valid for Italy. A recent article by Giannecchini and Moggi-Cecchi, using skeletal evidence from exactly the areas covered by Kron, has brought further methodological difficulties to light. Giannecchini and Moggi-Cecchi argue that proper analysis of the bone evidence from central and southern Italy leads to the conclusion that Italian males of the Roman period were actually a bit shorter than those of the Iron Age and also shorter than those of medieval times. It should be noted that Giannecchini and Moggi-Cecchi based their anthropometrical calculations on the Pearson statistical method. Kron, however, used the Olivier method, which they regard as amongst the least reliable available.62 It is difficult to avoid the conclusion that Kron’s finding that the Italians of Roman times were considerably taller than those of the mid-nineteenth century rests on a fragile methodology. This new twist in the debate has led Scheidel to adjust his earlier argument. He now thinks that the osteological data from Italy support an essentially Malthusian scenario in which Roman economic growth, 60 62

Scheidel (2008), 46–8. 61 Geertje Klein-Goldewijk, pers. comm. Giannecchini and Moggi-Cecchi (2008). One of their findings is that adult male Italians of the Roman period were on average 2.2 cm shorter than those of the Iron Age and 2.5 cm shorter than those of medieval times. Their figure for Roman Italy (164.4 cm) is close to Kron’s figure for the nineteenth century.

New approaches

19

if indeed any occurred, did not lead to gains in material well-being amongst any large part of the population.63 Even this modified conclusion is not immune to criticism. In an article that extends Giannecchini’s and Moggi-Cecchi’s data set and backs up their conclusions, Gowland and Garnsey argue that the decrease in mean Italian height that seems to have occurred after the Iron Age can be attributed to the fact that a significant proportion of the population of Roman Italy lived in towns. In their view, the relatively high level of urbanization that seems to have characterized late-republican and early-imperial Italy must have meant that large numbers of Italians were exposed to various diseases tending to flourish in densely settled areas. Conversely, the decline of towns during the early Middle Ages might help to explain why this period witnessed an increase in mean body height.64 These recent findings undermine any attempt to use changes in average or mean body height as straightforward evidence for changes in material as opposed to biological well-being, and thereby also any attempt to link changes in human stature to specific levels of demographic pressure. In this context it seems worth pointing out that research into the historical evolution of body height in various parts of medieval and early-modern Europe has brought further interpretational difficulties to light. It has been demonstrated, for instance, that during the Dutch Golden Age, which is generally seen as a period of high per capita income, the population of the western provinces of the Dutch Republic became on average significantly shorter. As in the case of late-republican and early-imperial Italy, one possible explanation is that an increase in levels of urbanization resulted in a decrease in average medical well-being. A recent study by Van Bavel, however, calls attention to another possibility. Since reconstructed levels of per capita income tell us nothing about income distribution, the Dutch data are compatible with the view that large sections of Dutch society actually became poorer during the Golden Age. In other words, the bone evidence from the Netherlands could be telling us that the increase in average material well-being that undoubtedly took place during this period should not be taken to mean that the wealth of the majority increased.65 Up to a point this reading of the Dutch bone evidence resembles Scheidel’s most recent model for early-imperial Italy. There is, however, a crucial difference. Since Van Bavel identifies changes in income distribution as the crucial variable, his model is essentially non-Malthusian. 63 65

Scheidel (2009a), 66. 64 Gowland and Garnsey (in press). Van Bavel (2008), 14, referring to Maat (2005).

20

Evidence, theories and models

The only possible conclusion seems to be that, since the assumption that an increase in levels of material well-being ought to be reflected by an increase in average or mean body height is demonstrably problematic, the idea that osteological evidence offers a reliable basis for conclusions about population size is fundamentally flawed. Viewed in this light, it is clearly no coincidence that almost all publications that explore the multiple and problematic connections between demographic developments and changes in body height use existing data concerning population size to arrive at an interpretation of the bone evidence, rather than the other way round.66 Since osteo-archaeology is a very young field, it would be unwise to rule out the possibility that new osteological evidence will eventually shed more light upon the relationships between population levels, settlement patterns and levels of material and biological well-being in Roman Italy. At the moment, however, we cannot go beyond the conclusion that the findings of Giannecchini and Moggi-Cecchi and of Gowland and Garnsey are compatible with a high-count model in which population growth eroded per capita income, but also with a low-count model in which urbanization, by exposing a growing proportion of the Italian population to a variety of old and new diseases, had a negative impact upon biological standards of living. These considerations suggest to me that, without additional osteological or other evidence by means of which one of these competing models can be confirmed or refuted, studies of Roman body height will not help us to determine whether early-Augustan Italy had 6 million or 15 million inhabitants. 1.4.4 Estimates of carrying capacity One of the first attempts to relate population levels to agricultural output in late-republican and early-imperial Italy may be found in Jongman’s book on the economy and society of Pompeii. Jongman’s main starting assumptions were that about 40 per cent of the land surface of mainland Italy was suitable for agriculture, that 80 per cent of this land was used for cereals and that every year half the land was left to lie fallow. If net output stood at 400 kilograms of wheat equivalent per hectare and annual consumption per head at 200 kilograms of wheat equivalent, it can be calculated that the agricultural system posited by Jongman would have been capable of 66

The bone evidence from Roman Italy could also be used to trace the spread and prevalence of malaria, which is thought to have been linked to land reclamation (Sallares 2002). Since our evidence concerning land reclamation is compatible with a wide range of population estimates, however, further research along these lines cannot be expected to resolve the low count/high count controversy.

New approaches

21

sustaining some 8 million people. This is close to Brunt’s estimate, which suggests that a population of 7 million could have been supported in the time of Augustus.67 Although this attempt to explore the economic foundations of the low count was strikingly innovative at the time, later publications have highlighted serious weaknesses in the underlying calculations. One such publication, an article by Lo Cascio, points out that there is no evidence for the assumption that biennial fallowing was the norm in the agrarian economy of late-republican and early-imperial Italy.68 Lo Cascio also thinks it entirely possible that average seed:yield ratios in Roman arable farming were considerably higher than the ratio of 1:4 underlying Jongman’s calculations. Finally, he points out that as much as 55 per cent of the land surface of Italy appears to have been used for agricultural purposes during the late nineteenth century and suggests that the proportion of land used to grow grain is likely to have been much higher than the 80 per cent proposed by Jongman. Lo Cascio demonstrates that had 55 per cent of its land surface been exploited using a short-fallow system, early-imperial Italy would have been able to feed about 12.4 million people. The remainder of the Italian population, he suggests, may have eaten imported grain.69 Other recent publications have suggested further revisions. Morley, for example, has argued that ancient historians have underestimated the role of barley in the agrarian economy of Roman Italy, and that an all-barley Italy could have sustained the population proposed by the high count.70 Adding another twist, Jongman has drawn attention to the fact that the populations of certain parts of late-medieval and early-modern Europe obtained a considerable proportion of their daily calories by consuming large amounts of beer. Inspired by this observation, he suggests that olives, olive oil and wine may have met 50 per cent of the daily calorific needs of the population of Roman Italy. If we were to accept this theory, it would follow that the carrying capacity of Italy’s soil was far greater than has previously been thought, for the obvious reason that output in calories per hectare is much higher in arboriculture than in cereal farming.71 67

68 71

Jongman (1988), 81. Morley (1996), 49, presents an almost identical set of calculations, but lowers the proportion of agricultural land actually cultivated to 75 per cent. This explains why he ends up with a hypothetical carrying capacity of 7.5 million people. Lo Cascio (1999b), 239. 69 Ibid. 240. 70 Morley (2001), 56. Jongman (2007a), 602–5. Of course, calorific output per hectare in viniculture and olive cultivation fluctuated enormously according to the type of cultivation employed. For instance, the discrepancy between these and cereal cultivation would have been significantly less on farms where intercultivation of grain between rows of trees was practised and in areas in which planting densities were low. For an illuminating discussion of some of these issues, see Mattingly (1993).

22

Evidence, theories and models

Not all of these modifications are equally compelling. To begin with Lo Cascio’s arguments, the work of both recent and earlier researchers into agricultural production in pre-modern and modern Italy suggests that the proportion of land brought under cultivation is unlikely to have been greater than 50 per cent.72 With regard to Morley’s calculations, it has been pointed out that his assumption that barley yields are higher than wheat yields is incorrect and that the sources do not in any way support the idea of an all-barley Italy.73 Finally, the little evidence we have regarding the dietary habits of Roman Italy does not support the idea that consumption of wine and olive oil accounted for more than 20 per cent of the average calorific intake of the majority of the population.74 A far more fundamental problem, however, affects any attempt to use ‘carrying capacity’ as a starting point for estimates of the Italian population in Roman or indeed later times. This problem has to do with the fact that the concept of ‘carrying capacity’ was first developed by biologists seeking to establish how many animals of a particular species could be sustained by the food resources of a given piece of territory.75 Although even in eco-biology the application of this concept is far from unproblematic,76 it is at least agreed that most animals are not capable of manipulating their environment in such a way as to double or treble the amount of food available for consumption. It is immediately obvious that this is not always true of humans. As we have already seen, this insight inspired Esther Boserup to question the validity of the essentially Malthusian view that before the Industrial Revolution human populations tended to expand at a faster rate than their agrarian resource bases, so that phases of rapid demographic expansion were always cut short by famine. If we accept her central thesis 72

73 74

75

The estimate of 55 per cent (e.g. Hopkins 1978, 7 n. 13) is based on Nissen (1883), I, 227, where we read that 2/15 of the territory of the Kingdom of Naples was completely useless, while 1/5 was used to graze animals and 1/8 consisted of woodland. All other land was used as farmland. While these figures, if taken at face value, would imply that 54 per cent of the land was suitable for agriculture, a close inspection of the underlying data (Nissen 1883, I, 436) reveals that the correct figure is c. 50 per cent. Cf. e.g. Witcher (2005), 130. Scheidel (2004), 7. According to Foxhall and Forbes (1982), 71 and 74, followed by Horden and Purcell (2000), 201, cereals are likely to have met about 70 per cent of the ancient world’s calorific needs. For Rome, cf. Garnsey (1991/1998), 239–40. Note his observation (ibid. 236–7) that the rations that Cato gave to his slaves contained a high proportion of cereals (Cat. Agr. 56–8). As demonstrated by Jongman (1988), 80–1, the rations of wine and olive oil given to these slaves would have covered at most 20 per cent of their daily calorific needs. Cf. also Gallant (1991), 66, for the fact that during the 1930s Sardinian peasants obtained 78 per cent of their daily calories from cereal products. In those areas in which the dominance of cereals was less overwhelming, legumes, rather than wine and olive oil, were the most important complement. Cf. Livi-Bacci (2001), 190. 76 E.g. Lewontin (2004), 9–10.

New approaches

23

that increases in population density are likely to have promoted the adoption of more intensive forms of agricultural production, the resultant increase in output may well have been enough to sustain the growing population. In other words, since population growth can be seen as a major stimulus to economic development, the whole idea of a fixed Malthusian ceiling beyond which pre-industrial populations were unable to expand may well be illusory. If we apply these ideas to late-republican and early-imperial Italy, it becomes possible to argue, as Geoffrey Kron has done, that the rapid increase in the Italian population implied by the high count stimulated the widespread adoption of rotation systems in Italian agriculture and also the introduction of further innovations such as ley-farming. The latter development would have been particularly important because it would have meant that many Roman farmers owned cattle, thereby acquiring large amounts of manure that could be used to fertilize arable land. The introduction and spread of sophisticated strategies of agricultural production could thus have made it possible for Roman farmers to achieve the high grain yields required to sustain a hypothetical population of between 15 and 16 million.77 Unfortunately, it is extremely difficult to reconstruct levels of productivity and production on the basis of the surviving literary and archaeological evidence. On the one hand, certain passages from the Roman agronomists suggest that a high proportion of agricultural land routinely lay fallow.78 On the other, Pliny the Elder details various rotation systems under which no more than one-third of the land was left uncultivated.79 Those with a low opinion of Roman agriculture tend to argue that the more advanced systems described by Pliny were to be found only in central-western Italy, in particular in Campania.80 This makes sense to the extent that the presence in this region of urban markets would no doubt have stimulated the adoption of rotation systems and the cultivation of fodder crops to be used in rearing animals for their meat. As we have seen, however, high counters such as Lo Cascio and Kron have defended the view that levels of urbanization in early-imperial times were similar to those proposed for the late-medieval and early-modern periods.81 If we combine this idea 77

78

79 80

For an eloquent defence of this theory, see Kron (2008). Cf. also (2000) for an interesting discussion of literary evidence for ley-farming in Roman Italy. E.g. Col., RR 2.9.4; 2.9.15; 2.10.7; 2.12.7–10; Plin., NH 18.187; Varro RR 1.44.2–4; 3.16.33. Cf. Brunt (1971/1987), 194; Duncan-Jones (1982), 49. Plin., NH 18.187, 191. Cf. White (1970), 110–24; Spurr (1986), 117–22. E.g. Jongman (1988), 80–1, 104. 81 Chapter 6.

24

Evidence, theories and models

with the high-count view that mainland Italy had about 15 or 16 million free and non-free inhabitants during the time of Augustus, we must conclude that at this time the urban market for commercial crops, meat and dairy products was enormous. As we have just seen, any attempt to verify this optimistic view of Roman agriculture by referring to the Roman agronomists is impeded by the fact that it is generally impossible to determine whether the practices that they describe were used only by a relatively small group of landowners and in certain areas, or by significant sections of the free farming population across Italy. It should be noted that certain other types of evidence pose similar problems of interpretation. One of Kron’s arguments is that, judging by the archaeozoological evidence, the mean height at the withers of Roman cattle considerably exceeded the mean heights calculated for the Iron Age and for the early Middle Ages.82 Since the size attained by cattle is determined largely by nutrition, it seems reasonable to conclude that a substantial group of Roman landowners used some of their holdings to breed cattle in a system of convertible husbandry. Since, however, neither the archaeological evidence nor the fragmentary literature on this subject allows us to determine how common such systems were, this finding does not necessarily imply that the majority of the farming population had easy access to large amounts of manure. In my view, the only way to get to grips with this problem is to investigate the extent of the urban market for meat during this period. In a later chapter I shall argue that there is good reason to believe that the joint population of the towns and cities of Italy was significantly smaller at the time of Augustus than in early-modern times.83 If I am correct, the superior levels of land productivity allegedly achieved in late-republican and early-imperial Italy cannot be explained in terms of a larger urban market. We could also consider the radically different hypothesis that a population of between 15 and 16 million might be sustained if the vast majority of its rural members were subsistence-orientated peasants who obtained very high yields by working long hours on tiny plots of land. One obvious weakness of this reconstruction is that it logically implies the existence of a steeply stratified rural society.84 Although the archaeological evidence suggests that some parts of Italy were densely populated at this time, no 82 84

Cf. Kron (2002). 83 Chapter 5. For a brief discussion of rural stratification in the Roman world, see De Ligt (1990), 49–51. DuncanJones (1976) documents differences in wealth between well-off landowners in Italy, but his discussion of this topic sheds no light upon the living conditions of the poorer sections of the rural population.

New approaches

25

one has as yet argued that the society of early-Augustan Italy was one of widespread rural impoverishment. Moreover, while it is historically true that the growing populations of certain Asian countries whose principal crop is rice have been sustained by the adoption of increasingly labourintensive forms of cultivation, it is generally agreed that the cultivation of cereals offers less scope for development along these lines.85 In any case, it would be difficult to prove that the rural society of late-republican and early-imperial Italy was more steeply stratified than that of late-medieval and early-modern Italy, which never sustained more than c. 12 million inhabitants.86 Finally, it will be argued in a later chapter that the period 30 bc–ad 100 witnessed vigorous population growth in various parts of Italy.87 If this is correct, it becomes even more difficult to explain the supposed capacity of Roman Italy’s agricultural system to sustain a population well in excess of 15 or 16 million by postulating the adoption of hyper-intensive forms of arable cultivation that have no parallel in later Italian history. 1.4.5 Comparative evidence In view of the inadequacy of the ancient evidence and the insuperable difficulties that would remain even if the proportion of land used for agricultural purposes could be established with complete certainty, it is tempting to reject any attempt to estimate the number of people sustainable solely by the produce of Italian soil during Roman times as completely useless. In my view, this pessimistic conclusion is correct with regard to any attempt to estimate the carrying capacity of Italy on the basis of hypothetical crop yields, reconstructed patterns of land use and probable levels of consumption. We may still, however, legitimately examine the Italian population levels of later periods. According to a recent reconstruction of the demographic history of Italy between 1300 and the early 1990s, mainland Italy (in fact somewhat larger than Roman Italy, if we except the territories of the ‘attributed’ tribes in the foothills of the Alps88) never had more than 12.1 million inhabitants before 85

86 87 88

The classic Asian example is Indonesia. See Geertz (1963). For the limited scope for intensification offered by arable cultivation in the Mediterranean world, see e.g. De Ligt (2007b), 6. For Italian population levels in medieval and early-modern times, see section 1.4.5. Chapter 6. In discussing medieval and early-modern times, Lo Cascio and Malanima (2005), 211, use an estimate of 310,000 km2. In the time of Augustus, between 225,000 km2 and 230,000 km2 of the land surface of mainland Italy were inhabited by Roman citizens. See Beloch (1886), 321.

26

Evidence, theories and models

ad 1700. During the eighteenth and nineteenth centuries ad its population began to increase far more rapidly than ever before, so that by ad 1850 it numbered c. 22 million.89 The crucial question here is, of course, how this steep rise can be explained. Various improvements in social and economic organization, along with important advances in the control of at least some infectious diseases, may be dated to the first half of the nineteenth century.90 It is, however, generally accepted that the growth in population that took place after ad 1700 may largely be attributed to the arrival and diffusion of new crops from the New World, of which maize was by far the most important.91 The main reason why the diffusion of this crop had a dramatic impact upon Italy’s ‘carrying capacity’ was simply that a field sown with maize provided twice as many calories as the same field sown with wheat. At the same time, maize was about half as expensive as wheat, so that the price of a ‘subsistence basket’ was considerably reduced. In the words of Paolo Malanima, ‘If the Italian population [of mainland Italy plus the islands: LdL] was able to attain more than 18 million inhabitants in 1800, this depended primarily on the spread of maize.’92 This simple observation raises the question of which special features of Roman agriculture, or of the economy of Roman Italy generally, could possibly have enabled the Italian population to climb to 15 or 16 million long before the arrival of any such crops. As we saw in the previous section, Kron attempted to solve this problem by positing the existence of a large urban market for meat that permitted the agricultural system of early-Augustan Italy to develop to a greater degree of sophistication than some later systems were able to achieve. Other participants in this debate have, however, suggested completely different solutions. It has been argued, for instance, that favourable climatic conditions could in part explain how the farmers of the late Republic were able to feed a population of the dimensions suggested by the high count. It has also been suggested that late-republican Italy reaped the benefits of a trading system whereby its goods were exchanged for provincial grain, or that most provincial taxes were paid in kind and that huge amounts of tax grain were shipped to Italy. The merits and demerits of these theories will be examined in the next three sections. 89 90 92

Del Panta et al. (1996), 277. Cf. also Malanima (1998). Del Panta et al. (1996), 143–5; Livi-Bacci (2001), 96. 91 Livi-Bacci (2001), 61. Lo Cascio and Malanima (2005), 221. I do not understand why Kron (2008), 72, makes no reference to maize whilst suggesting (on the grounds that both economies were overwhelmingly agrarian) that the agrarian economy of late-republican Rome may properly be compared to that of Italy in 1861.

New approaches

27

1.4.6 Climate change Scholarly interest in climate change was greatly stimulated by the French Annales school, one of whose favourite topics was change on a time-scale of hundreds or even thousands of years. The best-known study in this field is perhaps Le Roy Ladurie’s Times of Feast, Times of Famine, ambitiously subtitled A history of climate change since the year 1000. Among ancient historians, the theme of climatic developments was picked up by Shaw, who applied it to Roman North Africa.93 A few years later, Greene brought out an excellent survey of the existing literature in which he paid special attention to the time span covered by the Roman empire.94 One of his most interesting conclusions was that the emergence of Rome as a Mediterranean superpower coincided with a worldwide glacial retreat, accompanied by accelerated tree-ring growth in California and by a rise in Carbon 14 emissions, thought to reflect an increase in solar radiation levels.95 Until recently, these interesting insights played no part in any discussion of demographic developments in Roman republican Italy. In an article published in 2005, however, Lo Cascio and Malanima pointed out that even a very modest rise in average temperatures would have had a noticeable effect upon potential population growth. Their most important observation was that a temperature rise of one degree Celsius would have been enough to push up the maximum altitude at which cereal farming was possible by between 100 and 200 metres. In a hilly country such as Italy, the increase in the amount of cultivable land brought about by this relatively minor change would have exceeded 5 million hectares. Even if we assume that the grain yields of hilly districts were somewhat lower than those of the plains, it would have been possible to harvest enough additional crops to feed between 3 million and 4 million people.96 Although this attempt to link the rapid expansion of the Italian population implied by the high count to long-term climatic developments is both highly original and superficially persuasive, a closer examination of the basic argument exposes certain difficulties. The least serious of these is perhaps the fact that the positive effects of any rise in temperature cannot be assessed without detailed information about 93 95

96

Shaw (1981). 94 Greene (1986a), 83. See also Foraboschi (1992), 14–16, for an attempt to apply these insights to the history of Cisalpine Gaul. Her graph shows a striking rise in temperatures during the first centuries ad, but does not accurately reflect the fact that this upward trend began in around 300 bc. Lo Cascio and Malanima (2005), 219. Cf. Sallares (2002), 102, for the view that mean summer temperatures in early-imperial Italy were at least 0.5°C higher than those of the early-modern period.

28

Evidence, theories and models

levels of precipitation.97 The obvious reason for this is that if the said rise coincided with a period of desiccation, the beneficial effects of any increase in the amount of land cultivated would have been to some extent undone. The very few indications currently available do not suggest that this was the case, but much more research is clearly needed before any far-reaching conclusions can be drawn. Slightly more important is the fact that rising temperatures would have promoted the spread of certain diseases. The importance of this causal connection has recently been emphasized by Sallares. In his view, a combination of natural factors and human activities (deforestation, for example) stimulated the spread of malaria, seriously affecting the life expectancy of every age group in the areas affected.98 The message is clearly that it would be naïve to see a rise in temperatures as an entirely beneficial development in the context of this discussion. A third problem is that, even taking the most optimistic view of the environmental conditions created by an increase in average temperatures, the fact remains that better conditions for arable farming only created a potential for demographic growth. In other words, even if we accept the view that climatic conditions favoured population growth during the last three centuries bc, we still have to establish whether or not such an expansion actually occurred. In this context it might be pointed out that, despite the existence of the favourable circumstances implied by climatic models, the last two or three centuries bc are widely believed to have witnessed a population decline in many parts of Greece.99 Recent research has nuanced the picture of near-total desolation painted by some of the literary sources,100 but this critical reassessment has not resulted in any radically new vision of strong population growth similar to that proposed by Lo Cascio for Italy. The lesson to be learned is that, regardless of the importance we choose to assign to climate change as a stimulus to population growth, its actual impact can be assessed only by looking at how its effects were mediated by the specific economic and social structures that existed in various parts of republican Italy. Finally, it must be emphasized that climate change operates on a timescale spanning many centuries. Current research tends to confirm Greene’s provisional conclusion that average temperatures in the Mediterranean rose 97 98 100

As is explicitly acknowledged by Lo Cascio and Malanima (2005), 218. 99 Sallares (2002), 101–3. E.g. Bintliff (1997). Alcock (1993). Note, however, that her alternative model of urban growth is problematic with regard to many parts of Greece (De Ligt 1994b).

New approaches

29

by approximately one degree Celsius after about 300 bc and that these warmer climatic conditions persisted until the early fourth century ad.101 As we have seen, the high counters have argued that this change made it possible for the Italian population to grow to about 15 or 16 million by the time of Augustus. What tends to be overlooked, however, is the fact that the low count also identifies the late Republic as a period of population growth. It is true that in most low-count reconstructions this growth is attributed in part to the expansion of urban and rural slavery. It should not, however, be forgotten that even the most pessimistic version of the low count implies that immediately before the large-scale emigrations of the period 49 bc–28 bc the free population of Italy was about 10 per cent larger than it had been in 225 bc.102 For this reason alone it is difficult to see how climatological arguments can be used to buttress any specific model of population growth. Perhaps the only way forward is to pay more attention to demographic developments during the first 150 years of the Empire. As we have seen, the high count implies that the population of Italy peaked under Augustus, since all are agreed that it would have been difficult to fit more than 15 or 16 million people into the landscape of early-imperial Italy. If, however, we accept the climatological approach as basically sound, there is no reason why population growth should not have continued after 28 bc or ad 14. This problem is all the more urgent because at least some specialists think that average temperatures in the western Mediterranean peaked during the first century ad rather than in late-republican times.103 Viewed in this light, the finding that the period between 300 bc and ad 300 was characterized by higher temperatures actually supports the low-count theory that the Italian population grew from about 4.5 million in 225 bc to about 6 million in 28 bc, then continued to expand to about 7 million in ad 50 and to approximately 8.6 million by ad 160.104 These critical observations should not, of course, be taken to imply that research into the climatic history of the Mediterranean is a useless undertaking. My point is merely that, despite its undeniable attractions, the climatological approach cannot be expected to shed much light on demographic developments during a period that can only be called ‘short’ from the perspective of climatic history. Although some ancient historians may find this a disappointing conclusion, there is actually an important lesson to

101 103

See the excellent discussion in Hin (2009), 46–7. 102 Cf. De Ligt (2004), 732. E.g. Heide (1997), 88. 104 Beloch (1886), 437; Frier (2000), 812, 814.

30

Evidence, theories and models

be learned. Precisely because the kind of climate change that is of interest to ancient historians took place over a time-scale of about six centuries, we should pay more attention to developments after the Augustan period when trying to assess the strengths and weaknesses of the low and high counts.105 1.4.7 Commercial grain imports In a recent article, Lo Cascio calls attention to the potential benefits that could have been reaped from specialization and division of labour between Italy and the provinces. As he observes, there is good archaeological evidence that large amounts of wine and pottery were exported from Italy; we also know that large amounts of grain were imported. In Lo Cascio’s view, this pattern is similar to that of the High Middle Ages, when manufactured goods were sold outside Italy and wheat was imported.106 Against this suggestion, it can be pointed out that the basis of the ‘division of labour’ that Lo Cascio discerns in republican times was completely different from that of the inter-regional trade of medieval Florence or Venice. While at least some Italian cities of the High Middle Ages earned a lot of money by exporting manufactured goods and imported large amounts of grain on a commercial basis, most of the grain carried to republican and early-imperial Italy appears to have been tax grain.107 Interestingly, there are reasons for thinking that not only the grain for the grain dole in Rome but even some of the grain which was sold on the free market of the capital came from public granaries, both during the late Republic and in early-imperial times.108 Of course, Lo Cascio is absolutely right that, at least during the final century of the Republic, large amounts of wine were exported to southern Gaul. As far as we can tell, however, most of this wine was exchanged for slaves and metals.109 While this trade was no doubt profitable, it does not explain where or how Italy obtained the additional grain needed to feed 3 or 4 million consumers.

105 107

108

109

Cf. Chapter 6. 106 Lo Cascio and Malanima (2005), 219–20. On the weakness of the private grain market even in early-imperial times, see De Ligt (1991), 71–5; Erdkamp (2005), Bang (2008), 173–9. Garnsey (1988), 215, suggests that Roman magistrates may have sold Sicilian grain to traders or consumers at Rome between 123 bc and 62 bc. After the annexation of Egypt, Rome received far more tax grain than was needed for the distributions. It has long been suspected that some of the surplus grain was sold on the private market. See, for example, Duncan Jones (1990), 193–4, followed by De Ligt (1997), 517, Erdkamp (2005), 251, and Bang (2008), 177, n. 115. Tchernia (1986), 90–3.

New approaches

31

1.4.8 Imperial expansion and population growth In yet another attempt to explain how a population of between 15 and 16 million could have been sustained by the economic system of Augustan Italy, the high counters have called attention to the scope for further demographic growth created by the huge increase in imperial revenues that took place during the last two centuries of the Republic. For obvious reasons, the acquisition of additional food resources in the form of tax grain is seen as particularly important. In an article that appeared some ten years ago, in fact, Lo Cascio went so far as to claim that ‘the population of Rome and in all probability that of many other urban agglomerations on the coast was fed with provincial grain’.110 It would, at least in theory, be possible to widen the scope of this type of enquiry to include all the material benefits associated with the emergence of Rome as a Mediterranean superpower. About 30 years ago a comprehensive discussion of these benefits was offered by Hopkins, first in the opening chapter of his Conquerors and Slaves and then in his famous article on the intimate link between (monetary) taxes and trade in the Roman empire. With regard to the second century bc, Hopkins emphasized the importance of irregular forms of income such as booty and war indemnities. As time went by, these were gradually replaced by taxation.111 Although some provinces, such as Sicily and Sardinia, had been paying taxes since the late third century, an enormous increase in tax revenues took place after Pompey’s conquests in the East between 63 and 60 bc. In his Life of Pompey, Plutarch appears to say that before 63 bc state revenues had stood at 200 million sesterces and that Pompey’s annexations added 340 million sesterces. The text is, however, ambiguous, and most scholars assume that Plutarch was referring – or should have referred – to a rise from 200 million to 340 million.112 Whatever the correct interpretation, it is generally agreed that a series of further increases took place after 60 bc, with total imperial tax income rising to approximately 800 million sesterces by ad 25 and to 900 million sesterces by ad 50.113 As Hopkins noted, this spectacular increase in public revenues is only part of the story, since we must also take into account private income from government and, at a later stage, from landed estates in the provinces owned by Roman aristocrats. Like income from taxation, these private forms of 110 112 113

Lo Cascio (1999b), 239–40. 111 Hopkins (1978), 37. Plut. Pomp. 45. Hopkins (1978), 37, n. 49; Duncan-Jones (1990), 43. Hopkins (1980), 119: 800 million sesterces in c. 25 bc; (1995–96/2002), 200: 900 million in c. 50 ad; Duncan Jones (1994), 45: between 832 and 983 million in c. 150 ad.

32

Evidence, theories and models

income boosted the spending power of the Italian elite, thereby creating a market for the sale of goods of provincial origin. Adding a final twist, Hopkins observed that the benefits of empire appear to have percolated far beyond elite circles.114 Two illustrations of this are the booty acquired by common soldiers and the increasingly generous cash handouts given to veterans. Another is the grain dole for the citizen population of Rome that was gradually introduced. Initially grain was sold at a subsidized price to a limited number of beneficiaries, but from 58 bc the wheat was distributed without charge and the number of recipients quickly climbed to 320,000, until Caesar reduced it to 150,000.115 Added together, these snippets of information point to a vast increase in the resources that could have been used to sustain the population of Italy, suggesting that income from empire could partly explain an expansion in this population beyond the ‘ordinary’ pre-modern levels of the sixteenth and seventeenth centuries. While the overall validity of this line of argument cannot be denied, it must be emphasized that in other pre-industrial societies the link between increased material resources and population growth has often been tenuous and that the (limited) demographic effects of such an increase often became visible only after a couple of decades. In the case of early-modern England, for instance, painstaking statistical work has revealed that increases in average per capita income were followed by periods of more rapid population growth, but only after a delay of about 40 years. In most cases the annual rate of population growth crept up by no more than 0.1 per cent.116 Although this is only one example, it suggests that it might take some considerable time for a population to adapt itself to a higher level of material well-being. It has been argued that, partly because of large-scale emigration, the free population of Italy remained more or less constant during the last century of the Republic. According to this theory, the failure of the free population to ‘catch up’ with the considerable increase in material resources that took place between 133 and 31 bc resulted in a noticeable improvement in the material standards of living of a large proportion of the Italian population.117 114 115

116

117

Hopkins (1978), 38. Garnsey (1988), 211; Morley (1996), 36–7; Erdkamp (2005), 241. Under Augustus the number of recipients seems to have grown again, but in 2 bc they seem to have been limited to 200,000 (ibid. 236–7). Wrigley and Schofield (1981), 438. For critical discussion, see e.g. Flinn (1982); Laqueur (1993), 107–8. For an attempt to narrow the time lag between wage increases and increases in the marriage rate, see Wrigley (2004), 76–7. Scheidel (2007a).

New approaches

33

Of course, the theory works only if the low count is assumed to be the only possible interpretation of the demographic history of Italy. Even if this methodological drawback is acknowledged, nothing in the theoretical or comparative literature contradicts the basic idea that pre-industrial societies could have gone through phases in which population growth failed to keep pace with considerable increases in the resources available for consumption. In assessing the overall plausibility of the theory that additional income from empire helps to account for the high population levels envisaged by the high count, we must also ask how many additional Italians were actually supported by specific types of government revenue and expenditure. As far as military spending is concerned, it seems reasonable to assume that, at least after the suspension of tributum in 167 bc, provincial revenues such as the tax grain from Sicily and Sardinia were used to keep the legions of the Republic on the march.118 It should not, however, be forgotten that between 167 bc and the outbreak of the Social War in 91 bc there were never more than 50,000 Roman men on active service in the legions.119 Even during the first decades of the first century ad, when more than half of the state budget of c. 800 million sesterces was used for military purposes, the number of legionaries and auxiliaries sustained by the public purse was only about 300,000. A similar caveat must be expressed with regard to the grain dole. While the 320,000 free handouts of the period following 58 bc might have been enough to feed some 600,000 people, even this figure falls far short of explaining the discrepancy of at least 3 million between the high-count figure of 15/16 million for ad 14 and the population figures for mainland Italy before the eighteenth century. A far greater difficulty, however, stands in the way of any attempt to explain the large population posited by the high count as reflecting a high level of income from empire. As we have just seen, imperial tax revenues are thought to have increased from 200 million sesterces to 340 million sesterces by 60 bc, and it seems reasonable to suppose that this increase made it possible to introduce free handouts of wheat on a large scale. These developments are certainly spectacular. They came, however, far too late to account for the supposed existence of about 15 million free and non-free Italians in 28 bc. It is also important to bear in mind that the upward trend in imperial 118

119

As Erdkamp (2000) has pointed out, during the first half of the second century bc most of the grain from Sicily and Sardinia appears to have gone to the armies in the provinces rather than to the city of Rome. Hopkins (1978), 33. Allied units were fed at the expense of the communities that had supplied them. See Nicolet (1978/2000).

Evidence, theories and models

34

tax revenues continued after 60 bc. The most important step in this process was the annexation of Egypt in 30 bc, which might have contributed more than 250 million sesterces to the hypothetical early-imperial state budget of 800 million.120 Private income from the provinces presents an analogous problem. There is plenty of evidence that Roman senators and equites owned estates outside Italy from the early Principate onwards. It must, however, be emphasized that our abundant sources of information concerning the last decades of the Republic hardly ever refer to such provincial estates. This curious silence in the sources has led some ancient historians to hypothesize that there may have been a legal ban on the acquisition of provincial land by members of the Roman elite.121 Although this theory appears very fragile, it remains extremely likely that income from landed property in the provinces did not begin to make a significant contribution to overall Italian elite income until the first century ad. In discussing these various forms of revenue from empire, my principal aim has been to demonstrate that it is extremely difficult to predict the demographic effects of any increase in resources. If we adopt a strictly Malthusian approach, we can predict that population levels will eventually catch up with any expansion in the material resource base. Even if we ignore Boserup’s warnings and opt for this simplifying approach, it remains impossible to predict how long it will take for the population of any given society to reach the higher ceiling implied by such an increase. What makes any analysis along these lines even more difficult in the case of republican and early-imperial Italy is that the additional resources made available by imperial expansion continued to increase until 30 bc and beyond. As in the case of climate change, it is impossible to use this insight to buttress any specific reconstruction of the demographic history of Roman Italy. It does, however, alert us to the possibility that the Italian population may have continued to grow after the Augustan period and perhaps even until the beginning of the second century ad. 1.5 old and new approaches to roman demography: strengths and limitations My overall verdict on recent attempts to infuse the old debate about Italian population numbers with new life by focusing on levels of land productivity, by looking at population levels in medieval and early-modern Italy, by 120

Duncan Jones (1994), 53.

121

Rawson (1976), 90–1. Cf. also Hopkins (1978), 105.

Old and new approaches to Roman demography

35

taking into account some previously unexploited types of proxy evidence and by bringing into play general theories about the causal connections between climate change, economic growth and demographic developments is mixed. Certainly, estimates of the ‘carrying capacity’ of Italy based on varying sets of assumptions about the extent of the area cultivated and about average yields and consumption levels help us to delineate what Braudel called ‘the limits of the possible’.122 As a next step, the range of possible outcomes can be narrowed down with the help of comparative data. As we have seen, this approach casts considerable doubt on the general viability of the high count. A further gain to be had from the comparative approach is that it alerts us to some difficulties raised by the low count that clearly deserve more attention than they have hitherto received. How, for instance, do we explain the agrarian and social ‘crisis’ that apparently lay behind the Gracchan land reforms if Italy was as under-populated as the low count implies? Moreover, why should various Roman politicians of the second half of the first century bc have decided to settle hundreds of thousands of veterans in the provinces if the ‘carrying capacity’ of Italy – however conceptualized – could easily have sustained a much larger population? Valuable insights have also emerged from the recent search for new types of proxy evidence relevant to demographic developments. From amongst the various new approaches that have been suggested, I would single out that focusing on the size of towns as particularly promising, partly because the low and high counts make very specific predictions about the number of residents in the towns of late-republican and early-imperial Italy and there is a wealth of unexploited archaeological evidence that can be used to test these predictions. Finally, an examination of some of the recent literature on the potentially stimulating effects of climate change and of rising per capita income helps us to identify some promising avenues for research. As we have seen, the demographic impact of favourable economic and climatic conditions cannot be studied without taking into account the specific economic, social and military structures of Roman-republican society. Even if we keep this caveat in mind, however, it cannot be denied that climate change and economic growth seem to have stimulated population growth in some other premodern societies. Why then should the Roman citizen body, whilst continuing to reap extensive benefits from imperial conquest, have ceased to

122

Braudel (1981).

36

Evidence, theories and models

grow after the mid-160s bc, as Brunt and Hopkins have asked ancient historians to believe? If we apply the same line of reasoning to the final decades of the Republic and the early Empire, it is difficult to avoid the conclusion that climatic and economic circumstances continued to favour population growth throughout the first century ad. Of course, that does not prove that further demographic growth took place. Even if we go no further than the conclusion that a rise in temperatures and an increase in revenues from empire created the potential for further population growth, it would seem that far too little attention has been paid to the possibility that the first 100 or 150 years of the Empire witnessed an expansion in the populations of at least some parts of mainland Italy. Although these advances in understanding are clearly significant, it is important not to lose sight of the drawbacks and limitations of some of the new approaches that have found their way into the recent literature on the population history of Roman Italy. As we have seen, at least some of the archaeological data that have been produced as proxy evidence for population growth do not support any particular reconstruction of demographic developments. We should, moreover, have an eye to the dangers posed by those approaches that explore the demographic implications of general demographic theories and of quantitative models. Although engagement with such theories and models has the useful effect of forcing ancient historians to put their starting assumptions on the table, there remains a risk that the enthusiastic adoption of top-down methods by students of Roman history will lead to abstract discussions that are completely divorced from the surviving literary and archaeological evidence. It is easy, for instance, to see how approaches that seek to delineate the complex relationships between levels of productivity, patterns of consumption, increases in per capita income, climate change and demographic developments might encourage ancient historians to focus exclusively on the logical connections between statements made by other ancient historians, without any attempt being made to back up any specific reconstruction with the help of the surviving evidence, or, to use a more accurate formulation, with any concrete interpretations of that ‘evidence’. In my view, this particular kind of top-down approach can only lead to an infinite variety of ‘if-then’ formulations, which are ultimately sterile. I am particularly sceptical about attempts to estimate the ‘carrying capacity’ of Roman Italy. As we have seen, the main value of such estimates is that they establish the limits of the possible. This could, however, be seen as a fundamental weakness. The obvious reason for this is that even in

Old and new approaches to Roman demography

37

pre-modern societies, such as that of late-republican Italy, the limits of the possible were determined by a fairly large number of variables that can be quantified only very approximately, either because we simply have too little information or because at least some input values (such as seed:yield ratios) are known to have varied between regions. We therefore have to face the fundamental fact that hypothetical reconstructions of levels of production and consumption cannot be expected to reveal what happened within the limits of what was possible in the specific context of Roman-Italian society during the last two centuries of the Republic.123 I conclude that, while comparative evidence can be used to highlight some of the weaknesses of existing reconstructions, most recent approaches perform a better job in generating interesting new research questions than in providing us with plausible answers. Clearly it is an illusion to think that the demographic history of Roman Italy can be written without reference to the information contained in the literary sources or in modern archaeological reports. As noted in an earlier section of this chapter, at least some ancient historians are inclined to reject almost all types of literary and archaeological ‘evidence’ as unreliable or useless. If we push this approach to its logical conclusion, however, we shall end up with a tabula rasa offering room for an infinite number of competing reconstructions that cannot be verified. The pessimistic view that all things that can be known about ancient history are futile and that none of the really important questions can ever be answered has an obvious appeal (even to the author of this book). Nonetheless, it should be pointed out that this methodological stance is justified only if we assume that all interpretations of the ‘evidence’ are ultimately arbitrary. In my view, this assumption is unsustainable to the extent that no one will be convinced by any interpretation seen to ignore the fundamental ‘rules of the game’ that define ancient history as a scholarly discipline.124 Needless to say, the alternative to a non-committal theoretical or topdown approach need not be a return to old-fashioned inductivism. As we have seen, the principal weakness of approaches that focus on logical connections between variables is that they leave us with an infinite number of options. In discussing Roman population history we are, however, faced 123 124

For an in-depth discussion of this methodological problem, see De Ligt (1994a), 10–12. I realize, of course, that at least some of these ‘rules’ are open to negotiation or subject to developments over time. To me, the concept of compatibility is important (that is to say, the compatibility or otherwise of certain interpretations with other interpretations, and also with comparative data, some of which may themselves be interpretations). Proper treatment of this topic would require another book.

38

Evidence, theories and models

not with an endless variety of interpretations, but with only two conflicting reconstructions.125 In practical terms, this means that we can study the problems posed by the demographic history of republican Italy by starting from these two rival theories and examining the mutually exclusive interpretations of the written and archaeological evidence that they imply. This approach is based upon the idea that even if nearly all types of ‘evidence’ are soft in the sense of being compatible with multiple interpretations, it is still possible to assess the explanatory power of rival explanations of our evidence by looking at the number of ‘contrived’ interpretations required to make them work. In adopting this method, I fully accept that the distinction between ‘natural’ and ‘contrived’ interpretations will always be to some extent subjective. It is, however, to be hoped that at least some of my conclusions will be shared by those other ancient historians who play the game of ancient history under the rules upon which my investigation will be based. To conclude, I should like to add a few words about the range of topics that will be dealt with in this book. Some recent publications on Roman history have narrowed down the debate about Roman population levels to a discussion of the meaning of the Augustan census figures.126 In these publications written and archaeological evidence relating to the Republic recedes into the background, or even disappears altogether. This book will, however, pay close attention to written sources dealing with the third and second centuries bc, as well as to the republican material brought to light by the South Etruria surveys and by other field-survey campaigns. As in the case of the literary sources, I shall be asking what assumptions must be made about this material in order to produce a satisfactory fit with the frameworks supplied by the low and high counts. I shall also attempt to determine how these competing theories affect our interpretations of the development of the urban network of Italy as it took shape during the last two centuries bc. My decision to adopt this wide-ranging approach arises from my conviction that an understanding of the far-reaching implications of the low 125

126

For an attempt to argue in favour of a ‘middle count’ (which assigns between 7.5 and 10 million inhabitants to mainland Italy in 28 bc), see Hin (2008) and (2009). Although these publications contain a great many innovative insights, I cannot accept the author’s ingenious reinterpretations of the republican and Augustan census figures. In any case, Hin’s thesis that Augustus counted all men, women and children sui iuris because he wanted to introduce a tax on property logically implies that he also counted married women sui iuris. This means that she cannot assign to early Augustan Italy any more than 7.5 million inhabitants, which is very close to the estimates of 7 million and 6 million put forward by Brunt and Hopkins. For further discussion, see Chaper 3, at notes 193–200. E.g. Harris (2007), 516.

Old and new approaches to Roman demography

39

count/high count debate is of considerable importance to a very wide range of inter-related topics. In my view, no reasoned choice can be made between the completely different late-republican histories logically implied by these competing models until these implications have been set out in their full complexity.

chapter 2

Polybius’ manpower figures and the size of the Italian population on the eve of the Hannibalic War

2.1 introduction An important starting point for all reconstructions of the history of the Italian population during the Republic is the famous list of Roman and Italian manpower resources that appears in the second book of Polybius’ Histories.1 As Polybius explains, his collection of figures is meant to give an impression of Rome’s enormous power at the onset of the Second Punic War. His account pertains, however, to 225 bc, when a large-scale Gallic invasion threatened central and southern Italy. We are told that as soon as the Romans became aware of this threat they began to prepare themselves by stockpiling large quantities of grain, missiles and other resources required for war. They also carried out a survey to establish how many troops they and their allies would be able to mobilize in the event of war. According to Polybius, ‘the inhabitants of Italy, terror-struck by the invasion of the Gauls, no longer thought of themselves as the allies of Rome or regarded this war as undertaken to establish Roman supremacy, but every man considered that the peril was descending on himself and his own city and country. So there was great alacrity in obeying orders.’2 The results of the survey are shown in Table 2.1. 2.1.1 Low count Of the proponents of the low-count theory, Brunt has supplied the most detailed analysis of these figures. The basic assumptions behind his reconstruction are, however, identical to those to be found in Chapter 8 of Beloch’s Die Bevölkerung der griechisch-römischen Welt.3 It therefore seems appropriate to begin with the latter’s views, before discussing those of 1

Plb. 2.23–4.

2

Plb. 2.23.12.

3

Beloch (1886), 356–60.

40

Introduction

41

Table 2.1 Roman and allied manpower figures in 225 bc according to Polybius Foot soldiers

Horsemen

Total

A. Men under arms 1. Romans: With the consuls In Sicily and Tarentum Reserve at Rome Total

20,800 8,400 20,000 49,200

1,200 400 1,500 3,100

22,000 8,800 21,500 52,300

2. Allies: With the consuls Reserve at Rome Sabines and Etruscans Umbrians and Sarsinates Veneti and Cenomani Total

30,000 30,000 50,000+ 20,000 20,000 150,000+

2,000 2,000 4,000 8,000

32,000 32,000 54,000+ 20,000 20,000 158,000+

23,000 5,000 7,000 16,000 3,000 4,000 58,000

273,000 85,000 77,000 66,000 33,000 24,000 558,000

B. Men on the registers 250,000 80,000 70,000 50,000 30,000 20,000 500,000

Romans and Campanians Latins Samnites Iapygians and Messapians Lucanians Marsi, Marrucini etc. Total

Brunt’s arguments that altered, nuanced or strengthened the findings of his German predecessor. Beloch observes that the force available to the Romans and their allies numbered 699,200 foot and 69,100 horse. This is compatible with Polybius’ statement that the total number of men serving in the army or listed on the registers (apographai) amounted to ‘more than 700,000 foot and almost 70,000 horse’. Beloch also points out that Diodorus’ account of the invasion of 225 bc gives precisely the same figures, while Pliny the Elder states that at this time Rome could call upon 700,000 foot and 80,000 horse.4 Eutropius and Orosius, both of whom draw on Livy, give a round total figure of 800,000. Interestingly, it appears from these late authors that Livy himself took his figures from Fabius Pictor.5 As Beloch notes, this leaves us in no doubt that Polybius’ figures also came from this source. We 4

Diod. 25.13; Plin. HN 3.138.

5

Eutr. 3.5; Oros. 4.13.6.

42

Polybius’ manpower figures

are therefore dealing with material put together by a contemporary of the Second Punic War. Unfortunately, this does not mean that the figures given by Polybius are entirely unproblematic. There are certainly some minor problems, such as the horse:foot ratios given for various groups of allies. As Beloch points out, this ratio lies between 1:10 and 1:16 for the Romans and most of their allies, but is only 1:3 for the Apulians and 1:5 for the Marsi and their Oscanspeaking neighbours. Beloch solves this problem by amending the number of Apulian horse to 6,000 and by increasing the number of foot supplied by the Marsi and the other Abruzzi peoples to 40,000,6 thereby changing the ratios for these two groups to 1:8.3 and 1:10 respectively. More important is the discrepancy between the total suggested by Polybius and the census figure for 234 bc. If we are to believe Polybius (Fabius Pictor), the Romans were theoretically capable of mobilizing 325,300 men of military age in 225 bc. The census figure for 234 bc is, however, only 270,713.7 Since this figure is very close to the figure given by Polybius for the number of Roman men ‘on the lists’ (273,000), Beloch concludes that Fabius must have overlooked the fact that men already under arms were included on the lists. In other words, Polybius’ figure can be reconciled with the census of 234 bc only by assuming that he counted no fewer than 52,300 men twice. A completely different problem concerns the number given for allied troops, which is lower than the usual ratio between legionaries and allies in the Roman armies of the third and second centuries bc would lead one to expect. The best treatment of this difficulty is that of Brunt, whose attempt to explain these figures focuses on the Latins.8 Polybius states that the Latins were capable of furnishing 80,000 foot and 5,000 horse. Brunt ingeniously combines these figures with the number of Latin communities, of which, he asserts, there were thirty-six in 225 bc (twenty-eight Latin colonies and eight other towns of Latin status). Observing that the implied average of 2,300 fighting men per Latin community was much lower than the average number of male colonists (c. 3,800) sent out to each of six Latin colonies 6

7

Cf. Brunt (1971/1987), 49, who opts for 30,000. Beloch (1886), 360, justifies the correction by pointing out that Polybius’ figures for Roman and allied infantry do not add up to his total of ‘more than 700,000’ (2.24.16). Since the Etruscans and Sabines are credited with ‘more than 50,000 foot soldiers’, however, this particular argument is weak. La Regina (1970/1971), 447, points out that the corrections suggested by Beloch, Afzelius and Brunt are arbitrary and thinks that the figure of 20,000 can be retained. Since the separate figures for foot and horse add up to the total given by Polybius, it is surely most probable that Fabius did write 20,000, although it remains possible that he misread xxxmilia or xlmilia as xxmilia. Per. Livy 20. 8 Brunt (1971/1987), 56–7.

Introduction

43

established before 225 bc, he concludes that Beloch must have been right in suggesting that, unlike the Roman figures, the allied figures cannot have included all adult males. Brunt, like Beloch, takes the Latin figures to refer solely to iuniores (men aged between 17 and 45).9 Since this age group made up c. 75 per cent of the adult male population, his interpretation raises the number of adult males per Latin community to c. 3,000. If we assume that the returns submitted by the allies were c. 20 per cent defective, we can raise this figure to c. 3,700. This figure is very close to the average number of colonists suggested by the literary sources.10 In order to arrive at a grand estimate of the entire population of Italy at this time, we need estimated populations for Bruttium, for the Greek cities of southern Italy and for Cisalpine Gaul. As Beloch and Brunt have noted, the territory inhabited by the Bruttians and Greeks was roughly equal in size to that of the Lucanians.11 On the other hand, Magna Graecia seems to have been more fertile than either Lucania or Bruttium. From these facts Brunt infers that the Greeks and Bruttians together must have outnumbered the Lucanians. This leads him to conjecture that there must have been c. 210,000 Bruttians and Greeks, of whom c. 65,000 would have been adult males.12 Most difficult to estimate is the population of Cisalpine Gaul. The only ‘evidence’ that we have concerning this consists of scattered references to the number of troops fielded by various Gaulish tribes between 228 and 223 bc. One example is Polybius’ statement that the Cenomani and Veneti supplied c. 20,000 men to support the Romans in 225 bc.13 Another is his assertion that the Insubres mobilized 50,000 men in 223 bc, when their territory was invaded by a Roman army of c. 40,000 men.14 Beloch is inclined to accept the figure for the Veneti and the Cenomani, but dismisses all other references as unreliable. He therefore makes no attempt to estimate the population of Cisalpine Gaul in 225 bc.15 Despite sharing Beloch’s scepticism regarding the reliability of Polybius’ Gallic manpower figures, Brunt does attempt a conjectural estimate. He argues that the Insubres would surely have responded to the Roman attack of 223 bc by mobilizing every man they had. If he is right, the iuniores amongst the Insubrian population must have numbered c. 50,000.16 Brunt 9

10 14 15

Beloch’s and Brunt’s interpretation of the figures given for the allies has found wide acceptance. See e.g. Ilari (1974), 64; Wulff Alonso (1991), 158; Baronowski (1993), 183, 187; Lo Cascio (1999a), 168; Hin (2008), 191–3. Brunt (1971/1987), 56. 11 Beloch (1886), 358. 12 Brunt (1971/1987), 52, 59. 13 Plb. 2.24.7. Plb. 2.32.6. Cf. Beloch (1886), 428; Brunt (1971/1987), 185. Beloch (1886), 428. But cf. Chapter 1, at note 21. 16 Brunt (1971/1987), 185.

44

Polybius’ manpower figures

goes on to argue that the Boii were no less powerful than the Insubres and must therefore also have had c. 50,000 iuniores. By assigning a conjectural 100,000 iuniores to the Veneti and the lesser Gallic tribes (including the Cenomani) and a further 100,000 to the Ligurians, he arrives at a rough estimate of c. 300,000 iuniores in Cisalpine Gaul and Liguria. This would imply a total population of about 1 million. Having offered this estimate, however, Brunt admits that ‘a total of 400,000 or 500,000 could not be disproved’.17 In the end, he opts for a total northern population of the order of 1.4 million. Combining this estimate with those for central and southern Italy, we end up with a free Italian population of c. 4.5 million. If we follow Brunt in assuming that there were c. 500,000 slaves in Italy at the time of the Gallic invasion,18 we must conclude that the combined free and non-free population of Italy stood at about 5 million at this time. 2.1.2 High count If indeed mainland Italy had about 13.5 million free inhabitants at the time of Augustus,19 it is very hard to believe that it had a free population of only 4.5 million in 225 bc, since if we accept both figures we must assume an annual growth rate so high as to be markedly out of step with demographic data relating to other pre-modern societies.20 In other words, those opting for a high-count interpretation of the Augustan census figures must either reject Polybius’ manpower figures or reinterpret them. In an article published in 1999, Lo Cascio offered a re-analysis of the Polybian data designed to meet this challenge.21 Lo Cascio’s discussion starts from the observation that Polybius establishes a clear distinction between the numbers of men already on active service and the numbers reported by the Latin and Italian allies to be ready for duty. It appears clear that the former category of troops consisted of the two consular armies, the reserve legions in Rome, the legions stationed in Sicily and at Tarentum and the troops raised by emergency levies in Etruria, Sabinum and Umbria. The second category consisted of the fighting men to be fielded, in case of need, by the allied communities of central and southern Italy, with the exception of the Greeks and Bruttians. In Lo Cascio’s view, the obvious conclusion is that, contrary to the theories of Beloch and Brunt, Polybius was quite correct in calculating the total manpower 17 20

Ibid. 189. 18 Brunt (1971/1987), 67, followed by Hopkins (1978), 68. Cf. Chapter 3, section 3.7.2. 21 Lo Cascio (1999a).

19

Chapter 1, at note 31.

Introduction

45

resources available to Rome by adding together the figures given for these two sub-groups. Lo Cascio further modifies the low-count reading of the Polybian figures by questioning the traditional assumption that the figure for ‘the Romans and Campanians’ should be taken to include all Roman citizens, regardless of where they were domiciled. In his view the fact that Polybius groups the Sabines (who were Roman citizens) together with the Etruscans (most of whom were of allied status) proves that the levy of 225 bc was carried out on a purely regional basis, without regard to the juridical status of the regions’ inhabitants. If this is correct, the figures provided by Polybius should be interpreted as including all fighting men, whether Roman citizens, Latins or allies, living in the various regions to which he refers. This assumption is the background to Lo Cascio’s suggestion that the emergency army of the Etruscans and Sabines ‘was formed by 2/5 Roman citizens and 3/5 allies’. A corollary of his interpretation is that the 85,000 ‘Latins’ listed cannot have included the fighting men of the Latin communities in Samnium, Apulia and Etruria. It follows that Polybius’ figure for ‘Latins’ must refer solely to the inhabitants of the Latin communities of Latium Vetus and Latium Adiectum.22 As we have seen, the low-count interpretation favoured by Beloch and Brunt is based partly on the observation that Polybius’ figure for ‘the Romans and Campanians’ is very close to the census figure for 234 bc. Since Lo Cascio’s re-interpretation involves the assumption that the regional figures include many tens of thousands of Roman citizens, he cannot accept this traditional reading. In order to obviate this difficulty, he argues that Beloch and Brunt were wrong in regarding the census figure and Polybius’ ‘Romans and Campanians’ figure as in pari materia. According to Lo Cascio, there are good grounds to believe that the Roman censors targeted only that part of the adult male citizen population belonging to one of the thirty-five Roman tribes, ignoring the numerous cives sine suffragio who had no tribe. If we accept this theory, we must conclude that Polybius’ figure for the ‘Romans and Campanians’ included a substantial group of citizens not covered by the census figures.23 Lo Cascio also believes that the republican census figures and the tally of 225 bc included different age groups. Like Beloch and Brunt, he thinks that the censors were expected to register both iuniores and seniores. In interpreting the tally of 225 bc, however, he argues that the aim of the Roman government must have been to establish the total number of men available for active field service against the Gauls. Since there is good evidence to suggest that only 22

Ibid. 168.

23

Ibid. 167–8.

46

Polybius’ manpower figures

iuniores were expected to serve in military campaigns, Lo Cascio concludes that all of the figures given by Fabius/Polybius, including that for the ‘Romans and Campanians’, should be interpreted as referring only to men aged between 17 and 45.24 In his view, this interpretation is supported by the fact that Polybius uses the expression ‘those of military age’ (hoi en tais hêlikiais) to refer to the age groups covered by the allied manpower figures. As far as the military forces of the allied communities are concerned, then, Lo Cascio’s interpretation is identical to that of Beloch and Brunt. Unlike these two scholars, however, Lo Cascio thinks that the figures relating to Roman citizens, including those subsumed in the regional totals, must also be interpreted as referring exclusively to men aged between 17 and 45. Lo Cascio founds part of his argument upon a famous passage from Caesar’s De Bello Gallico that briefly mentions certain tabulae listing the members of the Celtic tribe of the Helvetii.25 These tabulae appear to have listed the names of ‘those able to bear arms’ (qui arma ferre possent) separately from those of children, old men (senes) and women, suggesting that (amongst the Helvetii, at least) senes were not classed as ‘able to bear arms’.26 If we accept Lo Cascio’s reinterpretation of Polybius’ figures, we arrive at a total of c. 355,000 Roman iuniores, pointing to an adult male citizen population of c. 473,00027 and to an aggregate citizen population of c. 1.56 million. Meanwhile, the figures for the allied communities (including those that can be reconstructed for the Greeks and Bruttians) suggest that there were at this time some 525,000 adult men of Latin and allied status and thus that the areas covered by Polybius’ figures had a total free population of roughly 3.5 million.28 In a more recent publication, Lo Cascio uses these figures to arrive at a rough estimate of the population of Italy as a whole. He begins by demonstrating that his reading of the Polybian manpower figures implies a population density of thirty-two people per square kilometre in the areas affected by the measures of 225 bc. If we apply this density to the territories of the Greek cities, to Bruttium and to Cisalpine Gaul, we obtain a total population of c. 8 million.29 Even if we assume that these areas had an 24 27

28 29

Ibid. 25 Caes. BG 1.29. 26 Lo Cascio (1999a) 168, followed by Hin (2008), 192. 325,000 Roman and Campanian iuniores + 10,000 Volsci + 20,000 Etruscan cives sine suffragio (Lo Cascio 1999a, 168). Scheidel (2008), 40, seems to overlook the fact that the figure of 514,000 adult males given by Lo Cascio (1999a, 169) includes c. 40,000 adult male Etruscans of allied status. Lo Cascio (1999a), 168–9. For southern and central Italy, assuming a density of thirty-two people per hectare would give a free population of 4.16 million, implying a total population of at least 4.3 million (including slaves). This is only about 20 per cent lower than the population of the same areas in ad 1600 (Del Panta et al. 1996, 277).

Some weaknesses of existing interpretations

47

average population density of the order of only eighteen people per square kilometre, we will obtain a total free population figure no lower than 6 million.30 On the basis of these estimates, it is possible to account for the existence of c. 14.75 million free people of citizen status at the time of Augustus by assuming that the number of free Italians grew at an average annual rate of about 0.45 per cent during the last two centuries of the Republic.31 2.1.3 Other interpretations In a recent article, Walter Scheidel has tried to cast doubt upon the value of Polybius’ manpower figures by drawing attention to the surprisingly neat ratios built into them.32 As he points out, the total number of allied infantry listed on the katagraphai is exactly equal to the number of Roman infantry ‘on the lists’, while the allied cavalry total amounts to one and a half times the Roman figure. This observation leads Scheidel to infer that Fabius’ breakdown of allied manpower figures must have been constructed ‘from the top down’ (that is, on the basis of known active and passive Roman troop strength).33 If this theory is correct, clearly Polybius’ figures (especially those relating to the allies) cannot be used as a basis for any demographic reconstruction. 2.2 some weaknesses of existing interpretations Since there would be little point in analysing the Polybian figures if these could be shown to be the results of imaginative calculations based on the composition of the Roman armies of the 220s bc, we must first try to assess the strengths and weaknesses of Scheidel’s theory that these figures can shed no light upon the demographic composition of Italy on the eve of the Second Punic War. Although Scheidel’s critical approach may have a certain appeal to those who argue that it is impossible to achieve any credible reconstruction of the demographic history of Roman Italy, the arguments that he uses are, in this 30 31

32 33

Lo Cascio and Malanima (2005), 201. The hypothetical figure of 14.75 million includes citizens living in the provinces. All forms of nonnatural growth resulting from the manumission of slaves and the enfranchisement of provincials are, of course, included in this calculation. Cf. De Ligt (2007c), 170–1. As Lo Cascio (2008), 242, n. 11, points out, his figures actually imply an average annual growth rate lying somewhere in the range between 0.31 and 0.49 per cent. My calculation is based on a starting figure of 6 million, which appears considerably more realistic than 8 million. Scheidel (2004), 4. It is not clear to me what Scheidel means by ‘passive troop strength’. If he means manpower reserves as revealed by military records, there would have been no need for Fabius to adopt a top-down approach.

48

Polybius’ manpower figures

specific case, weak. We are asked to believe, for instance, that Fabius Pictor simply invented his account of the submission of katagraphai recording allied manpower strength in 225 bc.34 As many scholars have pointed out, it is far more plausible that Fabius’ survey was based on an administrative document closely resembling the formula togatorum. Although very little is known either about the contents of this document or about how it was put together, there is no reason to doubt the widely held view that it was a survey of allied manpower resources.35 Secondly, the ‘smooth ratios’ observed by Scheidel appear only if we exclude the figures given by Polybius for the Etruscans and the Umbrians. As has often been observed, we have also to consider the Bruttians and the Greek cities of southern Italy. In short, Scheidel’s theory requires us to believe that Fabius did not merely extrapolate his total manpower figures from active or passive Roman troop strength, but did so in circumstances under which four important groups of allies had for some reason offered no troops to join the non-Roman contingents. Thirdly, Fabius/Polybius tells us that the two consular armies sent to Etruria to meet the Gauls were made up of 22,000 Roman citizens and 32,000 men of allied status.36 These figures demonstrate that Fabius regarded it as usual for the allies to provide one and a half times as many foot soldiers as the Romans.37 It is therefore unclear why he should have concocted hypothetical tallies based upon the assumption that the number of men supplied by the joint allies would exactly match the number of legionaries. Fourthly, if we accept the theory that Fabius estimated Roman and Italian manpower reserves on the basis of the size of the allied contingents already serving in the Roman armies of 225 bc, his figures can be rejected only if it is demonstrated that his method was clearly incorrect. Given this, it is odd that, having rejected Fabius’ testimony as worthless, Scheidel should go on to argue in favour of an estimate of 4 million free Italians, exactly the estimate implied by Beloch’s and Brunt’s interpretation of the 34

35 36 37

Scheidel tries to buttress his sceptical approach by pointing out that Fabius was happy to present the undoubtedly invented census figure of 80,000 for the reign of Servius Tullius as reliable (Livy 1.44.2). This argument glosses over the fact that, while Fabius is unlikely to have had access to reliable information regarding the censuses of the regal and early-republican period, he lived at the time of the Second Punic War. Cf. Forsythe (2005), 365. E.g. Toynbee (1965), 424–37; Brunt (1971/1987), 545–8; Ilari (1974), 57–85; De Ligt (2007a), 116–17. Above, Table 2.1. In a completely different context (3.107.12), Polybius claims that the number of allied foot soldiers roughly equals the number of legionaries, but his use of the present tense shows that at this point he is referring to the mid-second century bc. Cf. Chapter 3, at notes 58–67.

