International migration and inequality across nations

Transcription

1 fondation pour les études et recherches sur le développement international International migration and inequality across nations Policy briefs from the conference "International labor mobility and inequality across nations" organized by Ferdi and Ires, January, , Clermont-Ferrand LA FERDI EST UNE FONDATION RECONNUE D UTILITÉ PUBLIQUE. ELLE MET EN ŒUVRE AVEC L IDDRI L INITIATIVE POUR LE DÉVELOPPEMENT ET LA GOUVERNANCE MONDIALE (IDGM). ELLE COORDONNE LE LABEX IDGM+ QUI L ASSOCIE AU CERDI ET À L IDDRI. CETTE PUBLICATION A BÉNÉFICIÉ D UNE AIDE DE L ÉTAT FRANCAIS GÉRÉE PAR L ANR AU TITRE DU PROGRAMME «INVESTISSEMENTS D AVENIR» PORTANT LA RÉFÉRENCE «ANR-10-LABX-14-01» policy brief October 2014 note brève

2

3 Overview Simone Bertoli CERDI,Université d Auvergne and CNRS Frédéric Docquier FNRS and IRES, Université Catholique de Louvain, FERDI The absence of detailed and high quality data was for a long time a major obstacle preventing academics from quantifying the extent of human capital mobility, its determinants, and its consequences for origin and destination countries. The increasing availability of new databases has enabled economists and demographers to characterize the global patterns of human migration and their impact on the world economy. Given the growing interest for these questions and their implications for development, FERDI and IRES organized an international conference on "International labor mobility and inequality across nations". The event was held at the Université d Auvergne in Clermont-Ferrand on January 23-24, This collection of policy briefs summarizes the findings of five invited contributions, which shed light on the effects of international migration on the geographic distribution of human capital and its implications for the world economy. In the first article, Çağlar Özden and Christopher Parsons use two recent bilateral data sets to identify statistical regularities in international migration. They show that the world distributions of immigrants and emigrants follow Zipf s law: the relative size of the nth largest corridor equals 1/n of the largest one, at least in the upper tail of the distribution. This property holds true for total and high-skilled migration and for all censuses, confirming that human capital tends to agglomerate in a small set number of rich countries where it is already abundant. The second article by Michel Beine, Simone Bertoli and Jesús Fernández-Huertas Moraga discusses the advantages of using these new bilateral databases, but it also highlights the dangers of misusing them. Recent gravity-based studies have quantified the effects of distance, visa restrictions, migrant networks, observed and expected income, and employment disparities on migration decisions. A sound identification of these effects requires use of micro-based specifications and appropriate econometric techniques. The authors provide a practitioners guide for future studies. In the third article by Maëlan Le Goff and Sara Salomone, the link between migration patterns and remittances is investigated. Remittances have become a key resource for many developing countries. This article first shows that the shares of college graduates and women have increased in migration flows from poor to rich countries. Using a unique bilateral database on bilateral remittances over the period , they also show that, on average, remittance receipts increase with education and with the share of women in the bilateral diaspora. In particular, college-educated women have the greatest propensity to remit. 3

4 4 The next two studies investigate the effect of migration policies and future population changes on global inequality. The fourth article by Frédéric Docquier and Joël Machado characterizes the global patterns of high-skilled migration and quantifies the effects of an increased competition for talents on the spatial distribution of income. The comparison of original data for actual, desired, and potential migrations reveals that Europe has been less effective than the US in mobilizing its potential high-skilled labor force, and is also globally less attractive. However a fierce competition for attracting talent would have a limited impact on the income gap between the EU15 and the United States. Some European countries such as Austria, Belgium, Germany, Greece and the Netherlands are less attractive than the US. If high-skilled migration was liberalized, they would see their income gap with the US increase. A fierce competition for attracting talent would adversely affect all developing regions, although skill-biased migration prospects stimulate human capital investments. The fifth paper by Andrew Mountford and Hillel Rapoport focuses on Africa. It examines the effects of rich countries immigration policies on fertility and education decisions in Africa. The percentage of the world population living in Africa is expected to increase drastically during the 21st century. Policies affecting Africa s population growth will have strong implications for the future changes in global inequality and poverty.

5 Sommaire Çağlar Özden, Christopher Parsons Zipf s law, Gibrat s law and international migration (B 101)... 7 Michel Beine, Simone Bertoli, Jesús Fernández-Huertas Moraga A practitioners guide to gravity models of international migration (B 102) Maëlan Le Goff, Sara Salomone Changes in Migration Patterns and Remittances: Do Female and Skilled Migrants remit more? (B 103) Frédéric Docquier, Joël Machado Global competition for attracting talents and the world economy (B 104) Andrew Mountford, Hillel Rapoport Migration Policy, African Population Growth and Global Inequality (B 105)... 31

