Migration and Development: Sonderausgabe Heft 6/Bd. 229 (2009) Jahrbücher für Nationalökonomie und Statistik 9783110507522, 9783828204973

Citation preview

Migration and Development

Herausgegeben v o n Jürgen M e c k l

M i t Beiträgen v o n Bakewell, Oliver, Oxford, UK Bauer, Thomas K., Essen Czaika, Mathias, Freiburg Docquier, Frédéric, Louvain-La-Neuve, Belgium Egger, Hartmut, Bayreuth Felbermayr, Gabriel, Stuttgart

Lucius &c Lucius • Stuttgart 2 0 0 9

Lindley, Anna, London, UK Rapoport, Hillel, Ramat Gan, Israel Sinning, Mathias G., Canberra, Australia Ziesemer, Thomas H.W., Maastricht, The Netherlands

Anschrift des Herausgebers des Themenheftes Prof. Dr. Jürgen Meckl Fachbereich Wirtschaftswissenschaften Justus-Liebig-Universität Gießen Licher Strasse 66 35394 Glessen E-Mail: [email protected]

Bibliografische Information der Deutschen Nationalbibliothek Die Deutsche Nationalbibliothek verzeichnet diese Publikation in der Deutschen Nationalbibliografie; detaillierte bibliografische Daten sind im Internet über http://dnb.d-nb.de abrufbar ISBN 978-3-8282-0497-3

© Lucius 8c Lucius Verlagsgesellschaft mbH · Stuttgart · 2 0 0 9 Gerokstraße 51, D-79184 Stuttgart Das Werk einschließlich aller seiner Teile ist urheberrechtlich geschützt. Jede Verwertung außerhalb der engen Grenzen des Urheberrechtsgesetzes ist ohne Zustimmung des Verlags unzulässig und strafbar. Das gilt insbesondere für Vervielfältigungen, Übersetzungen und Mikroverfilmungen sowie die Einspeicherung und Verarbeitung in elektronischen Systemen.

Satz: Mitterweger & Partner Kommunikationsgesellschaft mbH, Plankstadt Druck und Bindung: Neumann Druck, Heidelberg Printed in Germany

Jahrbücher f. Nationalökonomie u. Statistik (Lucius & Lucius, Stuttgart 2009) Bd. (Vol.) 229/6

Inhalt / Contents Editorial: Migration and Development

676

Abhandlungen / Original Papers Docquier, Frédéric, Hillel Rapoport, Documenting the Brain Drain of "La Crème de la Crème". Three Case-Studies on International Migration at the Upper Tail of the Education Distribution Egger, Hartmut, Gabriel Felbermayr, Endogenous Skill Formation and the Source Country Effects of Skilled Labor Emigration from Developing Countries Bauer, Thomas K., Mathias G. Sinning, The Purpose of Remittances: Evidence from Germany Ziesemer, Thomas H.W., Worker Remittances and Growth: The Physical and Human Capital Channels Lindley, Anna, Remittances and Conflict: Some Conceptual Considerations Bakewell, Oliver, Migration, Diasporas and Development: Some Critical Perspectives Czaika, Mathias, The Political Economy of Refugee Migration Bandinhalt des 229. Jahrgangs der Zeitschrift für Nationalökonomie und Statistik Contents of Volume 229 of the Journal of Economics and Statistics

679-705 706-729 730-742 743-773 774-786 787-802 803-821

Jahrbücher f. Nationalökonomie u. Statistik (Lucius & Lucius, Stuttgart 2009) Bd. (Vol.) 229/6

Editorial: Migration and Development Impacts on economic development emanating from international, interregional or seasonal migration movements are discussed controversially both in the public and in the scientific community. While we expect overall global gains from increased integration of labor markets, migration produces a series of distributional effects that are not yet fully understood. Especially the topical discussion about selective migration that has been enforced by the industrialized countries' increased application of selective standards generating a bias of immigration policies toward skilled applicants is characterized by a lack of consensus. Additionally, involuntary migration in cases of conflict and/or prosecution induces considerable humanitarian, social and economic challenges. With the emergence of the brain-drain literature in the 1960s, migration has at the latest been recognized as a critical factor for the development of economically backward countries. Whereas early contributions emphasize the negative impact of emigration of highly skilled labor on the economic growth and development of those countries (cf., e.g., Bhagwati/Hamada 1974), more recent contributions (cf. Mountford 1997, Beine et al. 2001) take a different stance and highlight several positive effects of high-skilled emigration, such as (i) brain gain by comprehensive incentives to acquire education in emigration countries, (ii) overall brain gain in less-developed countries by brain circulation in case of re-migration, (iii) financial remittances of successful emigrants to their home countries and (iv) transfer of technology and knowledge to less-developed countries (including institutional change in financial markets and social security systems) especially in the case of temporary migration. As a result, new challenges arise with respect to the design of migration policies both in developed immigration countries and less-developed sending countries in order to support the development process in the poor sending countries. Challenges also arise with respect to the evolution of institutions (financial markets, education, provision of social security) in less-developed countries and their support by specific policy measures. In order to promote research in that topical and wide field, the Justus-Liebig University (Glessen) awards the Developing Countries Prize for outstanding contributions. On occasion of awarding the Developing Countries Prize 2008 (that time sponsored by Kreditanstalt für Wiederaufbau, Frankfurt a.M.) to the laureates Frédéric Docquier and Hillel Rapoport (main price), Anna Lindley (best PhD thesis), and Benjamin Etzold and Cristian Vasco (best student theses), an interdisciplinary symposium on "Migration and Development" has been arranged bringing together economists, political scientists, lawyers and geographers with politicians and members of nongovernmental organizations. The present special issue of the Journal of Economics and Statistics has been stimulated by this symposium and contains several papers that have been presented there. The laureates Frédéric Docquier and Hillel Rapoport contribute a survey-like paper complementing existing empirical cross-country studies on brain drain and brain gain. The authors present general data on the international migration of the very upper tail of the skill and education distribution and then concentrate on prominent cases of high-skill emigration: the African medical brain drain, the exodus of European researchers to the United States, and the contribution of the Indian diaspora to the rise of the IT sector in India. The choice of these case studies is highly persuasive since they illustrate the ambiguity of effects of a brain drain on source countries derived in theoretical studies: they find unambiguously negative effects in the case of European researchers, clearly positive effects in case of Indian diaspora, and mixed effects for the African medical brain drain.

Editorial: Migration and Development · 677

Their findings document the difficulties to capture the effect of skilled emigration on source countries. Based on these results, any derivation of uniform policy recommendations seems highly illusory. The paper of Hartmut Egger and Gabriel Felbermayr provides a valuable theoretical contribution that integrates the traditional and the new approach to skilled emigration. The authors emphasize the role of factor complementarities and associated wage-structure effects for the prospect of source countries gaining from emigration of skilled workers. These factor complementarities already played a central role in the traditional braindrain literature that assumed exogenous skill supplies. The new literature endogenizes skill supplies and integrates externalities from educational decisions that are affected by emigration possibilities, but it typically abstracts from factor complementarities by restricting the analysis on linear technologies. The authors' integration of the old and the new approaches gives rise to ambiguous theoretical effects from increased labor-market integration. Applying a calibrated version of their model, the authors show for four prominent source countries (Mexiko, Turkey, Morocco and Philippines) that a brain-gain scenario would require implausible high externalities. In the end, the paper nicely complements the new literature from the methodological perspective; from the perspective of application, however, it casts considerable doubts on the optimistic view of labor-market integration produced by the brain-gain hypothesis. Three papers address the remittances problems each of them emphasizing different aspects. Thomas Bauer and Mathias Sinning examine the determinants of remittance behavior of immigrants living in Germany empirically. Differentiating between temporary and permanent migration they find that the probability to accumulate wealth in the home country as well as the amount of transfers to family members in the source countries are higher for temporary migrants. Applying a decomposition analysis, they find that the observed differences in remittance patterns are not due to differences in observable characteristics (such as the migrants' economic characteristics, the composition of households in home and host countries) between these two immigrant groups. Rather the various remittance patterns accrue from behavioral differences between the two types of migrants originating in type-specific preferences. Their result thus provides an incentive to better account for individual heterogeneity in analyses of migration. The paper by Thomas Ziesemer investigates the impact of remittances on economic growth of migrants' home countries via two channels: the accumulation of physical and of human capital. By accounting for human-capital accumulation the paper contributes to the existing literature that concentrated on the accumulation of physical capital and found ambiguous effects of remittances on GDP growth. The author estimates a system of seven equations by the general method of moments with heteroscedasticity and autocorrelation correction using a pooled data set of different samples of countries receiving remittances in 2003. He finds that the poorest countries in the sample show the highest effect of remittances on GDP per capita while the effect for richer countries is generally small. Additionally, he finds that savings react more strongly than investment implying that remittances contribute significantly to reduce the debt problem of home countries. The paper by Anna Lindley deals with the question how the inflow of remittances is related to violence and policy instability. Based on a field study in Somaliland, the paper addresses a set of economical (esp. distributional) and political questions regarding the dynamics and impact of remittances in conflict-affected settings. Based on this field study, the author works out that the implications of the violent causation of migration, the ongoing conditions in the migrants' home country and the post-migration situation of refugees are the key determinants of

678 · Editorial: Migration and Development how remittance dynamics in conflict situations differ from those in more peaceful settings. The paper of Oliver Bakewell provides a critical overview of the link between migration, diasporas and development based on results from migration studies from the perspective of political science. He questions why governments and development organizations increasingly stress the role of diasporas in the development process, especially whether diasporas share a common set of interests with general development concept. He then focuses on the nature of development and on the problem of how can diasporas and migration play a role in the development process. The author ends up with a rather critical view on the idea of utilizing diasporas (and migration) as an important cornerstone of a development concept. He emphasizes several shortcomings of the spatial development concepts in the context of migration and formulates the need for new concepts in an increasingly integrated world. The range of problems listed above is rounded out by Mathias Czaika's analysis of refugee migration. Although the migration decision per se is not primarily an economic but a political issue in cases of conflict and prosecution, refugees still have choices that are governed by economic forces: whether they apply for asylum in a Western country, whether they (legally or illegally) move to a neighboring country or whether they stay at home. The theoretical analysis of the paper shows that asylum seeking in Western countries is biased towards comparatively less prosecuted people. The subsequent politicaleconomy analysis of asylum policies argues that host countries are likely to end up in a race to the bottom of respective asylum policies. Pro-active refugee-related aid transfers, however, are shown to have the potential of being an effective economic instrument to reduce asylum pressure of potential host countries. Overall, the contributed papers provide a rather impressionistic picture of the problems stated above with emphasis on a considerable range of topics with many open questions. Particularly the implications for economic development of countries experiencing emigration of skilled people are far from being solved comprehensively and there is a lack of clear general policy recommendations. Although considerable progress has been achieved by the stimulus of the new brain-gain literature, applied research - especially research bridging the gap between economic, political and sociological analyses - in that field is still in its infancy. Hopefully, the papers in this special issue contribute to take us one step ahead in the field of migration and development.

References Beine, M . , F. Docquier, H. Rapoport (2001), Brain Drain and Economic Growth: Theory and Evidence. Journal of Development Economics 64(1): 275-289. Bhagwati, J . N . , K. H a m a d a (1974), The Brain Drain, International Integration of Markets for Professionals and Unemployment: A Theoretical Analysis. Journal of Development Economics 1(1): 19-42. Mountford, A. (1997), Can a Brain Drain be G o o d for Growth in the Source Economy? Journal of Development Economics 53(2): 287-303. Jürgen Meckl

Jahrbücher f. Nationalökonomie u. Statistik (Lucius & Lucius, Stuttgart 2009) Bd. (Vol.) 229/6

Documenting the Brain Drain of "La Crème de la Crème" Three Case-Studies on International Migration at the Upper Tail of the Education Distribution By Frédéric Docquier, Louvain-La-Neuve, Belgium, and Hillel Rapoport, Ramat Gan, Israel* JEL F22, J24, 0 1 5 Brain drain, international migration, African medical brain drain, European brain drain, Indian diaspora.

Summary Most of the recent literature on the effects of the brain drain on source countries consists of theoretical papers and cross-country empirical studies. In this paper we complement the literature through three case studies on very different regional and professional contexts: the African medical brain drain, the exodus of European researchers to the United States, and the contribution of the Indian diaspora to the rise of the IT sector in India. While the three case studies concern the very upper tail of the skill and education distribution, their effects of source countries are contrasted: clearly negative in the case of the exodus of European researchers, clearly positive in the case of the Indian diaspora's contribution to putting India on the IT global map, and mixed in the case of the medical brain drain out of Africa.

1

Introduction

Most of the recent literature on the effects of the brain drain on source countries consists of theoretical papers (e.g., Mountford 1997, Vidal 1998, Beine et al. 2001) and of crosscountry empirical studies on the "brain gain" (Beine et al. 2 0 0 8 ) and the diaspora networks channels (Kerr 2 0 0 8 , Agrawal et al. 2 0 0 8 , Kugler/Rapoport 2 0 0 7 , Docquier/Lodigiani 2009). The main novelty of the recent literature is to show that under certain circumstances, the brain drain may ultimately prove beneficial (but of course is not necessarily so) to the source country, and to do so while at the same time accounting for the various fiscal, technological and Lucas-type externalities that were at the heart of the pessimistic models of the 1970s. Another novelty is that it is evidence-based, something which was out of reach until not long ago due to the lack of decent comparative data on international migration by educational attainment. 1 By nature, theoretical models and cross-country comparisons cannot account for the intricacies and details which are context specific. They have also abstracted (so far) from * This paper is part of a broader survey project on the brain drain under preparation for the Journal of Economic Literature. Docquier acknowledges financial support from ARC convention on "Geographical mobility of workers and firms" (convention 0 9 / 1 4 - 0 1 9 ) and from the Marie-Curie research and training network T O M (Transnationality of migrants). Rapoport acknowledges support from the Adar Foundation at Bar-Ilan University. 1 See Docquier and Rapoport (2009) for a broad survey of this literature.

680 · F. Docquier and H. Rapoport accounting for the huge heterogeneity among skilled workers, aggregating flows of workers with intermediate skills (e.g., less than 4 years of college education) and high skills (e.g., PhD holders). In this paper we complement the recent literature in that we focus on "the cream of the cream", that is, the upper tail of the skill and education distribution. We first present general data on the international migration of very highly educated individuals, and then investigate in more details three very different regional and professional contexts: the African medical brain drain, the exodus of European researchers (mainly to the United States), and the contribution of the Indian diaspora to the rise of the I T sector in India.

2

Data: the brain drain at the upper tail of the education distribution

2.1 General figures International migration of highly-skilled professionals (or brain drain) has increased tremendously over the last few decades, at about the same pace as trade, and has recently increased even more rapidly (by 7 0 percent during the 1990s only). 2 By 2 0 0 0 , there were sixty million highly-skilled (tertiary educated) immigrants in the O E C D area, or about one third of total immigration. These highly skilled immigrants represent a tiny three percent of the European skilled workforce against more than ten percent of the skilled labor force in countries such as the United States, Canada and Australia. Given that the vast majority of these immigrants come from developing countries where human capital is very scarce, it often represents a significant loss of human capital for source countries. And indeed, some developing countries exhibit brain drain rates frequently higher than fifty percent (which is typically the case for Sub-Saharan African countries) or even eighty percent (in countries such as Jamaica and Guyana) (Docquier et al. 2 0 0 9 ) . 3 However, general emigration rates may hide heterogeneity across sectors and occupations. If emigration is concentrated in certain fields and the domestic supply of these skills is inelastic, then emigration can induce occupational shortages that may be particularly harmful for economic development. In this paper, we focus of the upper tail of the skill and education distribution: PhD holders, researchers in Science and Technology, medical doctors, information technology specialists. These professions are crucial for the R & D sector and for technological innovation (in the case of already advanced countries) and adoption (which is more relevant for developing countries), not be mention the fact that health care is a complement to human capital, implying that the quantity and quality of the medical staff strongly conditions the productivity of all other professions (Kremer 1993). Before turning to our three case studies, we first present more focused data on PhD holders and researchers in science and technology, on the one hand, and on the medical brain drain, on the other hand.

2.2 PhD holders and researchers in science and technology Table 1 focuses on the emigration of PhD graduates. For 82 origin countries, we provide (i) the numbers of PhD graduates working in the US, (ii) the shares of these PhDs among US post-secondary educated immigrants by country of origin, (iii) the ratio of PhD hol2

3

The total number of highly educated immigrants living in the O E C D member countries has increased by 7 0 percent during the 1 9 9 0 s (and has doubled for those originating from developing countries) against just a 3 0 percent increase for unskilled immigrants. See Figure 1.

Documenting the Brain Drain of "La Crème de la Crème" • 681

682 · F. Docquier and H. Rapoport Table 1 Top-30 suppliers PhD's to the US PhD graduates in the US China United Kingdom Canada Germany Russia South Korea Iran France Poland Japan Mexico Nigeria Egypt Israel Argentina Romania Italy Brazil Turkey Colombia Cameroon Ukraine Philippines Spain Ireland Cuba Greece Ghana Hungary Australia

63153 24482 19122 17840 12835 12172 8996 7277 6488 6478 5693 4862 4725 4694 4405 4122 3997 3952 3798 3787 3714 3701 3658 3435 3294 3246 2948 2909 2877 2477

Share in graduates in the US Slovenia Cameroon Georgia Tunisia Saudi Arabia Iceland China Estonia Uzbekistan Azerbaijan Switzerland Croatia Finland Czech Republic Slovakia Austria Israel Hungary Ghana Romania Turkey Russia Ethiopia Spain Argentina Armenia France Brazil United Kingdom Sweden

71,4% 51,7% 46,1 % 31,8% 26,8 % 21,5% 21,3 % 19,6% 19,6% 19,6% 18,1 % 18,1 % 17,8% 17,6% 17,6% 17,4% 16,5% 16,3% 15,9% 15,8% 15,4% 15,2% 12,5% 12,0% 12,0% 11,9 % 11,6% 11,4 % 11,3 % 11,2 %

Estimated mig. rate to the US Panama Ethiopia Colombia Honduras Iceland Uruguay Tanzania Cyprus Macao Trinidad and Tobago Argentina Cuba Cameroon China Cambodia Bangladesh Ghana Ireland Israel Canada Iran Croatia Jordan Mexico Armenia Hungary Bulgaria Estonia Lebanon Philippines

93,2 % 91,3 % 84,4 % 78,5 % 72,9 % 71,8% 65,8 % 49,2 % 49,1 % 47,2 % 37,0 % 30,7 % 23,7 % 22,8 % 22,7 % 21,7% 16,6% 16,0% 15,9% 15,7% 15,1 % 14,4% 14,4% 13,4% 12,8% 12,5% 11,7% 11,2 % 10,7% 10,2 %

Sources: SESTAT-NSF and UNESCO

ders living in the US to the estimated number of PhD holders trained in their country (an estimate of the emigration rate to the US of PhD holders by country of origin). To compute (i) and (ii), we use the SESTAT database of the National Science Foundation. To calculate (iii), we use UNESCO data on the flow of PhD graduates trained at origin (average 2002-2004) and assume that the flows of new PhD graduates represent 5 percent of the stock in developing countries and 4 in developed countries. The estimated emigration rate is obtained by dividing the stock living in the US by the estimated stock domestically trained. The highest numbers of foreign PhD holders are obtained for developed countries and large developing countries such as China, Russia, Iran, Nigeria, Egypt. As a proportion of tertiary graduates living in the US, the proportion of PhD is extremely high in the cases of Slovenia, Cameroon, Georgia and Tunisia. The last columns indicates that the estimated emigration rate of PhD holders is high for Latin American countries and some African countries. Regarding the capacity to innovate, it is also interesting to focus on researchers employed in S&T. This includes many PhD holders but also many other college graduates employed

Documenting t h e Brain Drain of "La Crème de la Crème" · 6 8 3

Table 2 Researchers employed in Science and Technology in t h e US in 2 0 0 3 Developing countries Birth

High-income countries

S&T reS&T researchers searchers in the US at home

Algeria 1242 Bolivia 2214 Brazil 10980 4497 Bulgaria Myanmar 1727 Cambodia 3030 Cameroon 3643 Chile 5496 China 158524 19362 Colombia Costa Rica 4659 Cote divoire 288 Croatia 1666 Ecuador 7012 2549 Ethiopia Guatemala 1415 Indonesia 5163 Kazakhstan 1108 Latvia 2728 Lithuania 2285 Macedonia 80 Madagascar 166 Malaysia 7955 Malta 452 Mexico 46356 Nepal 1739 Pakistan 14682 Panama 7498 Paraguay 335 Romania 10900 Russia 35588 South Africa 5906 Sri Lanka 4652 Thailand 7781 Tunisia 2003 Turkey 8878 Uruguay 1625 Venezuela 8058 Vietnam 44236

5678 1140 79600 9400 732 239 472 10120 907743 4487 529 1292 6722 595 1649 398 45567 10339 3291 7105 1147 887 10419 359 42953 1627 12919 307 489 20761 478090 16248 2703 18430 11805 31587 1244 3537 9863

Average

Brain Birth drain to US in %

S&T reS&T researchers searchers in the US at home

Brain drain to US in %

17.9 66.0 12.1 32.4 70.2 92.7 88.5 35.2 14.9 81.2 89.8 18.2 19.9 92.2 60.7 78.1 10.2 9.7 45.3 24.3 6.5 15.8 43.3 55.7 51.9 51.7 53.2 96.1 40.6 34.4 6.9 26.7 63.3 29.7 14.5 21.9 56.6 69.5 81.8

Australia Austria Belgium Canada Hong Kong Cyprus Czech Republic Denmark Estonia Finland France Germany Greece Hungary Iceland Ireland Italy Japan Kuwait Luxembourg Netherlands New Zealand Norway Portugal Singapore Slovakia Slovenia South Korea Sweden Switzerland United Kingdom

4889 3815 4767 72584 26602 591 2455 2561 813 791 16072 59213 6554 4986 1002 9270 15022 34757 1118 100 7616 3217 3291 2581 3397 1227 202 50605 3585 3768 72396

5.8 12.6 12.9 37.1 68.2 52.6 12.5 9,3 21.0 1.9 7.6 18.0 28.4 24.9 33.0 46.3 17.0 4.9 84.7 4.5 15.6 16.8 13.4 11.4 13.5 10.9 4.3 24.6 6.7 12.8 29.0

45.6

Average

79919 26563 32229 122809 12410 532 17232 25035 3063 39897 195638 269703 16546 15001 2034 10741 73181 677723 202 2108 41082 15911 21339 20067 21821 10008 4455 154884 50091 25616 177625

21.4

Sources: SESTAT-NSF and UNSECO in this sector. Table 2 c o m p a r e s migration of researchers e m p l o y e d in the US R & D sector (using the SESTAT database) t o U N E S C O data o n researchers nationally e m p l o y e d in S&T. We will provide researchers' emigration numbers and rates to the US for 7 0 countries, including 3 9 developing states. T h e average emigration rates of d e v e l o p i n g coun-

6 8 4 · F. Docquier and H. Rapoport

tries (45.6 percent) exceeds that of developed countries (21.4 percent). The rate is particularly high (above 80 percent) in the cases of Cambodia, Cameroon, Colombia, Costa Rica, Ecuador, Panama or Vietnam. 2.3 The medical brain drain

In developing countries, the size and quality of the medical sector is a key determinant of human development and economic performances (see Bhargava et al. 2001, Hagopian et al. 2004, Cooper 2004, Bhargava/Docquier 2008). While the number of physicians per 1,000 people is greater than 3 in most industrialized countries, it is lower than 0.25 in many developing countries (see Figure 2a). Many observers and analysts have pointed to the medical brain drain as one of the major factors leading to the under-provision of healthcare staff in developing countries (see Bundred/Levitt 2000, or Beeckam 2002) and, ultimately, to low health status and shorter life expectancy - hence Michael Clemens's (2007) provocative question: do visas kill? Two data sets can be used to document the international migration of physicians: • Clemens and Pettersson (2006) collect data on foreign physicians and nurses from nine important destination countries (UK, US, France, Australia, Canada, Portugal, Belgium, Spain and South Africa) and compute the stock of African-born physicians living abroad by country of birth in 2000. They then evaluate the medical brain drain in relative terms, dividing the number of physicians abroad by the total number of physicians born in each origin country. • Docquier and Bhargava (2006) use the same methodology but collect data from 17 countries (16 OECD countries and South Africa) and define migrants according to their country of training. Such data can be obtained from national medical associations and are available on an annual basis. They come up with 14 yearly observations per country covering all the countries of the world for the period 1991-2004. Regional comparisons reveal that the medical brain drain is highest in Sub-Saharan Africa (with average rates above 20 % against 13 % in South-Asia and less than 10 % in all the other regions); the figures are relatively stable over the period. Focusing on the year 2000, the comparison of these two data sets reveals important differences, with a correlation between the two of only .23. The "bilateral" correlations between physician immigrants stocks in the eight common destination countries are much higher (from 55 percent for South Africa to 97 percent for France and the United States). However, the stock based on country of training is usually much lower than the stock based on country of birth (e.g., 10 % in France, 4 45 % in South Africa, 77 % in the United Kingdom, and 8 2 % in the United States). Figure 2b shows the geographical distribution of the medical brain drain computed in Docquier and Bhargava (2006). The average medical brain drain is particularly severe in Sub-Saharan Africa, South Asia, East Asia and Latin America. The most affected countries exhibiting emigration rates above 40 percent are Grenada, Dominica, Saint Lucia, Ireland, Liberia, Jamaica and Fiji. Using the same dataset, Figure 2c reveals that the medical brain drain rates have increased dramatically in many African countries but also in Lebanon, Cuba, Cyprus, or the Philippines. 4

Licensure requirements for foreign physicians are more stringent in France t h a n in most other host countries.