Some weaknesses of existing interpretations

49

Polybian manpower figures. Why, we are entitled to wonder, are we asked to set aside figures that are in perfect harmony with an estimate based on deductive reasoning?38 Finally, Scheidel’s inferences concerning the methods used by Fabius to construct his allegedly unreliable tallies presuppose that Fabius assumed, or thought that he could persuade his readers to believe, that the relative numbers of fighting men supplied by various groups of Italian allies mirrored differences in the free populations of the areas concerned. It is in my view highly significant that a well-informed contemporary observer should under any circumstances have made this assumption. In other words, even if Fabius could be shown to have constructed his tallies on the basis of known active Roman troop strength, a strong case could still be made for accepting the outcome of his calculations as reliable. For all these reasons, Scheidel’s attempt to set aside the Polybian manpower figures fails to convince. In fact, if we take into account not only the Umbrians and Etruscans, but also the Bruttians and the Greeks, whose manpower reserves may have amounted to some 65,000 men,39 the ratio of Romans to their peninsular allies rises to approximately 1:1.5. Since this ratio is also implied by the composition of the consular armies (and that of the reserve army near Rome) mobilized on the eve of the Gallic invasion, it is extremely odd to ignore it in favour of a purely hypothetical construct that seeks to establish a relationship of numerical parity between Romans and allies. It can therefore be argued that Roman demands for allied manpower were based on the principle that all contributions to Roman armies should be roughly proportional to the number of adult men (or to the number of iuniores) the communities concerned were able to field. According to this view, the ratio of Roman to allied troops revealed both by the katagraphai and by the composition of the Roman armies of 225 bc must roughly have mirrored the demographic realities of the 220s bc. Even if this moderately optimistic conclusion is accepted, it remains difficult to achieve a satisfactory interpretation of Polybius’ manpower figures. There can be no doubt, for instance, that Lo Cascio has identified 38

39

Scheidel (2004, 4) notes that his own top-down estimate of 3 million inhabitants for peninsular Italy in 225 bc ‘tallies well with Afzelius’ estimate of 3.1 million plus slaves that is not simply derived from Polybius but also takes account of likely carrying capacity and [. . .] comparative data for the agricultural population in 1936’. Afzelius did not, in fact, derive any population estimates from these comparative data, but used them to demonstrate that the regional variations in population density implied by Polybius’ figures were mirrored by similar variations in the 1930s. His aim was to confirm that Polybius’ account was reliable. Brunt (1971/1987) 59.

50

Polybius’ manpower figures

some serious weaknesses in the traditional low-count reading of these figures. His most forceful criticism is perhaps that Beloch’s and Brunt’s claim that Fabius Pictor is a first-rate source sits very uneasily with their assertion that this very same source made the huge mistake of counting 52,300 men twice. The traditional low-count interpretation, moreover, rests on the assumption that the figures for the ‘Romans and Campanians’ and those for the allies should be interpreted differently, so that the former are taken to refer to all adult male citizens, while the latter should be regarded as including only adult men aged between 17 and 45. Lo Cascio’s view that all of these figures should be interpreted in the same way (as referring solely to iuniores) is certainly more elegant, if only because it dispenses with the need to charge Fabius/Polybius with the mistake of comparing apples with oranges. Finally, Lo Cascio’s theory provides us with an interesting explanation for the fact that Polybius gives a separate figure for ‘the Etruscans and the Sabines’, despite the fact that the Sabines were of citizen status (either as cives optimo iure or as cives sine suffragio). These important advantages of Lo Cascio’s reinterpretation should not blind us to its weaknesses. As we have seen, the basic assumption underlying his reading is that the Polybian figures refer to a form of registration that was carried out on a strictly regional basis, without regard to the legal status of the inhabitants. While this assumption makes it possible to explain the grouping together of the Etruscans and the Sabines, it creates insuperable difficulties in the case of the Latins. As we have seen, Lo Cascio identifies the ‘Latins’ referred to by Fabius/Polybius as the inhabitants of Latium Vetus and Latium Adiectum. A fundamental problem with this theory is that both in Livy and in epigraphic documents of the second century bc the terms Latini and nomen Latinum invariably refer to those communities whose relationship with Rome was governed by the ius Latii, regardless of their whereabouts.40 It is clear that this undermines Lo Cascio’s theory that Fabius/Polybius paid no attention to juridical status when compiling his regionally based survey of Italian manpower, since if the Latins are to be identified as those governed by the ius Latii, it is surely most natural to suppose that the figure for the ‘Romans and Campanians’ included all fighting men of citizen status (with the possible exception of the Sabines). 40

In Livy 29.7 the thirty Latin colonies existing at the time of the Second Punic War are referred to as a coherent group. Cf. also the legationes socium nominis Latini referred to in Livy 41.8.6–8 (177 bc). Similarly, the expression ‘allies of the Latin name’ in line 21 of the epigraphic lex agraria of 111 bc (Crawford 1996, 115) clearly covers all Latin colonies. The only debatable point is whether or not the term ‘Latins’ included the prisci Latini of Cora, Tibur and Praeneste and the Hernician towns of Aletrium, Ferentinum and Verulae. Cf. below, note 105.

Some weaknesses of existing interpretations

51

Another weakness of Lo Cascio’s interpretation of the Polybian figures has to do with the balance between people and land. As we have seen, Lo Cascio begins from the assumption that the manpower figures for 225 bc are purely regional. This leads him to infer that the figure for ‘the Etruscans and the Sabines’ is to be read as including not only men of allied status but also those Roman citizens living in South Etruria and Sabinum. If, however, we accept that the citizen populations of Etruria and Sabinum were not included in the ‘Romans and Campanians’ figure, we must surely draw exactly the same conclusion regarding those adult male citizens (with or without the vote) living in southern Umbria, in western Samnium, in the Ager Picentinus or in the territories of the Praetuttii, the Vestini and the Aequi. A simple calculation is enough to show that this leaves a territory of at most 12,500 square kilometres for ‘the Romans and Campanians’.41 Since 325,300 adult male citizens would imply a total population of 1,073,490, the average population density for the area inhabited by ‘the Romans and Campanians’, as interpreted by Lo Cascio, would have been c. 86 people per square kilometre, an implausibly high figure.42 It would seem to follow from this that Lo Cascio’s regional reading of the Polybian manpower figures cannot be correct. In the light of these considerations, the population figures and population densities calculated by Lo Cascio on the basis of Pol. 2.23–4 must be regarded as extremely uncertain. More generally, it seems fair to conclude that neither the low counters nor the high counters have managed to come up with any convincing interpretation of the Polybian manpower figures to support their overall reconstructions of demographic developments between 225 bc and 28 bc.

41

42

According to Afzelius (1942), 192, the Ager Romanus covered some 26,805 km2 in 225 bc. From this area, the following subareas must be subtracted: Caere (840 km2), half of the territory of Tarquinii (640 km2), half of the territory of Falerii (345 km2), a very large part of the former territory of Vulci (940 km2), four communities in southern Umbria (860 km2), the Ager Praetuttianus (1,390 km2), upper Sabinum (3,305 km2), four praefecturae in western Samnium (1,715 km2), the territories of the Aequi and the so-called Vestini Romani (1,225 + 640 km2), and the Ager Picentinus, regardless of whether or not this was really ager Romanus (1,000 km2). If we also subtract the Ager Gallicus and Picenum, neither of which is mentioned by Polybius, only 7,670 square kilometres are left for the ‘Romans and Campanians’. In my view, this ‘omission’ militates against the theory that the Polybian figures should be read as regional totals. As will be demonstrated in Chapter 3 (at notes 44–51), this type of calculation can also be used to prove that the cives sine suffragio were included in the republican census figures for the third century bc. Lo Cascio (1999a), 168, himself argues in favour of an average population density of fifty-six free inhabitants per km2 for the pre-Hannibalic Ager Romanus. This hypothetical figure excludes slaves and assumes that, unlike the censors of the third and second centuries bc, those responsible for the survey of 225 bc achieved full registration of proletarians.

Polybius’ manpower figures

52

2.2.1 Towards a new interpretation: the background to the tally of 225 bc The sheer size of the scholarly output concerning the correct interpretation of the Polybian manpower figures strongly suggests that the chance of arriving at a new reading more convincing than the existing ones is remote. There is, however, one element of Polybius’ account that does not appear to have received the attention it deserves. It is my contention that, by giving this element its due weight, it is in fact possible to arrive at a coherent reading that avoids most of the difficulties surrounding the interpretations offered by Beloch, Brunt and Lo Cascio. My first step will be to take a closer look at the nature of the preparatory measures taken in 225 bc. Here the passage referring to the manpower contributed by ‘the Etruscans and the Sabines’ deserves special attention. According to Polybius, more than 54,000 Etruscan and Sabine foot soldiers and horsemen came to the assistance of Rome ek tou kairou (that is, in response to the emergency created by the impending Gallic invasion). It has long been recognized that the phrase ek tou kairou must here refer to a tumultuary levy, the tumultus in question being a tumultus Gallicus.43 What has generally been overlooked, however, is the fact that Polybius places the figure for the Etruscans and the Sabines on a par not only with that for the Umbrians and the Sarsinates (also mobilized via a tumultuary levy), but implicitly also with the manpower reserves whose existence was revealed by the list of Romans and Campanians compiled by the Roman authorities and by the lists provided by the Latin and Italian allies. In my view, this is the only plausible explanation for the fact that those allied peoples already mentioned in Polybius’ description of the tumultuary levies of 225 bc are missing from his survey of the remaining allied manpower reserves as revealed by the katagraphai, and vice versa (Table 2.1). If the information that Polybius provides is to be taken seriously, the complementary relationship that clearly exists between the two parts of his account can only mean that he (and Fabius) interpreted all the figures relating to the allies (and also those relating to the Romans and Campanians) as referring to the same categories of men. In other words, like the figures for the Etruscans and the Umbrians, those for the remaining allies and those for the Romans and Campanians should be interpreted as referring to the total number of men available for military service in the event of a tumultus Gallicus.

43

E.g. Toynbee (1965), 483 n. 3.

Some weaknesses of existing interpretations

53

This interpretation is supported by Pliny the Elder, who concludes his survey of the peoples and towns of Italy with the following reference to the events of 225 bc: Moreover, this is that Italy which, in the consulship of Lucius Aemilius Papus and Gaius Atilius Regulus, on receipt of news of a Gallic rising (tumultu Gallico nuntiato), single-handedly and without any alien auxiliaries, and moreover at that date without aid from Gaul north of the Po, equipped an army of 80,000 horse and 700,000 foot.44

It appears quite clear from this that Pliny interpreted the figures given by Fabius Pictor as referring to the results of a survey of Italian manpower resources carried out in the specific context of a Gallic tumultus. Like the complementary relationship between the two parts of Polybius’ survey of Roman and allied manpower resources, this implies that all of the figures given by Polybius/Fabius, including that for the ‘Romans and Campanians’, should be read as referring to the total number of citizen soldiers available for mobilization in the event of a Gallic attack. These considerations, in my view, shed new light on the technical implications of Polybius’ description of the preparations undertaken on the eve of the Gallic attack. As he tells us, ‘there was great and general alarm in Rome, as they thought they were in imminent and serious peril’. For this reason, he continues, the Romans ordered all their subjects to supply lists of their men of military age (hoi en tais hêlikiais) ‘as they wanted to know what their total forces (to sympan plêthos) amounted to’.45 Read in the light of the other indications already alluded to, this passage can only mean that the prospect of a Gallic invasion led the Romans to declare a tumultus Gallicus and to order an immediate survey to establish the number of fighting men available to them in an emergency of this specific type.46 Building on this simple observation, it is possible to circumvent a number of difficulties that have remained unresolved by the existing literature. To begin with, many adherents of the low count have been perturbed by the discrepancy between the total number of adult male Romans implied by 44

45 46

Plin. HN 3.138: super haec Italia, quae L. Aemilio Paulo C. Atilio Regulo cos. nuntiato Gallico tumultu sola sine externis ullis auxiliis atque etiam tunc sine Transpadanis equitum lxxx, peditum dcc armavit. Plb. 2.23.9. Gabba (1976, 176, n. 58) correctly points out that Polybius does not refer to a general tumultuary levy (my italics). Pliny’s statement that the preparations of 225 bc were made tumultu Gallico nuntiato, however, surely suggests that they took place in a general atmosphere strongly resembling that created by the declaration of a pan-Italian tumultus. There are, in any case, no good grounds for rejecting Polybius’ statement that the Roman authorities perceived a serious need to discover ‘what their total forces amounted to’.

54

Polybius’ manpower figures

the Polybian figures (325,300) and the census figure for 234 bc (270,713). This has prompted them to charge Fabius/Polybius with the gross error of overlooking the fact that the c. 52,300 Romans already mobilized must also have been listed as part of the manpower reserves of the Romans and Campanians. The underlying assumption here is that there would have been no need for the Roman authorities to compile new lists of citizen soldiers, since they already had the results of the census of 229 bc. Interestingly, however, Polybius explicitly states that 273,000 Romans and Campanians ‘had been listed’ (katelechthêsan) as part of the preparations of 225 bc. Although his language may be inaccurate, the most natural reading of this expression is that the Roman authorities had in fact carried out an update of their military records. If one accepts the view that the atmosphere at this time was that of a tumultus Gallicus, this is readily understandable. It is generally accepted amongst modern scholars that under normal circumstances the Roman censors made no attempt to register all proletarians,47 and it does not seem far-fetched to suppose that the Roman authorities were aware of this basic fact. Under normal circumstances, again, the existence of numerous unregistered proletarians is unlikely to have worried any Roman magistrate, for the obvious reason that proletarians did not normally fight in the legions and were not required to pay tributum.48 It must nevertheless be remembered that proletarii could be called up during emergencies, and in particular during the sort of emergency constituted by a tumultus Gallicus.49 It follows from this that if the Roman authorities really wanted to know ‘what their total forces amounted to’, they might well have supplemented the existing lists of adult male citizens with the names of as many proletarians as they could find. This means that there is no compelling reason to reject Polybius’ claim that even after the mobilization of 52,300 Roman citizens some 273,000 men remained on the lists; in other words, that a grand total of 325,300 adult male citizens was available for service in the legions. Needless to say, it does not follow that the tally of 225 bc was 100 per cent accurate. Be that as it may, the specific background to this tally presents us with good reason to think that it may well have been more accurate than any of the censuses taken before the Second Punic War. If we assume that 5 per cent of the target population went unregistered, it would

47 49

Cf. Chapter 3, at notes 90–1; Chapter 4, at notes 138–9. 48 Northwood (2008). Proletarians had been called up in 281 bc. See Oros. 4.1.3; Aug. CD 3.17; Cass. Hem., fr. 21, Peter; Ennius ap. Gell. 16.10.1. Cf. Brunt (1971/1987), 395, n. 6.

Some weaknesses of existing interpretations

55

appear that there must have been about 340,000 adult male Roman citizens in 225 bc.50 The discrepancy between the Polybian figure for the Romans and Campanians and the census figure for 234 bc can therefore be explained if we assume that a normal census would have revealed the existence of c. 280,000 adult male citizens51 and that approximately 17.5 per cent of the target population remained unregistered during the censuses of the third quarter of the third century bc. These assumptions are, in my view, entirely unproblematic, and certainly preferable to the alternative theory that Fabius, a well-informed senator and a contemporary of the Second Punic War, was incapable of distinguishing between men in the field and men on the registers.52 2.2.2 Age groups in the armies of the Republic If we read the Polybian figures as referring to adult men registered and mobilized in the context of a tumultus Gallicus, it is also possible to draw certain important conclusions concerning the age groups affected by the preparations of 225 bc. As we have seen, all of the adherents of the low-count theory seem to accept the view that the figure for the Romans and Campanians should be regarded as including both iuniores and seniores (in other words, all men aged between 17 and 60). Against this, Lo Cascio has pointed out that we are very clearly informed that the Latin and Italian allies were required to list ‘men of military age’ (hoi en tais hêlikiais). This expression is widely interpreted as referring exclusively to men eligible for active service in the 50

51

52

325,300 x 1.05 = 341,565. If we assume that the two ‘armies’ (stratopeda) stationed at Tarentum and in Sicily were mixed forces made up of 4,400 citizens and 4,400 allies (thus Ilari 1974, 68–9, n. 27; cf. Prag 2007, 72, n. 24), the number of adult male citizens drops to about 335,000. A further 21,500 men will have to be deducted if Ilari (1974, 69, n. 27) is right in suggesting that the four ‘urban’ legions referred to in Plb. 2.24.9 are to be identified with the four consular legions and thus never existed. On the other hand, my assumption that only 5 per cent of the target population remained unregistered in 225 bc may be too optimistic. I would, in fact, be happy to accept an estimate of c. 350,000 adult male citizens. This would imply that the Roman censuses of this period missed some 20 per cent of the target population. This assumes that the number of adult male citizens grew at an average annual rate of 0.3 per cent from 234 bc–225 bc. During the decades following the end of the First Punic War, the real growth rate may well have been higher. Cf. Chapter 4, section 4.2.2. Brunt’s argument that Fabius is unlikely to have gone through the complicated procedure of deducting the numbers of men already under arms from the total numbers available for service (Brunt 1971/1987, 47) can easily be countered. The fact that Fabius gave round figures for the numbers of allied foot soldiers still ‘on the lists’ suggests that he may simply have deducted the estimated number of troops already in the field and rounded the outcome of this simple operation to the nearest multiple of 10,000. His aim was clearly not to engage in complex or precise calculation.

56

Polybius’ manpower figures

field: the iuniores.53 Building on this interpretation, Lo Cascio reasons that the figure for the ‘Romans and Campanians’ should also be interpreted as including solely iuniores, since Fabius’ calculations must assume that all of his manpower figures refer to the same age groups. Lo Cascio also finds it significant that, after adding up the figures for the Romans and their allies, Polybius ends up with a total of more than 770,000 men ‘able to bear arms’ (dynamenôn hopla bastazein). Since under the Roman system of recruitment only men aged between 17 and 46 were expected to ‘bear arms’ by serving in the legions,54 we seem to be faced with the inescapable conclusion that in 225 bc the Roman authorities were interested only in the number of men eligible for ordinary service in the legions or in the allied contingents. Although this reading of the Polybian figures has a superficial attraction, it suffers from certain lethal weaknesses. One of these has to do with the meaning of the expressions hoi en tais hêlikiais and dynamenôn hopla bastazein. Polybius does not, it must be stressed, restrict his use of the former expression to his accounts of the composition of Roman or Italian armies; he also employs it in describing various episodes that took place in Hellenistic Greece. Now, according to the Greek tradition of warfare, all men aged between 18 and 60 were required to make themselves available for military service.55 One illustration of this fact is provided by the Athenaiôn Politeia, in which we read that ‘the heroes giving their names to the tribes are ten in number and those of the years of military age (tôn hêlikiôn) fortytwo’.56 As is universally agreed, the forty-two age classes referred to in this passage comprised all adult Athenian men aged between 18 and 59. Greek literature abounds in specific references to elderly men fighting on the battlefield. In the general levy of 418 bc, for example, all the citizens of Sparta, together with their serfs, marched to the border. Here the oldest and youngest were left to defend the border, while those aged between 20 and 54 went on to fight the battle of Mantinea.57 Socrates and Demosthenes both continued to participate in military expeditions abroad until they were over 50.58 The Greeks, we must conclude, regarded all men aged up to 60 as able to ‘bear arms’, not only as last-ditch defenders on the home front, but also as active members of expeditionary forces. There were, of course, many military campaigns in which the oldest and youngest citizens did not participate. Thucydides tells us that of the 29,000 53 55 57

58

The same interpretation may be found in Hin (2008), 191–3 and (2009), 166–7. 54 Plb. 6.19.5. Van Wees (2004), 46. 56 Ath. Pol. 53.4. Thuc. 5.64.2–3. Cf. Van Wees (2004), 46. According to Xen. Hell. 6.4.17, the levy for the battle of Leuctra also included men aged up to 54. Hanson (1989), 89–95.

Some weaknesses of existing interpretations

57

hoplites available to the Athenians in 431 bc, 16,000 men were used for garrison duty. As he explains, this body included the oldest and youngest citizens, along with the metics.59 The age group to which the phrase ‘the oldest citizens’ refers remains unknown. More detailed information exists concerning the fourth century. For example, an Athenian fleet of 353/2 bc was manned by citizens aged up to 45, and all Athenian hoplites aged under 40 were mobilized for the Lamian War of 323 bc.60 These examples suggest that it was unusual for men older than 45 to be called up for active service in the field. They do not, however, contradict the general principle that in classical Greece all adult men aged up to 60 were regarded as of military age. There are several indications that the military levies of the Hellenistic world were carried out according to the same general principles as those of fifth- and fourth-century Greece. Normally only men aged up to 45 were called up for frontline service, but in exceptional circumstances the call-up was extended to all men of military age. In 220 bc, when King Philip of Macedon invaded the Peloponnese, three of the Spartan ephors ordered ‘all men of military age’ (tous en tais hêlikiais) to assemble, armed, at the temple of Athena Chalkioikos ‘since the Macedonians were advancing on the city’.61 In this context it must be remembered that in classical and Hellenistic Sparta it was quite normal for men aged up to 60 to see active service in the field.62 In this specific case, moreover, we are clearly dealing with an emergency levy calling upon both the youngest and the oldest men regarded as capable of military service.63 An interesting parallel is provided by an incident that occurred ten years later in Acarnania. When their land was attacked by a large Aetolian army in 210 bc, the Acarnanians decided to send their women, their children, and men aged over 60 to Epirus, while every man aged between 15 and 60 took an oath not to return home unless victorious. This led the Aetolians to postpone their offensive, enabling Philip of Macedon to come to the rescue of the Acarnanians.64 The Roman system of recruitment must, then, be interpreted against this background. It is generally agreed that under normal circumstances only Roman citizens aged between 17 and 45 were expected to serve in the legions; older men either performed no military duties at all, or were used as a home guard to defend the city.65 Since most of the abundant 59 61 63

64

Thuc. 2.13.6–7. Cf. Van Wees (2004), 241. 60 Diod. 18.10.2. Cf. Van Wees (2004), 242. Plb. 4.22.8. 62 E.g. Singor (2002). Cf. also Plb. 3.86.11, according to which Hannibal had all men of military age (hoi en tais hêlikiais) killed as he marched through Umbria after the battle of Lake Trasimene. Although it is theoretically possible that men aged between 46 and 60 were spared, this seems unlikely. Livy 26.25. 65 Mommsen (1887), ii.1, 409, n. 5; Brunt (1971/1987), 21.

58

Polybius’ manpower figures

literary evidence relating to military matters refers to military campaigns, it comes as no surprise that the Greek expression hoi en tais hêlikiais is often used to denote Roman men aged between 17 and 45. A well-known example of this usage may be found in Polybius’ description of the Roman levy, in which the expression denotes the iuniores required to present themselves for recruitment.66 In passages such as this the phrase hoi en tais hêlikiais is synonymous with the expression hoi tên strateusimon hêlikian echontes (‘those of the right age to participate in military campaigns’). As in the case of classical and Hellenistic Greece, however, it would be completely wrong to infer from this that men aged between 46 and 60 were never regarded as ‘of military age’.67 As so often, Mommsen summarizes the basic rules with exemplary clarity. According to him, ‘the completion of one’s forty-sixth year brought exemption from service in the field (Felddienst), that of one’s sixtieth year from any form of military service’.68 This rule was still in operation in 49 bc, when Atticus ‘took advantage of the exemption based on age and did not stir from the city, as he was about sixty years old’.69 Interestingly, Livy reports that the consuls of 171 bc were authorized to recruit both veteran centurions and soldiers aged up to 50 for the Third Macedonian War.70 Since the Third Macedonian War was clearly not a matter of life or death for the Republic, we would a fortiori expect Roman seniores to have been called up in the event of any more serious threat. Livy tells us, indeed, that the highly threatening circumstances of 296 bc, when Rome faced simultaneous attacks by the Samnites, the Etruscans, the Umbrians and the Gauls, prompted the Senate to close the courts and to order the conscription of ‘men of every category’. As a result, not only were freeborn iuniores conscripted, but contingents of seniores were formed.71 A similar story is to be found in Appian’s account of the Hannibalic War. According to the passage in question, when Hannibal had destroyed two 66 67

68 69

70 71

Plb. 6.19.5. Cf. Cornell (1995, 183): ‘it is likely that the first census made no distinction between seniores and iuniores, but simply counted all men of military age’. The creation of these separate classes would thus have been a later development (ibid. 186). Mommsen (1887), ii.1, 409 and iii.1, 242. Cf. Nicolet (1988), 97. Nepos, Atticus 7.1: cum haberet annos circiter sexaginta usus est aetatis vacatione neque se quoquam movit ex urbe. Cf. Lammert (1948), 2029. Livy 41.33; 42.31. Livy 10.21.3–4. Cf. Nicolet (1988), 93–4. See also Livy 6.9.5, in which Camillus comes to the rescue of Sutrium with the legiones urbanae (contingents made up of seniores and of men invalided out of the army, originally entrusted with the defence of the city; cf. 6.6.14).

Some weaknesses of existing interpretations

59

Roman legions at Trasimene, along with a detachment of 8,000 men blocking the way to Rome, the Romans feared that his next move would be to march upon their city; they therefore ‘collected stones upon the walls and armed the old men (tous gerontas)’.72 An explicit formulation of the principle that men aged between 46 and 60 could be called up for active service in an emergency may be found in the second book of Appian’s Bella Civilia, which contains the following summary of Caesar’s exploits in Gaul: He fought thirty pitched battles in Gaul alone, where he conquered forty nations so formidable to the Romans previously that in the law which exempted priests and old men (hiereôn kai gerontôn) from military enrolment a formal exception was made ‘in the event of a Gallic incursion’; for then both priests and old men were required to serve.73

Since this was exactly the event that threatened Rome in 225 bc, this passage has a direct bearing upon our interpretation of the Polybian manpower figures.74 Most writers on this subject are agreed that the Greek expression hoi en hêbêi (‘the adults’) was generally used to denote men aged up to 45, but could sometimes mean men aged under 60.75 Meanwhile, the expression hoi en tais hêlikiais is considered to be an entirely unambiguous term exclusively denoting men aged up to 45.76 As we have seen, this assumption is demonstrably incorrect, for the simple reason that under certain circumstances older citizens (whether in the Greek or in the Roman world) were considered to be of military age.77 The only possible conclusion is that while the phrase hoi en tais hêlikiais normally referred to the younger citizens, it could also include fighting men aged between 45 and 60. As we have seen, the expression seems to have this wider meaning in Polybius’ account of the Macedonian invasion of Laconia in 220 bc. In interpreting his description of the military preparations made on the eve of the Gallic invasion of 225 bc, we must remember that this part of his 72 74

75 76

77

App. Hann. 11. 73 App. BC 2.150. Lo Cascio (2001b), 586, accepts that the manpower figures would have included proletarians, since these men could be mobilized in an emergency, but fails to note that seniores must also have been included, for the same reason. E.g. in DH 5.75.3–4. Cf. Bourne (1952), 130; Brunt (1971/1987), 21. E.g. Ilari (1974), 64 and 70; Hin (2008), 191–3. One of the few ancient historians who have stressed the ambiguity of the expression hoi en tais hêlikiais is Toynbee (1965), 458, n. 3 and 480, n. 4. Since accounts exist of seniores going out on military campaigns during the early Republic (above, note 71), even the expression hoi tên strateusimon hêlikian echontes (‘those of the right age to serve in military campaigns’) is not completely unambiguous, although it normally denotes iuniores. See e.g. DH 11.63.2, where this expression is used in a context in which we would normally expect hoi en hêbêi.

60

Polybius’ manpower figures

account was based on a passage from Fabius Pictor. We must therefore reckon with the possibility that Fabius used the expression hoi en tais hêlikiais to denote all ‘men of military age’, including the seniores, and that Polybius copied this phrase. Alternatively, it is possible that Fabius used another phrase (such as hoi en hêbêi) and that Polybius replaced this with his standard term for ‘those of military age’. To round off my argument, I want to take a brief look at the expression ‘those able to bear arms’ (tôn dynamenôn hopla bastazein), used by Polybius to describe the total manpower reserves of the Romans and their allies. It has, of course, long been recognized that this Greek phrase translates the Latin expression qui arma ferre possent, ‘(those) who are able to bear arms’. The technical meaning of the latter phrase is to some extent illuminated by Livy’s statement that Fabius Pictor interpreted the no doubt fictitious census figure for the reign of King Servius Tullius as referring to the number of ‘those able to bear arms’.78 Livy gives this piece of information in the context of his description of the system of census classes allegedly introduced by Servius. From his account, it appears that the first three classes contained 120 centuriae of foot soldiers. Of these, 60 were made up of iuniores and 60 of seniores. We are also told that all members of the first three classes, whether iuniores or seniores, were equipped with helmets, shields, spears and javelins.79 In conformity with his descriptions of certain early-republican levies, Livy states that the seniores were used for civil defence and the iuniores for service in the field. Taken in conjunction, these indications suggest that, unlike some modern scholars, Livy regarded ‘home defence’ as a genuine form of military service.80 78

79

80

Livy 1.44.1. As Rathbone has noted (1993a, 124, n. 2 and 136, n. 11), it is not entirely clear whether Fabius was the common source for Livy’s and Dionysius’ accounts of the Servian system of classes. This uncertainty does not, however, change the fact that Livy’s reference to the results of the first census is clearly linked to his account of the Servian system. Rathbone (1993a, 11) and Lo Cascio (2001b), 569–70, interpret Livy’s appeal to Fabius’ authority as suggesting that other authors had proposed a different reading of the regal and early-republican census figures. If we accept this theory, it is tempting to interpret Livy as alluding to the alternative theory that the early census figures represented the number of adult men sui iuris who had registered themselves, their property and their children, a tradition which has left some traces in the surviving sources (e.g. DH 9.36.3). For a possible explanation of why some of Livy’s predecessors preferred to interpret the early census figures in this way, see Chapter 3, note 34. Contra Hin (2008), 206, the emphatic reference to Fabius need not imply that Livy wanted to draw attention to differences that he believed to have existed between the regal census and that of his own time. Livy 1.42. According to Livy, only the pedites of the first class were equipped with breastplates (loricae) and round shields (cf. Plb. 6.19.3). Those of the second and third classes carried long shields; foot soldiers of the third class did not wear greaves. Cf. Livy 5.10.5, where the custodia urbis with which the seniores were charged is explicitly called a form of opera militaris.

Some weaknesses of existing interpretations

61

Another instructive passage is Livy 3.4.9–10. Here we are told that after one of the consuls had suffered a crushing defeat against the Aequians in 464 bc the Senate declared a state of emergency by ordering Postumius, the other consul, ‘to ensure that the state would not be harmed’.81 It was then decided that Postumius should remain in Rome to enrol ‘every man capable of bearing arms’ (omnes, qui arma ferre possent). The most obvious interpretation of this is that Livy was describing a tumultus during which all adult men, including those aged between 46 and 60, were mobilized.82 On the basis of this handful of texts alone, it is difficult to avoid the conclusion that, at least to Livy, the expression qui arma ferre possent meant all of the fighting men included in the various census classes, including the seniores.83 Confirmation of this interpretation is provided by certain passages from Dionysius of Halicarnassus that are rarely referred to in this context. One of the clearest examples is to be found in his account of the events of 487 bc. According to Dionysius, this year witnessed two simultaneous wars: one against the Hernicians, the other against the Volscians. In order to cope with this double threat, the Roman field army, consisting of the younger men (neôteroi), was divided into three parts, one of which was entrusted with the task of defending the districts surrounding the city. At the same time, ‘those too old to appear on the list of legionaries’ (hyper ton stratiôtikon katalogon) but ‘still capable of bearing arms’ (hosoi dynamin eichon eti bastazein hopla) were organized under their standards and ordered to guard the city’s citadels and walls.84 In a very similar account, we are told that ‘those who were too old to appear on the list of legionaries but still had sufficient strength to bear arms’ (dynamis hoplôn chrêseôs) were left to guard the citadels and walls during the campaign of 480 bc.85 Here Dionysius is undoubtedly following some annalistic predecessor who clearly believed

81

82

83

84

85

The reference to a senatus consultum ultimum is, of course, completely anachronistic. Cf. Mommsen (1887), I, 689, n. 1. As noted by Mommsen (1887), iii.1, 241–2, followed by Liebenam (1905), 603, all vacationes, including those of the seniores, could be suspended during a tumultus. Thus, correctly, Mommsen (1887), ii.1, 408–9 (‘alle dienstberechtigten Bürger’). Cf. also Livy 5.39.13, where the senes triumphales consularesque (aristocrats older than 60) are said to have bodies quibus non arma ferre, non tueri patriam possent. DH 8.64.3. Note that this passage demonstrates that the expression qui arma ferre possent does not always refer to the same group as the Greek phrase hoi tên strateusimon hêlikian echontes (‘those of the right age to participate in military campaigns’). The latter expression excludes the military duty of home defence and therefore specifically denotes the iuniores. Cf. DH 5.75, where hoi tên strateusimon hêlikian echontes are a sub-group of ‘the adults’ (hoi en hêbêi). DH 9.5.3.

62

Polybius’ manpower figures

that a significant number of those not normally liable for service in the legions (that is, the seniores) were fully ‘capable of bearing arms’.86 As we have seen, the only textual evidence cited by Lo Cascio to buttress his narrow interpretation of the expression qui arma ferre possent is Caesar’s reference to the somewhat mysterious tabulae discovered in the camp of the Helvetii.87 Since these tabulae listed ‘those able to bear arms’ separately from women, children and old men (senes), this passage would seem to prove that the Roman expression qui arma ferre possent did not include adult men aged between 46 and 60. One obvious weakness of this interpretation is that nothing is known about the age limits for active military service observed by this Celtic tribe. The only inference that can safely be drawn is therefore that the Helvetii, like the Romans, had an upper age limit for active military service. Although it is entirely possible that this age limit was 45, it may equally well have been considerably higher. More importantly, Lo Cascio seems to overlook the fact that while in Latin ‘seniores’ is a technical term for men aged over 45, the connotation of the term senex is much vaguer. As many scholars have pointed out, the latter term is often used to refer to various categories of ‘elderly’ men, the threshold for ‘old age’ varying between 45 and 60. Within this range men younger than 50 are very unusual, while it is possible to detect a majority view that the age of 60 marked the beginning of senectus.88 When read in this light, the passage from the De Bello Gallico actually supports the view that men aged up to 60 were considered to be ‘capable of bearing arms’.89 86

87

88

89

Note that these passages militate against Gabba’s theory (1976, 8, and 176 n. 54) that the formula qui arma ferre possent refers to the economic capacity of the assidui to provide their own arms. In my view, the age limit of 60 was an essential component of the Roman concept of ‘being able to carry arms’. On the other hand, Gabba (1949/1976, 8) is right to state that this formula does not refer directly to the physical strength of those liable for legionary service. As he points out, physical fitness was scrutinized not by the censors, but at the time of the levy. Hin’s argument (2008, 195–7; 2009, 169– 70) that a census figure disclosing the total number of adult males would not have revealed the military potential of the Republic therefore has no force. Invalids would have been given a permanent vacatio, relieving them of the obligation to attend the levy. According to Schulz (1937), 168, n. 2, followed by Hin (2008), 192 and (2009), 166, the apparently close correspondence between the social and military structures of the Helvetii and those of the Romans, along with Caesar’s use of technical Roman terminology, feeds the suspicion that Caesar is projecting Roman habits onto a barbarian tribe. That does not, however, diminish the importance of this passage in understanding the terminological problems discussed here. See the illuminating discussion in Parkin (2003), 16–26. According to Varro ap. Censorinus, De Die Natali 14.2, the term seniores referred exclusively to men aged between 45 and 60 and the term senes to men older than 60. In late antiquity, Augustine (PL 34.566) was to explain, ‘seniorum aetas minor est quam senum, quamvis et senes appellentur seniores’. Exactly this conclusion was drawn by Schulz (1937), 168, n. 2. The Acarnanians’ decision to mobilize all men aged between 15 and 60, while evacuating those over 60 (above, at note 64), is a perfect parallel.

Some weaknesses of existing interpretations

63

If these linguistic indications are combined with Polybius’ description of the specific background to the preparations undertaken in 225 bc, it will appear that, at least in this particular passage, the expression hoi en tais hêlikiais is synonymous with the phrase hoi dynamenoi hopla bastazein, and that both expressions are to be understood as referring to all men aged between 17 and 60. 2.2.3 Some further implications If the conclusions of the foregoing sections are accepted, it is possible to shed new light on the nature of the formula togatorum, a list of allied manpower resources that was used by the central Roman government to apportion military burdens amongst the Latin and Italian allies.90 It has often been observed that Polybius’ account of the preparations of 225 bc implies that Rome had at that time no reliable data concerning the manpower resources of its allies. From this, some have inferred that no such thing as a formula togatorum existed before 225 bc and that it was precisely the information supplied by the allies on the eve of the Gallic invasion that enabled the Roman government to create such a list for the first time.91 Finding it difficult to believe that Rome can have known nothing of its allies’ resources before 225 bc, others have argued that the Roman government used the returns supplied by the allies to revise an existing list.92 What these rival explanations have in common is that they are based on the assumption that the Latin term togati and the Greek expression hoi en tais hêlikiais denote the same category of men. The nub of the question is, what does the term togati mean? As many scholars have observed, it seems unlikely that all of Rome’s Italian allies had adopted the toga at this early date. We are, in fact, explicitly told that some groups among the allies (for example, the Greeks of southern Italy) clung to their old dress codes for a very long time after they had been drawn into the Roman system of alliances.93 In order to circumvent this difficulty, it has been conjectured that the term togati originated as a designation used in relation to the Latin allies and was subsequently retained as a general term for any allied citizen when the Romans conquered all of peninsular Italy.94 As Lo Cascio’s brilliant article on the subject points out, however, there is a 90 92 93

94

Above, at note 35. 91 Ilari (1974), 79; cf. Lo Cascio (1991/1994), 312–13. E.g. De Sanctis (1907), 453; Brunt (1971/1987), 547. Mommsen (1887), iii.1, 673–7. Cf. App. Samn. 7; Suet. Aug. 98.3. Mommsen held that the Greeks of southern Italy were excluded from the formula togatorum. Against this theory, see Ilari (1974) 114. Salmon (1982), 169–70.

64

Polybius’ manpower figures

much simpler solution. In his view, the term togati must be understood against the background of the Roman system of according the toga virilis to young men when they reached the age of 17.95 In other words, the term togati can be interpreted as referring to an age group rather than to people who had adopted Roman clothing customs. In my view, this solution is both elegant and convincing. Unlike Lo Cascio, I cannot think of any reason why the age group denoted by the term togati should be narrowed down to those aged between 17 and 45.96 Since there is nothing in this term to suggest an upper age limit, it is surely natural to conclude that the formula togatorum began its life (quite possibly in 225 bc) as a survey of the numbers of adult men, including those aged over 45, the Latin and Italian allies could supply.97 This interpretation of the term togati is, of course, in complete accordance with my suggestion that, in the specific context of Polybius’ account of the preparations of 225 bc, the Greek expression hoi en tais hêlikiais should be understood as denoting all adult males. The specific circumstances of 225 bc also help us to resolve another problem that has been under discussion since the late nineteenth century. As Beloch notes, it is odd that Fabius/Polybius should have grouped the Sabines together with the Etruscans despite the fact that the former, unlike the latter, had been given citizenship in 290 bc. Various explanations have been offered. Beloch’s own solution is to assume that Fabius/Polybius might inadvertently have mentioned the Sabines together with the Etruscans because he was drawing on a tradition according to which a tumultus had been declared in both regions. If we accept this view, we must interpret the figure of 50,000+ as referring exclusively to the Etruscans.98 In other words, we must regard the information given by Fabius/Polybius as inaccurate. Another popular theory is that the ‘Sabines’ of this passage are to be identified not with the inhabitants of Sabinum, but with those of the tribus Sapinia, an area to the northwest of the Umbrian town of Sarsina.99 95 96

97

98

Lo Cascio (1991/1994), esp. 320–2. Contra Lo Cascio (1991/1994), 322–3. We read, it is true, of tabulae iuniorum being used to trace Roman citizens who had evaded their military duties (Livy 24.18.7) and of units of allied soldiers being called up pro numero cuiusque iuniorum (Livy 34.56.6; cf. Toynbee 1965, 430–1; Ilari 1974, 73–5). As Schulz (1937), 166, long ago pointed out, however, the centuriate assembly, as described by Livy 1.42–3, cannot have functioned without tabulae seniorum. Since seniores were called up only in exceptional circumstances, it would have made perfect sense to keep separate lists of iuniores and seniores. According to this view, the age groups covered by the formula togatorum were identical to those targeted by the Roman censors. Cf. Brunt (1971/1987), 48–9. 99 Toynbee (1965), 484–8, followed by Torelli (1987), 49.

Some weaknesses of existing interpretations

65

The main weakness of this ingenious theory is that the sources leave us in no doubt that the tribus Sapinia was, like Sarsina, in Umbrian territory.100 This makes it difficult to understand why Fabius/Polybius should have grouped the hypothetical Sapini with the Etruscans rather than with the Umbrians and Sarsinates. In my view, it is much better to accept Fabius/Polybius’ description of the events of 225 bc as broadly reliable. What Polybius tells us is that a tumultus was declared in the three frontline regions of Umbria, Etruria and Sabinum. Those Etruscans and Sabines who were then called up (in reality a relatively small force)101 were stationed in northern Etruria. If this reading is correct, Polybius grouped the Sabines with the Etruscans simply because this was the only part of the Ager Romanus in which a tumultus had been declared. Part of the explanation must be that the geographical location of Sabinum made it relatively easy to direct fighting men from this region to the area through which the Gauls were expected to launch their attack on central Italy. An additional consideration may have been that at the time of the Gallic attack the Sabines of Upper Sabinum were still cives sine suffragio102 and were therefore not yet included amongst the tribes covered by Rome’s central levy. As we shall see in a later chapter, this theory does not imply that the citizens of Upper Sabinum never served in the legions between 290 bc and the date of their full enfranchisement.103 It does, however, imply that these men could well have been called up as a separate group and directed to the northwestern front together with the Etruscans.

100 102

103

Livy 31.22. 101 According to Plb. 2.25.9, 6,000 Romans were killed near Faesulae. According to Afzelius (1942), 21–3, Taylor (1960), 60–5 and Toynbee (1965), I, 377–86, the Sabini who received the civitas sine suffragio in 290 bc and the ius suffragii ferendi in 268 bc (Vell. 1.14–15) should be identified as the people of Cures, whose territory is sometimes referred to as the ager Sabinus (e.g. Cic. Leg. Agr. 2.66). The best argument in favour of this theory is perhaps that the tribus Velina and Quirina were created only in 241 bc, making it difficult to see to which tribe the indigenous inhabitants of upper Sabinum would have belonged had they received the suffragium as early as 268 bc. Brunt (1969), 124, suggests that the Sabines of upper Sabinum might have been provisionally assigned to an unknown tribe (the Sergia?) in 268 bc and might then have been transferred to the Velina and Quirina in 241 bc. As Humbert (1978), 235, n. 110, points out, however, there are no parallels for entire peoples being first assigned to one tribe and then transferred to another. Brunt is right to point out that the viritane settlers sent out to Sabinum must have been transferred from their old tribes to the Velina and Quirina in 241 bc, but in this case an entire ethnic group was not being transferred en bloc from one tribe to another. In all likelihood, the inhabitants of upper Sabinum went through a stage of citizenship without the vote (Taylor 1960, 65; Toynbee 1965, 382), but they may not have acquired this status until some time after 290 bc. According to Velleius (1.14–15), Capua and a number of Samnites (perhaps the Sidicini) received the civitas sine suffragio in 334 bc, four years after the end of the Latin war. Chapter 3, at note 68.

Polybius’ manpower figures

66

If we accept the overall veracity of Polybius’ account, the figure for the Etruscans and Sabines should include an unspecified number of the latter. It is in my view significant that the total figure for these two groups is given as ‘more than 50,000’, especially since no such formulation is used of any other allied contingent. This suggests to me that Fabius/Polybius did not go to the trouble of working out a separate figure for the Sabines. Instead, he seems to have taken 50,000 as a round figure for the Etruscans, then simply given the total size of the combined tumultuary force after the addition of the Sabines as ‘more than 50,000’. A corollary of this reading is that, precisely because Fabius/Polybius did not go to the trouble of calculating a separate figure for the citizens of Upper Sabinum, he cannot have subtracted them from his aggregate figure for the Romans and Campanians. In other words, even if we assume that a tumultuary levy was held in Sabinum (as I think we should), the specific formulation used by Polybius to indicate the number of Etruscans and Sabines strongly suggests that the number of Sabines played no part in his overall calculations. We are now in a position to explore some further implications of my reading of the Polybian manpower figures. As we have seen, most of the existing literature sees no problem in charging Polybius with the gross mistake of counting more than 50,000 men twice. The main argument for this interpretation is that the total number of citizen soldiers implied by Polybius’ survey (325,300) is much higher than the census figure for 234 bc. Against this I have argued that this high figure can be accounted for by assuming that the survey of 225 bc was more successful than most of the censuses of the third century bc, and that Polybius was correct in adding the men already in the field to the numbers of men revealed by the katagraphai. If this reading is accepted, we must reinterpret not only the manpower figures for the Romans, but also those for their Latin and Italian allies. The relatively low figure of 85,000 given by Polybius for the Latins has long puzzled scholars. It has therefore frequently been asserted that this figure makes sense only if it is taken to refer to iuniores alone.104 If, however, Fabius/Polybius was correct in adding the troops already in the field to the number of men still left on the registers, some 13,300 Latins (20.8 per cent of 64,000) must already have been mobilized. If we then accept the general view that the figures for the allies were approximately 20 per cent defective (partly because they are likely to have been less complete than the Roman tally), we end up with a rough estimate of 117,960 Latins of military age, or 104

Above, at notes 8–9.

Some weaknesses of existing interpretations

67

c. 3,470 per Latin town.105 This comes close to the average of 3,800 colonists sent out to each of six Latin colonies founded before 225 bc.106 All of Polybius’ figures for the Latin and Italian allies’ infantry are multiples of 10,000, indicating that these are rough approximations that may well be seriously defective.107 It also seems reasonable to suppose that in most Italian communities those rich enough to serve as cavalry were registered more efficiently than the poorer sections of the adult male population. Taken together, these distorting factors may help to explain certain anomalies such as the odd ratio between cavalry and infantry (1:5) among the Abruzzi peoples.108 Only one figure is entirely impossible: the 16,000 cavalry to be supplied, supposedly, by the Iapygians and Messapians.109 The most probable explanation is that this figure is the result of a scribal error.110 The only other oddity is the absence of the Bruttians and the Greeks. According to some scholars, the Greeks do not appear because they

105

106

107 108

109

110

Unlike Afzelius (1942), 134, and Brunt (1971/1987), 56, I do not think that there are any grounds for classifying the Volscian towns of Fabrateria and Aquinum as Latin. After this adjustment, we are left with thirty-four Latin towns. Their number can be brought down to thirty-one by assuming that (despite Livy 34.42.5–6) the three Hernician towns of Aletrium, Ferentinum and Verulae were not technically Latin (Bispham 2007, 85, n. 58), and to only twenty-eight by assuming that the nomen Latinum consisted solely of the Latin colonies (e.g. Salmon 1982, 53–4; Bispham 2007, 29 and 74, n. 2, pointing to Livy 23.20.2). Note that our idea of the average populations of Latin colonies would probably have to be revised downwards if we had any data about small Latin towns such as Nepet, Sutrium and Narnia. As noted by Brunt (1971/1987), 56, the isolated colonies of Placentia and Cremona, to each of which 6,000 colonists were sent, are likely to have been exceptionally large. Cornell (1989), 405, reckons that between 334 bc and 263 bc 71,300 settlers were sent out to nineteen Latin colonies, giving an average figure of 3,753 per colony. Four Latin colonies founded between 193 bc and 181 bc had an average complement of 3,400 (Brunt, see above.). We must also keep in mind the demographic impact of the First Punic War. The census of 234/233 suggests that at that time the Roman citizen body had not yet fully recovered from the demographic setback caused by this war. The same may have been true of the Latin and Italian allies. Afzelius (1942), 100. It must also be remembered that the ratio between horse and foot in the allied units already in the field was 1:15. This means that a higher proportion of allied foot soldiers had already been called up. Although Toynbee (1965), 499, is right to point out that Apulia contained the excellent pastureland of the Tavoliere, a ratio of 1:3 for horse and foot seems too high even for a horse-breeding region. In favour of Toynbee’s view, see Lippolis (1997), 83, n. 20; Lo Cascio (1999a), 169; Forsythe (2005), 365. Against it, see Brunt (1971/1987), 49; Yntema (2008), 373. Since the figures for Roman and allied cavalry will no longer add up to the total of 70,000 given by Polybius (2.24.16) if we correct this error, it may well have originated with Fabius Pictor. Did he perhaps misread xmiliaequitum as xvimiliaequitum? Of course, another possibility is that a numerical error had crept into the Fabian manuscript used by Polybius. Had Fabius reckoned with 6,000 Apulian cavalry, but included 2,000 Umbrian horsemen and 6,000 or 7,000 cavalry of the Veneti and Cenomani (cf. Plb. 1.24.1 for the high ratio of horse to foot in the army of the invading Gauls), he would have arrived at the same round figure for the number of Roman and allied horse as Polybius. Needless to say, all attempts to account for the scribal error that is probably responsible for this problematic figure must remain purely speculative.

68

Polybius’ manpower figures

were expected only to supply ships and crews, rather than foot soldiers.111 Against this, it has been pointed out that Greeks appear to have fought at Cannae, and that the obligations to provide ships’ crews and foot soldiers cannot have been mutually exclusive.112 It is possible, of course, that Fabius/Polybius simply overlooked the men supplied by the southernmost allies,113 or that no reliable records concerning their numbers were available. Building on the observation that all of the preparations of 225 bc can be understood as having been made in the context of a tumultus Gallicus, it is also possible to conjecture that it made sense to approach only those communities that could reasonably be expected to feel threatened. In principle, this would have excluded the Apulians, the Greeks and the Bruttians. Apulia, however, had been penetrated by Gaulish bands in 367 bc and in 366/5 bc, a fact that could have been seen as rendering it appropriate to request the Apulians’ manpower figures.114 Yet another possibility is that the Roman government never considered the option of mobilizing large numbers of Greeks and Bruttians for an expedition in northern Italy because it had to reckon with the possibility of a Carthaginian attack on Sicily and southern Italy. The Bruttians and the Greeks would have been ideally placed to support the Roman legions against any such attack. Some Bruttian and Greek contingents may, indeed, already have been serving in the two small armies that were stationed in Sicily and Tarentum at the time of the Gallic invasion. Which, if any, of these explanations is correct must remain a matter of speculation. The only safe conclusion is that Fabius/Polybius gave no figures for the Bruttians or the Greeks, presumably because he did not have any, and that several tens of thousands of fighting men belonging to these groups must be added to the manpower reserves upon which the Romans would have been able to draw in the event of a Gallic invasion. Following Brunt and Lo Cascio, I have reckoned with some 65,000 adult male Bruttians and Greeks aged between 17 and 60.115 The foregoing interpretation of the Polybian figures implies the following estimates of the manpower reserves of the Latin and Italian allies:

111 112 113 115

Thus Mommsen (1887), iii, 673–7. E.g. Ilari (1974), 113–14, referring to Livy 23.1 and 24.13 and to Sil. Ital. Pun. 8.534. This is the solution preferred by Ilari (1974), 114. 114 Livy 6.42.7–8; 7.1.3. Note that Toynbee (1965), I, 494–5, assigned only 40,000 adult men under 46 to the Bruttians and Greeks.

Some weaknesses of existing interpretations

69

Table 2.2 Allied manpower in 225 bc

Latins Samnites Apulians Lucanians Abruzzi peoples Umbrians Etruscans Greeks/Bruttians Total

Still available

+ 20 %

Already serving

Total

85,000 77,000 56,000 33,000 24,000 22,000 54,000 (48,000) 399,000

102,000 92,400 67,200 39,600 28,800 26,400 64,800 (57,600) 478,800

13,635 12,350 8,980 5,295 3,850 3,530 8,660 (7,700) 64,000

115,635 104,750 76,180 44,895 32,650 29,930 73,460 (65,300) 542,800

if this reconstruction is accepted, it is possible to draw some interesting conclusions about Rome’s method of distributing military obligations among the Italian communities that made up the Roman-Italian system of alliances during the second half of the third century bc. As we have seen, Scheidel has suggested that Polybius’ manpower figures may have been constructed from the top down, on the basis of the assumption that the ratio between allies and legionaries approximated parity. The foregoing re-analysis of the Polybian data suggests, however, that the ratio between legionaries and allies was actually approximately 1:1.6.116 In my view, it can be no coincidence that according to Polybius’ account of the preparations of 225 bc each Roman legion of 5,500 men was accompanied by an allied force of 8,000, implying a ratio of 1:1.45.117 The most natural explanation for this is surely that the military obligations imposed on the members of the Roman alliance system mirrored underlying demographic realities.118 Another way of assessing the plausibility of the foregoing reconstruction is to examine variations in regional population density. If the Ager Romanus indeed contained roughly 340,000 iuniores and seniores of citizen status, its average number of free inhabitants per square kilometre was 43.8.119 Two 116 117

118

119

542,800:340,000 = 1.596. Plb. 2.24.3–4. According to Plb. 2.24.9, the reserve army near Rome consisted of 21,500 citizens and 32,000 allies. These figures imply a ratio of 1:1.49. In theory it now becomes possible to argue that Fabius Pictor constructed the manpower figures for 225 bc using a ratio of approximately 1:1.6, but this alternative hypothesis seems unnecessarily contrived. Note that the census figures for 115/114 bc and 70/69 bc, taken together, point to a ratio between citizens and allies of roughly 1:1.25. Since in the aftermath of the Second Punic War Rome had confiscated large amounts of allied territory, a significant amount of which was now inhabited by Roman colonists and by viritane settlers, this later ratio is in perfect harmony with my reading of the Polybian manpower figures. This figure uses Beloch’s estimate that the pre-Hannibalic Ager Romanus covered 25,615 km2.

70

Polybius’ manpower figures

other areas for which high densities can be calculated are the territories of the Latins and those of the Samnites, where we find 35.9 free people per km2 and 33.5 free people per km2 respectively. By contrast, much lower densities (between twelve and fifteen free inhabitants per square kilometre) are implied by the figures for the Etruscans, the Umbrians, the Abruzzi peoples, the Apulians and the Lucanians. In my view, these discrepancies can easily be accounted for.120 The high population density of the Ager Romanus, for instance, is likely to reflect the social and economic makeup of the Roman citizen body, a very large proportion of which consisted of farmers who owned small or mediumsized farms. This social and economic group had been greatly reinforced by the assignationes viritanae of the fourth and third centuries bc. The high figure for the territories inhabited by Roman citizens also, of course, reflects the size of the city of Rome and the demographical importance of northern Campania, whose population densities are likely to have been much higher than those in most other parts of the peninsula.121 The high densities in the Latin territories again reflect a pattern of widespread land ownership, and we must remember that most Latin colonies were founded in areas containing a disproportionate amount of good arable land. In the case of the ‘Samnites’, it must be borne in mind that these included the inhabitants of southern Campania, where crop yields approximated those of the Ager Campanus. Arable cultivation was also the mainstay of the economies of Etruria and large parts of Apulia. Viewed in this light, the population densities implied by the Polybian manpower figures for these areas appear unexpectedly low. Part of the explanation must be that the average densities for Etruria and Apulia as a whole conceal large intra-regional discrepancies, with southern Etruria and southern Apulia characterized by much higher densities than the northern districts.122 Polybius’ figure for the Apulians’ manpower resources is in any case compatible with the archaeological data that we currently have.123 In the case of Etruria, we must also reckon with the probability that a significant proportion of the rural population consisted of ‘serfs’, most of whom may well not have been included amongst the c. 50,000 able-bodied ‘Etruscans’ listed 120 121

122

123

Scheidel (2004) notes these variations, but makes no attempt to explain them. The exceptional fertility of the Campanian soil is commented on by many ancient authors. On the other hand, the density of between 213 and 320 inhabitants calculated for the pre-Hannibalic Ager Campanus by Savino (1997), 186, on the basis of Livy 23.5.15, is almost certainly too high. Cf. Chapter 4, at notes 14–16. Note that the towns of South Etruria and Messapia were far closer together than those of the northern districts of Etruria and Apulia. Yntema (2008). I am not, of course, suggesting that the archaeological evidence proves the Polybian figure to be correct. For my purposes, it is enough that the archaeological data do not contradict this figure.

The population of Italy in 225 bc

71

by the Etruscan towns in 225 bc.124 If this assumption is correct, these ‘serfs’ should be added to my hypothetical tally for 225 bc. Unfortunately, their number cannot be estimated. Finally, population densities of between twelve and fifteen persons per square kilometre seem plausible for the societies and economies of the Abruzzi peoples, and also for those of Umbria and Luciania. Although there can be no doubt that arable farming played an important part in the economies of these areas, it remains likely that these were regions in which various forms of pastoralism were important, especially in the mountainous districts, and that average population densities were far lower here than in central-western Italy.125 2.4 the population of italy in 225 bc On the basis of my re-interpretation(s) of the Polybian manpower figures, it is possible to construct a tentative estimate of the size of the Italian population in 225 bc. If my reading of these figures is correct, there were about 340,000 adult male citizens and approximately 540,000 adult men of allied status in peninsular Italy on the eve of the Gallic invasion. These figures imply a total free population of about 2.9 million. To this basic figure we must add the population of Cisalpine Gaul and an unknown number of slaves. As we have seen, the only ‘evidence’ that we have concerning Cisalpine Gaul consists of some scattered data concerning the sizes of armies fielded by various Cisalpine tribes between 225 bc and 222 bc. Starting from Polybius’ statement that the Insubrians were able to field 50,000 men in 223 bc, Brunt argues that the Celtic tribes of northern Italy plus the Veneti and the Ligurians must have been able to field a theoretical total of 300,000 iuniores.126 The total population implied by this estimate is 1 million. There is no need to point out the unsatisfactory character of this numerical exercise. Perhaps the only way of testing the outcome of Brunt’s calculations is to look at the average population density that it implies. If we consider only 123

124

125 126

Yntema (2008). I am not, of course, suggesting that the archaeological evidence proves the Polybian figure to be correct. For my purposes, it is enough that the archaeological data do not contradict this figure. According to Zonaras 8.7 (= Cassius Dio 10.42), the Etruscans of Volsinii used their ‘servants’ (oiketai) to fight their wars. It seems likely, however, that most Etruscan contingents consisted of freeborn men. Cf. Dench (1995), 111–53. Above, at note 17. It may be noted that Brunt’s assumption that the figure of 50,000 for the Insubrian forces refers solely to iuniores is of questionable validity. If, as Brunt suggests, the Insubrians really put ‘every available man’ into the field, this figure would surely have included both iuniores and seniores.

72

Polybius’ manpower figures

those parts of northern Italy whose inhabitants held citizenship after 49 bc, Cisalpine Gaul covered an area of approximately 100,000 km2.127 The average population density implied by Brunt’s figures is therefore c. 10 persons per square kilometre. This is lower, but not much lower, than the density of 13.7 persons per square kilometre that can be calculated on the basis of Polybius’ figure for the Umbrians. Since in any view Cisalpine Gaul was at least somewhat less developed than most parts of peninsular Italy (including Umbria), this outcome seems acceptable.128 Although this is not to deny that our scanty data can easily be massaged in such a way as to produce a fairly wide range of equally ‘acceptable’ outcomes, it seems at least possible to conclude that there is nothing to contradict the view that the free population of mainland Italy numbered about 3.9 million on the eve of the Hannibalic War. If we add 300,000 hypothetical slaves, we arrive at a grand total of 4.2 million free and non-free inhabitants.129 Since at least some of the underlying figures are clearly unreliable, this estimate comes with a considerable margin of error. Despite their obvious limitations, however, the Polybian manpower figures would seem to narrow down the range of possible population estimates for central and southern Italy to manageable proportions. This very approximate total should, therefore, at least be of the correct order of magnitude. 2.5 the polybian manpower figures and roman mobilization rates In an important article published in 2001, Lo Cascio argues that the low count implies military participation rates so high as to be out of line with comparative data.130 His arguments can be summarized as follows. If we follow Brunt in assuming that there were about 300,000 adult male Roman citizens at the outbreak of the Second Punic War,131 the number of iuniores must have been about 216,000. Of these iuniores, about 50,000 had been mobilized at the time of the battle of Trasimene. In other words, in 217 bc the ratio between legionaries and iuniores stood at 1:4.32. In the years that followed, especially after Cannae, the number of legionaries increased 127

128

129 130

120,000 km2, less the c. 20,000 or 25,000 km2 inhabited by attributi in 28 bc (Beloch 1886, 321; cf. Chapter 1, at note 88). Kron (2005a) argues that many parts of Cisalpine Gaul had been brought under cultivation by the time of the first Augustan census. He does not, however, deny the presence of extensive swamps and forests in the final decades of the third century bc. For the hypothetical tally of 300,000 slaves in pre-Hannibalic Italy, cf. Scheidel (2005a), 76. Lo Cascio (2001a). 131 Brunt (1971/1987), 62, actually uses a figure of 325,000.