6

7 FONDATION POUR LES ÉTUDES ET RECHERCHES SUR LE DÉVELOPPEMENT INTERNATIONAL Zipf s law, Gibrat s law and international migration * Çağlar ÖZDEN Christopher PARSONS Çağlar ÖZDEN, Development Economics Research Group, World Bank. Christopher PARSONS, Department of International Development, University of Oxford. This note details recent efforts to appraise the existence of Zipf s Law and Gibrat s Law when applied to international migration. In comparison with previous work, we examine aggregate and high-skilled immigration and emigration in levels and as percentages of the population. We provide evidence that a number of statistical regularities, including Zipf s Law, hold, which although often documented elsewhere in a variety of different contexts, nevertheless provide several statistical curiosities. These are interesting not least because the theoretical models used to explain Zipf s and Gibrat s laws elsewhere are not truly applicable when the unit of observation is a sovereign country, the borders of which are arguably controlled by the Government. LA FERDI EST UNE FONDATION RECONNUE D UTILITÉ PUBLIQUE. ELLE MET EN ŒUVRE AVEC L IDDRI L INITIATIVE POUR LE DÉVELOPPEMENT ET LA GOUVERNANCE MONDIALE (IDGM). ELLE COORDONNE LE LABEX IDGM+ QUI L ASSOCIE AU CERDI ET À L IDDRI. CETTE PUBLICATION A BÉNÉFICIÉ D UNE AIDE DE L ÉTAT FRANCAIS GÉRÉE PAR L ANR AU TITRE DU PROGRAMME «INVESTISSEMENTS D AVENIR» PORTANT LA RÉFÉRENCE «ANR-10-LABX-14-01» policy brief 101 October 2014 note brève * We are grateful to Ferdinand Rauch for useful and timely comments and would like to extend our gratitude to Simone Bertoli and Frédéric Docquier for their combined encouragement and for writing the technical background paper for this note. The usual disclaimer applies. The opinions expressed here are those of the authors and do not reflect the views of the World Bank, its executive directors or the countries they represent. Corresponding author: Chris Parsons, University of Oxford, Queen Elizabeth House, 3 Mansfield Road, Oxford OX1 3TB, United Kingdom. addresses: cozden@worldbank.org and christopher.parsons@qeh.ox.ac.uk.

8 Policy brief n 101 Çağlar Özden, Christopher Parsons 8 In urban economics, Zipf s Law and Gibrat s Law occupy a special place in explaining the relationships between cities in terms of their relative sizes and growth rates. The role of this note is to establish some stylized facts and preliminary analysis to link international migration to urban economics, especially the two issues of population growth and distribution. More specifically, we draw on two of the most recent advances in the development of bilateral migration data, Özden et al. (2011) and Artuç et al. (2014), to examine the relevance of Gibrat's and Zipf's Laws for immigrant and emigrant numbers and densities in aggregate, as well as examining migration patterns for the high-skilled. Zipf's law for cities states that the city size distribution within a country can be approximated by a power law distribution. More specifically, if we were to rank the cities in terms of their sizes, the nth largest city is 1/n of the size of the largest city. Another way to look at this rule is by running a regression with the natural log of city size rank as the dependent variable and the natural log of city size as the main explanatory variable. A coefficient of 1 provides evidence that Zipf s Law holds. This is typically the case with a high level of precision, especially when larger cities are considered. Gibrat's law, on the other hand, is about growth rates and was initially noted for French firms (Gibrat, 1931). When applied to cities, it states that growth processes have a common mean and are independent of initial sizes. The seminal paper of Gabaix (1999) establishes that, when cities grow according to Gibrat's law, then, in steady state, their size distribution will follow a Pareto distribution with a power exponent of one. This, as well as numerous results from across the urban economics literature, which also show that Zipf's Law holds, provides some evidence against the applicability of Gibrat's Law, which predicts that the resulting distribution is log-normal. These patterns are observed across the world and are discussed in great detail in many other places (such as Gabaix, 2004). The natural question then is to ask what mechanisms exist that might link Gibrat's and Zipf's laws. This question is especially important when divergent rates of population growth across locations are the norm. Among the causes that would lead to differing growth rates are climate, natural disasters and resource endowments. There are numerous examples of cities that disappear from history or boom after the discovery of a natural resource. People however, have the ability to move from one location to another, especially when some locations become overcrowded and resources are restrained due to higher natural growth rates. Under these circumstances, the convergence of growth rates via migration would naturally take place within a larger geographic area (such as a country) with no internal barriers to population mobility. There have been few studies that take the analysis beyond national boundaries since the theoretical models used to explain Zipf's and Gibrat's laws are not truly applicable when the unit of observation is a sovereign country. Governments exercise considerable power over their international (as opposed to internal) borders and can dictate who can enter and, to a certain extent, exit the country. The most prominent study is Rose (2005), who tests these theories using countries as the relevant geographical entities. He concludes that the `hypothesis of no effect of size on growth usually cannot be rejected. When testing for Zipf's law for countries, he shows that it also strongly holds in the upper tail of the distribution of country sizes, as is the case with city sizes. The only study that links Gibrat's and Zipf's laws to international immigration is the work by Clemente et al. (2011), which uses aggregate immigration numbers and densities and finds, for the biggest 50 countries, that Zipf's Law holds only for immigrant numbers. We draw upon two recent bilateral migration databases, Özden et al. (2011) and Artuç et al. (2014). These yield two advantages. First, global bilateral migration data allow us to calculate country level emigrant stocks, as opposed