Documenting the Brain Drain of "La Crème de la Crème" • 685

3

Africa's medical brain drain

As explained above, Clemens and Peterson (2006) and Docquier and Bhargava (2006) use different definitions of the medical brain drain, by country of birth (for the former) and by country of training (for the latter). This leads to important differences in their respective estimates of the medical brain drain, as we have seen. Interestingly, the main culprit for such differences is Africa. Indeed, due to absence of local medical schools, eleven African countries have no domestically trained physician emigrants living abroad while they exhibit medical brain drain rates between 5 to 15 percent if one uses the country-of-birth criterion. Figures 3a and 3b illustrate the difference between these two definitions of physicians' brain drain in the case of Africa. 3.1 Determinants of the medical brain drain As for general migration, it is obvious that the emigration of physicians is not an exogenous process. Individual-level surveys in six African countries indicate that more than half of all physicians would like to emigrate to developed countries, in search of better working conditions and more comfortable lifestyles (Awases et al. 2003). The risks associated with caring for HIV/AIDS patients and the possibility of children of healthcare staff contracting HIV as they enter adolescence may exacerbate the medical brain drain (Awases et al. 2003; Bhargava 2005). Using their data set by country of training, Bhargava and Docquier (2008) estimated the determinants of the African medical brain drain. Consistently with Awases et al. (2003), countries with higher physician wages have lower emigration rates. Net enrolment in secondary education is also a positive and significant predictor of the medical brain drain, with an estimated short-run elasticity of 0.12. This result is not surprising, as higher enrolments in secondary education entail greater expenditures on education; physicians educated in such environments are likely to have better emigration prospects. More importantly, the HIV prevalence rate is a significant predictor of the medical brain drain, with a short-run elasticity of 0.07 and a long-run elasticity of 0.80; this means that a doubling of the HIV prevalence rate implies an 80 percent increase in the medical brain drain rate in the long run. This is a large effect, with important policy implications. Using the same data set, Moullan (2008) recently analyzed the effect of bilateral health assistance on the bilateral medical brain drain. The rationale is that, by increasing health capital and infrastructure, health assistance can improve the working conditions of health professionals. His cross-section and panel analyses show that health assistance is an effective tool to retain doctors at home. However, elasticities are relatively low, suggesting that a huge amount of health assistance would be required to reduce the medical brain drain. Interestingly, total bilateral aid (health + non-health) seems to stimulate the medical brain drain under most specifications. 3.2 The case for a medical brain gain In the spirit of the recent literature on endogenous human capital in a context of migration, we may ask whether there is a chance for a net medical brain gain. Regressing the log of domestic physicians per capita on the log of physician emigrants per capita, Clemens (2007) found a positive correlation of about 70 percent. Clearly, this correlation can be driven by the simultaneous effects of observed variables (GDP per capita, school enrolment conflicts, etc.) or unobserved variables. However, after controlling for obser-

686 · F. Docquier and H. Rapoport

Documenting the Brain Drain of "La Crème de la Crème" · 687

o o

(N O O

morosmO ι Ο ιλ co κ ι I τ- Q (• II m Φ 'f ΙΟ \(Ν 0Ο

ο ο.

• ino Λ Γ ( • in to iti ^ ι

υ bo te -C m Jt -O (β

a Ξ

s

S Ρ £ 2j O z î o ïΛ s

#5115 ? i g

3E_ ¿ Z Q£ Q ¡«"OjS 5 c S f i l i l i ;

Documenting the Brain Drain of "La Crème de la Crème" • 695

σ> : 1

η rv rs

« in

$

S

^

R§ S ™ 5S$ Κ τ-

5

o

$

σ\

ι 00

VO

- ?

ΙΓ

rv

ι r--

\ο

g8

• «t oo

q>

)

σι

in

S IS 8 íS

εe s 8! e ¡ó

'

s 3«

5$ S as S Sg • (fi «-

Î = Lo«of number of researchers pei 1,000 jobsiii 1995 (busines sector)

Number of researchers per 1,000 jobs (EU, Japan and US)

Figure 4 Numbers and growth rate of researchers per 1,000 jobs ( 1 9 9 5 - 2 0 0 5 )

2,0

Documenting the Brain Drain of "La Crème de la Crème" · 699 Table 5 Number of H1B visas delivered to Indian immigrants, 1998-2008 11 Year

India as country of Citizenship Residence

Total

Percentage of total Citizenship Residence

2008 2007 2006 2005 2004 2003 2002 2001 2000 1999 1998

154726 157613 125717 102382 83 536 75964 81 091 104543 102453 85012 62544

409619 461730 431853 407418 387147 360498 370490 384191 355605 302326 240947

37.77 34.14 29.11 25.13 21.58 21.07 21.89 27.21 28.81 28.12 25.96

78913 81 584 67292 55873

19.26 17.67 15.58 13.71

Source: DHS Yearbook of Immigration Statistics for years 1998-2008, http://www.dhs.gov/ximgtn/statistics/ publications/yearbook.shtm

aimed at skilled professionals in sectors with occupational shortages (in practice, software engineers and programmers). Given its high ranking and standing as an exporter of skilled professionals and talented individuals, India has been the subject of a large amount of brain drain oriented research. The presence of highly educated Indians among the business, scientific and academic elites of England, the US, and other Western countries, is impressive and has long been both a matter of national pride and of persistent concern. Echoing this ambivalence, Desai et al. (2009) evaluated the fiscal cost of the brain drain for India at 0.5 percent of Indian GDP or 2.5 percent of total Indian fiscal revenues, a "conservative" estimate in their view. However, their computations are based on the assumption that all Indian engineers abroad would have worked as engineers in India, and would have engaged in engineering studies in the first place, which is disputable. While it is clear that many of them would not have worked as engineers if it was not for the possibility of migration, the nomigration counterfactual is not clear. If one assumes that in alternative occupations their wages would have been lower, then Desai et al. (2009) fiscal loss estimates could instead be seen as an upper bound. In fact, many of them end up in managerial jobs (for example, 52 percent of the graduates of IIT-Bombay of 2005-6 ended up in consulting and finance), which are much higher paying occupations in India than engineering, and accounting for this would indeed push the fiscal loss estimates upwards. Perhaps more importantly, if the loss is not that of engineers per se but a selection bias in which entrepreneurial talent is lost, 1 2 then the tax losses are on corporate and VAT/sales taxes and not income taxes on which Desai et al. (2009) focused on. In any event, the last years have seen a gradual reversal in media and public attitudes in India, and it is now common to celebrate the contribution of the Indian diaspora to the country's industrial and economic success. India has already been frequently cited in the recent literature to exemplify the potential for a diaspora to foster technology and knowledge diffusion (Kerr 2008, Agrawal et al. 2008) or the contribution of return mig-

11 12

Courtesy of Devesh Kapur As evidenced for example by Saxenian (2006).

700 · F. Docquier and H. Rapoport

ration to the home economy (Agrawal et al. 2008, Saxeenian 2006). In what follows we will focus on the role of the Indian diaspora, especially that established in the Silicon Valley, in the rise of the Indian IT sector in India. We will base our account mainly on the works of Saxenian (1999, 2002), Arora and Gambardella (2005), Kapur and McHale (2005), Commander et al. (2008), and Kapur (2009, Chapter4). The first study to point to the potential role of the Indian diaspora in the rise of the software industry in India is the well known work of Saxenian (1999), who noted the large implication of Indian (and Chinese) entrepreneurs in the Silicon Valley: Indians were shown to run 9 percent of Silicon Valley start-ups from the period 1995 to 1998, with a majority of these start-ups (nearly 70 percent) in the software sector. A more recent survey ( Wadhwa et al. 2007) shows the last decade has been even more impressive in terms of Indian-born entrepreneurs' share in the US high-tech sector: it shows that out of an estimated 7,300 U.S. tech start-ups founded by immigrants between 1995 and 2005, 26 percent have Indian founders, CEOs, presidents or head researchers - more people than from the four next biggest sources (United Kingdom, China, Taiwan and Japan) combined. Indian immigrants outpaced their Chinese counterparts as founders of engineering and technology companies in Silicon Valley, with Indians being key founders of 15.5 percent of all Silicon Valley startups, mainly in the fields of software (for 46 percent of them) and innovation/manufacturing-related services (44 percent). Saxenian (2002) then proceeded to explore not just the potential but the actual business links with India. In her survey of Indian (and Chinese) members of professional associations in the Silicon Valley, she shows that these links are indeed important: for instance, 77 percent of the respondents had one or more friends who returned to India to start a company, 52 percent used to travel to India for business purposes on a regular basis (at least once a year), 2 7 percent reported regularly exchanging information on jobs/business opportunities with those back home, 33 percent reported regular exchanges of information on technology. In addition, 46 percent had been a contact for domestic Indian businesses, and 23 percent claimed to have invested their own money into Indian startups. Last but not least, when asked about the possibility of return migration, 45 percent reported returning as somewhat or quite likely. Such results must be taken with caution as they are based on a non-representative sample (due to self-selection into the professional associations surveyed and to the choice to respond to the survey). 13 As Kapur and McHale (2005) note, "these figures contradict what is known about the activities of Indian diaspora from other sources, so that the survey's results need to be treated with some caution. One problem is that the investment data is silent on the magnitude of investments. Foreign direct investment from the Indian diaspora is less than 5 percent of its Chinese counterparts - even though the propensity to invest is comparable for the two diasporas in Saxenian's survey. Similarly, the finding that 45 % would consider returning is belied by reality. While aggregate data on return migration is unavailable, segment specific data such as NSF longitudinal data on PhD students suggests a number closer to 5 percent." Still, Saxenian's results are suggestive of strong connections between the Silicon Valley resident Indians and those in India. And indeed, the role of the Indian diaspora has been singled out as a primary factor of India's emergence on the global IT scene. As Kapur (2002) put it, "One of the puzzles about the explosive growth of India's IT sector is how

13

The overall response rate was 21 percent.

Documenting the Brain Drain of "La Crème de la Crème" · 701

and why India has emerged as a global leader in a leading edge industry when, despite strenuous (and, in retrospect, misguided) policies, it failed to achieve such leadership in any other technology intensive sector. The issue is even more puzzling if one keeps in mind that conventional indicators of IT penetration, such as personal computers (PCs) per thousand population, internet subscribers, telephone connections, scientists and engineers per million, all make India look decidedly mediocre". To solve this puzzle, Kapur (2002,2009) first reviewed what he presents as proximate causes of the Indian IT sector success, namely, the lack of State intervention and the flexibility of the labor market in the IT sector, and then turns to what he sees as the root causes. Chief among them is... the brain drain, whose beneficial effects, he argues, have been multifaceted. Paraphrasing Kapur's account and linking his analysis to the general arguments put forward in the recent literature on the effects of emigration on home countries, the following channels may be emphasized: • A first windfall from India's brain drain is that it has provided prospective investors with information on the quality of the Indian labor force and created virtuous reputational spillovers, sparking demands for Indian IT specialists in countries without previous Indian migration experience (e.g., Germany, Japan) as well as international demand for IT services exported from India. 14 This is very much in line with the general argument about an information and transaction cost channel, especially with the argument that migrant workers, skilled or unskilled, can convey information and reduce transaction costs through their sheer presence in the host countries labor markets. Evidence of such information and transaction cost effects contributing to foster FDI from host to home countries can be found in studies using bileral (Kugler/Rapoport 2007, Javorcik et al. 2006, Buch et al. 2006) as well as aggregate (Docquier/Lodigiani 2009) data. • Second, the overseas Indian presence has helped in the diffusion of knowledge through a variety of mechanisms: substantial skill upgrades for those who worked in the US, with diffusion to India through return migration and brain circulation. This is a a perfect illustration of another channel put forward in the recent brain drain literature, namely, the knowledge and technology diffusion channel, as well as additional evidence of the brain circulation (or return migration with additional skills and human capital). As such, this confirms recent studies using patent citation data to measure the international diffusion of knowledge and innovation through diaspora networks (Kerr 2008, Agrawal et al. 2008). • Third, the diaspora has been an effective partner in setting up sectoral institutions and networks who successfully lobbied the Indian government to change the regulatory framework for venture capital in India. While this example is restricted to a particular sector, it is not difficult to imagine that once such lobbying organizations are in place, with their set-up costs already incurred, they can also be activated towards achieving broader political and institutional reforms. This exemplifies the type of institutional reform towards better regulations and more effective economic and political institutions emphasize in the recent brain drain literature in the effects of skilled emigration and foreign students on home countries institutions and governance (Li/McHale 2006, Docquier et al. 2009, Spilimbergo 2009).

14

This echoes Banerjee and Duflo's (2000) evidence that reputation affects the form of contracts that firms outsourcing customized software enter into with Indian software firms.

702 · F. Docquier and H. Rapoport

• And fourth, instead of developing a protectionist attitude by trying to keep engineers and IT specialists at home, the Indian industry has realized the potential gains from foreign experience and supported an increase in the number of H l - B visas for Indian professionals in the US. The reason lies in changes in the market structure of the global IT industry, itself a lagged effect of previous brain drain. Ten of the largest twenty-five companies hiring foreign nationals with Η-IB visas are IT firms based in India or U.S.based IT firms run by Indian nationals. This may clearly be interpreted along the lines suggested in our introduction about the endogenous human capital formation in a context of migration, often referred to as the brain drain v. brain gain debate, and further adds to the recent evidence (e.g., Beine et al. 2008) on endogenous human capital formation and return migration as potential mechanisms possibility leading to a beneficial brain drain (or net brain gain). Kapur's account demonstrates the crucial role played by the Indian diaspora at the onset of the IT revolution which took place in the 1990s as well as in the later phases and goes beyond the general effects on knowledge diffusion and technology diffusion emphasized for example in the papers by Kerr (2008) and Agrawal et al. (2008). This assessment is confirmed by other surveys and analyses. For example, a recent comprehensive survey of India's software industry showed that 30 to 40 % of the higher-level employees have relevant work experience in a developed country (Commander et al., 2008). Similarly, Nanda and Khanna (2009) used a survey sent to all the CEOs of Indian software firms to study the role of diaspora links and found that entrepreneurs who live in hubs, where the local networking environment is stronger, rely on local networks and do not necessarily gain significantly from diaspora networks. More specifically, for those entrepreneurs based in smaller cities with weaker networking and financing environments, having a personal experience abroad allows for gaining access to business and financial opportunities through diaspora networks. They conclude that brain circulation is crucial as such networks, it is argued, are successful not just because of the expatriates who live abroad, but because some of the expatriates have returned back home and know how to effectively tap into the diaspora. 6

Conclusion

Most of the recent literature on the effects of the brain drain on source countries consists of theoretical papers and cross-country empirical studies. In this paper we complement the literature through three case studies on very different regional and professional contexts: the African medical brain drain, the exodus of European researchers to the United States, and the contribution of the Indian diaspora to the rise of the IT sector in India. While the three case studies concern the very upper tail of the skill and education distribution, their effects of source countries are contrasted: clearly negative in the case of the exodus of European researchers, clearly positive in the case of the Indian diaspora's contribution to putting India on the IT global map, and mixed in the case of the medical brain drain out of Africa. These contrasted experiences also illustrate how difficult it is to capture the effect of skilled emigration on source countries using uniform approaches leading to uniform policy recommendations. The recent brain drain literature shows that the brain drain has a potentially strong incidence on between-country inequality. In other words, there are winners and losers, and the brain drain may in some cases contribute to speed up the pace of convergence for some countries while contributing to increased divergence in

Documenting the Brain Drain of "La Crème de la Crème" · 703

t h e c a s e o f o t h e r c o u n t r i e s . 1 5 T h e c a s e studies p r e s e n t e d in this p a p e r c o m p l e m e n t a n d s t r e n g t h e n this v i e w in t h a t t h e y s h o w similar p a t t e r n s f o r r e g i o n s a n d / o r p r o f e s s i o n s . A s t r a i g h t f o r w a r d i m p l i c a t i o n o f the a b o v e analysis is t h a t c u r b i n g skilled e m i g r a t i o n m a y be a s o u n d p o l i c y o b j e c t i v e in t h e c a s e o f E u r o p e ( a s s u m i n g it d o e s so by b e c o m i n g m o r e " t a l e n t f r i e n d l y " ) b u t w o u l d clearly be c o u n t e r p r o d u c t i v e in t h e c a s e o f India. R e g a r d i n g specific p r o f e s s i o n s , the m a i n p o l i c y discussions so far h a v e f o c u s e d o n p r o p o s a l s t o c r e a t e b l a c k l i s t s o f high-risk o c c u p a t i o n s a n d / o r origin c o u n t r i e s (e.g., p h y s i c i a n s a n d nurses o r i g i n a t i n g f r o m high m e d i c a l b r a i n drain c o u n t r i e s w i t h less t h a n 0 . 5 h e a l t h c a r e p r o f e s s i o n a l s per 1 , 0 0 0 p e o p l e ) . O u r analysis s h o w s t h a t such p r o p o s a l s , w h i c h p r i m a rily t a r g e t the A f r i c a n m e d i c a l b r a i n drain (see e.g., B e e c h a m 2 0 0 2 ) , s h o u l d a l s o be reev a l u a t e d in t h e light o f t h e c o m p l e x r e l a t i o n s h i p s b e t w e e n t h e m e d i c a l b r a i n d r a i n , the e n d o g e n o u s f o r m a t i o n o f m e d i c a l h u m a n c a p i t a l , a n d the h e a l t h i n f r a s t r u c t u r e a n d gen e r a l e n v i r o n m e n t in A f r i c a .

References Aghion, Ph., D. Cohen ( 2 0 0 4 ) , Education et croissance. Paris: La Documentation Française. Rapports du Conseil d'Analyse Economique. Agrawal, Α., D. Kapur, J . M c H a l e ( 2 0 0 8 ) , Brain Drain or Brain Bank: The Impact of Skilled Emigration on Poor-Country Innovation. N B E R Working Paper N o 1 4 5 9 2 , December. Arora, Α., A. Gambardella (eds.) ( 2 0 0 5 ) , From Underdogs to Tigers: T h e Rise and Growth of the Software Industry in Some Emerging Economies. Oxford: Oxford University Press. Awases, M . , A. Gbary, J . Nyoni, R. Chatora ( 2 0 0 3 ) , Migration of health professionals in six countries: A synthesis report. World Health Organization, Regional Office for Africa. Banerjee, A.V., E. Duflo ( 2 0 0 0 ) , Reputation effects and the limits of contracting: A study of the Indian software industry. Quarterly Journal of Economics 1 1 5 : 9 8 9 - 1 0 1 7 . Beecham, L. ( 2 0 0 2 ) , UK government should stop recruiting doctors from abroad. British Medical Journal 3 2 5 . Beine, M . , F. Docquier, H. Rapoport ( 2 0 0 1 ) , Brain drain and economic growth: theory and evidence. Journal of Development Economics 6 4 : 2 7 5 - 8 9 . Beine, M . , F. Docquier, H . Rapoport ( 2 0 0 7 ) , Measuring international skilled migration: new estimates controlling for age of entry. World Bank Economic Review 2 1 : 2 4 9 - 2 5 4 . Beine, M . , F. Docquier, H. Rapoport ( 2 0 0 8 ) , Brain drain and human capital formation in developing countries: winners and losers. Economic Journal 1 1 8 : 6 3 1 - 6 5 2 . Bhargava, A. ( 2 0 0 5 ) , AIDS epidemic and health care infrastructure inadequacies in Africa: A socioeconomic perspective. Journal of AIDS 4 0 : 2 4 1 - 2 4 2 . Bhargava, Α., D. Jamison, L. Lau, C. Murray ( 2 0 0 1 ) , Modeling the effects of health on economic growth. Journal of Health Economics 2 0 : 423—440. Bhargava, Α., F. Docquier ( 2 0 0 8 ) , H I V Pandemic, Medical Brain Drain, and Economic Development in Sub-Saharan Africa. World Bank Economic Review 2 2 : 3 4 5 - 6 6 . Buch, C . M . , J . Kleinen, F. Toubal ( 2 0 0 6 ) , Where Enterprises Lead, People Follow? Links between Migration and FDI in Germany. European Economic Review 5 0 : 2 0 1 7 - 3 6 . Bundred, P., C. Levitt ( 2 0 0 0 ) , Medical migration: who are the real losers? The Lancet 3 5 6 , 9 2 2 5 : 245-46. Chauvet, L., F. Gubert, S. Mesplé-Somps ( 2 0 0 8 ) , Are remittances more effective than aid to improve child health? An empirical assessment using inter and intra-country data, paper presented at the Annual Bank Conference on Development Economics, Cape Town, South Africa.

15

See, e.g., Beine et al. (2008), Mountford and Rapoport (2007), and, for a survey, Docquier and Rapoport (2009).

704 · F. Docquier and H. Rapoport

Clemens, M . (2007), D o Visas Kill? Health Effects of African Health Professional Emigration. Working Paper 114, Center for Global Development. Clemens, M.A., G. Pettersson (2006), A N e w Database of Health Professional Emigration from Africa. Working Paper, 95, Center for Global Development. Commander, S., R. Chanda, M . Kangasmieni, A.L. Winters (2008), The Consequences of Globalisation: India's Software Industry and Cross-Border Labour Mobility. World Economy 31: 187-211. Cooper R.A. (2004), Weighing the evidence for expanding physician supply. Annals of Internal Medicine 141: 7 0 5 - 1 4 . Coulombe, S., J.-F. Tremblay (2009), Migration and skills disparities across the Canadian provinces. Regional Studies 43: 5 - 1 8 . Defoort, C. (2009), To educate or not to educate: the impact of migration perspectives in the medical sector. Working Paper, EQUIPPE, University of Lille. Desai, M . , D. Kapur, J. M c H a l e , K. Rogers (2009): The fiscal impact of high-skilled emigration: Flows of Indians to the U.S. Journal of Development Economics 88: 3 2 ^ 4 . Docquier, F., A. Bhargava (2006), Medical brain drain - A N e w Panel Data Set on Physicians' Emigration Rates ( 1 9 9 1 - 2 0 0 4 ) . Research Report, The World Bank: Washington D C . Docquier, F., E. Lodigiani (2009), Skilled migration and business networks. Open Economies Review, forthcoming. Docquier, F., H . R a p o p o r t (2009), Skilled migration: the perspective of developing countries. Pp. 2 4 7 - 2 8 4 in: J. Bhagwati, G. H a n s o n (eds.), Skilled immigration: problems, prospects and policies. O x f o r d University Press. Docquier, F., B.L. Lowell, A. M a r f o u k (2009), A gendered assessment of the brain drain. Population and Development Review 35: 2 9 7 - 3 2 1 . Docquier, F., E. Lodigiani, H . R a p o p o r t , M . Schiff (2009), Migration and home-country institutions and governance. M i m e o , World Bank. D u m o n t , J.C., G. Lemaître (2004), Counting immigrants and expatriates in O E C D countries: a new perspective. Mimeo, O E C D . D u m o n t J-C., G. Lemaître (2007), Enjeux et limites des politiques migratoires sélectives à des fins d'emploi, Actes du 17e congrès des économistes belges de langue française. Cifop, Charleroi. H a g o p i a n Α., M.J. T h o m p s o n , M . Fordyce, K.E. Johnson, G.L. H a r t (2004), The migration of physicians f r o m sub-Saharan Africa to the United States of America: measures of the African brain drain. H u m a n Resources for Health 2 - 1 7 . Javorcik, B.S., C. Ozden, M . Spatareanu, I.C. Neagu (2006), Migrant N e t w o r k s and Foreign Direct Investment. M i m e o , Rutgers University, November. Kangasniemi, M . , L.A. Winters, S. C o m m a n d e r (2007), Is the medical brain drain beneficial? Evidence f r o m overseas doctors in the UK. Social Science and Medicine 65: 9 1 5 - 9 2 3 . Kapur, D. (2002), The causes and consequences of India's IT boom. India Review 1: 9 1 - 1 1 0 . Kapur, D. (2009), The impact of migration f r o m India on India. Princeton University Press, forthcoming. Kapur, D., J. M c H a l e (2005), Sojourns and Software: Internationally Mobile H u m a n Capital and High Tech Industry Development in India, Ireland and Israel. In: A. Arora, A. Gambardella, op. cit. Kerr, W.R. (2008), Ethnic Scientific Communities and International Technology Diffusion. Review of Economics and Statistics, 90: 5 1 8 - 5 3 7 . Kremer, M . (1993), The O-ring theory of economic development. Quarterly Journal of Economics 108: 5 5 1 - 7 5 . Kugler, M., H . R a p o p o r t (2007), International labor and capital flows: complements or substitutes? Economics Letters 94: 1 5 5 - 6 2 . Li, X., J. M c Hale (2006), Does brain drain lead to institutional gain? A cross country empirical investigation. M i m e o , Queens University. Moullan, Y. (2008), Can Health Foreign Assistance break the Medical Brain Drain? M i m e o , University of Paris-1 Sorbonne.

Documenting the Brain Drain of "La Crème de la Crème" · 705

M o u n t f o r d , A. (1997), Can a brain drain be good for growth in the source economy? Journal of Development Economics 53: 2 8 7 - 3 0 3 . M o u n t f o r d , Α., H . R a p o p o r t (2007), Brain Drain and the World Distribution of Income and Population. C R E A M Working Paper N o 04/07, M a r c h . N a n d a , R., T. Khanna (2009), Diasporas and Domestic Entrepreneurs: Evidence from the Indian Software Industry. Working Paper, H a r v a r d Business School. Saxenian, A. (1999), Silicon Valley's N e w Immigrant Entrepreneurs. Public Policy Institute of California, San Francisco, CA. Saxenian, A. (2001), Bangalore, the Silicon Valley of India? CREDPR Working Paper N o 91, Stanford University. Saxenian, A. (2002), Local and Global N e t w o r k s of Immigrant Professionals in Silicon Valley. Public Policy Institute of California. Saxenian, A. (2006), The N e w Argonauts: Regional Advantage in a Global Economy. Cambridge, MA: H a r v a r d University Press. Spilimbergo, A. (2009), Democracy and foreign education. American Economic Review 91: 528-543. Tritah, A. (2008), The Brain Drain between Knowledge Based Economies: the European H u m a n Capital Outflows to the US. CEPII Working paper N o 2008-08. Vidal, J.P. (1998), The effect of emigration on h u m a n capital formation. Journal of Population Economics. 11: 5 8 9 - 6 0 0 . W a d h w a , V., A.L. Saxenian, B. Rissing, G. Gereffi (2007), America's new immigrant entrepreneurs. Working Paper, U C Berkeley. Weizsäcker, J. von (2006), A European Blue Card Proposal, Horizons Stratégiques. N o 1, July. Weizsäcker, J. von (2008), Strait is the gate: Europe's immigration priorities. Bruegel Policy Brief 2008/05, July. Frédéric Docquier, 1RES, Université Catholique de Louvain (Office D.232) 3, Place Montesquieu, B-1348 Louvain-La-Neuve, Belgium, and IZA and CReAM. E-Mail: [email protected] Prof. Hillel R a p o p o r t , Dept. of Economics, Bar lian University, 5 2 9 0 0 R a m a t Gan, Israel, and EQUIPPE, Université de Lille II, CReAM and CEP REMAP. E-Mail: [email protected]

Jahrbücher f. Nationalökonomie u. Statistik (Lucius & Lucius, Stuttgart 2009) Bd. (Vol.) 229/6

Endogenous Skill Formation and the Source Country Effects of Skilled Labor Emigration from Developing Countries By Hartmut Egger, Bayreuth, and Gabriel Felbermayr, Stuttgart* JEL F22, J24 Emigration, endogenous skill formation, source country effects.