Manpower figures and mobilization rates

73

until a maximum of 80,000 was reached in 212 bc. In this calculation, it must be remembered that by that time the heavy casualties suffered between 218 bc and 214 bc had reduced the number of iuniores to c. 126,000. The low count therefore implies that by 212 bc c. 75 per cent of all iuniores were on active service. Lo Cascio also argues that the 126,000 iuniores of 212 bc corresponded to a total of 190,000 adult male citizens, implying a Roman citizen population of 632,700. With 80,000 legionaries in the field, this would imply a mobilization rate of 12.6 per cent.132 Lo Cascio goes on to compare this rate with the ratios between men under arms and total population that can be calculated for various parts of early-modern Europe. Here a rate of 7.7 per cent was achieved by Sweden in 1709, while in 1760 7.1 per cent of Prussia’s population served in the armies of Frederick the Great. These were high percentages; in most other countries military participation rates were much lower, with rates of under 2 per cent being the norm in Spain, France, Great Britain and Russia.133 Lo Cascio interprets these data as indicating that the population levels postulated by the low counters cannot possibly be correct. Although this argument may at first sight appear convincing, a closer inspection reveals serious weaknesses. The most important of these is perhaps that Lo Cascio’s estimates of the total Roman population are misleadingly low. His figures for 212 bc, for instance, are based on the improbable assumption that between 218 bc and 213 bc Rome had lost not merely 35 per cent of its iuniores, but 35 per cent of its entire population. Brunt’s version of the low count, in fact, assumes that between 218 bc and 212 bc the free Roman population was reduced from about 1 million to c. 840,000, by the loss of some 50,000 men on the battlefield and as a result of the defection of the Campanians, whose total population may have numbered c. 110,000.134 If we consider the free citizen population alone,

132

133

134

If we add an estimated 15,000 rowers of citizen status (Lo Cascio 2001a, 135), this percentage rises to 15 per cent. Lo Cascio (2001a), 137, Table 2.1. According to Gat (2006), 503, a mobilization rate of 3 per cent was achieved in revolutionary France in 1794, but this high rate could not be sustained for long. In 1813 about 2 per cent of the French population was on active service. For casualties, see Brunt (1971/1987), 419, n. 4. His estimate of 50,000 refers to the period 218 bc–215 bc. As he points out elsewhere (1971/1987, 64 and 422), however, the ranks of the iuniores were continually replenished with large cohorts of boys as they reached the minimum age for active service, and there were few military disasters after Cannae. Brunt (1971/1987), 18–19 and 64, believes there to have been 34,000 adult male Campanians before Cannae (cf. Livy 23.5.15). As he notes, however (1971/1987, 51), Strabo (6.3.4) asserts that Tarentum was able to field exactly this number of foot and horse. Brunt also observes that Livy’s figure of 30,000 Tarentines captured and sold into slavery in 209 bc (Livy 27.16.7) may well have been purely conventional. The same may be true of the figure for the Campanians. If we combine the probable extent of the Ager Campanus with the (high)

74

Polybius’ manpower figures

the mobilization rate for 212 bc would thus appear to have been 11.3 per cent.135 While so high a rate seems acceptable for the most critical period of a war in which Rome was fighting for its survival, the real mobilization rate must have been significantly lower. As we have seen, Polybius’ manpower figures suggest that there were as many as 340,000 adult male citizens on the eve of the Second Punic War, implying a free Roman population of c. 1,122,000. To this figure must be added the slave population. If we assume, conservatively, that 10 per cent of people were slaves, the total population rises to 1,234,200. With about 95,000 adult male citizens and slaves serving as legionaries or as rowers in the fleet, the mobilization rate for 212 bc works out as 7.7 per cent. Although this rate is certainly high by comparative standards, it is not much higher than the 7.1 per cent calculated for Prussia in 1760. If we ignore the servile population, the rate of military participation at the height of the Second Punic War rises to 8.5 per cent. Even this rate, however, is only marginally higher than the top rate of military participation achieved by the Confederacy during the American Civil War. Here (if we exclude slaves) 8 per cent of the population saw active service at the peak of the war.136 In assessing the wider significance of these estimates, we must bear in mind the highly specific character of the network of alliances by means of which Rome dominated Italy during the middle republic. It has often been observed that during their conquest of Italy the Romans imposed no direct taxes on any Italian community. Instead they created a complicated system of juridical statutes and a network of military alliances that gave them access to an ever-growing pool of military manpower. In developing and extending this network, both Rome itself and most of its Italian allies successfully maintained the high levels of military participation that had characterized the city-states of archaic and early-republican times. As Scheidel notes, all of this means that republican Rome cannot directly be compared with the emerging nation-states of early-modern Europe, where taxation primarily

135 136

population densities that can be calculated for the late sixteenth century, we end up with significantly fewer than 110,000 Campanians. Cf. Chapter 4, at note 16. For a general discussion of conventional figures in the ancient sources, see Scheidel (1996b). With 80,000 citizens in the legions and 15,000 in the fleet. Gat (2006), 527. If slaves are included, the highest rate achieved by the Confederacy drops to 5 per cent. In comparing this to the Roman rate, it must be borne in mind that my hypothetical figure for the slave population of the pre-Hannibalic Ager Romanus is almost certainly too low. Note that the mobilization rate of 11 per cent given by Scheidel (2008), 39, as the highest level achieved by the Confederacy is based on a misinterpretation of the estimate of 850,000 to 900,000 southern soldiers given by McPherson (1988), 306, n. 41. Only half of these men were on active service at any given time.

Manpower figures and mobilization rates

75

meant the imposition of demands for material resources.137 For this reason alone, it is difficult to maintain that the mobilization rates that can be calculated for the Hannibalic War are ‘impossibly’ high. At the same time, we must pay close attention to economic and social structures. Literature on pre-modern and modern warfare generally explains the unusually high mobilization rate (8 per cent of the free population) achieved by the southern states during the American Civil War as reflecting the fact that the South had a relatively unsophisticated economy in which slaves made up a considerable proportion of the labour force.138 There can be no doubt that this was also the case in Roman-republican Italy. In the absence of reliable figures, it is extremely difficult to come up with even a very approximate estimate of the number of urban and rural slaves in pre-Hannibalic Italy. It is nonetheless generally agreed that slaves were widely used in Roman and Italian agriculture from the late fourth century bc.139 What is particularly interesting is that the little evidence we do have suggests that slave ownership was very widespread at the time of the Hannibalic War. We are told, for instance, that when the Senate faced a shortage of oarsmen for a newly launched fleet of ships in 214 bc, senators were ordered to supply eight slaves each, while those citizens whose property had been rated at 1 million asses were expected to produce seven, those worth 300,000 asses five, those worth 100,000 three and those worth 50,000 one.140 As Rosenstein has recently observed, the census threshold of 50,000 asses is that of the third class of the centuriate assembly. The measures of 214 bc therefore imply the existence of a very substantial group of moderately well-off farmers, all of whom were assumed to own at least one slave.141 From this it can be inferred that, even in times of heavy conscription, production would have carried on normally on the slave-staffed farms of the rich and also on many larger family farms.142 It would, of course, be absurd to maintain that the majority of the free farming population owned slaves at the time of the Hannibalic War. In analysing the effects of widespread conscription on the poorer section of the free rural population, it must, however, be remembered that the agrarian economy of republican Italy can be described as a ‘peasant economy’, characterized by a high degree of underemployment. This implies that military service at this time should be seen not primarily as a disruptive 137 141 142

Scheidel (2008), 39. 138 Gat (2006), 527. 139 E.g. Finley (1980). 140 Livy 24.11.7–8. Rosenstein (2008), 6–7. From a functional point of view, widespread slave ownership can even be seen as an economic adaptation that made possible high military participation rates. Up to a point, this is similar to the state of affairs in classical Athens, where widespread slave ownership facilitated political participation.

76

Polybius’ manpower figures

force, but as a withdrawal of surplus labour that would otherwise have remained unused.143 Another important point is that the economic feasibility of massive mobilization is determined in part by the makeup of rural households and by their size. Although little is known about the size of rural households during the Roman republic, there is some reason to think that, unlike most urban families, many rural ones were probably of the extended type, including the members of three generations and in some cases perhaps two co-resident families.144 As several scholars have pointed out, such extended families would have been in a good position to cope with the conscription or death of one or two young men.145 It could, in fact, be suggested that the burdens imposed by the Roman system of conscription might have been a major stimulus to the formation of extended households in the Italian countryside. The same point can be made in another way. In one of the footnotes to the first chapter of Conquerors and Slaves, Hopkins claims that the Romans cannot be compared with notoriously warlike tribes such as the Zulus or the ‘Red Indians’, whose rates of military participation were much higher than those of the Roman Republic.146 Although I share Hopkins’ view that these tribal societies are not directly relevant to the Romans, they illustrate a general principle that is often ignored. I mean by this simply that the economic and demographic feasibility of a given mobilization rate depends upon the economic structures of the society in question.147 In my view, there can be no doubt that mid-republican Italy’s economy was far less sophisticated than those of most parts of early-modern Europe. I believe that this basic economic fact made it much easier for the Romans to withdraw large amounts of manpower from the rural districts of republican Italy.148 This is another reason why any direct comparison between Roman and early-modern military participation rates is potentially misleading. It seems reasonable to conclude that my re-interpretation of the Polybian manpower figures implies military participation rates during the Hannibalic 143 144

145 148

E.g. Hopkins (1978), 24, followed by Erdkamp (1998), 264–5 and 267, and De Ligt (2007a), 121. As is so often the case, we have almost no evidence from republican or early-imperial Italy. Extended and multiple families were, however, very common in the rural districts of Roman Egypt and of fifteenth-century Tuscany; see Bagnall and Frier (1994), 60. One indication that many Romans of the republican era lived in households of this sort is the legal construct of the societas ercto non cito (also known as the consortium), an arrangement whereby children succeeding their father did not divide up their inheritance, but continued to enjoy it in common. See e.g. Jolowicz and Nicholas (1972), 296. Cf. Plut. Aem. 5.4–5, for the famous case of the sixteen Aelii Tuberones who lived together in a small farmhouse. E.g. De Ligt (2007a), 121. 146 Hopkins (1978), 11, n. 19. 147 Andreski (1954). This would be a worthwhile angle from which to re-examine the transition to a standing army that took place during the imperial period.

Conclusions

77

War that are entirely plausible in light of the highly specific structures of Italian society during the second half of the third century bc. If this is true of the Hannibalic War, the same conclusion necessarily obtains for the early decades of the second century bc, when military commitments were still high, but nonetheless significantly lower than those of the decisive years of the Second Punic War. 2.6 conclusions The most important finding of this chapter is perhaps that, although Polybius’ manpower figures were clearly intended as rough approximations, there are no good grounds to reject them as completely unreliable. At the same time, we have seen that the correct interpretation of these figures is bedevilled by a number of technical and terminological problems. Despite the uncertainties resulting from these technical difficulties, it seems possible to conclude that many earlier scholars have been misled by the fact that Polybius’ figure for those ‘Romans and Campanians’ who remained on the reserve list is almost identical to the census figure for 234 bc. Lo Cascio is undoubtedly right to insist that this similarity may well be fortuitous. On the other hand, Lo Cascio’s theory that Polybius’ survey reflects a regional pattern of mobilization that took no account of legal status runs up against insuperable difficulties. It seems more appropriate to interpret these figures as referring to the numbers of men available for mobilization in the event of a tumultus Gallicus. According to this interpretation, the figures should be taken to include both iuniores and seniores, not only in the case of the Romans and Campanians, but also in that of the allies. If we read the figures in this way, the weaknesses that vitiate the traditional low-count reading can be avoided. The size of the free Italian population of central and southern Italy implied by this new interpretation (c. 2.9 million) is nonetheless more or less identical to the estimates suggested by Brunt and Hopkins. With a hypothetical 1 million Gauls, Ligurians and Veneti in the north, this implies that mainland Italy as a whole had a free population of c. 3.9 million. If we reject Lo Cascio’s regional interpretation and opt for 3.9 million as the most reasonable estimate of the size of the free population of mainland Italy at the time of the Gallic invasion, we must posit an almost fourfold increase in the free Italian population to account for the c. 14.75 million free Italians implied by the high-count reading of the census figure for 28 bc.149 149

14.75:3.9 = 3.78. Cf. Chapter 1, at note 29.

78

Polybius’ manpower figures

In principle, the enormous gap between the figures for 225 bc and 28 bc can be closed by assuming that the number of free Italians increased rapidly over these years by means of a combination of natural and non-natural factors and that extensive enfranchisement of provincials during the last 50 years of the Republic created many hundreds of thousands of new citizens. We could assume, for instance, that the number of free Italians grew at an average annual rate of 0.6 per cent and that the number of ex-slaves, enfranchised foreigners and descendants of these groups had risen to approximately 2 million by 28 bc. Alternatively, we could assume an average natural growth rate of 0.5 per cent per annum and put the number of citizens created by means of legal devices at 4.2 million.150 As we shall see in a later chapter, however, the natural growth rates required by these scenarios seem much too high for a society that was not only engaged in quasicontinuous warfare abroad, but had to cope with the impact of a number of vicious wars between various groups of Italians.151 In this sense, the findings of this chapter can be seen as an argument against the high-count interpretation of the Augustan census figures. 150

151

Scheidel (1997), 166, n. 140, suggests that as many as 500,000 slaves may have been manumitted in Italy between 81 bc and 49 bc. As Brunt long ago pointed out (1971/1987), 143–6, however, there are strong indications that freedmen had fewer children than ingenui, partly because most female slaves were freed at an advanced age. There is in my view no good evidence for the massive enfranchisement of provincials posited by Crawford (2008). Cf. Chapter 1, note 35. Chapter 4, at notes 217–18.

chapter 3

Census procedures and the meaning of the republican and early-imperial census figures

3.1 introduction As long as the low-count theory remained dominant, with Polybius’ manpower figure for the Romans and Campanians generally regarded as representing the number of adult male citizens registered in the census of 229 bc, ancient historians could feel free to reconstruct the demographic history of the Roman citizen body on the basis of the census figures for the third to the first century bc. There were, of course, endless disputes about the exact meaning of these figures and (to a lesser extent) about the degree of accuracy achieved by the republican censors.1 Nonetheless, even those who held that the censors were only expected to register certain categories of citizen were at least inclined to accept that the census figures for the last 200 or 250 years of the Republic gave some rough idea of long-term changes in the number of Roman men of citizen status. Against this view, Lo Cascio has recently argued that the censuses of republican times are likely to have netted a relatively small proportion of their target population.2 If this theory could be proved to be correct, it would open up the possibility that the Roman citizen population (and thus the population of late-republican Italy as a whole) was far larger than previously thought. This is precisely what Lo Cascio and some other high counters have argued. If we take this sceptical attitude towards the reliability of the census figures seriously, as I think we should, we can no longer simply build our demographic conclusions upon these data. Even if the legitimacy of a purely inductive approach to the republican census figures has become uncertain, we can still try to assess the various assumptions that must be made in order to fit these figures into the general demographic frameworks implied by the low and high counts. While such an exercise cannot be expected to provide us with any incontestable 1

Cf. below.

2

Lo Cascio (2001b).

79

80

Meaning of the census figures

conclusions in favour of either theory, it should at least help us to see more clearly which basic assumptions must be taken on board by those tempted to espouse one or the other interpretation. The recent debate over the reliability of the republican census figures does not, therefore, in any way diminish the importance of investigating in detail the registration procedures followed by the censors and the relationship between the census figures and the size of the population targeted. In fact, a thorough understanding of the technical and terminological problems surrounding the republican census figures must be regarded as a vital prerequisite for an overall assessment of the respective merits of the low and high counts.3 3.1.1 A very short account of the republican census It must be emphasized that no coherent account of the republican census is to be found in any Latin or Greek text of the republican or imperial period. This means that the workings of the census during the third, second and first centuries bc must be reconstructed from scattered references in historical and antiquarian works, in various of Cicero’s writings and in a few surviving statutes, of which the late-republican Tabula Heracleensis is the most important. As Northwood has recently noted, a more or less comprehensive reconstruction can be achieved only if we assume that some practices of imperial date were continuations from an earlier stage and that certain reports concerning the early Republic are retrojections of later procedures. The obvious danger of this approach is that we end up with a static picture in which the republican census has no history.4 Despite these methodological difficulties, there is substantial agreement about the most basic features of the census procedures of the middle and late Republic. Every 5 years, all male citizens who were sui iuris were obliged to declare themselves and their property before the censors’ iuratores. We also read of citizens declaring their wives and their sons in potestate, and the ages of these sons, at least, seem to have been reported.5 Finally, the fragmentary evidence suggests that declarants were obliged to give a monetary valuation

3

4

Unlike the authors of some recent publications (cf. Chapter 1, at note 126), I do not think that the viability of the low and high counts can be assessed solely on the basis of Augustus’ use of the traditional phrase capita civium censa in the Res Gestae. Even Lo Cascio (1994), 31, concedes that the demographic debate cannot be settled by means of this philological argument. Northwood (2008), 257. 5 DH 5.75.3, with Northwood (2008), 258, n. 5.

Target population of the census

81

of all items of property declared and that these valuations were expected to reflect real market rates.6 3.1.2 The central problem: census procedures and registration rates While scholars such as Brunt and Hopkins assume that the republican censuses succeeded in registering between 75 per cent and 90 per cent of the target population (defined by them as all adult men of citizen status), the high counters argue in favour of a deficiency rate of 45 per cent or higher. In support of this claim, Lo Cascio has developed two arguments. His first contention focuses on the various groups of citizens targeted by the republican censors and challenges the low-count view that the censors of the last three centuries bc were expected to register all adult male citizens. Lo Cascio argues that there is good reason to believe that these magistrates were not expected to register either citizens without the vote or legionaries serving outside Italy, and that they made no attempt to achieve full registration of the no doubt numerous proletarians.7 These omissions would, in his view, go a long way towards explaining how about 45 per cent of all adult male citizens might have remained unregistered by the censuses of the third and second centuries bc. Lo Cascio also claims that up until the time of Caesar the Roman government maintained a fully centralized census procedure.8 This would have meant that citizens optimo iure who lived in the peripheral parts of the Ager Romanus (for example, the viritane settlers sent out to Upper Sabinum, the Ager Praetuttianus and Picenum during the third century bc) were expected to travel to Rome in order to present themselves before the censors. During the second century bc, viritane settlers and colonists of citizen status were also sent out to Apulia, Samnium, Aemilia and Liguria, and by the mid-80s bc all of the formerly allied communities south of the Po and all of the Latin colonies in Transpadana had been enfranchised. The corollary of Lo Cascio’s theory is therefore that the registration rates achieved by the censors became progressively lower towards the end of the Republic. It has, in fact, been calculated that his interpretation of the census figure for 70/69 bc implies a registration rate as low as 30 per cent.9 According to Lo Cascio, this trend continued until 45 or 44 bc, when Caesar introduced the fully decentralized census procedure described in the

6 7 9

Mommsen (1887), ii, 388–96; Brunt (1971/1987), 15–16; Northwood (2008), 258–60. Lo Cascio (1999b), 232–3. 8 E.g. Lo Cascio (1994); (1999b); (2001b); (2008). Scheidel (1996a), 168.

82

Meaning of the census figures

Tabula Heracleensis.10 Although Lo Cascio assigns a certain importance to the enfranchisement of the allied communities of Transpadana (49 bc) and of large numbers of provincials, he identifies Caesar’s administrative reform as the main reason why the Augustan census figures are between four and a half and five and a half times higher than the census figure for 70/69 bc.11 In the sections that follow, I shall first examine the arguments that have been adduced in favour of or against the theory that various groups of citizens were excluded from the republican censuses. I shall then devote a separate section to the problem of (de)centralization. Finally, I shall briefly examine some of the strengths and weaknesses of competing interpretations of the Augustan census figures. 3.2 the target population of the republican censuses: iuniores and seniores Since the mid-nineteenth century, some ancient historians have maintained that the republican census figures are to be interpreted as referring exclusively to those adult men who could be called up for active military service (in other words, as referring solely to iuniores).12 In recent years, this interpretation has found very few adherents. A partial exception must be made for Lo Cascio, who has suggested that originally the magistrates responsible for the census might have counted only men younger than 46. His main argument is that Fabius Pictor explained the (no doubt fictitious) census figure for the reign of Servius Tullius as referring to those qui arma capere possent.13 If we follow Lo Cascio in interpreting this expression as denoting only those liable for legionary service under normal circumstances, we must conclude that there was a time when only the number of iuniores was recorded.14 It should be emphasized that Lo Cascio fully endorses the prevailing view that by the late fourth century bc (possibly much earlier) the scope of the census had been widened to include all adult men of citizen status, including seniores. 10

11 14

The decentralized nature of the census procedure described by the Tabula emerges clearly from lines 142–6, which read as follows: ‘Whatever municipia, colonies or prefectures of Roman citizens are or shall be in Italy, whoever in those municipia, colonies or prefectures shall there hold the highest magistracy or the highest office, at the time when a censor or any other magistrate shall hold the census of the people at Rome, within the sixty days next after he shall know that the census of the people is conducted at Rome, he is to conduct a census of all his fellow municipes and colonists and those who shall be of that prefecture, who shall be Roman citizens.’ (Translation: Crawford and Nicolet, in Crawford 1996, 377.) Lo Cascio (2001b), 591. 12 Mommsen (1887), II, 411. 13 Livy 1.44.2. Lo Cascio (2001b), 569–73; cf. Lo Cascio (1999a), 163.

Target population of the census

83

As Beloch pointed out long ago, one weakness in the theory that the census figures refer solely to iuniores is the simple fact that the phrase civium capita contains no reference to any age limit.15 In the second half of the twentieth century this argument was endorsed by Brunt, who observed that one of the aims of the census was to create a reliable basis for the assessment of tributum, which seniores were clearly expected to pay. Since it is difficult to think of any reason why the censors should have registered tax-payers over 45, then ignored them when reporting the total number of civium capita, the most natural assumption is that seniores were included in the published census figures.16 In my view, Beloch’s and Brunt’s arguments are not in any way contradicted by Fabius’ use of the expression qui arma ferre possent in relation to the census allegedly carried out by Servius Tullius. As we saw in the previous chapter, the fact that the Roman legions were normally made up of men aged between 17 and 45 does not alter the principle that, at least in theory, any man below the age of 60 could be called up for military duties.17 This explains why Livy’s description of the comitia centuriata (which must originally have been an assembly of all those capable of bearing arms) refers to the possession by seniores and iuniores of identical military equipment. For the purposes of the present discussion, the most important conclusion is that there is no need to posit any shift from an early census that registered only iuniores to a later system covering both iuniores and seniores. Moreover, if (as may well be the case) the practice of census taking evolved during the first century of the Republic, neither Fabius Pictor nor any of his successors is likely to have had any reliable information about the nature or scope of the census in its earliest forms. As Northwood has pointed out, the few pieces of evidence available tend to suggest that all reconstructions of the early-republican census were actually retrojections of administrative practices of the third, second and first centuries bc,18 and as we have seen, there can be no doubt that during this later period the censors were expected to register both seniores and iuniores. 3.2.1 Citizens sui iuris and citizens alieni iuris Another theory, also dating back to the nineteenth century, postulates that all republican census figures should be interpreted as referring to the number of adult male citizens sui iuris.19 15 18 19

Beloch (1886), 317–18. 16 Brunt (1971/1987), 22. 17 Chapter 2, section 2.2.2. Northwood (2008), 257. Zumpt (1840); Hildebrand (1866). In the twentieth century this theory was revived by Bourne (1952). A recent adherent is Hin (2008) and (2009).

84

Meaning of the census figures

At first sight, this theory is attractive. As we have already seen, the republican censors compiled their lists of citizens using the information supplied by citizens sui iuris (or their representatives) and it seems clear that these were the only declarants. The reasons for this are obvious. Since all Roman citizens were either sui iuris or subject to another citizen who was sui iuris (that is, to a pater familias), accurate information collected from all of the citizens sui iuris would have included every male citizen and could in theory have included the entire citizen population, if those supplying the information were also expected to list every female subject to their potestas. Another practical reason for requiring only citizens sui iuris to present themselves before the censors was that under Roman law these were the only citizens who could own property. Full registration of the citizens sui iuris would therefore have given the censors all of the information they needed to assign adult males to the appropriate property classes. The same information could, of course, also have been used to determine the amount of tributum to be imposed and (in the case of widows and orphans) to determine liability to the aes equestre and the aes hordiarium, used to defray the costs incurred by the equites equo publico.20 Despite these attractive features, the theory that the republican census figures refer to adult men sui iuris raises certain questions that are difficult to answer. It is, for instance, not easy to understand why the censors should simply have published the total number of citizens whose property had been assessed. It is true that such a figure would have revealed the number of adult male citizens upon whom tributum could be imposed, but it seems likely that the Roman authorities would have been more interested in a figure that revealed Rome’s military strength. This reading of the republican census figures would be in keeping with the prevailing view that the census must have originated as a periodic review of the republican army.21 The view that the censors were expected to report the number of men who qualified for military service rather than the number of adult men sui iuris is supported by certain indications in the literary and epigraphic sources. As Beloch points out, one of these is the existence of the technical term duicensus, which Festus explains as cum altero, id est cum filio census.22 The most natural interpretation of this term is surely that it referred to the counting of both a father and his dependent son.23 It is also striking that the 20 21 22

23

E.g. Mommsen (1887), iii, 236, 256–7; Gabba (1949/1976), 8; Kaser (1966a), 105. E.g Wiseman (1969), 59; Lo Cascio (2001b), 569–74. Festus 58L. Cf. Beloch (1886), 316; Brunt (1971/1987), 22; Nicolet (1988), 68; Lo Cascio (2001b), 575; Northwood (2008), 259. Note that Hin (2008), 216, n. 96, is unable to come up with a convincing interpretation of the term duicensus.

Target population of the census

85

Tabula Heracleensis orders the local magistrates of Italy to register all those (males) ‘who shall be Roman citizens’ (quei cives Romanei erunt), without narrowing down the target population to citizens sui iuris.24 Similarly, in Cicero’s De Legibus the duties of the censors are described as follows: censoris populi aevitates, suboles, familias, pecuniasque censento (‘the censors are to record the ages, the offspring, the families, and the property of the populus’).25 Here not only the declarants but also their offspring (suboles) are the object of censento. Such passages show that in technical contexts to do with the census the Latin verb censere need not (and often does not) refer specifically to a valuation of property. It is, in fact, clear that in such contexts censere often means ‘to register’ (one of its primary meanings, according to the Oxford Latin Dictionary).26 In my view, this is the key to Livy’s brief reference to the census of 465 bc, of which he reports that ‘the number of civium capita registered (censa), apart from orphans and widows, is said (dicuntur) to have been 104,714’.27 We must surely infer that although orphans and widows were not included in the republican census figures, they were regarded as censi in the sense of ‘having been registered’, evidently because financial burdens could be imposed on them.28 Similarly, dependent sons aged 17 or over had to be registered because they had the right to vote in the centuriae iuniorum of the comitia centuriata.29 Another important argument is that Livy clearly interpreted the republican census figures as referring to all adult men. The best illustration of this is perhaps his well-known digression on the likely outcome of a hypothetical war between the army of Alexander the Great and the Roman armies of the later fourth century bc. As Livy explains, this question can only be answered by comparing the ‘opposing’ forces, taking into account numbers 24 25 27

28

29

Tabula Heracleensis, lines 145–6 (cf. note 10), as interpreted by Northwood (2008), 259. Cic. Leg. 3.3.7. 26 OLD s.v. censeo. Livy 3.3.9. In my view, the term dicuntur militates against Scheidel’s theory (Scheidel 2009b, 656–7) that the phrase praeter orbos orbasque was added by Livy himself when he wrote Book 3. The fact that Livy uses a different expression (praeter pupillas et viduas) to denote orphans and widows in referring to the census figure for 131/130 bc (Per. Livy 59) is another hurdle for Scheidel’s theory. Instead of focusing on the dates of composition of Books 3 and 59 of the Ab Urbe Condita, we should consider the possibility that Livy copied these expressions from his annalistic predecessors. For this reason, these passages cannot be used to buttress Hin’s theory that, unlike the census figures for the last two centuries of the Republic, those for 28 bc, 8 bc and ad 14 must have included all citizens sui iuris, including orphans and widows. For the financial obligations of orphans and widows cf. above, at note 20. According to some sources (e.g. Plut. Cam.2.2), these taxes were introduced in the early fourth century, but Livy 1.43.9 states that they dated from the reign of Servius Tullius. E.g. Livy 1.43. Cf. Nicolet (1988), 49. Northwood (2008), 259, n. 8, suggests that sons were assigned to the same classes as their fathers.

86

Meaning of the census figures

of soldiers, types of soldier and the sizes of auxiliary contingents.30 He then notes that at the time of Alexander 250,000 capita were registered and that ten Roman legions were levied to meet the revolt of the Latin allies.31 In my view, the only possible reading of this passage is that the 250,000 capita meant the total number of adult fighting men Rome was able to field.32 The same conclusion emerges from Livy 29.37.5, where we read that the censors of 204/203 bc sent agents out through the provinces ‘to report on the number of Roman citizens who were in the armies in the various locations (ut civium Romanorum in exercitibus quantus ubique esset referretur numerus). With these included, the census numbered 214,000 souls.’ Since there is nothing in this text to suggest that only legionaries sui iuris were to be included in this enumeration, we must surely infer that the census figure for 204/203 bc included all Roman citizens serving in the legions. As we have just seen, this reading is in perfect harmony with the much later Tabula Heracleensis, which also refers to a general registration of all cives Romani. Finally, it is instructive to compare Livy’s use of language with that of Dionysius of Halicarnassus. When writing about the censuses of the fifth century bc, Livy invariably uses the traditional expression censa (sunt) civium capita (for instance, when he reports that 104,714 civium capita were registered in 465 bc).33 It is in my view highly significant that Dionysius uses the expression hoi en hêbêi (‘those in their prime’, or ‘the adults’) to describe the category of men referred to by two of the census figures that he gives for the early decades of the republic.34 Although it 30 31

32

33 34

Livy 9.19.1. The figure of 250,000 may well be unrealistically high (e.g. Brunt 1971/1987, 27–8; Hopkins 1978, 20, n. 26), but this does not affect my argument. Hin (2008), 202–3, and (2009), 173, suggests that the term capita may refer to the number of units (i.e. households headed by a pater familias) registered. This meaning is not found in any other source, however, whereas there are many texts (especially amongst the works of the Roman jurists) in which caput means ‘individual’ or ‘person’. See Heumann-Seckel (1926), s.v. caput. The latter meaning may also be found in Livy 22.57.11 (a shortage of capita libera after Cannae), in Livy 40.38.6 (40,000 libera capita of the Apuani, including women and children, deported to Samnium), and in many other passages (cf. Giovannini 2008, 52–3). The term capita therefore supports the view that the republican census figures should be read as ‘head counts’ of those able to bear arms, i.e. all men aged 17 or older (cf. below, at note 112), although during the middle and late Republic the number of ‘heads’ was derived from the declarations made by the patres familiarum rather than counted in a literal sense. Livy 3.3.9. E.g. DH 5.20.1 (508 bc) and 5.75.3 (498 bc). In 11.63.2 Dionysius says that in 443 bc the number of hoi en hêbêi was unknown because no census had been taken for 17 years. Bourne’s theory (1952, 130) that Dionysius must have calculated the number of adults from the number of men sui iuris enumerated by the censors strikes me as highly implausible. It is true that in another passage (9.36.3) Dionysius interprets the relatively low census figure for 474 bc as referring only to adult men sui iuris, but I believe that here he is drawing on a source which attempted to explain why the figure for this year is so much lower than those for 508 bc, 498 bc and 493 bc. The annalist

Target population of the census

87

would be possible to hypothesize that Dionysius (or his predecessors) knew, or thought they knew, that the censors of the middle and late Republic had abandoned an earlier procedure aimed at establishing the total number of adult males, it is surely far more probable that the annalistic tradition projected later administrative practices back onto the late sixth and early fifth centuries bc.35 In any case, even if we were to take on board the theory that the census figures for the period 234–69 bc refer solely to adult men sui iuris, this would push up the number of adult male citizens by no more than between 10 and 13 per cent.36 3.2.2 Cives optimo iure and cives sine suffragio During the second half of the twentieth century, some scholars developed the theory that the Roman census figures for the third and second centuries bc did not include the cives sine suffragio (inhabitants of those Italian communities that had been given Roman citizenship, but not the right to vote or to present themselves as candidates for Roman magistracies).37 Although this theory has never attracted many adherents, it has recently been revived by Lo Cascio. An early attempt to disprove the theory that the republican census figures should be interpreted as including only cives optimo iure is to be found in Chapter Eight of Beloch’s Die Bevölkerung der griechisch-römischen Welt. His main arguments were (a) that Polybius gives a single manpower figure for the Romans and the cives sine suffragio of Campania and (b) that citizens without the vote must have been registered because they served in the legions and paid tributum.38

35

36

37

38

responsible for this ingenious reading must have been inspired by the fact that, even in his own time, the census records were compiled on the basis of declarations made by citizens sui iuris. This anomalous passage does not, of course, diminish the crucial significance of the phrase hoi en hêbêi. The census figures for the late sixth and early fifth centuries bc are widely regarded as artificial creations (e.g. Brunt 1971/1978, 26–7; Alföldi 1977, 125–7; Scheidel 2004, 4, n. 17). This makes it difficult to maintain that the annalists of the second and first centuries bc had access to reliable information regarding the census procedures of the early Republic. Nor are they likely to have invented the existence of a procedure completely unrelated to the administrative practices of their own time. Precisely for this reason, it is quite reasonable to use references to the early republican censuses to illuminate later practices (contra Hin 2008, 204–6). Calculated from Hin (2008), 199, Table 1. The hypothetical figure of 13 per cent applies only if we assume that during the Republic adult men whose fathers were still alive never became independent by means of emancipatio. This assumption is clearly unrealistic. E.g. Shochat (1980), 13–16, 36. Shochat supported his theory by pointing to the discrepancy between the census figure for 234 bc and Polybius’ figure for the Romans and Campanians. See Chapter 2, at notes 47–9, for a completely different explanation. Beloch (1886), 318–19.

88

Meaning of the census figures

Neither of these arguments is entirely convincing. As various scholars have pointed out, the Romans and Campanians are not the only groups listed together by Polybius; he also groups together the Etruscans and the cives sine suffragio of (Upper?) Sabinum. As we saw in Chapter 2, Lo Cascio has inferred from this that the Polybian manpower figures are based on regional groupings without regard to the juridical distinctions between citizens, Latins and Italian allies. This theory can easily be refuted.39 Even if we interpret Polybius’ account as being based on a clear distinction between Romans and allies, however, it might perhaps still be possible to argue that, precisely because his basic categories seem to be juridical, his description of the preparations of 225 bc need not reflect the workings of the normal Roman levy. The argument that the censors must have counted cives sine suffragio because these citizens served in the legions is equally insecure. As Mommsen observes, a passage in Festus explicitly states that citizens without the vote served in legione, suggesting that they had the technical status of legionaries.40 Mommsen finds it significant that we never read of alae or turmae (allied units) made up of citizens without the vote, and that Polybius gives a single figure for ‘the Romans and the Campanians’. He also notes that the Roman levy as described by Polybius seems to have been based exclusively on the thirty-five tribes from which the half-citizens were excluded.41 From these facts he concludes that cives sine suffragio and full citizens must have been recruited in separate levies and that the former group must have performed some special form of military service of which no trace survives in the sources.42 With some minor variations, this overall interpretation has been adopted by many twentieth-century ancient historians.43 Against this theory, Lo Cascio has argued that the year 332 bc witnessed the introduction of a new census-taking procedure based on the tribes. Although he concedes that the censors may have been expected to register cives sine suffragio before this reform, Lo Cascio reasons that the fact that these citizens had no tribe must mean that they cannot have been registered by the censors after 332 bc. Instead, he suggests, they must have been registered by local magistrates who carried out censuses independent of those periodically undertaken by the Roman censors.44 Building on these observations, Lo Cascio goes on to construct a very similar argument with 39 41 43

44

Chapter 2, sections 2.2 and 2.2.1. 40 Festus 117 L. Plb. 6.20. Cf. Mommsen (1887), iii, 586. 42 Mommsen (1887), iii, 587. Humbert (1978), 319–20, argues that the cives sine suffragio served in separate units, but nevertheless had the same legal status as Roman legionaries. Lo Cascio (2001b), 579.

Target population of the census

89

respect to the system used to recruit citizen soldiers for the legions. In his view, since the cives sine suffragio could not have been levied by means of the tribe-based system, they must have been expected to serve in the allied contingents. This alternative interpretation implies that Festus’ statement that citizens without the vote were expected to serve in legione must refer to the system of recruitment used before 332 bc 45 and that all other references to cives sine suffragio serving in legions should be regarded as non-technical.46 One obvious weakness of this ingenious theory is that the suggestion that local magistrates were responsible for the registration of cives sine suffragio is entirely compatible with the view that the results of these local censuses were sent to the censors in Rome, who could then have combined them with their own figures to calculate the total number of Roman citizens. In this context we should note the existence of the tabulae Caeritum, which contained not only the names of the cives sine suffragio, but also the names of those citizens whom the Roman censors had deprived of the right to vote.47 Since these lists were kept in Rome, it may be inferred that the Roman censors had access to tabulae listing Rome’s half-citizens. Since almost all of the technical clues we have either are open to alternative interpretations or can be eliminated by assigning them to purely hypothetical earlier stages in the history of the census, the only way to resolve this old debate is to take a fresh look at the balance between people and territory implied by the theory that citizens without the vote were not included in the census figures for the third and second centuries bc. The census figure for 280/279 bc, when 287,222 capita civium were registered, provides us with a useful starting point.48 In this year the Ager Romanus covered some 15,295 square kilometres. A very large proportion of this area consisted, however, of regions that had recently been annexed and whose inhabitants had been given the status of cives sine suffragio. Adding up Afzelius’ figures for the areas inhabited largely or exclusively by citizens without the vote gives us an estimate of no less than 12,750 km2, leaving only about 2,545 km2 for those areas originally inhabited by cives optimo iure.49 45

46 48

49

It may be objected that before 332 bc there were very few cives sine suffragio apart from the Campanians. Note that Velleius Paterculus 1.14.3 dates the bestowal of citizenship without the vote on the Campanians to 334 bc rather than to 338 bc, the date suggested by Livy 8.14.10 (Taylor 1960, 60). Lo Cascio (2001b), 582–3. 47 Strabo 5.2.3; Gellius 16.13.7. Per. Livy 13. Of course, all census figures relating to the period before the First Punic War can be dismissed as fictitious (Chapter 1, note 12), but Lo Cascio (2001b), 566, thinks it possible that at least some of the early census figures were based on written records that had survived in family archives. Afzelius (1942), 181, 190–1. At this early date, the main areas in which citizens without the vote are likely to have made up the majority of the population were the Ager Gallicus (2,580 km2), Sabinum

90

Meaning of the census figures

If we begin from the (simplifying) assumption that all of the c. 948,000 people of citizen status implied by the census figure for 280/279 bc lived within this area,50 we arrive at an average density of c. 372 inhabitants per square kilometre, an absolutely incredible figure. In reality, a significant number of citizens could have benefited from various viritane assignations carried out in areas also inhabited by cives sine suffragio. The assignations carried out by Dentatus in Sabinum, in the Ager Praetuttianus and perhaps also in other areas spring to mind.51 Even if we assume, generously, that 50,000 viritane settlers (men, women and children) had already been sent out by 279 bc, though, the average density for the areas inhabited by the remaining 898,000 Roman citizens who supposedly had the vote would still be as high as c. 353 citizens per square kilometre. If we repeat this exercise for 234 bc, it would appear that c. 890,000 people of citizen status must have been accommodated within the approximately 9,700 km2 of the Ager Romanus.52 This time the average density implied by these figures is about ninety-three persons per square kilometre. If we optimistically assume that the number of Roman settlers in Sabinum,

50

51 52

(4,340 km2), the areas confiscated from the Aequi and the Aequicoli (1,225 km2), Aveia and Peltuinum (640 km2), the Ager Praetuttianus (1,390 km2), the Ager Campanus (1,075 km2) and the areas inhabited by the cives sine suffragio of Latium (at least 1,500 km2). I have excluded Forum Clodii and the praefectura Statoniensis in South Etruria (annexed between 284 bc and 280 bc, according to Afzelius 1942, 69, and Forsythe 2005, 349), and the Umbrian praefecturae of Tadinum, Nuceria, Fulginiae and Plestia, where prefects seem to have dispensed justice mainly to the descendants of viritane settlers. See Afzelius (1942), 117; Brunt (1971/1987), 528, n. 5; Sisani (2007), 218. 287,222 x 3.3 = 947,833. Those subscribing to Lo Cascio’s theory that the republican censuses missed a large proportion of the target population must necessarily posit a much larger citizen population. We must also take into account an unknown number of slaves. Mine are minimum figures. E.g. Toynbee (1965), 104. 270,713 x 3.3. = 893,353. According to Afzelius (1942), 192, the Ager Romanus (not counting the Ager Picentinus south of Salernum) comprised some 25,800 square km in 234 bc. From this area the following regions must be subtracted: Caere (420 km2), the Ager Praetuttianus (1,390 km2), Upper Sabinum (3,305 km2), the Ager Gallicus (1,930 km2) and Picenum (2,900 km2), the four ex-Samnite praefecturae (1,715 km2), the territories confiscated from the Aequi, the Aequicoli and the Vestini (1,225 + 640 km2), the Ager Campanus (1,075 km2) and the areas belonging to the cives sine suffragio of Latium (at least 1,500 km2). Total: 16,100 km2. Since the evidence for the early incorporation of the territory of Aufidena is very uncertain (e.g. Sherwin-White 1973, 208, n. 1; Galsterer 1976, 24), I have excluded it from my calculations. In the case of Upper Sabinum, I have followed Afzelius, Taylor and Toynbee in assuming that the original inhabitants of Upper Sabinum were still cives sine suffragio in 225 bc (cf. Chapter 2, note 102). In my view, this remains the most natural explanation for the fact that Polybius 2.24 groups the Sabines together with the Etruscans and also for the curious fact that Livy 28.45.19 mentions Reate, Nursia and Amiternum alongside Caere and a number of allied towns in a list of communities that provided Scipio with men, equipment and grain in 205 bc. Even if the indigenous inhabitants of Upper Sabinum had already been given the vote by this date, this region is unlikely to have had more than 50,000 inhabitants (based on a density of 15 inhabitants per square km). It follows that the uncertainties surrounding the legal status of the Sabines cannot affect my argument, especially since my hypothetical figure for the number of viritane settlers sent out to the areas conquered by Dentatus is very high.

Target population of the census

91

the Ager Praetuttianus, southern and eastern Umbria and the praefectura Statoniensis in Etruria had risen to 100,000 by this date,53 we still end up with a density of c. eighty-one people per square kilometre for the remaining parts of the Ager Romanus. Although this figure is much lower than the density that can be calculated for 280/279 bc, it is still formidably high,54 especially if we take into account the presence of numerous slaves and the likelihood that all republican census figures for the third and second centuries bc were seriously defective. This last argument ought to be particularly compelling to the high counters, whose overall reconstruction of the demographic history of Italy is based on the assumption that a very large proportion of those adult male citizens who had the vote remained unregistered by the censors of this period.55 In the case of the high count, these difficulties are compounded by the fact that the very high population densities implied by the theory that the cives sine suffragio were not included in the republican census figures would make it impossible to model any further population growth in centralwestern Italy. This would create serious conflict with the archaeological evidence from southern Etruria and Latium.56 For all these reasons, the only possible conclusion is that the Roman censors of the third and second centuries bc were in fact expected to report a total figure covering all adult male citizens, including those who were cives sine suffragio. As we have seen, the problem of whether or not cives sine suffragio were included amongst the number of capita civium reported by the censors is closely bound up with the question of whether or not these citizens served in the legions. This notorious conundrum can, in my view, be resolved by taking a fresh look at the date of the recruitment procedure described by Polybius. As we have seen, Mommsen assumes that Polybius’ account must 53

54

55

56

For viritane assignations carried out in Umbria and Etruria and in the territories of some Volscian towns, see Taylor (1960), 84–6; Galsterer (1976), 32–3; Bradley (2000), 139–44; Sisani (2007), 134–5, 214–16. Judging from Beloch (1937–1961), i, 254 and 268, ii, 211, and iii, 380–1, densities of between thirty-five and fifty-five persons per km2 were the norm in central and southern Italy before the arrival of New World crops. The areas corresponding to Roman Campania had higher densities (e.g. Beloch 1937– 1961, i, 235), but since Lo Cascio thinks that the Campanians were excluded from the census, this does not affect my argument. Note that Lo Cascio himself (1999a, 168) operates with a density of fifty-six free inhabitants per square kilometre of ager Romanus in 225 bc. Since this time Campania is included in his calculations (cf. Chapter 2, at notes 22–7), those areas inhabited by cives optimo iure ought to have had fewer than fifty-six free inhabitants per square km. Lo Cascio’s theory that there were some 474,000 adult male citizens in 225 bc (1999a, 169) implies that the censors of the third century bc registered approximately 55 per cent of the target population. His estimates for the second century imply a registration rate of between 43.5 and 53 per cent. Cf. Chapter 4, at notes 149–50. Chapter 6.

92

Meaning of the census figures

have been valid not only for his own time but also for the third century bc.57 A variant of this theory has been defended by Brunt, who argues that Polybius’ description of the levy contains a number of anachronistic elements. The most obvious of these is perhaps that Polybius gives the standard size of a Roman legion as 4,200 men, whereas the evidence supplied by Livy suggests that most or all Roman legions were 5,200 strong from the early second century bc.58 Brunt also draws attention to various indications suggesting that Polybius must have adapted an existing written source. One clue pointing in this direction is that Polybius’ description of the levy as ending with the raising of the cavalry is followed by the statement that the cavalry is ‘nowadays’ actually reviewed first.59 Brunt’s conclusion is that Polybius must have used an antiquarian description of the levy that did not accurately reflect the conditions of his own time.60 Although some of Brunt’s arguments are sound, his final conclusion cannot be accepted. At this point, I should like to underline the crucial importance of an article by Elizabeth Rawson that has not really been digested by the recent literature, despite the fact that it is often referred to. To my mind, Rawson has convincingly demonstrated that even though Polybius’ account is clearly based on an earlier written source, there is good reason to believe that this source cannot have been more than 20 years old. More specifically, she interprets the curious fact that the levy is consistently described from the point of view of the military tribunes as an indication that this source must have been a handbook for this class of officer, rather than some antiquarian account containing many anachronistic elements. This implies that, contrary to Brunt’s contention, Polybius’ description is in fact a more or less reliable guide to the realities of his own time.61 In this context it may be noted that it is highly misleading to insist on the fact that Polybius’ figure for the size of a normal legion (4,200 men) falls short of the 5,200 men implied by Livy’s figures for the period 184 bc–167 bc. The simple reason for this is that, although Polybius begins by speaking of 57 60 61

Above, at notes 40–2. 58 Brunt (1971/1987), 674–6. 59 Plb. 6.20. Brunt (1971/1987), 627, followed by Galsterer (1976), 106. Rawson (1971). As she points out (ibid. 14), it is doubtful whether any antiquarian account of the levy existed as early as 160 bc, and even if one did, Brunt’s hypothesis would not explain the curious emphasis on the activities of the military tribunes. Dalby (2002, 215, n. 16) suggests that Polybius may have composed his account of the levy and the Roman army at the time when his friend and patron Scipio Aemilianus was serving as a military tribune. This alternative hypothesis would imply that Polybius is describing the levy as it was in the 160s bc. Using a slightly different approach, Dobson (2008), 55, argues that Polybius used a source that can be dated to c. 216 bc, but that his account is broadly valid for his own time because he took account of later developments.

Target population of the census

93

legions of 4,200 men, he explicitly refers to the custom of levying legions of 5,200 men in the event of a serious threat.62 In other words, his reference to legions of 4,200 men implies no more than that some legions of this size were still being levied in his own time, as indeed may well have been the case.63 Interestingly, the theory that Polybius is describing conditions that prevailed in his own time enables us to shed new light on a problem that has never received the attention it deserves: Polybius’ statement that the number of troops provided by the allies roughly equalled the number of legionaries provided by the Romans.64 As some scholars have observed, this claim is completely out of line with the figures we have for the sizes of Roman legions and allied contingents during the period 225 bc–180 bc.65 What is particularly puzzling is that some of the figures given by Polybius himself, such as those relating to the preparations of 225 bc, point to a ratio between allies and citizens of approximately 3:2. What has been overlooked is that Polybius’ claim that the aggregate size of the allied contingents was ‘about equal’ (parison)66 to that of the legions is entirely correct of most armies raised during the 170s bc and early 160s bc, when the ratio of allies to citizens had declined to little above parity.67 In my view, this observation

62 63

64 66 67

Plb. 6.20.8. The consistent use of the present tense in Plb. 1.16.1–2, where the standard size of a legion is also given as 4,200, supports Rawson’s view that Polybius’ account is valid for his own time. Toynbee (1965), II, 50–1, argues that only those legions that are explicitly stated to have been made up of 5,200 men had been reinforced, with legions of 4,200 men still being the norm during the first half of the second century bc. Although this theory seems over-ingenious, there is some reason to believe that the praesidial legions sent out to Sardinia and Cisalpine Gaul numbered no more than 4,200 men (Afzelius 1944, 77; cf. Livy 32.8.7 for the size of the allied force guarding Sicily in 198 bc). Supplementa of 4,400 and 4,200 men for the legions in Spain are recorded for 195 bc, 184 bc and 182 bc (Livy 33.43.7; 39.38.10–11; 40.1.6–7). It must also be realized that the reliability of Livy’s Roman army figures is open to doubt. According to the sceptical German tradition exemplified by Gelzer (1935/ 1964) and Gschnitzer (1981), all references to military arrangements in the western Mediterranean stand a good chance of being late-annalistic inventions, with explicit statements about troop strength and casualties being particularly suspect. Note that the late-republican annalists were capable of inventing legions of 5,300 men for the mid-fourth century bc. See e.g. Livy 8.8.14, where (pace Daly 2002, 73) the accensi are clearly a regular component of the legions rather than ‘hastily armed servants’. The most reasonable conclusion may be that although legions of 5,200 men had become common by around 160 bc, the notion of a standard legion of 4,200 men was still alive at this time. Cf. Dobson (2008), 49, for the suggestion that legions of 4,200 continued to be levied during the second century bc, although the larger legion size came to be more usual. Plb. 3.107.12. 65 E.g. Afzelius (1944), 62–3; Ilari (1974), 174–5; Brunt (1971/1987), 681. For the meaning of this term cf. Marquardt (1881), 380. As observed by Brunt (1971/1987), 681. Note his finding that ‘Polybius’ generalization that Romans and allies served in equal numbers is seldom borne out by specific instances except c. 170 b.c .’ (1971/1987, 683, my italics). The seemingly enigmatic ratio of 3:1 between allied and Roman cavalry can be explained by assuming that in Polybius’ time allied contingents of cavalry included Numidian and Gallic auxiliaries. For Numidian cavalry serving in Roman armies operating outside Africa, see e.g. Livy

94

Meaning of the census figures

reinforces Rawson’s view that Polybius’ account of the levy is a reliable guide to the practices of the period 180 bc–160 bc. For those interested in the relationship between the republican census and the levy, the most important advantage of Rawson’s thesis is that it enables us to get round the old problem of exactly where the cives sine suffragio should be fitted into the military structures of the third and early second centuries bc. In my view, there is nothing to contradict the theory that, before and during the Second Punic War, the cives sine suffragio were levied by local magistrates and then sent to a point of assembly where they joined the Roman legions.68 In other words, just as the number of cives sine suffragio registered by local magistrates was added to the number of capita civium already registered by the censors in Rome, any adult male citizens sine suffragio recruited by local magistrates could be added to those legionaries already recruited in Rome. This theory, of course, raises the question of why Polybius’ account never mentions citizens without the vote. Brunt claims that this is simply an accidental omission, and his theory cannot be disproved. There is nevertheless an alternative explanation that is far more attractive. As has often been noted, we can find no definite references to communities of cives sine suffragio existing after 180 bc.69 Up to a point this is not surprising, especially since the literary record for the period 167 bc–91 bc contains so many lacunae. Despite this methodological difficulty, it remains tempting to infer that, with the possible exception of the Campanians,70 all of the citizens sine suffragio were enfranchised during the early decades of the second century bc.71 If this widely held theory is correct, the fact that

68

69 70 71

31.19.3–4 (Macedon, 200 bc: 1,000 Numidians); 42.35 (Macedon, 171 bc), Sall. BJ 7.1 (Spain, 134 bc). The troops supplied by Masinissa for the Spanish war in 150 bc (Val. Max. 5.2, ext. 4) must also have included cavalry. We also read of Gallic cavalry serving in the Third Macedonian War (Livy 42.58.13, 171 bc). Increasing reliance on non-Italian cavalry would help to explain why we have no references to Italian communities contributing three-quarters of the cavalry after the end of the Second Punic War (Brunt 1971/1987, 683). For auxiliaries being accommodated in the Roman camp, cf. e.g. Livy 21.48.1 (Gauls, 218 bc); Livy 26.12 (Numidian deserters, 211 bc). Cf. the allied units that were combined into larger fighting forces at points of assembly determined by the consuls (Plb. 6.21.4–5). The fact that the survivors of the legio Campana that had massacred the men of Rhegium and seized their town were scourged and beheaded in the Forum Romanum (Plb. 1.7.12; DH 20.5.5) also suggests that these men had the status of legionaries. E.g. Brunt (1965/1988), 103–4, and (1988), 136; Humbert (1978), 346–54; Nicolet (2001), 730. Cf. Chapter 4, at note 25. Mouritsen (2006), 421, n. 19, and (2007), 143–8, correctly insists that we hear only of the enfranchisement of Fundi, Formiae and Arpinum (Livy 38.36.7–9). Cf., however, Taylor (1960), 93, for the plausible suggestion that this unique report was preserved because the proposed assignment of these communities to the tribus Aemilia and Cornelia was controversial. We must also consider the relevance of practical considerations such as the desirability of simplifying the recruitment of legionaries in areas with mixed populations of cives optimo iure and citizens without the vote.

Target population of the census

95

Polybius describes a system of recruitment exclusively based on the tribes can be explained as reflecting the recent enfranchisement of all or almost all cives sine suffragio and their inclusion in the thirty-five Roman tribes. Although the usual dearth of conclusive evidence makes it impossible to prove this reading of the evidence, it is at least possible to conclude that Polybius’ description of the levy is compatible with the view that the cives sine suffragio were liable to be called up for legionary service, and also compatible with the view that these citizens were included amongst the capita civium counted by the censors in Rome. 3.2.3 Legionaries serving outside Italy It is often assumed that the censors were not normally expected to register citizens who were on active service in the legions, especially if these were operating in the provinces. Although this theory would seem to be supported by a considerable amount of evidence, it is in fact extremely fragile. In his Italian Manpower, Brunt calls attention to Livy’s account of the year 378 bc.72 As Livy explains, in this year the main problem on the domestic front was that many citizens were deeply in debt. We are also told that Spurius Servilius Priscus and Quintus Cloelius Siculus were appointed censors to clarify the financial position of the Roman citizens. When the Volscians invaded Roman territory, however, and two legions needed to be enrolled, apparently no census could be taken. Brunt interprets this story as meaning that it was not possible to count Rome’s citizens when most of them were already under arms. In my view, this inference gives Livy’s story a general significance that it does not have. What Livy means is quite clearly that, with a large proportion of the adult male population in the field, it was impossible for the censors to call in property-owners in order to determine their assets and liabilities. He is certainly not implying that it would have been impossible to determine the number of men aged between 17 and 60. Two other passages that are often referred to in this context are Livy 27.36, where we read that only 137,108 capita civium were counted by the censors of 209/208 bc, and Livy 29.37, which records a figure of 214,000 capita civium for 204/203 bc. As we have already seen, the censors of 204/ 203 bc sent special officials out to the provinces ‘in order to determine the

72

The fact that many communities that had obtained civitas sine suffragio between 338 bc and 268 bc continued to use Oscan right until the Social War cannot, of course, be used as an argument against early enfranchisement. As Mouritsen (2007), 148–50, has convincingly demonstrated, the bestowal of the right to vote cannot be seen as part of a general policy of Romanization, for the simple reason that there was no such policy. Livy 6.31.2. Cf. Brunt (1971/1987), 70, n. 2.

96

Meaning of the census figures

number of Roman citizens serving in the army’.73 Brunt suggests that the low figure for 209/208 bc can be explained if we suppose that some 75,000 adult male citizens serving in the legions or in the fleet were omitted.74 He regards the decision of the censors of 204/203 bc to register soldiers serving in the provinces as unusual. One obvious weakness of this theory is that Livy describes the census figure of 209/208 bc as ‘somewhat lower’ (minor aliquanto) than the pre-war figure, which must have been higher than the census figure for 234/233 bc. Since the discrepancy between the figures for 234/233 bc and 209/208 bc is more than 130,000, we must surely reckon with the possibility that we are dealing with a scribal error.75 Moreover, even if we accept Brunt’s explanation of the census figure for 209/208 bc, it does not follow that legionaries serving outside Italy must normally have been excluded from the census. It must be borne in mind that the censorship of 209/208 bc appears to have been the first during which a very substantial proportion of the adult male population was serving outside central and southern Italy. For this reason alone it is unwarranted to assume that the approach taken by the censors of these years was typical of later censorial policies.76 In this context it must also be borne in mind that the years between 209 bc and 207 bc were still a time of military crisis. It can therefore be argued that the line taken by the censors of 209/208 bc was atypical, and that it was the all-encompassing approach adopted by the censors of 204/203 bc that set the norm for the next couple of decades. In considering the question of whether legionaries serving in the provinces were normally omitted from the census, we must also pay attention to the practical aspects of the process of census-taking. As Mommsen noted more than a century ago, it seems extremely improbable that every new census would have been begun entirely from scratch. Instead, each new pair of censors is likely to have used the records assembled by their predecessors as their starting point.77 In my view, this hypothesis is supported by a feature of the census figures for the third and second centuries bc whose significance has never been fully appreciated. A look at the list of census 73 75

76

77

Above, at note 32. 74 Brunt (1971/1987), 63, 68. As argued by Beloch (1886), 349–50. If we want to uphold the census figure for 209/208 bc, Brunt’s explanation is clearly superior to Frank’s suggestion (1924, 330) that Livy’s phrase is ironic. Brunt’s own theory implies that legionaries serving in Italy were omitted in 209/208 bc, whereas they were normally included (cf. Brunt 1971/1987, 68). Mommsen (1887), II, 370, followed by Suolahti (1963a), 33, Nicolet (2001), 723, and Northwood (2008), 265. The privately kept documents referred to by DH 1.74.5 must be the underlying documentation rather than the final census lists.

Target population of the census

97

figures put together by Brunt is enough to show that, if we except the years 209/208 bc, there are no sudden decreases in the number of capita civium registered.78 Conversely, we do observe a number of sudden increases (for instance, in 204/203 bc, in 169/168 bc and in 125/124 bc). This suggests to me that during certain periods a considerable proportion of the target population remained unregistered, but that it was difficult for those who had been registered in one of the more successful censuses to escape registration during the next two or three censuses. The most obvious explanation for this is that, with the possible exception of the censors of 209/208 bc,79 all of the censors of the third and second centuries bc made use of the records compiled by their predecessors. In the case of soldiers serving overseas, we must also consider the potential political side effects of omitting these men from the census records. Can it reasonably be supposed that all men on active service outside Italy were invariably excluded from the census lists and thereby deprived of the opportunity to exercise their political rights (for instance, as voters in the comitia centuriata) for a period of 5 years? It is in my view far more likely that the names of those doing legionary service abroad were copied from the records compiled during the previous census, and that they were assigned to the same class to which they, or their fathers, had belonged during the previous 5 years. It is against this general background that we must interpret the case of the poet Archias, a client of Cicero, who was accused under the lex Papia of 65 bc of having illegally usurped Roman citizenship. In his speech on behalf of Archias, Cicero admitted that his client had been registered neither in the census of 86/85 bc nor in that of 70/69 bc, since he had been abroad first with Sulla and then with Lucullus.80 Various twentieth-century ancient historians have claimed this example as proof that Roman citizens on legionary service in the provinces were not normally registered.81 In fact, 78 79

80

Brunt (1971/1987), 13–14. The census figure for 194/193 bc is generally held to be corrupt. In principle the censors of 209/208 bc could have built on the records compiled by the censors of 220/219 bc, but in the specific circumstances in which they had to operate it must have been extremely difficult for them to find out how many of the men sent out to Spain, Sardinia and Cisalpina were still alive, and military concerns are likely to have over-ridden purely administrative considerations. Another problem must have been that the enormous casualties of the first years of the Hannibalic War had resulted in redistributions of property on an unprecedented scale, making it impossible for the censors to achieve a reliable distribution among the classes. The census of 204/203 bc can be seen as an attempt to make a fresh start after a period of partial administrative breakdown. In my view, this explains why the Roman government took the census of 204/203 bc as its starting point when it ordered all Latins who had illegally been enrolled as Roman citizens between 204 bc and 187 bc to return to their home towns (Livy 39.3.4–5). Cic. Arch. 11. 81 E.g. Jones (1948); Brunt (1971/1987), 36, 41.

98

Meaning of the census figures

the case of Archias does not permit any such inference. As Cicero explains, the local census records of Heraclea (presumably those relating to the period before 86 bc 82) had been destroyed during the Social War, making it impossible for the local (or Roman) authorities to use these records as a basis for the census of 86/85 bc.83 As we have just seen, the reason why Archias had not been registered either in the first pan-Italian census or in the next one was simply that he had been abroad when they were taken. This case undoubtedly proves that, following the destruction of Heraclea’s local records, Archias would have had to present himself either before the census officials of his own town or before those in Rome if he wanted to be registered.84 There is, however, no indication that his absence from Italy would automatically have resulted in his being omitted from the Roman census lists if there had been any record of an earlier census that could be used as a starting point by those responsible for the Roman censuses of 86/85 bc and 70/69 bc. All in all, there appear to be no good grounds for arguing that it was Rome’s standard policy to exclude legionaries abroad from the census. Instead, all the evidence we have is compatible with Mommsen’s view that the censors of the middle and late Republic normally used the records compiled by their predecessors as their starting point, and that this helped them to register a high proportion of the target population. 3.2.4 Assidui and proletarii Another important question concerns the registration of those citizens whose assets fell short of the property qualification for military service. As we have seen, Fabius Pictor states that the legendary first census supposedly held by Servius Tullius registered only those citizens ‘capable of bearing arms’. Although the reliability of this tradition seems doubtful, it might be taken to support the view that in early-republican times those too poor to provide their own arms (that is to say, the proletarians) were excluded from the census.85 There are, however, reliable indications that the censors of the 82 84

85

Cf. Brunt (1971/1987), 42; Lo Cascio (2001b), 595. 83 Cic. Arch. 8. Kron (2005a), 453, n. 69, claims that Cicero’s allusion to the destruction of the Heraclean records proves that copies of such records were not transferred to Rome for incorporation into the Roman census. Since the said destruction took place before 86 bc, however, in this case no copies could have been transferred. The surviving records of the other towns at which Archias was enrolled could not be used to prove his Roman citizenship because his registration as a Roman citizen (during the failed census of 89 bc) had been based exclusively on his enrolment at Heraclea, in conformity with the provisions of the lex Plautia Papiria. Thus Herzog (1877), followed by Gabba (1949/1976), 8.

Target population of the census

99

middle and late Republic did in fact register a significant number of proletarian citizens. The most important clue here is of course the existence of the term capite censi, which is synonymous with proletarii.86 The absolute number of proletarians remains elusive. In his description of the centuriate organization allegedly created by Servius Tullius, Cicero says that the single century of proletarii contained more men than almost the whole of the first class. In two passages also purporting to describe the ‘Servian’ system of classes, Dionysius writes that Servius put the proletarians, who were apparently ‘more numerous’ or at least ‘not less numerous’ than the five property-owning classes combined, into a single century and exempted them from service in the army and from every form of tax.87 These passages have been used as evidence that proletarians made up more than half of the citizen body. In my view, this inference is insecure. As Rich has noted, we can only guess whether the statements of Cicero and Dionysius were true of any historical period. According to him, they were not true of the Second Punic War, when most freeborn citizens (and almost all of the freeborn citizens living outside the city of Rome) seem to have been assidui.88 It is also worth noting that Cicero and Dionysius write as if all adult male Roman citizens turned up at the meetings of the comitia centuriata, and that they take the numerical strength of the various voting groups as accurately reflecting the social composition of the Roman citizen body. It may therefore be speculated that their accounts, or the accounts that they found in their sources, were inspired by the composition of the crowds attending the meetings of the centuriate assembly. We can be certain that the social makeup of these crowds did not in fact mirror the social composition of the citizen body. It is clear, for instance, that large numbers of proletarians lived in Rome. At the same time, there are good grounds to believe that the highest classes contained relatively few citizens and that many rurally based members of the third to fifth classes did not always take the trouble to attend elections or legislative assemblies in Rome.89 Given all this, urban proletarians must

86

87 88

Nicolet (1988), 78, suggests that a distinction was made between capite censi, who had only themselves to declare, and those who also had children to declare, but this theory has been universally rejected. See e.g. Toynbee (1965), I, 455, n. 1; Gabba (1949/1976), 173, n. 23; Rich (1983), 299. Rathbone (1993a) leaves open the possibility that the term capite censi might have been used specifically of those proletarians who had no property to declare, but rejects the theory that there were distinct property thresholds for proletarians and capite censi. Note that Festus 253 L (proletarium capite censum) identifies both terms as describing the same group. Cic. RP 2.40; DH 4.18.2 and 7.59.6. Cf. DH 13.12.2, with Northwood (2008), 268, n. 39. Rich (1983), 294 and 291, n. 21. 89 E.g. Mouritsen (2001).