9 to simply immigrant stocks. Secondly, the latter database, which reports bilateral migration stocks by education levels, allows us to further delineate between patterns of migrants with low and high education levels. Zipf's law implies a concentrated distribution of the total population among a few large cities. We observe similar patterns in immigration patterns but the converse over time in emigration patterns. In 2000, the top ten receiving countries accounted for no less than 57 percent of the world migrant stock, which is approximately equivalent to the total emigrant stocks of the top 25 emigration countries in the same decade. In 1960, the figure for the top ten receiving nations was 54 percent, which is equivalent to less than the total emigrant stock of the top 9 sending countries in A similar concentration is observed for individual corridors. In 2000, 90 percent of all bilateral corridors (over 40,000 in total) comprised fewer than 50 migrants each. Together, they accounted for only 0.1 percent of total migrant stock. On the other hand, in the same year, just 505 corridors accounted for over 80 percent of the global migrant stock. Whether the sum of these bilateral trends holds across countries is the subject of our analysis. We delve deeper into the observed patterns of international migration over the period , by examining the existence of Zipf's and Gibrat's Laws, two empirical regularities that are ubiquitous in the urban economics literature. We therefore adopt an alternative perspective in analyzing to what extent the underlying trends in global migration patterns are converging or diverging. The results are both interesting and encouraging. With regards to Zipf's law, we find that it holds in the upper tail of the distribution with the Pareto coefficient very close to unity for both immigration and emigration for all time periods. The results are less precise for high skilled migration but even so we cannot reject the hypothesis of the coefficient being unity. When the whole sample is included, we find a log-normal distribution for immigrant numbers, which are very similar to the results found for cities and countries in the literature. With respect to Gibrat's law, our results are somewhat less uniform. We find evidence of convergence with the growth rates being linked negatively to initial levels for both aggregate and high-skilled immigration and emigration numbers and densities. In our non-parametric analysis, we find some evidence in favor of Gibrat's Law holding for immigration stocks, i.e. that the growth in stocks is independent of their initial values and stronger evidence that immigration densities are diverging over time. Conversely, emigrant stocks are converging in the sense that countries with smaller emigrant stocks are growing faster than their larger sovereign counterparts. These are all surprising regularities given that government policies and other physical and cultural barriers impose strong restrictions on international migration patterns. Zipf s law in migration We begin our analysis with an examination of the existence of Zipf's Law, the rank-size rule, which describes the (upper tail) of the population distribution of the geographical entity under investigation. Typically the existence of Zipf's Law is analyzed graphically and/or using regression techniques. Both approaches first rank the (population) size variable of interest S, from highest to lowest. Then the natural logarithm of the rank variable is analyzed with respect to the natural logarithm of the (population) size variable. The left hand two panels in Figure 1 show the graphical scatter plots of the natural logarithms of the rank of total immigrant and emigrant levels for the 50 countries in the top tail of the various distributions. Although we only plot these for the year 2000, these graphs are also representative of other years. The data in the right hand two panels refer to high-skilled migrants, defined as having completed at least Policy brief n 101 Çağlar Özden, Christopher Parsons 9

10 Policy brief n 101 Çağlar Özden, Christopher Parsons 10 one year of tertiary education. Clear linear trends are evident, but the line imposed demonstrates that some deviations from Zipf's Law clearly exist in the data. For a more robust analysis, simple regression analysis is performed the details of which are presented in the background paper of this note. The results from the regression analysis show that the Pareto coefficient is remarkably close to unity for the 50 largest countries in the upper tail of the size distribution, although Zipf's Law is resolutely rejected across the entire distribution since the coefficient is for the year Turning to the high-skill migration numbers, the point estimates on the Zipf coefficient are substantively different from unity but still within the confidence intervals. When we consider the entire distribution of bilateral migrant stocks, we easily reject Zipf s law. The question is whether the distribution would satisfy log-normality assumptions. This would imply that Gibrat s law held and that an independent growth process might lead to this distribution. Our empirical analysis shows that both immigrant numbers and densities are often log-normally distributed. growth and distribution processes of populations over national boundaries and across the globe. In our empirical analysis, we estimate parametric and non-parametric kernel regressions to test for the applicability of Gibrat's Law. These regress the size of migrant populations (or in our case densities) on their growth. Across all decades and across all measures, the parametric regression results are negative and statistically significant or else statistically insignificant. These results are indicative of convergence over time or indeed of Gibrat's Law holding. The major exceptions are those countries in the lower end of the distribution, which tend to grow faster than the other countries. The composition of high skilled immigrant and emigrant levels and densities exhibit signs of convergence, albeit to differing degrees. These patterns are consistent with the onset of the global competition for international talent, an increasing global supply of high-skilled workers across all origins, limited supplies per origin as well as the imposition of selective immigration policies by an increasing number of destination countries. Gibrat s law in migration As opposed to the static analysis presented when discussing Zipf's Law, an examination of Gibrat's Law with growth rates, requires a dynamic analysis of global migration movements. Should the growth of geographical entities be independent of their size, their growth will subsequently result in a log-normal distribution. Of course, this does not mean that a log-normal distribution implies that Gibrat's Law necessarily holds. Both immigration and emigration are determined by government policies and limited to the extent by which individuals are willing and able to move. Internal mobility, on the other hand, is far less restricted and is typically thought to follow Zipf's and Gibrat's laws. Thus, it is difficult to argue that some underlying law of nature might determine the underlying