Summary In this paper we set up a simple theoretical framework to study the possible source country effects of skilled labor emigration from developing countries. We show that for given technologies, labor market integration necessarily lowers GDP per capita in a poor source country of emigration, because it distorts the education decision of individuals. As pointed out by our analysis, a negative source country effect also materializes if all agents face identical emigration probabilities, irrespective of their education levels. This is in sharp contrast to the case of exogenous skill supply. Allowing for human capital spillovers, we further show that with social returns to schooling there may be a counteracting positive source country effect if the prospect of emigration stimulates the incentives to acquire education. Since, in general, the source country effects are not clear, we calibrate our model for four major source countries - Mexico, Turkey, Morocco, and the Philippines - and show that an increase in emigration rates beyond those observed in the year 2000 is very likely to lower GDP per capita in poor economies.

Introduction H o w to design migration rules is one of the most controversially discussed issues in policy circles and the general public alike. Accounting for concerns of domestic unskilled workers, many industrialized countries nowadays apply selective standards that bias immigration quota toward skilled applicants. That such standards are beneficial for the host (i.e., the immigration) country is by now broadly accepted in the economics discipline. To the contrary, the consequences for the source (i.e., the emigration) country are not settled yet. The rather small existing literature has focused on the effects that emigration of educated workers has on those who do not emigrate. The traditional view is that such a brain drain is harmful for the source country. As pointed out by Freeman (2006), with exogenous schooling, a fall in the supply of skilled labor raises the wages of skilled non-migrants, while at the same time it reduces the wages of unskilled workers, with negative consequences for income of those who stay in the source country. 1 * We thank t w o anonymous referees for helpful comments and suggestions. Felbermayr gratefully acknowledges financial support from the Fritz Thyssen Foundation (grant no. Az. 1 0 . 0 6 . 1 . 1 1 1 ) . This paper is an updated and extended version of CESifo Working Paper Nr. 2 0 1 8 . 1 Based on the insight that brain drain harms those w h o are left behind, Bhagwati and Hamada (1974) have called for a tax on educated emigrants in order to counter the negative effect.

Endogenous Skill Formation and the Source Country Effects of Skilled Labor Emigration · 707

Despite the loss of labor, a brain drain can as well be beneficial for the source country. In particular, those who emigrate may pay remittances to their relatives at home or they may bring with them knowledge and experience from the host country when returning in later stages of their working career. A more recent literature has emphasized a further channel through which the source country can benefit even if remittances and return migration do not materialize. The prospect of emigration may raise the expected return to schooling and thus render education more attractive. To the extent that only a small fraction of those who are educated actually emigrates and that there is a high social rate of return to education, those who are left behind may actually be better off after the brain drain (see Mountford 1997, Beine et al. 2001). This argument has sparked a lot of interest in recent years, because it indicates that selective immigration policies in the industrialized world may be beneficial for developing countries as well. In this paper, we reconsider important aspects of losses and benefits for the source country of emigration. For this purpose, we set up a neoclassical model with a single production sector and emigration from a developing into an industrialized economy. To facilitate our analysis, we associate the developing country with a small economy, implying that factor prices in the industrialized country are independent of the prevailing emigration rates. Along the lines of the traditional migration literature, we consider two types of workers who differ in their education levels. We enrich this simple framework with several features that have been emphasized to be important in the new brain drain literature. In particular, we assume that migration is represented by a lottery with exogenous emigration probabilities (which can be skill-specific). Furthermore, we assume that the agents differ in their abilities and we account for adjustments in the schooling decision that can either be triggered by changes in the emigration probabilities or changes in foreign wages. In contrast to recent work on the brain drain, however, we do not associate schooling with a simple increase in the effective labor supply of workers (see, e.g., Stark/Wang 2002) but instead assume that individuals who acquire education become skilled and thereby complements to unskilled agents in the production process (see, e.g., Egger et al. 2007). In this setting, we show how the prospect of emigration distorts the private schooling decision with negative consequences for GDP per capita in the developing country. Clearly, this distortion is not confined to the case of selective migration in the brain drain context but it also materializes if emigration probabilities are the same for skilled and unskilled workers. Hence, in order to highlight the distortion of private schooling decisions as an important aspect of negative source country effects of emigration, we choose non-selective emigration as the baseline scenario of our analysis. In this case, a higher non-selective emigration propensity would not affect GDP per capita in the source country if schooling were exogenous. This corresponds to the traditional view on possible migration effects (see Freeman 2006). However, with endogenous schooling, individuals have an incentive to adjust their decision upon acquiring education if relative wages in the host country differ from relative wages in the source country. In the empirically relevant case of higher wage inequality in developing countries the incentives for participating in schooling decline with a higher emigration probability. This result is in contrast to one of the key findings of the new brain drain literature that a higher probability of emigration boosts the incentives for education in the source country (see Stark et al. 1998, Beine et al. 2001, 2008).

708 · H. Egger, G. Felbermayr

To see whether and to what extent the results from our results are driven by the assumption that emigration rates are the same for skilled and unskilled workers, we extend our model to selective migration with a positive emigration rate of skilled workers and no migration of unskilled ones. This is the scenario typically accounted for in the mew brain drain literature (see Mountford 1997). With selective migration, the prospect of emigration indeed increases the incentives for acquiring education, irrespective of the prevailing international differences in relative wages. Despite this difference to our baseline scenario, GDP per capita also deteriorates if emigration has an education bias. Similar to models with exogenous schooling, the loss of educated workers lowers skill intensity in production. This raises wages of skilled workers and lowers wages of unskilled workers, with negative consequences on GDP per capita. While in our setting, the former effect is counteracted by an increase in the incentives to acquire education, this second indirect effect cannot dominate and hence GDP per capita is lower in the migration scenario than in a scenario without migration. This is intuitive because the skill intensity typically deviates from its optimal level and less able agents acquire education if migration is possible for skilled workers. With the no migration equilibrium being first best, by construction, our baseline model rules out any positive efficiency effects of skilled labor emigration in the source country. To account for such positive effects, that have been highlighted in the new brain drain literature, we introduce a social return to schooling. Similar to existing studies on the matter, we assume that the social return arises due to a positive externality of education on total factor productivity. To be more specific, we assume that total factor productivity depends positively on the skill intensity in production (see Siiekum 2006, 2008). In this case, the negative GDP per capita effects of emigration are reinforced if schooling incentives fall, while the externality counteracts the negative effect from the distortion in schooling, if acquiring education becomes more attractive. Since it is in general not clear which of the two effects dominates, we calibrate our model for four major source countries: Mexico, Turkey, Morocco, and the Philippines. To give the new brain drain literature a fair chance, we choose a rather accommodating parameterization of the human capital externality. The calibration results indicate that even in this case there is little reason for being overly optimistic about emigration of highly educated agents to boost GDP per capita through adjustments in the schooling decision. However, it may still be possible that emigration leads to benefits through other channels, which are excluded from our analysis, such as remittances or return migration. The remainder of the paper is structured as follows. In Section 2, we review stylized facts on emigration patterns, complementarities between skilled and unskilled workers, wage inequality and social returns to schooling in order to motivate the key assumptions in the theoretical model that we set up in Section 3. In Section 4 we present the main analytical results for the baseline scenario with identical emigration probabilities for skilled and unskilled workers. Section 5 discusses selective migration in which only skilled workers are allowed to emigrate, while unskilled workers are immobile. Section 6 introduces social returns to schooling by accounting for an externality of education on total factor productivity. Section 7 reports the results from several numerical simulation exercises and provides insights on whether positive GDP per capita effects of emigration can be expected when relying on realistic parameter domains. Section 8 relates the results from our analysis to existing insights from the empirical brain drain literature. Finally, Section 9 concludes with a brief summary of the most important results.

Endogenous Skill Formation and the Source Country Effects of Skilled Labor Emigration · 709

2

Stylized facts

In this section, we present a number of stylized facts which will guide our modeling approach. First, recent evidence suggests that emigration of both educated and uneducated workers from poor countries is quantitatively important. Hence, a model with identical emigration probabilities for skilled and unskilled workers may indeed characterize an interesting benchmark case. Second, we briefly review recent evidence on the elasticity of substitution between different education groups and come to the conclusion that technological complementarity between workers with different levels of education indeed exists. Third, we rely on several data sets and different methods to show that real wage income is higher in the host country of emigration, while the education premium is typically much higher in source than in host countries. Fourth, we review evidence on social returns to schooling, since these returns have been emphasized to be important in the recent theoretical brain drain literature. 2.1 Emigration rates

Docquier and Marfouk (2006) provide an extensive data set on migration patterns, which has the unique advantage that it differentiates migrants according to their education levels. Unfortunately, the data do not provide information on non-OECD host countries, and they are only available for the year 2000. Docquier and Marfouk (2006) distinguish migrants with a low, medium and high schooling level. Following the existing literature, we combine individuals with low and medium education levels into the group of uneducated (or unskilled) workers, while a high level of schooling refers to educated (or skilled) workers in our model (see below). Based on this data set, Figure 1 depicts emigration rates for a collection of poor countries (with per capita incomes below the world average in the year 2000) that together account for about 75 percent of immigration into the OECD. The right-hand-side axis records the contribution of each country to the overall stock of emigrants: 18 percent of all emigrants are Mexicans, about 5 percent are Turks, about 4 percent are Indian, Chinese or from the Philippines, and so on. The left-hand-side axis shows the skill-specific emigration rate for educated (p e ) and uneducated (p") individuals. While, pe > p" holds for almost all listed countries, the difference in emigration rates, pe — p " , is small on average 2 and it is particularly small for the two largest source countries, Mexico and Turkey. 3 From this, we conclude that identical propensities for both skill groups are a good description of real world migration patterns (at least for large source countries) and hence provide an interesting benchmark for our theoretical analysis.

2

3

The overall migration rate of educated workers is 5.4 percent while that of uneducated workers is 1.4 percent. The data by Docquier and Marfouk (2006) does not account for illegal and/or undocumented migration. The stock of illegal immigrants is estimated to comprise about 30 to 40 million people worldwide, with approximately 10 million of these illegal immigrants living in the U.S. and a similar number living in Europe (Papademetriou 2005). Information about the educational background of illegal migrants is scarce. A survey carried out by the Pew Institute however shows that about two thirds of undocumented Mexican migrants in the U.S. have only very low levels of education (Kochhar 2005). Hence, the omission of illegal immigration from the data shown in Figure 1 may well bias the emigration rates of uneducated workers downwards, possibly by large amounts.

710 · H. Egger, G. Felbermayr

ρ ' (%)

•

ρ " { % | — · — S h a r e in world stock of migrants (%, right hand axis)

Figure 1 Skill-specific emigration rates and the cumulated stock of emigrants

2.2 Substitutability of different skills in production One important assumption which distinguishes our analysis from previous contributions to the new brain drain literature is that workers with different education levels are imperfect substitutes in the production of goods. This feature of technology is widely documented for OECD countries and it is indeed central to many explanations for rising wage inequality (see, e.g., Autor et al. 2008). For non-OECD source countries of emigration, evidence is rather scarce. However, a recent comparative study by Aydemir and Borjas (2007) shows that the elasticity of substitution between different education levels is about 3 in Mexico. 4 This is probably the best available estimate and we take it as representative for other poor source countries in our numerical simulation exercises in Section 7.

2.3 Inequality measures In the presence of migration opportunities, the economic incentives for skill acquisition depend on the expected return to education, which, in turn, is a function of withincountry skill premia and the real labor income between countries. 5 We start with the latter.

4

5

The hypothesis of workers being perfect substitutes up to efficiency differences is rejected at high degrees of statistical significance. Clearly, in a model with labor market imperfections, skill-specific unemployment rates at home and abroad would influence the education decision as well. Unfortunately, we do not have access to reliable data on skill-specific unemployment rates in developing countries and, hence, emanate from the assumption that the presented figures on wage differntials are a good proxy for differences in expected labor income.

Endogenous Skill Formation and the Source Country Effects of Skilled Labor Emigration · 711

Wage differentials between countries. Real wages in typical source countries of emigration are much lower than those in host countries. This is not surprising, since wage differentials are major determinants of emigration (Borjas 1987). However, the importance of the gap is striking. Using PPP adjusted GDP per capita as a proxy for real wages, 6 Table 1 displays information on the most important source counties of emigration (column 4). 7 Moreover, the last column presents GDP per capita of the source country relative to the average GDP per capita of the host country of emigration. The host country averages are thereby computed for each source country separately, using bilateral emigration rates from Docquier and Marfouk (2006) as weights. The per capita income gap ranges between 1.43 for Taiwan to 14.36 for Vietnam in our data set. Wage inequality within countries. A large empirical literature provides details on crosscountry differences in the returns to schooling. Psacharopoulos and Patrinos (2004) provide a recent survey on Mincerian wage regressions. They report that "[t]he highest returns are recorded for low-income and middle-income countries" (p. 112). For example, the average private return to an additional year of schooling ranges between 11.3 and 13.4 percent in the OECD, between 13.6 and 18.8 in emerging non-OECD countries, and between 24.6 and 37.6 in sub-Saharan Africa. Furthermore, returns also differ significantly within these groups of countries. In particular, they are smaller in Europe than in the U.S. (See Wössmann 2 0 0 3 , for a similar assessment.) Since the schooling differential in years between educated and uneducated workers is roughly 8 to 10 years, the data suggest that the wage premium should be about 100 percent in continental Europe, somewhat higher in the U.S. and above 150 percent in many source countries. This is in line with Fernandez et al. (2005, Table I) who provide direct evidence on the skill premium for 34 developing and industrial countries and find that it is much bigger in developing countries (e.g., Mexico: 3.162) than in developed countries (e.g., U.S.: 1.743; Germany: 1.489). A recent book edited by Lazear and Shaw (2009) discusses the structure of wages in several OECD countries. While not directly informative on the wage differential between educated and uneducated individuals, their evidence suggests that wage differentials exhibit important variation across countries. For example, the ratio of the 75th to the 25th percentile of the wage distribution in year 2 0 0 0 is 2.72 in the U.S. and 1.91 in Germany. We may take these percentiles as representative for the upper and the lower tier of the wage distribution and identify their ratio as a measure of inequality between educated and uneducated individuals. Data from existing country reports indicate that wage inequality importantly contributes to overall income inequality, so that aggregate indicators of inequality can be used to infer cross-country differences in the education premium. Data column 2 in Table 1 shows the most recent available Gini coefficients of overall income inequality for urban populations in developing countries taken from the World Income Inequality Database.

6

7

Focusing on low-skilled workers, Ashenfelter and Jurajda (2004) compare real wages in the fast food sector across a wide range of countries. They find that purchasing power parity (PPP) adjusted U.S.$ wages for identical jobs in the U.S., Japan, and Western Europe are four to five times higher than in Eastern Europe, Korea, or Brazil. The respective difference is even more pronounced between rich industrialized economies and China, India, or Colombia. However, the country coverage in this study is fairly limited so that we prefer to use the real GDP per capita proxy. The information is taken from the Penn World Tables mark 6.2 and refers to year 2 0 0 0 since for this year we have data on skill-specific emigration rates (see above).

712 • H. Egger, G. Felbermayr

ω S o oo ^t· to τ- όν τ— (Ν (Ν (Ν ι - m to σ\ ιν o α\ in αϊ m IN (Ν in IN IV τ— rri to oo «3 rri H ον 00 κ Τtri iri rri Έ- iri in to (Ν to iri to

(Ν *— τ— σι ^f m Ο (ν m (Ν

co to m m oo m ΓΜ o σ\ O in σ\ τ— σ\ O (Ν m o t IV to 00 o 00 to in tv m ιν to 00 o o s 00 *— o Γν ΓΝ ιν to to IV σι τ— tv σ\ IV m in to co m ΓΜ m m (Ν ΓΜ ΓΜ ΓΜ ΓΜ m ΓΜ ΓΜ ΓΜ m ΓΜ (Ν

tv σ> m to σ\ m tv ΓΜ m tD 00 •Ψ m ΓΜ o s Κ X ori 00 Μ iri ο τ- ΓΜ ΓΜ to Ω " ^ m Γ m m m m m m m

in ' t σ> o 00 Ol to in ΓΜ τ— o oo m •3· σ tv ΓΜ m

τX

— —

oo O in 00 ΓΜ σ\ tv oo

IV m o m

to t o to

ΓΜ m 00 oo o ΓΜ tp ΓΜ tv to ai rri to iri ò oo ó m m m m ΓΜ m

o 00 o to

τ— m tv IV τ— to tv σ\ ΓΜ ΓΜ •Ί· ιη σ> ΓΜ 3

in to to O to tv IV m o to O O σ\ ι f tv ΓΜ tv τ— to O m σ\ *— m ΓΜ m m ΓΜ ΓΜ m

in IV ori m

m in σ\ σ\ in tv τ— m ο to τ- ó m m m

IV to m in to Vtv 00 oi IV m m m m

in rn tv m

to ΓΝ tv to oo ΓΜ

m ΓΜ ιη m ο ΓΜ σι to ΓΜ in ιη m ΓΜ σ\ Ο - 8 X 2 = J - c .2 δ ΐ Ξ ί I i i

Endogenous Skill Formation and the Source Country Effects of Skilled Labor Emigration • 713

Across the main source countries of emigration, the Gini coefficient varies between 27 percent (Cuba) and 59.8 percent (Brazil). We also compute the average Gini coefficients in the host countries by weighing host country Ginis with source-country specific bilateral emigration rates (column 6): Aside from a few exceptions (most notably Cuba), inequality is larger in the source than in the host countries of emigration. Finally, we use information on wage inequality across occupations in industrial sectors from Galbraith and Lu (2001) to check the robustness of our above findings. The Theil index on wage inequality reaches an average value of 7 percent in the 119 poorest countries and of 3 percent in the richest 25 economies. Data column 3 in Table 1 reports further details on this index for the key source countries of emigration. While the Theil indices are only partly informative about cross-country differences in returns to education they are highly correlated with Gini coefficients (for which data coverage is higher). We can thus conclude that wage inequality is considerably higher in poor source countries than in the rich host countries of emigration. 2.4 Social returns to schooling It is by now broadly accepted among economists that private and social returns to education differ. For instance, Schultz (1988: 545) writes "education is widely viewed as a public good (with positive externalities), which increases the efficiency of economic and political institutions while hastening the pace of scientific advancement on which modern economic growth depends." Theoretical work has also pointed to the relevance of human capital externalities as a source of economic development (see, e.g., Lucas 1990). However, reliable empirical work on the magnitude of these externalities is scarce and only available for industrialized economies (mainly the U.S.). Rauch (1993) uses micro-level data to quantify human capital externalities. Taking the average schooling gap between educated an uneducated workers to be approximately 8 years, the numbers provided by Rauch imply that a 10 percent increase in average schooling (evaluated at the minimum schooling requirement of 8 years) increases productivity of all workers by about 2.4 to 4.1 percent. Acemoglu and Angrist (2000) criticize that Rauch's estimates are likely to be upward biased due to problems of regressor endogeneity. They propose to use a natural experiment to quantify the size of human capital externalities in the U.S. In their analysis the externalities are rather small (not larger than 1 percent), and they cannot reject the null hypothesis of returns being zero. The final word on the size of schooling externalities is still out, in particular for our sample of poor source countries. However, recent evidence for the U.S. suggests that human capital externalities are probably not too large.8

8

As pointed out by Wössmann (2003), any reliable estimate of the social returns to schooling should capture "the full social benefits of education which may contain any externalities arising from education, and the full social costs of education, including any public funding" (p. 373). Hence, by abstracting from the social costs of education the results in Rauch (1993) clearly overestimate the true social returns to education also from this perspective. However, calculating reliable estimates for the social costs of schooling is difficult, because these costs critically depend on both the prevailing schooling system and the prevailing tax system and may therefore differ substantially between countries. For a first attempt to assess how the existence of social costs affects the returns to education, see Psacharopoulos and Patrinos (2004).

714 • H. Egger, G. Felbermayr

3

An emigration model with endogenous skill supply

We consider a small one-sector economy, 'South', which is populated by a unit mass of individuals. The representative firm in this economy employs skilled, £ , and unskilled, U, labor to manufacture a homogeneous numéraire good Y, which is internationally tradable without any impediments. Considering a linear-homogeneous technology, we can write the production function in intensive form as Y = AUf(e), where e = E/U denotes the skill intensity in production and A measures total factor productivity (TFP). 9 f(-) has the usual properties, />(·) > 0 and f(-) < 0, and it satisfies the Inada conditions. Hence, in line with the empirical evidence in Section 2, skilled and unskilled labor are complementary production factors. All markets are perfectly competitive, so that workers are paid their marginal products. The local supply of skilled and unskilled labor is endogenous and depends on both the individual education decision (discussed in detail below) and the emigration rates of the two skill groups. Following the new brain drain literature (cf. Mountford 1997, Beine et al. 2001), emigration is modeled as a lottery outcome, with all workers of a specific skill group facing the same probability of successful emigration, p' 6 (0,1), i — E, U. We assume that the host country of emigration, 'North', is large and sufficiently rich to render emigration always beneficial for southern workers. Intuitively, this is the case if both TFP in South and emigration propensity p' are sufficiently small. Individuals in the southern economy differ in their innate learning abilities a € [0,1]. These abilities are distributed according to a c.d.f. G(a), with G/(a) > 0. Educated agents supply a efficiency units of skilled labor E, while uneducated agents supply one unit of unskilled labor U.w Hence, 1 — a describes the private cost of education in terms of lower working time. Risk neutral agents maximize expected income by choosing whether or not to get educated. The expected income crucially depends on learning abilities and the skill-specific propensities to emigrate. Using an asterisk to indicate northern variables and denoting skilled and unskilled wages per efficiency unit by r and w, respectively, the expected return to a unit of skilled and unskilled labor is given by r = peer* + (1 — pe)r and w = p"w* + (1 — p")w, respectively. The marginal individual, that is indifferent between education and non-education, is determined by the cutoff ability condition ä = 1 /ω, where ω ξ r/w is the expected return to education. One can rewrite that condition in terms of within-country wage inequality measures (i.e., skill premia) (a = r/w and w*=r*/w*, and the between-country unskilled wage differential q = w* jw : a = - = ω (1 -pe)œ

+ peœ*q

(1) '

κ

With a share pe of educated and a share p" of uneducated individuals leaving South to work in North, the supply of unskilled and skilled labor that is available for production

9

10

For the moment, we treat A parametrically, while in Section 6 we assume that T F P depends positively on the skill intensity in production in order to account for social returns t o schooling. We do not account for the possibility that educated individuals execute unskilled tasks (see Davidson et al. 2 0 0 8 ) . This presumes that the returns to skilled labor are higher than the returns to unskilled labor, which we take for granted in the subsequent analysis.

Endogenous Skill Formation and the Source Country Effects of Skilled Labor Emigration · 7 1 5

in the source country is given by U = (1 - pu)G(a) and E = (1 - pe) J^1 adG(a), respectively. Labor market clearing hence implies that the skill intensity in southern production is given by (l-p

(1 -p")G(a)

'

(¿)

where limä_o+ e(à) = oo, lim ä _i- e(a) = 0, and et {a) < 0 hold. Intuitively, a higher ability threshold means that less individuals participate in education and hence the skill intensity in southern production falls, all other things being equal. Using the skill intensity from Eq. (2) in the solution to the profit-maximization problem of the competitive producers, which is characterized by r = Afi(e) and w = Af(e) — Aefi(e), it is straightforward to determine the skill premium (per efficiency unit of labor) in South as well as the between-country unskilled wage differential. The equilibrium values of these two variables are given by f'jeja)) f(e(â))e(â)f,(e(â))

, q

w* Af(e(ä))-e(ä)Af,(e(ä))>

{Λ)

respectively. The first expression determines a positive relationship between the skill premium ω and marginal ability â. This is intuitive, as a higher cutoff threshold reduces the number of educated workers and thereby lowers skill intensity, with a positive impact on the skill premium. Furthermore, under a linearly homogeneous production technology the higher skill premium is associated with a decline in the absolute return to unskilled labor in South and, hence, between-country unskilled wage inequality must increase for a given w*. This explains the positive relationship between a and q as determined by the second expression in Eq. 3. The equilibrium cutoff ability, skill intensity and wage inequality are jointly determined by Eqs. (l)-(3). Figure 2 illustrates the equilibrium in the (l/,ä)-space. The upward-sloping 45-degree line depicts the left-hand-side of Eq. (1). Furthermore, substituting Eq. (2) into Eq. (3) and using the resulting expression in the right-hand-side of Eq. (1) gives a function Ω(α), which is positive and strictly decreasing in a. It is convex to the origin and satisfies lim^o Ω{α) = oo.11 When a goes up, a smaller fraction of agents invests into schooling and the education intensity in production goes down (see Eq. 2). This puts downward pressure on wages for uneducated agents, while those of educated individuals rise. Hence, both the inverse of the skill premium, 1 /ω, and 1 ¡ ω fall. Figure 2 shows that the model exhibits a unique equilibrium, with the equilibrium cutoff ability level being denoted by a(pe ,p") to indicate the dependence of the cutoff level on the prevailing emigration rates, pe and p". In the next two sections, we are particularly interested

11

An explicit solution for Ω (a) is given by = K>

puw* + {\-p")A\f{e{a))f'{ 0 and p" = 0 in Section 5. This corresponds to the case that has been put forward in the new brain drain literature. 4

The case of identical emigration probabilities

We now consider emigration probabilities that are the same for both education groups. Then, the impact of a higher propensity to emigrate on cutoff ability a and skill premium ω can be summarized in the following way. Proposition 1. An increase in the emigration propensity p lowers (raises) the incentives for education if the skill premium in the host country is lower (larger) than the skill premium in the source country. The skill premium in South increases (declines) in this case. Proof. Analysis in the text. The impact of an increase in emigration propensity p(= pe = p") on a and ω can be read off from Figure 2. A higher p increases the expected income of both skill groups, because oír* > r and w* > w. The expected income shift is however more pronounced for uneducated workers if the skill premium in North is lower than the skill premium in South and, hence, schooling becomes less attractive. Graphically, the ß(ä)-locus shifts outwards in Figure 2, implying an increase in ä. Since not all agents who change their schooling decision in view of the higher emigration propensity can indeed emigrate, the prospect to emigrate induces an underinvestment in education and hence it leads to an increase in the southern skill premium. Conversely, if the skill premium in North is higher than the skill premium in South, the incentives for acquiring education increase. In this case, the ß(ä)-locus shifts inwards in Figure 2, so that ä increases. Since not all newly educated workers can actually emigrate, there is overinvestment in education with a negative effect on the southern skill premium. The result in Proposition 1 is remarkable, because it qualifies the insight from the new brain drain literature that higher emigration probabilities render the participation in schooling more attractive and, hence, can be interpreted as a substitute for an education subsidy in developing countries (see Stark/Wang 2002). As it turns out from our analysis, if the increase in emigration propensity is non-selective, the schooling incentives fall in the empirically relevant case of a higher skill premium in the poor source than the rich host country of emigration. 1 3 Even though the case of a higher southern skill premium seems to have considerable empirical support, we do not need to impose any assumption concerning the relation between ω and ω*, when deriving our results on GDP per capita effects. The reason is that except of the borderline case with ω = ω* emigration distorts the private schooling decision and, hence, always exhibits a negative effect on per capita income. For a more formal treatment of the respective GDP per capita effect, we can note that in the case of identical emigration rates, Eq. (4) simplifies to V = G(a)Af (e(a)). Then, the following result is immediate.