100

Meaning of the census figures

easily have outnumbered voters belonging to one of the five property classes, even if they did not constitute the majority of the electorate. In the absence of any reliable information that could be used to estimate the number of proletarians, it is difficult to determine what proportion of the proletarians was registered by the censors of the third and second centuries bc. As has often been pointed out, the censuses of the republican period served a threefold purpose: they revealed the number of adult males who could be called up for military service, provided a basis for impositions of tributum and permitted the construction of lists of voters for the use of those responsible for the smooth functioning of the comitia centuriata and the comitia tributa.90 Since proletarians were not normally called up for legionary service and were not expected to pay tributum, it seems reasonable to suppose that most censors of the third and second centuries bc did not attempt to achieve full registration of this particular section of the citizen body. On the other hand, it is possible to identify some factors likely to have pushed up the number of proletarians on the census records. To begin with, at least some sub-groups amongst the proletarians are likely to have displayed an active interest in being registered. It does not seem far-fetched to suppose, for instance, that many urban proletarians might have registered themselves with the censors in order to be able to give their votes to powerful patrons from whom they hoped to secure favours.91 Another factor we must consider is that although proletarians were not normally called up for service in the legions, they were liable for service as rowers in the fleet.92 It is therefore to be expected that the Roman government would have been interested in the number of proletarians available for service not only in times of exceptional danger (such as that of the Gallic invasion), but also in those periods during which large Roman fleets were needed for active warfare or to guard the coasts of Italy and Sicily against the Carthaginians or other enemies. Finally, and perhaps most importantly, it must always be borne in mind that the number of proletarians was determined not only by the distribution of wealth, but also, and primarily, by the threshold for membership of the fifth class of the comitia centuriata. In practical terms, this meant that the 90

91

92

For the practical purposes served by the republican census, see e.g. Mommsen (1887), ii, 388–412; Nicolet (1988), 82–6; Hin (2008), 207–14. The numerous leges de ambitu of the second and first centuries bc (e.g. Nadig 1997) presuppose that financial generosity could buy a large number of votes. For a good discussion of the factors encouraging and discouraging registration, see Northwood (2008), 264, 268. Plb. 6.19.3.

Target population of the census

101

number of proletarians could be reduced simply by lowering the minimum amount of property qualifying a Roman citizen for legionary service. Whenever such a step was taken, any proletarians now artificially transformed into assidui must have become a target group of some interest to the censors. The overall effect would have been a significant decline in the rate of under-registration. In my view, it is actually possible to draw some interesting conclusions about the number of proletarian citizens and about their share in the citizen population as a whole by taking a closer look at the historical evolution of the property qualification for legionary service. As is generally known, this is an extremely difficult topic, not only because our evidence is fragmentary, but also because the few snippets of information that do happen to have survived must be interpreted against the background of the historical development of the Roman monetary system, whose reconstruction is bedevilled by a number of technical difficulties. Fortunately, we can rely on a detailed study by Rathbone that sets out the technical problems with exemplary clarity and offers a comprehensive reconstruction of the financial aspects of the Roman system of classes and of its development during the third and second centuries bc. Building on an earlier article by Lo Cascio,93 Rathbone argues that before the outbreak of the Second Punic War the property qualification for membership of the fifth class (the threshold for access to the legions) must have stood at 1,100 libral asses, corresponding to 5,500 sextantal asses of the period 212 bc–140 bc.94 We also happen to know that by the time at which Polybius wrote the sixth book of his Histories this threshold had been reduced to 4,000 sextantal asses.95 According to Rathbone, the most likely explanation for this discrepancy is that the threshold for legionary service was lowered in about 212 bc. It seems plausible that this change was prompted by the introduction of a new monetary system in which the heavy asses of the pre-Hannibalic period were replaced by asses that were five times lighter. Since the new threshold for membership of the fifth class was 27 per cent lower in terms of silver, it seems reasonable to suppose that the Roman government was also trying to increase its recruiting capacity during a critical phase of the Hannibalic War by transforming a substantial number of proletarians into assidui. 93 94

95

Lo Cascio (1988). Rathbone (1993a), 139–41, followed by Rosenstein (2002), 168–9. Cf. also Lo Cascio (1988), 289–90. Gabba (1949/1976), 5–6, believes that a reduction from 11,000 to 4,000 sextantal asses took place in 214 bc. For further discussion of this topic, see Chapter 4, at notes 142–7. Plb. 6.19.3.

102

Meaning of the census figures

As a second step, Rathbone points out that the settlers sent out to a number of Roman colonies during the first decades of the second century bc received between five and ten iugera (1.25 to 2.5 ha) of land apiece. Since these allotments must have been large enough to make the recipients liable for military service, it can be inferred that the threshold of 4,000 sextantal asses indicated by Polybius corresponded to the value of four or five iugera of arable land.96 If we accept Rathbone’s theory that the threshold for membership of the fifth class stood at 1,100 libral asses before 212 bc, it follows that at the outbreak of the Hannibalic War the ownership of seven iugera was enough to qualify a Roman citizen for legionary service.97 If this reconstruction is correct, the number of proletarian citizens equalled the number of urban proletarians plus the number of rural citizens owning less than 1.75 hectares of land (during the third century bc) or less than 1.25 hectares (during the first half of the second century). Of those citizens belonging to the former category, the vast majority must have lived in the city of Rome. Unfortunately, the surviving evidence does not permit any safe conclusion as to the size of Rome’s population at any point during the second century bc. This explains why the few estimates that have been attempted vary from between 150,000 and 200,000 in around 200 bc to between 200,000 and 400,000 at the time of the Gracchi.98 If we start from the highest figure ever suggested, reckon the servile population at 25 per cent of the total urban population and assume that adult male proletarians made up one-third of the free population, we obtain a rough estimate of c. 100,000 urban proletarians for the final decades of the second century bc.99 With reference to rural citizens, Rosenstein has pointed out that precisely because the property qualification for military service was so low, large numbers of poor citizens are likely to have benefited from the colonial schemes of the late fourth and third centuries bc and from the viritane distributions carried out by figures such as Dentatus and Flaminius. In his view, the proportion of the adult male citizen population qualifying for 96 97

98 99

Rathbone (1993a), 145, followed by Rosenstein (2002), 167–9. It is, of course, tempting to speculate that the numerous accounts of viritane settlers receiving seven iugera (e.g. Val. Max. 4.3.5b; Vir. Ill. 33; Plin. NH 18.4.18), along with the tradition that this was the size of the farm owned by the Roman general Atilius Regulus (Val. Max. 4.4.6), reflect the existence of this hypothetical threshold. See De Ligt (2004), 742 for further discussion and references. If the population of Rome was characterized by a skewed sex ratio (as suggested by Beloch 1886, 400–1), adult male citizens would have made up a larger share of the free population, but of course not all urban citizens were proletarians.

Target population of the census

103

service in the legions is likely to have been closer to 90 than 50 per cent throughout most of the third century bc.100 As we shall see in the next chapter, the large-scale viritane assignations carried out in the years 200–199 bc and 173 bc, along with the foundation of a number of Latin colonies and coloniae civium Romanorum between 194 bc and 177 bc, must have raised many former proletarians to the status of assiduus and must also have helped recipients who were already of that status (or whose fathers were already of that status) to maintain it. If the minimum amount of private land required for legionary service had indeed been reset at five iugera by this period, the proportion of the adult male population belonging to the five classes must have been even higher than it had been before the Hannibalic War. In my view, these interlocking developments must be part of the key to an enigmatic passage in Polybius in which it is claimed that in around 160 bc the Romans were no longer able to man as many ships as they had during the First Punic War.101 As is well known, Polybius says that he will explain this in his account of the Roman constitution, but no such explanation has ever been identified. Polybius’ account of the First Punic War suggests that he is speaking of fleets made up of between 330 and 350 ships. A fleet of this size would have required at least 100,000 men.102 Just before making his surprising comment on Rome’s reduced capacity to assemble large fleets, Polybius notes that some 700 Roman quinqueremes were lost between 264 bc and 241 bc.103 Let us assume that this is the number of ships he has in mind and that some 200,000 rowers were required to man them.104 Even if the majority of rowers came from the socii navales and other allied communities, the number of rowers of citizen status cannot have been much lower than 80,000.105 As we have just seen, Polybius appears never to have returned to the subject of the decreased number of rowers that must lie behind his surprising observation of Book 1. He does, however, explicitly state that Roman 100

101 102

103 105

Rosenstein (2002), 166, 177. As he points out, the surviving sources strongly suggest that the largescale distributions of land carried out after the Roman conquests of the late fourth and early third centuries bc ended the Struggle of the Orders. For the number of assidui at the time of the Second Punic war, cf. above, at note 88. Plb. 1.64.1–2. Plb. 1.25.7; 1.36.10. In 1.26.7 Polybius assigns 140,000 men to a fleet of 330 ships, but his estimate may be on the high side. Plb. 1.63.6. 104 If we opt for a lower figure, this only strengthens my arguments. The proportion of citizens implied by this guesstimate is much lower than the 55 per cent assumed by Brunt (1971/1987), 65. Cf. ibid. 669.

104

Meaning of the census figures

proletarians could be called up for service in the fleets.106 If we combine this statement with his claim that Rome was no longer able to man fleets of the same size as those sent out during the First Punic War, the inescapable conclusion is that at this time the number of registered proletarians was considerably lower than, say, 80,000. Polybius’ statement is often interpreted as suggesting that he believed the number of adult male Roman citizens to have declined.107 As we have just seen, however, the shortage of proletarians eligible for service in the fleet to which he alludes may simply have reflected the fact that the reduction of the threshold for membership of the fifth class and the land distributions carried out between 200 bc and 173 bc had transformed large numbers of proletarians into assidui. If this is the case, the only demographical inference that can be drawn here is that proletarians cannot have made up more than half of the registered census population in the mid-second century bc. Of course, this leaves open the possibility that large numbers of proletarians remained unregistered. In my view, the only way to come to grips with this problem is to ask how many landless proletarians and poor peasants owning less than one or one and a quarter hectares of arable land could reasonably have been accommodated within the agrarian economy of central-western Italy during the second century bc. Since estimates of the total number of adult male citizens living in Italy in the mid-second century bc range from 400,000 in the case of the low count to 700,000 in the case of the high count, the traditional view that proletarian citizens made up more than half of the Roman citizen population necessarily implies that between 100,000 and 250,000 adult male proletarians made a living by working other people’s land as wage labourers or as tenants.108 Since wage labour in the agrarian economy of republican Italy was almost exclusively seasonal, the former scenario seems unrealistic.109 The idea that tenancy was already a phenomenon of some importance in

106 107

108

109

Plb. 6.19.3. E.g. Brunt (1971/1987), 88–9 (admittedly arguing against a scenario of fast growth); Crawford (1978), 102. On the assumption that some 100,000 adult male proletarians can be assigned to the city of Rome. Rosenstein (2004), 77–8, suggests that many rural citizens may have had access to plots of ager publicus. As Rathbone (2003) and Roselaar (2010), 206–7, have pointed out, however, there is absolutely nothing to suggest that ager publicus was readily available to poor peasants of citizen status, especially in central-western Italy, where most citizens lived. De Ligt (2007b), 7–8.

Target population of the census

105

the second century bc is less problematic.110 Nonetheless, it is difficult to believe that a large proportion of the citizen population derived most of its sustenance from leaseholdings at this early date. As we have just seen, this problem is all the more acute because even during the first half of the second century bc many assidui appear to have owned tiny plots of land that cannot possibly have covered their families’ subsistence needs. This means that, even without the very large number of poor rural citizens postulated by the high counters, a large number of poor citizens must be fitted into our reconstruction of the agrarian economy of the second century bc. Another important argument against the existence of a very large number of rural proletarians is the rapid expansion of rural slavery during the last two centuries bc. As we shall see in the next chapter, the most convincing explanation for the brisk growth of rural slavery after the Second Punic War is that during the first decades of the second century the free population was thin on the ground. In other words, the setting up of slave-based estates for the production of wine, olive oil and grain can be seen as a rational economic strategy in a period during which free labourers were relatively scarce and therefore expensive.111 It should be noted that this is precisely the demographic and economic situation envisaged by the low count. The importation of many tens of thousands of foreign slaves cannot easily be explained if a very large proportion of the Roman citizen population was made up of proletarians owning fewer than four or five iugera of land. Before we round off this brief discussion, it must be conceded that it remains extremely difficult to assess the strengths or weaknesses of a theory positing the existence of several hundreds of thousands of people who were, by definition, invisible. Nonetheless, it seems possible to conclude that the demographic impact of the Second Punic War, the large-scale land distributions carried out between 201 bc and 173 bc and the rapid expansion of rural slavery sit very uneasily with the theory that even after the threshold for membership of the fifth class had been lowered to 4,000 sextantal asses more than half of the Roman citizen body was made up of urban and rural proletarians, most of whom remained unregistered by the censors of the second century bc. 110

111

Rosenstein (2004), 181–2, suggests that there were almost no tenants before the first century bc, but his arguments are weak. In my view, we must begin from the assumption that tenancy was always an option (De Ligt 2000) and then try to explain why this option was used more often in some historical periods than in others. Unlike Rosenstein, I do not think that a preference for social and economic independence can explain the failure of tenancy to take off in the second century bc. Cf. Chapter 4, at notes 78–82. Chapter 4, section 4.2.4.

106

Meaning of the census figures 3.3 centralized and decentralized census procedures before the social war

It has often been suspected that the original aim of the republican census was to establish Rome’s military potential and that this aim was originally achieved by ordering all men aged between 17 and 60 to assemble in the campus Martius, whereupon the number of ‘those able to bear arms’ would be established by means of a simple head count.112 If this hypothesis is correct, the lustratio of the populus with which the census was still concluded in historical times must originally have had a military connotation. Since these inferences are based on interpretations of certain features of the census as it existed during the middle and late Republic, they are necessarily speculative. Despite this, it seems likely enough that the census in its original form required the personal attendance of those required to register. There are several indications that the registration in absentia of certain categories of citizen was permitted at a later stage. It is, for example, generally accepted that in historical times a son-in-power was normally registered by his pater familias.113 It seems reasonable to infer from this that such sons did not have to present themselves before the censors. From the limited perspective of Italy’s demographic history, the most important question is whether the basic rule of personal attendance was maintained after the Ager Romanus had begun to expand beyond centralwestern Italy. If it could be proved that citizens living in the peripheral districts of the Ager Romanus could be registered locally by magistrates who were then expected to send the results of their local censuses to Rome, it would follow that the republican census figures, for all their obvious shortcomings, cannot have been hopelessly inaccurate. On the other hand, if it could be demonstrated that all those under the obligation to register themselves (and their sons in potestate) had to present themselves before the censors in Rome, it would become easier to posit a deficiency rate of the order of 50 per cent.114 According to this view, the census figures for the 112

113 114

According to Varro, LL 6.86, the republican censors summoned omnes Quirites pedites armatos to attend the census. Gabba (1949/1976), 8, suggests that originally only assidui were taken into consideration. Above, at note 5. Note, however, that this issue is intimately linked to the number of proletarians. According to the high counters, many of those who did not travel to Rome are likely to have been citizens whose assets fell short of the (low) property qualification for legionary service. Lo Cascio (1999a), 167–8, and (1999b), 232–3, claims that such people might even have been formally released from the obligation to register themselves with the censors.

Centralized and decentralized census procedures

107

third and second centuries bc can be used only to establish the minimum number of adult male citizens alive at the time of any specific census.115 Unfortunately, we have almost no evidence that might help us to decide which of these conflicting theories is correct. One of the few certainties in this field is that the citizens of the maritime colonies of the fourth and third centuries bc were excused from legionary service because they were supposed to guard strategically important points on the coast.116 It is generally accepted that they must therefore also have been exempt from the customary obligation to appear before the censors in Rome.117 Brunt has argued that this arrangement must have been extended to the non-maritime citizen colonies founded in northern Italy during the 180s bc, and subsequently to provincial coloniae civium Romanorum such as Narbo Martius in Gaul. In fact, we are told that Gaius Gracchus’ plan to found a colony of Roman citizens on the site of Carthage was opposed on the grounds that this initiative would constitute a break with the timehonoured policy of requiring all Roman citizens to travel to Italy at the time of the census.118 With the self-governing municipia, a category that contained both communities of cives optimo iure and communities of citizens without the vote, we move into far more difficult territory. Imperial-era censores are recorded in some former communities sine suffragio (such as Caere), and from this fact it has plausibly been inferred that these communities must have conducted local censuses before the Social War.119 This inference is accepted by Lo Cascio and by some of his supporters. In their view, however, these censuses might have been purely local affairs that had nothing to do with the Roman census. This view is in line with Lo Cascio’s theory that a centralized census procedure was maintained until 45 bc or 44 bc and also with his contention that the census figures for the third and second centuries bc do not include cives sine suffragio. Against the latter theory, I have already pointed out that the hypothetical population densities that it implies are implausibly high. For this reason alone, it seems preferable to adopt the view that the censores of Caere and other municipia were expected to send the results of their local censuses to the censors in Rome, who then added 115 118

119

Lo Cascio (2008). 116 Livy 26.3.38. 117 Brunt (1971/1987), 40; Kron (2005a), 452. Brunt (1971/1987), 40; Galsterer (2006), 300. Vell. 2.7.7: civis Romanos ad censendum ex provinciis in Italiam revocaverant. The explicit reference to Italy, not Rome, seems significant. Wiseman’s suggestion (1969, 60, n. 18) that the phrase in Italiam is an anachronism is clearly inspired by the preconceived idea that there was no decentralized census under the Republic. E.g. Brunt (1971/1987), 40–1; Ilari (1974), 81–2; Moatti (1993), 73–4. In those communities accorded citizenship in or after 90 bc, the standard term for local censorial magistrates was iiiiviri quinquennales. See Brunt, ibid.; Bispham (2007), 337–64.

108

Meaning of the census figures

them to the results of their own census. As we have seen, this would explain the presence of the tabulae Caeritum in Rome. A similar discussion centres on the municipia of cives optimo iure. In this case, too, the existence of local magistrates responsible for census operations is widely agreed upon.120 Again the high counters have argued that the results of these local censuses cannot have been transmitted to Rome because the Roman census was fully centralized. Since all of the municipia optimo iure that existed before the Social War were in central-western Italy, it would perhaps not have been unreasonable to require all Roman citizens belonging to these communities to travel to Rome once every 5 years. The force of this argument is, however, surely undermined by the fact that no such obligation was imposed on the cives sine suffragio of Caere. It seems, moreover, most unlikely that Rome would have deprived the new municipia created after 90 bc of the local administrative machinery that had been in operation for centuries. As has often been observed, the new municipalities continued to perform the function of finding soldiers for the Roman army.121 It would seem to follow that they must have had lists of those qualified to serve. Such lists can only have been created by means of local censuses, and it remains hard to understand why men who had already been registered locally should still have been required to travel to Rome for a second registration.122 Finally, a few words must be said about the praefecturae established in areas to which viritane settlers had been sent. We know that there were praefecturae in Sabinum, in Picenum and in the territories of the Vestini, and they must surely also have existed in other areas subject to large-scale assignationes viritanae, such as Liguria, southern Samnium and Apulia.123 If a wholly centralized census procedure had indeed been maintained until the time of Caesar, Roman citizens living in these areas would have been obliged to travel hundreds of kilometres to present themselves before the censors in Rome. Such an arrangement would not have been very practical. At the same time, viritane settlers and their descendants must have been a target group of some interest to the central Roman authorities, if only because a high proportion of them would have been assidui. Under these circumstances, it would have made sense to create some kind of local or 120 122

123

E.g. Kron (2005a), 452, n. 60. 121 Brunt (1971/1987), 41. According to Kron (2005a), 450–3, this requirement would have served to prevent the vast majority of the former allies from exercising their political rights as citizens, but in my view any policy of exclusion (if one was ever implemented) must have run out of steam by the time of the census of 70/69 bc. Cf. below, at notes 143 and 157. For traces of praefecturae in Apulia, see Grelle (1999a) and (2009).

Centralized and decentralized census procedures

109

regional census procedure. An explicit reference to such a procedure, run by the highest-ranking magistrates of the praefecturae, may be found in the Tabula Heracleensis.124 Since we have, however, no evidence whatsoever for the involvement of prefects or other magistrates in local or regional censuses at any earlier date, we cannot categorically rule out the possibility that before the time of Caesar Roman citizens who lived in areas not administered from self-governing coloniae or municipia were required to travel to Rome for the census.125 Conversely, it is equally impossible to rule out the possibility that the system described in the Tabula was rooted in much earlier practices.126 Since the ancient sources have little to say about such mundane topics as the practical aspects of the census, it is extremely difficult to back up any assertion concerning the existence of a centralized or decentralized census procedure with textual evidence. The most significant exception to this rule is Livy’s description of the measures taken in 169 bc to improve the efficiency of the levy.127 As Livy explains, in this year there was great concern about the levy for the Macedonian campaign because even the iuniores were not answering the usual call-up. The task of remedying this problem was entrusted to the praetors, assisted by the censors elected for the census of 169/168 bc. These censors added to the customary oath taken by all citizens a special oath for the iuniores in which they were required to affirm that they were under 46 years of age, that they had presented themselves for enlistment in accordance with the edict issued by the censors and that they would continue to do so for the next five years. The censors also issued an edict concerning soldiers who had been enlisted for Macedonia in or after 172 bc. Any of these men who were in Italy (qui eorum in Italia essent) were ordered to return to Macedonia within thirty days, after first appearing for assessment before the censors themselves (censi apud sese), while anyone who was alieni iuris was required to be registered by his pater familias. The censors also announced that they would personally investigate the reasons for any discharge that might have been given to any of these soldiers, and that they would order the re-enlistment of anyone who had been discharged without valid grounds. 124 125

126

127

Tab. Heracl. 142–6. As noted by Crawford and Nicolet in Crawford (1996), 388, this uncertainty also exists in the case of the fora and conciliabula. In favour of early decentralization: Giovannini (2008), 61. According to Humbert (1978), 313–17, praefecti iure dicundo must have been responsible for regional censuses as early as the third or second century bc, but as Bispham (2007), 96, n. 108, points out, we have no evidence that praefecti performed any non-jurisdictive administrative task. Certainty here appears impossible. Livy 43.14.2–10.

110

Meaning of the census figures

When this edict, accompanied by letters, had been sent to all market centres and places of assembly (per fora et conciliabula), a large number of iuniores assembled in Rome, so that the unusual crowd (insolita turba) caused a great deal of inconvenience.128 This unique account contains some interesting details. It seems significant that there is no indication whatsoever that the new oath imposed on iuniores had to be sworn in Rome.129 Livy’s account is therefore compatible with the view that large numbers of iuniores were normally registered by local magistrates, and that it was before these magistrates that they were expected to swear the new oath formulated by the censors. Nevertheless, Livy’s text suggests that a different procedure was followed in the case of soldiers who were on leave and those claiming to have been discharged, since the edict concerning these men clearly ordered them (or their patres) to come to Rome. We are also told that this second measure applied to any soldiers from the Macedonian legions who were currently in Italy. It is therefore tempting to interpret this as a general measure affecting the entire Ager Romanus. One small detail, however, seems to militate against this interpretation. As various scholars have observed, Livy reports that the second edict was circulated per fora et conciliabula, rather than per totam Italiam. The significance of this lies in the fact that Livy and other sources often use the expression fora et conciliabula as a technical term to denote those parts of the Ager Romanus that were not administered by self-governing coloniae or municipia.130 From this terminological detail it has been inferred that the second edict affected only those parts of the Ager Romanus for which the censors were directly responsible (in other words, the fora et conciliabula). It has also been pointed out that Livy’s account of the measures taken by the central Roman government in 212 bc to stamp out draft-dodging suggests a similar geographical division of labour between central and local magistrates. On this occasion, two triumviral boards were instituted to inspect the entire range of free men in pagis forisque et conciliabulis. Again, the municipia and coloniae are missing.131 In my view, this makes it difficult to argue that Livy’s consistent failure to mention these communities in accounts

128 129

130 131

Livy 43.14.8–10. Lo Cascio’s claim that Rome was the only place where this oath could be sworn (Lo Cascio 2008, 252) has no support in the text. Beloch (1880), 104–5; Galsterer (1976), 25–9; Bispham (2007), 93. Livy 25.5.6. Cf. Galsterer (1976), 106; Humbert (1978), 323–4; Crawford and Nicolet, in Crawford (1996), 388; Bispham (2007), 89.

Centralized and decentralized census procedures

111

concerning measures dealing with recruits and legionaries is attributable to the repeated use of inaccurate formulations.132 As noted above, the only suggestion that the second edict affected the entire Ager Romanus is offered by the phrase qui eorum in Italia essent. As Galsterer demonstrated more than 30 years ago, however, there is good reason to believe that during the second century bc the term Italia was sometimes used to denote the non-municipalized area immediately around the city.133 One interesting example of this usage may be found in Livy’s account of the activities of the censors of 204 bc. According to Livy, ‘In Rome and throughout Italy (per totam Italiam) salt was selling for a sixth of an as. The censors put out contracts for it now to be sold at the same price as before in Rome, at a higher price in the fora and conciliabula, with prices varying from place to place.’134 In this text, the area denoted by the term Italia is clearly identical with the fora et conciliabula. Another instance is to be found in Livy 40.19 and 40.36–7: when pestilence raged in the fora et conciliabula, the Roman authorities ordered a supplicatio to be held per totam Italiam.135 This brief terminological survey demonstrates that Livy’s use of the term Italia cannot be used to set aside the reference to the fora et conciliabula. Even if anyone should want to accord a broader meaning to the term Italia in the specific context of the measures taken in 169 bc, several parts of Livy’s text would continue to militate against the conclusion that it was normal for all adult male citizens to be ordered to travel to Rome for the census. It could be argued, for instance, that Livy’s use of a number of phrases emphasizing the personal involvement of the censors (in particular, censi apud sese) indicates a departure from standard administrative procedures. It might be speculated that the censors took the unusual step of ordering all soldiers on leave and all those who had been discharged from the legions in Macedonia to come to Rome because they wanted to make it entirely impossible for any of those enlisted for service in Macedonia in or after 172 bc to escape their military duties. 132 133

134

Cf. Crawford (1996), 388. Galsterer (1976), 37–41. Note Galsterer’s suggestion that this usage lies behind Appian’s claim that the Gracchan land reforms were meant to stem the numerical decline of the Italiôtai. Galsterer acknowledges that in some other passages Livy uses the expression Italia to denote the entire Ager Romanus. More recently, Bispham (2007), 57–68, has convincingly demonstrated that from the second or third quarter of the third century bc onwards official documents began to use this term to refer to Italy as far as the Arno-Rubicon line and occasionally as far as the Alps. The epigraphic lex agraria of 111 bc (Crawford 1996, 113–80) is an obvious example. The only possible conclusion is that the meaning of the term Italia was ambiguous during this period. Livy 29.37.3. 135 On these texts, see Bispham (2007), 89 and n. 81.

112

Meaning of the census figures

One advantage of such a reading is that it helps to explain Livy’s reference to the inconvenience caused by the unusual crowd that assembled in Rome as a result of the censorial edict. Regardless of how we prefer to interpret the term Italia, this detail strongly suggests that ordering all legionaries and ex-legionaries from Macedonia who were in Italy to come to Rome was abnormal. A fortiori, we must conclude that it was unusual for all adult male citizens to be ordered to come to Rome for the census, since such an order would have resulted in crowds many times larger than those of 169 bc. Even if we should decide to set aside Livy’s reference to the fora et conciliabula, then, his account of the levy of 169 bc continues to suggest that the census procedures of this time were no longer fully centralized.136 3 . 4 t h e c e n s u s f i g u r e f o r 8 6 / 8 5 bc During the past century, those interested in the workings of the republican census during the second and early first centuries bc have put a great deal of intellectual energy into attempting to interpret the census figures for 86/85 bc and 70/69 bc. As is generally known, the first of these figures relates to the first census taken after the end of the Social War, during which Roman citizenship, initially offered only to those allies who had remained loyal or had laid down their arms, had eventually been extended to all of the Italian communities. The last two groups to receive citizenship seem to have been the Samnites and the Lucanians, who are thought to have received it in 87 bc.137 It follows from this that, if the censors of 86/85 bc were attempting to register every citizen they could, the results of their operations should shed some light on the size of the free population of mainland Italy south of the Po. It is extremely regrettable that Livy’s epitomator does not record the census figure for 86/85 bc. The only figure we have is the one that appears in Jerome’s Chronicon, in which we read that the census tally for 85 bc was 463,000.138 Since Livy’s epitomator reports census figures of c. 395,000 for 136

137 138

In my view, the time limit of 30 days cannot explain why a crowd of ‘unusual’ size should have appeared in 169 bc (contra Kron 2005a, 452, n. 59). Since the second edict applied only to legionaries and ex-legionaries from Macedonia, it would have affected no more than a few thousand men. Even if we take into account the fact that many men would have come to Rome for the ordinary dilectus, the unusual nature of the gathering described by Livy remains unexplained. According to the high count, there were about 475,000 adult male citizens as early as 225 bc. With a hypothetical registration rate of 55 per cent (Chapter 4, at notes 149–50), this would imply that some 260,000 adult male citizens had to come to Rome within the 18 months during which the censors held office, an average of about 14,450 per month. E.g. Bispham (2007), 187. Hier. Chron. p. 151, Helm (1926): descriptione Romae facta inventa sunt hominum cccclxiii milia.

The census figure for 86/85 bc

113

the years 125/124 bc and 115/114 bc, when citizenship had not yet been extended to the allies, this tally seems impossibly low. It is also odd that Livy’s epitomator should inform us that the Senate bestowed the suffragium (that is, the right to vote) on the new citizens in 84 bc.139 The obvious problem created by this statement is that one would expect the new citizens to have been given the right to vote and to have been registered as belonging to tribes before the start of the census of 86 bc. Since the late nineteenth century, various solutions to these anomalies have been proposed. The simplest of these is that Jerome’s figure is corrupt and should be emended to 963,000.140 One argument in favour of this solution is that many of the other census figures reported by Eusebius and Jerome, or by later chroniclers drawing on them, are demonstrably incorrect.141 There is therefore a real possibility that the same is true of the figure for 86/85 bc. On the other hand, it seems methodologically unwarrantable to reject a census figure just because it is anomalous, without having explored the viability of other possible explanations. A fortiori, we should resist the temptation to get rid of any problematic figures by emending them in such a way as to make them compatible with some preconceived theory about the numerical development of the Roman or Italian population, or the efficiency of the census. For these reasons, I suggest that we should take the figure for 86/85 bc seriously, without losing sight of the possibility that we may be dealing with a scribal error. Perhaps predictably, many adherents of the high count have claimed that the low number of capita civium registered by the censors of 86/85 bc can be fully explained if we assume that the Roman census continued to be fully centralized until the mid-40s bc. In their view, a variety of factors such as the persistence of parochialism, the negligible influence of the majority of the population in the comitia centuriata, the prospect of loss of work and the difficulties and costs of overland travel can be expected to have dissuaded about two-thirds of the theoretical target population from travelling to 139 140

141

Per. Livy 84. Beloch (1886), 352; Brunt (1971/1987), 92. But then, why not 863,000 (with dccc misread as cccc ), as suggested by Bennett (1923), 44–5? The figure of 463,000 is accepted by Frank (1924), 336, Taylor (1960), 105–6, Galsterer (2006), 299–300 and Bispham (2007), 198, n. 185. In the manuscripts of Jerome’s Chronicon, the census figure for ad 14 is distorted to 9,370,000 (cf. below, at note 173) and that for ad 47/48 (5,984,072, according to Tac. Ann. 11.25) to 6,944,000 in the Latin text and to 6,941,000 in the Armenian version (Hier. Chron. p. 180, Helm (1926) and in Syncellus (p. 404, Mosshammer). As Beloch (1886), 352, points out, most manuscripts containing the text of the Periochae of Livy give ccccl (instead of dcccc ) as the census figure for 70/69 bc, evidently because the copyist overlooked or misread the initial d . Of course, Eusebius/Jerome may faithfully have copied a figure that had already been distorted.

114

Meaning of the census figures

Rome, if this was the only place where adult male citizens could register themselves.142 It has, in fact, even been suggested that a fully centralized census procedure requiring all household heads to come to Rome was deliberately kept in place after the conclusion of the Social War in order to prevent as many as possible of the new citizens from exercising their newly won political rights.143 The main weakness of these arguments is that they assume what has yet to be proved, namely the existence of a fully centralized census procedure, without looking for alternative explanations. Focusing on the enigmatic reference to the bestowal of the suffragium in 84 bc in the text of Livy’s epitomator, other ancient historians have suggested that the various grants of citizenship laid down in the lex Iulia of 90 bc and in the later legislation dealing with the dediticii populi were not immediately implemented, in the sense that it took some years of political wrangling to sort out the distribution of the new citizens among the thirtyfive tribes. In the view of these historians, the information supplied by Livy’s epitomator implies that it was only in 84 bc that the majority of Italians were given the opportunity to exercise their political rights as Roman citizens.144 This assumption allows it to be argued that it was a delay between the various grants of citizenship made in abstracto and the implementation of these grants that was responsible for the omission of the vast majority of the new citizens by the censors of 86/85 bc. Against this interpretation it has been pointed out that, following Sulla’s march on Rome and his departure for Asia, Marius and Cinna had formed an alliance with the Samnites and recaptured the city with the help of troops and money supplied by the Italian allies.145 It seems likely that the allied communities in question were won over by the promise that, once in power, Marius and Cinna would rescind Sulla’s annulment of the lex Sulpicia of 88 bc (which had prescribed the enrolment of the new citizens in the old 142

143

144

In favour of a fully centralized census: Taylor (1960), 105; Wiseman (1969), 67–9; Lo Cascio (1994), (2001b), 587, and (2008); Kron (2005a), 450–3; Bispham (2007), 360, n. 141. Against: Mommsen (1887), ii, 368–9; Brunt (1971/1987), 38–43; Humbert (1978), 322–5; Crawford and Nicolet, in Crawford (1996), 388–9; Mouritsen (1998a), 96, n. 29; Northwood (2008), 263. Kron (2005a), 450–3. According to Kron, ibid., 448, Lepidus’ claim that Sulla was barring a considerable number of the allies and of the people of Latium from citizenship (Sall. Hist. 1.55.12 = 1.48.12 McGushin) must mean that no census was held in order to prevent the allies from exercising their newly won political rights. This would certainly be the most natural reading if we could be sure that the census of 86/85 bc was a failure. But perhaps Lepidus’ words are to be understood as referring to the fate of those communities that Sulla had deprived of citizenship (as suggested by McGushin 1992, 118–19), which may have included Praeneste (cf. Santangelo 2007a, 137–42). Alternatively, the passage can be read as referring to the large number of prominent men who had individually lost their rights as citizens (cf. App. BC 1.96). In any case, this passage has no bearing on the interpretation of the census figure for 70/69 bc. E.g Harris (1971), 233–6; Mouritsen (1998a), 168, n. 45; Lovano (2002), 62. 145 App. BC 1.65–8.

The census figure for 86/85 bc

115

tribes).146 From this chain of events, it has plausibly been inferred that it would have been politically impossible for Marius and Cinna to renege on their earlier promise and to exclude the majority of Italians from the census by postponing their registration in the tribes.147 It would then follow that the new citizens must already have been distributed between the tribes before the census of 86/85 bc, and that the grant of suffragium made in 84 bc must have been no more than a senatorial decree confirming the right of the new citizens to vote amongst the thirty-five tribes.148 Although these objections must be taken seriously, it would be wrong to conclude that no satisfactory explanation for the low census figure reported by Jerome can be found unless we accept the existence of a fully centralized census procedure. In my view, there is absolutely nothing to contradict the view that the censors of 86/85 bc had asked all or most of the self-governing communities between the Po and the Ionian Sea to hold local censuses and to send the results to Rome. If such local censuses were indeed held, it seems reasonable to suppose that all municipal lists would have had to be verified, either by the censors themselves or by other magistrates assisting them.149 Since this would have been the first pan-Italian census, a whole range of practical problems could have arisen at this point. For instance, the censors or their assistants might well have rejected the results of local censuses on the grounds of inaccuracy, deviation from normal Roman procedures or suspected invalid grants of citizenship to foreigners. This hypothesis is particularly plausible in the case of the census of 86/85 bc because one of the censors, L. Marcius Philippus, is known vehemently to have opposed Livius Drusus’ proposed extension of citizenship to the Italian allies.150 It could therefore be argued that the figure for 86/85 bc was low because most of the locally compiled census records had been rejected and that the grant of suffragium was a follow-up measure that validated these local lists (or improved versions of them) as entitling the men on them to cast their votes at Roman assemblies. Since the censors of 86/85 bc would already have assigned the 463,000 citizens registered on local lists to the classes, the 146 147 149

150

Law of Sulpicius: App. BC 1.55–6; annulment by Sulla: BC 1.59. Taylor (1960), 104–5; Bispham (2007), 178–80, 193–4. 148 Bispham (2007), 194. It appears from Cic. Arch. 9–10 that the censors of 89 bc had entrusted three of the praetors with the task of verifying the registration of people who had received honorary citizenship in more than one Italian town, the so-called adscripti. Wulff Alonso (2002), 150–3, argues that these praetors must also have been responsible for the verification of all municipal lists sent to Rome. In his view, some of the praetors of 86 bc and 85 bc might also have performed this task. Although this theory is speculative, the underlying idea that local census lists would not have been incorporated into the Roman census without prior verification seems plausible. Brunt (1971/1987), 93; Bispham (2007), 195.

116

Meaning of the census figures

practical outcome might have been that until the census of 70/69 bc the vast majority of the new citizens could vote in the comitia tributa but not in the comitia centuriata.151 Faced with this plethora of possible interpretations, we can only say that the census figure for 86/85 bc is entirely compatible with the existence of a centralized census procedure, but equally so with the theory that the censors of these years were facing a whole range of problems resulting from the introduction of a fully decentralized system of registration across the entire Italian peninsula. At the same time, the poor quality of many of the census figures given by Jerome makes it impossible to know for certain whether or not his low figure for 86/85 bc can be relied upon. For these reasons, the best solution is perhaps to eliminate the figure for 86/85 bc from all discussions concerning the practicalities of the republican census. 3.5 some other pieces of evidence Apart from the figure in Jerome’s Chronicon, our only snippet of evidence concerning the census procedures followed in the years after the end of the Social War is a passage from Cicero’s speech on behalf of Cluentius. In this passage Oppianicus, who offered his services to Sulla in 83 bc, then installed himself and three of his friends as quattuorviri of his home town after Sulla’s victory, is accused of having falsified the public census records of Larinum (tabulas publicas Larini censorias).152 Lo Cascio argues that the census referred to in this passage must have been a local one conducted by the town of Larinum in 82 bc, when Oppianicus could have been performing censorial duties in the role of a quattuorvir quinquennalis.153 The difficulty here is that there is absolutely nothing to suggest that the magistracy held by Oppianicus in 82 bc or 81 bc was that of a quinquennial quattuorvir; we have, moreover, no evidence that any local census was taken in Larinum during his term of office. It seems, therefore, more plausible to suppose that the tabulae censoriae to which Cicero refers were local tabulae compiled for the panItalian census of 86/85 bc.

151 152 153

Taylor (1960), 105–6, followed by Bispham (2007), 198, n. 185. Cic. Clu. 41, discussed in detail by Moreau (1994). For the quattuorviri quinquennales of late-republican and early-imperial Italy and their censorial duties, see Bispham (2007), 337–79. There are no certain attestations before the 60s bc (ibid. 374). Note that Bispham, ibid., 200–1, dates Oppianicus’ quattuorvirate to 81 bc, the year after the Battle of the Colline Gate.

Some other pieces of evidence

117

If the system described by the Tabula Heracleensis was already in operation by this time, the results of any local census would have had to be sent to Rome, but the Tabula makes it clear that the information collected on the local tabulae censoriae was copied onto libri (presumably in the form of papyrus rolls) and that only these documents were transmitted.154 This implies that the municipal authorities of Larinum and indeed of every other town would have retained the original records for their own use (for instance, in regulating access to the local town council, or determining liability to conscription). Since Lo Cascio is thinking in terms of a fully centralized census procedure, he cannot date the tabulae censoriae of Larinum to 86 bc or 85 bc. He therefore has to assume that the local authorities of Larinum had no access to the results of the pan-Italian census of 86/85 bc, and that this made it necessary for them to organize a completely new census 3 or 4 years after the completion of the lustrum of 85 bc. There can be no doubt that this would have been an extremely impractical arrangement. Besides, as we have just seen, there is no evidence whatsoever that any local census took place in Larinum in 82 bc or 81 bc. At the very least, we must conclude that the evidence contained in the Pro Cluentio is fully compatible with the existence of a decentralized census procedure. Regarding the second pan-Italian census (that of 70/69 bc), the only clue of any importance is a famous passage from Cicero’s first speech against Verres, from which it appears that a large crowd came to Rome in 70 bc ‘for the elections, for the games and for the census’ (comitiorum ludorum censendique causa).155 Commencing with Tenney Frank, many scholars have claimed this passage as proof that any Roman citizen who wanted to register himself had to travel to Rome.156 As Crawford and Nicolet pointed out some years ago, however, the crowd referred to by Cicero as having come for the census could well have consisted of representatives of the Italian municipia who had had to travel to Rome in order to present the results of their local censuses.157 It is true that these representatives alone are not likely to have formed a huge crowd.158 As Cicero explains, however, the visitors in question included a lot of people who had come to Rome for the games or for the elections. In other words, the only piece of evidence 154 156 157

158

Tab. Her. 148–50. 155 Cic. I Verr. 15.54. Frank (1924), 334; Wiseman (1969), 68–9; Lo Cascio (2001b), 596–7; Kron (2005a), 452–3. Crawford and Nicolet, in Crawford (1996), 389. Pace Bispham (2007), 360, n. 141, Cicero neither says nor suggests that most or all of those who had travelled to Rome had done so in order to register themselves with the censors. Thus, correctly, Kron (2005a), 453.

118

Meaning of the census figures

that has been claimed as incontrovertible proof of the existence of a fully centralized census procedure cannot bear the heavy burden placed upon it by the high counters. One indication suggests that the censors of 70/69 bc did in fact manage to register a very large proportion of the target population. This indication has to do with the demographic balance between citizens and allies before and after the Social War. If we are to believe Livy’s epitomator, roughly 395,000 adult male citizens were registered in 125/124 bc and in 115/114 bc.159 Under normal circumstances the corresponding figure for 70/69 bc would presumably have been considerably higher than this. It must, however, be remembered that between 110 bc and 80 bc some 85,000 men of citizen status had perished during the Cimbrian Wars, the Social War and the Civil War of the late 80s bc. It therefore seems reasonable to suppose that the descendants of those registered in 125/124 bc numbered approximately 400,000 in 70/69 bc.160 If we combine this estimate with the number of citizens registered at this time, it follows that there must have been at least 500,000 adult male citizens of formerly allied status, implying a ratio of 4:5 between old and new citizens. As we have already seen, the ratio between citizens and allies was approximately 4:6 on the eve of the Second Punic War, and it seems reasonable to suppose that the large-scale confiscations carried out after the Hannibalic War, along with the viritane assignations that took place during the early decades of the early second century bc, would have had the effect of increasing the relative share of the Roman citizens. It should be noted that this inference is in perfect harmony with the fact that the ratio between legionaries and allies appears to have been close to parity at the time of Polybius.161 Now, there is perhaps a theoretical possibility that the balance between legionaries and allies had absolutely nothing to do with underlying demographic realities.162 It might be conjectured that the long-term stability in the ratio between Romans and allies that can be observed between 225 bc and 69 bc is attributable to the fact that, by a remarkable coincidence, the proportions of old and new citizens captured by the census of 70/69 bc were almost identical.163 Yet the most natural interpretation is surely that 159 162

163

Per. Livy 59 and 60. 160 Cf. Chapter 4, at note 88. 161 Above, at note 67. According to Velleius 2.15.2, the allies had to bear a disproportionate share of the military burden. This, if true, would strengthen my argument. We must, however, surely reckon with a fair amount of rhetorical exaggeration. Since most of the old citizens lived closer to Rome than the former allies, one would expect the former to have been over-represented, if all adult male citizens sui iuris were required to travel to Rome.

Some other pieces of evidence

119

the consistency with which these ratios appear in completely different contexts reflects a genuine demographic pattern. If we accept this conclusion, it becomes difficult to reject the census figures for the period 204 bc– 69 bc as completely unreliable. Apart from the census figures for 86/85 bc and 70/69 bc and a few references in the literary sources, our most important source of information about the history of the census during the first century bc is the Tabula Heracleensis. Some scholars argue that this text was inscribed some time before 62 bc, when Heraclea is first recorded as a municipium.164 In that case, it could belong to the period before the census of 70/69 bc or even to that before the census of 86/85 bc. It seems more likely, however, that the text was inscribed in 45 bc.165 The Tabula is the first document that unambiguously refers to a decentralized census procedure, and Lo Cascio and other scholars have claimed that this proves their theory that no such procedure existed before the mid-40s bc. An obvious weakness of this argument is that the text of the Tabula is characterized by a striking lack of unity. As Frederiksen observed many years ago, the most probable explanation for this seems to be that the text is a digest of material copied or adapted from a variety of earlier sources.166 Precisely for this reason, many scholars have identified the section concerning the census as belonging to an earlier period, with most of them preferring a date between the conclusion of the Social War and the census of 70/69 bc. In my view, the mixed contents of the Tabula Heracleensis strongly suggest that the section on the census was indeed copied from an earlier document, meaning that the decentralized procedure it describes is unlikely to have been an innovation of the mid-40s bc.167 Admittedly, the provisions concerning the census contain no clues that unambiguously point to the 80s or 70s bc, which makes it impossible to go beyond the conclusion that such a procedure was introduced some time before Caesar’s dictatorship.

164

165 166 167

Giovannini (2004), 203, and (2008), 47–8, suggests that the provisions concerning the census were copied from a lex Iulia municipalis of 90 bc. Cf. also Galsterer (2006), 300, who favours a date shortly after the end of the Social War. Crawford and Nicolet, in Crawford (1996), 360–2. Frederiksen (1965), followed by Crawford and Nicolet (1996), 358. Unless, of course, this earlier document was a Caesarian lex Iulia municipalis of 45 bc. The majority view is, however, that Caesar never issued a matrix law regulating municipal government and administration. See e.g. Frederiksen (1965); Crawford and Nicolet, in Crawford (1996), 359.

120

Meaning of the census figures 3.6 republican census procedures: some provisional conclusions

In view of the fragmentary nature of the surviving evidence, there is no room for any far-reaching conclusions. There are nonetheless at least some good reasons for arguing that from the early third century bc onwards it was normal for cives sine suffragio who lived in self-governing communities to be registered by local magistrates, following which the results of these local censuses were transmitted to Rome. I have also argued that the technical language used in Livy’s account of the levy of 169 bc indicates the existence of a partially decentralized census procedure for registering citizens living within the confines of the Ager Romanus during the second quarter of the second century bc. It seems a reasonable supposition that this decentralized system was developed and extended to Italy as a whole soon after the Social War, that is to say, either between 89 bc and 86 bc or between 85 bc and 70 bc. As we have just seen, all of our current evidence is compatible with this view. It must, however, be conceded that compatibility does not amount to proof, and that the evidence supplied by the Tabula Heracleensis does not allow us to trace the evolution of the census between 89 bc and 45 bc. The only real certainty is that there is no secure proof of the existence of a fully centralized census procedure at any time between the end of the Social War and Caesar’s dictatorship. Although this conclusion may seem minimal, it at least shows that the theory that the use of a fully centralized census procedure resulted in a very low rate of registration is built on very weak foundations. 3.7 a change in registration or reporting practices under augustus? In his Res Gestae, Augustus reports the results of three censuses carried out at his behest between 28 bc and ad 14. In 28 bc 4,063,000 capita civium Romanorum were registered (censa). This figure had risen to 4,233,000 by 8 bc and to 4,937,000 by ad 14.168 In the Fasti Ostienses, we find a lower figure for the census of ad 14. This figure has been read as 4,100,900.169 The numerical notae, however, appear at the right-hand edge of the inscription, and approximately eight characters have been lost at the beginning of the next line. It is therefore possible to restore [xxxvii] in the lacuna and to read 4,100,937 instead of 4,100,900.170 168 170

RG 8.2–4. 169 E.g. Brunt (1971/1987), 113. Seston (1954), followed by Nicolet (1991/2000), 191–2.

A change in practices under Augustus?

121

On this basis, it has plausibly been conjectured that the Fasti Ostienses may contain a slightly corrupted version of the figure found in the Res Gestae, the discrepancy between 4,937,000 and 4,100,937 having resulted from the sort of mistake often found both in literary texts and in epigraphic documents. Interestingly, the censuses of 28 bc and ad 14 are also referred to in Jerome’s translation of Eusebius’ Chronica and in a number of other late sources, all of which seem to depend on the Eusebian tradition. In Jerome we read that the census of 28 bc revealed the existence of 4,164,000 Roman citizens (civium Romanorum).171 Exactly the same figure appears in the tenth-century chronicle of George Syncellus, who reports that 4,164,000 ‘Romans’ (Rhômaioi) were recorded in 28 bc.172 Jerome reports that 9,370,000 ‘people’ or ‘men’ (hominum) were ‘found’ when Augustus and Tiberius carried out a census in ad 14. The Armenian version of Eusebius reports a census figure of 4,001,917 for the same year. Both of these figures must be corrupt versions of the 4,937,000 mentioned in the Res Gestae.173 Further variants are to be found in the Byzantine Suda and in Syncellus. In the Suda, we read that Augustus had all of the inhabitants of the territories of the Romans counted person by person (kata prosôpon) and that the number of those living in these territories was found to be 4,101,017 men (andres).174 This passage seems to represent another branch of the Eusebian tradition. The same tradition must lie behind Syncellus’ statement that Augustus, ‘having counted the inhabitants of Rome on a personby-person basis (kata prosôpon), found the number of those inhabiting it to be 131,037 men (andrôn)’.175 3.7.1 Interpreting the Augustan census figures: philological and technical arguments Since the low counters tend to regard the census figures for the period 234–69 bc as giving at least a rough idea of real demographic developments, they cannot accept the three figures reported in the Res Gestae as having exactly the same meaning as the republican figures. They therefore have to 171 172 173

174 175

Hier. Chron. p. 163, Helm (1926). Georgius Syncellus p. 378, Mosshammer; cf. Adler and Tuffin (2002), 452. Hier. Chron. p. 171, Helm (1926). For the Armenian version, see Helm (1926), 500. Cf. Seston (1954), 20. Suidas, vol. I, p. 410, Adler (s.v. Augoustos Kaisar). Georgius Syncellus, p. 386, Mosshammer; cf. Adler and Tuffin (2002), 459. Contrary to the claim of Seston (1954), 20, the text does not give the figure 131,017.

122

Meaning of the census figures

hypothesize that in the case of the Augustan figures the phrase civium capita refers to all people of citizen status, despite the fact that the same phrase refers exclusively to adult male citizens in the written sources relating to the Republic. Starting from the premise that the Augustan censuses were the first to take women and children into account,176 most low counters argue that the figures recorded during these censuses are likely to be at least as deficient as the republican ones. Operating with a hypothetical deficiency rate of 25 per cent, Brunt arrived at an estimate of c. 5 million people of citizen status in 28 bc. With an equally hypothetical slave population of c. 2 million and with c. 1 million citizens in the provinces, this would imply an Italian population of c. 6 million (excluding foreigners).177 Various explanations for a change in census practices under Augustus have been advanced. It has been suggested, for instance, that, having noticed that men, women and children were registered in Ptolemaic Egypt, Octavian decided that he too was going to register and report the total number of citizens instead of merely the number of adult males. In support of this argument, it has been pointed out that in republican times the term civis was already sometimes used to denote any person of citizen status, including women and children.178 The alleged break with the republican tradition has also been played down in other ways. Two examples are Brunt’s claim that the language of the Res Gestae is not technical and Scheidel’s recent suggestion that the three Augustan census figures might not have been published in the form in which they appear in this document.179 It has also been pointed out that it must have been relatively easy to justify the inclusion of women and children as perfectly traditional by retrojecting the Augustan census procedure into the distant past. In this context we may instance a strange story in Dionysius of Halicarnassus according to which King Servius Tullius instituted a system under which the number of men, women and children 176

177

178

Since Dionysius of Halicarnassus claims that women and children were registered in the censuses of the sixth and fifth centuries bc, this premise is open to challenge. Cf. De Ligt (2007c), 179, for the suggestion that we may be dealing with a change in reporting practices rather than with a new policy of general registration. We must also, however, reckon with the possibility that Dionysius was simply retrojecting the census practices of his own time (cf. below, note 182). Cf. Chapter 1, at notes 23–4. Note that Brunt (1971/1987), 115–16, explains this high deficiency rate of 25 per cent as reflecting the unpopularity of the citizen census (especially among potential tax-payers) and a massive under-registration of women and children. The second of these suggestions is supported by the Egyptian evidence, which indicates heavy under-reporting of girls under the age of 15. See Scheidel (2001a), 149–50. Brunt (1971/1987), 114. 179 Scheidel (2004), 5.

A change in practices under Augustus?

123

could be established by means of coins contributed at the Paganalia.180 This system seems to foreshadow the census procedure allegedly introduced by Servius Tullius and maintained throughout the early decades of the Republic, since Dionysius repeatedly asserts that during this period all male citizens sui iuris were required to register not only themselves and their possessions, but also their women and children.181 These descriptions have been interpreted by Pieri as a retrojection of practices of the Augustan period.182 A closely related question concerns the motive or motives that might have prompted Octavian/Augustus to include women and children in the published census totals. As many low counters have pointed out, the Augustan marriage laws and Augustus’ decision to institute a system of birth registration for legitimate children point to a strong interest in demographic matters.183 More generally, the Augustan age seems to have been characterized by a vigorous policy of mapping countries, measuring land (especially in Italy) and counting people. The decision to register and report all men, women and children of citizen status might have been an offshoot of this general policy.184 Some other scholars have speculated that Octavian/Augustus had a strong interest in covering up the manpower losses of the recent civil wars, for which he was partly responsible. He may therefore have felt that inconvenient public attention to this matter could be avoided by adopting a broader definition of civium capita.185 Finally, we could consider the possibility that the liberal bestowal of citizenship upon large numbers of provincials during the last decades of the Republic had created unprecedented opportunities for non-citizens and for citizens who had married women of peregrine status to claim citizenship for themselves or their children. Only a complete registration of women and children of citizen status could solve this problem.186 Although none of these hypotheses is inherently implausible, the fact remains that there is not a single explicit reference to any change in census procedures or reporting practices under Augustus. To put it another way, a fundamental weakness of the low-count interpretation of the Augustan 180 183

184 186

DH 4.15.4. 181 DH 4.15.6. Cf. Northwood (2008), 258. 182 Pieri (1968), 15. Brunt (1971/1987), 114. As he notes, both the lex Aelia Sentia of ad 4 and the lex Papia Poppaea of ad 9 required registration of children. These laws did not only relate to the wealthier classes. Nicolet (1988). 185 Schulz (1937), 183–5. Under the Republic only adult male citizens, widows and orphans seem to have been registered (Northwood 2008, 258, n. 5, contra Mommsen 1887, ii, 362). For Dionysius’ statement that King Servius Tullius counted men, women and children, cf. above, at notes 180–2.

124

Meaning of the census figures

census figures has always been that it requires us to assume that in his Res Gestae Augustus was using the phrase civium capita in a sense without parallel in the sources for the Republic. Although the low counters are correct that the term civis could be used of women and children, this does not alter the basic fact that there is nothing to indicate that this less specific usage affected the meaning of the traditional expression civium capita. Arguing against the theories of Beloch and Brunt, the high counters have pointed out that there is no reason to think that the three figures given in the Res Gestae should be interpreted differently from the republican census figures. As they point out, Augustus went out of his way to emphasize continuity with the institutions of the Republic, which suggests that the phrase civium capita probably retained its traditional republican meaning.187 In various publications, Lo Cascio has claimed that this interpretation is supported by the evidence contained in the Suda and in the Chronicle of George Syncellus. As we have seen, both of these sources specify that the census figure for ad 14 referred to the number of men (andres) registered by Augustus and Tiberius.188 The quantitative implications of this reading of the Augustan census figures are enormous. In a recent article, Lo Cascio claims that the census figure for 28 bc points to a total citizen population of between 12.7 and 13.5 million. Of these, he assigns c. 1.25 million to the provinces, leaving between 11.5 and 12.25 million people of citizen status for mainland Italy. If the number of urban and rural slaves stood somewhere between 1.5 and 2.5 million, this theory leaves us with an Italian population of between 13 million and 14.75 million (excluding foreigners).189 Of course, this reading of the evidence raises the question of why the census figure for 28 bc is more than ten times higher than that for 125/124 bc and about 4.5 times higher than that for 70/69 bc. It is generally agreed that this enormous gap cannot be closed by adding the newly enfranchised population of Transpadana. Scholars such as Frank, Wiseman and Lo Cascio have therefore developed the theory that the censuses of the period 234–69 bc must have missed at least half of the target population. In their view, it was only after the institution of a decentralized census procedure in about 45 bc that accurate registration of the adult male citizen population could be achieved. In other words, unlike the censuses of the preceding two

187 188 189

Frank (1924), 338; Wiseman (1969); Lo Cascio (1994), 31, n. 52; Kron (2005a), 454–7. Lo Cascio (1994), 32; (1999a), 162. Lo Cascio, Malanima (2005), 203. For the number of slaves, cf. Chapter 1, at notes 32–3.

A change in practices under Augustus?

125

centuries, those carried out in 28 bc, 8 bc and ad 14 must have resulted in near-complete registration of the adult male citizen population. Not all of the philological arguments developed by the high counters are equally cogent. In particular, no independent value can be attached to the fact that George Syncellus and the Suda identify the target population of the census of ad 14 as men (andres). A glance at these late sources and at the corresponding passage in Jerome leaves us in no doubt that all of the writers active after ad 350 derived their information about the census of ad 14 either directly from Eusebius’ Chronica or from other writers who had used Eusebius and, as Seston has demonstrated, Eusebius himself must have drawn on a historiographical tradition that passed on the figure contained in the Res Gestae, or a distorted version of that figure.190 In view of these historiographical connections, we must surely conclude that neither Eusebius nor any of his successors had access to any information not contained in the Res Gestae. For this reason, their testimony cannot be used to shed light on the meaning of the phrase civium capita. Another weak spot in the high-count theory is that its estimates for the Augustan period are based on the assumption that the local magistrates and officials responsible for the task of registering the adult male citizen population in 28 bc, 8 bc and ad 14 managed to register all, or nearly all, men of citizen status. Since no census carried out in the pre-modern world is known to have captured more than 90 per cent of its target population, this assumption is clearly unrealistic. If we apply this adjustment without changing any of the other figures, Lo Cascio’s estimate of the size of the free Italian population in 28 bc rises to about 13.5 million and the figure for ad 14 to about 16 million. With between 1.5 and 3.5 million slaves, the corresponding figures for the entire free and non-free Italian population range from 15 million to 19.5 million. The minimum figure for ad 14 is 17.5 million.191 On the other hand, these critical observations do not change the fundamental fact that from a purely philological point of view the high-count reading of the Augustan census figures is clearly superior to the interpretation of the low counters, since they can only point to the dubious evidence supplied by Dionysius, who may or may not have projected a new policy of including women and children back into the distant pre-republican past.192 It follows from this that if we confine ourselves to the information supplied by the ancient sources the only serious defect in the high-count theory appears to be that it stands or falls by the assumption that no 190

Seston (1954).

191

Chapter 1, at note 34.

192

Above, at note 182.

126

Meaning of the census figures

decentralized census procedure existed before 45 bc. As we have seen, this crucial premise is not supported by the surviving evidence. In the final analysis, then, the low and high counts turn out to suffer from exactly the same methodological weakness. Those favouring the low count must assume a change in registration or reporting practices for which there is no evidence in order to explain the enormous jump in the census figures after 70/69 bc. High counters must make a number of unproven assumptions about the aims and methods of the republican censors in order to explain away the inconveniently low census figures for the last two centuries of the Republic. Viewed in this light, the arguments used by both sides can be seen to be evenly balanced. This, of course, is the main reason why this debate has been going on for more than a century. To conclude this brief discussion of the philological and technical problems raised by the Augustan census figures, a few words must be said about a new interpretation developed by the Dutch ancient historian Hin.193 The starting point of her theory is the observation that two of the republican census figures are accompanied by phrases from which it appears that orphans and widows were not covered by the figures relating to the capita civium censa. In the first of these cases, we read that orbi and orbae (orphans and widows) were not covered by the census figure for 465 bc. In the second, it appears that the census figure for 131/130 bc did not include pupillae and viduae.194 Partly on the basis of these indications, Hin argues that under the Republic the term censi must have referred exclusively to citizens sui iuris. In her view this interpretation is supported by Livy’s description of the levy of 169 bc, in which the term censi clearly denotes independent adults capable of declaring themselves, their children and their property before the censors.195 Starting from this premise, Hin goes on to argue that the high census figure for 28 bc can be explained by assuming that Octavian decided to include widows and orphans in the published census totals, presumably because he wanted to impose an inheritance tax comparable to the later vicesima hereditatium on all property-owning citizens. Since c. 40 per cent of the Roman citizen population can be shown to have consisted of men and 193 194

195

Hin (2008), 218–33 and (2009), 183–95. Livy 3.3.9 and Per. Livy 59. As pointed out above (note 27), the fact that Livy indicates that he took his information about the census of 465 bc from an earlier source makes it difficult to maintain that he added the phrase about the exclusion of orphans and widows in order to alert his readers to a recent change in census procedures. Hin (2008), 202, (2009), 173.

A change in practices under Augustus?

127

women sui iuris (if it is assumed that all married women were in potestate and all widows sui iuris), the census figure for 28 bc would then imply a citizen population of at most c. 10 million. In reality large numbers of married women were sui iuris. As Hin realizes, however, her theory logically implies that these women were included in the Augustan census figures, for the obvious reason that they too were property-owners. With this modification, the number of citizens of both sexes and of all ages drops to 7.0 million, and the Italian population (including slaves and foreigners) to about 7.5 million.196 Needless to say, the main interest of this imaginative reconstruction lies in the fact that it sidesteps the old dichotomy between low counters and high counters by presenting us with an intermediate scenario in which the population of Italy is roughly 50 per cent lower than the level envisaged by the high count, but also 20 per cent higher than the most recent estimates of the low counters. The ingenuity with which this intermediate position has been constructed should not, however, blind us to the weakness of some of its underlying arguments. One of these weaknesses has to do with the workings of the republican census procedure. While it is generally accepted that heads of households were expected to declare not only themselves but also their adult sons in potestate, it does not follow that only the declarants themselves were regarded as censi.197 For the reasons set out earlier in this chapter, there can be no doubt that the census figures for the middle and late Republic are to be interpreted as referring to the total number of adult male citizens.198 This makes it impossible to posit a basic continuity between a period in which all men sui iuris were enumerated and a period in which the census figures began to include all men, women and children sui iuris. A closely related problem is that it is not immediately apparent how women of citizen status fit into Hin’s interpretation of the Augustan figures. Here one difficulty is that married women sui iuris, one of the groups supposedly covered by the Augustan figures, were evidently neither pupillae nor orbae. In other words, the two phrases that are the starting point of Hin’s analysis do not actually support her re-interpretation of the Augustan census figures. Thirdly, and finally, Hin’s emphasis on the fiscal aspects of the census leads her to suggest that Octavian’s decision to include women and children 196

197

Hin assigns 1 million citizens to the provinces and 1.5 million slaves and foreigners to mainland Italy. Above, at note 29. 198 Above, section 3.2.1.

128

Meaning of the census figures

sui iuris in the published census figures had to do with his plan to impose a tax on those inheriting property from deceased citizens.199 Since detailed information concerning the assets owned by widows and orphans (and perhaps also married women sui iuris) had already been gathered during the Republic, their hypothetical inclusion in the published census figures would not have served any practical fiscal purpose. It may also be observed that there are no indications that the vicesima hereditatium, which was introduced in ad 6, could not be collected during the period ad 15–47, when no censuses were taken. This fact casts considerable doubt on the theory that Octavian’s decision to include widows and orphans was prompted by fiscal considerations. For all these reasons it seems fair to conclude that, while the low and high counters are faced with considerable problems in explaining the Augustan census figures, the interpretational difficulties posed by Hin’s intermediate scenario are even more numerous.200 3.7.2 Comparative perspectives on the Augustan census figures Although the text-based arguments of previous scholarship should not be set aside as uninteresting or useless, the fact that the low and high counts have been on the table for more than a century suggests that this debate cannot be settled by means of philological and technical argument alone. For this reason, we must try to place the low-count and high-count interpretations of the Augustan census figures (or, to be more precise, the general demographic trajectories implied by these interpretations) within a wider context. One step in this direction was taken in Chapter 1, in which Lo Cascio’s estimate that the population of mainland Italy was between 15 and 16 million at the time of Augustus was compared with recent estimates for the latemedieval and early-modern periods. One of the conclusions that emerged from this comparison was that, even if we take into account grain imports from the provinces, it remains very difficult to explain how the 15 or 16 million Italians of the high-count theory could have been fed in the absence of the new crops (especially maize) that helped to sustain the significantly smaller Italian population of the seventeenth and eighteenth centuries.201

199 200

201

Hin (2008), 232–3; (2009), 194–5. I hasten to add that Hin’s chapter on the Augustan census figures is the only weak part of an otherwise excellent book that offers a large number of valuable new insights into Roman demographic history. Chapter 1, at notes 91–2.