11 Figure 1. Zipf Plots for Aggregate and High-skilled Immigrants and Emigrants, 2000 Log Rank Log Rank Aggregate Immigrant Levels Log Stock Aggregate Emigrant Levels Log Rank Log Rank High-skill Immigrant Levels High-skill Emigrant Levels Log Stock Policy brief n 101 Çağlar Özden, Christopher Parsons 11 Log Stock Log Stock Conclusion Gibrat's and Zipf's laws are among the most studied and well established phenomenon in various contexts including linguistics, firm sizes and urban agglomerations. The linkages between population growth and the distribution across geographic space are key to all economic analysis since economic activity cannot be analyzed in isolation from location. Thus, it is important to study and identify the underlying processes that determine the growth rates of populations over time and their allocation across various locations. Even though Zipf's and Gibrat's laws have been extensively analyzed in the literature, there are fewer studies on the role of population movements and there are even fewer studies that focus on mobility across national boundaries. It is therefore natural to search for a relationship between such population laws and migration patterns.

12 Policy brief n 101 Çağlar Özden, Christopher Parsons 12 References Artuç, E., Docquier F., Özden C., Parsons C. (2014). A Global Assessment of Human Capital Mobility: the Role of non- OECD Destinations. World Development. Clemente, J., Gonzalez-Val R., Olloqui I. (2011). Zipf s and Gibrat s laws for migrations. The Annals of Regional Science 47, Gabaix, X. (1999). Zipf's Law for Cities: An Explanation. Quarterly Journal of Economics 114, Gabaix, X., Ioannides Y. M. (2004). The evolution of city size distributions. In: J. V. Henderson and J. F. Thisse (eds.), Handbook of Regional and Urban Economics, vol. 4, chap. 53, , Elsevier. Özden, C., Parsons C. R., Schiff M., Walmsley T. L. (2011). Where on Earth is Everybody? The Evolution of Global Bilateral Migration World Bank Economic Review 25(1), Rose, A. K. (2005). Cities and Countries. NBER Working Papers

13 FONDATION POUR LES ÉTUDES ET RECHERCHES SUR LE DÉVELOPPEMENT INTERNATIONAL policy brief 102 October 2014 note brève A practitioners guide to gravity models of international migration * Michel BEINE Simone BERTOLI Jesús FERNÁNDEZ-HUERTAS MORAGA Michel BEINE, CREA, University of Luxembourg. Simone BERTOLI, CERDI, Université d Auvergne and CNRS. Jesús FERNÁNDEZ-HUERTAS MORAGA, FEDEA and IAE, CSIC. In international macroeconomics, bilateral data refers to data capturing flows of economic activity between a specific origin and a specific destination. Origin and destination refer to a given geographical area, such as a city, a region, a country or even a set of countries, which share some common features. The scope of economic activities involving bilateral flows is quite large. It concerns, for instance, international trade, international migration, foreign direct investments, and portfolio investments, just to name the most prominent examples. This paper deals with the use of bilateral data in the context of international migration. The academic literature has recently made extensive use of this type of data to identify the potential determinants of the magnitude and the composition of the international flows of economic migrants between countries. To that aim, the literature has mainly relied on a very popular tool, the gravity model. * Corresponding author: Jesús Fernández-Huertas Moraga, FEDEA, Jorge Juan, 46, E-28001, Madrid, Spain. addresses: michel.beine@uni.lu, simone.bertoli@udamail.fr and jfernandezhuertas@fedea.es. LA FERDI EST UNE FONDATION RECONNUE D UTILITÉ PUBLIQUE. ELLE MET EN ŒUVRE AVEC L IDDRI L INITIATIVE POUR LE DÉVELOPPEMENT ET LA GOUVERNANCE MONDIALE (IDGM). ELLE COORDONNE LE LABEX IDGM+ QUI L ASSOCIE AU CERDI ET À L IDDRI. CETTE PUBLICATION A BÉNÉFICIÉ D UNE AIDE DE L ÉTAT FRANCAIS GÉRÉE PAR L ANR AU TITRE DU PROGRAMME «INVESTISSEMENTS D AVENIR» PORTANT LA RÉFÉRENCE «ANR-10-LABX-14-01»