13

Clearly, in our model cross-country differences in the skill premium are only a g o o d proxy for the cross-country differences in income inequality if the effective skilled labor supply per educated worker does not differ too much between the source and the host country of emigration.

718 · H. Egger, G. Felbermayr

Proposition 2. If an increase in the emigration propensity p leads to an adjustment in the schooling decision and thus to a change in cutoff ability a and in skill intensity e (a), GDP per capita falls. In contrast, an increase in ρ does not exhibit an impact on GDP per capita if the schooling decision and thus cutoff ability a as well as the skill intensity e(a) remain unaffected. Proof. We can consider dV/dp = dV/da χ da/dp. Noting that dV/da — Gi(a)Af(e(a)) +G{ä)Afi(e(ä))e/(ä) and substituting ef(a)G(a)/e(a) = — G/(ä)[l + a/e(a)\, according to Eq. (2), implies

Using Eqs. (1) and (3), we further obtain ^

=

AGr(ä)\f(e(ä))-fHe(ä))e(ä)}

pq ω[(1 - ρ)ω +

pœ*q]

(ω* — ω).

Noting from Proposition 1 that dä/dp > , = , < 0 if ω >, =, < ω*, we can conclude that dV/dp < 0 if ω φ ω* and dV/dp = 0 if ω = ω*. This completes the proof of Proposition!. The prospect of emigration implies that the incentives for education become increasingly dependent on northern relative factor prices. However, only local technology conditions are relevant for maximizing GDP per capita in South. Thus, an increase in p widens the gap between the incentives for schooling and the optimal southern skill intensity, as long as the factor price differential in North and South do not coincide (i.e., ω* φ ω). If the incentives for education do not change, i.e., if a is constant, the relative skill supply remains unaffected and so does GDP per capita. This follows immediately when noting that changes in p affect V = G(a)Af(e{a)) only through adjustments in the schooling decision, i.e., through adjustments in â. Hence, the negative GDP per capita effect of an increase in the emigration propensity p does not materialize in a model with exogenous skill supply. This suggest that accounting for endogenous schooling is indeed important for drawing a comprehensive picture about the possible source country effects of emigration. Beyond the GDP per capita effects of emigration, we can also determine the distributional consequences in the southern economy by looking at the impact that a change in the emigration propensity exhibits on income inequality R. The following proposition summarizes these effects. Proposition 3. An increase R(äp), if the skill premium i.e., if ω > ( 0. Furthermore, we can note that dp,¡da = G/(â) [a - a]dG{a)j / [ I - G (Λ)] > 0. Putting together, we thus obtain dR/dä > 0. Noting finally da/dp > , = , < Oifcu > , = , < ω* from Proposition 1, completes the proof of Proposition 3. By virtue of Proposition 3, the distributional consequences of emigration depend on whether the skill premium in South is lower or higher than the skill premium in North.

Endogenous Skill Formation and the Source Country Effects of Skilled Labor Emigration · 719

Emigration into an egalitarian economy (with ω > ω*) raises both the skill premium ω and the income ratio between educated and uneducated workers, R. The opposite holds true for emigration into a non-egalitarian country (with ω < ω*). 1 4 As empirical stylized facts indicate that North is more egalitarian, we can conclude that an increase in emigration propensity p not only lowers GDP per capita but it also raises income inequality in South, if schooling is endogenous.

5

Selective migration

In contrast to the case of identical emigration propensities for both skill groups in the last section, we now assume that the respective propensities are skill-specific, with pe > 0 and p" — 0. Unfortunately, with selective migration the formal analysis turns out to be quite complicated with a general linear-homogeneous production technology.15 Therefore, we rely on a Cobb-Douglas technology in the subsequent analysis. In particular, we assume Y = AG(ä)ea, with a € (0,1). Furthermore, in the interest of readability we discuss our results in an intuitive way and refer the reader interested in the formal derivation of our results to the working paper version of this manuscript in Egger and Felbermayr (2007). If North introduces a (quality-)selective migration policy, there is a direct negative effect on southern GDP per capita due to a loss in the mass of educated workers and, hence, a fall in skill intensity for given private schooling decisions. However, there is also an indirect effect, because the expected income for skilled workers and, therefore, the incentives to acquire education increase (recall r* > r). As a consequence, cutoff ability à falls and e(a) increases ceteris paribus. Graphically, a higher pe shifts the i2( )-locus inwards in Figure 2. The increase in e(a) counteracts the negative direct effect of a pe increase on skill intensity e and, in fact, it may overturn it if the elasticity of labor supply with respect to cutoff ability a is sufficiently large. Irrespective of whether e is higher or lower in the scenario with selective migration than in a scenario without migration, GDP per capita will definitely fall when skilled workers obtain a prospect of emigration. The reason for a negative impact on GDP per capita is twofold. On the one hand, with selective migration, the private schooling decision cannot be expected to establish an efficient skill intensity outcome and, on the other hand, less able agents acquire education in the source country. It is notable that a negative GDP per capita effect of selective migration is not specific to a scenario with endogenous schooling, but can also materialize in a setting in which schooling is exogenous. However, with exogenous schooling a negative GDP per capita effect is only realized if the Cobb-Douglas parameter a is sufficiently large, while the respective effect is positive for low levels of a 1 6 . Furthermore, if cutoff ability ä is 14

15

16

Notably, changes in the emigration propensity p exhibit an impact on income inequality only if schooling is endogenous. With exogenous schooling, ä does not change in response to a higher p and, hence, income inequality R remains unaffected. As compared t o the case of identical emigration propensities, there exists an additional effect on GDP per capita under selective migration: output per capita declines due to a direct negative effect of a pe increase on skill intensity e (see E q . 2 ) . With ä being constant and pu = 0, the first derivative of V with respect to pe is given by dV

dp

(l-a)[(l-pg)+fi£G(á)]-G(5) (1 ~ P c ) [ ( l

—

Pe)

+

according to (4). It is now possible to define a unique a e e while dV/dp > 0 if a < a 0 and dV/dp < 0 if a >a°.

0

P'G(a)] e

'

( 0 , 1 ) , such that

dV/dpe =

0 if

a

= a0,

720 • H. Egger, G. Felbermayr

exogenous, the skill intensity in the southern economy will be definitely lower with selective migration than without migration. This effects of selective migration indicate a crucial difference to the case of non-selective migration, where changes in the emigration propensities would neither affect skill intensity nor GDP per capita in the southern economy, if schooling were exogenous. While the GDP per capita effect of selective migration is unambiguous, the distributional consequences turn out to be less clear. This has the following reason. As pointed out above, there are two counteracting effects of selective migration on skill intensity in production, with the sign of the overall impact being unclear. However, since under a Cobb-Douglas technology the skill premium is inversely proportional to skill intensity and given by co = [α/( 1 — a)][e{a)\ the impact of selective migration on the skill premium is unclear as well. With the skill premium per efficiency unit being a crucial determinant of factor return R, it is hence not surprising that in response to selective migration income inequality may decrease or increase. The compositional effect, however, reinforces the indirect negative effect on ω due to changes in the schooling decision, so that a decline in the skill premium is sufficient for a decline in relative factor return R. To the contrary, both the skill premium and the income inequality effect would be unambiguously positive if schooling were exogenous and hence cutoff ability ä constant. This underpins the relevance of considering endogenous adjustments in the schooling decision when drawing a comprehensive picture of the source country effects of skilled labor emigration.

6

Social returns to education

The analysis in Sections 3 - 5 , relies on the assumption that the no-migration equilibrium is first best. Hence, by construction emigration cannot be efficiency enhancing in the source country. However, positive source country effects have been put forward to be important in the recent brain drain literature. A common approach to modeling positive source country effects of emigration is to allow for social returns to education (see Beineet al. 2001). We follow this approach and assume that total factor productivity, A, depends on the skill intensity in the production process. (See Siidekum 2006, 2008, for a similar approach.) More specifically, we postulate that A = e(a)r, where y is a positive parameter. This opens a second channel through which a change in schooling incentives affects per capita output in the source economy. Quite generally, whenever e (a) falls due to emigration - which happens, e.g., if pe = p" and ω > ω* - the negative GDP per capita effect derived above is magnified. In contrast, if e(a) rises - which may happen, e.g., in the case of selective migration with pe > p" = 0 - the negative GDP per capita effect is counteracted and, possibly, average income in the source country increases. Hence, at least in this case, accounting for a social return to education may generate an ambiguity in the GDP per capita effects of emigration in the source country. In order to assess the likelihood of a positive GDP per capita effect, we use numerical simulation techniques to quantify the positive and negative implications of emigration in the next section. 17 17

Clearly, social returns to education need not be positive (see the respective discussion in Subsect i o n ! ^ ) . For instance, one may assume that part of the education expenditures are borne by the public sector and financed through an income tax. In this case, a higher emigration propensity would raise the tax burden of non-migrants, thereby providing an additional channel through which the source country of emigration may lose.

Endogenous Skill Formation and the Source Country Effects of Skilled Labor Emigration · 721

7

Quantifying the effects of the brain drain

In order to assess whether, in the presence of social returns to education, GDP per capita can be stimulated by a brain drain, we conduct a numerical simulation analysis for four important source countries: Turkey, Mexico, Philippines and Morocco. The first two economies feature a very low gap in the propensities to emigrate for educated and uneducated workers and their average income differential vis-á-vis the host countries is not too big and similar in size. The latter two economies both display a strong bias of emigration rates toward educated workers. They are poorer than the former two ones, but between them, they are comparable regarding their relative average income gap to their respective average destination countries. To calibrate our model, we need to parameterize the production function. For this purpose, we consider a (CES) technology with a constant elasticity of substitution between educated and uneducated workers of σ. In line with evidence for Mexico discussed in Section 2.2 we set σ = 3. 18 Our discussion of empirical evidence on the quantitative importance of human capital spillovers in Section 2.4 leads us to compute two scenarios, one for y = 0 and the other for γ > 0. In the simulation exercises presented in Subsection 7.1, we set γ = 0.36, which corresponds to the highest value that has been found in the literature and may therefore provide an upper bound for the true social benefits. Note, however, that the more recent empirical literature strongly points toward social benefits that are closer to zero. Furthermore, we do not account for social costs of education and, hence, by setting γ = 0.36 we are likely to significantly overestimate the possible GDP per capita gains from brain drain. This is particularly true for very poor economies, in which social returns are considerably smaller than in mediumincome countries. To account for this fact, we set γ = 0.1 in the calibration exercise executed in Subsection 7.2. To determine the supply of skilled and unskilled labor, we proceed in two steps. First, we calibrate the share of educated workers in total labor income such that, at the observed emigration rates 19 shown in Table 1, the respective economies reproduce empirically reasonable education premia of 3.5 in Mexico, 3.2 in the Philippines, 2.6 in Turkey and 2.5 in Morocco. 2 0 Second, regarding abilities we simply consider a uniform distribution. This seems the most natural choice given the absence of direct evidence. Since our theoretical results do not hinge on specific functional forms of G(a), we can be optimistic that the choice of a uniform distribution does not restrict the generality of our results in qualitative terms. In all numerical exercises, we fix the education bias of emigration b = pe /p", by the empirically observed value for the year of 2000 and plot the cutoff ability a, the skill premium ω (whose benchmark value at observed emigration rates turns out to be higher 18

19

20

Choosing σ = 1 would lead to a Cobb-Douglas technology as the one considered in Section 5. Also note that our model would converge to the standard new-brain-drain literature benchmark case without losses from emigration due to changes in skill intensity (see, e.g., Stark 1998) when σ —> oo. The emigration rates in 2000 are pe = 15.3, p" = 12.07 for Mexico; pe = 5.82,p" = 5.59 for Turkey; pe = 16.96,p" = 7.00 for Morocco; and pe = 13.73,p u = 2.17 for Philippines. We compute the ratio between Gini coefficients of income inequality and the 75th to the 25th percentile ratio of the earnings distribution (which we take as a proxy for the wage premium). Data on percentiles is available for the OECD, so that we use the OECD multiple to calibrate wage premia for our source countries using their reported Gini coefficients.

722 · H. Egger, G. Felbermayr

than ω* in all of our exercises), GDP per capita V (relative to its initial level without migration) and income inequality R as a function of pe.21 7.1 Countries with a small bias of emigration propensities toward skilled workers

Figure 3 shows the results f r o m our numerical exercises for Turkey and Mexico when γ = 0. In the case of Turkey, there is almost no education bias (b = 1.04), since pe and p" are of approximately similar magnitudes according to Figure 1. Hence, an increase pe induces an almost pari passu increase in p" and we therefore expect that the numerical analysis replicates the results in Propositions 1 - 3 . In the case of Mexico, educated workers have an emigration probability that is about 3 percentage points larger than the one for uneducated agents and the education bias of emigration b is equal to 1.27. Income in both source countries is four to five times smaller than the emigrationrate weighted average income in the respective host countries. In both economies the degree of domestic inequality is larger than the weighted foreign counterpart. The numerical exercise clearly shows that, given the absence of an education bias, an increase in the emigration rate lowers education incentives and, hence, drives up ä in Turkey. This is because the expected return to schooling falls, once foreign employment opportunities are taken into account. Notably, with a higher cutoff ability ä the skill intensity in Turkey's production falls and hence the skill premium ω has to increase. This effect is well understood from the analysis in Section 3. Furthermore, we find that along with an increase in the cutoff ability level, emigration reduces GDP per capita. In the status quo, where pe is about 6 percent, GDP per capita is 0.1 percent lower than in the counterfactual situation with no migration. It would be 0.25 percent lower if the emigration rate were to double. Proposition 3 suggests that the factor income ratio R should increase if b is identical to unity. Our simulation exercise confirms this and shows that emigration has a sizable effect on inequality. The case of Mexico is different since b is substantially above unity. In this case, an increase in pe boosts education incentives because it increases the likelihood of emigration overproportionally for educated workers. As a consequence, a falls and a higher fraction of individuals receive schooling, when pe increases. Abstracting from a positive externality of schooling, the distortion in the education decision leads to a fall in GDP per capita of Mexico. In contrast to the case of Turkey, the effect of an increase in pe on Mexican income inequality is inversely hump-shaped. At the observed level of pe = 0.15, the predicted income inequality is lower than in the counterfactual where pe = 0, but variation in R due to migration is much smaller than in the case of Turkey. Figure 4 allows for a positive externality of human capital formation on total factor productivity. In both countries, we set the size of the externality to the highest justifiable level and assume γ = 0.36. Since the education bias in emigration rates is fairly low in Turkey, we would expect from our discussion in Sections 4 and 6 that the negative GDP per capita effect is even more pronounced if education exhibits an externality on TFP. The reason is that with wage inequality being higher in the source than in the host country of emigration, a pari passu increase in the propensities to emigrate for both skill groups lowers the incentives to acquire education and hence reduces 21

Holding pe/p" constant, we presume that changes in emigration propensities are interdependent. As pointed out in Section 3, this need not be the case in reality. Hence, at least in this respect our numerical results have to be interpreted with care.

Endogenous Skill Formation and the Source Country Effects of Skilled Labor Emigration • 723

TURKEY

MEXICO

Figure 3 Turkey and Mexico: GDP per capita and distribution effects of emigration without social returns to schooling the skill intensity in source country production. This intuition nicely bears out in Figure 4. Furthermore, the impact of a change in the propensity to emigrate on income inequality in Turkey remains almost unaffected by the introduction of a Hicks-neutral positive externality of schooling on TFP. This is also true for Mexico, where the education bias of emigration is more pronounced than in Turkey. Due to this bias, a higher participation in schooling induces an increase in TFP, which counteracts the negative impact on GDP per capita from the distortion in the private education decision. For relatively low emigration rates, the first effect dominates and hence GDP per capita increases when the propensity to emigrate goes up. Evaluated at status quo emigration rates (pe = 0.15), however, it is the second effect that dominates and GDP per capita turns out to be slightly lower (—0.2) than in the counterfactual situation with no migration. From Figure 4 we can conclude that the GDP per capita effect of an increase in the emigration propensity is hump-shaped and reaches a maximum at rather low levels of emigration rates. Interestingly, at status quo emigration rates, γ = 0.36 turns out to minimize the loss, while lower or higher values of γ turn out to further increase it. The reason is that any change in TFP affects the relative incentive to acquire human capital through its impact on between country inequality. This can further dissociate actual enducation decisions from those that would maximize GDP per capita.

724 · H. Egger, G. Felbermayr

TURKEY

MEXICO

Figure 4 Turkey and Mexico: GDP per capita and distribution effects of emigration with soocial returns to schooling 7.2 Countries with a strong bias of emigration propensities toward skilled workers We now turn to a numerical assessment of the impact of emigration on Morocco and the Philippines. In both countries, emigration rates are strongly biased toward educated workers (see Footnote 19). Hence, the emigration patterns differ substantially from the stylized setup in Section 4. However, the empirical patterns also differ from those assumed in Section 5, as emigration of uneducated workers seems to be important as well. Furthermore, the elasticity of substitution between educational classes is assumed to be 3 in our numerical exercises, while it was set equal to one in the selective migration scenario studied in Section 5. Figure 5 depicts the situation without any externality in human capital formation (i.e., γ = 0). As expected, from the analysis in Section 5, (quality-)selective migration raises the incentives for acquiring education and hence it lowers the ability cutoff for education, ä. While this is true for either economy, the effect turns out to be stronger in the Philippines, whose average host economy (of emigration) features a larger degree of earnings inequality. In both countries, overall income inequality, R, falls for relevant magnitudes of emigration propensities. Turning to GDP per capita, the numerical exercise confirms our insights from the analysis in Section 5, at least for the emigration rates depicted in Figure 5. In both countries, GDP per capita is lower in the status quo than in

Endogenous Skill Formation and the Source Country Effects of Skilled Labor Emigration • 725

MOROCCO

PHILIPPINES

Figure 5 Morocco and Philippines: GDP per capita and distribution effects of emigration without social returns to schooling the reference case of no emigration. The loss is quite dramatic: 2.5 % in the case of Morocco (at pe = 0.17) and 2.9 % in Philippines (at pe = 0.14). Next, we analyze a scenario with a positive externality of schooling. The respective results for Morocco and the Philippines are depicted in Figure 6. As explained above, we set γ = 0.1 to parametrize the schooling externality in the two very poor developing countries under consideration in this subsection. For this parameter value, we find that a higher propensity to emigrate not only raises the incentives to acquire education but it may indeed stimulate GDP per capita due to an increase in TFP. In the case of Morocco, a positive GDP per capita effect materializes only for rather small emigration rates, while at the status quo rates its GDP per capita is lower than in the counterfactual situation of no migration. In contrast, in the Philippines the GDP per capita is higher for almost all considered emigration rates and it reaches a maximum at a rate that is not too far away from its status quo. Similar to Turkey and Mexico, the inclusion of a schooling externality does not influence the distributional effects of emigration significantly.

726 · H. Egger, G. Felbermayr

8

A comparison with the empirical brain drain literature

In a final step of our analysis, we compare the insights from the numerical simulation exercises in the former section with results from the empirical brain drain literature. First of all, there is convincing evidence that the international wage differential is indeed a crucial factor for explaining emigration of skilled workers from developing to developed countries (see Beine et al. 2001). Second, there is also empirical support for the hypothesis that selective emigration of skilled workers increases the incentives for acquiring education in the source countries (see, e.g., Beine et al. 2008, Docquier et al. 2008). This finding is well in line with the simulation results in Figures 5 and 6. 22 However, to the best of our knowledge there is so far no discussion to what extent the education bias of emigration affects these adjustments in the schooling decision. From the numerical results in Figures 3-6, we can conclude that this education bias is indeed important for the respective impact of changes in the emigration propensity on human capital formation in the source country. Furthermore, the respective figures also indicate that the impact of emigration rates needs not be monotonie. This insight may provide an intuition for the empirical observation from cross-section data that schooling incentives do not increase in all source countries in which emigration propensities go up (Beine et al. 2008). Our results also indicate that the impact of emigration on the incentives for acquiring education depends on the prevailing wage inequalities in the source and the destination country, in particular if the education bias in emigration rates is not too strong. However, the empirical relevance of this mechanism has so far neither been tackled in work on the brain drain nor in other subdisciplines of the migration literature. A further aspect that features prominently in our analysis is the impact of selective emigration on GDP per capita. In particular, we distinguish two channels of influence. On the one hand, there is a direct negative effect, due to a reduction in the skill intensity for given education decisions, and, on the other hand, there is an indirect positive effect, if emigration prospects increase the incentives to acquire education. Beine et al. (2001) tackle the GDP per capita effects empirically and find support for the positive indirect effect, while the negative direct one turns out to be insignificant. However, this result is not at odds with ours, because in their data set the authors cannot distinguish between skilled and unskilled migration. Beyond that, Schiff (2006) presents evidence for a humpshaped impact of the brain drain on annual GDP growth rates in developing countries, with growth rates being highest for intermediate levels of emigration rates. This provides support for the respective hump-shaped relationship between GDP per capita and skilled emigration rates in Figures 5 and 6. 9

Concluding remarks

Can skilled labor emigration really be beneficial for the source country, as claimed by the new brain drain literature? To address this question, we have set up a simple model to study the potential losses and benefits of skilled labor emigration in a unified framework. In particular, we have shown that in the absence of social returns to schooling, emigration always lowers GDP per capita in the source country, if schooling is endogenous. This effect arises, irrespective of whether emigration is biased toward skilled workers or not. 22

Schiff (2006) concludes form a review over the existing empirical literature that the existence of a positive brain drain effect on schooling is not settled yet.

Endogenous Skill Formation and the Source Country Effects of Skilled Labor Emigration · 727

MOROCCO

PHILIPPINES

Figure 6 Morocco and Philippines: GDP per capita and distribution effects of emigration with social returns to schooling Furthermore, accounting for social returns to schooling, which are assumed to arise due to a positive externality of skill intensity on total factor productivity, we show that the negative effect on the source country is amplified if the prospect of emigration reduces the incentives to acquire education, while a positive effect of a brain drain possibly materializes if education is stimulated and not all newly educated workers can actually emigrate. Since with social returns to schooling the effects of a brain drain are not clear in general, we rely on numerical simulation techniques to assess the relative strength of the counteracting forces for four important source countries of emigration. Our numerical exercise highlights that GDP per capita effects of skilled labor emigration critically depend on a number of important source country features, such as the education bias of emigration, the degree of income inequality relative to the average host country, and, most importantly, the size of the human capital externality. Relying on rather large values for the size of this externality, we find that in only one out of four cases, the source country may indeed have a higher GDP per capita under status quo emigration rates than in the counterfactual no migration scenario. However, even this result may be too optimistic, because, aside from considering a parameterization of the externality which is at the upper bound of empirically realistic levels, we have totally ignored social costs of

728 · H. Egger, G. Felbermayr

e d u c a t i o n , w h i c h m a y arise d u e t o p u b l i c f i n a n c e of t h e s c h o o l i n g s y s t e m . I n d e e d , a c c o u n t i n g f o r t h e s e costs will s u b s t a n t i a l l y r e d u c e t h e social r e t u r n s t o e d u c a t i o n a n d , h e n c e , w e c a n c o n c l u d e t h a t in c o n t r a s t t o t h e m a i n insight f r o m t h e n e w b r a i n drain literature our analysis does not provide strong support for a positive brain drain e f f e c t o n t h e s o u r c e c o u n t r y of e m i g r a t i o n - at least w h e n a b s t r a c t i n g f r o m r e m i t t a n c e s and return migration.

References Acemoglu, D., J. Angrist (2000), H o w large are h u m a n capital externalities? Evidence f r o m compulsory schooling laws. NBER M a c r o a n n u a l 2000: 9 - 5 9 . Ashenfelter, O.S. J u r a j d a (2004), Cross-country comparisons of wage rates: The Mcwage index. Mimeo: Princeton University. Aydemir, Α., G.J. Borjas (2007), Cross-country variation in the impact of international migration: C a n a d a , Mexico, and the United States. Journal of the European Economic Association 5(4): 6 6 3 - 7 0 8 . Autor, D . H . , L.F. Katz, M.S. Kearney (2008), Trends in U.S. wage inequality: Revising the revisionists. Review of Economics and Statistics 90(2): 3 0 0 - 3 2 3 . Beine, M . , F. Docquier, H . Rapoport (2001), Brain drain and economic growth: theory and evidence. Journal of Development Economics 64(1): 2 7 5 - 2 8 9 . Beine, M . , F. Docquier, H . R a p o p o r t (2008), Brain drain and h u m a n capital formation in developing countries: winners and losers. Economic Journal 118(528): 6 3 1 - 6 5 2 . Bhagwati, J.N., K. H a m a d a (1974), The brain drain, international integration of markets for professionals and unemployment: A theoretical analysis. Journal of Development Economics 1(1): 1 9 - 4 2 . Borjas, G.J. (1987), Self-selection and the earnings of immigrants. American Economic Review, 77(4): 5 3 1 - 5 5 3 . Davidson, C., S.J. Matusz, A. Shevchenko (2008), Globalization and firm level adjustment with imperfect labor markets. Journal of International Economics 75(2): 2 9 5 - 3 0 9 . Docquier, F., O. Faye, P. Pestieau (2008), Is migration a good substitute for education subsidies? Journal of Development Economics 86 (2): 2 6 3 - 2 7 6 . Docquier, F., A. M a r f o u k (2006), International migration by education attainment, 1 9 9 0 - 2 0 0 0 . Pp. 151-199 in: C. Ozden, M . Schiff (eds.), International migration, brain drain and remittances. N e w York: McMillan and Palgrave, Chapter 5. Egger, H., J. Falkinger, V. Grossmann (2007), Brain drain, fiscal competition, and public education expenditure. IZA Discussion Paper N o . 2747. Egger, H., G. Felbermayr (2007), Endogenous Skill Formation and the Source Country Effects of International Labor M a r k e t Integration. CESifo Working Paper N o . 2018. Fernandez, R., N . Guner, J. Knowles (2005), Love and Money: A Theoretical and Empirical Analysis of Household Sorting and Inequality. Quarterly Journal of Economics 120(1): 273-344. Freeman, R.B. (2006), People flows in globalization. Journal of Economic Perspectives 20(2): 145-170. Galbraith, J.K., J. Lu (2001), Measuring the evolution of inequality in the global economy. Pp. 1 6 1 - 1 8 5 in: J.K. Galbraith, M . Berner (eds.), Inequality and industrial change: a global view. Cambridge University Press. Kochhar, R. (2005), Survey of Mexican migrants, part III: The Economic Transition to America. Pew Hispanic Center. Lazear, E.P., K.L. Shaw (2009), The structure of wages: An international comparison. Chicago: University of Chicago Press. Lucas, R.E. (1990), W h y doesn't capital flow f r o m rich to poor countries? American Economic Review 80(2): 9 2 - 9 6 . M o u n t f o r d , A. (1997), Can a brain drain be good for growth in the source economy? Journal of Development Economics 53(2): 2 8 7 - 3 0 3 .