A change in practices under Augustus?

129

Conversely, the low-count reading of the Augustan census figures would seem to make Roman Italy rather under-populated by late-medieval and early-modern standards. This problem can, however, be solved by assuming that certain parts of Augustan Italy were in the middle of a process of economic development and demographic growth that may have continued until the early decades of the second century ad. In other words, the wide gap between early-imperial and late-medieval Italy could well have diminished considerably after 28 bc.202 Another way of assessing the general plausibility of the low and high counts is to look at the widely diverging rates of population growth during the period between the outbreak of the Second Punic War and the beginning of the Principate that they imply. Any attempt to calculate the average rates of natural demographic growth required by the high and low counts is, of course, complicated by the fact that the citizen population of Augustan times consisted not only of the descendants of the free Italian population of 225 bc, but also of the offspring of manumitted slaves and a significant number of provincials enfranchised by Sertorius, Pompey, Caesar and the triumvirs. Since female slaves tended to be manumitted in their thirties, their contribution to the long-term expansion of the free population was probably limited.203 The number of newly enfranchised provincials in 28 bc cannot be determined. If we take on board Brunt’s suggestion that some 110,000 adult males were enfranchised in provincial colonies and municipia before 28 bc, the total number of enfranchised provincials (including women and children) would have been approximately 330,000.204 While some scholars think this figure far too high, others have argued that the real number of enfranchised provincials was several times higher.205 The only acceptable conclusion is that no certainty is possible. In an earlier chapter, we saw that the Polybian manpower figures indicate a free Italian population of approximately 3.9 million in 225 bc. If we combine this estimate with the c. 5 million citizens implied by the lowcount interpretation of the census figure for 28 bc, we end up with an average annual growth rate of about 0.1 per cent. Of course, this rate drops significantly if we take non-natural growth into account. For instance, if we assume (arbitrarily) that the number of Roman citizens would have stood at 4.5 million if no slaves had been manumitted and no provincials enfranchised, we will end up with an average annual growth rate of only 0.07 per cent. This rate is very low compared with the rates calculated for various 202 204

Cf. Chapter 6. 203 E.g. Brunt (1971/1987), 143–5, followed by Scheidel (1997), 160–2. Brunt (1971/1987), 262–3. 205 Cf. Chapter 1, note 35.

130

Meaning of the census figures

parts of early-modern Europe (cf. below). A higher growth rate can, of course, be obtained by lowering the assumed starting population of 3.9 million. In this context, it must be remembered that we have no reliable evidence regarding the size of the indigenous population of pre-Roman Cisalpina.206 Even if we adopt the very low estimate of 750,000 for this population, we still arrive at an average growth rate of no more than 0.17 per cent per annum, including all forms of non-natural growth. How can so low a rate be accounted for? In my view, the only possible answer to this question is that the demographic gains achieved by the expansion of some subsections of the free Italian population had been annulled by a decline in other groups.207 As we shall see in the next chapter, there is good reason to believe that the Roman citizen body recovered quickly from the setbacks suffered during the Second Punic War and continued to grow throughout the second century bc. If the low count is correct, the census figures for the mid- and late second century bc should give us some idea of the average annual rate at which the Roman citizen body was expanding.208 For instance, if we take the census figures for 163 bc and 124 bc as our starting points, we arrive at an average annual growth rate of about 0.3 per cent.209 This relatively slow expansion would have included all accretions resulting from the manumission of slaves and from the legal or illegal acquisition of Roman citizenship by Latins and other non-Romans. It must also be remembered that this was a time at which the foundation of Latin colonies had come to a halt. In periods during which more Latin colonies were founded, the expansion of the citizen body is likely to have proceeded at a significantly slower pace. Up to a point, the expansion of the citizen body can be seen as mirroring the growth of the Ager Romanus. In this context, the extensive confiscations that took place after the Hannibalic War and the colonial foundations and viritane assignations of 201 bc–173 bc spring to mind. It seems likely that these gains were large enough to offset any loss of land resulting from the spread of slave-staffed estates in central-western Italy.210 206 207

208

209

210

Chapter 2. In order to avoid overly complicated formulations, I have used the term ‘free Italian population’ to denote all free persons of Italian descent, including those living in the provinces. For reasons that will be explained in the next chapter, the figures for the early second century bc cannot be used for this purpose. Of course, this type of calculation is possible only if we assume that the registration rates achieved in these censuses were roughly equal (De Ligt 2004, 741, n. 34). If we repeat this exercise for the period 225 bc–124 bc, assuming that there were 340,000 adult male citizens in 225 bc and that the census figure for 124 bc was 10 per cent defective, the long-term growth rate drops to 0.2 per cent. For further discussion, see Chapter 4.

A change in practices under Augustus?

131

Another demographic sub-group that is likely to have expanded between 225 bc and the Social War is the Latins. Between 201 and 177 bc five new Latin colonies211 were founded, and the Latins also benefited from at least one viritane distribution in Cisalpine Gaul.212 There is, moreover, an impressive amount of archaeological evidence to suggest that many Latin towns and their territories continued to flourish during the first century BC.213 By contrast, many other sub-groups of the free Italian population had lost extensive tracts of territory. In Aemilia the Boii had been deprived of half their territory and almost certainly of a higher proportion of their arable lands.214 According to some sources, many of those who survived left Italy and settled in Bohemia.215 The void thus created was, of course, partly filled by immigrants of Roman, Latin and allied status. Nonetheless, it remains likely that it took some time for the demographic losses resulting from warfare and deportation to be made good. In the long term, the net result must have been an expansion of the Romans and Latins at the expense of the pre-Roman population. Other areas in which large-scale confiscations were carried out between 201 bc and 170 bc included Apulia, Bruttium, Lucania and Liguria. Some of the Samnite tribes (in particular the Hirpini) were deprived of a large part of their territories. Like the former territories of the Boii, many of these areas were designated for settlers of Roman or Latin status at various points during the second century bc, and the towns of western Samnium experienced a further round of confiscations and assignations to Roman veterans at the time of Sulla.216 In assessing the general plausibility of a model of slow growth in the free Italian population, we must also take into account the probable effects of the Hannibalic War, the Social War and the civil wars of the first century bc. Drawing on Nefedov’s recent research into the relationships between political instability, internal warfare and demographic development in premodern China, Turchin has recently argued that instability in pre-modern empires has generally been associated with demographic stagnation or decline, and that republican Italy is unlikely to have been an exception to 211 212 214 215

216

Copia, Vibo Valentia, Bononia, Aquileia and perhaps Luca (cf. Salmon 1933; Toynbee 1965, II, 538). Livy 36.39.3. 213 Pelgrom (in press). Livy 37.2.5 shows that the Boii were forced to leave the confiscated parts of their territory. Strabo 5.1.6 and 5.1.10. Williams (2001) is optimistic about the fate of the surviving Boii, whom he sees as having quickly adopted Roman/Latin habits, but since the Celtic-speaking population of Aemilia is archaeologically invisible after 190 bc, the number of survivors cannot be determined. Santangelo (2007a), 148–54.

132

Meaning of the census figures

this rule.217 Needless to say, it is always possible to play down the amount of disruption actually caused by the Second Punic War or by the various conflicts fought out between Italian armies between 91 bc and 31 bc.218 Even if we take a very optimistic view of the Italian potential for recovery, however, it remains true that the highly unstable conditions that existed in Italy throughout several decades of the period between 225 bc and 28 bc were hardly conducive to fast population growth. Although the quantitative effects of these complicated and mutually interdependent developments cannot be gauged, it seems reasonable to conclude that the net outcome might have been a slow increase in the number of free people of Italian descent. The high-count theory faces us with the task of modelling an increase from c. 6 million to c. 14.75 million free people between 225 bc and 28 bc.219 The average annual growth rate required to produce this increase would have been about 0.45 per cent. Since the early-imperial citizen population included manumitted slaves and their descendants, along with a significant number of enfranchised provincials, the underlying rate of natural growth would have been significantly lower than this. Despite this caveat, there can be no doubt that the high-count theory requires an average annual growth rate far in excess of 0.3 per cent per annum (often given as the highest rate that the populations of pre-modern Europe were able to sustain for any longer than 150–200 years in the absence of non-natural increases resulting from immigration).220 In a recent article, Morley has called attention to a handful of latemedieval and early-modern examples that seem to depart from the normal European pattern.221 He points out, for instance, that the population of Holland increased at a rate of 0.8 per cent per annum between 1514 and 217 218

219

220

221

Turchin (2005). Cf. also Turchin and Scheidel (2009). See Brunt (1971/1987), 269–77, for an attempt to minimize the long-term impact of the devastation wrought by Hannibal and the Romans during the Second Punic War. For a much bleaker view of the impact of the Hannibalic War, see e.g. Cornell (1996a). 4.063.000 x 3.3 x 1.1 = 14,748,690. Again, citizens living in the provinces are included. For 225 bc, Lo Cascio and Malanima (2005), 201–2, assume a starting population of between 6 and 8 million. These figures are, however, derived from an erroneous interpretation of the Polybian manpower figures. See Chapter 2. According to Livi-Bacci (2000), 89, the average annual rate of population growth in Europe as a whole was somewhat lower than 0.3 per cent in the period 1550–1800. During this period, the population of England expanded at an average annual rate of 0.4 per cent (ibid. 9). Much higher growth rates can be calculated for the frontier regions of the United States and Canada, but these reflect an exceptionally favourable balance between people and territory. See e.g. Easterlin (1971), 399; Sallares (1991), 75, 86; Livi-Bacci (2001), 48–52. Cf. also Hin (2009), 159, for the observation that even a long-term annual growth rate of 0.2 per cent is quite high for a pre-industrial population. Morley (2001), 53.

A change in practices under Augustus?

133

1622. Similarly, the populations of France and Sicily are thought to have expanded at an average annual rate of 0.7 and 0.8 per cent respectively between 1450 and 1560.222 In the case of Holland, it is generally agreed that the growth rate of 0.8 per cent was sustained only because of large-scale immigration.223 The same explanation seems to hold for Sicily, where a substantial proportion of the population gains of the late fifteenth and early sixteenth centuries can be attributed to immigration from the southern mainland, from northern Italy and from Albania.224 The high growth rate that Morley calculates for France is of doubtful value. Recent literature concerning French demographic history abounds in warnings that all estimates relating to the size of the French population before 1600 are highly unreliable. The standard view is that the population of France (within its present borders) stood at approximately 20 million immediately before the Black Death, fell to about 12 million between 1350 and 1450 and climbed back to 20 million between 1450 and 1560.225 During the recovery period, the average annual growth rate implied by these figures is 0.45 per cent rather than 0.7 per cent. Of course, even this lower rate is still impressive. It must, however, be emphasized that the French recovery occurred over a far shorter period than the 197 years between 225 bc and 28 bc and that it took place at a time during which France underwent no large-scale devastation resulting from internal or external warfare. It seems significant that French population growth slowed down to approximately zero during the second half of the sixteenth century, when large parts of the country were affected by the Wars of Religion.226 Against this general background, it is difficult to see how an annual growth rate any higher than 0.3 per cent could have been maintained in Italy during a period that witnessed extensive destruction in southern Italy during the Hannibalic War, in Cisalpine Gaul during the Roman conquests of the early second century bc and throughout Italy during the Social War and the civil wars of the first century bc. Although perhaps this counter-argument 222

223 226

For rapid population growth in late-medieval Sicily, see also S. R. Epstein (1992), 72–3, and L. Epstein (1991), 17 (with excessive reliance on the estimates of Russell 1972). E.g. Van Zanden (1993), 26. 224 Epstein (1992), 73. 225 Dupaquier (1988), vols. i and ii. According to Livi-Bacci (2000), 8, the French population only grew from 19.5 million to 19.6 million between 1550 and 1600. If France did indeed have 12 million inhabitants in 1450 and 20.3 million in 1650 (Livi-Bacci, ibid.), its population must have grown at an average rate of 0.27 per cent per annum over a period of 200 years. If the figures given by Del Panta et al. (1996), 277, are reliable, the joint population of Sicily and Sardinia expanded at an average annual rate of 0.45 per cent during the same period. As we have seen, however, the rate of natural increase was in this case significantly lower. For both islands this was a period of peace.

134

Meaning of the census figures

cannot be regarded as decisive, it certainly speaks in favour of the low count, which can point to these disruptions as part of the reason why natural population growth in late-republican Italy fell short of the more usual premodern rate. 3.6 conclusions As we saw during our discussion of the Augustan census figures, the highcount reading of the figure for 28 bc can be supported with excellent philological arguments. Nonetheless, it seems fair to conclude that major difficulties appear once the demographic implications of this seemingly attractive reading are examined in the light of comparative data relating to Italian population levels reached before the arrival of maize, and to plausible rates of population growth during periods including several decades of severe military and economic disruption. In the next three chapters I shall argue that further weaknesses in Lo Cascio’s attempt to revive Frank’s interpretation come to light if we examine the capacity of the high-count theory to make sense of various other categories of evidence. These categories include the literary tradition concerning the demographic background to the Gracchan land reforms, the abundant archaeological data relating to the development of Italy’s urban network and a fast-growing body of rural survey data. If the arguments to be developed in these later chapters are broadly valid, the conclusion that Octavian/Augustus did indeed decide to register and report the existence of every man, woman and child of citizen status cannot be avoided. Even if we accept that the low count performs a better job in accommodating large amounts of literary and archaeological material and that this feature of the low count allows us to cut the Gordian knot presented by the early-imperial census figures, we can only speculate as to the reason or reasons that may have prompted the first emperor thus to expand the target population of his census.

chapter 4

Peasants, citizens and soldiers, 201 bc–28 bc

4.1 introduction Investigators of demographic developments in Italy between the end of the Second Punic War and the beginning of the Social War can seek enlightenment in a variety of sources. These include the census figures reported by Livy and his epitomator, a relatively rich body of evidence relating to the sending out of viritane settlers and the foundation of Roman and Latin colonies, and a continually growing body of archaeological data. Amongst these sources, only the census figures refer solely to Roman citizens. Despite this drawback, they have attracted a disproportionate amount of attention, for the obvious reason that they appear to offer a sound basis for detailed quantitative conclusions. As we saw in the previous chapter, it is possible to detect a growing reluctance among ancient historians to use these figures as their starting point and to force all remaining types of evidence into interpretative frameworks derived from this specific data set. Indeed, if one accepts the high-count reconstruction of Italy’s demographic history it would appear that no republican census succeeded in capturing more than 55 per cent of its target population.1 Since we must also reckon with considerable variations in the degree of efficiency with which censuses were carried out, it would then follow that the census figures are essentially worthless as a starting point for any demographic inquiry. Once the reliability of the census figures for the second century bc is called into doubt, the possibility of anchoring any reconstruction of the economic, social, military and political history of post-Hannibalic Italy within the parameters of a widely accepted demographic framework disappears. The effect of this is to undermine the validity of almost all previous attempts to establish causal relationships between various long-term developments such as the spread of agricultural slavery and the development of 1

Below, at notes 149–50.

135

136

Peasants, citizens and soldiers, 201 bc–28 bc Table 4.1 Roman census figures (204 bc–114 bc) Year 204/3 bc 194/3 bc 189/8 bc 179/8 bc 174/3 bc 169/8 bc 164/3 bc 159/8 bc 154/3 bc 147/6 bc 142/1 bc 136/5 bc 131/0 bc 125/4 bc 115/4 bc

Census figure 214,000 143,704 258,318 258,794 269,015 312,805 337,022 328,316 324,000 322,000 327,442 317,933 318,823 394,736 394,336

Source Liv. 29.37 Liv. 35.9 Liv. 38.36 Per. Liv.41 Liv. 42.10 Per. Liv.45 Per. Liv.46 Per. Liv.47 Per. Liv.48 Euseb. Armen. Ol. 158.3 Per. Liv.54 Per. Liv.56 Per. Liv.59 Per. Liv.60 Per. Liv.63

the free rural population, and thereby also of any theory that seeks to explain the actions and decisions of some Roman politician or general as an attempt to deal with the social or military consequences of these trends. Up to a point this is a welcome development, since at least some of the assumptions underlying the overall reconstructions of Brunt and Hopkins (and their interpretations of certain specific episodes in late-republican history) can now be shown to be problematic. At the same time, this loss of innocence makes it almost impossible to construct anything even approximating to a running narrative delineating the principal outlines of the demographic, economic and social history of Italy between the end of the Second Punic War and the early years of the Principate. In the sections that follow, I shall make no attempt to tell a linear story on the basis of the surviving literary and archaeological evidence. Since it is my intention to call attention to the weaknesses of previous attempts to describe the history of Roman Italy in terms of more or less linear developments within a time-scale of 170 or more years, however, I have subdivided the period 201 bc–28 bc into three sub-periods. My first period starts with the end of the Second Punic War and ends in 163 bc, when the republican census figures reached a temporary high. My second period spans the 30 years between 163 bc and the beginning of the Gracchan land reforms. Finally, I shall examine certain important developments that took place between 133 bc and 28 bc, such as the military reforms introduced by

Developments between 201 bc and 163 bc

137

Marius, the growth of towns and the emigration to the provinces of hundreds of thousands of free Italians. In accordance with this book’s general approach, my main focus will be upon the assumptions that must be made about various categories of evidence in order to produce a satisfactory fit with any of the competing demographic reconstructions. One of my findings will be that the quantitative and qualitative evidence supplied by the surviving literary sources, along with the numerical trends that may be observed in rural sites dating from the late Republic, is compatible with the low-count theory. I shall, however, also argue that in opting for the low count as the most accurate guide we have to the demographic, social and economic history of Italy during the second and first centuries bc we should avoid falling into the trap of accepting the reconstructions of Beloch, Brunt and Hopkins as correct in each and every respect. I shall, in fact, try to demonstrate that their theory that the Roman citizen body began to stagnate or decline from the midsecond century bc onwards rests on weak foundations and that a more convincing reconstruction of demographic developments can be achieved if we assume that this particular sub-group of the Italian population continued to expand until the outbreak of the Social War. 4.2 developments between 201 bc and 163 bc: four questions Since the high counters reject all republican census figures as hopelessly defective, the number of capita civium registered by the censors of the second century bc cannot be used to put their theories to the test. This is not true of the low count. Most low counters are, of course, aware of the strange fluctuations that can be observed in the census figures for the period 203 bc–114 bc, and none of them has suggested that these figures are straightforward reflections of real demographic developments. The endless methodological reservations that have been expressed since the late nineteenth century do not, however, alter the fundamental fact that the low count logically implies that the number of adult male citizens cannot have been much higher than that recorded in the census figures.2 For this reason the low counters, unlike the high counters, cannot avoid the question of whether or not the census figure for 204/203 bc can be connected in any significant manner to the number of adult male citizens recorded in Polybius’ manpower figures. 2

The highest deficiency rate envisaged by the low counters is 33 per cent (Brunt 1971/1987, 43).

138

Peasants, citizens and soldiers, 201 bc–28 bc

Another interesting problem concerns the astonishing pace of demographic recovery suggested by the census figures for the period 203 bc–163 bc. Taken at face value, the discrepancy between the census figure for 203 bc and that for 163 bc would seem to suggest an average annual growth rate of about 1.1 per cent.3 How can so steep an increase be explained? It is also striking that the Roman government should have maintained a policy of sending out large numbers of viritane settlers and colonists during what must, if the low count is correct, have been a demographic slump, and that these state-sponsored population transfers stopped almost exactly when the census figures had climbed back up to the highest level recorded before the Hannibalic War. How can we explain this curious relationship between demographic developments and levels of state-organized migration? Finally, we must take a closer look at the spread of agricultural slavery. As we shall see, there is considerable reason to doubt the traditional view that a rapid increase in the number of rural slaves pushed tens of thousands of free Italian peasants off the land and that this caused the social and military problems that led to the Gracchan land reforms. Nonetheless, it cannot be denied that the expansion of the urban market in Rome stimulated the creation of larger estates whose permanent workforce consisted of slaves. How can this development be fitted into the low-count and high-count reconstructions of the demographic history of Italy during the first decades of the second century bc? 4.2.1 The impact of the Hannibalic War and the census figure for 203 bc Using the (sometimes conflicting) casualty figures reported by Polybius and Livy, Brunt calculates that between 218 bc and 215 bc some 50,000 adult male citizens had been lost on the battlefield.4 It should be noted that his calculations assume that Polybius was wrong to assign eight legions to the Romans at Cannae, an assumption based solely upon the questionable idea that Hannibal’s tactics at Cannae could not have worked had the Romans had an army of c. 85,000 men of whom roughly 40,000 were legionaries.5 This view is not shared by most specialists in ancient military history.6 If we reject 3

4 6

337,000:214,000 = 1.57. A growth rate of 1.1 per cent per annum results in 55 per cent growth after 40 years. Brunt (1971/1987), 419–20. 5 Brunt (1971/1987), 419, n. 2. Ilari (1974), 154, n. 18; Lazenby (1978), 75–6; Rich (1983), 292, n. 26; Goldsworthy (2001), 64–6; Daly (2002), 25–9. Polybius’ figure for the total strength of the Roman army at Cannae makes it clear that his eight ‘armies’ (stratopeda) should be interpreted as eight legions, rather than as four legions plus four allied armies.

Developments between 201 bc and 163 bc

139

it, there is no good reason to set aside Livy’s statement that some 50,000 Romans and Italians were killed at Cannae alone.7 With this revision, the number of Romans lost by the end of 215 bc rises to about 60,000. For the years 214 bc–203 bc, Brunt estimates the number of legionaries who fell in battle at 75,000.8 Although some of his assumptions, such as his idea that most or all Roman legions were operating at under full strength during this period, are debatable, this estimate does not seem unreasonable.9 On top of these losses, the number of adult male citizens liable for legionary service was depleted by the defection of the Campanians in 216 bc. In a no doubt fictitious speech that Livy puts into the mouth of the Roman consul Terentius Varro, the total number of Campanianinfantry and cavalry is said to have been 34,000.10 It is unclear whether this figure was intended to refer to all men of military age, or only to the iuniores.11 A more important problem is that it is identical to the number of fighting men which Strabo attributes to Tarentum, suggesting that we are dealing with a symbolic figure meant to convey the notion ‘an awful lot’.12 Finally, any attempt to assess the reliability of Livy’s figure by calculating the average population density it implies for the Ager Campanus is bedevilled by the fact that we do not know whether Livy is referring to the number of fighting men from the town of Capua and its territory (c. 500 km2), those from Capua and the neighbouring towns of Atella, Calatia and Suessula (700 km2), or those living in the Campanian prefecture (1,000 km2).13 If we take this last figure as our starting point and interpret the figure of 34,000 as referring to the entire adult male population, we end up with an average population density of c. 112 free inhabitants per square kilometre. This result is in line with the average population densities calculated by Beloch for the same area at the beginning of the seventeenth 7

48,200 killed and 18,700 captured: Livy 22.49.15; 50,000 killed: Livy 22.59.5 and 60.14, 25.6.13. Polybius (3.117.4) assumes 70,000 casualties, but this figure is generally considered to be too high (e.g. Goldsworthy 2001, 155). It can be explained as a round figure for the total number of men killed or captured. 8 Brunt (1971/1987), 422. 9 Another uncertainty is whether or not there was only one battle of Herdonia. If we accept the first battle of Herdonia as genuine, some 10,000 dead legionaries must be added. For discussion and references, see e.g. Brunt (1971/1987), 652; Ilari (1974), 155, n. 19; Yardley (2006), 662. 10 Livy 23.5.5. 11 While Brunt (1971/1987), 18–19, 64–5, interprets the figure for the Campanians as referring to all adult males, many other scholars believe it to relate only to iuniores (e.g. Savino 1997, 176–7). 12 Strabo 6.3.4. Cf. Chapter 2, note 134. 13 For a clear discussion of the various possibilities, see Savino (1997), 177–9. Brunt (1987), 19, n. 4, interprets Livy as referring to the adult male citizen population of the Campanian federation, excluding Cumae, Acerrae and Suessula.

140

Peasants, citizens and soldiers, 201 bc–28 bc

century.14 At this point, it should be emphasized that this convenient outcome by no means proves the figure of 34,000 to be reliable. It has been pointed out, for instance, that the Campanian prefecture was probably created after the Romans had recaptured Capua in 211 bc.15 On this basis, it can be argued that Livy must be referring to the military potential of Capua and a handful of neighbouring towns. If we begin from the estimate of 700 km2 and assume, optimistically, that the average number of free inhabitants per square kilometre equalled the highest level observed in Campania in around ad 1600, there would have been no more than 28,850 adult male Campanians (assuming slaves to have made up only 15 per cent of the population).16 Although almost any figure relating to the demographic impact of the Second Punic War (and especially that for the Campanians) can be called into question, these estimates suggest that no fewer than 170,000 adult male citizens may have been killed or otherwise lost between 218 bc and 203 bc. Before trying to relate these estimates to the number of adult male citizens before the Second Punic War (as implied by the low count), we must consider Rome’s potential for demographic recovery. One mechanism coming under this heading is the cancelling out of natural deaths by new births. According to modern life tables relating to societies with demographic structures similar to that of republican Rome, between 27.4 and 31.6 per cent of men aged between 25 and 30 would have died of non-military causes within 15 years and 8.6 to 10 per cent within 5 years.17 If we assume that most legionaries were aged between 25 and 30, simplify the actual pattern of military mortality by assigning all casualties to the years 217 bc and 207 bc and posit natural mortality rates of 30 and 10 per cent for the periods 217 bc–203 bc and 207 bc–203 bc, we end up with the conclusion that of the c. 135,000 Romans killed in battle between 218 bc and 203 bc

14

15 17

In around 1600 the Terra di Lavoro had fifty-four inhabitants per square kilometre, if we exclude the city of Naples from the calculations. In the plains around Vesuvius, however, densities were sometimes as high as 160 people/km2. See Beloch (1937–61) I, 235; III, 378–80. The land distributions of 59 bc, which gave 10 or 12 iugera to a family of five (Cic. Att 2.16.1; Suet. Caes. 20.1; cf. Cic. Leg. Agr. 2.78–9, 85), imply densities of between 167 and 200 persons/km2 for the areas concerned, but not all of the Ager Campanus consisted of arable land, and densities may have increased during the last two centuries of the Republic. In the light of the comparative data, the average density of between 213 and 320 inhabitants per km2 calculated by Savino (1997), 186, seems too high for the time of the Second Punic War. Savino (1997), 178. 16 0.85 × 700 × 160:3.3 = 28,848. Coale and Demeny (1983), 43 (Model West, level 3, males).

Developments between 201 bc and 163 bc

141

some 25,000 men would in the normal course of events have died of natural causes during this period.18 It is also important to bear in mind that there is nothing to suggest that ‘normal’ levels of Roman warfare ever led to population decline. It has, for instance, been calculated that more than 130,000 Romans and Italians who would not normally have died were killed during the wars fought out in Greece, Asia Minor and Spain between 200 bc and 168 bc.19 If 40 per cent of these men were citizens, 52,000 citizens were lost. The census figures for this period, however, continue to rise rapidly. Clearly the demographic system was capable of replacing everybody who died of natural causes plus at least 1,575 additional annual casualties resulting from ‘normal’ warfare.20 At least in theory, then, the demographic system of late-third century bc Rome should have been able to produce c. 25,000 replacements in addition to absorbing the loss of the c. 25,000 men representing ordinary mortality from non-military causes.21 The ‘net’ loss of manpower resulting from the Second Punic War may therefore have been of the order of 120,000 men, if there were really as many as 34,000 adult male Campanians at the outbreak of the war.22 If we combine this hypothetical figure with the c. 340,000 adult male citizens implied by the Polybian manpower figures,23 we end up with a rough estimate of 220,000 adult male citizens at the end of the Second Punic War. This is very close to the census figure for 204/203 bc, when 214,000 capita civium were registered. Unquestionably, the many uncertainties surrounding most of the figures that have been fed into my calculations make it impossible to build any far-reaching conclusions on this neat correspondence. We may, however, conclude that the low counters should have little difficulty in reconciling the census figure for 204/203 bc with the demographic picture that seems to emerge from Polybius’ survey of Roman manpower resources on the eve of the Gallic invasion of 225 bc. 18

19

20 21

22

There is no need to underline the crudeness of this calculation. Nonetheless it shows that Brunt’s much higher estimate of 70,000 for those who would have died in the normal course of events (Brunt 1971/1987, 422) cannot possibly be correct. Rosenstein (2004), 136. As Rich (2007), 165, points out, the surviving sources may have exaggerated Roman casualties during the first decades of the second century bc. Since I am concerned only with very rough orders of magnitude, however, this criticism does little to undermine my general argument. 52,000:33 = 1,576. Cf. my remarks in De Ligt (2007a), 120–1. 16 × 1,575 = 25,200. Of course, the extreme disruption caused by the Hannibalic War might temporarily have reduced fertility to abnormally low levels. 135,000 + 35,000 – 25,000 – 25,000 = 120,000. 23 Chapter 2.

142

Peasants, citizens and soldiers, 201 bc–28 bc 4.2.2 Demographic recovery after the Second Punic War

The census figures for the period 203 bc–163 bc also raise another difficult question that merits more attention than it has received. As we have already seen, the figures for the beginning and the end of this period, if taken at face value, seem to point to an annual rate of population growth of 1.1 per cent per annum, much higher than any of the rates reconstructed for pre-modern Europe.24 This staggering figure can be reduced by adopting the widely held theory that the surviving Campanians re-entered the citizen body in 189 bc as cives sine suffragio.25 Even if we raise the figure for 203 bc by assuming 25,000 surviving adult male Campanians, the annual rate of increase must still have been as high as 0.9 per cent per annum. In principle, further modifications can be achieved by assuming that no republican census ever captured more than 90 per cent of its target population. Livy’s testimony, however, suggests that the census of 204/203 bc was carried out with exemplary efficiency, so that one would expect the census figure for 163 bc to have been less rather than more accurate than that for 203 bc.26 We would then end up with a rate of population growth among Roman citizens significantly higher than 0.9 per cent. We could also reduce the apparent rate of natural growth by assuming large-scale naturalization of Latins and other non-Roman immigrants, or by hypothesizing a high rate of non-natural increase as a result of manumission. The sources do indeed report large-scale Latin immigration, but we also read that 12,000 Latins were sent back to their home towns in 187 bc.27 Even if many of them managed to escape expulsion from Rome, it seems unlikely that their inclusion can explain the extraordinary rise in the census figures. As for the hypothesis of manumission on a massive scale, this runs up against the difficulty that there were almost certainly far fewer slaves in post-Hannibalic Italy than has previously been thought.28 There are, moreover, strong indications that female slaves were usually manumitted towards the end of their reproductive lives. If this was the case, most of their children

24 25

26 27

28

Cf. Chapter 3, section 3.7.2. Brunt (1971/1987), 63–4; Lo Cascio (2008), 248–9. Strictly speaking, Livy 38.36.5–6 only says that the Campanians were given back the ius conubii (Humbert 1978, 352, n. 55), but the Senate’s decision that they should be registered by the censors at Rome (38.36.5) seems to presuppose that they were to be registered as citizens. Livy 29.37. Livy 39.3.4–6. On this episode see Brunt (1971/1987), 72; Broadhead (2008), 461; and especially Coşkun (2009), 160–8. De Ligt (2004), 745–7, and below, section 4.3.3.

Developments between 201 bc and 163 bc

143

would not have been born as citizens.29 For both these reasons, manumitted slaves are likely to have made a relatively minor contribution to the expansion of the Roman citizen body. One of the few ancient historians who have tried to think through the problem of fast population recovery is Nathan Rosenstein. In his view, the growth rate of 1.1 per cent per annum implied by the census figures for the early second century bc must be taken very seriously. He thinks, in fact, that without heavy military mortality the annual rate of population growth might have climbed as high as 1.3 or even 1.5 per cent.30 Rosenstein believes that a completely satisfactory explanation for these high rates can be found by looking closely at the inner workings of the Roman demographic system as it had been developing since at least the third century. As he observes, the Roman conquests of the fourth and third centuries bc did not only result in a massive number of deaths among young men, but also opened up huge amounts of land for occupation by Roman citizens. In his view, these two developments must have worked to loosen the social and cultural constraints on population growth that typically operated in pre-industrial societies.31 The enormous manpower losses suffered during the Second Punic War and the colonization schemes of the early second century bc would then have pushed these tendencies even further, to the point where all constraints on population growth temporarily disappeared. In other words, if we are to believe Rosenstein, the demographic setbacks suffered between 218 bc and 201 bc had the seemingly paradoxical effect of pushing up rates of population growth to unprecedented levels during the first half of the second century bc. Although this reconstruction is very ingenious, it also raises some difficult questions. Why did no similar population explosion occur after the Samnite Wars or the First Punic War? And why should the large-scale military losses of the early second century bc, when tens of thousands of men were lost in Spain and Greece,32 suddenly have ceased to be an effective check on demographic expansion? In a nutshell, while Rosenstein has certainly identified some of the factors that made it possible for the Roman citizen body to expand, his analysis does not really explain the extreme acceleration suggested by the census figures. In my view, a convincing solution to this problem can be found by exploring the probable relationship between the census figures and general demographic 29

30 32

Cf. also Chapter 3, at note 203, for the likelihood that manumitted slaves tended to have smaller families than freeborn Romans. Rosenstein (2004), 146; (2006), 237. 31 Rosenstein (2004), 153–4; (2006), 239. Cf. Rosenstein (2004), 136.

Peasants, citizens and soldiers, 201 bc–28 bc

144

developments. Let us begin from the simplifying assumption that the Roman population can be described with the help of modern life tables.33 The table most commonly used by ancient historians is Coale and Demeny’s Model West, level 3. If we use this life table as a starting point, the adult male citizen population of 225 bc would have looked more or less like this: Table 4.2 Age structure of a population of 340,000 adult males Age groups (0–16) 17–19 20–4 25–9 30–4 35–9 40–4 45–9 50–4 55+

% of male population

Number of men

(37.06) 5.67 8.81 8.1 7.36 6.62 5.91 5.22 4.52 10.73 100.00

(-) 30,629 47,591 43,756 39,759 35,610 31,926 28,198 24,417 57,963 339,849

The next step is to look at the number of fertile women. Under Roman law girls reached adulthood at the age of 12, and this was the age at which marriage was permitted. In reality, few Roman women seem to have married before the age of 17.34 If we use this age as the commencement of female fertility and regard all women aged between 17 and 45 as capable of child-bearing, we end up with about 230,000 fertile women in Roman territory in around 225 bc, assuming the sex ratio to have been 1:1.35 What the existing literature seems to have overlooked is that the female part of the citizen population survived the Second Punic War more or less intact. It is true that in other pre-industrial societies warfare very often led to a significant increase in mortality from non-military causes (for example, because warfare resulted in food shortages and/or because it encouraged the spread of all kinds of infectious disease).36 During the Second Punic War, 33

34 35

36

Scheidel (2001d) argues that the life tables most commonly used overestimate child mortality and underestimate adult mortality. Neither Scheidel nor anyone else seems, however, to have constructed a more realistic set of life tables for the ancient world. E.g. Treggiari (1991), 399–400; Scheidel (2001c), 33. 30,629 + 47,591 + 43,756 + 39,759 + 35,610 + 31,926 = 229,271. In reality, selective infanticide might well have resulted in a smaller number of fertile women. Cf. Hin (2009), 67, n. 285. E.g. Erdkamp (1998), 278.

Developments between 201 bc and 163 bc

145

however, the Ager Campanus was the only part of the Ager Romanus to experience prolonged warfare, and we hear nothing of deadly diseases spreading among the civilian population of central-western Italy.37 It would seem to follow from this that the number of fertile women of citizen status cannot have been much lower at the end of the Hannibalic War than it had been in 225 bc. The crucial question is therefore whether enough Roman men survived the Second Punic War to make it possible for all or most fertile women to marry. At this point it must be remembered that under normal circumstances Roman men tended to marry at a considerably more advanced age than women. Although there has been some debate about the average age of Roman women at first marriage, it is generally agreed that most women married before the age of 20. Since in general men began to enter marriage only in their late twenties, they would often have been 5–10 years older than their wives.38 Up to a point this pattern of marriage can be understood as a selfperpetuating system sustained by cultural factors. It does not seem far-fetched to suppose, for instance, that many Romans felt that wives ought to be younger than their husbands, although the size of the age gap deemed appropriate would presumably have been open to continual reinterpretation. It is also easy to see how a deeply rooted system in which men aged between 25 and 30 tended to marry girls aged between 17 and 22 would have made it difficult for men younger than 25 to find wives a couple of years younger than they were themselves. In this way patterns of marriage may have become deeply ingrained in Roman society. Another important factor was the Roman system of military recruitment. As Rosenstein has demonstrated, the legions of the Roman republican army were made up largely of men who were not yet 30. As he points out, Roman patterns of recruitment seem intentionally to have minimized men’s military obligations precisely at the point in their lives when they were likely to marry.39 Viewed in this light, the Roman system of marriage can be seen as being partially driven or sustained by the Roman habit of mobilizing large numbers of young men for warfare. On the other hand, 37

38

39

According to Livy 29.7 (209 bc), many soldiers of Latin and allied status had been killed in the field or carried off by disease during the first 9 years of the Second Punic War. In 205 bc the Roman army in Bruttium was afflicted by disease (Livy 28.46 and 29.10). Only one epidemic appears to have struck the population of Rome, however (Livy 27.23), and this led to chronic rather than fatal illness. E.g. Scheidel (2001c), 33. According to Lelis, Percy and Verstraete (2003), average age at first marriage was 14 or 15 for Roman women and under 22 for Roman men. These figures would imply an age gap of about 6 to 7 years. Their figure for Roman men is, however, almost certainly a couple of years too low. See the critical discussions in De Ligt (2005) and Scheidel (2007b). Rosenstein (2004), 86.

146

Peasants, citizens and soldiers, 201 bc–28 bc

the importance of the military factor should perhaps not be overrated. After all, under normal circumstances at least half of those men aged between 17 and 30 would not have been serving in the legions, and although there were some significant exceptions (such as the soldiers serving in Spain), most legionaries were not expected to serve in many consecutive campaigns. Instead, we must think in terms of a system under which periods of service in the legions alternated with years of agricultural or other work.40 Finally, we must take into account the impact of economic considerations and of inheritance customs. It has often been observed that one of the reasons why men in traditional Mediterranean societies tended to marry late was that it took some time for young men to build up sufficient economic assets (often meaning ‘enough land’) to set up independent households.41 Up to a point the necessary resources could be accumulated by performing physical labour on other people’s farms (or in the city of Rome) or, in the case of republican Rome, by participating in lucrative military campaigns. While building up some resources independently, the younger generation was also waiting for landed property to become available through inheritance. Viewed in this light, the Roman tradition of late marriage for men can also be seen as an adaptive strategy attuned to the prevailing pattern of property transfer from one generation to the next.42 An interesting implication of this pattern of marriage is that it suggests the existence of a large pool of adult unmarried men. For example, if men did not normally marry before the age of 24, and if about half of those men aged between 25 and 29 were married, there would have been approximately 40

41

42

As Ilari (1974), 93, points out, most legions of the period 200 bc–168 bc were disbanded after 1, 2 or 3 years’ service. If we ignore the Spanish legions, the average legion existed for only 2.5 years. Admittedly, the picture is somewhat blurred by the practice of immediately enrolling the youngest soldiers from disbanded legions in freshly created ones (Ilari 1974, 102–3). Nonetheless, it was clearly very unusual for young men aged between 17 and 30 to spend more than 5 or 6 years in the legions. Many of those serving in Cisalpine Gaul (and perhaps also those stationed in Sicily or Sardinia) must have been able to visit their farms and families (if they had any) when on leave. Some of the exhortative speeches that Livy puts into the mouths of Roman generals assume that a significant proportion of serving soldiers had wives and children (e.g. Livy 28.28.7; 30.32.10). E.g. Kertzer and Bretell (1987), 95–9 (emphasizing the crucial importance of patterns of inheritance); Oppo (1990), 487. In areas dominated by large estates worked by wage labourers, men tended to marry much earlier, since they would never be able to accumulate many resources by means of inheritance and could find salaried jobs at an early age. See e.g. Rettaroli (1990), 411–12. Of course, the impact of economic factors does not mean that the age at first marriage for men always coincided with the age of economic independence. As countless studies (e.g. Rettaroli, ibid. 423) have pointed out, purely cultural factors were at least as important. Judging by modern life tables, roughly half of all young men might have lost their fathers before reaching the age of 20, but we must also consider the rights of surviving wives (who might receive the usufruct of part of the land as a legacy) and the claims of brothers and sisters; see Rosenstein (2004), 83. As Hin (2009), 97, points out, Roman mortality patterns must significantly have diminished competition for inheritable assets amongst those aged between 20 and 30.

Developments between 201 bc and 163 bc

147

100,000 young unmarried men at the outbreak of the Second Punic War.43 Of course, this pool of potential husbands is likely to have been much reduced by the heavy casualties of 218 bc–209 bc. The other side of the coin is that their number was continually being replenished by young men born between 235 bc and 226 bc. Especially after 209 bc, by which time the worst of the fighting was over, natural growth might well have been sufficient to offset losses on the battlefield. This means that by the end of the war there must have been a very substantial group of unmarried men aged between 17 and 28. In the years after 201 bc, this group would have been swelled by the addition of young men born after 218 bc.44 While most of these men would normally have postponed marriage, the extreme skewing of the sex ratio brought about by the Hannibalic War and the presence of an unusually large number of unmarried women must have created unprecedented opportunities for earlier marriage. One factor pushing matters in this direction was the fact that the deaths of tens of thousands of legionaries must have made it difficult for women aged between 17 and 22 to find enough husbands aged 25 to 30. Another was that large numbers of young men had inherited substantial amounts of property from fathers who had been killed in battle. At the same time, the realignment of property rights resulting from the Hannibalic War must have increased the amount of property that guardians or surviving fathers could offer to prospective husbands. In other words, it is not difficult to imagine a situation in which the exceptional distortion of the sex ratio created by the Second Punic War resulted in a temporary lowering of the average age of men at first marriage. As we have just seen, this would have provided tens of thousands of women who would otherwise have remained unmarried with the opportunity to find husbands and start families.45 An interesting parallel is provided by demographic conditions in Tuscany in the decades following the Black Death. Here the ‘favourable’ economic prospects created by the wiping out of one-third of the population appear to have resulted in a drastic lowering of the average age at first marriage of both men and women and in a temporary reduction of the customary age gap between husbands and wives.46 Interestingly, a return to the old pattern can 43 44

45

46

30,629 + 47,591 + (0.5 × 43,756) = 100,098. Of course, any decrease in marital fertility caused by the extremely severe casualties suffered between 218 bc and 215 bc would have slowed this trend down. It is true that between 25,000 and 33,000 men were still serving as legionaries in 199 bc (e.g. Brunt 1971/1987, 424), but, as we have seen, most of these men served for brief periods and some of them were stationed in Italy. In any case, it is simplistic to think of all legionaries as being unmarried. Note, however, that, unlike the Black Death, the Hannibalic War had killed off far more men than women. In my view, this explains the astonishing speed of the Roman recovery.

148

Peasants, citizens and soldiers, 201 bc–28 bc

be observed during the fifteenth century, when the average age at first marriage went up again, more steeply for men than for women.47 Various other adaptive mechanisms can be envisaged. It might, for instance, be hypothesized that, with so many women on the marriage market, the number of widowed men entering second or third marriages increased. We should also consider the possibility that the huge imbalance between land and people created by the Second Punic War led many women to give birth at shorter intervals.48 Like my proposal of a reduced age at first marriage for men, these suggestions are speculative. They are nonetheless important because they show that, even after a net loss of some 95,000 men,49 the demographic system of republican Rome allowed room for rapid recovery. In order to flesh out this scenario of a quick demographic rebound, I want to return to my schematic reconstruction of the Roman population (Table 4.2). As we have seen, the low count suggests that there should have been about 340,000 women of citizen status at the outbreak of the Second Punic War (if the sex ratio was perfectly balanced). Of these women, about 230,000 would have been aged between 17 and 45. As noted above, their number is unlikely to have been much lower at the time of the census of 204/203 bc. At this time the number of adult male citizens would have been about 225,000, if the census figure for 203 bc was only slightly defective. To these surviving adult male citizens we must add about 25,000 surviving Campanian men who were not registered by the censors. This brings the total number of adult men to about 250,000. From these bare figures it immediately appears that, if all men aged 17 or over were prepared to enter the marriage market, it would have been possible to find husbands for all of the fertile women who had survived the war. Of course, the reality is likely to have been much fuzzier. Certainly, the starting assumption that the sex ratio was perfectly balanced may well be unrealistic in view of the fact that female children stood a greater chance of being disposed of by means of infanticide or abandonment.50 Another caveat is that a scenario based on the assumption of universal marriage for all women aged 17 years and over is out of line with the prevailing view that 47 48

49

50

Herlihy and Klapisch-Zuber (1978), 205–8; Livi-Bacci (2001), 42. For the spacing of births as a (partly unconscious) adaptive mechanism, see Hin (2009), 105–7. As Bagnall and Frier (1994), 148–51, point out, breast-feeding had the effect of lengthening the intervals between births. 120,000 less the 25,000 hypothetical Campanians who are thought to have re-entered the citizen body in 188 bc. According to the no doubt fictitious ‘law of Romulus’ (DH 2.15.2), Roman citizens were under an obligation to raise all male children, but only the first-born of the females.

Developments between 201 bc and 163 bc

149

most Roman women tended to marry after reaching the age of 17 but before their twentieth birthdays.51 It seems reasonable to infer that my estimate of the number of fertile women looking for husbands is too high. On the other hand, it seems likely that even with so many women on the marriage market many young men remained reluctant to marry before the age of 20, and we can also be certain that many older men remained married to women no longer capable of bearing children.52 This means that it is completely unrealistic to assume that all or almost all of those adult citizens and ex-citizens who had survived the war were available for marriage.53 Even with these important qualifications, however, there can be no doubt that since the demographic system of republican Rome was characterized by the existence of a substantial age gap between husbands and wives, it had huge potential for recovery after such major demographic disasters as the Second Punic War.54 An important advantage of this reconstruction is that it enables us to get round the problem posed by the extremely high growth rate of 1.1 per cent per annum suggested by the census figures for 203 bc and 163 bc. If my general model of demographic recovery is correct, we can argue that the first decades following the end of the Second Punic War witnessed a fast recovery in the number of adult male Romans and Campanians, but not an equally fast recovery of the citizen population as a whole. The central message of my model is that instead of looking at male numbers we should focus upon the capacity of the c. 340,000 adult women who are likely to have survived the war to reproduce their own gender. If most women aged between 17 and 45 managed to find husbands and if the female population reproduced itself only at replacement level (for example, because husbands were in short supply), the number of adult male citizens would have approached 340,000 after 40 years. If we compare this figure with the roughly 250,000 adult Romans and Campanians who can be hypothesized for 203 bc, we end up with a growth rate for males of about 0.8 per cent per 51 52

53

54

Above, at note 34. The customary age gap of 5–10 years between Roman spouses implies, however, that many women older than 55 are likely to have been widows. The difficulties that (the fathers of) young women must have experienced in finding husbands could go a long way towards explaining why it took about three decades for the number of adult males to recover to pre-war levels. The same mechanism must have been at work in some other Mediterranean societies. We are told, for instance, that it took only twenty-six years for the kingdom of Macedonia to recover from the acute shortage of iuniores that existed in 197 bc (Livy 33.3.2–4; 42.11.6; 42.52.1–2). Admittedly, the Macedonian recovery was speeded up by the forced settlement of large numbers of Thracians in the mid-180s bc (Livy 39.24.4).

150

Peasants, citizens and soldiers, 201 bc–28 bc

annum. If we look at the size of the adult population as a whole, however, this rate drops to 0.4 per cent, and if we focus exclusively on the number of adult women, we end up with a growth rate of zero. These considerations suggest to me that the impression of very strong population growth in the early decades of the second century bc may well be an illusion created by the highly selective mortality that had taken place during the Second Punic War and by the simple fact that the republican census figures relate exclusively to the numerical fate of the adult male citizen population. If we correct this bias by concentrating on the number of adult women of citizen status, a far more realistic picture emerges. For instance, if we assume that the census figure for 163 bc was 10 per cent defective (implying the existence of about 370,000 adult males) and that the sex ratio had completely balanced out by the mid-160s bc, the discrepancy with our hypothetical starting population of 340,000 adult women in 203 bc would imply an average annual growth rate for adult women of only 0.2 per cent, this percentage including all forms of natural and non-natural population growth.55 It appears, therefore, that if we accept the low count the anomalous growth rate suggested by the census figures for the first decades of the second century bc can easily be explained. The pattern suggested by these figures cannot therefore be used as an argument against the reliability of the republican census figures or against the general plausibility of a low-count reconstruction of the demographic history of Italy during the last two centuries bc. 4.2.3 The rationale behind viritane distributions and colonization After the end of the Second Punic War, large-scale viritane distributions were carried out in southern Italy. In Livy’s account of the year 201 bc, we read that ager publicus in Samnium and Apulia was distributed among the African veterans of Scipio Africanus.56 We are also told that other legionaries who had seen long service in Spain, Sardinia and Africa received allotments of ager publicus in 199 bc.57 This time no information is given 55

56

57

If we assume that the number of adult women remained stable between 225 bc and 203 bc, the average annual growth rate for the period 225 bc–163 bc drops to about 0.15 per cent. Livy 31.4.1 and 31.49.5. Within Samnium and Apulia, the districts of Hirpinia, the territory of the Caudini and the Salento have been identified as areas settled by some of Scipio’s veterans. See e.g. Silvestrini (2001) and Grelle (2009), 320–3. Unfortunately, the indications commonly used to trace pockets of Scipionic veterans are less than completely reliable. See below, note 58. Livy 32.1.6.

Developments between 201 bc and 163 bc

151

about the areas in which these distributions were made.58 Kromayer and Brunt argue that as many as 40,000 adult male citizens may have been entitled to receive allotments under these schemes. The number of those who actually benefited from them has been estimated at 30,000 or more.59 Although these viritane assignations have often been noted in the existing literature, they have rarely been discussed from a demographical point of view. Some kind of explanation is, however, surely required, particularly of those favouring a low-count interpretation of the agrarian and social history of Italy. After all, if the number of adult male citizens stood at c. 340,000 in 225 bc and had dropped to about 225,000 in 203 bc (or to 250,000, if the Campanians are included), it follows that the Hannibalic War must significantly have increased the amount of land available to the free rural population of central-western Italy. With so many new opportunities opening up in the pre-Hannibalic Ager Romanus, why should the Roman government have decided to send out tens of thousands of viritane settlers to some of the newly confiscated districts of Apulia and Samnium? Before suggesting a possible answer to this question, I should like to draw attention to the fact that, as far as we can tell, the viritane distributions of 201 bc and 199 bc targeted exclusively men who had been on legionary service outside Italy. As Rosenstein has recently argued, there is good reason to believe that the recruitment of large numbers of soldiers did little longlasting harm to the farms from which these men had been withdrawn. In most cases enough family members stayed behind to carry out all necessary agricultural labour, and even if farms did suffer from neglect, returning soldiers often brought home enough cash to support themselves until new crops could be harvested.60 Even if this reconstruction seems eminently likely for long periods of republican history, the veterans of Sicily, Sardinia, Africa and Spain were clearly a special category. The legionaries who had fought with Scipio at 58

59

60

Sisani (2007), 59–60, 135–9 and 218–21, suggests that some Umbrian land was confiscated in 206 bc and that this land was assigned to veterans of the Hannibalic War in 199 bc. His main argument is that in the northern parts of Umbria some communities belonged to the tribus Cornelia or the tribus Aemilia and that these tribes are anomalous. He also suggests that these tribes were controlled by the Scipiones and their friends. Against this theory, it may be pointed out that exactly the same claim has been made regarding the tribus Galeria (Silvestrini 2001) in the case of Hirpinia and for the tribus Cornelia and tribus Claudia in that of Apulia (Grelle 2009, 322–3; but see the cautionary remarks in Grelle 2001, 24). The fact that as many as four different tribes have been identified as ‘Scipionic’ casts considerable doubt on the use of ‘anomalous’ tribes to trace Scipionic veterans and their descendants. E.g. Kromayer (1914), 150; Brunt (1971/1987), 69, n. 1; Gabba (1989), 202. Pina Polo (2006), 174, postulates several tens of thousands (‘einigen Zehntausend’) of beneficiaries. Rosenstein (2004), 101–2.

152

Peasants, citizens and soldiers, 201 bc–28 bc

Zama included the survivors of the legiones Cannenses recruited in 217 bc and 216 bc. By 201 bc these men had seen at least 16 years’ service. Similarly, many veterans of the Spanish legions had served for 15 consecutive years, while those stationed in Sardinia had done 13 years’ service.61 These periods of active service are longer than those of most of the legionaries who served in Campania, Etruria, Apulia and Cisalpine Gaul.62 It must also be borne in mind that even during the Second Punic War soldiers stationed in Italy had greater opportunities to stay in touch with their farms and families. It is true that the legions stationed in Italy during the Second Punic War are likely to have spent the winter season in camp,63 and there are indications that most Roman armies of the third century bc served all year round, but many soldiers stationed in Italy are likely to have been able to obtain leave to attend to pressing personal business. This would have enabled them to prevent their holdings being taken over by other people and to do at least some of the work required to keep their farms going.64 For the soldiers serving in Spain, Sicily and Africa between 218 bc and 201 bc, it must have been impossible to make even occasional visits. It seems reasonable to suppose that the prolonged absence of these men would have caused a significant number of their farms to deteriorate, and in at least some cases their properties may have been usurped by other people who could not easily be expelled after cultivating the land for many years.65 Another problem facing many veterans after 13 to 16 years’ service must have been that they had had little opportunity to accumulate resources by means of labour or marriage. If they had lost a lot of relatives during the Second Punic War the prospect of taking over inherited properties might certainly have attracted them back to central-western Italy, but even in 201 bc there must have been many veterans whose return was complicated by the survival of too many fathers, brothers or sisters. In this context, it must be remembered that even before the reduction of the property qualification for military service thought to have taken place in 212 bc, the ownership of 7 iugera (1.75 ha) of land seems to have been enough to make a citizen liable

61 62

63 65

De Sanctis (1916–1917), vol. ii, 632–3. Some examples: Legions 30 and 31 (disbanded 205 bc) existed for 7 years, Legions 21 and 22 (disbanded 203 bc) for 12 years, Legions 28 and 29 (also disbanded 203 bc) for 10 years, Legions 32 and 33 (disbanded 202 bc) for 9 years. For Roman armies spending the winter season in camp, cf. Rosenstein (2004), 31. 64 Ibid. 103. This problem would have been particularly pressing for those peasants whose holdings included plots of public land held on the basis of occupatio. As Rathbone (2003), 149–50, and Roselaar (2010), 208, point out, however, few peasants living in central-western Italy are likely to have had access to tracts of state-owned arable land.

Developments between 201 bc and 163 bc

153

for service in the legions.66 After 212 bc, fewer than 5 iugera seem to have been required.67 These figures suggest that even after the thinning out of the adult male citizen population many returning legionaries are likely to have welcomed the prospect of receiving farms much larger than those they had owned before the war. Of course, there must have been many veterans whose families had already owned sizeable farms before the war and whose farms had not deteriorated much. With so many variables, it is impossible to estimate the number of veterans likely to have taken up the offer of a viritane plot in Samnium or Apulia. As we have seen, some scholars have estimated their number at between 30,000 and 40,000. In my view, these estimates are far too high.68 Most importantly, the number of veterans qualifying for viritane allotments appears to have been no higher than 28,000, even if we take no account of casualties and natural deaths.69 Moreover, for the reasons already given, it seems extremely unlikely that none of these potential beneficiaries would have wanted to return to their ancestral holdings in central-western Italy, very possibly augmented by plots inherited from relatives who had died during the war. All in all, the viritane distributions of 201 bc and 199 bc are best understood as offering the opportunity for a fresh start to all those citizens for whom the prospect of a return to central-western Italy held little appeal. As we have seen, two obvious target groups would have been those whose land could not easily be brought back under cultivation and those whose families had owned little land even before the war. Of course, poor citizens could also gain access to additional land by renting from wealthier neighbours. For reasons that will be explained in a later section of this chapter, however, tenancy is likely to have had few attractions for wealthy landowners during this period, and many poor peasants are likely to have welcomed the prospect of acquiring a substantial amount of privately owned land for free. From the point of view of the Roman government too this would have been a most welcome development, since it would have increased the likelihood that the recipients and their descendants would remain assidui. It seems likely that this was seen 66

67 68

69

One thinks here of Atilius Regulus, the consul of 256 bc who was allegedly called away from his seven-iugera plot (Val. Max. 4.4.6). Further references in Rosenstein (2004), 234, n. 68. Cf. also Roselaar (2010), 205–6. Cf. Chapter 3, at note 96. It has often been pointed out that many of those eligible might not have received or wished to receive allotments. For this reason, Toynbee (1965), ii, 240, n. 7, estimated the number of recipients at between 10,000 and 15,000, while Camodeca (1991), 20, suggests that some 15,000 veterans received land in Apulia and some 5,000 in Samnium. Cf. also Gargola (1995), 108. Cf. De Ligt (2009), 266–7: 8,000 veterans from Africa plus at most 20,000 potential recipients from Spain, Sicily and Sardinia.

154

Peasants, citizens and soldiers, 201 bc–28 bc

as an urgent priority at a time when so few potential legionaries were available for service. The decision to use large tracts of state-owned land for viritane settlement thus makes perfect sense in the context of a temporary shortage of adult male citizens. The foundation of at least eight maritime colonies between 199 bc and 184 bc and of between three and six larger coloniae civium Romanorum between 184 bc and 181 bc can be seen in the same light.70 At least some of the colonists sent out to these settlements are likely to have been proletarians, while many others may have been (the sons of) assidui who owned small plots of land or had large families. Although the allotments assigned to the colonists were sometimes no larger than 5 iugera,71 they must have been large enough to make the recipients liable for legionary service. In the case of the four or five72 Latin colonies founded between 193 bc and 177 bc the outcome would have been slightly different, since men of Latin status would have served in the allied contingents. Since those sent out to Latin colonies included many men of allied status, however, even these colonies would not have drained away much manpower from the legions, and again these schemes are likely to have helped to prevent downward social mobility amongst Roman and allied assidui. 4.2.4 The low count and the spread of agricultural slavery If the low count is correct, it is impossible to avoid the conclusion that the thinning out of the (adult male) Roman population by the Second Punic War and the viritane distributions and colonization schemes of the period 201 bc–173 bc must have resulted in a population whose proportion of proletarian citizens was low. For a proper understanding of this development, it is of vital importance not to lose sight of the lowering of the property qualification for legionary service that seems to have taken place during the Second Punic War. As we saw in the previous chapter, the threshold for membership of the fifth class seems to have been reduced to 4,000 sextantal asses in 212 bc, and the size of the allotments given to Roman citizens sent out to Roman colonies in the 70

71

For these colonies see e.g. Toynbee (1965), ii, 145–6; Salmon (1969), 96–106; Laffi (2001/2007), 39– 41. While Toynbee classifies Pisaurum, Potentia and Graviscae as small maritime colonies of the old type and posits another maritime colony at Castrum Hannibalis (cf. Livy 32.7.1–3), Salmon argues that all citizen colonies founded after 194 bc received 2,000 settlers apiece and identifies Castrum (Livy 32.7.3) with the maritime colony of Salernum. Another colonia civium Romanorum was founded at Luna in 177 bc. Livy 39.55.7–9 (Mutina) and 40.29.1–2 (Graviscae). 72 Chapter 3, note 211.

Developments between 201 bc and 163 bc

155

190s and 180s bc suggests that this sum corresponded to the value of 5 or 6 iugera of colonial land. Immediately after the end of the Second Punic War most rural citizens must have owned enough land to fall into this category. As we have just seen, moreover, the further wave of land distributions must have helped to keep down the number of rural proletarians. It may be no coincidence that references to freedmen being used as rowers in Roman fleets begin to appear in Livy’s account of the years 201 bc–167 bc.73 The steady rise in the census figures makes it impossible to maintain that freedmen were being used for this purpose because freeborn citizens were in short supply. Instead, it seems reasonable to suppose that the favourable economic conditions that existed after the Second Punic War and the lowering of the threshold for legionary service had resulted in an unprecedented shortage of freeborn proletarians. As we saw in the previous chapter, Polybius’ statement that Rome was no longer able to man very large fleets could be interpreted in the same way.74 An important advantage of this reconstruction is that it makes it easier to understand the advance of slavery in the Roman countryside in the decades following the end of the Second Punic War. Most accounts of the agrarian history of Italy during the second century bc account for the spread of slavery by pointing out the various advantages to be gained by using non-free labourers. It has been stated, for instance, that slave-ownership conferred status and that slaves could be made to work longer hours than wage labourers. Moreover, unlike free labourers, slaves could be made to work in gangs.75 Finally, the flow of booty generated by the successful wars of the early second century bc boosted elite incomes and thus made slaves more affordable. While all these observations are valuable, I feel that too little attention has been paid to the bargaining position of free labourers. Even if we know almost nothing about rent levels and wages during the second century, the fact remains that the Second Punic War had severely reduced the number of adult male Romans and Campanians. If the low count is correct, in fact, the number of adult male citizens and ex-citizens had dropped by about 30 per cent. Some of the probable economic effects of this demographic downturn can be reconstructed with the help of comparative evidence. One obvious parallel is the demographic void created by the Black Death across large 73

74

Brunt (1971/1987), 423, 669–70. The Roman government’s decision to compel the citizens of the coloniae maritimae to serve as rowers in the fleet (Livy 36.3.4–6, 191 bc) points in the same direction. During the Second Punic War slaves had been used for this purpose (Livy 24.11.7–9). Plb. 1.64.1–2. Cf. Chapter 3, at notes 101–7. 75 E.g. Hopkins (1978), 111.

156

Peasants, citizens and soldiers, 201 bc–28 bc

parts of Europe in the second half of the fourteenth century. As many studies have shown, the dramatic population decline that took place in this period resulted not only in lower grain prices and higher real wages, but also in lower land rents. The obvious reason for this is that the collapse of the rural population put serfs and tenants in a strong bargaining position in relation to their landlords.76 It does not seem far-fetched to suppose that this was also the case in Roman Italy after the Second Punic War. In short, even without good evidence concerning rent levels and grain prices, it remains plausible that the use of tenants and wage labourers was an unattractive option during the first decades of the second century bc. At the same time, there are good grounds for thinking that slaves were relatively cheap during this period.77 In my view, the interplay between these two factors goes a long way towards explaining the rapid expansion of rural slavery in the early second century bc. Admittedly, we must not lose sight of some other important factors. It is, for instance, important to keep in mind Appian’s claim that wealthy landowners preferred slaves because they could not be called up for military service.78 In principle the practical disadvantages of using free men as labourers could have been avoided by leasing out plots of land to rural citizens whose assets fell short of the property qualification for military service.79 As we have just seen, however, the reduction in the property qualification for legionary service to only 4,000 sextantal asses had artificially reduced the number of rural proletarians who would have made ‘safe’ tenants. The relative dearth of rural proletarians not liable for legionary service can therefore be identified as another factor that must have made the use of free labourers unattractive, especially during the first decades of the second century bc. Although this overall reconstruction of economic and social developments in post-Hannibalic Italy is ultimately speculative, it highlights some advantages of the low count. As we have seen, the low count implies a shortage of proletarian citizens during the early decades of the second century bc that makes it possible to explain certain military developments such as the increasing use of freedmen as rowers in Roman fleets. More importantly, the low count helps to explain the spread of the slave-run villa 76

77 79

For a quick orientation, see the valuable collection of essays in Aston and Philpin (1985). In latemedieval and early-modern eastern Europe, landlords managed to avoid giving their tenants a better deal by gradually reducing large parts of the rural population to serfdom. In my view, the Roman elite of the early second century bc solved the ‘problem’ posed by the low profitability of tenancy after the Second Punic War by setting up slave-run villae. Cf. Scheidel (2005b). 78 App. BC 1.7. Although proletarians could be made to serve as rowers in the fleet.