14 Policy brief n 102 M. Beine, S. Bertoli, J. Fernández-Huertas Moraga 14 Bilateral data on migration are a blessing Compared to data that are specific only to one destination or to one origin country, the use of bilateral data yields a set of obvious advantages for the analysis of international migration, and it allows capturing specific phenomena of interest. In other words, the use of bilateral data is a blessing. Using bilateral data allows for analysis of the impact of factors that are pair specific in nature. Examples of this include the analysis of migrants networks in shaping future migration flows. The analysis of the influence of bilateral migration policies, such as the impact of visa restrictions, the effect of bilateral agreements, or even the influence of multilateral treaties such as the Schengen agreement, is only possible if bilateral measures of flows and policies are available. Furthermore, the combination of information at the origin level, at the destination level, and over time yields econometric models with many dimensions. These dimensions, in turn, permit the use of econometric panel data approaches with a large set of so-called fixed effects, which capture the role of specific factors that are either difficult to measure or even impossible to quantify in a satisfying way (such as cultural proximity between countries). As a consequence, when treated in an appropriate way, the use of bilateral data yields econometric models that are less subject to modelling errors. Bilateral data on migration are a curse While the access to bilateral data is a blessing in many respects, it is nevertheless also a curse. Indeed, the analysis of bilateral data requires appropriate statistical tools such as the gravity model that make the best use of the information embedded in the dyadic dimension. The breaking down of the migration process into various destinations of emigration calls for a sound theoretical treatment of the individual choices of the prospective migrants in terms of optimal destination. In that respect the dominant theoretical approach has been the Random Utility Maximization (RUM) approach, which has been developed over the last few decades by a number of scholars. In a nutshell, the RUM model assumes that agents maximize their utility, which varies across a large set of domestic and foreign destinations. In the canonical RUM model, location-specific utility contains both a deterministic and a stochastic component. The deterministic component, which describes the attractiveness of each location for potential migrants, is modeled as a function of income variables, amenities, and the specific migration costs. The stochastic components are individual-specific, and the modeling approach assumes convenient statistical properties such as the type of stochastic distribution and a constant substitution rate between alternative destinations. The key implication of our paper is that the econometric treatment must be based on the relationships as implied by the theory, and must account for any deviation between the underlying hypotheses of the theory and what the data do and do not capture. Our practitioners guide lists the main dangers faced by an econometric approach using the dyadic data for international migration, in the same spirit as what has recently been done for trade by Head and Mayer (2014). The dangers of falling with gravity The main issues and dangers with the use of a gravity model are the following. First, gravity equations used without reference to the theory face the danger of not using the appropriate form of the key variables, such as the bilateral migration rate. In general, our paper argues that a micro-based econometric model should be favored instead of a so called a-theoretical specification. Another danger is that the inclusion of the deterministic components might be insufficient to make the stochastic component

15 consistent with the underlying assumptions of the theory. Fortunately, econometric solutions to this issue exist, but imply that researchers invest beyond the simple econometric methods that have been mainly used so far. One obviously desirable development of the literature is to include approaches that account in a satisfying way for the issues of multilateral resistance of migration (Bertoli and Fernández-Huertas Moraga, 2013). Finally, there are specific econometric issues such as the large presence of zeroes in the bilateral migration flows. This calls for the use of specific estimation methods such as the Pseudo Poisson Maximum Likelihood approach. What gravity models have produced so far While many challenges are still ahead of us, the existing literature has not been inactive in terms of pinning down the determinants of migration flows. The empirical analyses can be classified into three major areas depending on the dimension of the factors that capture the limelight in the analysis. A first strand looks at the factors that are specific to the origin countries, a second one focuses on the destination-specific determinants, and thirdly a large body of the literature looks at the role of factors that are specific to each migration corridor. Origins matter One of the most important factors is the income at origin. Income at origin has an ambiguous role on the magnitude of migration. On the one hand, an increase in income at origin reduces the wage differential with all foreign destinations, so lowering the magnitude of emigration. On the other hand, in poor origin countries, low income is often associated with liquidity constraints that prevent many candidates moving out of the country. In many countries, these liquidity constraints are strengthened by important credit constraints affecting the prospective migrants. In this context, an improvement in the economic conditions at origin might therefore lead to a surge in the emigration rate. Another factor that has received little attention in the academic literature is the unemployment rate. While some papers have allocated some analysis to the specific role of unemployment at origin, refinements constitute a desirable avenue of research. One of the interesting directions would be to look at the specific effect(s) of unemployment benefits in the origin country. Other factors at origin are potentially important. One kind of determinant includes the so-called environmental and climatic shocks that affect origin countries, especially developing ones. A growing literature has been devoted to their global role, highlighting some existing indirect effects (e.g. through wages), rather than direct ones. In particular, adverse climatic shocks and natural disasters tend to lead to more emigration, mainly through a deterioration of the labor market conditions in the origin countries, rather than through a more direct reaction of the affected people. Other potential determinants that have so far received little attention include the institutional factors or the demographic structure of the origin countries. Destinations matter Variations of income and unemployment at destination also play a very important role in the explanation of migration flows. With respect to income, wage differential can be regarded as one of the most robust determinants found in the literature. This does not mean however that all studies find the same magnitude of the effect. There are several reasons for this. A first main reason is related to the empirical measurement of wage differentials. In order to have a large geographical coverage researchers have often to rely on GDP per head as a proxy of wages. This overlooks the possible discrepancy between the two measures, and the variation in wages within a given country. A second main reason for the variation in the findings relates to the specific form of the econometric model used. The literature assumes that what matters is either the relative difference in wages, or in contrast the absolute difference. Policy brief n 102 M. Beine, S. Bertoli, J. Fernández-Huertas Moraga 15