Endogenous Skill Formation and the Source Country Effects of Skilled Labor Emigration · 729

Papademetriou, D.G. (2005), The global struggle with illegal migration: N o end in sight. Migration Policy Institute, Migration Information Source Sep. 2005. Psacharopoulos, G., H.A. Patrinos (2004), Returns to investment in education. A further update. Education Economics 12(2): 1 1 1 - 1 3 4 . Rauch, J.E. (1993), Productivity gains from geographic concentration of h u m a n capital: Evidence from the cities. Journal of Urban Economics 34(2): 3 8 0 - 4 0 0 . Schiff, M . (2006), Brain gain: claims about its size and impact on welfare and growth are greatly exaggerated. Pp. 2 0 1 - 2 2 6 in: Ç. Özden and M . Schiff (eds.), International migration, remittances and the brain drain. N e w York: Palgrave Macmillan and World Bank, Chapter 6. Schultz, T.P. (1988), Education investments and returns. Pp. 5 4 3 - 6 3 0 in: H . Chenery, T.N. Srinivasan (eds.), H a n d b o o k of development economics vol. I, Amsterdam: N o r t h - H o l l a n d Publishing, Chapter 13. Stark, O., C. Helmenstein, A. Prskawetz (1998), H u m a n capital depletion, h u m a n capital formation, an migration: a blessing or a curse? Economics Letters 60(3): 3 6 3 - 3 6 7 . Stark, O., Y. Wang (2002), Inducing h u m a n capital formation: Migration as a substitute for subsidies. Journal of Public Economics 86(2): 29—46. Siidekum, J. (2006), H u m a n capital externalities and growth of high- and low-skilled jobs. IZA Discussion Paper N o . 1969. Siidekum, J. (2008), Convergence of the skill composition across German regions. Regional Science and Urban Economics 38(2): 1 4 8 - 1 5 9 . Wössmann, L. (2003), Returns to education in Europe (Book Review Essay). Review of World Economics (Weltwirtschaftliches Archiv) 139(2): 3 4 8 - 3 7 6 . Prof. Dr. H a r t m u t Egger, University of Bayreuth, Universitätsstrasse 30, 9 5 4 4 7 Bayreuth, Germany. E-Mail: [email protected] Prof. Ph.D. Gabriel Felbermayr, Universität Hohenheim, Box 520E, 70593 Stuttgart, Germany. E-Mail: [email protected]

Jahrbücher f. Nationalökonomie u. Statistik (Lucius & Lucius, Stuttgart 2009) Bd. (Vol.) 229/6

The Purpose of Remittances: Evidence from Germany By Thomas K. Bauer, Essen, and Mathias G. Sinning, Canberra* JEL F22, D12, D91 International migration, savings, remittances.

Summary This paper examines the purpose of remittances using individual data of migrants in Germany. Particular attention is paid to migrants' savings and transfers to family members in the home country. Our findings indicate that migrants who intend to stay in Germany only temporarily have a higher propensity to save and save larger amounts in their home country than permanent migrants. A similar picture emerges when considering migrants' payments to family members abroad. The results of a decomposition analysis indicate that temporary and permanent migrants seem to have different preferences towards sending transfers abroad, while economic characteristics and the composition of households in home and host countries are less relevant.

1

Introduction

According to the World Bank (2008), recorded remittances to developing countries are estimated to reach $US 283 billion worldwide in 2008, growing by 6 . 7 % relative to 2007. Even though this number is regarded to represent a lower bound because of unrecorded remittance flows, therewith remittances are more than twice as large as total development aid and represent the largest source of foreign exchange for many countries. The growing importance of remittances for many developing countries generates a strong interest of policy makers and scientists on the motivation of migrants to transfer financial resources to their home countries and the impact of these resources on the home countries' economy. Policy makers are in particular interested in potential measures to channel these remittances towards a productive use. Yet, the scientific literature on remittances has largely focused on the motives of migrants to remit (see, among others, Lucas/Stark 1985, Bernheim et al. 1985, Cox 1987, CoxJ Rank 1992, Ilahi/Jafarey 1999, Amuedo-Dorantes/Pozo 2006). A substantive number of theoretical papers has identified a variety of motives that may induce migrants to send remittances to their countries of origin. The main empirical challenge of this line of the literature lies in the differentiation of the explanatory power of different theories, since many of them come to similar testable hypotheses (Rapoport/Docquier 2006 provide an extensive recent survey of this literature). M u c h less research exists on the impact of remittances on the individuals or families and the overall economy of the countries w h o receive the remittances. The focus of interest of * We thank Julia Bredtmann, Sebastian Otten and Matthias Vorell for helpful comments on an earlier draft of the paper. Financial support of the German Israeli Foundation (G.I.F. Research Grant N o . G - 9 8 9 - 2 3 2 . 4 / 2 0 0 7 ) is gratefully acknowledged.

The Purpose of Remittances: Evidence from Germany · 731

this strand of literature concerns the use of remittances, in particular whether they are used for investments or consumption. If remittances are mainly used for investments, they may foster growth and the development of the economies receiving the remittances. However, remittances may be used predominantly for consumption purposes, reducing the incentives to work and slowing down the development of the receiving economy. Microeconomic studies on the use of remittances are predominantly descriptive, relying mostly on individual and household surveys collected in the countries that receive remittances. The majority of studies on the use of remittances, however, is based on growth theories and use macroeconomic data to analyze the effects of remittances (see Rapoport/ Docquier 2006). Using individual data of foreign-born staying in Germany, this paper combines both strands of this literature by analyzing the extent and the determinants of remittances focusing in particular on the intended purpose of these transfers as well as the role of return intentions. As a major immigration country, Germany represents an excellent example for the analysis of remittances. Even though the majority of migrants in Germany does not originate from developing countries, they remit a substantial part of their income. In 2004, remittance flows from Germany amounted to about S US 10.4 billion (World Bank 2006). Our paper contributes to the existing literature in several respects. First, while most of the studies on remittances concentrate on migrants' transfers to developing countries using predominantly data collected in the country of origin of a migrant, the analysis in this paper focuses on remittances of migrants from traditional labor-exporting countries, such as Turkey, Italy, and Greece as well as refugees originating from former Yugoslavia using data obtained in Germany rather than in the countries of origin. Second, utilizing information on migrants' return intentions, the differences in the extent and the purpose of financial transfers to the home country between temporary and permanent migrants are analyzed. 1 Finally, different to most of the existing literature, the analysis does not only concentrate on the extent of financial transfers to the home country and its determinants. The data used in our analysis rather allows us to differentiate between purposes of these transfers, i.e. whether they constitute savings in the home country or payments to family members. Existing studies on the use of remittances typically differentiate whether remittances are used for consumption, social purposes such as weddings or funerals, housing, debt repayment, or investments, i.e. investments in the human capital of family members or the establishment of a firm. In this context, the use of remittances for consumption or social purposes, for example, is often classified as being non-productive. However, in many cases even the consumptive use of remittances could be seen as being productive, if, for example, better food or the acquisition of air-conditioning increases the productivity of the household members. Although the information on the purpose of remittances in our data is rather limited in comparison to surveys available in many emigration countries, important insights may be gained from the analysis of migrants' savings, which foster investments in the countries of origin.

1

Since economic theory suggests that the savings behavior depends on expectations about the future economic situation, return intentions may be considered as a strong predictor of migrants' savings. At the same time, it seems likely that migrants who intend to return have strong incentives to send remittances to their country of origin.

732 · T.K. Bauerand M.G. Sinning

The paper proceeds as follows. In the next section we describe the data used in the empirical analysis and discuss the empirical strategy. Section 3 provides a detailed discussion of our results. Section 4 concludes. 2

Data and empirical strategy

The following analysis of the extent and purpose of remittances is based on the German Socio-Economic Panel (SOEP), a representative longitudinal study that includes German and immigrant households and started in 1984. The SOEP contains information about socioeconomic and demographic characteristics, household composition, occupational biographies etc. The following analysis is restricted to immigrant workers aged between 18 and 65 years, where immigrants are defined as foreign-born persons who immigrated to Germany since 1948. Due to the small number of observations, the sample does not include ethnic migrants from Central and Eastern Europe who received German citizenship after immigration. Furthermore, since less than five percent of the migrant population live in East Germany, the analysis concentrates on immigrants residing in West Germany. Finally, since lag variables have to be generated for some of the explanatory variables of our model, the year 1984 may not be considered in the empirical analysis. For the period between 1985 and 1995, the SOEP provides detailed information about transfers of foreigners to their home country. Immigrants were asked whether they sent any financial transfers to their home country. Additionally, the amount of transfers for two different purposes of the remittances is observed, i.e. savings and support for the family. After 1995, only the amount of transfers to persons abroad is available. Hence, given the particular research question of this paper, the sample is limited to the period from 1985 to 1995. We follow the existing empirical literature on remittances and apply a binary Probit model to investigate the effects of relevant determinants on the propensity to transfer money to the home country and use a Tobit model to account for the censored nature of the outcome variable when investigating the amount of these transfers (Merkle/Zimmermann 1992, Rodriguez 1996, Cox et al. 1987, Bauer/Sinning 2009). In order to provide a comprehensive descriptive analysis of behavioral differences between temporary and permanent migrants, particular attention is paid to the isolation of the part of the differences in remittances that can be explained by differences in socioeconomic characteristics from the part attributable to differences in coefficients, using the decomposition method proposed by Blinder (1973) and Oaxaca (1973). To perform this decomposition, the empirical models are estimated separately for temporary (t) and permanent (p) migrants. For the linear regression models of groups g = (t, p), yg* = x'gyg +

(1)

Blinder (1973) and Oaxaca (1973) propose the decomposition·,2 y* - yP* = {xt - χ ρ ) + Xp(% -%)•

2

(2)

Note that an alternative decomposition exists for equation (2). The choice of the decomposition equation, however, did not affect the results of the empirical analysis qualitatively. Consequently, the estimates of the alternative decomposition are not presented in this paper. They are available from the authors upon request.

The Purpose of Remittances: Evidence from Germany · 733

Bauer and Sinning (2008) show that a decomposition of the outcome variable similar to equation (2) is not appropriate for nonlinear regression models, because the conditional expectations E(yg\xg) may differ from x'gyg· They propose to decompose the mean difference of y using conditional expectations evaluated at different coefficient estimates, i.e. Atp = Eh(y,\x,) - Eïp{yp\xp) = \Ey,(yt\xt) - Eyt(yp\xp)}

(3) + [Eyt(yp\xp) -

Eyp{yp\xp)}.

To apply this decomposition to different nonlinear models, one has to estimate the sample counterparts S(yg, xg) and S(jh, xg) of the conditional expectations E-/K(yg\xg) and Eyh {yg\xg) for g, h = (í, p) and g φ h. In the empirical analysis, the decomposition results of Probit and Tobit models will be reported. 3 We differentiate between temporary and permanent migrants using information on return intentions, with temporary migrants being defined as the group of migrants who intend to return to their country of origin, while permanent migrants are considered as those who intend to stay in Germany forever. The set of explanatory variables considered include socioeconomic and demographic characteristics such as age, gender, years of education, and current income. Economic theory suggests that wealth accumulation depends on permanent rather than current income. Therefore, we follow Blau and Graham (1990) and add a measure of the predicted current income as a proxy variable for permanent income to our specification. It can further be expected that migrants are more likely to send transfers abroad if they face higher income risks in their host country (Amuedo-Dorantes/Pozo 2006). To account for this possibility, we use the standard deviation of the average net income in the last five years as a proxy for income risk. Following the analysis of Lucas and Stark (1985), a number of empirical studies have shown that marital status as well as household size and household composition in the migrants' home and host country are decisive determinants of remittances (see, among others, Hoddinott 1994, de la Briere et al. 2002). Unfortunately, the SOEP does not provide information about the household size of migrants in their home country. Apart from the marital status and the household size in Germany, the set of regressors includes dummy variables indicating whether the spouse or children of the respondent live abroad. Using data on El Salvador and Nicaragua, Funkhouser (1995) has shown that the selfselection of migrants has an important influence on remittances. Other studies indicate that the savings behavior may also be affected by cultural background (Carroll et al. 1994, 1999). Based on these results, we include source country indicators to control for variations in the remittance behavior across countries of origin. Finally, since migrants' remittances typically decline as the duration of residence in the host country increases (DeVoretz/Vadean 2005) and savings of more established immigrants in their host countries tend to be higher than those of more recent immigrant cohorts (Bauer et al. 2009), differences between immigration cohorts are taken into account by controlling for the number of years since migration. Descriptive statistics and a detailed description of the definitions of all variables used in the analysis are given in Tables A l and A2 of the Appendix. After excluding all observations with missing values on one of the variables used, the data set contains 7,976 person-year-observations of 1,535 individuals.

3

A detailed discussion of the decomposition for these models is provided by Bauer and Sinning (2008).

7 3 4 · T.K. B a u e r a n d M . G . Sinning

Table 1 Savings and remittances: Descriptive Statistics Temporary

Proportion of migrants saving abroad

migrants

Permanent migrants

0.08 0.27) 19.34 (115.00) 237.69 (332.92) 0.34

0.03 0.16) 4.53 ( 48.63) 178.78 (250.58) 0.27

(

(

( Savings abroad (in € per month) Savings abroad if > 0 (in € per month) Proportion of migrants sending remittances to family members abroad Payments to family members (in € per month) Payments to family members if > 0 (in € per month) Observations

0.47) 55.08 (120.52) 160.71 (159.41) 5,195

(

0.45) 44.96 (120.67) 165.01 (183.50) 2,781

Table 1 provides some descriptive statistics on the remittance behavior of temporary and permanent migrants. It shows that temporary migrants have a higher propensity to transfer money to their home country than permanent migrants and that they transfer higher amounts. Specifically, 8 % of the temporary migrants report savings in their home country compared to 3 % of the permanent migrants. Conditional on saving abroad, temporary migrants save almost 238 € per month or 21.9 % of their income while permanent migrants save only 179 € or 16.8 % of their income. The picture appears to be different when considering transfers to family members abroad. About one third of temporary migrants and about 27 % of the permanent migrants send remittances to family members abroad. Conditional on sending money to their family members, temporary and permanent migrants send almost a similar amount. 3

Estimation results

To investigate differences in the propensity to save and the amount of savings between temporary and permanent migrants, Probit and Tobit regressions were estimated using a pooled sample of the two groups. The estimates of these models - which are available upon request - indicate that the probability to accumulate wealth in the home country is 4.4 % higher for migrants who intend to return to their home country in comparison to those who decide to stay in Germany permanently. Temporary migrants further save about 10 % more than permanent migrants. Both results are in accordance with theoretical models on the impact of return intentions and the intended duration of stay on migrants' savings behavior (Galor/Stark 1990, Dustmann 1997). A similar picture emerges when considering payments to family members in the home country: temporary migrants have a higher probability (7.6 %) to support their family members abroad and transfer more money (15.9 %) than permanent migrants. Panels A and Β of Table 2 show the estimation results of the Probit and Tobit model for savings in the home country separately for temporary and permanent migrants. The results of the Probit estimations indicate significant differences in the effects of the determinants of savings in the home country between the two groups. While the savings

The Purpose of Remittances: Evidence from Germany · 735

Table 2 Savings Abroad: Temporary vs. Permanent Migrants - 1985-1991, 1993, 1995

Marginal effect Age Female Current income x10 2 Variation in past income x10 2 Permanent income x10 2 Years of education Household size Married Spouse lives abroad(t_ V) Children live a b r o a d ^ ) Years since migration Country of origin: Turkey (Reference group) Country of origin: Italy Country of origin: Greece Country of origin: Ex-Yugoslavia Country of origin: Other Observations

0.001*** 0.001 0.006*** -0.007* 0.010 - 0.005 - 0.001 0.016 -0.013 0.030** - 0.001

0.001 0.026 0.001 0.003 0.008 0.003 0.002 0.012 0.018 0.014 0.001

- 0.002 -0.010 0.003 0.018

0.011 0.012 0.011 0.015

Marginal effect Age Female Current income x10 2 Variation in past income x10 2 Permanent income x10 2 Years of education Household size Married Spouse lives abroad(t_t) Children live abroad(t_i) Years since migration Country of origin: Turkey (Reference group) Country of origin: Italy Country of origin: Greece Country of origin: Ex-Yugoslavia Country of origin: Other Observations

A. Temporary Migrants Probit Tobit Standard Marginal Standard error effect error 0.379 - 2.566 1.202* -0.191 1.959 - 1.185 - 0.789 9.031* - 0.272 8.836 0.611 - 5.712 - 11.413*** - 5.751 - 0.414 5,195

0.300 10.912 0.731 1.012 3.824 1.439 1.215 5.087 8.925 7.332 0.488 5.216 3.785 4.152 6.473

B. Permanent Migrants Probit Tobit Standard Marginal Standard error effect error

- 0.001 -0.019 0.001 - 0.001 0.002 - 0.001 -0.003** 0.011 -0.016 0.007 - 0.001

0.001 0.015 0.001 0.002 0.006 0.002 0.001 0.006 0.006 0.012 0.001

- 0.051 -0.916 0.205 - 0.345 1.321* -0.541* - 0.574* - 0.729 -2.329*** 0.384 0.027

0.071 2.136 0.177 0.331 0.766 0.318 0.313 1.677 0.679 1.521 0.115

-0.012* - 0.001 0.003 0.008

0.005 0.009 0.007 0.010

- 1.320 - 1.283 -0.264 1.845

1.087 0.946 1.037 2.702

2,781

Notes: Weighted estimates based on weights provided by the SOEP. Standard errors are adjusted in order to take repeated observations of households into account. The regression further includes year dummies. * significant at 10%-level; ** significant at 5 %-level; ** significant at 1 %-level.

736 · T.K. Bauerand M.G. Sinning

propensity of temporary migrants increases significantly with age and current income, these factors have no significant influence on the savings propensity of permanent migrants. Variation in past income has a negative impact on savings of temporary migrants, with the coefficient being significant at the 10 %-level. This result may indicate that a higher income uncertainty leads migrants to hold their savings in the host country in order to have quick access to savings in the case of an income reduction. This would also imply that these migrants hold predominantly liquid assets. The data set, however, does not provide the information necessary to test this hypothesis. Finally, the propensity of temporary migrants to save aborad is increasing when children are living in the country of origin. For permanent migrants, the propensity to save abroad is only affected significantly by the household size, with savings abroad decreasing with the household size in the host country. The Tobit estimates indicate that among temporary migrants the amount of savings is increasing with current income. Furthermore, compared to unmarried temporary migrants, migrants being married save more abroad. Note, however, that both coefficients are statistically significant at a 10 %-level only. For permanent migrants the amount of saving is decreasing with education and the household size and increasing with higher permanent income. Again, these coefficients are statistically significant at a 10 %-level only. Finally, permanent migrants save significantly less when the spouse resides in the home country. In sum, both the economic situation and the composition of the household in home and host countries seem to have a considerable though heterogenous impact on savings of temporary and permanent migrants. At the same time, the explanatory power of our empirical models is rather low, with only a few coefficients being significant at conventional levels. This result is not surprising, given the relatively low proportion of migrants saving abroad (see Table 1). The overall picture changes substantially when considering transfers to family members abroad. The Probit estimates reported in Table 3 suggest that temporary and permanent migrants are much more similar concerning these transfers if compared to saving abroad. For both groups the propensity to transfer to family members abroad is increasing with age, current income and when the spouse or children are living in the home country, while it is decreasing with years since migration and household size and it is lower for females. Note that all estimated effects are in line with standard theories on the determinants of remittances (see Rapoport/Docquier 2006). A similar pattern may also be observed when analyzing the amount of transfers to family members. The Tobit estimates shown in Table 3 indicate that the amount of transfers is increasing with age and are higher when children live abroad for both groups of migrants. For temporary migrants the amount of savings is decreasing with household size and variation in past income and increasing with current income. Different to permanent migrants, temporary migrants whose spouse lives abroad also transfer significantly more money. Note that female permanent migrants transfer a significantly lower amount, while this coefficient is insignificant for temporary migrants. Finally, source country effects on the propensity to remit and the amount of remittances are similar for temporary and permanent migrants. Table 4 reports the results of our decomposition analysis. The estimates of the Probit decomposition provide evidence for significant differences in the propensity to save or remit between temporary and permanent migrants. The major part of the gap in the propensity to save (almost 90 %) may be attributed to different coefficients, suggest-

The Purpose of Remittances: Evidence from Germany · 737

Table 3 Payments to Family Members Abroad: Temporary vs. Permanent Migrants - 1 9 8 5 1991, 1993, 1995

Marginal effect Age Female Current income x10 2 Variation in past income x10 2 Permanent income x10 2 Years of education Household size Married Spouse lives a b r o a d ^ ) Children live abroad (t _i) Years since migration Country of origin: Turkey (Reference group) Country of origin: Italy Country of origin: Greece Country of origin: Ex-Yugoslavia Country of origin: Other Observations

0.006*** -0.123** 0.013*** - 0.007 - 0.003 - 0.006 -0.032*** - 0.004 0.325*** 0.336*** - 0.005***

0.000 0.045 0.002 0.004 0.016 0.005 0.004 0.023 0.040 0.023 0.001

1.729*** 1.663 3.637*** - 3.701 * 5.765 - 0.880 - 7.829*** - 8.874 46.635** 63.091*** - 1.212

0.450 20.972 1.319 1.958 7.106 2.322 2.064 9.186 18.250 11.323 0.941

-0.193*** 0.009 0.056*** -0.129***

0.016 0.022 0.020 0.020

- 16.578 - 0.962 12.560* -25.968*** 5,195

10.829 9.816 7.420 6.224

Marginal effect Age Female Current income x10 2 Variation in past income x10 2 Permanent income x10 2 Years of education Household size Married Spouse lives abroad (t _i) Children live abroad(t_i) Years since migration Country of origin: Turkey (Reference group) Country of origin: Italy Country of origin: Greece Country of origin: Ex-Yugoslavia Country of origin: Other Observations Notes:} See Notes to Table 2.

A. Temporary Migrants Probit Tobit Standard Marginal Standard error effect error

B. Permanent Migrants Probit Tobit Standard Marginal Standard error effect error

0.006*** -0.144*** 0.011*** -0.012 - 0.026 0.001 -0.014*** 0.032 0.059 0.281*** - 0.005***

0.000 0.047 0.003 0.008 0.020 0.006 0.004 0.023 0.051 0.042 0.001

1.604*** - 32.424* - 1.457 2.727 -2.018 0.772 - 2.079 - 7.703 17.848 56.120*** - 0.672

0.525 18.595 2.689 4.021 8.330 3.592 2.940 11.935 18.054 14.712 0.778

-0.168*** - 0.059** 0.034* -0.157***

0.015 0.022 0.021 0.014

- 34.207*** - 16.047** 12.155 -31.796*** 2,781

6.715 7.544 9.957 4.703

738 · T.K. Bauer and M.G. Sinning

Table 4 Decomposition Analysis - 1985-1995 2Probit Explained Part Unexplained Part 2"Tobit Explained Part Unexplained Part

N, Np

Savings

Remittances

0.055*** [0.007] 0.006** [0.003] (11.5) 0.049*** [0.007] (88.5)

0.069*** [0.016] 0.021** [0.007] (30.9) 0.047*** [0.015] (69.1)

16.739*** [2.497] 12.631 [12.547] (75.5) 4.108 [11.396] (24.5) 5,195 2,781

9.282 [6.416] - 4.686 [8.756] ( - 50.4) 13.968** [6.722] (150.4) 5,195 2,781

Notes: Bootstrapped (50 replications) standard errors in brackets. Percentages of the raw differential are reported in parentheses. * significant at 10 %-level; ** significant at 5 %-level; *** significant at 1 %-level.

ing that the gap is a result of behavioral differences between temporary and permanent migrants rather than differences in observable characteristics between the two groups. A similar part of the gap in the propensity to remit (almost 70 % ) is explained by different coefficients, indicating that temporary and permanent migrants seem to have different preferences towards sending transfers to their country of origin, while economic characteristics and the composition of households in the home and host countries play a relatively minor role. At the same time, about~30 % of the overall gap in the propensity to remit may be explained by observed characteristics, suggesting that different economic restrictions or a different household compositions may at least explain a part of the differences in the propensity to remit. The estimates of the Tobit decomposition reveal that temporary migrants save significantly more (about 17 € per month) than permanent migrants, while the raw gap in the amount of remittances is insignificant. Although the major part of the savings gap between the two groups (about 75 % ) may be attributed to differences in observed characteristics, this part is not significant. This result is in line with the heterogenous effects and the low explanatory power of the Tobit estimates reported in Table 2. Since differences in the amount of remittances are relatively small (less than 10 € per month), we observe that a very large share of the gap (about 150 % ) is unexplained. This result may be explained by relatively large variations in the Tobit estimates presented in Table 3 and small differences in the observed characteristics between the two groups (see Appendix Table A l ) .