Developments between 163 bc and 133 bc

157

in the period immediately following the end of the Second Punic War. If the Roman and Italian population was as large as the high counters have suggested, free labour should have been relatively cheap, while the chance of free tenants being dragged off to the legions would have been minimal. If we accept this view we will be forced to attribute the spread of rural slavery to purely cultural factors, such as a deeply felt reluctance on the part of free citizens to accept employment as permanent wage labourers or to work other people’s land as tenants. If pressure on the land was really as intense as the high count implies80, however, it would soon have become impossible for many citizens to act on such cultural preferences. It seems more plausible that attitudes of this sort remained the norm because they were economically realistic. Since our evidence regarding the agrarian and social history of Italy during the second century bc is extremely meagre, it would be foolish to claim that any of the foregoing arguments proves the high-count theory to be incorrect. What can be demonstrated is that the low counters have an easier job in explaining certain developments that are relatively uncontroversial, such as the spread of rural slavery. It is only in this limited sense that the low count may be called superior in terms of its capacity to make sense of some of the general economic and social trends that can be observed during the first forty years of the second century bc. 4.3 developments between 163 bc and 133 bc: the background to the gracchan land reforms All existing reconstructions of the demographic history of the free Italian population, especially that of the Roman citizen body, during the second and third quarters of the second century bc are intimately linked to competing attempts to explain the social tensions that seem to have lain behind the Gracchan land reforms, universally seen as a decisive episode in the history of the Roman republic.81 It is evident, for instance, that the particular version of the low count championed by Brunt and Hopkins cannot be separated from their view that Tiberius Gracchus introduced his programme of agrarian reform with the aim of halting a perceived decline in the free rural population.82 80

81

82

Regarding the Gracchan period, Lo Cascio (2004) even detects signs of the onset of a Malthusian crisis. Cf. below, section 4.3.7. For a detailed reconstruction of the juridical, economic and social background to the Gracchan land reforms, see Roselaar (2010). Brunt (1971/1987), 74–7; Hopkins (1978), 56–64.

158

Peasants, citizens and soldiers, 201 bc–28 bc

It will immediately be clear that this theory cannot be maintained if we accept the high-count view that the Ager Romanus of the mid-second century bc was much more densely populated than has previously been thought and that the last two centuries of the Republic were characterized by strong and continuous population growth. This explains why some adherents of the high count have developed the strikingly novel theory that, rather than being prompted by a downturn in the free rural population, the Gracchan reforms were an attempt to deal with the social consequences of a Malthusian overpopulation crisis. These completely divergent interpretations provide us with an opportunity to assess the explanatory power of the low and high counts by looking at their capacity to make sense of the literary tradition concerning the background to the lex Sempronia agraria of 133 bc. 4.3.1 The Gracchan land reforms in the literary tradition Since the nineteenth century, most attempts to elucidate the social history of the Roman citizen body during the three or four decades before 133 bc have taken their lead from the first book of Appian’s The Civil Wars and from Plutarch’s Life of Tiberius Gracchus. The fullest account is that of Appian, who traces the social and demographic problems encountered by the Gracchi all the way back to the time when the Romans were conquering Italy.83 As he explains, the cultivated land captured by the Romans was either assigned to colonists or sold or leased out. When large tracts of land remained uncultivated after the war, however, general permission was given to use it at a rent of onetenth of the produce from arable land and one-fifth of the crops harvested from vineyards and olive groves. According to Appian, the original aim of this policy was ‘to increase the population of Italy’. In fact it had the opposite effect, since most of the uncultivated land was seized by the rich, who used it to set up large estates and ranches whose workforce consisted of slaves. The next step was that many poor peasants were bought out or simply driven off their holdings. As a result of these developments, ‘the Italian people began to suffer from depopulation and a shortage of men (oligotês kai dysandria), worn down as they were by taxes and military service’. 83

App. BC 1.7–8. For a penetrating analysis of Appian’s tendency to represent earlier conflicts and policies as leading ‘naturally’ to the Gracchan land reforms, and of the distortions that are likely to have resulted from this tendency, see Gargola (2008).

Developments between 163 bc and 133 bc

159

In order to remedy this situation, a tribunician law was passed that laid down a maximum of 500 iugera (125 ha) for holdings of ager publicus. Unfortunately, this law soon began to be evaded or even completely ignored, so that the displacement of the free Italian population by slaves continued. Appian claims that it was for this reason that Tiberius Gracchus decided to launch his programme of agrarian reform, his aim being not so much to increase rural prosperity (euporia) as to promote the availability of sufficient military manpower (euandria). Essentially the same picture emerges from Plutarch’s Life of Tiberius Gracchus. Plutarch says that Tiberius Gracchus became concerned about the fate of the rural poor when travelling through the coastal parts of Etruria, where ‘barbarian slaves introduced from abroad’ had replaced the native Italian population.84 In other words, slave-staffed estates were pushing the free Italian peasantry off the land. Although many adherents of the low count have stressed the fact that we are dealing here with rhetorical simplifications of a no doubt very complex reality, the downward trend in the census figures after 163 bc seems to confirm the general picture of demographic stagnation or decline as broadly reliable. At the same time, twentieth-century scholarship has remained divided as to the causes of this negative demographic trend. 4.3.2 Theories of demographic decline: poverty and its consequences In his Italian Manpower, Brunt identifies the progressive impoverishment of a growing proportion of the free rural population as one of the main factors behind the picture of demographic decline or stagnation that seems to emerge from the literary tradition and from the census figures for the period 163 bc–130 bc. In his view, Appian’s assertion that an increase in rural poverty made it more and more difficult for rural citizens to reproduce their kind at replacement level is well supported by comparative data. He points out, for instance, that the peasants of medieval England did not normally marry until they had acquired a farm of their own. The same picture seems to emerge from data relating to eighteenth-century Sweden and Iceland and to late nineteenth-century Russia. In all these societies there appears to have been a strong link between material well-being and marriage rates, with men belonging to the lowest income groups marrying at much later ages than the better-off members of society. In Brunt’s view, these comparative data suggest that poor people tended to postpone marriage and to have fewer 84

Plut. TG 8.7.

160

Peasants, citizens and soldiers, 201 bc–28 bc

children.85 Partly on this basis, he concludes that from about the midsecond century bc the citizens ‘of the old stock’ (that is, those of free birth) began to decline, although he modifies this conclusion by pointing out the positive demographic effects of the large-scale manumission of slaves.86 In Brunt’s view, the net outcome of these processes might have been demographic stagnation rather than population decline. This superficially attractive theory has several serious problems. Foremost amongst these is the fact that the relatively low census figure for 130 bc is followed by much higher figures for the years 124 and 114 bc. If we are to believe these figures, the censors of 124 bc managed to register c. 395,000 adult male citizens, some 75,000 more than their predecessors of 6 years before (above, Table 4.1). Needless to say, these data are incompatible with the theory that the Roman citizen body was in continual decline from the late 160s bc onwards. This explains why some of Brunt’s predecessors who also believed in demographic decline took the drastic step of lowering the census figures for these years by eliminating one of the initial Cs. In other words, the traditional low-count interpretation of the Gracchan land reforms can be made to work only if two out of the fourteen census figures preserved for the second century are discarded.87 The view that the low census figure for 131/130 bc is more reliable than the higher figures for 125/124 bc and 115/114 bc sits very uneasily with Brunt’s own interpretation of the census figure for 70/69 bc, when some 900,000 adult male citizens were registered. Building on scattered data relating to the number of troops fielded by the Romans and their Latin allies during the Social War, he argues that about 412,500 of those registered in the census of 70/69 bc (or their fathers) must already have possessed citizenship in 91 bc.88 In comparing this hypothetical figure with the census figure for 131/130 bc, we must bear in mind that some 85,000 adult male citizens ‘of the old stock’ are likely to have been killed during the Cimbrian Wars and the Social War.89 Even if we assume that these losses had been made good by 70 bc, the inescapable conclusion remains that there must have been at least 400,000 adult male citizens during the final decades of the second century bc. In short, if one accepts Brunt’s arguments concerning the numerical balance between ‘old’ and ‘new’ citizens in around 85 87

88

Brunt (1971/1987), 138–40. 86 Ibid. 76–7, 81. Beloch (1886), 351; Toynbee (1965), ii, 471. The figure for 115/114 bc is accepted as reliable by Nicolet (1989), 606, De Ligt (2004), 740–1, Rosenstein (2004), 145–6, and Lo Cascio (2008), 252–3. Brunt (1971/1987), 82–3, contemplates the possibility that the census figures for 125/124 bc and 115/114 bc may have been transmitted correctly, but glosses over the fact that acceptance of this possibility would undermine his overall scenario of demographic stagnation. Brunt (1971/1987), 97. 89 Ibid. 685, 697.

Developments between 163 bc and 133 bc

161

70 bc, it becomes very difficult to set aside the high census figures for 125/124 bc and 115/114 bc. Another weakness in Brunt’s theory is that almost all of his comparative evidence comes from countries and areas whose populations were growing despite the fact that many people were living in poverty. This observation applies not only to large parts of late-medieval England, but also to eighteenthcentury Sweden and pre-revolutionary Russia.90 The only exception is eighteenth-century Iceland, where a combination of economic oppression and legal restrictions on marriage caused the population to decline.91 Since the Icelandic anti-marriage laws are, however, without parallel in mid- or late-republican Italy, this curious case does little to illuminate the demographic developments allegedly lying behind the Gracchan land reforms. Finally, and perhaps most importantly, Brunt’s thesis is based on preconceived ideas about a ‘natural’ link between poverty and population decline that are highly debatable in the light of modern demographic theory.92 For obvious reasons, most of the recent literature on this topic centres on the Third World, where widespread poverty has not prevented populations from expanding rapidly. Various competing explanations for this have been offered. One of these is Caldwell’s theory that it is economically advantageous for poor people to have a lot of children because the costs of bringing up an additional child are low compared to the additional income that can be obtained by putting him or her to work, often at an early age.93 Another popular theory is that having a lot of children can be seen as a form of old-age security, especially in countries with limited capital markets and few opportunities to accumulate resources by means of saving or insurance schemes. As some recent publications have pointed out, this form of old-age security is often particularly important to women, for instance in cultures where women tend to marry older men and in societies where job opportunities for older women are limited. Interestingly, some studies have found that fertility is highest in those parts of the Third World where the fate of a widow without a mature son is worst.94 Up to a point, the relevance of these theories to second-century-bc Italy is open to dispute. It has been pointed out, for instance, that Caldwell’s theory has little relevance to economies dominated by smallholder agriculture. The main reasons for this are that in such economies it is often difficult 90

91 92 93

For fast population growth in eighteenth-century Sweden and pre-revolutionary Russia, see e.g. LiviBacci (2000) 10, 132–3. Brunt (1971/1987), 139. For more extensive discussions of this topic, see De Ligt (2004), 749–51; Hin (2009), 110–24. Caldwell (1982). Cf. Mamdani (1972). 94 E.g. Cain (1981) and (1988).

162

Peasants, citizens and soldiers, 201 bc–28 bc

to gain access to additional land and that the additional income that can be obtained by working existing holdings more intensively is negligible.95 Under these circumstances, it is virtually impossible for young children to make a significant contribution to household incomes. Serious reservations have also been expressed concerning the idea that having a lot of children can be seen as the Roman equivalent of a modern pension scheme. It has been observed that considerations of this kind are likely to have been of limited importance in a society characterized by low life expectancies, in which older people seem to have continued to work for as long as their physical condition allowed.96 The other side of the coin is that there are clear indications that in both Greek and Roman society children were expected to look after their parents in case of need.97 For those who survived into old age, then, children may indeed have been an important source of support, although it cannot be determined how many people adopted reproduction as a strategy to cope with this eventuality. Recent literature concerning levels of fertility in Third World countries puts increasing emphasis on the relevance of purely cultural factors, such as the wish to ensure the continuity of some family cult or simply of the family name.98 This sort of approach is certainly relevant to the Graeco-Roman world, in which having many children was a source of social prestige and perpetuation of the family name was widely acknowledged as an important social goal.99 At the most general level, it can be said that all known pre-industrial societies are characterized by high levels of fertility and that there is no empirical evidence to back up the theory that poor people are either reluctant or unable to reproduce at replacement level. This fact is enough to shed considerable doubt on Brunt’s theory that the second half of the second century bc witnessed an increase in rural poverty that pushed up the average age at first marriage of both men and women and ultimately reduced levels of fertility to such an extent as to bring about a decline in the number of freeborn citizens. 4.3.3 More slaves = fewer rural citizens? In his Conquerors and Slaves, Hopkins accepts Brunt’s general reconstruction of the demographic development of the Roman citizen body, but pays almost no attention to rural poverty as a driving force behind the negative 95 97 99

De Ligt (2004), 750, n. 64; Hin (2009), 111. 96 Hin (2009), 114–17. E.g. Parkin (2003), 213–16. 98 E.g. Mukhopadhyay and Savithri (1998). E.g. Dixon (1988), 71; Hin (2009), 122–4.

Developments between 163 bc and 133 bc

163

trend suggested by the census figures for the mid-second century bc. This shift in emphasis has to do with the fact that Hopkins’ account of developments during the second and first centuries bc centres on the economic and social effects of Roman imperialism.100 One of his central observations is that the lucrative wars of the early second century bc boosted the income of the Roman elite. Since there were few alternative opportunities for investment, a large proportion of this newly acquired wealth was invested in Italian land, especially land within easy reach of the growing market in Rome. Up to a point such acquisitions were made possible, or at least facilitated, by the fact that so many Roman soldiers were serving abroad. Simultaneously, the crucial importance of the free rural population as a source of recruits made it impossible for the elite to rely on wage labour, which was moreover very unpopular in Roman society. Partly for this reason, wealthy landowners preferred to use slaves, many of whom seem to have been prisoners of war enslaved during the Roman wars in the East. Another important building block of Hopkins’ reconstruction is his claim that there was an important difference between the wars of the fourth and early third centuries bc and those of the second and first centuries. His main argument is that while the Romans and their allies were fighting on Italian soil it remained possible for most peasant soldiers to return to their farms at the end of the fighting season. When the theatre of war moved to Greece, Asia Minor, Africa and Spain, this was no longer a realistic option. Unlike the peasant soldiers of the fourth and third centuries bc, many of those recruited for service in these areas returned after spending many years in the legions only to find that their families had fallen into debt or that their farms had been taken over by creditors.101 For many of these men and their families, migration to Rome was the only realistic option left. Here these migrants joined the urban proletariat that constituted the most important market for the wine and olive oil produced on the slave-operated estates of the rich. In this way, the deracination of the free rural population stimulated the spread of slavery in the countryside. When the number of displaced peasants continued to grow during the first century bc, a new solution was found: mass migration to newly established colonies in the provinces.102 Interestingly, this general theory appears to have guided not only Hopkins’ reconstruction of the numerical fate of Italy’s free population during the last two centuries of the Republic, but also his interpretation of the Gracchan ‘crisis’. If we are to believe Hopkins, the free Italian rural population collapsed from about 4.1 million in 225 bc to about 2.9 million 100

Hopkins (1978), 11–13.

101

Ibid. 30.

102

Ibid. 64–7.

164

Peasants, citizens and soldiers, 201 bc–28 bc

in 28 bc, a drop of 30 per cent.103 Hopkins recognizes, of course, that these estimates refer to developments over a period far longer than the three or four decades preceding the Gracchan land reforms, and to population trends in a geographical area much larger than the Ager Romanus of the mid-second century bc. This chronological and geographical mismatch does not, however, stop him from characterizing the second century bc as the period during which slaves began to push large numbers of free peasants off the land.104 In view of Hopkins’ scepticism concerning the reliability of the literary sources, it is striking that his overall reconstruction of developments in the Roman and Italian countryside remains very close to Appian’s and Plutarch’s accounts. Since the expansion of rural slavery plays such a large part in both ancient and modern reconstructions, it is of the utmost importance to obtain some idea, however vague, of how many rural slaves there actually were in the midsecond century bc. A useful starting point for such an inquiry is Jongman’s demonstration that at the time of Augustus fewer than 200,000 hectares of Italian land were needed to produce all the wine and all the olive oil consumed annually by the joint populations of all the Italian cities, including Rome.105 If this estimate is combined with a labour-to-land ratio of one slave per two hectares (and this is higher than the ratios reported by Columella and Pliny the Elder), we obtain a rough estimate of only 100,000 rural slaves directly engaged in the cultivation of grapes and olives during the early years of the Principate.106 This basic figure can, of course, be pushed up by adding assistants, supervisors, female slaves and children. Moreover, several hundred thousand slaves could well have been employed on estates whose principal cash crop was grain.107 Another important caveat is that slave ownership seems to have been widespread across Roman society, so that large numbers of slaves are likely to have been employed on family farms of modest size.108 If we take all these considerations into account, an estimate of about 800,000 rural slaves can be defended for 28 bc.109 In considering the implications of these findings for developments in the Italian countryside during the first two-thirds of the second century bc, it 103 105 106 107

108

109

Cf. Chapter 1, Table 1.1. 104 Ibid. 56–64. Jongman (1990), 50–1, restated in Jongman (2003), 113–14. Labour ratios in Roman agriculture: Col. RR 3.3.8; Plin. NH 17.215. Cf. Duncan Jones (1982), 327. As Carrié and Rousselle (1999), 541, point out, the Dutch ancient historian De Neeve was the first to draw attention to the widespread use of slaves in grain farming. See De Neeve (1984), 94, 107, 113, 130 and especially 215–16 (on the use of slaves in cereal farming in Sicily). Cf. also Spurr (1986); Scheidel (1994). Rosenstein (2008), 5–7. In an earlier discussion of rural slavery (De Ligt 2004), I paid insufficient attention to the use of slaves by many moderately well-off farmers. For further discussion see Appendix iv.

Developments between 163 bc and 133 bc

165

must be remembered that the population of Rome roughly doubled between 133 bc and 28 bc.110 This can only mean that the urban market for wine and olive oil was much smaller in the second century bc than it was to become during the last years of the Republic and the early years of the Principate.111 For this reason alone, the number of slaves employed in the countryside of central-western Italy before the time of the Gracchi should not be exaggerated. The archaeological record suggests that even in central-western Italy there were few large slave-staffed estates before the early decades of the first century bc. It is possible that most of the earlier villae were small and that some of them have therefore been misclassified as small sites belonging to free farmers, but even those taking this view accept that the first century bc witnessed a very significant acceleration in the spread of slave-run estates.112 Even if we accept the existence of a causal connection between the spread of rural slavery and the decline in the free rural population, then, we must reckon with the possibility that this development did not really get under way until the first quarter of the first century bc. Taken in conjunction, these observations make it difficult to maintain that the spread of rural slavery had already pushed a large proportion of the free rural population of central-western Italy off the land by the time of the lex Sempronia agraria. 4.3.4 Recruitment and losses on the battlefield A third theory, first developed by Rathbone, postulates that the decline of the free peasantry can be explained largely as the inevitable result of the high levels of mortality associated with the large-scale wars of the third and second centuries bc. Rathbone’s central observation is that the Latin colony of Cosa, which may have had about 4,000 adult male colonists at its foundation in 273 bc, had to be reinforced with 1,000 new colonists in 197 bc, suggesting that its population had declined by about 25 per cent.113 Rathbone claims that a decline of this order can be explained by looking at average call-up and mortality rates. For instance, we could account for a 110

111

112

The population of Rome may have grown from c. 200,000 to c. 400,000 between 201 bc and 133 bc and from c. 400,000 to c. 900,000 between 133 bc and 28 bc. For discussion and references, see De Ligt (2004), 741–2. As Hopkins (1978), 73–4, observes, the institution of a permanent grain dole must have enabled more people in Rome to buy wine and olive oil. Subsidized sales of grain and free distributions took place only on a small scale before the early 50s bc, however. Rathbone (1993b). 113 Livy 33.24.8–9.

166

Peasants, citizens and soldiers, 201 bc–28 bc

decline from 4,000 to 3,000 colonists by supposing that some 700 men from Cosa were on military service at any given moment, that about 7 per cent of serving soldiers were killed and that the death of a soldier resulted in the abandonment of his farm in one in four cases.114 One weakness in the theory that conscription pushed many rural families over the edge is that it assumes that most legionaries were already married. As Rosenstein has shown, the Roman legions of the third and second centuries bc were made up largely of young men, only 14 per cent of legionaries being over 30.115 If we accept the widely held view that most Roman men did not get married until their late twenties, it follows that a very large proportion of those serving in the legions had no wives or children to leave behind. If such men died while on active service, therefore, the only households liable to disruption would have been those of their parents (if still alive). A second problem arises when we consider the possible effects of the prolonged absence or death of a legionary who had already started a family. In evaluating the impact of military mortality on surviving spouses and relatives, we must bear it in mind that many rural households are likely to have consisted of an extended multi-generational family, or possibly two co-resident nuclear families. Such households would have been in a better position to cope with the absence or death of a male family member than a single-nuclear-family household.116 Finally, we must remember Hopkins’ observation that many rural households controlled limited amounts of land and therefore suffered from a structural labour overcapacity. From a narrowly economic point of view, the effects of conscription and even of military mortality could thus have been largely positive.117 Since Rathbone’s analysis is based on the low count, which assumes that most republican censors managed to register between 75 and 90 per cent of their target population, its viability can also be assessed by looking at the census figures for the second century bc. A good starting point for this exercise is Rich’s observation that Rome’s military commitments were far heavier between 201 bc and 163 bc than they were between 163 bc and 133 bc.118 Yet the census figures for the second century bc, however we choose 114 116

117

Rathbone (1981), 16–19. 115 Rosenstein (2004), 85–8, 141–2. Cf. above, at note 39. Cf. De Ligt (2007a), 121. As so often, the only area regarding which we have a large amount of solid evidence is Roman Egypt. Here c. 43 per cent of all rural households were of the extended or multiple type. See Bagnall, Frier (1994), 67. In the case of mid- and late-republican Italy, it does not seem farfetched to suppose that the high military participation rates characterizing these societies stimulated the formation of extended and multiple households. Hopkins (1978), 24. Cf. Chapter 2, at note 143. 118 Rich (1983), 292–3.

Developments between 163 bc and 133 bc

167

to interpret them, strongly suggest that the number of adult male citizens somehow managed to increase during the 40 years following the end of the Second Punic War. Why, therefore, should the number of free rural citizens have begun to decline after 163 bc, when levels of military mortality are likely to have been considerably less? All in all, the arguments in favour of the traditional low-count reconstruction of demographic developments between 163 bc and 133 bc are weak. First and foremost, the theory that an increase in poverty resulted in lower levels of fertility and eventually in population decline is not supported by comparative data. Secondly, there can be little doubt that the quantitative importance of rural slavery during this period has been greatly exaggerated, so that there can be no question of huge numbers of peasants being pushed off the land by the advance of slave-run estates before 133 bc. Thirdly, the surviving evidence, however fragmentary, would seem to contradict the notion that conscription and a heavy military mortality rate prompted a demographic downturn. In the light of these arguments, those favouring a low-count interpretation of mid- and late-republican history ought to take seriously the census figures for 125/124 bc and 115/114 bc, which strongly suggest that the citizen body continued to grow at least until the final decades of the second century bc. 4.3.5 An alternative low-count reconstruction If we accept the census figures for 125/124 bc and 115/114 bc as more or less reliable,119 we must conclude that the demographic expansion that can be observed during the first third of the second century bc continued until 125/124 bc and probably extended into the final quarter of the second century bc.120 How, then, do we explain Appian’s and Plutarch’s claim that Tiberius launched his programme of agrarian reform in order to halt a decline in the number of free citizens eligible for legionary service brought about by a steady expansion in rural slavery? It should be noted that this question is all the more urgent because the common source used by Appian and Plutarch seems to have drawn on various speeches made by Tiberius 119 120

That is to say, as being no more than 10 per cent defective. Cf. De Ligt (2003) and (2004). Rosenstein (2004) also believes that the population continued to grow, but (for the reasons given in an earlier section of this chapter) I cannot accept his conclusion ([2004], 146) that the census figures for the period 203 bc–124 bc suggest an average annual growth rate of 0.75 per cent. A closely related problem is that Rosenstein follows Brunt in accepting the census figures for the second century bc as more or less reliable while also expressing sympathy with the high-count interpretation of the Augustan census figures (e.g. [2004], 13). See my remarks in De Ligt (2007d).

168

Peasants, citizens and soldiers, 201 bc–28 bc

Gracchus and on a pamphlet written by his brother.121 All of these contemporary sources seem to have focused on the demographic and military rationale of the agrarian law of 133 bc. In my view, a plausible explanation compatible with the low count can be found by looking at the balance between population and agrarian resources in central-western Italy, which was home to the vast majority of the citizen population throughout the second century bc.122 If the low count is correct, the census figures for the early second century bc can be interpreted as indicating that the citizen body recovered from the losses suffered during the Hannibalic War within 30 or 35 years.123 For reasons that have already been explained, the number of rural slaves employed on estates in centralwestern Italy is also likely to have risen, even though this expansion has often been exaggerated. With both the free and the non-free rural population growing, one would expect pressure on agrarian resources to have increased. In this context, it must be remembered that several tens of thousands of viritane settlers and colonists were sent out to southern and northern Italy between 201 bc and 173 bc. The effect of this must have been to reduce competition for arable land in the pre-Hannibalic Ager Romanus and to prevent large numbers of rural citizens from falling into poverty.124 Between 173 bc and 133 bc we hear nothing of any new viritane assignation and have little evidence for the foundation of new colonies.125 If we accept the census figure for 125/124 bc as roughly correct, however, the number of adult male citizens was considerably higher at that time than in 169 bc or 163 bc. Since the continuing expansion of the urban market in central-western Italy is likely to have resulted in a further expansion of rural slavery after 163 bc, there are good grounds to believe that at least some sections of the free rural population were becoming progressively poorer during this period.126 121

122

123

124 125

126

For the sources used by Appian and Plutarch, see e.g. Stockton (1979), 6, 40; Bringmann (1985), 10; Roselaar (2008), 7–9. For the geographical distribution of the citizen population during the mid-second century bc, see De Ligt (2009). Cf. above for the suggestion that the number of adult male citizens must have grown much faster than the citizen body as a whole. Above, section 4.2.3. The only possible exception is Auximum, if indeed, as Velleius Paterculus (1.15.2) reports, this colony was founded in 157 bc. See, however, Salmon (1982), 112–15, for the suggestion that it was in fact founded in 128 bc. This reconstruction assumes that throughout the second century bc state-organized migration was the principal form of permanent migration and that spontaneous migration by individuals or families was limited. As Osborne (1991) says, pre-modern populations were sometimes very mobile, but his comparison with early-modern England is vitiated by the fact that most English migrants were landless tenants and labourers.

Developments between 163 bc and 133 bc

169

In order to refine this picture, we must also take into account the probable effects of the scaling-down of the Roman war effort that can be observed after 167 bc. As we have seen, some ancient historians have interpreted military recruitment as a disruptive force in the agrarian economy. Against this I have argued that many of the poorer rural families are likely to have welcomed the opportunity to send one or two young adults to the legions, since this would have meant a temporary reduction in the number of mouths to feed and perhaps additional income in the form of booty and donatives. If these arguments are accepted, the scaling down of military commitments after 167 bc can be seen as a negative development, in the sense that it reduced access to additional sources of income and increased population pressure in the Ager Romanus by reversing the temporary emigration of many thousands of legionaries. For some households this must have meant a further reduction in per capita resources and, in the event of the death of the pater familias, a further fragmentation of holdings. A promising feature of this hypothetical reconstruction is that it provides us with a convincing explanation for the decline in the number of assidui that seems to have worried Tiberius Gracchus and at least some of his contemporaries. As we have seen, ownership of just 4 or 5 iugera (1 or 1.25 hectares) of land seems to have been enough to meet the property qualification for legionary service at the time at which Polybius wrote the sixth book of his Histories (c. 160 bc). While the citizen body was still recovering from the setbacks of the Hannibalic War and while land distribution schemes were still active, the number of rural citizens whose assets fell short of this threshold must have remained limited. If, however, the three decades between 163 bc and 133 bc witnessed a simultaneous expansion of the free rural population and of the number of rural slaves in central-western Italy, it is easy to imagine how the holdings of an increasing number of peasants could have shrunk to fewer than 4 or 5 iugera. In other words, if the number of free country-dwellers continued to grow, this could have had the seemingly paradoxical effect of reducing the number of citizens eligible for legionary service. The upshot of my argument is that a low-count model that takes on board the idea of population growth makes it possible to take seriously Tiberius Gracchus’ claim that large numbers of rural citizens were declining into poverty and that this development was undermining Rome’s military potential by reducing the number of available assidui. This is not to say that Tiberius’ analysis was entirely accurate in each and every respect. According to Appian and Plutarch, he seems to have claimed that the number of Roman citizens had begun to decline in absolute terms.127 This could be explained as 127

App. BC 1.9; Plut. TG 8.3, 8.7, 9.4–5.

170

Peasants, citizens and soldiers, 201 bc–28 bc

a rhetorical exaggeration.128 Alternatively, it might be conjectured that Tiberius interpreted the steady decline in the census figures after 163 bc as indicating the beginning of a slow demographic downturn. If this is indeed what he thought, his understanding of recent demographic developments was flawed. Nevertheless, it must be remembered that, unlike modern ancient historians, Tiberius did not have the census figure for 125/4 bc, which was the first to reveal that the number of adult male citizens had not in fact declined. In any case, even if Tiberius’ understanding of the reasons for the decline in the number of registered citizens was wrong, his programme of agrarian reform would in fact have achieved his primary goals, namely a considerable reduction in the number of poor citizens and a corresponding increase in the number of those eligible for service in the legions. Up to a point, the tense situation implied by this reconstruction can be described as a Malthusian overpopulation crisis. It must, however, be emphasized that in the case of central-western Italy this ‘crisis’ developed not just as the inevitable result of endogenous population growth, but also at least partly as a result of the expansion of rural slavery.129 A more important qualification is perhaps that we cannot be sure that the simultaneous expansion of the free and non-free rural population really created circumstances under which the food requirements of the free rural population began to outstrip available resources.130 What we read in the sources is simply that many peasants were trying to maintain themselves on inadequate plots and that access to public land was largely monopolized by the elite. Although these snippets of information suggest that the amounts of privately owned land held by many rural citizens fell short of what was needed to cover the subsistence needs of their families, there is absolutely nothing to contradict the view that the free rural population of centralwestern Italy could easily have been supported if poor peasants had been given access to additional plots of land owned or controlled by other people. In particular, we must seriously consider the possibility that most of the tensions created by continuing population growth in the countryside could have been diffused by employing freeborn citizens as tenants on the holdings of wealthy landowners. From a strictly economic point of view, tenancy ought to have become a more attractive proposition as population growth gradually undermined the 128 129 130

Thus Bringmann (1985), 27–8. Against the idea of a purely Malthusian crisis, see De Ligt (2007c), 177. As we shall see in Chapter 6, the rural survey evidence strongly suggests that the rural population of the suburbium of Rome increased significantly during the first century ad.

Developments between 163 bc and 133 bc

171

bargaining position of the free rural population.131 A complicating factor in all this is that throughout the second century bc tenants would have been liable for legionary service if they owned sufficient property to qualify for membership of the fifth class. One interesting feature of a model of continuing population growth is that it implies a significant increase in the number of rural proletarians, who would by definition not have been liable for service in the legions. Ceteris paribus, this should have removed what was perhaps the most important obstacle to widespread tenant farming.132 As we shall see presently, however, there is some evidence (most of it, admittedly, open to alternative interpretation) to suggest that the property qualification for military service was significantly reduced either in around 140 bc or shortly after the census of 131/130 bc. If such a reduction did indeed take place, it would artificially have reduced the number of proletarians at exactly the time when taking on impoverished citizens as tenants was becoming an increasingly attractive option for rich landowners. In other words, it is possible to defend the view that, rather than being purely Malthusian, the ‘crisis’ behind the Gracchan land reforms was created, or at least exacerbated, by the Roman system of recruitment, which made it difficult even for very poor peasants to gain access to additional land as leaseholders. 4.3.6 Competing low-count readings of the census figures for the period 163 bc–124 bc Of course, the theory just outlined raises the question of how we can explain the decline in the census figures that can be observed after 163 bc (Table 4.1). One attractive possibility is the idea that this trend reflected the growing reluctance of many Roman citizens to serve in the army. The little evidence we have does not support the view that during this period a large proportion of Roman citizens became reluctant to fight in any war. There are clear indications that many were eager to enlist for the war against Carthage in 149 bc.133 On the other hand, we are told that between 173 bc and 169 bc many citizens successfully avoided enlistment for the Third Macedonian War, a campaign generally regarded as difficult.134 Similarly, there can be no doubt that the protracted and unrewarding wars fought in Spain between 154 bc and 133 bc aroused little enthusiasm amongst 131 132 133

Cf. above, section 4.2.4. The possibility of taking on proletarians as tenants is overlooked by Rosenstein (2004), 181–2. Thus, correctly, Rich (1983), 317. 134 Livy 43.14.2–6; 15.1.

172

Peasants, citizens and soldiers, 201 bc–28 bc

potential legionaries. It therefore seems reasonable to suppose that at least some adult male citizens tried to avoid military service by dodging the censors and that this had a negative effect on the efficiency of the census.135 It is striking, however, that no significant rise in registrations can be observed after the end of the Spanish Wars, when legionary service might be thought to have become a more attractive prospect.136 This suggests that a more structural explanation may be required. Did certain sections of the rural population perhaps grow unwilling to serve in the legions for reasons that had little to do with the rewards offered by military service? In my discussion of the impact of warfare upon the rural economy, I referred to Hopkins’ observation that Roman agriculture was characterized by a high degree of under-employment and that this feature of the Roman economy made it possible for the Roman government to mobilize a considerable proportion of the working population without doing much economic damage. Building on this idea, Erdkamp has recently argued that the expansion of the city of Rome must have encouraged many moderately well-off peasants to engage in the cultivation of cash crops for the urban market. Since such people would have had to work their land more intensively, a side effect of this development might well have been to reduce the attraction of other forms of employment, such as legionary service. Erdkamp argues that this helps to explain the two reductions in the census requirement for membership of the fifth class that seem to have taken place in around 212 bc and shortly after 140 bc. He argues that the aim of these reductions must have been to expand the pool of potential military recruits by targeting an existing group of very poor peasants. Since the size of the smallholdings owned by these peasants would have made it impossible for them to use much of their land for cash-crop production, the prospect of serving as legionaries would presumably have been more attractive to them. In other words, both Tiberius Gracchus’ worries about military manpower and the decision to lower the property qualification for military service can be explained without recourse to the traditional view that the mid-second century bc witnessed an increase in rural poverty.137 The main drawback of this attractive theory is that it combines a dynamic reconstruction of the fate of the more prosperous segments of the rural population with an essentially static view of the rural poor, the latter being 135 136 137

Thus Evans (1988), followed by De Ligt (2004), 742–3, and Rosenstein (2004), 157. Here one thinks of the war against Aristonicus, which was potentially very remunerative. Erdkamp (2006).

Developments between 163 bc and 133 bc

173

seen as a pre-existing group whose fate requires no further analysis. If, however, we accept that Livy’s epitomator transmitted the census figures for 125/124 bc and 115/114 bc correctly, and also take into account the spread of slavery, it seems impossible to avoid the conclusion that pressure on the land increased significantly after 163 bc, especially in central-western Italy. Regardless of the fate of those peasants who took advantage of the new opportunities offered by commercial farming, this change in the balance between land and population must on average have made Roman peasants poorer. Moreover, for all its obvious shortcomings, the literary tradition explicitly identifies an increase in rural poverty as one of the reasons that prompted Tiberius Gracchus to create his reform programme. It is, of course, always tempting to brush aside every literary account as utterly unreliable, but since this can only create a void that can be filled with an infinite variety of untestable theories, this approach should be avoided unless there are compelling reasons for discarding the few pieces of literary ‘evidence’ that happen to have survived. How, then, do we explain the slow decline in the census figures and the spectacular rise that can be observed after 130 bc? In my view, a satisfactory answer can be found by taking a closer look at some of the factors likely to have affected the efficiency of the census between 163 bc and 114 bc. The starting point of this explanation is the observation (by no means new) that the most important aim of the census was to register as many as possible of those adult male citizens owning sufficient property to be liable for military service.138 Amongst the censors’ assigned objectives, this goal must surely have gained yet more in relative importance when the suspension of tributum in 167 bc dispensed with the need to create an up-to-date record of those upon whom war taxes could be imposed. Be that as it may, it seems reasonable to suppose that (at least before the levy of 107 bc) assidui were always a more important target group than proletarians.139 If this insight is combined with a low-count reading of the census figures for the period 163 bc–133 bc, it becomes tempting to hypothesize that the number of assidui slowly shrank during these years.140 Many older publications assume that the number of assidui did indeed diminish, but explain this downward trend as part of a general decline in the overall number of freeborn citizens. If, however, we accept that the high census figures for 125/124 bc and 115/114 bc were transmitted correctly, it becomes difficult to avoid the conclusion that the said 138 140

Cf. Chapter 3. 139 For the background to the levy of 107 bc, see section 4.4.1. On this point I am in complete agreement with Rich (1983), 299, 304.

174

Peasants, citizens and soldiers, 201 bc–28 bc

diminution must have resulted from a combination of population growth and increased rural poverty. In other words, the census figures for the period 163 bc–130 bc make perfect sense if we assume that continuing population growth pushed an increasing number of citizens below the property threshold for legionary service and that this led to fewer adult male citizens being registered. If this hypothesis is correct, we are left only with the question of how to account for the steep rise in the census figures after 130 bc. The most attractive theory is perhaps that the partial implementation of the Gracchan land reforms between 131 bc and 129 bc persuaded tens of thousands of proletarian citizens to register themselves with the censors, for the obvious reason that they wanted to benefit from this scheme. According to those who support this reading of the census figures, the objection that no significant rise can be observed in 131/130 bc, when the Gracchan land committee had already been at work for some time, can be countered with the argument that the commissioners’ implementation of the reform programme took 3 years, so that its (partial) success did not begin to become apparent until 130 bc.141 Another possibility is that the jump in the census figures is to be connected with a further lowering of the threshold for military service. The main evidence for this hypothesis consists of Polybius’ statement that the threshold for legionary service stood at 400 drachmai (corresponding to 4,000 asses of the mid-second century) and two passages from Gellius and Nonius, both of whom define proletarii as people whose property is worth less than 1,500 asses.142 Somewhat curiously, Gellius also distinguishes between proletarians and capite censi, claiming that the latter term denotes citizens with property worth no more than 375 asses. It is generally agreed that this distinction is attributable to a mistake and that the terms proletarii and capite censi referred to the same group of people.143 An important argument against interpreting these snippets of information as evidence for a second reduction is that none of them unambiguously refers to the second century bc.144 One clue does, however, point to this

141 142

143

E.g. Brunt (1971/1987), 78–80; Crawford (in press). For doubts, see Stockton (1979), 49–50. Gell. NA 16.10.10, Nonius 228 L. Some scholars claim that the figure of 1,500 asses also appears in the second book of Cicero’s De Re Publica (2.40). It is, however, generally agreed that 1,100 asses was the original reading, and if a late-antique corrector changed this to 1,500 asses (which is by no means certain), there is no guarantee that this is what Cicero actually wrote. For good discussions of this notorious conundrum, see Lo Cascio (1988), 286–9; Rathbone (1993a), 140. Cf. Chapter 3, at note 86. 144 A point made by Rich (1983), 314–15.

Developments between 163 bc and 133 bc

175

period. If we accept the prevailing view that the terms proletarii and capite censi referred to the same class of person, it becomes possible to speculate that Gellius misread an account stating that the threshold for admission to the fifth class was 375 sestertii, the exact equivalent of 1,500 asses. If he did indeed do so, it would follow that the census rating of 1,500 asses must have been introduced in or after 141/140 bc, when the Roman state adopted the sestertius as the normal official unit of reckoning. If we also accept Rathbone’s suggestion that the hypothetical threshold of 375 sesterces may have represented the notional value of one iugerum of arable land during the mid-second century bc, we end up with the spectacular conclusion that during the final decades of the second century bc ownership of just one quarter of a hectare of arable land would have been enough to make a Roman peasant liable for military service.145 Rathbone believes that the new rating is likely to have been introduced simultaneously with the numismatic reform of 141/140 bc, and there can be no doubt that this is the most economical theory.146 Its only weakness is that a lowering of the census rating of the fifth class by more than 60 per cent would certainly have expanded the number of assidui to an extent that one would expect to see reflected in the census figures. Nonetheless, no significant increase in these figures is seen before 125/124 bc. In an influential study published in 1949, Emilio Gabba interprets the sudden jump in the census figures after 130 bc as an indication that the threshold for military service had been lowered during the early 120s bc.147 In formulating this theory, he begins from the assumption that the Roman census figures are to be interpreted as including only those Roman citizens with sufficient property to qualify for military service. In other words, Gabba holds that these figures do not include proletarians. As we saw in the previous chapter, this theory is unlikely to be correct. This explains why Gabba’s date for the hypothetical second lowering of the census rating of the fifth class is now widely rejected. Even if we can be certain that in theory the censors were expected to register all adult male citizens, it remains entirely credible that the names of many proletarians might not have been recorded. It is therefore still possible to defend the view that the discrepancy between the census figure for 131/130

145

146

Rathbone (1993a), 146, n. 20. I agree with Rich (1983), 313–16, that there is no solid proof of the theory that a reduction in the property qualification for legionary service from 4,000 asses to 1,500 asses took place. The other side of the coin is that Rathbone’s theory is compatible with the evidence and that there is nothing to contradict it. Rathbone (1993a), 144. 147 Gabba (1949/1976), 7–8.

176

Peasants, citizens and soldiers, 201 bc–28 bc

bc and that for 125/124 bc could reflect a lowering of the property qualification for legionary service. I conclude that two theories offer us some plausible explanation of the sudden rise in the census figures.148 The first makes excellent sense from a numismatic point of view, but must rely on the secondary assumption that the Gracchan land reforms prompted some 75,000 formerly unregistered citizens to present themselves for registration at the census of 125/124 bc. The second is purely hypothetical, but dispenses with the need to explain away the low census figure for 131/130 bc by assuming that it took some 3 years for the implementation of the Gracchan reforms to get under way. For the purposes of this chapter, it is not important which of these theories is more likely to be correct. What is more interesting is that any theory positing a considerable reduction in the census rating of the fifth class during the third quarter of the second century bc is compatible with a scenario of continuing population growth. After all, if the citizen population continued to expand after 163 bc and if this led to Roman citizens becoming on average poorer, it would have made excellent sense for the Roman government to broaden the basis for legionary recruitment by lowering the property requirement for military service. I should like to end this section by stressing the limitations of the foregoing reconstruction. As we have seen, the interpretation of almost all snippets of literary evidence is beset by deep uncertainties. Moreover, in developing an alternative model I have started from the low-count assumption that even the most unreliable censuses of the second century bc were defective by no more than 33 per cent. The moment this premise is rejected, all speculations based on a scenario of moderate population growth become entirely meaningless. For this reason alone, it is impossible to entertain the illusion that it will ever be possible to prove this scenario (or any other scenario) to be correct. What can, however, be demonstrated is that a lowcount reconstruction that takes on board the notion of moderate population growth in central-western Italy allows us to make sense of most of our fragmentary literary evidence. My next step will be to determine whether the same can be said of the high count. 148

Carcopino (1929), 14, suggests that unusually extensive manumissions of slaves and registrations of Latins and Italians on the census lists pushed up the number of citizens between 130 bc and 125 bc. Vanderspoel (1985) argues that tens of thousands of formerly unregistered citizens came forward at the census of 125/124 bc because they wanted to vote in the tribunician elections of 124 bc. While the two theories discussed in the main text attempt to connect the rise in the census figures with developments for which there is at least some evidence, however, none of the alternative explanations that have been advanced seems to have any support in the sources.

Developments between 163 bc and 133 bc

177

4.3.7 The high count and the Gracchan land reforms As we saw in Chapter 2, Lo Cascio’s interpretation of the Polybian manpower figures implies that by 225 bc there were approximately 475,000 adult male Italians of citizen status.149 If we compare this estimate with the census figure for 234 bc (c. 270,000), his interpretation suggests that the censors succeeded in registering only 55 per cent of the adult male citizen population. Similar calculations can be made for 130 bc and 124 bc. As we saw in an earlier chapter, Lo Cascio believes that in 225 bc Italy as a whole (including Cisalpina) had between 6 and 8 million free inhabitants. Since his highest estimate is based on the unrealistic assumption that in 225 bc population densities in Cisalpine Gaul already equalled those in peninsular Italy, it seems fair to start from the figure of 6 million. We have also seen that the high-count reading of the Augustan census figures implies a free Italian population of about 13.5 million in 28 bc.150 The average annual growth rate implied by these figures is about 0.4 per cent. Since the citizen body is unlikely to have expanded at a lower rate than the free Italian population as a whole, it may be inferred that the number of adult male citizens stood at approximately 730,000 in 124 bc. Since only 395,000 citizens were registered by the censors of 125/124 bc, again we end up with the conclusion that some 45 per cent of the adult male population somehow escaped registration by the censors. Although these calculations are necessarily imprecise, they clearly imply that the census figures for the second century bc cannot be used to reconstruct the demographic and social background to the Gracchan land reforms.151 Even if the high counters are inclined to dismiss the republican census figures as completely unreliable, their overall interpretation of the demographic history of Italy does in fact lead directly to a strikingly novel reconstruction of the circumstances that prompted Tiberius Gracchus to propose his agrarian law. The obvious reason for this is that the high starting population posited by the high counters, in conjunction with the relatively high annual growth rate of 0.4 per cent, implies a very large citizen population for the mid-130s bc and correspondingly high population densities for the Ager Romanus. In fact, if we accept the high count, the only possible conclusion is that by the end of the 130s bc central-western Italy was facing the beginning of what can only be called a full-blown Malthusian crisis.152 149 152

Chapter 2, at note 27. 150 Chapter 1. 151 Lo Cascio (2008). For a clear statement of this view see Lo Cascio (2004).

178

Peasants, citizens and soldiers, 201 bc–28 bc

In an attempt to back up this theory with some sort of evidence, Lo Cascio has offered an intriguing new interpretation of Appian’s statement that Tiberius Gracchus was worried by a lack of men (dysandria), presumably meaning ‘a lack of recruits’. His starting point is the philological observation that, since the Greek term dysandria often has qualitative overtones, Appian would surely have used the unambiguous expression oligandria had he intended to refer to a purely demographic phenomenon. This suggests to Lo Cascio that Appian’s description of the background to the Gracchan reforms should be interpreted as referring to a lack of healthy adult men endowed with the kind of physique required for military service, rather than to an absolute shortage of adult men eligible for call-up.153 The passage in question consequently appears to refer to what Lo Cascio calls ‘a situation of clear and structural population pressure in the Italian countryside’.154 The main weakness of this theory lies in the fact that, contrary to Lo Cascio’s assertion, Appian is in fact using purely demographical language. A particularly telling example is to be found in his summary of one of Tiberius’ speeches in defence of his proposal. According to Appian, Tiberius claimed that the free Italian population was being reduced to aporia (poverty) and to oligandria, while the number of agricultural slaves was constantly increasing (êuxêmenôn).155 In this context, the term oligandria can only mean ‘a lack of men’.156 In any case, as John Rich has recently observed, the ambiguity of some of Appian’s language is dispelled by the rest of his exposition, which makes it clear that Tiberius’ fear was that the free population had begun to decline because there were too many poor people who were no longer capable of rearing children.157 The most important conclusion to emerge from this discussion is that Appian’s description of the background to the Gracchan land reforms cannot really be fitted into a high-count interpretation of Italy’s demographic history. As always, this difficulty can be resolved by dismissing the accounts of Appian and Plutarch as distorted rhetorical representations reflecting the propaganda of the Gracchan age and/or the views of later writers eager to demonstrate the noble aims of politicians belonging to the ranks of the populares and the beneficial aspects of land distribution schemes. With every piece of ‘evidence’ discarded, however, the value of the high count as a narrative framework that can be used to make sense of 153 156

157

Ibid. 154 Lo Cascio (2008), 241. 155 App. BC 1.9. While Silvestrini (2004) is inclined to interpret the term dysandria as referring primarily to the dismal quality of life of the free rural population, she takes oligandria as a ‘chiaro accenno al calo demografico’ (2004, 79). Rich (2007), 162.

Developments between 163 bc and 133 bc

179

the few snippets of (no doubt tendentious) information that have been preserved is undermined. At least in the case of the Gracchan land reforms and their background, there can be no doubt that a low-count interpretation incorporating the idea of modest population growth in central-western Italy scores more highly in this context. 4.3.8 Survey archaeology and the Gracchan ‘crisis’ Since the early 1970s, survey archaeology has begun to play an increasingly important part in discussions of the demographic history of republican and early-imperial Italy. Initially it was claimed that the systematic collection of traces of human habitation by teams of field-walkers was an essentially scientific operation that could potentially shed radically new light on the fate of the free peasantry of Roman Italy and on the relative importance of peasant farms and slave-staffed estates during successive periods of republican and imperial history. In the final chapter of this book I shall examine the enormous methodological difficulties surrounding any attempt to use the results of rural survey campaigns as a basis for demographic conclusions. This later discussion will focus upon the capacity of competing models of demographic development to account for trends in rural sites numbers throughout Italy between c. 300 bc and c. ad 100. Here my aim is to say a few words about various attempts which have been made to use the survey data from one particular area, South Etruria, as a starting point for a reassessment of the demographic realities behind the Gracchan ‘crisis’. In what follows I shall briefly examine the degree of fit between the South Etrurian data and competing reconstructions of the demographic background to the Gracchan land reforms, whilst attempting not to anticipate all of the methodological reservations that will be expressed in my general discussion of the capacity of survey archaeology to contribute to the debate between low and high counters. As is generally known, the South Etruria Survey of the British School at Rome was the first project to attempt to illuminate the historical evolution of habitation patterns and levels of population across a large region of mainland Italy. Its principal aim was to reconstruct the history of settlement in large parts of the territories of Veii, Capena and Sutrium by collecting and registering all traces left by various categories of rural inhabitant (mainly pottery) and by assigning these traces to a number of chronological periods. Sherds of grey bucchero were thus interpreted as evidence for occupation during the fifth and fourth centuries bc, while Italian black-gloss and sigillata were used to assign sites to the last three centuries of the Republic

180

Peasants, citizens and soldiers, 201 bc–28 bc

(300 bc–30 bc) or to the early Empire (30 bc–ad 100) respectively. When this method was applied to the Ager Veientanus, it was found that the number of third-to-first century sites (242) was far higher than the number of late-Etruscan ones (127) and that the number of early-imperial sites (327) was higher still.158 Although some of the archaeologists who participated in the South Etruria project were wary of drawing any far-reaching demographic conclusions, it seemed difficult to avoid the impression that, at least in South Etruria, the rural population had ‘exploded’ rather than declined during the final centuries of the Republic. In the field of ancient history, the keenest supporter of this interpretation was Martin Frederiksen, who argued that the data from South Etruria proved the traditional picture of an agrarian and demographic ‘crisis’ caused by the spread of slavery and by heavy military recruitment to be completely false.159 Initially this challenging claim fell onto fertile ground, not only among archaeologists eager to free their discipline from interpretative templates derived from the literary sources, but also among ancient historians looking for new perspectives on Roman demographic history and on the Gracchan land reforms in particular. Unfortunately for those favouring a radical reinterpretation of laterepublican history on the basis of the survey data, it soon became clear that the archaeological evidence collected in South Etruria and in other parts of the peninsula posed far more problems than had initially been realized. In particular, it was felt that a better understanding of developments during the second century bc could be achieved only by breaking up the long period running from 300 bc to 30 bc into sub-periods. Once Morel had refined the chronology of republican black-gloss,160 the Italian archaeologist Liverani was able to carry out a pilot study on some of the material from South Etruria. His startling finding was that about 80 per cent of the black-gloss pottery collected in this area belonged to the fourth-to-second centuries bc and only 20 per cent to the second and first centuries.161 Inspired by this result, a group of British and Italian archaeologists decided to set up a new project aiming to re-study no fewer than 90,000 sherds of pottery from South Etruria. In order to bring developments within the republican period into sharper focus, a new chronological framework was developed, based on a distinction between mid-Republican pottery (350–250 bc) and two categories of late-Republican pottery that were dubbed Late Republican 1 (250–150 bc) and Late Republican 2 (150–30 bc). When this new typology was applied to the data from the middle Tiber 158

Potter (1979).

159

Frederiksen (1970–1971).

160

Morel (1981).

161

Liverani (1984).

Developments between 163 bc and 133 bc

181

valley, it turned out that 513 sites could be dated to the period 350–250 bc, 199 to the period 250–150 bc and 561 to the last 120 years of the Republic.162 At first sight this new pattern suggests a drastic decline in site numbers between 250 bc and 150 bc, and it would be tempting to connect this with the negative picture painted by Appian and Plutarch. There are, however, compelling reasons not to jump to any far-reaching conclusions. It has, for instance, been suggested that the scarcity of sites containing Late Republican 1 pottery may be attributable to the fact that this particular category of black-gloss was less widely circulated than some earlier and later types.163 If this hypothesis is correct, the apparent drop in site numbers must be regarded as an illusion. As Witcher has pointed out, this problem is all the more acute because the average number of black-gloss sherds at each southern Etrurian site was only six. If Late Republican 1 pottery was indeed distributed in smaller quantities than earlier and later types, so small a sample would not permit the conclusion that a given site ceased to be occupied between 250 bc and 150 bc. This suspicion is to some extent confirmed by the fact that about a third of those sites that were supposedly ‘abandoned’ after 250 bc appear to have been ‘reoccupied’ between 150 bc and 30 bc.164 Another way of looking at this problem is to check whether abandoned sites tend to be clustered in marginal areas, as might be expected if the rural population had really gone into decline. Viewed in this light, it is striking that continuing and abandoned sites are interspersed across the area investigated, again suggesting that the apparent downturn of the late-third and early-second centuries bc could be a purely archaeological artefact not corresponding to any genuine demographic development.165 While these methodological caveats make it impossible to translate the large fluctuations in site numbers that can be observed in southern Etruria into demographic trends, it is still possible to draw some interesting conclusions. One of these concerns the claim of some proponents of the high-count theory that the mid-republican and late-republican data from southern Etruria support their scenario of rapid population growth.166 Of course, if we accept the highly problematic premise that trends in site numbers reflect demographic developments, the apparent increase in site occupation after 150 bc can still be interpreted as fully compatible with a 162 163

164

Witcher (2008a), 275. Witcher (2008a), 276–7. Cf. also Rathbone (2008), 328, for the suggestion that the rural population of the late third and early second centuries bc may have preferred to use glass and metalware. Witcher (2008a), 279. 165 Ibid. 166 E.g. Lo Cascio (2008), 240.

182

Peasants, citizens and soldiers, 201 bc–28 bc

scenario of relatively fast demographic recovery after the Second Punic War. Furthermore, if we prefer to emphasize the importance of distorting factors, it is possible to interpret the archaeological record as pointing to a large degree of occupational and demographic continuity. The rub is that it is extremely difficult to see how the fluctuations in the visible sites that can be observed within the republican period can be made to support Lo Cascio’s view that the Gracchan land reforms were prompted by a Malthusian overpopulation crisis. In fact, the only archaeological ‘fact’ that can realistically be claimed to support the high count is the spectacular rise in site numbers after 30 bc, a phenomenon far too late to be of any relevance to Lo Cascio’s reinterpretation of the social and demographic background to the Gracchan land reforms. 4.4 developments between 133 bc and 28 bc Six census figures have been preserved for the period 133 bc–28 bc: those for 131/130 bc, 125/124 bc, 115/114 bc, 86/85 bc, 70/69 bc and 28 bc. Since the first three figures have a direct bearing on our understanding of the demographic background to the Gracchan land reforms, they have already been dealt with in the previous sections. The assumptions that must be made in order to fit the last three figures into the general frameworks provided by the low and high counts have been examined in my chapter on the meaning of the census figures. This leaves us with the task of examining the capacity of the low and high counts to make sense of three developments that feature prominently in almost all accounts of the final century of the Republic. The first of these is Marius’ decision to disregard the property qualification for legionary service during the levy of 107 bc. Although many recent publications play down the short-term and long-term effects of this episode, the question of how it can be fitted into the competing demographic theories remains an interesting one. The second and third topics to be examined are emigration from Italy and urbanization. It will immediately be apparent that the high-count theory can easily accommodate the settlement of tens of thousands of veterans in provincial colonies and the growth of towns. At first sight, it is less easy to see how large-scale migration from rural areas to towns and from Italy to the provinces can be reconciled with a model that assigns only about 6 million inhabitants to late-republican Italy. Why, for instance, should so many people have left mainland Italy if the population of this area was only half as large as it would become in the early fourteenth

Developments between 133 bc and 28 bc

183

and late sixteenth centuries ad? Moreover, does not a low-count model that incorporates urban growth and large-scale emigration necessarily imply a considerable imbalance of the Italian population between rural and urban areas? 4.4.1 Marius and the proletarianization of the legions Traditional accounts that begin from the assumption that the second and first centuries bc witnessed a steady decline in the number of free countrydwellers have identified this hypothetical downward trend as the main reason behind Marius’ decision to disregard the property qualification for military service when he needed to enlist soldiers for his campaign against Jugurtha.167 The idea underlying such accounts is that the impoverishment of large sections of the free rural population and the replacement of free peasants by slaves had eroded Roman manpower resources by diminishing the number of assidui available for military service. Marius’ decision to enrol proletarians apparently resolved this problem, but also created a new one, since poor legionaries expected to be rewarded with land following their term of service. In recent years, this overall interpretation has been undermined in several ways. It has been pointed out, for instance, that Marius’ approach to the levy was not immediately copied and that the large-scale recruitment of proletarians is unlikely to have become common before the Social War.168 In other words, between 107 bc and 91 bc the property qualification remained important. A highly relevant point is that the surviving evidence clearly demonstrates that throughout the final decades of the Republic large numbers of men were enlisted as conscripts rather than as volunteers.169 Finally, and perhaps most importantly, the theory that Marius’ action was prompted by a shortage of assidui founders on the fact that the property qualification for legionary service was already very low during the final decades of the second century, perhaps even as low as the notional value of a garden plot and a hut.170 It would seem to follow from this that many impoverished peasants were assidui and that the abandonment of the property qualification did not greatly affect the social composition of the legions.171 167 168 170 171

E.g. Gabba (1949/1976), 16–19; Brunt (1971/1987), 407; Hopkins (1978), 30–1. Rich (1983), 329. 169 E.g. Brunt (1962/1988), 255–6. Above at note 145. Cf. Brunt (1971/1987), 405–6. Cf. Keaveney (2007), 25, for the observation that in the armies of the first century bc proletarian soldiers appear to have formed no discernible interest group or lobby with aspirations different from those of their fellow legionaries.

184

Peasants, citizens and soldiers, 201 bc–28 bc

Given the many weaknesses surrounding the traditional interpretation of the military ‘reform’ of 107 bc, it is tempting to conclude from Sallust’s statement that the Senate reckoned that the conscription of even a few thousand legionaries would undermine Marius’ popularity, that Marius saw this danger a mile off and successfully avoided it by enlisting every man who was willing to serve.172 Alternatively, we could adopt Rich’s suggestion that Marius disregarded the property qualification because his proposed campaign against Jugurtha had aroused general enthusiasm, so that he did not want to disappoint any of those who were eager to serve under his command.173 Regardless of these interpretational problems, the fact that the surviving sources contain no reliable information about the demographic and social background of the dilectus of 107 bc means that Marius’ action is compatible with a variety of demographic scenarios. These include the orthodox theory of demographic decline, a revised low count assuming slow but steady population growth in central-western Italy, and also the Malthusian model of widespread rural impoverishment implied by the high count. This makes it difficult to use the levy of 107 bc as an argument in favour of any particular reconstruction of the demographic fate of the free population of the Ager Romanus, or, for that matter, of any other part of Italy. 4.4.2 Emigration from Italy At the time of the census of 28 bc many hundreds of thousands of Roman citizens lived in the provinces, most of them in Spain, southern Gaul and northern Africa. Before we examine the possible demographic implications of this phenomenon, it must be emphasized that the backgrounds of these provincial citizens varied enormously. It is clear, for instance, that Sertorius, Pompey, Caesar and Octavian bestowed Roman citizenship on large numbers of supporters of provincial extraction.174 Unfortunately, the surviving evidence does not allow us to estimate the number of non-Italians who became citizens in this way. Despite this difficulty, the majority view is that most citizens living in the provinces in 28 bc were people of Italian origin or their descendants.175 It is also thought that the 70,000 or 80,000 civilian colonists sent out by Caesar and the tens of thousands of veterans who 172 173

174 175

Sall. Iug. 84.3. Rich (1983), 326. Yet another interpretation in Evans (1994), 75. In his view, the primary motive for enlisting proletarians was simply that Marius was in a hurry and the usual method of levying troops would have taken too much time. The best recent discussion is that of Crawford (2008). For some tentative figures, see Appendix iv.

Developments between 133 bc and 28 bc

185

received provincial allotments between 49 bc and 28 bc accounted for a large proportion of all provincial citizens with Italian roots.176 Although the high counters do not seem to have paid much attention to the matter of migration from Italy, there can be no doubt that their favoured reconstruction provides them with a plausible explanation for it. If late-republican Italy was really as crowded as they think, it makes sense to suppose that intense competition for land and other resources would have persuaded large numbers of Italians to try their luck elsewhere, as traders, as investors in the Spanish mines, and especially as recipients of allotments of fertile land in civilian or veteran colonies. At first sight, large-scale emigration from Italy makes less sense from a low-count point of view. The obvious reason for this is that the Italy of the low count is strikingly under-populated compared to Italy in the early fourteenth or the late sixteenth century ad. Even if we ignore the enormous problems posed by the concept of ‘carrying capacity’, there can be no doubt that mainland Italy could easily have supported a population much larger than the 6 million that the low counters assign to this area during the final decades of the Republic. How, then, do we explain the fact that Caesar felt compelled to settle tens of thousands of proletarians in the provinces, and why did neither he nor any of those involved in the power struggles of the period 44 bc–31 bc manage to settle all of their veteran soldiers on unused land in Italy? In my view, a satisfactory answer to these questions can be found by looking at the amount of state-owned land available for distribution in Italy. In 81 bc Sulla managed to settle between 80,000 and 120,000 men on Italian soil.177 According to Appian, ‘he allotted to the twenty-three legions which had fought for him a large quantity of land belonging to the towns, some of which had never been distributed for cultivation and some of which was taken from them as penalty’.178 The said ‘land belonging to the towns’ may have been ager publicus populi Romani regarding which these towns had received the ‘right of enjoyment’ (ius fruendi).179 Alternatively, it may have been municipal land to which the Roman state had no vestigial rights. In any case, we are told that some cultivated land was also taken away, and it appears from other sources that the amount of land available for allocation was augmented by the confiscated properties of Sulla’s political enemies. 176

177

According to Suet. Caes. 42.1, Caesar settled 80,000 citizens in the provinces. This figure may, however, have included veterans (Brunt 1971/1987, 257). For settlement of veterans, Keppie (1983) remains fundamental. E.g. Santangelo (2007a), 147. 178 App. BC 1.100. 179 Cf. Roselaar (2010), 138.