16 Policy brief n 102 M. Beine, S. Bertoli, J. Fernández-Huertas Moraga 16 Expectations of income and unemployment are crucial to explain the flows of international migrants. In the presence of risk aversion, prospective migrants need to form expectations about the future state of the economy, at home and abroad. A few recent papers such as Bertoli et al. (2013) and Beine et al. (2013) have empirically looked at this aspect, connecting expectations with business cycles and aggregate fluctuations. These papers support a role for adverse economic shocks, such as the financial crisis, in explaining the magnitude and the direction of the international flows of workers between countries. Migration corridors definitely matter The empirical literature is more silent on the role of general immigration policies as determinants of international migration. The reason for that is the scarcity of empirical measures of restrictiveness that are policy based. Efforts are clearly needed to create indicators of stringency on international mobility of people that are comparable across countries and over time. In contrast, the recent literature has integrated new indicators of bilateral immigration policies, i.e. policies that are specific to a migration corridor. Two main indicators have been used recently. One is the existence of bilateral visa requirements (or waivers). The second one concerns multilateral agreements between a subset of countries, such as the Schengen agreement between some European countries. The recent literature has supported the significant impact of these policies. From that perspective, general integration processes such as trade agreements between countries, or monetary unification in a given area, have also been found to spur international mobility. At the bilateral level, other factors have also been found to be important to explain international migration. One of the main factors is definitely the size of the migrants network at destination. Diasporas tend to reduce migration costs for prospective migrants, especially the unskilled ones. The literature based on gravity models has confirmed that these diasporas have a significant role at the macro-economic level (see, for instance, Beine et al., 2011). Networks also substitute nowadays for factors that were at work in previous decades such as colonial links between countries. Nevertheless other dyadic factors are still important, as demonstrated by numerous empirical studies. Linguistic proximity stands high on this list. Even in a globalized world with English as the lingua franca, people tend to head to countries that share their mother language, or countries that belong to the same linguistic family. Cultural proximity is also at work in this context, although cross-country measures are less straightforward in this case. References Beine, M., Bourgeon P., Bricongne J.-C. (2013). Aggregate Fluctuations and International Migration. CESifo Working Paper No. 4379, Munich. Beine, M., Docquier F. and Özden Ç. (2011). Diasporas. Journal of Development Economics 95(1), Bertoli, S., Fernández-Huertas Moraga J. (2013). Multilateral Resistance to Migration. Journal of Development Economics 102, Bertoli, S., Brücker H., Fernández- Huertas Moraga J. (2013). The European Crisis and Migration to Germany: Expectations and the Diversion of Migration Flows. Working Paper No. 2013/21, CERDI, University of Auvergne. Head, K., Mayer T. (2014). Gravity Equations: Workhorse, Toolkit, and Cookbook. Chapter to appear in Handbook of International Economics, Vol. 4, eds. Gopinath, Helpman, and Rogoff, Elsevier.

17 FONDATION POUR LES ÉTUDES ET RECHERCHES SUR LE DÉVELOPPEMENT INTERNATIONAL policy brief 103 October 2014 note brève Changes in Migration Patterns and Remittances: Do Female and Skilled Migrants remit more? * Maëlan Le Goff Sara SALOMONE Maëlan LE GOFF, CEPII. Sara SALOMONE, IRES, Université Catholique de Louvain. Migrants remittances to developing countries have significantly increased and are the second largest source of finance for developing countries after foreign direct investment. Also, the composition of international migration flows has been characterized by a growing feminization and skill intensity. In reviewing the literature on remittances, we argue that these two recent phenomena cannot be ignored if one aims to explain the role of remittances as a lifeline for developing countries. Using an original dataset on bilateral remittances and estimating a gravity model in which the gender and skill dimensions of the migrants are accounted for, we show that on average, annual remittances received by origin countries increase with the share of women and college graduates in the bilateral corridor. This effect is mainly driven by the presence of high-skilled female migrants. * Corresponding author: Sara Salomone, IRES-UCLouvain, Place Montesquieu, 3, B-1348, Louvain-La- Neuve, Belgium. addresses: sara.salomone@uclouvain.be, maelan.legoff@cepii.fr. LA FERDI EST UNE FONDATION RECONNUE D UTILITÉ PUBLIQUE. ELLE MET EN ŒUVRE AVEC L IDDRI L INITIATIVE POUR LE DÉVELOPPEMENT ET LA GOUVERNANCE MONDIALE (IDGM). ELLE COORDONNE LE LABEX IDGM+ QUI L ASSOCIE AU CERDI ET À L IDDRI. CETTE PUBLICATION A BÉNÉFICIÉ D UNE AIDE DE L ÉTAT FRANCAIS GÉRÉE PAR L ANR AU TITRE DU PROGRAMME «INVESTISSEMENTS D AVENIR» PORTANT LA RÉFÉRENCE «ANR-10-LABX-14-01»