The Purpose of Remittances: Evidence from Germany · 739

4

Conclusions

This paper examines the extent and the determinants of remittances, using individual data of foreign-born staying in Germany. We differentiate between temporary and permanent migrants using information on return intentions, with temporary migrants being defined as the group of migrants who intend to return to their country of origin, while permanent migrants are considered as those who intend to stay in Germany forever. While most of the studies on remittances concentrate on migrants' transfers to developing countries using predominantly data collected in the country of origin of a migrant, the analysis of this paper focuses on remittances of migrants from traditional labor-exporting countries, such as Turkey, Italy and Greece as well as refugees originating from former Yugoslavia using data obtained in Germany rather than in the countries of origin. In order to provide a comprehensive descriptive analysis of behavioral differences between temporary and permanent migrants, particular attention is paid to the isolation of the part of the differences in remittances that can be explained by differences in socioeconomic characteristics from the part attributable to differences in coefficients using the decomposition method proposed by Blinder (1973) and Oaxaca (1973). Our findings indicate that the probability to accumulate wealth in the home country is higher for temporary migrants than for permanent migrants. Temporary migrants further save about 10% more than permanent migrants. A similar picture emerges when considering migrants' payments to family members in the home country. The results of a Probit decomposition suggest that the major part of the gap in the propensity to save or remit may be attributed to different coefficients, indicating that temporary and permanent migrants seem to have different preferences towards sending transfers to their country of origin, while economic characteristics and the composition of households in home and host countries play a relatively minor role. Due to the small unconditional gap in the amount of savings and remittances, evidence derived from a Tobit decomposition is rather mixed.

740 · T.K. Bauerand M.G. Sinning

Appendix Table A1 Descriptive statistics, 1985-1995

Variable Socioeconomic characteristics Age Female Current income Variation in past income streams Permanent income Years of education Household composition Household size Married Spouse lives abroad Children live abroad Migration background Years since migration Country of origin: Turkey Country of origin: Italy Country of origin: Greece Country of origin: Ex-Yugoslavia Country of origin: Other Ν

Temporary migrants Standard Mean Deviation

Permanent migrants Standard Mean Deviation

41.2 0.281 864.19 116.96 851.51 9.3

10.3 0.450 349.21 149.92 235.92 1.8

40.5 0.275 910.07 127.82 903.69 9.6

10.2 0.446 356.45 139.93 244.67 2.0

3.5 0.778 0.077 0.119

1.6 0.415 0.267 0.324

3.5 0.723 0.080 0.071

1.6 0.448 0.271 0.256

19.134 0.424 0.195 0.110 0.212 0.060 5,195

5.648 0.494 0.396 0.313 0.408 0.238

20.057 0.409 0.197 0.065 0.276 0.052 2,781

6.320 0.492 0.398 0.247 0.447 0.223

The Purpose of Remittances: Evidence from Germany · 741

Table A2 Definition of variables Variable Savings and remittances Savings abroad

Payments to family members abroad Socioeconomic characteristics Age Female Current income Variation in past income streams Permanent income Years of education Household composition Household size Married Spouse lives abroad Children live abroad Migration background Intended return migration Years since migration Country of origin: Turkey Country of origin: Italy Country of origin: Greece Country of origin: Ex-Yugoslavia Country of origin: Other

Description

Real average monthly amount of savings abroad in € (base year 2000). Real average monthly amount of payments to family members abroad in € (base year 2000). Age of respondent in years. 1 if respondent is female; 0 otherwise. Net real income last month in € (base year 2000). Standard deviation of current net income over the last 5 years. Estimated real permanent income in € (base year 2000). Education of respondent in years. Number of persons in household. 1 if respondent is married; 0 otherwise. 1 if spouse of respondent lives abroad; 0 otherwise. 1 if children of respondent live abroad; 0 otherwise. 1 if respondent intends to return to the home country, 0 otherwise. Duration of German residence in years. 1 if respondent originates from Turkey; 0 otherwise. 1 if respondent originates from Italy; 0 otherwise. 1 if respondent originates from Greece; 0 otherwise. 1 if respondent originates from former Yugoslavia; 0 otherwise. 1 if respondent originates from other OECD member country (reference category); 0 otherwise.

References Amuedo-Dorantes, C., S. Pozo (2006), Remittances as Insurance: Evidence from Mexican Immigrants. Journal of Population Economics 19: 227-254. Bauer, T.K., M.G. Sinning (2008), An Extension of the Blinder-Oaxaca Decomposition to NonLinear Models. Advances in Statistical Analysis 92: 197-206. Bauer, T.K., M.G. Sinning (2009), The Savings Behavior of Temporary and Permanent Migrants in Germany. Journal of Population Economics, forthcoming. Bauer, T.K., D.A. Cobb-Clark, V.A. Hildebrand, M.G. Sinning (2009), A Comparative Analysis of the Nativity Wealth Gap. Economic Inquiry, forthcoming. Bernheim, B.D., A. Shleifer, L.H. Summers (1985), The Strategic Bequest Motive. Journal of Political Economy 93: 1045-1076. Blau, F.D., J.W. Graham (1990), Black-White Differences in Wealth and Asset Composition. The Quarterly Journal of Economics 105: 321-339. Blinder, A.S. (1973), Discrimination: Reduced Form and Structural Estimates. Journal of Human Resources 8: 436-455. Carroll, C.D., B. Rhee, C. Rhee (1994), Are There Cultural Effects on Saving? Some Cross-Sectional Evidence. Quarterly Journal of Economics 109: 685-699. Carroll, C.D., B. Rhee, C. Rhee (1999), Does Cultural Origin Affect Saving Behavior? Evidence from Immigrants. Economic Development 48: 33-50. Cox, D. (1987), Motives for Private Transfers. Journal of Political Economy 95: 508-546. Cox, D., Z. Eser, E. Jimenez (1987), Motives for Private Transfers Over the Life Cycle: An Analytical Framework and Evidence for Peru. Journal of Development Economics 55: 57-80.

742 • T.K. Bauer and M . G . Sinning

Cox, D., M.R. Rank (1992), Inter-Vivos Transfers and Intergenerational Exchange. The Review of Economics and Statistics 74: 305-314. de la Briere, B.A., S.L. de Janvry, E. Sadoulet (2002), The Roles of Destination, Gender, and Household Composition in Explaining Remittances: An Analysis for the Dominican Sierra. Journal of Development Economics 68: 309-328. DeVoretz, D.J., F.P. Vadean (2005), A Model of Foreign-Born Transfers: Evidence from Canadian Micro Data. IZA Discussion Papers No. 1714: 1-34. Durand, J., W. Kandel, E.A. Parrado, D.S. Massey (1996), International Migration and Development in Mexican Communities. Demography 33: 249-264. Dustmann, C. (1997), Return Migration, Uncertainty and Precautionary Savings. Journal of Development Economics 52: 295-316. Funkhouser, E. (1995), Remittances from International Migration: A Comparison of El Salvador and Nicaragua. Review of Economics and Statistics 77: 137-146. Galor, O., O. Stark (1990), Migrants' Savings, the Probability of Return Migration and Migrants' Performance. International Economic Review 31: 463-467. Hodinott, J. (1994), A Model of Migration and Remittances Applied to Western Kenya. Oxford Economic Papers No. 46: 459-476. Ilahi, N., S. Jafarey (1999), Guestworker Migration, Remittances and the Extended Family: Evidence from Pakistan. Journal of Development Economics 58: 485-512. Lucas, R.E.B., O. Stark (1985), Motivations to Remit: Evidence from Botswana. The Journal of Political Economy 93: 901-918. Merkle, L., K.F. Zimmermann (1992), Savings, Remittances, and Return Migration. Economics Letters 38: 129-134. Oaxaca, R.L. (1973), Male-Female Wage Differentials in Urban Labor Markets. International Economic Review 14: 693-709. Rapoport, H., F. Docquier (2006), The Economics of Migrants' Remittances. Pp. 1135-1198 in: S. Kolm, J.M. Ythier (eds.), Handbook of the Economics of Giving, Altruism and Reciprocity Vol. 2, Elsevier. Rodriguez, E.R. (1996), International Migrants' Remittances in the Philippines. The Canadian Journal of Economics 29: 427-432. World Bank (2006), Global Economic Prospects 2006: Economic Implications of Remittances and Migration. World Bank. World Bank (2008), Outlook for Remittance Flows 2008-2010: Growth Expected to Moderate Significantly, but Flows to Remain Resilient. Migration and Development Brief 8. World Bank. Prof. Dr. Thomas Bauer, RWI, Hohenzollernstr. 1-3, 45128 Essen, Germany, Ruhr-University Bochum, and IZA Bonn. E-Mail: [email protected] Dr. Mathias Sinning, Social Policy Evaluation, Analysis and Research Centre (SPEAR), Research School of Social Sciences (RSSS), Australian National University, Fellows Road, Coombs Building (Building 9), Canberra ACT 0200, Australia, RWI, and IZA Bonn. E-Mail: [email protected]

Jahrbücher f. Nationalökonomie u. Statistik (Lucius & Lucius, Stuttgart 2009) Bd. (Vol.) 229/6

Worker Remittances and Growth: The Physical and Human Capital Channels By Thomas H.W. Ziesemer, Maastricht* JEL 015, J61 Remittances, growth.

Summary Remittances may have an impact on economic growth through channels to physical and human capital. We estimate an open economy model of these two channels consisting of seven equations using the general method of moments with heteroscedasticity and autocorrelation correction (GMM-HAC) with pooled data for four different samples of countries receiving remittances in 2003. The countries with per capita income below $ 1200 benefit most from remittances in the long run because they have the largest impact of remittances on savings. Their changes in remittances account for about 2 % of the steady-state level of GDP per capita when compared to the counterfactual of having no changes of remittances. Their ratio of the steadystate growth rates with and without changes of remittances is 1.39. Transitional gains are higher than the steady-state gains only for the human capital variables of this sample. As savings react much more strongly than investment an important benefit of remittances is that less debt is incurred and less debt service is paid than without remittances. All effects are much weaker for the richer countries.

1

Introduction

Besides foreign aid, trade and debt, remittances of former migrants have become a source of increasing amounts of foreign exchange for their country of origin. Most countries' remittances have remained below 10 % of GDP. For some countries as Jordan and Yemen they are structurally as high as 2 0 % . The highest values of more than 60 % are observed in years of serious trouble like Lebanon in the end of the 1980s, and more than 2 0 % in Albania, Cape Verde in the beginning of the 1990s, Bosnia in the end of the 1990s, and Haiti in the beginning of this millennium. One of the interesting related questions is how this affects growth. There are some macroeconomic papers on this question so far. 1 Glytsos (2005) estimates the impact of remittances on consumption, investment, output and imports for five countries in a traditional, dynamic Keynesian model 2 . He finds long term

* I am grateful to Femke Kramer for drawing my attention to the issue, to Pierre Mohnen for a very helpful discussion and two anonymous referees for useful comments. 1 For literature based on micro studies, see Rapoport and Docquier (2006) and Adams (2006). The results of the more recent household panels are very similar to those obtained here. However, a macroeconomic approach can work towards an analysis of the effects on levels and growth rates of the GDP per capita. 2 There is no price mechanism, no technology or resource constraint. Lagged dependent variables make the model dynamic.

744 · T.H.W. Ziesemer

multipliers of (on average) 2.3 for income (and .6 for investment). T h e paper is rich in discussing the related ups and d o w n s of remittances and other variables, but it does not consider the impact of remittances on h u m a n capital. C h a m i et al. (2003) have argued that remittances provide an incentive to reduce effort and thereby m a k e weak economic p e r f o r m a n c e s more likely. They find negative impacts of remittances on g r o w t h . Catrinescu et al. (2009) extend their a p p r o a c h t o include policy and institutional variables and run the estimation for a dynamic panel. They find some significantly positive results for the impact of remittances on g r o w t h , but these are not very robust. In these g r o w t h regressions, remittances and investment a p p e a r as right-hand side variable, which might cause collinearity and therefore a w r o n g sign. C h a m i et al. (2003) d o not discuss this possibility. Catrinescu et al. point out that there are endogeneity problems. T h e r e f o r e we think it is the best not to have remittances in the equations for g r o w t h or investment but rather in that of savings, which depend on disposable income of which transfers like remittances and aid are a crucial part. Solimano (2003) has included remittances in a time-series g r o w t h regression for C o l o m b i a a n d Ecuador finding a positive sign for both countries, which is insignificant t h o u g h for Ecuador. The regression does not include a labour g r o w t h variable. M u n d a c a (2005, 2009) adds remittances to a g r o w t h regression for some countries. It contains the s t a n d a r d deviation of G D P per capita g r o w t h a n d domestic bank credit as regressors. She finds a positive effect of remittances o n g r o w t h , which is higher t h a n w i t h o u t the credit variable. She interprets this as a higher impact of remittances in the presence of better financial development, because remittances are better channelled to their purposes. Giuliano and Ruiz-Arranz (2008) a d d remittances multiplied t o financial variables and find positive g r o w t h effects for financially less developed countries. This is the opposite of M u n d a c a ' s interpretation based on the idea that better financial development makes credit available and therefore remittances w o u l d have less of a n impact on investment a n d g r o w t h but rather replaces credit by o w n capital. However, the t w o papers have in c o m m o n t h a t the coefficients of remittances a n d financial indicators are positive. These latter four papers use g r o w t h regressions, which tell us w h e t h e r or not there is an effect on g r o w t h but not w h y and h o w it w o r k s (see Durlauf et al. 2005). Essentially, f r o m an economic point of view they are single equation models. They allow for the application of sophisticated econometric methods but d o not m a k e explicit the economic mechanisms which drive the g r o w t h . Besides d e m a n d , moral hazard and financial development as treated in these papers, the channels f r o m remittances to physical and h u m a n capital accumulation are certainly i m p o r t a n t a n d so are the economic mechanisms along which these w o r k . But these papers only contain the effect of the investments on g r o w t h but not f r o m remittances on the investments in h u m a n a n d physical capital. In order to get insights into the economic mechanism leading f r o m remittances t o g r o w t h via savings, interest rates, investments in physical a n d h u m a n capital, and the size of the effects we set up a simultaneous equation model. We are not afraid that one mis-specified equation contaminates the others 3 , because we use well-established equations, and get very plausible results. T h e model dealing with this consists of seven equations, six of which are available in the literature and slightly adjusted for our purpose. First, remittances as a share of G D P are explained by an equation similar to that of Chami et al. (2003) and others earlier 4 con-

3

4

This is the main reason why estimation of systems is less widely spread according to Akhand and Gupta (2002). However, in recent times more papers try this. See also El-Sakka and McNabb (1999) and the references there.

Worker Remittances and Growth: The Physical and Human Capital Channels · 745

taining the differences of income and interest rates in the host country and the country of origin. Second, remittances are added to an equation explaining the savings ratio similar to that in Loayza et al. (2000). Third, an increase in savings reduces the gap between investment and savings, which in turn reduces domestic interest rates as found by Obstfeld and Rogoff (2001). Fourth, an interest rate reduction has a positive impact on the investment/GDP ratio in a standard investment function. Fifth, enrolments in primary schooling are a non-linear function of their own past values and changes of development aid and, for poorer countries, the savings ratio. Sixth, a higher savings ratio (except for poor countries) together with higher enrolments in primary education leads to higher literacy five years later. 5 By implication, the concept is that remittances have an impact on human capital via savings as in the theory of Cinar and Docquier (2004) 6 and Bertoli (2006), with savings entering the enrolment equation for poorer countries and the literacy equation for less poor countries. 7 Seventh, thus enhanced investment shares and literacy may enhance transitional growth rates and the level of per capita income in standard growth equations related to the model by Mankiw et al. (1992) and linked to open economy situations by Barro et al. (1995) assuming that borrowing is proportional to physical capital. When world income growth is included as suggested by models with imported inputs the initial value of GDP per capita is insignificant and there are effects on the permanent growth rate. We concentrate on this latter case. We use data from the World Development Indicators for a sample with 96 countries, which had at least one dollar of remittances in 2003, and three sub-samples. We estimate the seven equations simultaneously for pooled data allowing for contemporaneous correlation between them. In both cases we use the General Method of Moments allowing for weak exogeneity. In section 2 we set up a model that explains our line of thought on how remittances have an impact on growth. In section 3 the data and the econometric method are explained. Section 4 explains the results of the estimates. In section 5 we calculate the direct, shortrun (similar to impact) effects of remittances on the endogenous variables for the permanent growth model. Section 6 presents the long run solution for the permanent growth model. Section 7 analyses stability and transitional gains. Section 8 summarizes and concludes. 2

The model

The starting point of the model is the equation explaining worker remittances as a percentage of GDP. This is formulated in equation (1). The first index of each coefficient indicates the number of the equation and the second that of the regressor. We drop time and country indices throughout. 5

6 7

Equations for literacy have been estimated by Akhand and Gupta (2002), Mazumdar (2005) and Verner (2005). See also the micro-evidence cited there. Chami et al. (2008) mention the impact of remittances on education only verbally but do not include it in their model. They conclude from household panel studies that effects of remittances on investment are not productive. This conclusion is actually static and partial because it ignores multiplier effects on investments (Taylor 1999) and effects from savings on investment (general equilibrium effects). Moreover, their model has a fixed capital stock excluding consideration of effects of remittances on investment. For migration effects other than remittances on education and health in panels of households, see McKenzie (2006).

746 · T.H.W. Ziesemer

wr/gdp = c n + c i 2 w r ( - l ) / g d p ( - l ) + c 13 log(OEC) + ci 4 (log(gdppc(-2))+ ci5 log(l + ri(-2)) + ci6 log(l + r i u s ( - l ) ) + e s t i m e + u j

(1)

Remittances, wr/gdp, are assumed to be driven by differences in the income per capita of the recipient and the sender. Therefore we include the income of the recipient country. The senders know their own current income. As most of the migrants go to the OECD countries we approximate their income by per capita income of the OECD, OEC. 8 The sender will have information on the recipient country only from data about earlier years because it takes about a year in many countries to make the data. An indicator of the recipients' income is therefore Gross Domestic Product per capita with two lags, gdppc(-2). The two income variables need not have the same coefficient because the OECD income is only a crude proxy that comes in because we use only one indicator for the host country of the senders. We do not use the Gross National Income as senders are more likely to receive information on GDP then those of GNI through the media. The sender might consider saving the amount of money rather than transferring it. Therefore we use the real interest rate of the USA, rius, as an indicator of these opportunity costs. On the other hand the sender might consider putting the money into a bank account in the recipient country. Therefore we also include the real interest rate of the recipient country, ri, with the same information lag as for the GDP per capita variable. Finally, remittances are assumed to depend on their own past value, a constant and a time trend, which will be dropped if insignificant. As real interest rates can be highly negative we add a value of 1 to it, before taking natural logarithms, because we use interest rates in their scientific notation, that is, 5 % is indicated by '.05'. Essentially equation (1) above is the one that appears also in Chami et al. (2003). 9 Using natural logs or not for the remittance variable does not matter for the results in this equation. Further below we will provide equations explaining the (growth of) GDP per capita and the dynamics of the interest rates. The US interest rate and the GDP per capita of the OECD will not be determined in the model. We add residuals, u, whose index is that of the equation. 10 The next step is to explain the impact of worker remittances on savings in equation (2). savgdp = C21 + c 2 2 savgdp(-l) + c 2 3 d(wr/gdp) + c 24 d(log(gdppc))+ c 25 log(l + ri(—1 )) + c 2 6 ((oda/gdp) - ( o d a ( - l ) / g d p ( - l ) ) ) + c 2 7 ((oda/gdp) - (oda(—l)/gdp(—1)))2 + u 2

8

9 10

(2)

Niimi and Özden (2006) provide some evidence that migration to Gulf countries does not yield different results than to the OECD in explaining remittances flows. Chami et al. (2003) use the real income of the USA instead of that of the OECD. Using other regressors leads to different endogeneity problems than those discussed below. They are discussed by Niimi and Özden (2006) in connection with a cross-country regression. For example, the income per capita of sender and destination countries used in this paper would also explain the number of migrants, which are a major determinant in their regression. The authors do not discuss the paper by Chami et al. (2003), which is closest to our approach. But they have interesting results in regard to the disaggregation with respect to education. Stock data on number of migrants are available only for 1990 and 2000 in Docquier and Marfouk (2006). Therefore we can't use them for our dynamic analysis. Since the work on this paper Docquier has made data for the stock of migrants in six OECD countries by 195 countries of origin, which can be found on the World Bank website.

Worker Remittances and Growth: The Physical and Human Capital Channels · 747

Basically, we assume that the savings ratio, savgdp, is driven by its own past value and, as in most of the literature (see Loayza et al. 2000, Table 1), by the growth of GDP per capita and by real interest rates. As disposable income is conceptually probably a better variable (see Bertoli 2006, eq. (6)) but also less available in terms of data we may add changes of worker remittances to the regression, which are part of disposable income but not part of GDP. The idea here is that higher disposable income and therefore remittances lead to a higher savings ratio as in models using the difference of consumption and a consumption minimum in the utility function when the country in question is close to that minimum. This is quite plausible here because remittances reduce poverty (see Adams/Page 2005). As an equation with a lagged dependent variable is similar to one on changes in savings here we take changes of remittances as a variable. Moreover, we add changes of official development aid and their squared term to the regression because aid might be significant (see Doucouliagos/Paldam 2006). If remittances enhance savings they should diminish the difference of investment and savings, which is the additional demand or flow variable of foreign debt. This should reduce interest rates as captured by equation (3). log(l + ri) = cai + caz log(l + r i ( - l ) ) + c 33 log(l + ri(-2))+

(3) 2

c 3 4 (invgdp(-l) - savgdp(-l)) + c 3 5 (invgdp(-l) - savgdp(-l)) + c 3 e d(log(OEC(-l))) + c 37 (log(l + rius(-l)) - log(l + rius(-2))) + u3 There are several possible rationales for this equation. First, Obstfeld and Rogoff (2001) have derived such a relation between the current account and the interest rate (without the other variables included here) from a two period model with transport costs. Second, in Bardhan (1967) and later publications on growth under capital movements by others one finds the assumption that large countries may have an impact on the world market interest rate and therefore on there own interest rate through a lower or higher stock of net debt per unit of GDP. If so, this should also hold for the flow of net debt. It is questionable here whether the countries involved have monopoly power. But they may have this as a group if their behaviour goes into the same direction. Third, it is plausible to relate domestic interest rates to the sum of LIBOR/EURIBOR or Prime Rate and a country specific spread or risk premium. Edwards (1984) has shown that they depend on the ratio of debt to GDP or GNI. This ratio is lower one period after investment net of savings less than the growth of the GDP. Banks and rating agencies then can verify that less new debt relative to GDP is incurred and may reduce spreads. Therefore we use the lagged variable of the current account deficit or investment minus savings. Moreover, Belloc and Gandolfo (2005) argue that this relation may be non-linear based on data analysis. Therefore we include a squared term of the investment-savings difference. Moreover, two lagged dependent variables, the change in the US interest rate, and the growth rate of the OECD are included. The change in the US interest rate will be highly insignificant in all but one of the estimates. But the growth rate of the OECD, which is highly correlated with the US interest rate, is significant. The reason probably is that it enhances exports and therefore less new debt has to be incurred leading to lower spreads, or alternatively an impact on the exchange rate. 11 11

For a growth model with spreads see Ziesemer (1998 or 1997: ch.8).

748 · T.H.W. Ziesemer

If remittances via enhanced savings and lower net debt demand reduce interest rates, the link to physical capital is gross fixed capital formation as a share of GDP, gfcfgdp. This is captured as in equation (4). log(gfcfgdp) = C4i + C42 log(gfcfgdp(-l)) + C43 log(l + r i ( - l ) ) + C44 log(l + r i u s ( - l ) ) + c 4 jd(log(gdppc(-l)))+ c 4 6 d(oda/gdp) + c 4 7 d((oda/gdp) 2 ) + u 4

(4)

Gross fixed capital formation as a share of GDP is assumed to depend on its own lagged value, interest rates and lagged growth rates as an indicator of the business cycle, expectations and the future need for investment. The lag in the interest rate variable indicates that it takes time to get the information on interest, order and deliver machines, and implement them. The domestic as well as the foreign interest rate indicate different types of opportunity costs. Moreover, as in the savings equation we add the changes of official development aid as a linear quadratic trend. Adding remittances directly here rather than only via the savings function makes the interest variable in this equation highly insignificant and violates the standard macroeconomic approach of making a clear assumption whether a decision is taken by firms or households. Households decide upon savings and firms decide where and how much to invest. But in open economy equilibrium savings are invested, at home or abroad. Therefore the fact that remittances are invested is not in contradiction with having remittances only in the savings equation. For development aid this may be different to the extent that donors can enforce that aid is invested without withdrawing domestic means. This is the reason why we have included aid here. Perfect withdrawal then should render aid insignificant. Besides the impact of remittances on physical investment via enhanced savings, reduced debt demand and interest rates, the higher savings from more remittances may complement primary school enrolments in their effect on literacy. This is captured in equation (5). lit - lit(—5) = cji + cs2Üt(-5) + C53sepri(—5) + cs4savgdp(-5)+ c 5 5 (lit(-5)) 2 + cjépeegdp(—5) + u j

(5)

Literacy, lit, is assumed to depend on its own lagged value in a linear-quadratic way, on enrolment in primary schooling five years earlier, sepri, and the savings available at the moment of enrolment. These can be used to avoid credit constraints (see Cinar/Docquier 2004, Bertoli 2006). Public expenditures on education as a share of GDP are also included. Enrolments are significant in the cross-country regression of Verner (2005), and Mazumdar (2005) has suggested public expenditure on education as a share of GDP. It is insignificant in his cross-country regressions but significant in our pooled estimate, which suggests that there is a dynamic impact. Literacy data are used as a proxy for human capital. They have a pretty good variation over time and across countries. In the working paper version we show the kernel density estimate using the EpanechnikovSilverman approach (see Silverman 1986). The distribution has decreasing maximum and increasing minimum values and goes from a slight twin peak structure to one that is increasingly skewed. Enrolment in primary schooling, sepri, is assumed to be a quadratic function of its lagged value and its square, and again savings at the moment of enrolment and the change of

Worker Remittances and Growth: The Physical and Human Capital Channels · 749

development aid, which is sometimes tied to investment in education through conditions imposed by donors. 1 2 sepri - sepri(—5) = cg¡ + Cé2sepri(-5) + c é 3 f s e p r i ( - 5 ) ) 2 + c é 4 ((oda/gdp) -

(oda(-5)/gdp(-5)))+

c 6 5 ((oda/gdp) - ( o d a ( - 5 ) / g d p ( - 5 ) ) ) 2 + c 6 6 savgdp + u6 (6) If remittances have increased fixed capital formation indirectly via enhancement of savings, reduction of net debt flows and reduction of interest rates and literacy via savings, both physical and human capital investments may have an impact on the growth rate. 1 3 Equation (7) endogenizes the growth rate. M g d p p c ) - log(gdppc(-5)) = c71 + c 7 2 log(gfcfgdp) + c 7 3

\og(gfcfgdp{-5))+

c 74 lit + c 7 J l i t ( - 5 ) + c 76 d(log(/)) + c 77 d(log(world))+ lagged dependent variables + m7

(7)

We use five-year intervals here for two reasons. First, we do want to get rid of business cycle effects. Second, we do not want to apply the method of using five-year averages for reasons given in Loayza et al. (2000) and Attanasio et al. (2000). In regard to the investment as a share of GDP variable Attanasio et al. (2000) have pointed out that growth regressions tend to use the investment data over the same period as the dependent variable whereas vector-autoregressive approaches use lagged investment and both get opposite signs. As the authors point out, this is hard to explain. We use both, current and lagged investments. Then, in a steady state both have equal values and can have the same role as the savings ratio in a Cass-Koopmans growth model if the difference of their coefficients is positive. They can differ, however, outside the steady state and will increase over time if the utility function has a consumption minimum to be reached for positive utility (see Dollar/Burnside 1997) 1 4 . Table 1 confirms this empirically for the past. Savings ratios for poorer countries had positive growth rates, whereas those of richer countries had negative growth rates. Investment rates are still growing in all samples. The literacy variable proxies for human capital and will have an impact on transitional growth and the long-run level of GDP per capita. 1 5 Bertoli (2006) has pointed out that it is desirable to have a feedback from human capital to remittances. This feedback is present in our model as literacy enters the growth equation, and growth is a variable in 12

Public expenditure on education as a share of GDP either has the wrong sign or is not significant in combination with making the intercept insignificant. In regard of the significance in the literacy equation this suggests that public expenditures are not so important for starting schooling but are important in succeeding to get basic education, here literacy.