186

Peasants, citizens and soldiers, 201 bc–28 bc

It has plausibly been inferred from this information that the amount of arable land owned by the state was much reduced as a result of this very ambitious programme of colonization.180 It is true that some arable ager publicus was still available in the late 60s bc, when Rullus proposed his agrarian law, and during his consulate Caesar pushed through a law prescribing the division of the Ager Campanus, large parts of which were held by long-lease tenants. Even in the mid-40s bc Caesar seems to have managed to find some tracts of undistributed public land in Italy that were suitable for farming.181 By the time of the triumviri, however, very little state-owned land was left for distribution to would-be farmers. In theory, this problem could have been solved by buying up tracts of land in thinly populated parts of Italy. As various ancient historians have pointed out, however, the large number of expectant recipients, the size of the plots promised and the price of Italian land meant that this was not a realistic option. In 43 bc Antony, Octavian and Lepidus found a simple solution. To spur on their armies, they selected eighteen Italian towns famous for their fine land and houses and promised to give this land to their men after the successful conclusion of their campaign against Brutus and Cassius.182 When this ruthless scheme was implemented in 42 bc and 41 bc, however, the huge political costs became apparent. Fearing that the strategically located towns of Vibo and Rhegium would go over to Sextus Pompeius, Octavian promised the inhabitants that he would remove their towns from the list. Soon afterwards, the remaining sixteen towns and a large number of other towns that feared like treatment pledged their support to Lucius Antonius, drove out those who were borrowing money from the temples for Octavian and manned the walls. In the end, a new civil war was only barely avoided.183 This sequence of events recalls that of the early 70s bc, when many of those dispossessed by Sulla joined Lepidus’ revolt.184 In short, during the final decades of the Republic the previous liquidation of almost all of the arable land previously owned by the Roman state had left the state in a position from which it could obtain suitable land in Italy only by means of purchase or confiscation.185 The former option entailed costs that were economically prohibitive; the latter course was politically dangerous. Settlement in thinly populated areas provided a partial solution 180 184 185

Ibid. 276–7. 181 Keppie (1983), 49. 182 App. BC 4.3. 183 BC 4.86; 5.19; 5.27. BC 1.107. Arable land belonging to Italian towns could also be acquired by exchanging it for provincial land or for land in other parts in Italy (e.g. Keppie 1983, 70–1; Biundo 2004; Roselaar 2010, 143), but this does not seem to have been done very often.

Developments between 133 bc and 28 bc

187

at best, since the disappearance of the arable ager publicus logically implied that all remaining farmland was either under private ownership or controlled by some town.186 Moreover, the promises made by the triumviri in 43 bc suggest that most veterans would not have accepted allotments in marginal areas. What they wanted was large plots of high-quality land in attractive locations.187 Given all of this, the decision to settle tens of thousands of civilian colonists and veterans outside Italy makes excellent sense. Large amounts of fertile land could easily be taken away from provincial communities at a very low political cost.188 Moreover, since provincial land was much cheaper than Italian land, large tracts of it could be snapped up at bargain prices.189 It can therefore be argued that, rather than pointing to high population pressure in Italy, the large-scale emigrations of the first century bc simply reflect a growing shortage of public land in Italy and the huge political and economic costs of any attempt to interfere with the property rights of Italian landowners. 4.4.3 Emigration, urbanization and the decline of the free population In an article published in 1994, Lo Cascio calls attention to an important weakness in the low-count models put forward by Brunt and Hopkins. Focusing on the convenient synopsis in the first chapter of Hopkins’ Conquerors and Slaves, he notes that Hopkins is asking his readers to believe that between 225 bc and 28 bc the free rural population of mainland Italy declined from c. 4.1 million to c. 2.9 million, while its free urban population went up from c. 400,000 to c. 1.1 million.190 In Lo Cascio’s view, this model of population development sits very uneasily with Esther Boserup’s finding that in pre-industrial societies no significant urbanization of any particular territory is possible unless it is accompanied by sustained growth in the rural population.191 This suggests that the low count implies a long-term change 186

187

188 189 191

In theory, it should have been possible for the triumviri and later for Octavian as sole ruler to buy up unused land (e.g. in marshy areas) at a low price and to make this land suitable for cultivation. Possible examples of this approach include the Piana delle Cento Fattorie near Lucca (Marzano 2007, 177) and the reclamation programme carried out in the Valli di Comacchio (Kron 2005, 480; cf. perhaps Tac. Ann. 1.17). Marshy land could not, however, be reclaimed overnight, and the veterans of the 40s and 30s bc were not prepared to wait. It is striking that no Caesarian or triumviral settlements are recorded in Apulia, despite the fact that this was the most thinly populated part of Italy (Cic. Att. 8.3.4). For confiscations of provincial land, see e.g. Fear (1996), 74–5, 79–80; Osgood (2006), 144. MacMullen (2000), 132. 190 Lo Cascio (1994), 38–9. Cf. Chapter 1, Table 1.1. See, however, my remarks in note 201.

188

Peasants, citizens and soldiers, 201 bc–28 bc

in the balance between rural and urban populations that is deeply implausible from a comparative point of view. When we take a closer look at Hopkins’ reconstruction, we discover that he regards this apparent decline in the free rural population as having been almost completely offset by a simultaneous expansion of rural slavery. At the same time, his model posits a rapid increase in the number of urban slaves. Taking all of this into account, we end up with a scenario in which the free and slave population of the Italian countryside declines from c. 4.35 million to c. 4.1 million, rather than from c. 4.1 million to c. 2.9 million.192 Lo Cascio’s claim that the low-count theory implies a ‘collapse’ of Italy’s rural population is therefore certainly exaggerated. On the other hand, it remains the case that the particular version of the low count championed by Hopkins posits a somewhat curious combination of strong urban growth and modest rural population decline. Although Hopkins does not explicitly confront this difficulty, he does identify a number of economic developments that made it possible for the urban population to expand at a much faster rate than its rural counterpart. One of these developments was an enormous increase in the amounts of grain imported from the provinces.193 Hopkins also made the interesting observation that the displacement of free peasants by slaves seems to have had the effect of increasing productivity. He supports this assertion by pointing out that according to the Roman agronomists the amount of land needed to support twenty peasant families could effectively be worked by only eight adult male slaves, the main reason for this striking discrepancy being that while free Roman peasants were under-employed, slaves could be forced to work long hours throughout the year.194 In short, the spread of rural slavery made it possible to produce a larger surplus with a much smaller workforce. Even if we take these important observations into account, the combination of fast urban growth and considerable rural decline indicated by 192

193

194

The figure of 4.35 million for 225 bc is based on the assumption that about half of the 500,000 slaves posited by Hopkins worked in the countryside. Hopkins (1978) 38–9, 73. In a recent article, Lo Cascio suggests that c. 700,000 consumers in Rome may have been fed on provincial grain during the time of Augustus (Lo Cascio 2009, 103). Elsewhere (Lo Cascio 1999b, 239–40) he claims that many of the towns on the Tyrrhenian coast also depended on imports. If these ideas are accepted, the problem of the seemingly excessive urbanization rate implied by Hopkins’ estimates largely disappears. Hopkins (1978), 24, 106–9, followed by Scheidel (2005a), 71. Hopkins’ arguments show that he is thinking in terms of an increase in per capita output rather than an increase in output per hour worked. Cf. Rathbone (1981) for the observation that the availability of a large pool of free peasants who could be hired during the harvest season made it possible to run slave-operated estates with a smaller permanent workforce.

Developments between 133 bc and 28 bc

189

Hopkins’ calculations remains at least a little awkward. It is, however, of the utmost importance to remember that the model that Lo Cascio sets out to attack is only one of many possible low-count reconstructions. At this point, I want to call attention to three basic uncertainties that have led me to propose an alternative version of the low count positing an increase of about 8.5 per cent in the number of free and non-free Italian country-dwellers between 225 bc and 28 bc.195 Let me begin by pointing out that the basis for Hopkins’ estimate of 4.5 million for the free Italian population in 225 bc is very fragile. As we saw in Chapter 2, there may well have been no more than 3.9 million free Italians on the eve of the Second Punic War.196 Since more than 90 per cent of the free population of pre-Hannibalic Italy is thought to have been engaged primarily in agricultural occupations, this downward adjustment translates into a much smaller decline in the free rural population. A second uncertainty stems from the fact that it is extremely difficult to estimate the total number of town-dwellers in late-republican and earlyimperial Italy, or the number of people primarily engaged in ‘urban’ occupations. In Chapter 5 it will be suggested that Hopkins’ estimate for the urban population of Italy in 28 bc may be a bit on the high side.197 A smaller number of townspeople translates into a larger rural population and also into a more balanced distribution of the Italian population between town and country (or between urban and rural occupations). Thirdly, and finally, Hopkins’ calculations are based on Brunt’s theory that in 28 bc there were about 5 million people of citizen status, of whom about 1 million lived in the provinces. At least some of the complicated arguments on which the latter figure is based can, however, be challenged. For instance, while it seems certain that many free Italians settled in the provinces during the last century of the Republic, Brunt’s suggestion that their number stood at c. 150,000 in 49 bc cannot be verified.198 Similarly, both the sizes of the provincial colonies founded between 49 bc and 28 bc and the numbers of enfranchised non-Italians in these colonies are disputed.199 Furthermore, there is no reliable evidence for the widely held theory that the wives and children of all those Roman citizens who happened to live in peregrine 195

196 199

Since I do not want to burden the readers of this book with a long exposition of technical arguments, my treatment here is deliberately superficial. For a full discussion of the strengths and weaknesses of Hopkins’ estimates, most of which are identical to those of Brunt, the reader is referred to Appendix iv. Chapter 2, at note 129. 197 Chapter 5, at note 173. 198 Brunt (1971/1987), 262. While Brunt (1971/1987), 262, reckons with 3,000 new settlers per colony, MacMullen (2000), 132, believes that a figure of between 1,500 and 2,000 is more realistic. Fear (1996), 93–100, points out that there is no good evidence for the liberal bestowal of citizenship upon indigenous inhabitants of provincial colonies.

Peasants, citizens and soldiers, 201 bc–28 bc

190

communities rather than in provincial coloniae or municipia had been given citizenship by the time of the census of 28 bc.200 In view of these uncertainties, it can easily be maintained that no more than 800,000 citizens lived in the provinces in 28 bc and that the free population of mainland Italy (excluding foreigners) stood at 4.2 million rather than at 4 million. This makes it possible to assign a further 200,000 free inhabitants to the Italian countryside, if all other variables are left unchanged. If these critical observations are combined, we end up with an alternative low-count model that looks like this:

Table 4.3 Alternative low-count model for the late Republic A – Population changes, 225 bc–28 bc Men, women and Adult males children (aged 17+ years) Free Slave Total

225 bc 3,900,000 300,000 4,200,000

28 bc 4,200,000 1,500,000 5,700,000

225 bc 1,180,000 90,000 1,270,000

28 bc 1,270,000 455,000 1,725,000

1,060,000 45,000 45,000 75,000 45,000 1,270,000

940,000 240,000 215,000 150,000 180,000 1,725,000

B – Rural/urban split Rural free Rural slaves Urban slaves Italian towns free City of Rome free Total

3,500,000 150,000 150,000 250,000 150,000 4,200,000

C – Migration from Italy overseas Adult males (aged 17+ years) Before 69 bc 69–49 bc 49–28 bc Sub-total: Killed/counted twice over: Net migration

200

85,000 15,000 120,000 220,000 40,000 – 180,000

3,100,000 800,000 700,000 500,000 600,000 5,700,000

D – Decline of free rural population Adult males (aged 17+ years) Natural growth

160,000

Migrants 225–28 bc to provinces: to Italian towns: Sub-total: Total loss:

180,000 100,000 280,000 120,000

Brunt’s arguments concerning the enfranchisement of the wives and children of Roman citizens sent out to provincial colonies and those of provincials incorporated into provincial municipia (1971/1987, 263–5) do not apply to citizens in peregrine communities.

Conclusions

191

Perhaps the most striking feature of this alternative reconstruction is that it implies a significant increase in the size of Italy’s rural population, which now rises from about 3.65 million in 225 bc to 3.9 million in 28 bc.201 If we take into account the increase in the number of rural slaves, we shall still end up with a decline in the free rural population. While the older model suggests a decline of about 30 per cent in this group, the new version implies a reduction of only 11.4 per cent. Although a decline of this order may still seem significant, it can easily be accounted for if we consider the rapid expansion of the city of Rome and the settlement of about 120,000 adult male Italians in provincial colonies between 49 bc and 28 bc. I do not, of course, claim that this alternative model can be proved to be a more reliable guide to the demographic history of Italy than the estimates of Brunt and Hopkins. My point is merely that while every version of the low count necessarily implies some measure of demographic decline in the free rural population of republican Italy, it is impossible to make any firm statement about the quantitative importance of this trend. We must at all times remain alert to the possibility that the fate of the free rural population varied geographically, from region to region, and also chronologically, from period to period. It could be suggested, for instance, that precisely because migration from rural areas to towns and from Italy to the provinces accelerated during the final decades of the Republic, this was the only period ever to witness a significant decline in the free rural population of Italy as a whole. 4.5 conclusions As we saw at the very beginning of this chapter, the emergence of the high count as an alternative interpretative framework for late-republican history makes it impossible to accept the census figures for the second and third centuries bc as a solid basis for inductive conclusions. It is nonetheless important that any adherent of either of the mutually exclusive reconstructions of Italian population history should examine the assumptions about the surviving ‘evidence’ required by his or her preferred interpretation. In my view, at least some of the problems raised by the traditional lowcount reconstruction of the demographic, social and economic history of central-western Italy during the last two centuries of the Republic cannot be 201

The rural growth rate of c. 6.8 per cent implied by these figures falls dramatically short of the urban growth rate of c. 227 per cent indicated by my estimates for the urban population, but since a large proportion of the urban population of 28 bc can be presumed to have been sustained by imported grain (Chapter 5, at note 174), this discrepancy is unproblematic. Cf. Scheidel (2005a), 71, n. 51, and (2008), 33, for the observation that between 1600 and 1750 the urban population of England increased by 260 per cent, compared to 20 per cent rural growth. Cf. Wrigley (1987), 162.

192

Peasants, citizens and soldiers, 201 bc–28 bc

resolved unless we accept that the Roman citizen body continued to expand throughout the second century bc. Once this idea is taken on board, all of the surviving evidence can be accounted for without resorting either to arbitrary textual emendations or to contrived interpretations. We have observed that, unlike the low count, the high count sits uneasily with the republican census figures, which suggest that there were no more than 400,000–435,000 adult male citizens during the final decades of the second century bc. This problem can, however, be circumvented by assuming these data to be hopelessly unreliable. We have also seen that the expansion of Rome and the growth of many smaller towns during the final decades of the Republic, as well as the settlement of at least 100,000 adult male Italians in provincial colonies, are compatible with the high count, but can also be fitted into a low-count reconstruction of the demographic history of mainland Italy. If we ignore the census figures and focus on what the literary sources have to say about the demographic and social history of Italy between 201 bc and 28 bc, it seems fair to conclude that the high counters have failed to come up with a reconstruction of the background to the Gracchan land reforms that does justice to the tradition represented by Appian and Plutarch. As we have seen, this tradition could indeed provide us with a distorted picture of conditions in the Italian countryside. Nonetheless, it seems certain that Tiberius Gracchus thought he could convince most of his fellow citizens that the free population had begun to decline. While Tiberius’ understanding of the relationship between poverty and demographic developments may have been imperfect, it would surely have been easy for his opponents to dismantle his arguments had the number of adult male citizens indeed risen from c. 485,000 in 225 bc to c. 730,000 in 133 bc. In the case of the archaeological data, the only safe conclusion is perhaps that, precisely because the interpretation of the republican material from southern Etruria is far more complicated than was realized until the early 1980s, this particular data set cannot be used to set aside the census figures and to discard the literary tradition concerning the Gracchan land reforms in favour of a model of fast population growth in which the only real problem was a lack of men with the right physique to qualify for service in the legions. At the same time, it remains possible to interpret the dramatic increase in site numbers after 30 bc as supporting the view that, at least during early-imperial times, population densities in southern Etruria may perhaps have reached the sort of level predicted by the high count. This topic will be examined in the final chapter of this book, which will look at a much larger set of rural survey data from the perspective of the Augustan census figures.

chapter 5

The Augustan census figures and Italy’s urban network

5.1 introduction In the previous chapters we have seen that adherents of the high count encounter considerable problems in interpreting certain pieces of textual and archaeological evidence. For the low-count theorists, the only problematic pieces of evidence are the Augustan census figures. I have tried to demonstrate that comparative data regarding population levels in medieval/early-modern Italy, along with general considerations concerning plausible rates of population growth, support the low-count interpretation of these figures. Since some ancient historians may be inclined to attach more weight to philological arguments than to comparative data, however, it is important that we should look for additional types of evidence that may corroborate this conclusion. We have seen that the stalemate that seems to have been reached in the existing literature cannot be broken by estimating the ‘carrying capacity’ of the Italian countryside, or by focusing on climatic or economic change. How, then, can the debate between low counters and high counters ever be resolved? This chapter aims to shed new light on the demographic history of Italy during the late Republic by looking at the shape of Italy’s urban network as it had become by the time of the census of 28 bc. In adopting this approach, I am not by any means suggesting that the archaeological evidence available to us allows us to gauge the size of the urban population of Roman Italy by means of inductive argument. In fact, my examination of a large amount of archaeological material, and of the recent literature concerning the methodological problems posed not only by the incomplete evidence supplied by excavations, but also by urban survey archaeology, has convinced me that any attempt to reconstruct ancient population figures on the basis of the surviving archaeological evidence is doomed to failure1 I have, nonetheless, 1

Cf. Chapter 6 for the very similar problems besetting the interpretation of rural survey data.

193

194

Augustan census figures and Italy’s urban network

come to the conclusion that archaeology holds the key to the solution of the demographic debate. In the sections that follow, I shall try to make good this claim by examining the geographical distribution of the Italian population implied by the low and high counts, the physical size of towns and levels of urbanization. 5.1.1 The geographical distribution of the population In recent attempts to clarify the geographical distribution of the free and non-free population of Italy in late-republican and early-imperial times, Cisalpine Gaul has frequently been the centre of attention. As has often been pointed out, the low-count interpretation of the republican and Augustan census figures implies that this particular part of Italy was (by comparison with its population in the High Middle Ages and during the early-modern period) thinly populated at this time. If one accepts the low-count interpretation of the census figure for 70/69 bc, the number of adult male citizens living south of the Po (and in those parts of Transpadana to which Roman or Latin colonists had been sent) cannot have been much higher than 910,000, implying a free population of about 3.6 million for central and southern Italy plus Cispadana.2 If we take into account the massive wave of emigration to the provinces that took place between 49 bc and 28 bc and interpret the census figure for 28 bc as pointing to a free Italian population of about 4.2 million,3 it follows that approximately 900,000 people, corresponding to c. 275,000 adult males, must have been enfranchised as a result of the bestowal of citizenship upon the free population of Transpadana in 49 bc.4

2

3

4

910,000 x 1.2 x 3.3 = 3,603,600 (on the assumption that the census of 70/69 bc missed about 20 per cent of the target population). A lower registration rate would imply a larger population, but in estimating the size of the free Italian population we must also reckon with the presence of numerous citizens domiciled in the provinces. The view that most soldiers serving in armies operating outside Italy remained unregistered (e.g. Brunt 1971/1987, 97) is without foundation, except insofar as these soldiers had not been registered in the census of 86/85 bc. See Chapter 3, section 3.2.3. Cf. Chapter 4, Table 4.3. Even if we scale down Hopkins’ estimate of 165,000 for the number of adult male Italians who moved to the provinces between 49 bc and 28 bc to 120,000, it remains difficult to avoid the conclusion that at least 300,000 Italian men, women and children of citizen status left mainland Italy during this period. If we subtract this figure from a hypothetical Italian citizen population of 3.6 million in 69 bc and add 0.9 million free Transpadani, we obtain a rough estimate of 4.2 million free Italians for 28 bc. Needless to say, all of these hypothetical figures come with a large margin of error. My only aim is to clarify the connections between estimates referring to different points in time. Cf. Brunt (1971/1987), 202. Brunt also envisages the possibility that there were only some 250,000 adult male citizens in Cisalpina as a whole at the time of the census of 28 bc. With 4.2 million citizens in mainland Italy, this estimate would imply a free population of about 3.4 million for the centre and south.

Introduction

195

In order to obtain an estimate for Cisalpine Gaul as a whole (that is, for Transpadana and Cispadana), we must add on the population of the region of Aemilia, the inhabitants of a handful of Transpadane towns that had already acquired citizenship before 49 bc and the non-free inhabitants of the four northern regiones.5 Taken together, the people belonging to these categories must have numbered at least half a million.6 At the same time, the logic of the low count requires that their number should not have exceeded 1 million, for the simple reason that a northern population of more than 2 million would imply an implausibly low population for central and southern Italy, especially if we assign between 800,000 and 1 million inhabitants to the city of Rome. In short, while the low counters must assign to Cisalpine Gaul a free and non-free population of at least 1.5 million, they cannot give it more than 2 million inhabitants. With a lowcount population of c. 5.7 million for mainland Italy as a whole, this would allow early-imperial Cisalpina between 26.3 and 35.1 per cent of the total Italian population.7 The high counters would argue that both the absolute figures and the relative shares implied by the low count are far too low. As Geoffrey Kron has pointed out in a recent article, throughout the early-modern period northern Italy never contained less than 43 per cent of the population of mainland Italy.8 This would seem to suggest that all figures falling within the range of estimates suggested by the low counters are unrealistically low. In Kron’s view, a far more convincing picture emerges if we follow Lo Cascio in assigning to early-imperial Italy a total population of c. 15 million. With 43 per cent of this population in the north, we then obtain a rough estimate of c. 6 million Cispadani and Transpadani at the time of the census of 28 bc. This would leave c. 9 million for the centre and south. 5

6

7

8

There were also some free non-Romans in the territories of certain communities that had citizenship in or shortly after 28 bc. Here one thinks of the Salassi incolae in the territory of Augusta Taurinorum (ILS 6753). The evidence for such groups is, however, limited, except in the attributed territories of the Alpine valleys, which play no part in the low count for Cisalpine Gaul. Cf. below, at notes 115–16, for the suggestion that the region of Aemilia is likely to have contained about a third of the northern Italian population at the time of Augustus. If only 750,000 northern Italians were enfranchised in 49 bc, if those who were already citizens before that date numbered 375,000 and if slaves made up 15 per cent of the population, Cisalpine Gaul would have had only c. 1.3 million inhabitants, but this alternative estimate seems too low. As elsewhere in this book, the attributed tribes of the sub-Alpine regions have been excluded from this calculation. With between 100,000 and 200,000 attributi in an area of between 20,000 to 25,000 square kilometres (Beloch 1886, 321, 434), the population of early Augustan Italy rises to between 5.8 million and 5.9 million, and Cisalpina’s share to at least 27.6 per cent. Kron (2005a), 462. Lo Cascio’s estimates for 225 bc (Lo Cascio and Malanima 2005, 201) imply that between 35 and 45 per cent of the population of mainland Italy lived in Cisalpine Gaul during the late third century bc. Cf. De Ligt (2008), 164.

196

Augustan census figures and Italy’s urban network 5.1.2 Italy’s urban network at the time of Augustus

The low and high counts suggest not only completely different overall population levels, but also widely diverging ideas of the size of urban agglomerations and of the contours of Roman Italy’s urban network. The most popular version of the low count, generally associated with the name of Keith Hopkins, assigns 0.9 million inhabitants to the city of Rome and posits the existence of another 1 million townspeople throughout the rest of Italy. This would imply an urbanization rate of 31.7 per cent (if Rome is included) or 20 per cent (if it is not).9 As Hopkins himself has admitted, however, Rome may have had no more than 800,000 inhabitants at the time of the census of 28 bc,10 and his estimate for the number of Italian townsmen outside Rome remains pure guesswork. A further complication is that his hypothetical figure of 1 million for this category refers to the number of people primarily engaged in ‘urban’ occupations rather than to the number of people living in large or mediumsized population centres.11 This means that his definition of ‘urban’ excludes town-based farmers, but includes rural craftsmen and the non-agrarian populations of very small towns. As we shall see in a later section of this chapter, most studies dealing with the urban populations of late-medieval and early-modern Italy exclude all centres with fewer than 5,000 or 3,000 inhabitants and ignore country-dwellers engaged in ‘urban’ occupations. Hopkins’ approach therefore makes it difficult to compare Roman levels of urbanization with those calculated for later periods of Italian history. If we use a uniform definition based on settlement size, the urbanization rates implied by the low count are significantly lower than Hopkins’ estimates would suggest. For instance, if we assign some 700,000 people to settlements with 3,000 or more inhabitants, ignore all towns whose populations fall short of this critical threshold and keep all other figures unchanged, the urbanization rate drops to 26.6 per cent if Rome is included and to only 11.7 per cent if it is not.12 Few have attempted to explain the high level of urbanization implied by the theories of Beloch, Brunt and Hopkins. One notable exception is Jongman, who has constructed a model that explains the high number of town-dwellers in Roman Italy as reflecting high levels of elite income and expenditure. According to him, the combined annual incomes of the Italian elite (that is to say, the senators, the equites and the decurions) must have 9 12

Hopkins (1978), 68–9. Cf. Chapter 1, Table 1.1. Cf. Table 4.3.

10

Hopkins (1978), 98.

11

Ibid. 69, note e.

Introduction

197

been enough to feed some 2.4 million people. In short, elite expenditure would have been more than enough to feed the joint populations of all of the Italian cities, including Rome.13 For our purposes, the most interesting aspect of Jongman’s approach is perhaps that it establishes a causal link between a high rate of urbanization and a consistent pattern of elite residence in towns. This is in line with the findings of the many specialists in medieval Italian history who have identified the habit amongst Italian landowners of living in towns as one of the factors that made it possible for between 17 and 21 per cent of the population of Renaissance Italy to live in towns with more than 5,000 inhabitants.14 Although there are conspicuous differences between the urban networks of Roman Italy and those of late-medieval Italy, it seems reasonable to conclude that the urbanization rates implied by the low count are not unrealistically high.15 I hasten, nevertheless, to add that this by no means proves the low count to be correct. Interestingly, the high counters share Hopkins’ view that urbanization levels in early Augustan Italy more or less matched those reached in medieval and early-modern times. To be more precise, they argue that the towns of Roman Italy were much larger than the low count would suggest and that this simple fact is enough to shed considerable doubt on the viability of the demographic model championed by Beloch, Brunt and Hopkins. One example of the high-count approach is Lo Cascio’s observation that many Italian towns covered 60 or more hectares. His examples include the Etruscan towns of Caere (120 ha) and Volsinii (80 ha), the Campanian towns of Capua (180 ha) and Neapolis (70–100 ha) and the northern towns of Bononia (80 ha) and Patavium (more than 85 ha).16 Lo Cascio also interprets the relatively abundant epigraphical evidence for distributions of food and money to the inhabitants of various Italian 13

14

15

16

Jongman (1988), 192–8. Note that Jongman’s estimate of annual subsistence costs (HS 115) is on the low side. On the other hand, the figure of HS 380 suggested by Frier (1993) is almost certainly too high for smaller Italian towns. For the early Roman empire as a whole, Scheidel and Friesen (2009), 68–9, base their calculations on a ‘bare bones’ subsistence basket of 390 kg of wheat equivalent. With hypothetical wheat prices of 6 HS per modius for Rome and 4 HS per modius for the remaining towns of Roman Italy (Duncan Jones 1982, 345–6), this would imply annual subsistence costs of between c. 235 HS and c. 350 HS. The best recent discussion of urbanization rates in medieval and early-modern Italy is Malanima (2005). For the significance of elite residence in towns, see e.g. Jones (1974), 1,679–81; Britnell (1991), esp. 26–30. Another factor was the pivotal role of some towns in a thoroughly commercialized economy. The presence of a very large imperial capital pushed up the urbanization rate of Roman Italy as a whole. Without Rome, urbanization rates might have fallen short of late-medieval and early-modern levels. Lo Cascio (1999a), 165. Note that his estimates for Caere and Volsinii are too high for the Roman period. His estimate for Patavium is, however, on the low side. See Appendices i and ii.

198

Augustan census figures and Italy’s urban network

communities of the early Empire as proving that towns such as Pisaurum, Comum and Spoletium had populations of between 9,800 and 17,850.17 In his view, such large towns cannot possibly have been sustained by the small rural population postulated by the low count.18 In a richly documented article, Geoffrey Kron has attempted to buttress Lo Cascio’s conclusions concerning the shape of the urban network of earlyimperial Italy by examining the towns of Cisalpine Gaul. Drawing on various recent publications concerning the development of the urban network of northern Italy during the late-medieval and early-modern periods, he calls attention to the fact that in late-medieval times northern Italy had only fiftythree towns with more than 5,000 inhabitants.19 As he points out, this is significantly fewer than the seventy-eight or eighty-two northern Italian settlements that enjoyed urban status during the early Empire. The only possible conclusion seems to be that if indeed early-imperial Cisalpina had more substantial towns, it must also have had a population matching or even surpassing the 6.3 million inhabitants of the early fourteenth century.20 As we saw in an earlier chapter, Kron has also given a new twist to the discussion of levels of urbanization by arguing that the existence of many large towns in early-imperial Italy does not merely suggest high overall population levels, but also helps to explain how it could have been possible to feed most of the large Italian population of the high-count theory entirely on Italian crops.21 He argues that there is good reason to believe that many Roman farmers combined arable cultivation and the raising of livestock in a ley-farming system. According to him, the large amounts of manure available on such farms would have made it possible for Roman farmers to achieve the high crop yields required to sustain the high population implied by the high-count reading of the Augustan census figures. As Kron explains, however, the viability of this system would have depended on the existence of a large urban market for meat. His challenging views on the levels of land productivity achieved in Roman agriculture are therefore intimately linked to the high-count theory that the aggregate size of the urban population of earlyimperial Italy was much larger than that envisaged by the low counters. 17 18

19 20 21

Ibid. In a recent article (2009) Lo Cascio claims that the low count implies an urbanization rate of 39 per cent if Rome is included and of 25 per cent if it is not. The calculations underlying his claim are, however, based on the guesstimates of Morley (1996, 181–2), the speculative nature of which has been emphasized by the author himself (e.g. Morley 2008, 123). No meaningful conclusions can be reached on this basis. Kron (2005a), 474–5, building on Ginatempo and Sandri (1990). For the size of the northern Italian population in around ad 1300, see Del Panta et al. (1996), 277. Chapter 1, at note 81.

Introduction

199

It will by now be clear that the low and high counters have completely different ideas about the defining features of Italy’s urban network at the time of Augustus. This provides us with an opportunity to examine the capacity of these competing models to make sense of the rich body of data relating to the development of Italian towns that has accumulated during many decades of archaeological research. 5.1.3 Definitional problems Before embarking on a study of urban networks, we must of course be clear about what we mean by the term ‘town’. As countless studies have pointed out, about 430 agglomerations in early-imperial Italy had the juridical status of towns during the first half of the first century ad.22 From an economic and demographic point of view, however, the ‘urban’ nature of many of these settlements is clearly dubious.23 Since one of my aims is to put the low and high counts for Roman northern Italy to the test by comparing the urban network that existed at the time of Augustus with that of latemedieval and early-modern times, it is absolutely necessary that we should use the same definition of ‘town’ for both periods. In principle, ‘towns’ can be distinguished from other types of settlement according either to juridical, administrative and political criteria or to population size. The former approach has the advantage of providing a clear distinction reflecting the views and definitions of the society under investigation. It can, moreover, be argued that the legal status of premodern settlements often had a real impact on the activities of their inhabitants, since certain functions were performed only or mainly by centres that were ‘urban’ in a legal or political sense.24 The obvious drawback of the legal and political approach is that formal definitions of what constitutes a town tend to differ from period to period and from region to region. Using this approach, it is therefore almost impossible to make any meaningful comparison between the towns and urban networks of different societies, especially if these are separated by a large gap in space or time. For this reason, most historians who are interested in the evolution of 22

23

24

This figure drops to about 420 if Arretium and Bovianum are counted as single towns rather than as multiple communities and if we ignore non-urban communities such as those of the Ager Labicanus, the Ager Hernicus and the Ager Latinus. On the crucial importance of distinguishing between towns in a legal and administrative sense and settlements that were ‘urban’ from a demographic and economic point of view, see my remarks in De Ligt (1990), 26–30. Epstein (2001), 2–3.

200

Augustan census figures and Italy’s urban network

urban networks on a supra-regional scale, or in the differences and similarities between the urban networks of completely different societies, prefer to use a purely quantitative approach, although the critical threshold that settlements must attain to qualify as ‘real’ towns differs from study to study.25 A third approach is to focus not so much on the numerical balance between those living in or outside ‘large’ settlements as on the proportion of the population not fully or primarily engaged in agricultural production. According to a widely held view, the low levels of (land) productivity that characterized most pre-modern economies meant that about four agricultural workers were needed to sustain one ‘urban’ worker, regardless of his place of residence. This would mean that few, if any, pre-modern societies could achieve an economic urbanization rate higher than 20 per cent, unless of course these societies had access to additional food resources produced in other areas.26 In some cases these three approaches can give rise to strikingly different results. One particularly illuminating example comes from classical Greece. Using a political and administrative definition of ‘town’, Hansen has argued that at least 60 per cent of the population of fifth- and fourth-century Greece lived in ‘urban’ settlements.27 Since the Greeks used the term polis for all settlements controlling a rural territory, this approach makes perfect sense from a cultural point of view. Hansen does not, however, mean to imply that 60 per cent of the Greek population was fully or primarily engaged in non-agricultural activities. His main point is, in fact, that farmers made up a large part of the urban populations of classical Greece, and that this society was unlike some parts of medieval Europe in that there was no clear separation between town and country. Like most publications that study towns in different areas or periods, my analysis of the urban network of early-imperial Italy will be based on a quantitative definition of the term ‘town’. Since we are about to adopt this approach, our next question must be how many inhabitants a given 25 26

27

This is the approach of De Vries (1984) and Bairoch (1988). Cf. below. Cf. Bairoch (1989), 247. Some indications suggest that the conventional figure of 20 per cent is on the low side. As Malanima (2005), 108, shows, during late-medieval and early-modern times between 17 and 21 per cent of the population of central-northern Italy lived in towns with at least 5,000 inhabitants. If we had reliable data concerning all settlements with 3,000 or more inhabitants, the proportion of the population engaged in non-agrarian occupations might rise to 25 per cent for at least some decades between 1300 and 1600. In 1861 42 per cent of the Italian population was engaged in non-agricultural activities, but only 20 per cent lived in towns with more than 5,000 inhabitants. In an intriguing article, Grantham (1993) demonstrates that the technology available to French peasants of the eighteenth century was theoretically capable of producing a surplus large enough to carry an urban ratio of 60 per cent. Hansen (2006).

Introduction

201

settlement must have in order to qualify as ‘urban’. In his study of the urban network of early-modern Europe, Jan de Vries applies the label ‘town’ only to those centres with 10,000 inhabitants or more.28 In my view, only a small minority of the Italian settlements that had the legal status of towns in earlyimperial times fulfilled this criterion. Other studies of the towns of medieval and early-modern Europe, such as that carried out by Bairoch, assume a critical threshold of 5,000 or 3,000 inhabitants.29 In the case of latemedieval and early-modern Italy we cannot go below the latter threshold, for the simple reason that there are no reliable lists of settlements with fewer than 3,000 inhabitants. If the towns that met this threshold had urban population densities of approximately 150 persons per hectare (a common density for medieval towns), they would have covered c. 20 hectares. It follows that if we want to make a rough comparison between the sizes of ‘urban’ populations in Roman and early-modern times, all Roman settlements that covered fewer than 20 hectares (including those that were towns in a legal and administrative sense) must be classified as ‘non-urban’. With these definitional problems in mind, I have created three categories of town. My first category consists of early-imperial Italy’s most important towns. These have been defined as all settlements that can be shown to have covered more than 40 hectares. Below this first tier, it is possible to identify a second group of towns that were considerably smaller but nonetheless substantial. This second group has been defined as including all settlements that covered between 20 and 40 hectares. Finally, I have assigned all settlements that enjoyed urban status in 28 bc but that covered fewer than 20 hectares to a separate category. In literature concerning the urban networks of earlymodern Europe, such agglomerations are often referred to as ‘small towns’.30 For the reasons just set out, I have followed De Vries and Bairoch in counting the populations of my third tier of settlement as ‘rural’. Needless to say, the use of this label does not in any way imply that the ‘small towns’ of Roman Italy performed no ‘urban’ functions for the populations of their catchment areas.31 Here we have only to read the extensive literature on the 28 30

31

De Vries (1984). 29 E.g. Bairoch (1988). E.g. Clark (1995). In British archaeology the term ‘small town’ is used to denote vici of the Roman period (e.g. Burnham and Wacher 1990). In an interesting study that centres on the small town of Tridentum, Migliario (2008) points out that the administrative function of minor towns presupposes the existence of a substantial rural population. This is undoubtedly correct, both of ancient and of early-modern times. Even if we had complete inventories of minor towns and large villages, however, any attempt to extrapolate the size of the total population from the number of townsmen and villagers would continue to be beset by major methodological difficulties (cf. above, at note 1). As Migliario realizes, the case of Tridentum is especially problematic because non-citizens made up a significant proportion of the rural population.

202

Augustan census figures and Italy’s urban network

small towns of medieval and early-modern Europe, from which it will appear that settlements with populations of between 1,000 and 3,000 played a vital role in the lives of the great majority of the population.32 This means that any reconstruction of an urban network that is based on a purely quantitative concept of ‘town’ can only provide us with a very partial glimpse of a complex settlement hierarchy. In other words, neither the urban networks reconstructed by De Vries and Bairoch nor my own reconstruction of the urban network of early-imperial Italy can be regarded as revealing any fully functional settlement system from an administrative, economic, social or cultural point of view. 5.1.4 Some methodological considerations In distributing the towns of early Augustan Italy among the three categories defined above, I have drawn mainly on a rich variety of archaeological publications. In some cases, however, valuable indications have also been supplied by the literary sources, in particular Strabo. Many examples of Strabo’s usefulness in this respect may be found in his description of northern Italy. Here the phrase polis axiologos is used to characterize Mediolanum and Dertona, while Patavium is singled out as an exceptionally wealthy and populous city.33 Strabo also calls Verona ‘a large city’ (polis megalê). He lists Placentia, Cremona, Parma, Mutina and Bononia among the ‘famous towns’ (poleis epiphaneis) of the north, but uses the term polismata (small towns) for Opitergium, Concordia, Atria, Vicetia, Regium Lepidum, Claterna, Forum Cornelii, Faventia and Caesena.34 Similar clues can be found in his survey of the towns of the central and southern regions. We are told, for instance, that Teanum is the largest town on the via Latina after Capua.35 Further clues may be gleaned from Pliny the Elder, who lists the ‘notable towns’ (nobilia oppida) of northern Liguria, and from Tacitus, who informs his readers that in the year of the four emperors Mediolanum, Novaria, Eporedia and Vercellae were ‘the strongest of the Transpadane towns’ (firmissima transpadanae regionis municipia).36 Unfortunately, we cannot always be sure that Strabo’s classifications and descriptions are valid for the late Republic and early Empire. In some cases 32 34 36

See the valuable collection of essays in Clark (1995). 33 Str. 5.1.6, 5.1.11, 5.1.7. Str. 5.1.6, 5.1.8, 5.1.11. 35 Str. 5.3.1. Plin. NH 3.49: Libarna, Dertona, Iria, Vardagate, Industria, Pollentia, Potentia, Forum Fulvii, Augusta Bagiennorum, Alba Pompeia, Hasta, Aquae Statiellae; Tac. Hist. 1.70.

Introduction

203

he can be shown to have missed or neglected recent developments, such as the expansion of Forum Cornelii and the establishment of veterans at Ateste after 30 bc.37 Similarly, it is not entirely clear what Tacitus means when he refers to the firmissima municipia of Transpadana. Does he mean that these were the largest towns in the region, or simply that they were the strongest towns in a military sense (for instance, because they were protected by formidable town walls)?38 Regardless of any remaining problems of interpretation, the impressionistic clues contained in the literary sources do not allow us to put any figure on the physical extent of the towns to which they refer. Fortunately, this problem can in part be resolved with the help of the many topographical studies on the cities of Roman Italy that have been published during the past 50 years. Many of these studies give a precise figure for the extent of the inhabited area, or contain maps from which an approximate figure can be derived. In other cases, we are given information only on the number of hectares enclosed by the town walls. This is not to suggest that these archaeological data are unproblematic. One obvious difficulty is that almost every town of the Roman period lies buried under modern agglomerations. There are, however, several reasons for thinking that the continued use of many of the sites occupied by Roman towns does not necessarily vitiate any attempt to gauge their physical extent during the early Empire. In many of the towns of central and northern Italy, a large number of scattered data have accumulated over the past century. Although this enormous body of evidence is in many ways incomplete and unsatisfactory, it certainly provides us with a rough idea of the extent of many Italian towns in Roman times. In many cases, natural boundaries (such as rivers) and the location of cemeteries shed additional light on the early extent of these towns.39 In quite a few cases it is also possible to reconstruct the course followed by the Roman town walls. Here a potential problem is that the size of the inhabited area may well have been either much smaller or substantially larger than the space enclosed by the walls. In those cases where the town wall was built during the third or second century bc, any suburbs that had developed before 28 bc must be added. The archaeological evidence 37

38 39

Chilver (1941), 54, infers that ‘Strabo is drawing on information about conditions in the region before the principate of Augustus, indeed before the colonizations of the triumviral period’. Cf. the data assembled in Appendix i. Cf. Panero (2000), 214, n. 501. Unfortunately, these observations are less true of the towns of Lucania, Apulia and Bruttium. Cf. below, at notes 129–30.

204

Augustan census figures and Italy’s urban network

demonstrates that such suburban districts could easily add 50 per cent to the original extent of the older towns.40 In the case of towns whose walls were built shortly before or after 28 bc it is possible to follow a different line of reasoning. Here I have assumed that the entire space enclosed by the late-republican or early-imperial walls was inhabited. One justification for this assumption is that the archaeological evidence does not appear to support the view that any significant proportion of the areas enclosed by town walls of late-republican or early-imperial date remained unoccupied. My personal impression (it is no more than that) is that most Roman towns founded or walled during the first century bc were more compact than most of the cities of Etruria and Magna Graecia during the archaic and classical periods.41 It seems significant that several towns whose walls were built during the late Republic acquired suburbs in earlyimperial times.42 Had there been large empty spaces within the town walls, one would expect most of these to have been built up before extramural quarters of any size began to develop. It is clear that the supposition that the inhabited areas of towns whose walls were built shortly before 28 bc equalled their walled areas is defensible. My main reason for taking this view, however, has to do with the general drift of my argument. As we shall presently see, most of the towns of Roman Italy appear to have been surprisingly small compared to their medieval and early-modern successors. In what follows, I shall argue that the archaeological evidence supports this conclusion even if we assume that the walled areas of these towns were fully built up. In other words, in the absence of secure evidence for the existence of substantial empty spaces within the town walls, it is methodologically preferable to assume that the entire walled area was inhabited, especially in an argument that will attempt to question the view that the towns of Roman Italy (and those of Cisalpine Gaul in particular) equalled or surpassed those of later periods in size and population.

40

41

42

Northern examples include Cremona and Aquileia. In peninsular Italy Pisaurum, Tibur and Tarracina are known to have acquired suburbs, but in all of these cases we have to reckon with the possibility of substantial growth after 28 bc. In most of central and southern Italy evidence for extramural suburbs is rare, but many towns appear to have expanded within the often spacious town walls built in earlier centuries. The expansion of Pompeii within the town wall of the sixth century bc is a classic example. For detailed discussion and references, see Appendix ii. On the walled towns of Cisalpine Gaul, see Chevallier (1983), 149: ‘Même si les plans d’urbanisme, conçues largement, prévoyaient des extensions futures à l’intérieur des murs, la totalité de l’espace urbain à été en general garnie.’ The same view may be found in Ortalli (2003), 96. E.g. Chevallier (1983), 149, and the essays in Antico Gallina (2000a).

The urban network of Cisalpine Gaul in 28 bc

205

A final difficulty is that the present archaeological evidence makes it difficult to follow the expansion of Roman towns over time.43 This is why archaeological literature tends to date the remains of urban buildings to broad time-spans such as ‘the Augustan period’ or ‘the early Empire’. In all such cases, I have opted for an estimate of the extent of the built-up area that includes all Augustan and early-imperial quarters. Since ideally we should operate with a chronological cut-off point coinciding with the census of 28 bc, the inclusion of all Augustan and early-imperial material almost certainly means that some of my estimates are too high. Again, the justification for this approach is that in assessing the viability of the low and high counts it is methodologically preferable to use high estimates for the areas covered by the towns of the late Republic and early Empire. 5.2 the urban network of cisalpine gaul in 28 bc As we have seen, one of the central tenets of the high count is that Cisalpine Gaul contained at least 40 per cent of the population of mainland Italy and that the towns of this region matched those of late-medieval times in terms of size and population. In complete contrast to this, the low count assigns to northern Italy between 25 per cent and 30 per cent of a much smaller Italian population. This alternative scenario can only work if the towns of early-imperial Cisalpina were much smaller than their latemedieval and early-modern counterparts. Precisely because the contrast between these models is so stark, northern Italy is in many ways an ideal testing ground for the low and high counts. Using the methods just outlined, it is possible to put an exact, or at least a rough, figure on the size of sixty-two northern agglomerations that enjoyed urban status in early-imperial times. In many cases the impressionistic indications supplied by the literary sources are confirmed. One example of this is Patavium, to which Beloch accords 85 ha, on the assumption that only the area between the two branches of the river Meduacus was built up during the early Empire.44 Later research has revealed this assumption to be incorrect. There was in fact a substantial built-up area to the east of the central ‘island’, possibly covering a further 40 or 45 ha. Early-imperial Patavium could therefore have covered some 130 ha, backing up Strabo’s 43

44

In principle, this observation also applies to cases of urban contraction. There is some evidence to suggest that certain towns in Apulia and Lucania declined during the second and first centuries bc. The data from Cisalpine Gaul suggest vigorous urban growth. Beloch (1898), 272.

206

Augustan census figures and Italy’s urban network

statement that it was ‘the best of all cities’ in the north.45 Another important centre was Mediolanum, whose town wall enclosed some 80 ha during the late Republic and early Empire. This confirms Strabo’s description of Mediolanum as an axiologos polis. It may, in passing, be noted that the figure of 133 ha given by Beloch refers to the area enclosed by the longer town wall of the late third century ad.46 A third example is Bononia, which covered some 50 ha in the age of Augustus.47 This time Beloch’s estimate (83 ha) turns out to be too high. Even with 50 ha, Bononia would have been one of the largest centres in the north, in accordance with Strabo’s statement that it was among the ‘famous’ cities of the north. According to the archaeological data, fifteen of those northern settlements that were towns in a juridical sense appear to have covered more than 40 hectares apiece. It is remarkable that the existing archaeological literature should permit us to put an exact or approximate figure on the size of each of these towns; the reason for this must be that archaeological research in the north has been biased towards the larger centres. The average number of hectares per town is 59.8; the total number of urban hectares is 896.4.48 As stated above, my second category consists of cities known to have covered or likely to have covered between 20 and 40 ha. Interestingly, three of the four firmissima municipia mentioned by Tacitus fall into this category, suggesting that even towns half the size of Mediolanum were regarded as substantial. All in all, twenty-nine towns can be assigned to this category on the basis of their physical extent. To these twenty-nine towns I have added Ateste and Atria. Ateste is poorly documented, but is known to have received a substantial influx of colonists after the battle of Actium. Atria possessed a theatre and a substantial amphitheatre, suggesting that it was far from negligible. In estimating the physical extent of the towns making up my second category, I have accorded to each of these towns 28.0 ha, the average for the twenty-nine towns for which we have secure evidence. The total number of hectares for my second category is 869.3.49 Finally, we arrive at the lowest tier in the urban hierarchy, for which I have used 19.9 ha as an upper limit. The sizes of eighteen of these smaller centres can be determined with a reasonable degree of confidence. They range from 2 ha in the case of Forum Novum to c. 15 ha in the case of Feltria, Potentia and Vardagate. The average for those towns whose physical extent can be determined is 10.4 ha. If we apply this figure to the centres for which 45 49

Str. 5.1.7. 46 Beloch (1886), 487. Appendix i.2.

47

References in Appendix i.

48

Appendix i.1.

The urban network of Cisalpine Gaul in 28 bc

207

no data are available (sometimes because their locations are unknown), we end up with a total of 332.4 ha for towns in this group.50 Before proceeding with my argument, I should like to draw attention to the fact that the foregoing analysis does not take into account the numerous vici that must have existed at this time. In this context a comparison with the settlement system of northern Italy at the beginning of the seventeenth century may be instructive. At that time the region appears to have had c. sixty cities and towns with 5,000 or more inhabitants apiece; there were also 150–200 ‘small towns’ with populations of between 2,000 and 5,000.51 Similarly, there must have been several large vici and numerous smaller lower-order settlements in the territories of each of the seventy-eight ‘towns’ of Roman Cisalpina. The sites of a considerable number of large vici, each covering between 5 and 15 hectares, have, in fact, already been located.52 Supposing these centres to have contained 150 persons per hectare, their populations would have ranged between 750 and 2,250.53 One lesson to be drawn from this is that at least some vici were bigger than some of the ‘small towns’ making up the third tier of my ‘urban’ hierarchy.54 At first sight this may seem to suggest that the distinction between ‘towns’ and ‘villages’ is largely arbitrary, but this difficulty arises only if we define towns on the basis of their administrative status. If we use 20 hectares as our cut-off point, the distinction between towns and villages becomes 50 51

52

53

54

Appendix i.3. For the large towns of northern Italy during the early-modern period, see Appendix iii; for the number of small towns, see Musgrave (1995), 254–5. See e.g. Zaccaria (1979), Strazzulla Rusconi and Zaccaria (1984), Gregori (1993), Maggi and Zaccaria (1994 and 1999), Sena Chiesa (1995 and 2003), Arnaud (2004). For the vici of the imperial province of Germania Superior, Wendt and Zimmermann (2008), 208, use an average density of between 70 and 140 inhabitants per hectare. Densities of up to 300 inhabitants per hectare have been calculated for some villages in Syria and Yemen, but in all these settlements two-storey houses were the norm. See e.g. Watson (1980), 59; Van Beek (1982); Rautman (2006), 162. I am grateful to John Bintliff for his helpful comments on this point. Unfortunately, the archaeological data that we currently have offer no sound basis for estimating the proportion of the ‘rural’ population living in the many lower-order centres that must have existed. It should be noted that during the early-modern period the vast majority of the rural population of northern Italy lived within walking distance of one of the 60 large and c. 200 small towns providing ‘urban’ goods and services, including administration (Musgrave 1995, 255). If the total number of settlements performing centralplace functions was roughly the same in the early years of the Empire, there would have been some 214 (260 less 46) small towns and large vici. Of these smaller settlements, the thirty-two ‘unimportant towns’ of Cisalpina appear to have covered c. 10.4 ha apiece on average. If each of the remaining 182 centres also covered 10.4 ha on average and if we assume that they had 150 inhabitants per hectare, the total population of the hypothetical 214 lower-level centres would have been 333,840. This is less than a quarter of the ‘rural’ population implied by the low-count model for Cisalpine Gaul (cf. below). In other words, even in a low-count model for Cisalpina there is room for a very large number of nonurban central places.

208

Augustan census figures and Italy’s urban network

unproblematic, for the simple reason that there are no known examples of northern Italian vici that covered more than 15 hectares in 28 bc.55 The net result of this exercise in quantification and classification is a network of forty-six settlements sufficiently large to qualify as urban at the time of the census of 28 bc. If we add together my estimates for these fortysix towns, we obtain a rough figure of 1,765 urban hectares in Cisalpine Gaul. Since some of the underlying data are rough approximations, there can be no doubt that there is a considerable margin of error, but I hasten to add that my estimate of the number of urban hectares in the north is almost certainly too high rather than too low. Firstly, as I have already pointed out, I have consistently assumed that the areas enclosed by the town walls of the north were entirely built up. Secondly, although my analysis focuses on the shape of the urban network of Italy in 28 bc, my list of northern towns includes some urban centres, such as Augusta Praetoria, that either did not yet exist or did not yet have citizenship in that year.56 Thirdly, in estimating the physical extent of the northern towns, I have included all town quarters that have been even vaguely dated to the Augustan era or to the early Empire. We can be certain that at least some of these town quarters did not yet exist in 28 bc. For all of these reasons, it seems highly unlikely that the forty-six towns that make up the first two tiers of my urban hierarchy covered more than c. 1,765 hectares in 28 bc. Regardless of the margins of error to which I have alluded, this crude figure is of considerable interest. As we saw at the beginning of this chapter, the high count for late-republican Italy assigns a population of c. 6 million to northern Italy. This is only slightly lower than the figure for ad 1600, when the regions corresponding to Roman Cisalpina were inhabited by some 6.5 million people.57 Of these 6.5 million inhabitants, at least 1.16 million, or 17.9 per cent, lived in cities and towns with a population of 3,000 or more. For the purposes of this chapter, I should like to draw attention to certain data concerning the sizes of the northern cities during the early-modern period. From the mid-sixteenth century, Milan, Venice and Bologna appear to have been the largest centres in terms of physical size, covering c. 794 ha, 55

56

57

Under the Empire the vicus of Bedriacum came to cover c. 30 ha (Sena Chiesa 1998, 345) and it seems likely that the vicus at Angera also reached 20 hectares during this period (Sena Chiesa 1995). These examples point to continuing population growth under the Empire (cf. Chapter 6, at note 83), but they do not undermine the validity of my three-tier model for 28 bc. Another possible example is Augusta Taurinorum, which may have been founded in 27 bc (e.g. Chilver 1941, 201). 6.5 million: Del Panta et al. (1996), 277. Beloch (1937–1961), iii, 352, followed by Belletini (1987), 29, and by Jongman (1988), 72, used a lower estimate of 5.4 million.

N W

E S

Augusta Praetoria Eporedia

Augusta Taurinorum Pollentia

Comum Bergomum

Mediolanum

Brixia

Opitergium Tarvisium Concordia

Vicetia Novaria Laus Vercellae Pompeia Verona Patavium Industria Pavia Cremona Ateste Hasta Dertona Hatria

Libarna Alba Pompeia Aquae Augusta Bagiennorum Statiellae Genua Pedo

Altinum

Placentia

Parma Regium Lepidi

Pola

Mutina Bononia Claterna Imola Faventia Forum Livii

Albintimilium

Aquileia

Ravenna Caesena

Ariminum

Adriatic Sea Ty r r h e n i a n S e a

Medium Large 0 0

50 25

100 50

150 75

200 km

100 miles

Map 5.1 The large and medium-sized cities of northern Italy, c. 28 bc

210

Augustan census figures and Italy’s urban network

c. 600 ha and c. 419.5 ha respectively.58 Alongside these giant cities were many other substantial towns such as Brescia, Cremona, Ferrara, Genova, Mantova, Padova, Piacenza, Verona and Vicenza. Of this second group of cities, Ferrara, Genova, Padova, Piacenza and Verona covered areas of 340 ha, 155 ha, 450 ha, 290 ha and 380 ha respectively.59 The total area for these five cities plus Milan, Venice, and Bologna is thus c. 3,428.5 ha. In other words, eight large northern Italian cities, inhabited by roughly 45 per cent of the urban population of the sixteenth century,60 were together in physical terms twice as large as the combined forty-six large and medium-sized towns of Cisalpine Gaul during the early Empire. These figures suggest that the total number of urban hectares in northern Italy during the early-modern period must have been at least 7,000, which would have been roughly four times higher than the estimated figure for 28 bc.61 Unless we assume that urban population densities were three or four times higher in Roman times than they were during the Renaissance, it follows that the high count for the north can be maintained only by positing a Roman urbanization rate far below the early-modern level. 5.2.1 Some ingredients for a more detailed analysis Although the main point of the first half of this chapter has already been made, it can, I think, be corroborated by taking a closer look at the size of the northern Italian populations predicted by the low and the high count and by confronting these competing models with archaeological evidence from the northern towns. In my view the most promising approach is to take a detailed look at the following formula: POPtot ¼ ð100: URB:RATEÞ  HECTurb DENSurb In this formula POPtot stands for the total population (including foreigners and slaves), URB.RATE for the percentage of the population living in the forty-six large or medium-sized towns of Cisalpine Gaul, HECTurb for the 58

59

60 61

Milan: 794 ha in 1565 according to Beloch (1937–1961, iii, 175); Venice: 600 ha according to Benevolo (1980, 600); Bologna: 419.5 ha in the early fourteenth century according to Beloch (1937–1961, ii, 91). Chandler and Fox (1974), 85: Ferrara 340 ha in 1500; Heers (1961), 45: Genova 155 ha in 1450; Benevolo (1980), 326: Verona 380 ha in the fourteenth century, Piacenza 290 ha in the fourteenth century, Padova 450 ha within the Venetian walls of the fifteenth century. Appendix iii. It follows that Kron (2005a), 474–5, is quite wrong in suggesting that the urban population of northern Italy in Roman times may have equalled that of the later Middle Ages.

The urban network of Cisalpine Gaul in 28 bc

211

number of hectares covered by these towns and DENSurb for the average number of inhabitants we can assign to each urban hectare. Although we do have some information about urban population densities during Roman times, none of it comes from Cisalpine Gaul. Moreover, as we shall presently see, we have to reckon with considerable variations in the number of inhabitants per urban hectare, and these variations can be analysed only by introducing comparative data. In other words, even if my estimates for the total area covered by the towns of Roman Cisalpina are accepted as approximately correct, it remains impossible to reconstruct the size of the northern Italian population by proceeding inductively.62 At this juncture it must be remembered that the low and high counts offer quite specific suggestions as to the size of the population of northern Italy at the time of the census of 28 bc. As we have seen, the low-count proposals range from 1.5 to 2 million, while the high-count estimate is approximately 6 million. Instead of using a bottom-up approach to calculate the population of early-Augustan Cisalpina, therefore, we can use these estimates as our starting points and explore the range of urban population densities and urbanization rates that they imply. This is the tactic that will underlie the investigations of the following sections. 5.2.2 Expected urbanization rates In an earlier section, we saw that at the beginning of the seventeenth century at least 17.9 per cent of the northern Italian population lived in towns (if we classify all settlements with 3,000 or more inhabitants as towns).63 All existing estimates of urbanization rates in Roman Italy at the time of Augustus are average figures for Italy as a whole. As noted in my discussion of the completely different views of the shape of the urban network of earlyimperial Italy implied by the low and high counts, Hopkins assumes a total urban population of 1.9 million, which would imply an urbanization rate of 20 per cent if Rome is excluded. Since Hopkins’ definition of ‘urban’ is based on occupation rather than on place of residence, however, it includes 62 63

Cf. De Ligt and Northwood (2008), 4. Above, at note 57. These percentages are based on an assumed northern Italian population of 6.5 million. During the early-modern period, northern Italy also had between 150 and 200 small towns with populations of over 2,000 (Musgrave 1995). Although some of these had more than 3,000 inhabitants, they do not appear on the lists compiled by Bairoch (1988). It follows that the proportion of the northern Italian population living in towns with 3,000 or more inhabitants was actually somewhat higher than 17.9 per cent.

212

Augustan census figures and Italy’s urban network

both rural craftsmen and the non-farming populations of minor towns. It follows from this that the application of the purely quantitative approach championed by De Vries and Bairoch would result in a significantly lower urbanization rate. In similar vein, Lo Cascio puts the share of the non-agrarian population outside Rome at about 20 per cent and infers from this that no more than 15 to 20 per cent of the Italian population can have lived in towns with at least 2,000 inhabitants.64 Since Hopkins and Lo Cascio are working from a pan-Italian perspective, their estimates offer little guidance to those interested in the urban networks of specific sub-regions. One of the very few things that can be said about Cisalpina is that most of its social and economic elites seem to have been living in towns by the time of the census of 28 bc. As Garnsey has noted, the modest size of the northern Latin colonies and the extent of the centuriated areas surrounding them suggest that most pedites were expected to live in the countryside. The other side of the coin is that the Roman government acted on the assumption that the larger landowners would be town-based.65 In other words, in setting up the Latin colonies of Cisalpina the Roman government was extending to the northern regions the central-Italian tradition of elite residence in towns. As far as we can tell, the indigenous elites of Transpadana were also largely town-based by the end of the Republic. From the early Principate onwards, they also took up the centralItalian model of urban euergetism.66 The establishment of veterans from central Italy in various northern towns between 44 bc and 25 bc can only have strengthened this pattern. This suggests that urbanization rates in Cisalpine Gaul are unlikely to have been dramatically lower than those in central and southern Italy. On the other hand, there is some reason to believe that Cisalpine Gaul was less urbanized than some other parts of Italy. Some 20 years ago, Bekker-Nielsen calculated average inter-city distances in various parts of 64

65

Lo Cascio (1999a), 164–5; Lo Cascio and Malanima (2005), 209. When Lo Cascio and Malanima compare this percentage with the urbanization rate of 16.2 per cent that has been calculated for northern and central Italy in 1861 using a threshold of 5,000 inhabitants (cf. Malanima 2005, 108), their definition of ‘urban’ seems to be similar to mine (although their cut-off point is different). When Lo Cascio uses the existence of 430 Italian ‘towns’ at the time of Augustus as an argument against the low count, on the grounds that it implies an incredibly large ‘non-agrarian’ population (e.g. Lo Cascio 2009, 98), he seems to switch to an economic definition. Since this wider definition underlies the urbanization rate of 39 per cent that he attributes to Morley and Scheidel, his comparison with the much lower rate revealed by the census of 1861 is invalid. For further problems, see above, note 18. Garnsey (1998), 123–6; cf. Gabba (1979), 34. 66 Frézouls (1990).

The urban network of Cisalpine Gaul in 28 bc

213

Italy. One of his findings was that the average distance between the cities of the north was between 24.9 and 35.6 km (except in the Via Aemilia corridor). The corresponding figure for Latium and Campania is only 11 km.67 In principle, we cannot rule out the possibility that urbanization rates in the north were similar to those in central Italy and that the towns of the north were simply bigger than those in Latium and Campania. One clue that points in this direction is Strabo’s statement that Cisalpine Gaul had a large free population and many large and wealthy cities; he asserts that northern Italy had surpassed the central and southern regions in these respects.68 However, even if we allow for the possibility that there was a positive relationship between the size of towns and the size of their territories, the simple fact that northern towns were spaced more widely apart than those in other parts of Italy still suggests that urbanization rates in the north were somewhat lower than the average rate for Italy as a whole. I conclude that (if only those settlements covering 20 or more hectares are counted as urban, but if Rome is left out of the calculation) both the low count and the high count point to an urbanization rate of between 15 per cent and 20 per cent for Italy as a whole. At the same time, there are good reasons for thinking that urbanization rates in Cisalpine Gaul were somewhat lower than the average Italian rate. It seem legitimate to infer that, regardless of which count we prefer, the proportion of the northern population living in medium-sized or large towns is likely to have been closer to 15 per cent than to 20 per cent. 5.2.3 Expected urban population densities Before trying to delineate a range of plausible urban population densities for the towns of northern Italy during the early years of the Principate, I should like to start with a brief survey of some methodological difficulties. The most obvious problem has already been noted: since the space enclosed by the walls of many pre-modern towns was not fully built up, it is often difficult to put any figure on the extent of the areas occupied by public or private buildings at any given time. The same difficulty has been noted with respect to Greek, Etruscan and Oscan towns and in my own research I have encountered quite a few examples of this phenomenon in other parts of Italy (for instance, in Latium and in Apulia). It must also be remembered that some ‘Roman’ towns developed within the wall circuits of pre-existing Greek and Oscan towns; here we could mention, for 67

Bekker-Nielsen (1989), 25.

68

Str. 5.1.12.

214

Augustan census figures and Italy’s urban network

instance, the Latin colonies of Luceria and Paestum.69 In all of these cases we have to reckon with the possibility that at least some of the walled space that started off empty was built up during later periods. We can therefore be certain that the number of inhabitants per hectare of walled space varied over time. Moreover, even if we assume that the entire area enclosed by the walls of an ancient town was built up, it was of course entirely possible for its population density to increase over time, for example because new buildings were fitted in among existing ones, or because it became more common to build houses with two or three storeys. A combination of these developments is known to have taken place in republican Pompeii, where we observe an increase in the number of buildings within the town walls and in the number of buildings with an upper floor.70 Several factors favouring such developments can be identified. It seems reasonable to suppose, for instance, that many inhabitants of urban agglomerations preferred to live within the town walls, if only because they appreciated the protection provided by these walls. We should therefore expect urban population densities to have increased before large suburbs began to develop. This phenomenon is well attested in medieval Europe.71 In the case of Graeco-Roman towns, we must also bear in mind that older towns were gradually boxed in by the cemeteries surrounding them, so that they were unable to expand in some directions. This may have been another factor tending to push up urban population densities in towns whose populations were expanding. It was, of course, also possible for urban population densities to decline. It has been pointed out that decreases in the urban population, even drastic ones such as the population collapse caused by the Black Death, were not usually reflected in any change in the physical make-up of towns. As we shall see later, this makes it difficult to follow demographic developments in some of the Greek towns of southern Italy. On the other hand, my investigations into the network of early-imperial Cisalpina have not brought to light any case of urban contraction during the late Republic or early Empire, so that this problem at least can be left aside. 69 70

71

For Luceria, see e.g. Antonacci Sanpaolo (1999); for Paestum, see Torelli (1999b). For the gradual filling up of empty spaces within the town walls of Pompeii, see Pesando (1997), 13–15, and Schoonhoven (2003), 236–8. For similar developments in second-century bc Cosa, see Brown (1980), 66. For more two-storey buildings in late-republican and early-imperial Pompeii, see Pesando (1997), 77, 207–8, and esp. Pirson (1999), 161, 171–3. This development is known to have taken place in many cities all over the world. See e.g. Chandler and Fox (1974), 5, for densities of up to 200/ha just before the building of a new town wall.