18 Policy brief n 103 Maëlan Le Goff & Sara Salomone 18 According to recent World Bank estimates, officially recorded remittances to developing countries were USD 401 billion in 2012, and remain a key resource flow that far exceeds official development assistance, private debt, and portfolio equity. Moreover, this flow is expected to grow by 12.3 percent and reach $515 billion in 2015, thus sustaining growth and development in developing countries and serving as a lifeline for the poor. However, to get a more comprehensive picture on the effects and the importance of remittances, the migrants gender and educational characteristics have to be taken into account as suggested by recent patterns of international migration. A growing share of female and skilled migrants In the last few decades, international migration has been characterized by a rapidly increasing feminization and skill intensity, as shown on Figure 2. On the one hand, women started to migrate more independently, in search of better work opportunities, rather than for family reunification reasons or as dependent of male migrants. This led the global number of females living outside their origin country to become almost equal to that of men. Using bilateral migration data presented in Artuç et al. (forthcoming), we observe that the mean level of the female share in migration equals 37 percent (48 percent in high income destinations and 16 percent in poor countries), showing a clear feminization of migration to developed countries. At the regional level, migrant women numerically outnumbered men in Europe, Latin America and the Caribbean, North America, and Oceania. In Africa and Asia they were generally under-represented as compared to men. In 2000 females represented the majority of migrants from several countries: El Salvador (60 percent), Philippines (64 percent), Belorussia (65 percent), Thailand (75 percent), among others. The most feminized corridor appears to be Dominican Republic-Greece, while the largest in terms of high-skilled women is Tajikistan-Portugal. On the other hand, the quality-selective immigration based policies implemented in many traditional destinations (such as Canada, United States, Germany and United Kingdom) boosted the immigration flows of highly skilled people. Data from Docquier et al. (2013) reveal that between 1990 and 2000, the number of skilled immigrants in OECD countries increased by 64 percent. The rise was stronger for immigrants from developing countries (+93 percent), especially from Africa (+113 percent) and Latin America & Caribbean (+97 percent). The mean level of the high-skill share in 2000 was 23 percent. This proportion increases slightly when considering only high-income countries (30 percent); conversely, it drops dramatically in low-income destinations (6 percent). Portrait of remittances drawn from a new bilateral database In this paper, data for bilateral remittances are drawn from Docquier et al. (2012). To the best of our knowledge, this is the most comprehensive bilateral dataset currently available, which documents bilateral remittances. This bilateral data set combines five existing bilateral databases constructed by other authors or organizations: the EU data set, the IMF database, the Romanian data, the IDB database built by the Inter-American Development Bank, and the ECB database. The merging procedure, as described at length in Docquier et al. (2012), prevents possible selfselection problems, and leads to an annual data set with 8,928 observations for 1,348 country pairs over the period This new data set reveals that the most important corridors are Philippines-United States, Bangladesh-Saudi Arabia, Philippines-Saudi Arabia and Philippines-Japan. The largest remittance

19 recipient is Philippines with more than 6 billion US dollars per year. As regards remittances per migrant, the data show that in 2000 the average level of remittances per migrant amounted to USD 1,023. However, the average amount remitted varies widely with the characteristics of migrants. On average, college-educated and female migrants remit more, with around USD 4,509 and USD 2,550 per individual, respectively. Also, variations in remittances per migrant are very large, depending on the development level of the host country: from USD 25 per migrant in low-income host countries to USD 1,025 dollars per individual in high-income host countries. New migration patterns and remittances A broad strand of literature on the determinants of remittances has focused on the role played by migrants characteristics. Several papers have studied the financial contribution of female migrant workers to their country of origin, emphasizing how men and women differ in terms of propensity to remit. On the one hand, females may remit more both in USD and as a percentage of income because they tend to maintain stronger ties with their restricted and extended family, and because their remitting behavior is driven more by altruism (Orozco et al., 2006). On the other hand, women may remit less because of disadvantaged labor market conditions (Cortes, 2013) and family reunifications. Policy brief n 103 Maëlan Le Goff & Sara Salomone 19 Figure 2. High-skilled and female migration in top host countries ( ) % High-skilled migrants % Female migrants

20 Policy brief n 103 Maëlan Le Goff & Sara Salomone 20 Table 1. Poisson estimations accounting for gender and education Poisson R ijdt (1) Number of migrants at dest. (log) 0.771*** (0.033) Share of females at dest *** (0.382) Poisson R ijdt (2) 0.820*** (0.036) Share of skilled people at dest * (0.552) Poisson R ijdt (3) 0.795*** (0.035) 1.161*** (0.375) (0.548) Poisson R ijdt (4) 0.795*** (0.033) Poisson R ijdt (5) 0.778** (0.033) Share of skilled females at dest *** (0.633) Share of unskilled females at dest * (0.388) GDP at origin (log) 2.417*** (0.875) GDP at dest. (log) 1.153*** (0.318) Interest Rate (0.015) Exchange Rate (0.019) Contiguity 0.962*** (0.311) Distance (log) (0.016) Common language (dummy) 0.417*** (0.159) Colony (dummy) 0.808*** (0.168) Constant ** (16.47) 2.336*** (0.894) 1.529*** (0.320) (0.015) (0.018) 1.123*** (0.289) (0.016) 0.375*** (0.168) 0.658** (0.167) ** (17.21) 2.444*** (0.876) 1.152*** (0.321) (0.015) (0.019) 0.944*** (0.311) (0.016) 0.337* (0.173) 0.812*** (0.168) ** (16.43) 2.510*** (0.887) 1.147*** (0.319) (0.015) (0.018) 1.134*** (0.289) (0.016) (0.171) 0.767*** (0.170) ** (16.71) 2.300*** (0.844) 1.554*** (0.314) (0.015) (0.019) 1.054*** (0.304) (0.016) 0.521*** (0.156) 0.740*** (0.163) ** (16.64) Pseudo log-likelihood -4.58E E E E E+13 Observations 8,888 8,888 8,888 8,888 8,888 R-squared Robust standard errors in parentheses below the estimates. Significant at 10% level ** 5% level *** 1% level. All regressions include origin, destination, year and database fixed effects. The effect of migrants education is also unclear. At the macro and micro level, it has been found that migrants remittances decrease with the proportion of skilled individuals among emigrants. Different conclusions were reached by Docquier et al. (2012) who highlighted the role of immigration policies in determining the sign and magnitude of the relationship between remittances and migrants education. At a micro level, Bollard et al. (2011) found, a mixed pattern between higher education and the likelihood of remitting, and a strong positive relationship between higher education and the amount remitted (conditional on remitting). Our paper sheds light on this issue, relying on bilateral remittance data and the same gravity model as in Docquier et al. (2012). Using the Poisson pseudo-maximum likelihood estimator, which addresses the issue of the large proportion of zeroes in our dependent variable, we regress the amount of remittances on the share of women, the share of college-educated migrants, and a combination of both. Our results, presented in Table 1, show that the share of females and the share of college graduates at destination are always positive and significant. We find that a 1 percent increase in the share of women in the migration diaspora leads to an 11