13

An early contribution to the relation between literacy and growth is Azariadis and Drazen ( 1 9 9 0 ) . The aspect cited here does not appear in the version published later. Illiteracy also captures inequality, because the illiterate are likely to be poor. In related work we found that Gini-cefficients of education get insignificant in growth regressions when literacy is included. Castellò and Doménech ( 2 0 0 2 ) found that Gini coefficients of income change sign in growth regressions when Gini coefficients of education are included. By implication of the t w o findings literacy is likely t o capture much of inequality.

14 15

750 • T.H.W. Ziesemer the remittance equation for the first three samples. Moreover, the growth rate of employment plus depreciation 16 , approximated here by that of the labour force, 17 has a negative impact on the transitional growth rate and the steady-state level of GDP per capita. Finally, we will add some lagged dependent variables as an autocorrelation correction hoping that this absorbs the business cycle effects and allows interpreting the other regressors as growth effects. In models with imported inputs (see Bardhan/Lewis 1970) one finds also the growth rate of exports at constant terms of trade. This should be an income term in an export demand function and therefore is approximated here by the growth rate of the world GDP. When using this variable the initial value of the GDP per capita becomes insignificant and current literacy becomes significant in addition to the lagged one in some samples. Moreover, the model containing the GDP of the world has a higher adjusted R-squared than the one containing lagged dependent variables. Constant long-run growth in the world economy or by the OECD allows for positive permanent growth in this model. Equation (7) will be used to calculate the impact of remittances on the longrun growth rate of the model. We call this the permanent growth model.

3

Data and econometric method

All data are taken from the WDI (World Development Indicators). We include 99 countries selected by the criterion of having at least one dollar of remittances received in 2003 according to the data reported. Other criteria yield a lower number of observations. From the complete sample of 99 countries we drop three not having GDP data and call the sample remit96. We generate a second sample by eliminating those twelve countries that did not receive development aid. This eliminates OECD countries. We call this sample remaid84. Next, we divide this sample into those above and under (constant 2000) $ 1200 GDP per capita. The reason is that we found in earlier work that the 70 countries below $ 1200 have no growth in a panel average when looking at the period 1960 to 2003. However, both samples have 42 countries only, because many of the poor countries do not provide the relevant data. Estimating the model for four different samples will tell us how robust our model is in regard to dis-aggregation or how differently poor and rich countries with and without OECD countries react to remittances in regard to the level or the rate of growth. 18 Countries not reporting data may behave differently in particular because they are probably among the poorest. As we have to exclude them,

16

17

18

This constant term changes nothing but the value and significance of the constant cj\ in the equation. This neoclassical approach is of course not adequate for labour surplus countries (see Ranis 2 0 0 7 ) , but yields significant results for our panel estimates below. We do not make a distinction between large and small countries or regional location in terms of dummies for Sub-Saharan Africa or other regions. Per capita income differences do have a clear interpretation in terms of different behavior when being poor. Therefore w e prefer to investigate this disaggregation rather than others. In particular there are enormous income differences between countries in Sub-Saharan Africa, with the poorest having about $ 1 0 0 (300 in PPP) and the richest $ 7 0 0 0 (15000) leading to a difference of income by a factor 70 (50) within SSA in 2 0 0 6 . For every other distinction we would have to make the analysis of the following sections again. Our four samples would then be doubled to contain eight samples by making the distinction between African and others and doubled again by making some distinction between small and large countries leading to analyses for sixteen samples. Moreover, some of these samples would be very small. A reduction of these huge numbers of possible samples would then only be possible by removing the other criteria like income differences, which in our view are the most interesting ones.

Worker Remittances and Growth: The Physical and Human Capital Channels · 751

our results may have a selection bias and differ from those of our samples as much as our samples have different results among each other. The data on remittances are official receipts in constant 2 0 0 0 US$. Unofficial receipts may be high - Freund and Spatafora (2005) estimate that informal remittances are between 35 and 75 % of the official ones - and important but we have no way to deal with the issue directly 19 (see Adams/Page 2 0 0 5 ) . 2 0 Data of the GDP per capita, gdppc and OEC are in constant 2 0 0 0 US$ and stem from national accounts. 21 Interest rates, π and rius, are real rates as obtained by use of the GDP deflator and taken from the IMF IFS Yearbook into the WDI data. Savings, savgdp, are gross national savings from national accounts, calculated as GDP minus consumption, plus net current transfers and factor income from abroad and expressed as a share of GDP. As investment, nvgdp, relates to the demand of net debt flows we use gross capital formation (formerly gross domestic investment) as a percent of GDP. The major difference with gross fixed capital formation as a share of GDP, gfcfgdp, is the inventories, which are not investments that add to the capital stock as usually written into a production function. All savings and investment data come from the national accounts. Literacy data, lit, from the UNESCO are available in the WDI for 75 of our 99 countries for more than 30 years. Data on public expenditure on education are from the UNESCO. Data on official development aid, containing at least a grant element of 25 % on the interest rate benchmark of 10 % , stem from the OECD. Finally, enrolment into primary schooling, sepri, refers to data from UNESCO on the gross enrolment of a vintage, that is, older people who go to primary school make it possible for this number to get above 100%. The average values of some of these data are presented in Table 1. These data show that all samples have positive growth rates of GDP per capita, but the poorest one has the lowest growth rate. Investment/GDP ratios are higher in richer samples and have higher growth rates in poorer samples. Savings/GDP ratios are highest in the middle-income groups and also have higher growth rates for poorer countries. Investment/GDP ratios are higher for all countries than savings/GDP ratios inducing higher indebtedness. Investment/GDP ratios have higher growth rates than savings/GDP ratios, implying that the indebtedness will also grow more quickly than in the past. Average remittances per unit of GDP are 2 - 3 % but with a growth rate of 2 - 5 % , which is larger for poorer samples. We estimate equations ( l ) - ( 7 ) as a system for pooled data. In the estimation of the system for pooled data we assume contemporaneous correlation, which means that the residuals of the equations may be correlated with each other for a given point in time. The reason for this may be that the variables do not only have growth effects but follow also a busi-

19

20

21

Panel data on remittance fees, which cause unofficial receipts, would be an interesting addition here. But we are not aware of their availability. We would like to point out though that GDP data also underestimate economic activity because of the neglect of the informal sector. Schneider and Enste ( 2 0 0 0 , Table 2 ) report values of 2 5 - 7 6 % of GDP for developing countries. This is the same order of magnitude as for remittances. For developed countries these values are lower. Informal remittances are falling as a share of the official ones. It is not clear though that the share of the informal sector is falling in developing countries over time. The imperfection of remittances data is broadly discussed in all related papers. That of GDP data is not discussed anymore although it may be as severe. We d o not use purchasing power parity data because they are available only since 1 9 7 5 in W D I 2 0 0 7 and since 1 9 8 0 in W D I 2 0 0 8 , whereas our data start in 1 9 6 0 .

752 · T.H.W. Ziesemer

Table 1 Data description of the four country samples 3 Panel average of Variable/Sample Remittances/GDP Growth rate of remittance/GDP ratio GDP per capita $ Growth rate of GDP per capita Investment/GDP Savings/GDP Growth rate of Investment/GDP Growth rate of Savings/GDP Literacy rate

Remit96 0.0235 .02 3884 .0155 .228 .181 .003 -.001 71.1

Aid84

Above $1200

.0263 .04 1660 .0127 .213 .171 .006 .00016 69.9

Under $ 1200 .02 .048

.032 .03 2860 .019 .227 .192 .0016 -.0014 81.8

500 .007 .199 .148 .012 .009 56.4

a Least-squares dummy variable regressions of the variable on a constant and, for growth rates, for the natural log of the variable on a constant and a time trend. Source: Author's calculations with data from World Development Indicators

ness cycle. Therefore the residuals are likely to m o v e together. Moreover, w e assume absence of serial correlation and w e a k exogeneity, w h i c h m e a n s that the residuals of an equation m a y be correlated w i t h future regressors, but not w i t h current or earlier ones. T h e interaction of these a s s u m p t i o n s m a k e s it possible that for e x a m p l e the remittances variable in e q u a t i o n (2) is correlated w i t h the residuals of equations (1) and (2). Therefore lagged regressors should be used as instruments for the remittance variables in these equations. Moreover, w e c a n n o t exclude the possibility that the residuals f o l l o w m o v i n g averages. This w o u l d make the first lags of all left-hand variables, w h e n they appear o n the right-hand side, also e n d o g e n o u s . W i t h or w i t h o u t m o v i n g average residuals, the higher lags than those o n the right-hand side variables will be admissible instruments. The c o m b i n a t i o n of endogeneity w i t h c o n t e m p o r a n e o u s correlation of the residuals leads to three-stage-least squares (3SLS), w h i c h in turn is a special case of the G M M - H A C estimator (see Greene 2 0 0 8 : 4 6 9 ) (General M e t h o d of M o m e n t s ( G M M ) in c o n n e c t i o n w i t h the heteroscedasticity and autocorrelation correction ( H A C ) for the covariance m a t r i x ) . 2 2 22

This improves the efficiency, but does not remove a potentially present serial correlation bias, which would also invalidate the used instruments. The only hint what to do about serial correlation, if anything, is to add lagged dependent variables (see Greene 2008). However, except for equations (7) we have already added all significant lagged dependent variables. There is nothing in addition we can do, but conceding that there may be an additional, hopefully small, serial correlation bias, which then is likely to exist also in the literature from which we have taken the specification of the equations. Introducing serial correlation processes by assumption leads us to a 'near singular matrix' warning, not when doing it only for one equation alone but when doing it for several. Therefore, and because the instruments approach would not avoid endogeneity, we abandon this possibility, hoping that a potential serial correlation bias, if any, is small in view of the fact that the cross section dimension is much larger here than that of time. Durbin-Watson statistics may serve as a crude indicator of serial correlation. They are not reported in the appendices, but they are close to two for all equations except those for literacy and primary school enrolment. Division of the number of observations by the number of countries both presented in the appendices suggests that the number of periods is between two and fifteen, which is very short even for testing of serial correlation. For literacy and primary school enrolment the time dimension is only two or three, indicating that little can be done against autocorrelation if it is a problem at all with such a low time dimension. Note though that the GMM-HAC method presents standard errors that are corrected for serial correlation.

Worker Remittances and Growth: The Physical and Human Capital Channels · 753

An alternative might be to estimate the equations single wise after checking for fixed effects redundancy. In case of redundancy, two-stage least squares m e t h o d s could be used perhaps in c o m b i n a t i o n with r a n d o m effects methods. If fixed effects are not red u n d a n t w e w o u l d have to employ dynamic panel data m e t h o d s as explained in Baltagi (2005: C h a p t e r 8). We prefer to take the interaction of the residuals of different equations on board, because they contain the business cycle effects and therefore will be correlated, and therefore we use only the systems a p p r o a c h sacrificing the fixed effects, which w o u l d a d d 9 5 coefficients t o each equation. We will leave the fixed effects app r o a c h for f u t u r e research. Some variables are between zero a n d unity. These are literacy, gross fixed capital f o r m a tion as a share of GDP, and remittances as a share of GDP. 2 3 For these one might think of using log(y/(l-y)) instead of y as the dependent variable. We will postpone this a p p r o a c h to f u t u r e research. We w o u l d like to point out t h o u g h that our variables are on the country level, are not discrete (let alone binary), have m a n y different values a n d are neither censored nor truncated. Moreover, the p r o d u c t of the regressors a n d their coefficients of the corresponding regressions are always smaller than those of the lagged dependent variables and ensure that there are n o negative values for the dependent variables. A referee has pointed out that the heteroscedasticity is k n o w n in these cases (at least at the zero and unity values) and one should be cautious a b o u t putting t o o m u c h weight o n test statistics. In general, running simulations with estimated results often turns out t o be an intuitively p o w e r f u l plausibility test especially in case where the significance of a regressor is more plausible in the cross-section dimension then in the time dimension. It therefore helps excluding highly implausible alternatives in the model selection procedure. 4

Estimation results

T h e results for the systems estimate are summarized for the f o u r samples in Appendix la-d. For the 96 countries receiving remittances the estimate in Appendix l a is done w i t h o u t the inclusion of an aid variable, because they are available only for 84 countries and the results for that sample are s h o w n in Appendix l b . All coefficients have the expected sign. T h e significance is worse t h a n 10% only for f o u r coefficients: the effect of domestic interest rates on receiving remittances in equation (1'); the effect of primary school enrolments on the change of literacy in equation (5'); and the constant of the g r o w t h equation (7'). Only the last of these exceeds the 20 % significance level slightly. 24 In the first equation, the positive sign of the O E C D per capita income and the negative one of the domestic G D P per capita are in accordance w i t h the altruistic a n d strategic motives of migration and with those motives, which d o n o t generate a clear expectation of the sign (see R a p o p o r t / D o c q u i e r 2006). T h e US interest rate and the O E C D income have a stronger impact t h a n the domestic counterparts. This will also be the case in all estimates for other samples given below. It confirms the result by Vargas-Silva a n d H u a n g (2006) for a

11

24

Savings ratios are in the interval (— 20, 70.5), school enrolment primary are in the interval (67, 139) as these are gross rates including older vintages going to school, and real interest rates are in ( - 97.8, 79). We do not drop variables with weak significance levels if this would decrease the adjusted R-squared strongly or yield a weak Durbin-Watson statistic indicating serial correlation bias.

754 · T.H.W. Ziesemer

smaller sample that home country variables have a weaker impact on remittances than host country variables. The main channel to physical capital has the expected, significant coefficients: remittances have a positive effect on savings, C23; savings have the expected negative effect on interest rates, C34, and also the quadratic term is significant; interest rates have a negative impact on gross fixed capital formation, C43; gross fixed capital formation has a positive impact on growth rates, which is larger than the negative effect of the lagged value. For the human capital channel, savings enhance literacy, C54, and literacy enhances growth rates as the positive current effect is larger than the negative lagged effect. How strong these effects are will be calculated in the next section. For the 84 countries receiving remittances and aid in 2003 the results can be found in Appendix l b . These are very similar to those of the larger sample. The insignificant variables now are US interest rates in equation (1"), again the enrolment variable in the literacy equation, the constant in the growth equation, and, though very close to the 10% level, the OECD growth rate in the interest equation. By implication the channels to physical and human capital have only significant variables although with slightly different values. The model gives reasonable results after the elimination of the OECD countries from the larger sample. The development aid variable appears only in the form of first differences. Under the steady-state assumption that aid as a share of GDP should be constant this result implies that aid has neither a level effect nor a growth rate effect in the long run. In spite of the similarity with the 'medicine model' of development aid defined by the squared term (see Doucouliagos/ Paldam 2005) we would like to caution that we do not include all the variables, especially for economic policy, which have featured prominently in the aid effectiveness debate. Our motivation to include aid does not stem from a desire to contribute to the aid effectiveness debate but rather from the desire not to underestimate the equations of our model. In the transition though aid has a positive effect on savings, investment and primary school enrolment as long as it is increasing and below 25 % (for investment even 50 %). When aid is a constant share of GDP though, there is no effect anymore. 25 For the 42 countries with GDP per capita above $ 1200 and receiving remittances and aid in 2003 we get quite a few insignificant results in Appendix lc. This may be partly due to the fact that now the number of observations is about half of what it was in the previous sample and less than half of the first one. Mostly, then the coefficients are also smaller. With some exceptions results do not improve (in terms of adjusted R-squared) if we take out these insignificant variables. As overestimation does not produce biases whereas underestimation does (see Davidson/McKinnon 2004) it is less risky to keep them on board. We have eliminated though the quadratic terms of the aid variable from the equations for savings and primary school enrolment. Also, the quadratic term of the investment-savings difference has been dropped in the interest equation, where the second lag is replaced by the change of the US interest rate. In the investment equation, the aid variable now has a higher peak at about .8 before it is getting negative. In the growth equation almost all variables have smaller coefficients in absolute terms now. The current interest rate rather than the lagged one is significant now in the savings equation, and for literacy in the growth equation we lag by one year more than in the other regressions. In regard to 25

When interpreting the equations in the spirit of first-differenced models - that is the case where the coefficient of the lagged dependent variable has a unit coefficient - one may be more positive about the long-run effects of aid. But here we stick to the exact formal derivation of steady-state results as carried out in the next section.

Worker Remittances and Growth: The Physical and Human Capital Channels · 755

the main channels, the impact of savings on interest rates is significant only at the 20 % level. For the growth equation the significance is even worse, both because the coefficients have become smaller and the standard errors are larger than in the larger samples. In the first equation for remittances the coefficient of the lagged dependent variable has gone down from .89 in the previous sample to .83 in this one. By implication we should expect that it goes up for the other half of the larger sample. For the 42 countries with GDP per capita below $ 1200 and receiving remittances and aid in 2003 the coefficient of the lagged dependent variable for remittances is - in line with the expectation three lines above - slightly larger than unity and differs insignificantly from unity. Therefore we have taken the first difference as the dependent variable. 26 There are five coefficients with marginal significance levels (p-values) between 10 and 20 percent and one worse than 20 %, which is the enrolment variable in the literacy equation. It has a very low coefficient too. Compared to the sample of countries with income above $ 1200, the quadratic term for aid in the savings equation is again significant and so is the quadratic investment-savings difference in the interest equation as they were in the sample of 84 countries receiving aid. The linear term for the aid variable in the investment equation has been taken out. Most importantly now the lagged savings variable does not appear in the literacy equation but rather the current one appears in the school enrolment equation. This suggests that in the richer countries one needs savings from earlier times to bring children through primary schooling, but in the poorest countries the bottleneck are current savings to start primary schooling. Comparing the results across the four samples also yields some interesting insights. We see larger coefficients of changes in remittances on savings in samples of poorer countries: a coefficient of .68 for the richest sample of 96 countries; 2 7 .69 for the second richest sample of countries with income above $ 1200; .88 for the 84 aid receiving countries; 1.85 and 1.91 respectively for the poorest sample. This may reflect the lower financial development of poorer countries as indicated theoretically by Cinar and Docquier (2004) and Bertoli (2006) and empirically by Giuliano and Ruiz-Arranz (2008). With less credit access people save more out of remittances. A low financial development may in principle also lead to a more intensive use of informal channels and less measured remittances and therefore to a higher coefficient. However, Niimi and Özden (2006) find no impact of financial development indicators on measured remittances though. Among the non-OECD countries, labour force growth has a more negative and world income growth a more positive sign in the growth equation the poorer the countries are. We would have expected that the public expenditure on education in the literacy equation has a weaker effect the richer the countries are as found by Otani and Villanueva (1990). This holds except for the poorest sample, which has the lowest coefficient - perhaps because dropout rates are higher.

26

27

The result can be improved by adding quadratic terms but then the forecast follows these terms and makes very unrealistic predictions for the steady-state analyzed below. Savings and investment rates are percentages multiplied by 100 in the WDI. Our own calculations of wr/GDP are not multiplied by 100. If they were, the coefficient would be lower by a factor 100. This explains the difference between the values in the Appendix such as 68 and the ones used here, 0.68.

756 · T.H.W. Ziesemer

5

The direct effect of a change in the rise of remittances on other endogenous variables

In order to understand the basic idea of the model it may be good to look first at the direct effects of changes in remittances on the endogenous variables, in particular the growth rate. For that purpose we abbreviate the variables as follows, w is worker remittances as a share of GDP. s is the savings ratio, f is gross fixed capital formation as a share of GDP. 1 + r is the gross interest rate. It is the literacy rate, p is the rate of primary school enrolment. g is the growth rate of GDP per capita, Λ: is a multiplication sign. The results for this part are collected in Table 2. We illustrate the derivation of the results in terms of the model equations (2) - (7) assuming in this section only that there are exogenous changes in remittances for these equations. We do this numerically for the estimates regarding the largest sample of 96 countries receiving remittances in 2003. In standard macroeconomic models one would speak of the impact effect. However, we have lags here and therefore call it direct effect. From equation (2), a one percent difference of dw, d(dw) = 1, the change in the remittance/GDP ratio, yields: ds = czsddw = .6S2ddw = .682. This means that an increase of dw by one percentage point increases savings by almost. 7 percentage points. Note that the panel average of ddw = .0138, and not unity as in our example. Therefore all effects could be multiplied by this number to get the realistic values according to the panel average. The effect of this change in the savings ratio on the interest rate according to equation (3) is: d(log( 1 + r)) = [C34 + c35l(f

- s)\d{f - s).

With an evaluation at a panel average of f-s = 4.06 and d(f-s) = -ds =-.682 from above 2 8 , the direct effect of remittances on interest rates using the numbers of Appendix l a is: Table 2 The direct effect of changes in remittances on the endogenous variablesa

Sample Change in

Remit96

Remaid84

Above $ 1200

Under $ 1200

savings (ds) interest in % investment in % literacy (dli) GDP per capita in % b ddw c

0.682 -1.048 χ 10" 3 5.9 χ 10- 5 1.54 χ 10"2 2.66 χ 10" 6 + 4.63 χ 10" 5 0.0138

0.88 -5.1 χ 10" 3 3.5 χ 10" 4 2.103 χ 10"2 (1.58 + 7) χ 10" 5

0.69 -4.58 χ IO" 4 8.4 χ 10" 5 2.33 χ 10~2 2.86 χ 10-6+ 4.7 χ 10" 5 -0.0004

1.845 -6.47 χ 10" 3 3.443 χ 10~4 9.1 χ 10~3 1.86 χ 10" 5 + 4.46 χ 10' 5 0.037

0.016

a

Calculated for ddw = 1, a one percent change of the rise in remittances per unit of GDP,dw b sum of investment and literacy effect c second difference of the remittance/GDP ratio in a panel average; multiplication factor for all effects above to get yearly effects Source: Author's calculations 28

For f-s the value for the sample with 84 countries is 4.8, for the u l 2 sample it is 6.075. For the sample above $ 1200 it is not needed because the quadratic term of investment minus savings is dropped from the estimate. The value is 3.6. N o t e that these values also represent the yearly additional foreign debt incurred as a percentage of the GDP. Whether this is suggestive of weak or strong capital movements is left to the reader.

Worker Remittances and Growth: The Physical and Human Capital Channels · 757

d(log( 1 + r)) = [C34 + cis2(f

- s)]d(f - s)

= [0.003426 - 0 . 0 0 0 2 3 χ 2 χ 4.06] χ ( - . 6 8 ) = - 1 . 0 4 8 χ IO" 3 . T h e gross interest rate changes by a b o u t (— .1 %). This change of the interest rate causes a percentage change of gross fixed capital f o r m a t i o n according t o equation (4) of dlogf = c4idlog(l

+ r) = - . 0 5 6 3 2 5 χ ( - 1 . 0 4 8 χ IO" 3 ) = 5.9 χ IO" 5 .

As w e are dealing with the direct effect only, this change in f has not been t a k e n into account in the previous step, w h e n trying to find the effect on interest rates. T h e effect of the change in savings on t h a t in literacy according to equation (5) is d{li) = c54ds = 0 . 0 2 2 6 2 2 χ .682 = 1.542 χ IO" 2 . Literacy changes by one a n d a half percent five years after the enhancement of savings. According to the g r o w t h equation we get the direct effect as (note that values f r o m five years before are given) dloggdppc = g = c72dlogf

+ cjnd(li) = 0.045 χ 5.9 χ IO" 5 + 0.0032 χ 1.542 χ ΙΟ" 2

= 2.6581 χ ΙΟ" 6 + 4.6284 χ ΙΟ" 5 = 4.8942 χ 1 0 - 5 . This is a very small percentage change of the G D P per capita. Interestingly t h o u g h , the effect via h u m a n capital or literacy is twenty times as large as that via physical capital. Effects calculated so far are rewritten in the first column of Table 2. For the other samples they can be f o u n d in the second t h r o u g h f o u r t h column of Table 2. For poorer samples the ratio goes f r o m 2 0 t o 2 . 5 for the poorest sample. W h e n dropping the O E C D countries, the sample remaid84 has larger effects in absolute terms t h a n the complete sample. Also the second difference of the remittance/GDP ratio, w, is larger. The relative strength of physical a n d h u m a n capital is n o w less strong; h u m a n capital has a g r o w t h effect that is m o r e than four times as large as t h a t of physical capital. W h e n splitting the sample we find for the p o o r countries that the effect of remittances on savings is m u c h larger; the fall in interest rates is larger t h a n in the samples considered previously; the effect on investment is a b o u t the same as for the 84 aid receivers; the effect o n literacy is m u c h smaller because the elasticity of primary school enrolment on literacy is very w e a k . H u m a n capital has an effect on g r o w t h t h a t is 2 . 5 times stronger t h a n t h a t of physical capital. T h e second difference of the remittance/GDP ratio is twice as large as for the r e m a i d 8 4 sample. For countries above $ 1200, things are quite different. T h e impact of remittances on savings is a b o u t the same as for the largest sample. T h e fall in interest rates is m u c h lower t h a n in all other samples. Correspondingly the effect on investment is weaker. T h e effect on literacy is a bit larger though. The total effect on g r o w t h is a b o u t the same as for the 96 countries. N o t e however, t h a t these are effects for ddw = 1, the second difference of the remittance/GDP ratio is taken t o be unity. T h e m a j o r difference between the sample with countries above $ 1200 a n d the others is t h a t the panel average of ddw is negative for this sample. T h e actual yearly effects therefore have the opposite signs and are m u c h smaller in absolute terms. This is the reason w h y it is useful t o split the analysis into the effects for ddw = 1 a n d the actual size of the change in the rise of remittances.