The urban network of Cisalpine Gaul in 28 bc

215

In addition to these methodological difficulties, we have to face the fact that (with the partial exception of Roman Egypt) we have hardly any data concerning urban population numbers that can be used to shed light on urban population densities in Roman Italy, or, for that matter, in any other part of the Roman world. In practical terms this means that the topic of urban population densities can be approached only by looking at evidence relating to other preindustrial societies. Since comparisons between Augustan Italy and late medieval/early-modern Italy are an important theme in the debate between low counters and high counters, it seems logical to start with the Italian towns of the High Middle Ages and the Renaissance. In passing, it must be emphasized that the scope of this type of inquiry can easily be expanded to other parts of pre-modern Europe and that the problems presented by the Italian material are similar to those that have occupied specialists in the urban history of pre-modern England, Germany, France, Flanders and Holland.72 Let me begin with a brief survey of urban population densities in Italian cities during the first half of the fourteenth century, when Italy’s urban population reached a temporary high. During this period most Italian cities appear to have had population densities ranging between 100 and 140 persons/ha. Cities falling into this category include Pistoia, Verona, Bologna and Padova, with 103, 105, 119 and 133 persons/ha respectively. Later on, in the sixteenth century, Milan had 126 inhabitants/ha.73 In some cities we find somewhat higher densities. Thus early fourteenth-century Arezzo is thought to have had some 168 inhabitants/ha, while a density of 175 persons/ha can be calculated for Florence in 1333.74 Finally, it is possible to detect a small group of cities with much higher densities. The most striking case is late-medieval Genova, with some 65,000 inhabitants on 72 73

74

Cf. note 76. Pistoia: 117 ha in the late thirteenth century according to Herlihy (1967), 74; Verona: above, note 59; Bologna: 419.5 ha in the early fourteenth century according to Beloch (1937–1961), ii, 91; Padova: 300 ha in 1320 judging by the map in Hyde (1966), 36; Milan: above, note 58. For all these cities I have used the population estimates given in Malanima (1998): Pistoia 12,000 in 1300, Verona 40,000 in 1300, Bologna 50,000 in 1300, Padova 40,000 in 1300, Milan 100,000 between 1500 and 1600. Arezzo: 107 ha according to Cherubini (2003), 140; Florence: 630 ha in 1333 according to Herlihy (1958), 35, n. 1 (the figure of 512 ha given by Beloch 1937–1961, ii, 128, is too low). Again the population estimates are those of Malanima (1998): Arezzo 18,000 in 1300, Florence 110,000 in 1300. If Pisa, whose third circuit of walls enclosed c. 185 ha according to Herlihy (1967), 74, had c. 38,000 inhabitants in around 1300 (thus Herlihy 1958, 36), it would have had 205 persons/ha. The estimates for early fourteenth-century Pisa are, however, controversial (Ginatempo and Sandri 1990, 259). Malanima (1998) gives this city only 30,000 inhabitants in 1300, implying an urban population density of 162 persons/ha.

216

Augustan census figures and Italy’s urban network

155 ha, or 419 persons/ha. Another is early fourteenth-century Siena, with 50,000 people on 165 ha.75 In this case the implied density is 303 inhabitants/ha. It should, however, be noted that the area enclosed by the town walls of Siena was not entirely built up. In other words, in the built-up area the number of people per hectare must have been considerably higher than 300.76 How can these differences in urban population densities be explained? In the case of medieval and early-modern Italy, the answer clearly lies in the architectural makeup of the cities and towns in question. Most medieval towns were agglomerations of one- and two-storey buildings.77 The population densities of such towns hardly ever exceeded 150 persons/ha. There were, however, some notable exceptions. One of these was medieval Genova, which was boxed in not only by its city walls but also by the surrounding mountains. Because of this, Genova had not only a large number of houses per hectare of built-up space, but also a disproportionate number of very tall buildings, many of which had six or more storeys.78 As we have seen, this is reflected in an unusually high urban population density.79 It can be no coincidence that Siena also had a lot of tall buildings,

75

76

77 78

79

For Genova, see above, note 59. According to Benevolo (1980), 326, Siena’s fourteenth-century walls enclosed c. 180 ha, but I have used the lower figure of 165 ha given by Bortolotti (1983), 30. Cf. Piccini (2003). The figure of 101 ha given by Beloch (1937–1961), ii, 150, and by Chandler and Fox (1974), 92, is far too low. For the estimated populations of Genova and Siena, see Malanima (1998). For the population of Genova, cf. also Ginatempo and Sandri (1990), 69–70, 248–9. The Italian pattern is very similar to that found in the northern Netherlands in the mid-sixteenth century. In 1560 the average urban population density for all towns in the northern Netherlands was c. 130 persons/ha. Nijmegen and Dordrecht, however, had 200 people/ha, while the fast-growing city of Amsterdam had 300 inhabitants/ha. See the valuable collection of data in Visser (1985), 15–17. For late-medieval France, see Mols (1956), iii, 189–92; Russell (1972) 146–59; and Dollinger (1977), 66, concluding that ‘la densité de population des villes médiévales variait entre 50 et 150 habitants à l’hectare’. The extremely high density of 500 persons per hectare often given for the town of Albi refers solely to the inner city. As noted by Russell (1972), 156, Albi as a whole had about 10,700 inhabitants on 100 hectares, implying a density of only 107 to the hectare. Pounds (1974/1988), 24, 275; cf. Pounds (1969). The rich collection of data assembled by Mols (1956), iii, 189–96, points to a strong correlation between the number of houses per hectare of urban space and urban population density. The considerable discrepancies revealed by his figures can only be explained by taking into account average building height. See e.g. Brondy (1974) for the exceptionally well-documented case of latemedieval Chambéry, which had only about twenty houses per urban hectare. Since many of these houses had three storeys, however, the number of inhabitants per hectare was still as high as 200. As noted by Pinol (2003), 474, late-medieval Marseilles and Genova both had about sixty houses per hectare. Because average building height was much higher in Genova, however, this town had about twice as many inhabitants per hectare. For the very unusual domestic architecture of Genova, see Heers (1962), 402; Chandler and Fox (1974), 5; Ginatempo and Sandri (1990), 69. As noted by Dollinger (1977), 66, n. 19, the presence of many tall stone-built houses also explains the high density (485 persons per hectare) of the urban population of eighteenth-century Bayonne.

The urban network of Cisalpine Gaul in 28 bc

217

four or five storeys being the norm in the central urban area.80 As a result of this, the streets of Siena feel like alleys, although they are not particularly narrow. A third example is late-medieval and early-modern Naples, where again we find a combination of tall buildings and high urban population densities.81 In many cases differences in average building height seem to have been connected with the social structures of the towns in question. The central areas of late-medieval and early-modern Naples, for instance, contained a large number of urban palazzi, but were also home to a huge urban proletariat living in appalling conditions. In such cases we must reckon not only with a high density of houses and rooms per hectare of urban space, but also with a large number of people per unit of domestic space. An extreme example of such conditions in recent times is provided by Bari vecchia, where 40,000 people were living in only 10,000 rooms in 1949.82 While the ratio of four to one implied by these figures is way out of step with the one person per room rule sometimes cited as a cross-cultural norm,83 it may well not be atypical of conditions in those medieval and early-modern towns or town quarters in which destitute proletarians made up a large proportion of the population.84 Although these exceptions are of great interest, it seems possible to conclude that across most of pre-modern Europe population densities of between 120 and 180 persons per hectare were normal in towns whose houses had no more than two storeys apiece on average. It must, however, be emphasized that this conclusion cannot mechanically be applied to towns in other parts of the pre-modern world. It has, for instance, been observed that significantly higher urban population densities appear to have characterized the towns of early-modern Syria and other parts of the Islamic world. One illustration of this is early-modern Aleppo, which seems to have

80

81

82

83 84

On the domestic architecture of medieval Siena, see Balestracci and Piccini (1977) and especially Sabelberg (1984), 49–57 (containing a good discussion of the ‘Massenwohnhäuser’, which had four to six storeys). In the 1530s, Naples had c. 150,000 inhabitants on 350 ha (Benevolo 1980, 326). The implied urban population density is c. 430 persons/ha. Cf. Beloch (1886), 409, for very high densities in the central districts of nineteenth-century Naples. Amendola and d’Elia Belli (1997), 56. As early as the 1790s, when Bari vecchia had 511 inhabitants per hectare, many people lived ‘nei sottoterranei, o nei sottani delle case, od in piccoli piani superiori’ (Petrignani 1981, 20). Wallace Hadrill (1994), 100, referring to Kolb (1985), 584. Very high densities (of up to 400–500 persons per hectare) are also recorded for the agro-towns of late-medieval Puglia (Vlora 1985, 24), where landless lavoratori made up a large proportion of the population.

218

Augustan census figures and Italy’s urban network

had about 300 inhabitants per hectare of urban space.85 The explanation for these higher densities seems to lie in a different tradition of domestic architecture. Both in Syria and in other parts of the early-modern Middle East, urban houses tended to have small ground plans, at least two storeys and a large number of small rooms. If these observations are combined with the (admittedly questionable) rule of thumb that in most pre-modern towns the number of inhabitants equalled the number of rooms, the relatively large numbers of people found in ‘eastern’ cities become intelligible. In view of the many factors known to have influenced urban population densities in the pre-modern world, we must be cautious in applying these general insights to the towns of the Roman empire. Even if we tread very carefully, however, it is difficult to avoid the impression that differences in building height, in house plans and in the social makeup of urban populations go a long way towards explaining the enormous variations in urban population density that seem to have existed both in Italy and in other parts of the Roman world. Let me begin with early-imperial Rome. Although some scholars have assigned fewer than 500,000 inhabitants to the city at this time,86 most ancient historians continue to support the old figure of between 800,000 and one million. If the entire population lived within the area later enclosed by the Aurelian wall, the average number of people per hectare would have been between 580 and 730. If we include the suburbs beyond the Aurelian wall, this figure drops to between 440 and 560.87 Between 20,000 and 58,000 people are thought to have lived within Ostia’s walled area, which covered 69 ha. The urban densities implied by these figures range from 290 to 840 persons/ha.88 The most recent estimates are closer to the lower end of this band.89 There can be no doubt that the key to these high figures lies in the architectural makeup of Rome and Ostia and in the social composition of their populations. Even though we know disappointingly little about the 85

86 87

88

Marcus (1989), 337–41. While the number of urban hectares stood at 365, estimates of the size of the urban population depend on the multiplier used to extrapolate the number of inhabitants from the number of households, thought to have been c. 15,000 in the mid-eighteenth century. Although the lowest population figure ever given implies a density of only 205 persons per urban hectare, an estimate of the order of 100,000 inhabitants and a density of c. 300 persons per hectare seem more realistic. As noted by Russell (1972), 200 and 208, densities of between 200 and 300 inhabitants to the hectare also characterized the Islamic cities of the medieval period. Storey (1997). Hopkins (1978), 97: 1,373 hectares; Morley (1996), 34: 1,783 ha; Lo Cascio (1999a), 165: 1,800 hectares. Duncan-Jones (1982), 276. 89 Storey (1997), 973–5: 319 persons/ha.

The urban network of Cisalpine Gaul in 28 bc

219

domestic architecture of late-republican and early-imperial Rome, the existence of tall buildings in this city is documented as early as the third century bc.90 We may also point to the Augustan regulation that established 70 feet as the maximum height for new buildings in Rome. If we assume that each floor had a ceiling height of between 12 and 15 feet, it follows that Augustus would have expected such buildings to have five or six storeys.91 Tall buildings were also characteristic of Ostia, where the average height of residential buildings was between 2.5 and 4 storeys.92 In the light of these data, it is quite reasonable to assign to Ostia and Rome population densities of between 300 and 600 persons per ha. Another interesting example is the Roman-Egyptian town of Hermopolis, which seems to have had some 37,000 inhabitants in the mid-third century ad.93 Although the exact number of urban hectares here cannot be determined, most recent publications opt for a (relatively low) figure of 120 hectares.94 If we accept this figure as at least approximately correct, we obtain a rough estimate of 300 inhabitants per urban hectare.95 Thanks to the rich records that have survived regarding the towns of Roman Egypt, it is possible to shed some light on the factors behind this relatively high figure. As Bagnall has observed, almost all of the houses in these towns seem to have had at least two storeys.96 The domestic architecture of Hermopolis conforms to this basic pattern to the extent that twostorey houses appear to have been the most common type. Recent analysis of the papyrological evidence has, however, revealed that in this particular town the average number of storeys per house was two and a half.97 It is also striking that the town houses of Roman Egypt were characterized by ground plans between three and four times smaller than those of the houses in Olynthus, Priene, Pompeii and Herculaneum.98 On the basis of these data, it seems possible to conclude that the high urban densities that can be calculated for Hermopolis and some other Egyptian towns reflect a tradition 90 91

92 93 94

95 96 98

Yavetz (1958), 506, and Patterson (2006b), 353, both referring to Livy 21.62.3. Yavetz (1958), 507, referring to Str. 5.3.7 and Suet. Aug. 89. For a good discussion of the archaeological evidence for multi-storey insulae in Rome, see Wallace-Hadrill (2000), 204–8. Duncan-Jones (1982), 277. Bagnall and Frier (1994), 55; cf. Tacoma (2006), 44: slightly more than 45,000 inhabitants. Although Van Minnen (2002), 286, n. 8, is right to point out that the extent of the built-up area of Hermopolis remains unclear, the discussion in Roeder (1959), 107, suggests to me that 190 hectares is a reasonable upper limit for the area covered by the four quarters of the city plus the suburbs. For an estimate of c. 120 hectares, see e.g. Bagnall (1993), 52. If we assume the town to have covered 190 ha, this figure drops to c. 200 persons per hectare. Bagnall (1993), 49 and 52–3. 97 Alston (2002), 59–60. Alston (2002), 53. Note that the papyrological evidence points to a high number of occupants per house (ibid. 72).

220

Augustan census figures and Italy’s urban network

of domestic architecture very similar to that typical of many Middle Eastern towns before the nineteenth century. Other towns in the Roman world had far fewer inhabitants per hectare of built-up space. A well-documented example is the Latin colony of Cosa, where twenty-four larger and 224 smaller houses were discovered.99 The ratio between the two types is approximately one to ten, suggesting that the bigger houses belonged to equites and the smaller ones to foot soldiers. In any case, it seems reasonable to assign a family of five or six to each of the smaller houses and ten to twelve people to each of the remaining twentyfour, which were approximately twice as big. This would give the colony between 1,360 and 1,632 free and non-free inhabitants, or between 101 and 121 persons/ha.100 It should be noted that this is the density for the entire area enclosed by the town wall. The density per hectare for domestic space alone would be approximately 25 per cent higher. Interestingly, the relatively abundant archaeological evidence from Cosa again suggests a connection between house types and urban population densities. From recent analyses of this town’s domestic architecture it would appear that the houses of the ordinary colonists who made up the majority of the population had no upper storeys.101 So far, the only domestic building in which traces of an upper floor have been detected is the so-called House of the Skeleton, dating from the early first century bc. It is perhaps no coincidence that this was a larger than average house, occupying the area of five former gardens.102 Cosa is therefore a clear example of an Italian town with low buildings and a correspondingly low urban population density. This combination seems to have been typical of many of the smaller towns of the ancient world.103

99 100

101 102

103

Fentress et al. (2003), 24. Assuming five people per household, Fentress (2009), 141, allows Cosa 1,360 inhabitants in the early second century bc. Brown (1980), 18, however, estimates its population at only 1,100. In Roman Egypt the average number of people per household in the metropoleis was 5.31. Egypt’s population structure is thought to have been similar to that of Italy. See Bagnall and Frier (1994), 68. Brown (1980), 64–7 and Fig. 81, 83, 85–7; Bruno and Scott (1993), esp. Fig. 6–9, 11–12, 20–1, 25–6. Brown (1980), 67–8; Bruno and Scott (1993), 142–3. According to Fentress et al. (2003), 34–42, another large house, the so-called House of Diana, had no upper floor. Hansen (2006), 61–3, assumes a density of 150 persons/ha of inhabited space in the Greek poleis of classical times. If we assume an average urban population density of 120 persons/ha for the towns of Roman Cisalpina and take it that 20 per cent of the area of towns in this region was occupied by public spaces and buildings, we too end up with a density of 150 people/ha for domestic space. For the Greek town of Halieis, Jameson, Runnels and Van Andel (1994), 549–52, assume 250 persons/ha for domestic space. If 20 per cent of this town’s built-up area was occupied by streets and public buildings, the implied density is 200 persons/ha. In the case of Hermion, the same authors assume that domestic space accounted for 75 per cent of the built-up area.

The urban network of Cisalpine Gaul in 28 bc

221

For the purposes of the present investigation, our central question must be whether the towns of Cisalpine Gaul were more like Rome or more like Cosa. In my view, there are good grounds to believe that the latter alternative is more likely to be correct. One of the reasons why Rome had so many tall buildings is that the oldest part of the city had been boxed in by the Servian wall for a very long time. Expansion was impeded by the peripheral location of cemeteries, which made it difficult for suburbs to develop to the west and south. A closely related factor is that the rapid expansion of the population had the inevitable effect of pushing up the price of land within and near the city walls. This means that the owners of rented apartments would have needed to charge a high rent per square metre. Their options would, therefore, have been to build expensive apartments that could be rented to wealthy people or to build high-rise insulae containing many cheaper domestic units. As several studies of the living conditions of the Roman plebs have demonstrated, many builders opted for the latter alternative.104 If we apply these ideas to the north, it is surely not without significance that many northern towns were recent foundations. Interestingly, many towns in Cisalpine Gaul appear to have acquired walls only in the final decades of the republican period or during the Augustan period.105 Many other towns remained unwalled under the Empire.106 Moreover, because most northern towns were recent foundations, they had not yet become locked in by their suburban cemeteries. If these towns expanded during the late Republic, then, one would expect them to have grown laterally rather than vertically. The archaeological evidence seems to support this inference. In a recent article that synthesizes the results of several decades of research into the domestic architecture of Aemilia, Jacopo Ortalli argues that the walled areas of the colonial towns south of the Po were not fully occupied by buildings from the moment of their foundation. At least during the early decades of their existence, some plots of land were left vacant (perhaps with the 104

105

106

Yavetz (1958). As Wallace-Hadrill (2000), 205, points out, many insulae are likely to have had a heterogeneous population made up of tenants of different social categories. For examples, see Conventi (2004): Mutina (42 bc), Alba Pompeia (Augustan), Brixia (Augustan), Comum (59 bc), Mediolanum (40–35 bc), Ticinum (Augustan), Verona (mid-first century bc), Vicetia (second half of the first century bc), Concordia (Augustan), Tergeste (33–32 bc), Tridentum (second half of the first century bc), Augusta Taurinorum (Augustan), Augusta Praetoria (Augustan) and Augusta Bagiennorum (vallum under Augustus). For general discussions of the town walls of Cisalpina, see Chevallier (1983), 104–6; Mansuelli (1971), 120–1, and Tiussi (2002/2003), 82. E.g. Libarna (Conventi 2004, 89), Ateste, Caesena, Forum Livii and most of the smaller towns listed in Appendix i.

222

Augustan census figures and Italy’s urban network

conscious aim of leaving room for later immigrants) and most or perhaps all blocks of buildings appear to have been surrounded by vacant strips of land.107 This arrangement is very similar to that in Cosa during the second century bc. At a later stage, which may have begun as early as the mid-second century bc, in at least some of the towns studied by Ortalli the strips separating the oldest blocks of houses gradually became built up, suggesting an increase in urban population densities. This trend seems to have continued under the early Empire, when many areas previously used as gardens were built on.108 Although the archaeological evidence is still fragmentary, it seems reasonable to conclude that the colonial towns of Cispadana were initially characterized by low urban population densities. During the late Republic the number of inhabitants per urban hectare seems to have risen, in conformity with the pattern that can be observed in many walled towns during periods of population growth. Given the archaeological evidence for low densities during the earlier stages of population growth, however, it remains plausible that even in the northern Italian colonies of the third and early second centuries the increasing density of the urban fabric had not yet reached levels any higher than, say, 150 inhabitants per urban hectare at the time of the census of 28 bc. Needless to say, lower densities are to be expected in unwalled towns and in settlements that did not develop an urban character until after the Social War. This impressionistic conclusion can be confirmed by looking at the architectural makeup of individual houses. In a valuable survey of the domestic architecture of Cisalpine Gaul as a whole, Daniela Scagliarini notes that the ground plans of northern houses tended to be spacious and that there is very little evidence for the building of upper storeys in the north. Her main finding is that the domestic architecture of northern Italy was ‘extensive’ compared to that of Pompeii.109 In other words, urban 107 109

Ortalli (2001), 44. 108 Ortalli (2001), 45–6, 48. Scagliarini (1983), 304: ‘Non vi è nulla che indichi concretamente una parcellazione delle insulae in piccole unità di abitazione con sfruttamento intensive dello spazio . . . Le planimetrie sono estensive, è scarsamente documentata la presenza di un secondo piano.’ Cf. Chevallier (1983), 147: ‘Elles [les maisons] ont en général un seul étage, même à basse époque, mais de grandes pièces (50 à 70 m2, très tôt)’. This conclusion was based on a corpus of 501 private houses (ibid. 148), of which only a small proportion had been properly published. Cf. Passi Pitcher (2003), 169: no evidence for second storeys in Cremona. These observations do not, of course, mean that there were no two-storey houses. See e.g. Ortalli (2000c), 513; Morandini (2001), 362; Maggi (2001), 302, 306; Manzelli and Grassigli (2001), 133–5, 147; Tirelli (2003a), 51, for some possible and definite examples from Ariminum, Brescia, Libarna, Ravenna and Opitergium. All known examples seem to be of imperial date. It does not seem far-fetched to interpret the gradual spread of two-storey houses as a sign of continuing population growth under the Empire. Cf. Chapter 6, at notes 81–4.

The urban network of Cisalpine Gaul in 28 bc

223

population densities in the north were low. In a more recent study, Michele George notes that the houses of northern Italy at this time were in many ways more similar to those of northern Africa and southern Gaul than to those in central Italy.110 Since urban population densities in Roman Gaul are thought to have been lower than 150/ha, this observation points in the same direction as Scagliarini’s earlier findings.111 Finally, it seems significant that during the extensive excavations carried out at Aquileia, a free-lying site, no traces of Ostia-type insulae were discovered.112 This suggests to me that even in Cisalpina’s larger towns the average height of buildings was low. Of course, it could be objected that our knowledge of the makeup of the towns of Cisalpine Gaul is based on a limited corpus of houses excavated in a large number of towns. Part of my response to this would be that our understanding of the domestic architecture of Cisalpine Gaul has advanced considerably in recent years and that there appears to be no reason to believe that the very numerous publications in this field offer us a misleading picture of housing conditions in the north. Be that as it may, we must view the picture that emerges from the comparative data as highly significant. As we have seen, these data suggest that in pre-modern Europe and in the Middle East high urban densities were associated with high-rise buildings, small ground plans and large numbers of small rooms. Since none of these features seems to be characteristic of the domestic architecture of late-republican and early-imperial Cisalpina, the onus of proof is surely on those who would argue that the northern towns of Roman Italy must have had a large number of inhabitants per hectare of built-up space. As long as no cogent argument in favour of this view has been put forward, the deductive-comparative argument must remain on the table, especially because there is nothing in the archaeological record to contradict it. Although any attempt to put a figure on population densities in the north must remain to some extent speculative, the fact that many northern towns had no walls and were of relatively recent creation, together with the 110

111

112

George (1997), 32–3. Cf. also Tosi (1992b), 382–3, Maioli (2000a), 183, Ortalli (2003), 96–7, Cavalieri Manasse and Bruno (2003), 47, and in particular the useful survey by Barbieri and Manzelli (2006), 340–6. For Transalpine Gaul see Goudineau (1980), 310: ‘une densité de 150 à l’hectare constitue un seuil qui ne fut sans doute franchi par aucune ville de province’, followed by Woolf (1998), 137 and n. 103. In my view, this claim is broadly valid for towns in the western half of the Empire during the late Republic and early Empire, but not for all of the eastern provinces. Cf. above, at notes 96–8. See e.g. Mirabella Roberto (1987), 357: ‘non sono noti . . . edifici di carattere popolare, magari a più piani’. The same observation may be found in Maselli Scotti (2002), 57, and Mian (2003), 84–6.

224

Augustan census figures and Italy’s urban network

archaeological evidence for the ‘extensive’ use of urban space, points to the conclusion that at the time of the census of 28 bc average urban population densities in Cisalpine Gaul did not exceed 150 inhabitants per hectare of built-up space. 5.2.4 A low-count model for Cisalpine Gaul We have now reached a point at which we can begin to feed some figures into our formula. Let me begin with the low count for northern Italy at the time of Augustus. As we have seen, the proponents of the low count have to assume that between 900,000 and 1 million free people were enfranchised in Transpadana in 49 bc.113 To this figure we must add the inhabitants of Aemilia and those of a handful of Transpadane communities that were given citizenship either at the date of their foundation or in 90 bc.114 The only way to get a rough idea of the number of northern Italians living in communities given citizenship before 49 bc is to look at the physical size of the towns concerned. From the archaeological data it would appear that the large and medium-sized towns belonging to this category covered some 605 hectares, approximately half the area covered by the remaining towns of Cisalpine Gaul.115 It therefore seems reasonable to assign to the region of Aemilia and the areas already inhabited by citizens before the enfranchisement of the remaining Transpadani a free population of between 400,000 and 500,000, about half the low-count figure for those parts of Transpadana and Liguria whose inhabitants were given citizenship in 49 bc.116 These estimates imply a free population of between 1.3 million and 1.5 million for Cisalpine Gaul as a whole. If we include slaves and a limited number of foreigners, these estimates can easily be increased to 1.5 and 1.7 million. If we feed these figures into our calculations and combine them with my estimate of the number of urban hectares, we can reconstruct a range of scenarios by varying urban population densities and/or urbanization rates. By way of illustration, I shall list eight possible scenarios giving a northern population of between 1.5 million and 1.7 million people. The first four of 113 114

115 116

Above, at note 4. Here one thinks of Eporedia and of those parts of Liguria whose viritane settlers were of citizen status. Appendix i. In extrapolating the population of Aemilia from that of Transpadana and Liguria, we must also bear in mind that the latter areas were characterized by greater inter-city distances and hence presumably by lower urbanization rates. See Bekker-Nielsen (1989), 25.

The urban network of Cisalpine Gaul in 28 bc

225

Table 5.1 Eight low-count models of urbanization in early Augustan Cisalpina

1. 2. 3. 4. 5. 6. 7. 8.

Target population

Number of urban hectares

Urban population density

Urbanization rate

1.7 million 1.5 million 1.7 million 1.5 million 1.7 million 1.5 million 1.7 million 1.5 million

1,765 1,765 1,5251 1,525 1,5902 1,590 1,370 1,370

150 p.ha 150 p.ha 120 p.ha 120 p.ha 150 p.ha 150 p.ha 120 p.ha 120 p.ha

15.6% 17.7% 10.8% 12.2% 14.0% 15.9% 9.7% 11.0%

1

With 120 inhabitants per hectare, only towns covering at least 25 ha would meet my critical threshold of 3,000 inhabitants. If we use this cutoff point, we are left with c. 1,525 urban hectares in early-imperial Cisalpina. See Appendix i. 2 0.9 × 1,765 ha = 1,588.5 ha.

these are based on the assumption that my estimated figures for the number of hectares covered by the large and medium-sized towns of Cisalpine Gaul at the time of the census of 28 bc are 100 per cent accurate; the last four make the perhaps more realistic assumption that they are c. 10 per cent too high.117 It is also possible to run a further test by considering urban population densities and urbanization rates in Transpadana alone. If we assume that between 900,000 and 1 million men, women and children living in this area were given citizenship in 49 bc and if we take it that slaves made up 15 per cent of the population, we end up with a total population of between 1.06 million and 1.18 million.118 These target populations are compatible with the eight scenarios listed in Table 5.2. As is immediately apparent, all of the northern Italian urbanization rates generated by these sixteen models fall short of the rate of (at least) 17.9 per cent that can be reconstructed for the same area at the start of the seventeenth century. They also fall short of the average ‘economic’ urbanization rate of 20 per cent suggested by Hopkins’ calculations for Italy as a whole. In fact, eight of our results are closer to 10 per cent than to 15 per cent. 117 118

For a justification of this assumption, see above, section 5.1.4. In this crude calculation, free residents of peregrine status have been ignored. Note, however, that the areas inhabited by attributi fall outside the scope of my inquiries.

226

Augustan census figures and Italy’s urban network Table 5.2 Eight low-count models of urbanization in early Augustan Transpadana

1. 2. 3. 4. 5. 6. 7. 8. 1

2

Target population

Number of urban hectares

Urban population density

Urbanization rate

1.18 million 1.06 million 1.18 million 1.06 million 1.18 million 1.06 million 1.18 million 1.06 million

1,160 1,160 1,0001 1,000 1,0452 1,045 900 900

150 p.ha 150 p.ha 120 p.ha 120 p.ha 150 p.ha 150 p.ha 120 p.ha 120 p.ha

14.7% 16.4% 10.2% 11.3% 13.3% 14.8% 9.2% 10.2%

With 120 inhabitants per urban hectare, we are left with c. 1000 urban hectares in early-imperial Transpadana. See Appendix i. 0.9 × 1,160 ha = 1,044 ha.

It also appears from these calculations that the low count can be made to generate plausible urbanization rates only if we assume that the average urban population density in towns covering 20 or more hectares was somewhere between 120 and 150 people per hectare. These values are somewhat higher than the density reconstructed for mid-republican Cosa, but roughly comparable to those found in early fourteenth-century Bologna and Padova and in sixteenth-century Milan.119 Although there is perhaps a theoretical possibility that the similarity between these later figures and the results generated by my sixteen low-count models for Roman Cisalpina is purely accidental, this finding makes it difficult to reject these models as unrealistic. I conclude that the low count for Cisalpine Gaul is associated with urbanization rates that are considerably lower than those calculated for early-modern times, but are nevertheless historically plausible. At the same time, the northern urban population densities implied by the low count are very similar to those that characterized the urban landscapes of Tuscany and Emilia-Romagna from the fourteenth to the sixteenth century, and are also compatible with the archaeological evidence that we currently have.

119

Cf. section 5.2.3.

The urban network of Cisalpine Gaul in 28 bc

227

5.2.5 A high-count model for Cisalpine Gaul? At this point we must ask whether it is also possible to construct an alternative model compatible with the high-count interpretation of Italian population development. Since the high count implies a total Italian population of at least 15 million people, of whom 40 per cent are assigned to Cisalpine Gaul, we are faced with the task of manipulating our formula in such a way as to obtain a northern population of approximately 6 million.120 Since the large and medium-sized towns of northern Italy cannot have covered an area of many more than 1,765 hectares, this can only be done by varying the number of people per urban hectare and/or the rate of urbanization. Just by way of illustration, I shall present four scenarios in which these two variables have been adjusted in such a way as to meet the requirements of the high-count model. Table 5.3 Four high-count models of urbanization in early Augustan Cisalpina Number of urban hectares 1. 2. 3. 4. 1

1

2,100 2,100 2,100 1,765

Urban population density

Urbanization rate

400 p.ha 300 p.ha 225 p.ha 150 p.ha

14% 10.5% 7.9% 4.4%

With 400 persons/ha, we must assume that all settlements that counted as towns in a juridical sense were densely built up. I have therefore added my estimate for the total area covered by the small towns to those for the large and medium-sized towns. Since there are no grounds to suppose that the villages of Roman Cisalpina had the kind of architecture required to support such high densities, however, the vici of this region can be ignored. Cf. note 53.

In other words, it is entirely possible to fit 6 million people into Cisalpine Gaul, but only at the cost of making this area very sparsely urbanized or giving its towns a very high urban population density. The first alternative is directly at odds with Lo Cascio’s assumption that between 15 per cent and 20 per cent of the population of Augustan Italy lived in towns with 2,000 or more inhabitants. As we have seen, it also sits very uneasily with the likelihood that a consistent pattern of elite residence in towns resulted in a relatively high urbanization rate. The second alternative raises the question of why the fledgling towns of Roman northern Italy should have had population densities two or three times as high as those of most of their counterparts in Renaissance Italy. One possible way round this would be to assume that the towns of Roman 120

Above, section 5.1.1.

228

Augustan census figures and Italy’s urban network

Cisalpina had many more three-storey and four-storey buildings than their late-medieval and early-modern successors. As we have seen, however, archaeological evidence regarding the urban makeup of the northern towns points in precisely the opposite direction. I conclude that our examination of the shape of Cisalpine Gaul’s urban network at the time of the first imperial census has provided us with a number of good reasons to believe that the high-count interpretation of the republican and Augustan census figures is unlikely to be correct. 5.3 the population of central and southern italy In the second part of this chapter, I shall extend the scope of my inquiries to include the shape of the urban network in the central and southern regions. Before doing so, I should like to make it clear that the archaeological evidence from these areas is generally more problematic than that from Cisalpine Gaul. To be sure, in many cases the problems posed by these archaeological data are similar to those we encountered in the north. For instance, just like most of northern Italy’s Roman towns, many Roman towns on the peninsula lie buried under modern agglomerations. I have already argued that this is not an insuperable obstacle. In many parts of central and southern Italy we are, however, faced with additional complications. As we have seen, there is reason to believe that many (perhaps most) northern towns were ‘compact’ in the sense that there were few large empty spaces within their walls. There can be no doubt that things were very different in other parts of early-imperial Italy. Both in Etruria and in Magna Graecia, for instance, many towns were surrounded by walls erected between the sixth and third centuries bc. It has long been recognized that as a rule only part of the area enclosed by old walls of this type was inhabited. For example, in many of the towns of Magna Graecia only between 25 per cent and 50 per cent of the walled area is thought to have been built upon.121 Similarly, pre-Roman town walls in Apulia often enclosed very large areas. In most cases indications suggest that only 50 or 60 per cent of these areas may have been inhabited.122 Less dramatic instances of this phenomenon may be found in Campania, Latium and Samnium. It is generally agreed, for instance, that a large proportion of the c. 66 hectares enclosed by Pompeii’s sixthcentury wall remained sparsely inhabited for a very long time, with many parts of the town becoming urbanized only during the second and first 121

Hansen (2006), 42.

122

E.g. Yntema (2008), 381.

The population of central and southern Italy

229

centuries bc.123 It follows from these observations that in many cases the course followed by the town walls is an unreliable guide to the extent of the built-up area. Another problem is that in some parts of southern Italy the archaeological evidence points to a process of urban contraction during the last two and a half centuries bc.124 It is widely accepted, for instance, that most pre-Roman towns in the Messapian districts of Apulia had far larger populations than their Roman successors, and the same is true of at least some of the Greek towns of southern Italy. Since in many cases the reduced populations of these towns continued to live in quarters that already existed in classical or Hellenistic times, it is often extremely difficult to put any figure on the extent of the areas still inhabited during the late Republic and early Empire.125 This problem is exacerbated by the fact that until recently archaeological research in the towns of Etruria, Magna Graecia and Apulia tended to focus on the pre-Roman period, for the obvious reason that this was the time at which most of these towns reached their acme in terms of building activity and population size. It seems possible to conclude that, generally speaking, it is more difficult to gauge the size of the areas covered by the towns of central and southern Italy at the time of the census of 28 bc than it is to estimate the size of the towns of Cisalpine Gaul. In this sense, the second part of this chapter will be more speculative than the first. On the other hand, the problems just described do not make it entirely impossible to reach any meaningful conclusion. To begin with, many of the large and medium-sized towns of Picenum, Samnium and northern Lucania were new foundations of the third-to-first centuries bc. Their development can thus be followed using the same methods that we applied to the northern towns. In some other areas, such as Campania, it is possible to find many towns in which a very large proportion of the area enclosed by the pre-Roman walls had become fully built up by the end of the Republic. Finally, and perhaps most importantly, it is now possible to discern a clear tendency among archaeologists working on the towns of Etruria, Apulia and Magna Graecia to pay more attention to the development of regional urban networks after the Roman conquests of the fourth and third centuries bc. In

123 125

E.g. Schoonhoven (2003). 124 For examples, see Appendix ii. This difficulty is exemplified by Roman Tarentum, Lupiae, Rhegium and many other towns. Cf. Appendix ii.

230

Augustan census figures and Italy’s urban network

at least some cases, this upsurge in interest in later developments has resulted in rough estimates of the extent of the areas that continued to be inhabited during the late Republic and early Empire. For all these reasons, it seems well worth collecting together the data that we already have, despite the fact that many of our preliminary findings will undoubtedly have to be revised in the light of future investigations. 5.3.1 Expected urbanization rates in central and southern Italy In ad 1600 the regions of central and southern Italy were inhabited by approximately 5.5 million people, 1,215,000 of whom lived in towns with 3,000 or more inhabitants. The urbanization rate implied by these figures is 22.1 per cent.126 This is considerably higher than the rate of 17.9 per cent that can be reconstructed for northern Italy during the same period. The main reason for this discrepancy is that many southern Italian towns of the early-modern period were ‘agro-towns’ in which small farmers, tenants and agricultural labourers made up a large proportion of the population. Without these groups, the central and south Italian urbanization rate would have been similar to that found in the northern regions.127 There are some indications that average urbanization rates in the central and southern regions also exceeded those in the north at the time of the census of 28 bc (even if we ignore Rome), mainly because northern Italy had fewer substantial towns at the beginning of the Principate than at the outset of the seventeenth century. In addition, as we have seen, average inter-city distances in Cisalpine Gaul were considerably greater than those in the centre and south.128 These figures are, of course, based on a juridical and administrative definition of ‘town’. There can nonetheless be no doubt that the towns of central-western Italy were closer together than those in most parts of Cisalpina.

126

127

128

Bairoch (1989). Since modern lists of towns that had populations of 3,000 or more in ad 1600 are likely to be incomplete, the real rate must have been somewhat higher. The rate of 16.2 per cent given by Lo Cascio and Malanima (2005), 209, is based on a cut-off point of 5,000 inhabitants. Cf. Del Panta et al. (1996), 43–4. In the southern half of the peninsula the average urbanization rate was also pushed up by the fact that grain imports from Sicily, Sardinia and France made a significant contribution to the food supply of Naples, which had c. 280,000 inhabitants in 1600. Since earlymodern Naples got about 60 per cent of its grain from the Terra di Lavoro (Alifano 1996, 118), and because large amounts of grain were also imported from Puglia, the distorting effect of external supplies should not be over-rated. If we assume that a sixth of the grain consumed by the population of Naples came from outside the peninsula, we need only deduct c. 50,000 Neapolitans. This would reduce the central and south Italian urbanization rate by less than 1 per cent. Above, at note 67.

The population of central and southern Italy

231

Essentially the same picture emerges if we begin from a quantitative definition of ‘town’ and calculate the number of large or medium-sized urban settlements per square kilometre of territory. If my assessment of the sizes of the northern towns is correct, Cisalpine Gaul had forty-six large or medium-sized towns in 28 bc. This works out as one large or medium-sized town per 2,109 km2. In the centre and south we find 104 large or mediumsized towns in an area of 133,500 km2, or one substantial town per 1,284 km2.129 These facts should lead us to expect a higher urbanization rate in central and southern Italy. For example, if we start from the average urbanization rate of 20 per cent that Hopkins suggests for Italy as a whole (without Rome) we will have to assign more than a fifth of the central and southern population to towns. If we opt for Lo Cascio’s range of 15 to 20 per cent for the whole of Italy without Rome, the average urbanization rate for the central and southern regions must have been much higher than 15 per cent and probably not much lower than 20 per cent.130 It must be emphasized that these suggested rates refer to central and southern Italy as a whole rather than to any specific region. If we focus on particular parts of the peninsula, we encounter many indications that urbanization rates varied enormously between regions and sub-regions. Again the number of large or medium-sized towns per square kilometre can provide us with some basic insights. In the first Augustan region (Latium and Campania) we find one large or medium-sized town per 470 km2, in the fifth region (Picenum) one per 722 km2 and in the sixth (Umbria and the Ager Gallicus) one per 714 km2. In the fourth and third regions (Samnium; Lucania and Bruttium), however, we find only one such town per 2,571 km2 and one per 3,438 km2 respectively. By assuming that the fourth and third regions had not only small urban populations but also small rural populations, we could still reconstruct relatively high urbanization rates for these areas. It seems, however, far more likely that, both generally speaking and in these specific cases, a scarcity of large and medium-sized towns indicates a regional urbanization rate falling short of the average figure for central and southern Italy as a whole.131 129 130

131

Appendix ii. This assumes that the average urbanization rate in northern Italy was lower than those of the centre and south. At the time of the census of 1861, only 12.2 per cent of the population of the Abruzzi and the Molise lived in towns with 5,000 or more inhabitants (Belletini 1987, 176; Malanima 1998). This is about a third of the average urbanization rate of 37.8 per cent calculated for southern Italy as a whole. Of course, the latter rate reflects the importance of Naples and the existence of many agro-towns in Puglia, Basilicata and Calabria. Cf. Del Panta et al. (1996), 43–4.

232

Augustan census figures and Italy’s urban network

In short, in calculating a range of urbanization rates for central and southern Italy as a whole we may well end up with figures that do not accurately describe the situation in any of its constituent parts. On the other hand, since exactly the same method has been used in the case of earlymodern Italy, it is still possible to compare Italian urbanization rates at the time of the census of 28 bc with those of ad 1600 without falling into the trap of comparing apples with oranges. Only in a very few cases is it possible to investigate the urbanization rates of specific areas. In the case of early-imperial Spoletium, for instance, valuable information can be gleaned from two imperial inscriptions concerning two foundations set up to finance free handouts of food and cash to the citizens (municipes).132 It has been demonstrated by Duncan-Jones that the number of people in receipt of these handouts was in the region of 4,700 and that this points to a citizen population of c. 16,450.133 Lo Cascio has used this to argue that the evidence relating to these early-imperial distribution schemes supports his thesis that many of the towns of Roman Italy were far more populous than is generally assumed.134 As Scheidel has pointed out, however, this conclusion follows only if we assume that communal cash handouts were offered exclusively to the urban citizens of the communities concerned. The only town in which such distributions are known to have been made to the plebs urbana alone is Saturnia, in Etruria. Here there were certainly fewer than 2,000 beneficiaries and quite possibly fewer than 1,000. These figures have been interpreted as supporting the low count.135 It may be possible to break the current deadlock by looking at Spoletium’s physical size. It has long been known that this town’s wall (of republican date) enclosed some 30 hectares of land, of which three-quarters were inhabitable.136 Nothing in the archaeological record suggests the existence of substantial suburbs of late-republican or early-imperial date.137 A simple calculation will show that even if we begin from the unrealistic assumption that the 30 hectares enclosed by the walls were fully built up, the hypothesis that all of the recipients of these handouts were urban implies an urban population density of roughly 550 people per hectare (excluding slaves). As we saw in the first half of this chapter, a figure of this order might be acceptable with respect to early-imperial Rome, but is clearly far too high for a moderately important town in which few houses are 132 133 135 137

CIL xi, 4789 and CIL xi, 4815. The latter inscription seems to belong to the second century ad. Duncan-Jones (1982), 267–8. 134 Lo Cascio (1999a), 165. CIL xi, 2650. Cf. Jacques (1993) and Scheidel (2008), 34. 136 The steep rocca occupies c. 7.5 ha. Morigi (2003), 143, detects no signs of suburban quarters of Roman date and concludes ‘che nella codifica del popolamento extramuraneo l’Antichità ha parte assai ridotta’.

The population of central and southern Italy

233

likely to have had more than two storeys. The only possible conclusion is that the handouts were given not only to members of Spoletium’s free urban population, but also to a large number of rural municipes. If we assume that there were between 150 and 200 people of citizen status per urban hectare, we can estimate the size of the free urban population at between 4,500 and 6,000. If Duncan-Jones is correct in stating that the entire citizen population numbered about 16,450, the urbanization rate amongst municipes and their families must have been somewhere between 27.4 and 36.5 per cent. Since slaves have been excluded from this crude calculation, the former percentage appears more realistic than the latter.138 In assessing the wider significance of these relatively high figures, it should be borne in mind that they refer to urbanization rates within the territory of a medium-sized town. Since, according to my definition, settlements covering fewer than 20 hectares do not count as ‘urban’, the inclusion of smaller centres and their territories would result in a substantially lower urbanization rate for central and southern Italy as a whole. 5.3.2 Expected urban population densities in central and southern Italy In my discussion of the northern towns, I suggested that the domestic architecture of early-imperial Cisalpina was characterized by an ‘extensive’ use of space. Two clues pointing in this direction are the apparently large size of the average room and the scarcity of archaeological evidence for the existence of multi-storey buildings. Building upon the observations of Chevallier, Scagliarini and Ortalli, I have interpreted these clues as suggestive of low urban population densities. It must be remembered that this argument rests on a contrast between the domestic architecture of the northern towns and that of towns such as Pompeii and Herculaneum. Some scholars have, however, suggested that Pompeii may have had a population of only 7,000 or 8,000, implying an urban population density of 110–25 persons per hectare.139 The main weakness of estimates of this low order is that they rest on the assumption that few Pompeian houses had upper storeys. The best discussion of this 138

139

A similar calculation could be made for the harbour town of Pisaurum. Here the annual revenue from a foundation of HS 400,000 may have been enough to provide between 2,500 and 5,000 adult male citizens with a free dinner, implying a total population of between 8,750 and 18,550 (Duncan-Jones [1982], 268). Unfortunately, any attempt to work out the approximate urbanization rate implied by these figures is complicated by the fact that Pisaurum seems to have acquired a substantial suburban vicus during the early-imperial period (e.g. Agnati 1999, 147–50). E.g. Eschebach (1975): 8,000 to 10,000; Russell (1977): 7,000 to 7,500.

234

Augustan census figures and Italy’s urban network

problem is that of Wallace-Hadrill. As he points out, the houses of Pompeii as a whole appear to have had between 9,000 and 10,000 ground-floor rooms.140 If we combine this estimate with comparative evidence suggesting an occupation density of one person per room for domestic space,141 we end up with 10,000 urban Pompeians and an urban population density of the order of 150 inhabitants per hectare. As Wallace-Hadrill goes on to explain, the figure of one person per room is almost certainly too high for the larger houses that were relatively numerous in Pompeii. On the other hand, he finds abundant evidence for the presence of second storeys.142 Although he refuses to put any figure on Pompeii’s population himself, the overall drift of his argument clearly indicates an estimate considerably higher than 10,000. If his arguments are correct, this particular town must have had substantially more than 150 inhabitants per hectare. As we saw during our discussion of the evidence from later pre-industrial Europe, a density of this order would not be surprising in a town whose entire walled area was already completely built over; it was even beginning to acquire a small suburb. The other side of the coin is that there is no evidence whatsoever that any free-standing house in Pompeii had more than two storeys.143 In Herculaneum the only multi-storey insula of the Ostian type was the three-storey apartment building in Insula Orientalis ii, which dates from the early first century bc.144 To the best of my knowledge no further archaeological evidence for the existence of three-storey or four-storey insulae has been discovered in central or southern Italy outside Rome and Ostia. Given the patchiness of the archaeological record, of course, it cannot be inferred from this that no such insulae ever existed.145 It seems, however, at least possible to conclude that there were very few such buildings in most central and southern Italian towns at the time of the census of 28 bc. Although the archaeological evidence that we currently have does not permit us to put any specific figure on urban population densities in the towns of Campania at this time, it certainly leads us to expect urban population densities of more than 150 persons (and perhaps even as many 140

141 143

144 145

Wallace-Hadrill (1994), 99–100: between 1,200 and 1,300 domestic units, with 7.5 rooms per unit on average. Ibid. 100. 142 Ibid. Of course, the multi-storey structures built onto the sloping terrain of western and southwestern Pompeii, such as the residential apartments of the Sarno Baths (e.g. Beard [2008], 40, 107, 113), do not belong in this category. Ibid. 104. The existence of three-storey and four-storey insulae in busy harbour towns such as Puteoli may be postulated.

The urban network in 28 bc

235

as 200 persons) per hectare.146 As we saw in the first part of this chapter, this hypothesis is in line with the evidence from Renaissance Italy, which leads us to expect urban population densities of up to 170 or 175 persons per hectare in towns containing a lot of two-storey buildings.147 On the other hand, it seems reasonable to suppose that the number of inhabitants per urban hectare was relatively low in areas such as Picenum and Samnium, where many of the Roman-style towns were still of relatively recent foundation. Similarly, one would expect to find low densities in those southern towns whose populations were in decline. The fact that many parts of central and southern Italy had a much longer history of Greek or Roman-style urbanism than the northern parts of Italy and the relatively abundant evidence for the existence of multi-storey houses in the large and medium-sized towns of Campania suggest to me that peninsular Italy must have been characterized by an average urban population density of at least 150 persons per hectare. 5.4 the urban network of central and southern italy in 28 bc: a brief outline If we list all of the towns in central or southern Italy that are likely to have covered more than 20 hectares in or shortly after 28 bc, we will arrive at an estimated total of 104 large or medium-sized towns. Of these, a disproportionately large number will be found in the first Augustan region (Latium and Campania). The sixth region (Umbria and the Ager Gallicus) also contained a considerable number of large and medium-sized towns. In the seventh region (Etruria) we encounter one very important town (Arretium) and nineteen others that appear to have covered between 20 and 40 hectares apiece. In absolute terms this is a high figure but it must be remembered that the area covered by the seventh region itself was very large (31,000 km2). If we examine numbers of large or medium-sized towns per square kilometre, we will find that Etruria trails behind Picenum (with eight large or medium-sized centres in an area of 6,500 km2) and Umbria (which had fourteen substantial towns in an area of 10,000 km2). Next come Apulia, with only thirteen substantial towns in an area of 25,000 km2, and Samnium, with seven medium-sized agglomerations in an area of 18,000 km2. At the bottom of the list we find Lucania and Bruttium, with 146

147

After considering a large number of methodological difficulties, Bowman and Wilson (2009b), 58, opt for 175 persons per hectare as the most reasonable estimate for Pompeii during the first century ad. Cf. my observations on Arezzo and Florence in the first half of the fourteenth century ad.

236

Augustan census figures and Italy’s urban network

only eight large or medium-sized towns in an area of 27,500 km2. Here the density of the urban network was between seven and eight times lower than that in Latium and Campania. If we look for the largest towns, we will find that (with the obvious exception of Rome) only twenty-two central or southern Italian towns are likely to have covered more than 40 hectares. This category can, of course, easily be expanded by adding a handful of urban centres that appear to have covered between 35 and 40 hectares. Nonetheless, the number of really large towns is surprisingly small.148 We will also find that (again with the exception of Rome) only five or six peninsular towns are likely to have covered more than 100 hectares. Four of these towns were in Campania: Capua (200 ha), Teanum (90–124 ha), Puteoli (100–20 ha) and Nuceria (116 ha). In the second region, Beneventum (100 ha?) may have belonged to this category, although there is some reason to believe that it may not have reached this size by the time of the census of 28 bc.149 Finally, the Lucanian town of Paestum is a possible candidate, although it must be stressed that we do not know how many of the 120 hectares enclosed by its town walls were inhabited during the laterepublican and early-imperial periods. This list shows us that Rome (with a built-up area of between 1,370 ha and 1,800 ha) was between seven and nine times larger than the second-largest Italian town in physical terms and perhaps twenty times larger in terms of population.150 I have already mentioned the matter of regional variations in the urbanization rates of central and southern Italy. A glance at my catalogue of central and southern Italian towns will show that these variations are connected with differences in the balance between large, medium-sized and minor towns. In Campania, for instance, most agglomerations that were towns in a juridical sense covered 20 or more hectares. This resulted in a dense network of ‘genuine’ towns. In every other area, including Latium, at least 50 per cent of all ‘towns’ were actually agglomerations covering fewer than 20 hectares apiece. The admittedly very deficient data from Apulia suggest that a staggering 80 per cent of ‘towns’ in this region were of this order. It follows from this that the large number of Apulian ‘towns’ should be seen largely as a juridical and administrative phenomenon that can tell us very little about the density of the ‘real’ urban network. In other areas we observe a 148

149

150

Note that Pompeii belongs to this select group. This casts considerable doubt on the traditional view that the finds from Pompeii shed no light on the character of the larger towns of Roman Italy. For signs of urban expansion in Beneventum during the first and second centuries ad, see Patterson (2006a), 110–13. For the high urban population densities likely to have characterized Rome, cf. above, at notes 86–92.

The urban network in 28 bc

237

N W

E

Luna Luca

Faesulae Florentia

Pisae

Arretium

Volaterra

S

Pisaurum Forum Sempronii Tifernum T.

Suasa

Ancona

Iguvium

Cupra Montana Helvia Ricina Cortona Tuficum Urbs Salvia Perusia Camerinum Firmum Clusium Asisium Fulginiae Rusellae Nursia Asculum Volsinii Tuder Novi Spoletium Saturnia Ameria Interamna Adria Nahars Vulci Ferentium Ocriculum Amiternum Teate Marrucinorum Lucus Graviscae Peltuinum Feroniae Carseoli Alba Fucens Tarquinii Veii Caere Marruvium Tibur Rome Praeneste Larinum Aletrium Ostia Ferentinum Velitrae

Adriatic Sea

Luceria

Aquinum

Fabrateria Venafrum nova Interamna Lirenas Teanum Allifae Herdonia Tarracina

Salapia (nova)

Antium

Cales Formiae Beneventum Capua Minturnae Nola Suessa Atella Abella Abellinum Cumae

Venusia Egnatia

Nuceria

Puteoli Neapolis

Brundisium

Salernum

Herculaneum

Pompeii Surrentum

Uria Paestum

Hydruntum

Velia

Medium

Ty r rh e n i a n

Lupiae

Tarentum

Grumentum

S e a

Thurii/Copiae

Large

Megapolis Scolacium Vibo Valentia

Rhegium

0 0

50 25

100 50

150 75

200 km

100 miles

Map 5.2 The large and medium-sized cities of central and southern Italy, c. 28 bc

238

Augustan census figures and Italy’s urban network

dearth of very large towns, but a dense network of medium-sized centres. This pattern is exemplified by Umbria and the Ager Gallicus, where we find twelve medium-sized towns, but only two centres covering more than 40 hectares. The urban network of Samnium is to some extent similar. Here we encounter seven towns covering between 20 and 40 hectares apiece, but not a single centre of any greater size. Because the medium-sized towns of Samnium were further apart than their Umbrian counterparts, however, the overall shape of the urban networks of these two regions is not identical. Although many more such observations could be made, I shall now move on to some quantitative features of the urban network of peninsular Italy as a whole. As in the case of Cisalpine Gaul, we can obtain a rough idea of the total number of hectares covered by large and medium-sized towns by adding together the regional totals for each category. The outcome of this exercise is an estimate of c. 3,935 hectares for the combined areas covered by the more substantial towns of the centre and south, with the exception of Rome and Ostia. It is also possible to establish a total figure for the 238 ‘small towns’ of the first seven Augustan regions, although the archaeological evidence for this category is much patchier than that for the larger centres. Taken together, the minor urban centres of central and southern Italy may have covered more than 2,700 hectares, approximately 70 per cent of the aggregate figure for the larger centres. For reasons set out earlier in this chapter, however, the undoubtedly substantial aggregate population of these minor centres has been classified as ‘rural’, regardless of their juridical status. 5.4.1 Some low-count models for central and southern Italy Using the figures calculated in the previous sections, we can construct a series of models to illuminate the range of central/southern Italian urban population densities and urbanization rates that the low count implies. In constructing these models, I have taken into account various indications suggesting that urban population densities in Campania and possibly also in Latium were higher than those in other parts of the peninsula. In practical terms, this means that some of my models will generate regionspecific urban population densities from which the total number of people living in large or medium-sized towns can be derived. Since the size of the rural populations of these areas cannot be determined, however, no attempt has been made to establish differential regional urbanization rates, for the simple reason that it is difficult to think of any sound basis for such a calculation. The only viable tactic is to use the low-count estimate of the

The urban network in 28 bc

239

population of central and southern Italy as our starting point and to work out an average urbanization rate for this area by comparing that figure with the total urban population implied by the regional figures for Latium, Campania and the remainder of central and southern Italy. Our population figure for the central and southern regions can be derived from the low-count model for Italy as a whole. If we assume that earlyAugustan Italy (excluding the areas inhabited by attributi) had a population of 5.7 million and assign 1.6 million free and non-free inhabitants to Cisalpine Gaul and 900,000 inhabitants to Rome and Ostia, we will find that the remainder of the central and southern regions must have had a population of about 3.2 million people, including slaves. As in the case of Cisalpine Gaul, I shall now present eight models capable of accounting for this figure, again differentiating between four scenarios that assume my grand estimate of the number of built-up urban hectares to be 100 per cent accurate and four scenarios that assume that this estimate is 10 per cent too high. Table 5.4 Eight low-count models of urbanization in central and southern Italy Number of urban hectares 1.

3,3951

2. 3.

3,935 3,9902

4.

4,1353

5.

3,0004

6. 7.

3,540 3,5955

8.

3,7406

1

Urban population density 150 p.ha in Campania, 120 p.ha elsewhere 150 p.ha throughout peninsular Italy 180 p.ha in Campania, 150 p.ha elsewhere 180 p.ha in Campania and Latium, 150 p.ha elsewhere 150 p.ha in Campania, 120 p.ha elsewhere 150 p.ha throughout peninsular Italy 180 p.ha in Campania, 150 p.ha elsewhere 180 p.ha in Campania and Latium, 150 p.ha elsewhere

Urbanization rate without Rome 13.8% 18.5% 19.9% 21.1% 12.2% 16.5% 17.9% 19%

Assuming urban population densities of 150 people per hectare in Campania and 120 people per hectare elsewhere, we lose some twenty-four non-Campanian towns covering between 20 and 24.9 ha apiece. Using 22.5 ha as the average area of these towns, we will find that about 540 hectares were covered by agglomerations that no longer had as many as 3,000 inhabitants. 2 With a density of 180 persons per hectare, towns covering 16.67 or more hectares would have had at least 3,000 inhabitants apiece. If we assume that towns covering fewer than 20 ha were spread evenly across the 4.0–19.9 ha range (i.e. that a quarter of them covered 4–8 ha, another quarter 8–12 ha, etc.), three minor Campanian towns would have reached this threshold. If we assume that these towns covered 18.33 ha apiece on average, they add 55 ha to the total number of urban hectares. 3 With 180 inhabitants per urban hectare in Campania and Latium, eleven minor towns covering 18.33 ha each on average (implying a total of c. 200 ha) must be added to the basic figure of 3,920. 4 (0.9 × 3,935) − 540 = 3,002. 5 (0.9 × 3,935) + 55 = 3,597. 6 (0,9 × 3,935) + 200 = 3,742.

240

Augustan census figures and Italy’s urban network

In the absence of reliable evidence, it remains extremely difficult to narrow down this range by arguing in favour of any particular model. As in the case of Cisalpine Gaul, however, any model based on an average population density of only 120 people per hectare results in an urbanization rate falling short of the lowest rate ever suggested. If we use a density of 150 persons per hectare, the urbanization rate implied by the low count falls into the central part of the range between 15 and 20 per cent, an interpretation that has found widespread support. Even this model is not, however, totally convincing. The main reason for this is that any model that simplistically assigns 150 persons per hectare to all of the towns of central and southern Italy does not sit well with the general expectation that the densely built up towns of Campania (and perhaps also many towns in Latium) would have had urban population densities considerably higher than those associated with the low-count model for Cisalpine Gaul. For these reasons, models 3, 4, 7 and 8 can be regarded as more promising than models 1, 2, 5 and 6. In the final analysis, the central and southern Italy of the low count, with its c. 4.1 million inhabitants (if we include Rome), is not radically different from its early-modern counterpart, which had a population of c. 5.5 million. The main difference has to do with the fact that late-republican earlyimperial Rome received large amounts of provincial tax grain.151 It was this factor that made it possible for Rome to become three times larger than early-modern Naples (which obtained most of its basic sustenance from the Terra di Lavoro and from Puglia). Without Rome and Ostia, the central and southern Italy of models 3, 4, 7 and 8 would have been characterized by an urbanization rate of between 17.9 and 21.1 per cent, which is close to the early-modern rate of 22.1 per cent.152 This observation makes it difficult to reject the low count for central and southern Italy on the ground that it results in an urbanization rate that is anomalous from a comparative point of view.153 At the same time, this group of models results in urban population densities matching those recorded in medieval Tuscany. That is certainly another argument for regarding these scenarios as realistic and defensible.

151

152 153

E.g. Erdkamp (2005). I am not, of course, denying that republican and early-imperial Rome also benefited from large amounts of produce grown in Italy. Cf. above, at note 131. As we have seen (note 18), Lo Cascio’s claim that the low count implies an urbanization rate of 39 per cent (Lo Cascio 2009) is based on a completely unrealistic reconstruction of Italy’s urban network.

The urban network in 28 bc

241

5.4.2 Some high-count models for central and southern Italy If we follow the high counters in assigning 6 million inhabitants to Cisalpine Gaul at the time of Augustus, their figure of 15 or 16 million for the whole of Italy implies a population of 9 or 10 million for the central and southern regions. If we credit early-imperial Rome with a population of 1 million, we are left with an estimate of 8 or 9 million for the rest of central and southern Italy. The following four models will accommodate these figures. Table 5.5 Four high-count models of urbanization in central and southern Italy Number of urban hectares 1. 2.

5,2501 5,2752

3.

5,4203

4.

5,9204

1

2

3

4

Urban population density 200 p.ha throughout peninsular Italy 250 p.ha in Campania, 200 p.ha elsewhere 300 p.ha in Latium and Campania, 200 p.ha elsewhere 300 p.ha in Latium and Campania, 250 p.ha elsewhere

Urbanization rate without Rome 11.7–13.1% 12.4–14.0% 14.3–16.1% 17.6–19.8%

With 200 inhabitants per urban hectare throughout peninsular Italy, seventy-five minor towns covering 17.5 ha apiece on average must be added. As in the case of Cisalpine Gaul, I have assumed that densities of 200 or more inhabitants per hectare are unrealistically high for rural vici. With 250 inhabitants/ha in Campania, we must add six Campanian towns covering 16 ha apiece on average. With 200 inhabitants/ha elsewhere, we must add seventy-one other towns covering 17.5 ha apiece on average. With 300 inhabitants/ha in Latium and Campania and 200 p/ha elsewhere, we must add twenty-nine towns covering 15 ha apiece on average and sixty towns covering 17.5 ha apiece on average. With 300 inhabitants/ha in Latium and Campania and 250 p/ha elsewhere, we must add twenty-nine towns covering 15 ha apiece on average and ninety-seven towns covering 16 ha apiece on average.

As in the case of Cisalpine Gaul, it appears that we must associate the high-count theory either with an urbanization rate far below the earlymodern level of 22.1 per cent, or with urban population densities far higher than those calculated for Florence and for many other Italian towns during the High Middle Ages and in early-modern times. It is true that higher densities have been recorded in medieval Siena and also in Genova. As we have seen, however, these densities appear to have been associated with the presence of a large number of tall buildings.154 154

For the irrelevance of the data from Roman Egypt, with its different tradition of domestic architecture, cf. above, at notes 96–8.

242

Augustan census figures and Italy’s urban network

Since very few domestic buildings with three or more storeys have been found outside Rome and Ostia, it remains difficult to argue in favour of the high urban densities required to obtain the high urbanization rates that are a central feature of all existing high-count reconstructions. 5.5 some general conclusions One of the most striking findings of this chapter is perhaps that throughout Roman Italy small towns (