21 percent increase in annual remittances. A 1 percent increase in the share of college graduates leads to an 8 percent increase in annual remittances. Moreover, the number of migrants at destination is also positive and significant. As regards the validity of the gravity factors (see Lueth and Ruiz-Arranz, 2008), geographical distance is negative but not significant, linguistic and cultural proximity are positive and highly significant. As far as nominal GDP variables are concerned, the positive sign for both the level of GDP at origin and that at destination confirms that larger countries receive (and send) larger amounts of remittances. Finally, as an additional empirical exercise, the shares of high-skilled and low-skilled women in the diaspora are presented in columns 4 and 5. It shows that the remitting propensity by females is clearly more for college-educated women than for less educated women. A single-unit change in the share of females and skilled in the migration diaspora leads to an increase of 2.3% and 1.6% in bilateral remittances. If skilled females are instead considered, remittances increase by 8.7%. In summary, the estimation of a gravity model of remittances reveals that the education level and gender of migrants matter significantly in determining the size of remittance flows. We show that the shares of females and collegeeducated in the diaspora significantly increase annual remittances received in the country of origin. Moreover, we find that the positive effect of women s share on remittances is mainly driven by high-skilled women relative to less educated ones. Given these findings, we can expect a positive effect of the recent feminization and the increasing education of migrants on annual remittance inflows received by developing countries. References Artuç, E., Docquier F., Özden Ç., Parsons C. (2014). A global assessment of human capital mobility: the role of non- OECD destinations. World Development, forthcoming Bollard, A., McKenzie D., Morten M., Rapoport H. (2011). Remittances and the Brain Drain Revisited: The Microdata Show that More Educated Migrants Remit More, World Bank Economic Review, World Bank Group, 25(1), Cortes, P. (2013). The Feminization of International Migration and its effects on the Children Left behind: Evidence from the Philippines, World Development, forthcoming. Docquier, F., Rapoport H., Salomone S. (2012). Remittances, Migrants' Education and Immigration Policy: Theory and Evidence from Bilateral Data, Regional Science and Urban Economics, 42(5), Lueth, E., Ruiz-Arranz M. (2008). Determinants of Bilateral Remittance Flows', B.E. Journal of Macroeconomics, 8(1), Orozco, M., Lowell B.L., Schneider J. (2006). Gender-Specific Determinants of Remittances: Differences in Structure and Motivation, Report to the World Bank Group Gender and Development Group, PREM, Policy brief n 103 Maëlan Le Goff & Sara Salomone 21

22

23 FONDATION POUR LES ÉTUDES ET RECHERCHES SUR LE DÉVELOPPEMENT INTERNATIONAL policy brief 104 October 2014 note brève Global competition for attracting talents and the world economy * Frédéric DOCQUIER Joël MACHADO Frédéric DOCQUIER, FNRS and IRES, Université Catholique de Louvain, FERDI Joël MACHADO, IRES, Université Catholique de Louvain. We examine the effects of a worldwide liberalization of high-skilled migration on the spatial distribution of income. Using Gallup survey data, we first identify the origin and preferred destination of all college-educated potential migrants in the world. The analysis of this database reveals that Europe has been less effective than the US in mobilizing its potential high-skilled labor force, but is also globally less attractive. Hence, a fierce competition for attracting talents would have a limited impact on the income gap between the EU15 and the US, and would be more beneficial for Canada and Australia. European countries such as Austria, Belgium, Germany, Greece, Luxembourg and the Netherlands are less attractive and would see their income gap with the US increasing. We also show that liberalizing highskilled migration would decrease income per worker in the developing world by 4.0 percent to 6.7 percent, although we account for a positive effect of skill-biased migration prospects on education decisions. * Corresponding author: Joël Machado, IRES-UCLouvain, Place Montesquieu, 3, B-1348, Louvain-La-Neuve, Belgium. addresses: frederic.docquier@uclouvain.be, joel.machado@uclouvain.be. LA FERDI EST UNE FONDATION RECONNUE D UTILITÉ PUBLIQUE. ELLE MET EN ŒUVRE AVEC L IDDRI L INITIATIVE POUR LE DÉVELOPPEMENT ET LA GOUVERNANCE MONDIALE (IDGM). ELLE COORDONNE LE LABEX IDGM+ QUI L ASSOCIE AU CERDI ET À L IDDRI. CETTE PUBLICATION A BÉNÉFICIÉ D UNE AIDE DE L ÉTAT FRANCAIS GÉRÉE PAR L ANR AU TITRE DU PROGRAMME «INVESTISSEMENTS D AVENIR» PORTANT LA RÉFÉRENCE «ANR-10-LABX-14-01»