758 · T.H.W. Ziesemer

Overall, we can say that poorer countries do not only have a stronger impact of remittances on savings as pointed out above, but also a stronger direct effect on growth. The direct effects considered here do little more than illustrating the basic ideas that drove the set up of the model. The indirect effects, as indicated above for the gross fixed capital formation on interest rates come from the calculated changes on all endogenous variables again. Multiplier effects then also take into account that all changing variables have an impact on their own future values and of course all the three sorts of effects then interact. What one would like to know than is the total long run effect on the level and the growth rate of the GDP per capita.

6

The long-run solution of the permanent growth model with and without remittances

In this section we first solve the model for its steady-state values and than do it again under the assumption of no change of remittances. A steady state is defined as follows: A constant growth rate of the GDP per capita for the receiving countries and constant interest rates, a constant but positive growth rate for the GDP of the OECD and the labour force variable. First, we have to find a steady-state growth rate of the world economy used in equations (7'"lv). We run an instrumental variable regression of that rate on its own lagged value using second lags as instruments. The result is a steady-state growth rate of about 3.4 %. Next, we assume a certain value of the growth rate of the domestic economy and go through a procedure explained below. If that one does not come out in the end, we adjust it, and go through the whole procedure again until the assumed growth rate equals the one coming out. This process stops when the rounding by the program used does not allow further refinements. The procedure referred to above is as follows. Equation (1) then implies a constant change of the remittance/GDP ratio, i.e. a constant d(wr/GDP). Constant aid and savings follow from equation (2), and constant investment and gross fixed capital formation from equations (3) and (4), all as a share of GDP; constant public expenditure on education as a share of GDP, enrolment in primary schooling and a constant literacy rate follow from equations (5) and (6). As an implication of this definition, our variables and there lagged values must then be identical, except for logOEC, which has a positive but constant time trend and therefore a constant growth rate. As the model has quadratic terms of the investment-savings difference, the enrolment in primary schooling and the literacy rate, we cannot solve it in 'one shot' after implementing the above assumptions but rather must proceed in certain steps. The first step is to take first differences of equation (1) and employ the steady-state assumptions. The result is a value for d(wr/gdp) depending on the growth rate of the OECD. For the OECD we assume a steady-state growth rate of 2 %. This is obtained from running an autoregressive instrumental variable regression of log(OEC) on a time trend and three lags and calculating the steady-state value of the growth rate. The results for the steady-state change of the remittance/GDP ratio are summarized in Table 3 for the first three samples. For the poorer sample, equation (l lv ) in the Appendix shows that we cannot solve independently for the change of the remittance/GDP ratio. It can be shown that for past OECD growth rates of about 2.5 % the time trend in log(oec) and the time trend in the regression cancel out. Using the abbreviations from the end of section 4 this leaves us with

Worker Remittances and Growth: The Physical and Human Capital Channels · 759

d(w) = e n + c u * log(OEC(0)) + c 1 5 * log(l + r)+ c16*(log(l+r)-log(l+rius(-2))) A regression of log(OEC) on a polynomial of time yields an initial value for log(OEC) of 9.12. An autoregressive instrumental variable model of order one for the US interest rate yields a steady state value of about 4.3 %. Using these values and the estimated coefficients from equation (l lv ) we get the results noted in Table 3, where the right-hand side remains a function of the domestic interest rate. However, the term with the interest rate is 2.05 χ 10" 4 χ log{ 1 + r). As log( 1 + r) is also a number like 8 % this expression is of the order of magnitude of 16 χ 10~ 6 and therefore can be dropped, leaving us with the number presented in Table 3. The yearly change in the remittance/GDP ratio lies between one tenth and nine tenths of a percent. For non-OECD countries, it is larger the poorer the sample is. Table 3 The steady-state solution of the permanent growth model in % Sample Variable publ.exp education /GDP labour force growth rate in %

Remit96

Remaid84

Ab.f 1200

3.97 1.93

3.79 2.07

4.57 1.95

3.22 2.20

0.1 20.1 14.4 10.7 88.0 100.5 0.3572

0.9 20.1 21.4 2.3 103.1 102.9 0.0275

0 20.1 13.6 10.9 88.0 100.3 0.3564

0 19.0 12.3 11.4 95.4 102.4 0.0198

0.041 0.79 -0.19 0 0.16 0.001 1.002

1.05 9.11 -9.03 7.654 0.49 0.008 1.393

(a)

With remittances

yearly difference in remittances/GDP Gross fixed capital formation/GDP Gross savings/GDP real interest rate enrolment rates primary school literacy rate GDP per capita growth rate

0.2 19.8 15.6 9.2 93.3 100.5 0.2296

0.2 20.4 14.7 9.9 97.3 100.6 0.2175

(b)

Without remittances

yearly difference in remittance/GDP Gross fixed capital formation/GDP Gross savings/GDP real interest rate enrolment rates primary school literacy rate GDP per capita growth rate

0 19.8 14.6 9.3 93.3 100.4 0.2291

(0

Perc.differences panel (a)-(b)

Gross fixed capital formation/GDP 0.015 Gross savings/GDP 0.99 real interest rate -0.15 0 enrolment rates primary school literacy rate 0.12 GDP per capita growth rate 0.001 Ratio of growth rates with and without 1.002 remittances

0 20.4 13.3 10.0 97.3 100.4 0.2167 0.014 1.47 -0.10 0 0.2 0.001 1.004

Und.S 1200

Steady state assumptions: OECD per capita income growth rate: 2 %; US interest rate: 4.3 %; World GDP growth rate: 4.3 %. Source: Author's calculations

760 · T.H.W. Ziesemer

T h e next step uses equations ( 2 ) - ( 4 ) , imposes the steady-state assumptions and results, inserts the estimated coefficients and solves for f , s, and r, the investment/GDP ratio, the savings/GDP ratio and the domestic interest rate. In doing so we equalize the investment/ GDP ratio with gross fixed capital formation per unit of GDP plus a constant of about 1.4% (different for each sample) from regressing them on each other, which represents the percentage share of inventories, as both follow a one-to-one correlation. T h e procedure in greater detail is to solve (4) for f , (2) for s and form f-s; together with (3) this gives two functions in f-s and η Solving, r can be inserted into (2) to get s and then f follows. For three of the samples we get two solutions, of which one makes no sense because interest rates and savings rates are highly negative and investment rates exorbitantly positive. Therefore these are ignored. T h e detailed procedure described above, results in the solutions of Table 3. All samples but the poorest have higher steady-state investment than savings in the steady state, which indicates that debt accumulation continues also in the steady state. The poorest sample may be credit rationed and has savings about as high as investment, but at a low interest rate, as savings are slightly higher. N e x t , we can go to the equation for enrolments. This can be solved independently except for the last sample where we will use the savings rate just derived. As this is an inverted u-shape function, the partially stable equilibrium is the one with higher enrolments. T h e panel average value is larger than the threshold value. Therefore we use this high value as the steady-state value. N o w we can solve for the literacy rate provided we have a long-run value for public expenditure on education. We run an autoregressive least-squares-dummy-variable 2 9 regression on the lagged value and its quadratic value, resulting in a steady-state value between 3 and 5 % as documented in the first line of Table 3. This equation also has an inverted u-shape form. T h e lower steady-state value though has negative numbers in all cases. The higher steady-state values for literacy are amazingly close to 100% and almost so for enrolments. Finally, using all results obtained so far we can calculate the steady-state growth rate of the G D P per capita from the last regression of the model, equations (7'" lv ), provided we have a steady-state value for the growth of the labour force. We run a least-squares-dummy-variable regression of the labour force growth rate on its lags, with linear and quadratic terms and calculate the steady-state values, which are around 2 % , with higher values for poorer countries. In panel (b) of Table 3 we present the numbers that are obtained when setting the changes in the remittances equal to zero when running through the whole calculation again. In panel (c) we take differences of panel (a) and (b) for all variables that are already percentage expressions. Steady-state growth rates are positive though small, .23 %, .22 %, and .36 % for the first three samples. 3 0 For the poorest sample we find a result of a 'po-

29

These estimates will be biased. Assuming fixed effects and using first difference methods does not provide us with a value for the constant, which we need in order to calculate a value for the steady state. Fixed effects L S D V results are very similar to those for the pooled regression. As they are multiplied by a coefficient of . 0 4 in the regression all these values have a small impact. The results should be taken as assumptions anyway, because we do not have an elaborate explanation for these exogenous variables by definition.

30

Of course, the growth rates would be larger if we would assume a lower growth rate of the labour force. N o t e however, that in times of increasing participation the labour force growth will be higher than the population growth.

Worker Remittances and Growth: The Physical and Human Capital Channels · 761 sitive-zero' growth rate, or more exactly less than 2.8 percent of a percent, 2.75 χ 10~ 4 . The counterfactual exercise of dropping remittances shows that the ratio of the growth rates obtained with and without remittances is less than 1.004 for the first three samples but 1.39 for the sample of the poorest countries below $ 1200 per capita income according to panel (c) of Table 3. Literacy goes to hundred percent anyway and remittances make a difference of less than a half percent. However, primary school enrolments are enhanced by remittances and speed up the move to the steady state, leading to transitional gains from remittances via savings to enrolments and quicker movement of literacy to hundred percent. For the poorest sample this effect is larger again than for the others. 31 Remittances have a very strong impact on savings of the sample with countries under $ 1200 and, less so, on the sample above $ 1200. This leads also to a fall in interest rates. Similarly, gross fixed capital formation per unit of GDP increases by one percentage point in the poorest countries but much less so in the other samples, because they react only weakly to the decrease in interest rates that is caused by the increase in savings. In sum, the transitional gains from higher enrolment rates going from 95 to 100%, the permanent gains from paying less interest to debtors because of a savings rate that goes from 12 to 20 % and the increase in the growth rate of the GDP per capita for the poorer sample may represent a considerable welfare gain. For countries above $ 1200 though these gains are fairly small. In sum, the major effects for the poorest sample are that remittances bring the steadystate primary schooling enrolment rates from 95 to 100 %, the savings rate from 12 % to 21 %, and the growth rate from 1.9 χ 10~ 4 to 2.75 χ 10~ 4 for the sample of countries with per capita income below (constant 2000) $ 1200 and reduce the foreign debt service. The ratio of these growth rates is 1.39, which means that the growth rate is 39 % percent higher than without remittances. For the richer sample these effects are much smaller. Transitional gains may be higher.

7

Stability and transitional gains

Steady-state results as presented in the previous section are only interesting if the steady states are stable. Therefore we should present a stability analysis.32 Moreover, besides the steady state, the transition is also interesting. In order to obtain a stability analysis and the transitional path, we iterate the estimated model of Appendix 1 forward in a deter31

This may seem to be partly be due to the fact that richer countries have higher literacy already. However, also the richer sample has an average of below 8 2 % of literacy (see T a b l e t ) and only 7 % of all observations in this sample are in the top 2 . 5 % bracket of literacy rates. For the poorer sample this rate is a very similar 2 % . An alternative human capital variable could be secondary school enrolment because it has a similar strong variation. Literacy varies from 7 1 0 0 % and secondary schooling from 1 . 4 % to 1 2 7 % in the sample of 8 4 countries. However, including secondary school enrolment may require an extension of the model by several variables explaining it. Then one could argue that poorer countries have stronger effects of remittances on human capital because they have much more of a lower basis to start from then for literacy. Gross enrolment rates much higher than 1 0 0 % would leave traces in the estimates with unclear consequences for the long run. The major difference between the samples in our view is the impact of remittances on the savings ratios.

32

Thereby we automatically include second and higher round effects, which are missing in many other types of studies (see Adams 2 0 0 6 ) .

762 · T.H.W. Ziesemer

C o u n t r i e s with per capita i n c o m e a b o v e $1200 14 12

t\ A ! \ ψ 'V ' ' ·ύ -JrV-. Γ • \ I Κ ·____— 8

10

._

sepri/10 loggfcfgdp loggdppc 100*wr/GDP 100*d(wr/GDP) sa^dp/10 100*log(1+ri) j lit/100

β j4 j

Í

2 :·ν-/Λ, «Í 10 19 28 37 46 55 64 73 82 91 100 109 118 127 136

Ì

years 1960-2100

C o u n t r i e s with per capita i n c o m e below $ 1 2 0 0 15 \ λ „ i\ 1 Ί'ιΛ * io · w V . i i

—sepri/io - loggfcfgdp -loggdppc 1000*d(wr/gdp) —savgdp/100 ^46 55 64 73 82 91 100 109 118 127 136

—°—100*log(1+ri) —ι—lit/100

Years 1960-2100

Source: Author's simulations Figure 1 Stability of the permanent growth model ministic way and repeat this after setting the remittance term in the savings equation equal to zero, in order to get the counterfactual of 'what would have happened hypothetically without changes in remittances'. This allows us to see the transitional path and whether or not it goes to the steady-state values. We do this for two samples of countries above $ 1200 and below $ 1200. To be able to do this, we have to construct initial values, because the data deliver them only per country whereas our model has estimated para-

Worker Remittances and Growth: The Physical and Human Capital Channels · 763

meters, which are averages across countries and over time. We construct the initial values by running fixed effect regressions of the variables on a constant and a polynomial of time. All other assumptions carry over from the previous analysis. We only have to add a data series on development aid per unit of GDP, which was not necessary for the steadystate analysis. In order to get that we run a regression of aid/GDP ratio on its own two or three lagged values. As in the steady-state calculations, we assume that the time trend in log(oec) and the time trend in the regression (1) cancel out because a slight difference could cause instability. The results for stability are summarized in Figure 1. In Figure 2 we plot the differences of the variables with and without remittances for the permanent growth model in order to make the undiscounted gains during the transition visible. Stability The strongest fluctuations can be seen in the series for real interest rates. After the phase of fluctuations they get smooth for the richer sample sample but turn into a zigzag pattern for the poorer sample. One can take it from the regression results of Appendix 1 that interest series are based on yearly data. The regressions based on five-year lags - enrolment, literacy and GDP per capita (growth) - have been turned in to yearly data by making five initial values. All zigzag patterns in figure 1 are getting smoother over time around 2010 as they should for unique, stable steady states. The strongly upward sloping line for the richer countries is the remittance/GDP ratio. In 2100 it has a value of about 9.3 %. If iterating forward to 2700 it would still be at reasonable values of about 88 % and 79 % respectively. Ultimately, these graphs show stability, which can be shown more exactly when we would show the forward iterations for some hundred years more. Transitional gains We consider the richer sample first. Figure 2a shows the difference for the endogenous variables for the iteration with and without the remittance term. For the savings/GDP ratio there is first a surprising fall. Once this period is over, the difference in the savings/ ratio goes to a value that is .9%, which means that the ratio is higher with remittances than without, and then approaches its steady-state difference of .8 %. In the transition, interest rates are more frequently higher than lower and then approach a very low difference of — .002. The more frequently positive differences are reflected in the difference for the gross fixed capital formation per unit of GDP. This first goes down before it goes up and approaches a positive long-run difference after a slight overshooting. However, this is not more than .04 percent in the long run and the transitional phase with higher values does not really more than the worse phases. The undiscounted transitional gains seem to be smaller than the undiscounted steady-state gains in regard to physical capital formation. In regard to literacy we see that it is higher in the long run, but also much less so in the transitional phase. For the poorer sample we do not get a counterintuitive phase first, but rather savings rates are higher right from the beginning. Interest rates are lower with remittances and gross fixed capital formation, as a percent of GDP are higher, as expected. Primary school enrolments are much higher with remittances. It is here for the first time that we see that transitional gains may be higher than the long-run gains if the discounting of the future is strong enough. Similarly, the differences in the literacy rate are stronger in the transition than in the long run.

764 · T.H.W. Ziesemer

T h e s a v l n g s / G D P ratio

Interest rate

*50

57

64

71

78

85

92

99

106 113 120 127 134

Years 1960-2100 Gross fixed capital formatlon/GDP ratio

G O P p e r capita (in n a t u r a l l o g s )

Literacy 0.2 0.15 0.1 0.05

0

-0.05

Years 1960-2100

Source: Author's simulations

Figure 2a Differences for endogenous variables with and without remittances of the permanent growth model for countries with per capita income above $ 1200

Worker Remittances and Growth: The Physical and Human Capital Channels · 765

Savings/GDP ratto

r u n 1960-2100

Interest rate·

G r o s s fixed capital formation/GOP

Y·*» 1900-2190 Primary school *iwolm*nt

Literacy

V a n 1000-2100 GOP per capita (In natural log·)

Source: Author's simulations

Figure 2b Differences for endogenous variables of the permanent growth model for countries with per capita income below $1200

766 · T.H.W. Ziesemer

8

Summary and conclusion

The innovations of this paper are as follows. The main idea is that remittances enhance savings; savings do two things. First, they reduce interest rates, which encourage investment. Second, savings enhance either school enrolment or keep the school participation in tact by ensuring finance, thus enhancing literacy. Investment and literacy then enhance the level and the growth rate of the GDP per capita. These are the main channels in the model. We built a model of six equations from recent modern literature and add a seventh one for enrolment in primary schooling. Then we enhance the savings equation to include remittances. We extend the equation relating the current account and the interest rate by a quadratic term, the OECD GDP per capita growth rate and, for one of the samples, changes in US interest rates. The growth equation is enhanced to include lagged in addition to current investment and includes literacy as a human capital variable. All equations are estimated jointly as a system with pooled data using the General Method of Moments with heteroscedasticity and autocorrelation corrected standard errors, allowing for contemporaneous correlation of the residuals of the seven equations, for four different samples. The estimates show that the model works well for all the samples with only minor modifications mostly related to the functioning of capital markets. The number of insignificant variables is very low for each sample, especially along the main channel of the argument from remittances via savings to investment and literacy, with a slightly weaker performance for the countries with per capita income above $ 1200. Estimation results are as follows. Remittances have higher growth rates in poorer countries. Their change has a positive impact on savings, which is stronger in poorer countries. Among the non-OECD countries, labour force growth has a more negative and world income growth a more positive sign the poorer the countries are. The direct effect along the main channel of the argument is very small. The largest are observed for countries below per capita income of $ 1200. With 39 % of the growth rate of the GDP per capita remittances make a strong contribution to the growth of the countries below $ 1200. However, this should not lead to any sort of development optimism because the steady-state growth rate is 2.75 χ 10~ 4 with remittances instead of 1.9 χ 10~ 4 without remittances. It remains very small. Stability is shown through forward iteration of the model. The undiscounted transitional gains are lower than the undiscounted steady-state gains for all variables except for primary school enrolment and literacy of the countries below per capita income of $ 1200. Long-run and transitional effects of savings investment and literacy are stronger in the poorer sample. For the GDP per capita the log difference in the poorer sample with and without remittance changes is 2 % of the GDP per capita. As savings react much stronger to remittances than investment does, less debt is accumulated and less debt service is paid. This paper has not suggested anything for policy. In particular remittances are only one aspect of the brain drain or gain debate. 3 3 However, given the moderate performance of official development aid one gets the impression that remittances are more effective in enhancing growth. As a suggestion for future research we therefore like to raise the questions (i) whether or not remittances should be taxed less on both sides, the sender coun33

See Schiff ( 2 0 0 4 ) and Docquier ( 2 0 0 6 ) for sophisticated and well balanced discussions on the issue.

Worker Remittances and Growth: The Physical and Human Capital Channels · 767

tries (Ranis 2007) and the receiving countries (Chami et al. 2008), and (ii) whether or not this should be financed through a reduction of official development aid or through other means like reduction of inefficient subsidies or increases of efficient taxes. Chami et al. (2006) suggest that remittances should induce higher labour income taxes and inflation taxes in a model with an exogenous capital stock and no education. They argue that countercyclical remittances weaken the volatility smoothing forces for example by encouraging more leisure in a downturn. As a suggestion for further research we would like to ask whether this also holds under an endogenous capital stock. Then remittances perhaps have a positive impact on investment (and education) and might strengthen volatility smoothing forces.

768 · T.H.W. Ziesemer

Worker Remittances and Growth: The Physical and Human Capital Channels · 769

770 · T.H.W. Ziesemer

W o r k e r R e m i t t a n c e s a n d G r o w t h : T h e Physical a n d H u m a n Capital C h a n n e l s · 7 7 1

+

s

nj "8

+

δ

m S fo o

•c +

ι

M

Τ Q.

O

ltT r-S 5 Ò o +

«J

•Ό C

0. The same asylum reducing effect holds for increasing migration costs M w . 1 3 13

If the Western asylum country held no 'bias' for asylum seekers from a specific background, that is if Vw(p) is equal for asylum seekers from different countries, then asylum recognition rates tend ceteris paribus to be lower for asylum seekers from countries close to the Western destination, since migration costs for them are more moderate (Proof: ,

w

> 0 for dVw

= 0).

814 · M. Czaika

Proposition 4

A more liberal asylum policy in the "Western destination alleviates the refugee situation in the cross-the-border country, while rising migration costs for realizing the asylum option to the Western asylum country deflects refugee flows towards the cross-the-border country. As a consequence of a more restrictive asylum policy in the Western country, the inflow of a large refugee population aggravates the public perception within the first asylum country of refugees as an economic and political burden and a threat for the internal security of the civil society. Consequently, the first asylum country might intensify encampment and reduce economic self-reliance, worsening the refugees' situation (Loescher/Milner 2005). This policy response of the first asylum country is driven by the expectation that a more liberal encampment policy with enhanced opportunities for the economic self-reliance of the refugees would rather protract the refugee situation within its territory. Potential refugees expecting to live under appalling encampment conditions are then ceteris paribus more likely to repatriate earlier or to choose immediately the asylum option in a Western country. As a consequence, the Western country is likely to respond in an analogous manner by restricting asylum conditions (i.e., reducing recognition rates). Finally, a race to the bottom is established with highly restrictive asylum policies in the Western world and appalling refugee and encampment conditions in the conflict-affected developing world. The challenge of international refugee politics is to solve this apparent dilemma of restrictive asylum policies, which is, in economic terms, an inefficient equilibrium. Deterrence, deflection, detention, and deportation of refugees are counterproductive measures for resolving the appalling conditions for refugees, including the waste of resources for long-distance asylum migration. But beyond this, without dealing with the root causes of the refugee movements, these measures are also costly to the respective asylum countries, e.g. by increasing costs for border control, administration and maintenance due to the increasing propensity of refugees and asylum seekers to stay irregularly in the respective asylum country. An alternative to this bilateral asylum restriction policy might be a more proactive approach, particularly on the side of Western countries, that might tackle the

Asylum Country

The Political Economy of Refugee Migration · 815

underlying causes of asylum-seeking in both the country of origin and the first asylum country. 3.2 Proactive asylum policy: migration-preventive aid transfers The Western country may invest resources into proactive measures for tackling the root causes in the conflict-ridden country of origin or sharing the refugee-burden in the first asylum country, assisting local integration that might also reduce the asylum migration pressure to the Western country. Although there are various proactive policies available, the focus in this analysis is on migration-preventive aid transfers from the Western asylum country to the country of origin or the first asylum country. The crucial question of any proactive asylum policy in terms of migration-preventive aid is whether aid can indeed reduce asylum migration flows? Or, in terms of the present model, does an aid-induced increase of income levels in the country of origin or the first asylum country relieve the asylum burden in the Western country? The following analysis presumes that aid is to some extent income-effective. A necessary condition for this to be true is that aid, when transferred from the donor to the respective recipient country, is indeed channeled within the recipient country to the refugee population where aid can develop the income-generating effect. This is a necessary condition of the subsequent discussion. Aid to the first asylum country First asylum countries are often overstressed and unable to tackle the political and economic challenges that large refugee inflows provoke. A common reaction of these crossthe-border countries is to confine refugees within camps, denying them freedom of movement, access to social services, or economic self-reliance. Aid targeted at refugees in first asylum countries shall promote the living standards of the refugees. But what is the effect on the first asylum country itself? According to the previous model, an aid-induced increase in the refugees' income level implies that ceteris paribus the total refugee stock in the first asylum country increases:

0y s

{yO-asf)2

s(l-a)[yf

dys

'

However, the net outcome for the first asylum country depends on the effect of asylum migration to the Western country: 8 Vs dp*{**)

_

dy

s

V

dys

> 0 for p*, < 0 for p**,

~dp

'

9p~

dVw dVs with eqs (15) and by taking into account that ^ s = 0 and -^-y > 0. Income-increasing aid towards the first asylum country has an unambiguous reducing effect on the number of asylum seekers in the Western country.

816 · M. Czaika

Proposition 5 By channeling aid to a refugee population in an aid recipient country (and by being effective in increasing their respective income level), the asylum pressure to the Western dip** - p*) country does not increase: N——-¡-= < 0. (See also Figure3). ay> While this policy is beneficial to the Western country, it leaves the first asylum country with a higher number of refugees. Interestingly, although this policy might work for the interests of Western donor countries, they are nevertheless reluctant to provide more resources for that purpose (UNHCR 2006a, Czaika/Mayer 2008). Why? Possibly, Western countries don't consider aid to be as effective in generating additional income for the refugees or they take the negative consequences for the first asylum countries into account. Obviously, these first asylum countries are even more reluctant for Western countries to adopt this policy, because it shifts and consolidates the refugee-burden onto their territory. As a consequence, the first asylum country would respond by imposing further restrictions on economic self-reliance and encampment conditions for refugees. Finally, refugee-hosting developing countries might oppose the diversion of aid from the needs of the native population to the refugees within their territory. Aid to the country of origin Alternatively, the Western asylum country could transfer aid towards the country of origin in order to reduce refugee outflow and facilitate a sustainable voluntary repatriation. Pre-conditioned by the fact that countries of origin do not hinder a voluntary repatriation of their population after mass emigration, rapid post-conflict reconstruction with appropriate levels of civil security, basic social services and economic perspectives for returnees might be a prior rationale for targeting foreign aid towards the country of origin (UNHCR 2006a). Obviously, cross-the-border asylum countries are also interested in sustainable return solutions with voluntary repatriation and reintegration of refugees that would otherwise stay within their territory. Thus, aid to the country of origin that ensures rising income levels for the conflict-affected population is also in the interest of first asylum countries, since - without considering asylum migration to the Western country - the total stock of refugees in the first asylum country decreases, with the less persecuted people going home first: Ì L dy°

=

a


0, the overall effect on the Western country is ambiguous:14

ev w

dp*(**) ^

dv°

S-y° >