Remittances and the Brain Drain: Evidence from Microdata for Sub-Saharan Africa Julia Bredtmann 1, Fernanda Martinez Flores 1,2, and Sebastian Otten 1,2,3 1 RWI, Rheinisch-Westfälisches Institut für Wirtschaftsforschung 2 Ruhr University Bochum 3 University College London This version: January 2016 Abstract Using comprehensive microdata from five Sub-Saharan African countries, we investigate the effect of education on migrants remittance behavior. By observing multiple migrants per household, we are able to control for unobserved characteristics of the household at the origin country and thus to identify an unbiased effect of education on migrants remittances. Our results show that migrants education has no significant impact on the likelihood of sending remittances. However, we find a strong positive effect of tertiary education on the amount of remittances sent, providing evidence that high-skilled migrants send more money to the households left behind. JEL Classifications: F22, F24, O15 Keywords: migration, remittances, skill level, brain drain, Sub-Saharan Africa Preliminary version: please do not cite. All correspondence to: Fernanda Martinez Flores, Rheinisch- Westfälisches Institut für Wirtschaftsforschung, Hohenzollerstr. 1-3, 45128 Essen, Germany, Email: fernanda.martinez@rwi-essen.de.
1 Introduction In 2013, one in every nine tertiary-educated persons born in Africa resided in an OECD country, representing the highest emigration rate among developing regions with the exception of Latin America and the Caribbean OECD-UNDESA (2013). In Sub-Saharan Africa, the emigration rate of high-skilled individuals increased from 19% to 25% between 1990 and 2010, surpassing the high-skilled emigration rate of East Asia and the Pacific (see Figure 1). The so-called brain-drain or the migration of high-skilled individuals to other regions where human capital is abundant is a major concern for developing countries with a relatively small number of highly educated individuals, as it represents the loss of their most talented workers. The most evident way through which the negative externalities of the brain drain can be somewhat offset are remittances. Theoretically, there are several reasons to expect differences in the remittance behavior of high-skilled and less skilled migrants. On the one hand, high-skilled migrants may remit less because they often come from better-off families with lower income constraints. Besides, high-skilled migrants may have a higher propensity to migrate with their entire household and they may have lower intentions to return to their origin countries, which decreases their incentives to invest in their home community. On the other hand, highskilled migrants face lower transaction costs because they are less likely to be illegal, more likely to earn higher wages and more likely to have access to bank accounts and other financial services. These factors may result in a higher transfer of resources. In addition, remittances may serve as a way of repayment if family members at the origin country funded the education of the migrant (Docquier et al., 2012). It is therefore not clear if highly educated migrants remit more or less than less skilled migrants. Most of the existing empirical evidence investigating the effect of education on remittance patterns is based on macrodata. These studies usually conclude that the negative impacts of the brain drain cannot be counterbalanced because high-skilled individuals send less remittances (e.g. Faini, 2007; Adams, 2009; Niimi et al., 2010). The main disadvantage of these cross-country studies is that they are only able to reveal evidence on whether countries sending a higher share of skilled migrants receive more or less remittances than countries sending fewer skilled migrants, instead of analyzing if more skilled migrants remit more or not. The few studies using microdata find, in general, inconclusive results: some studies suggest that there is a negative relationship between education and remittances (see e.g. Dustmann and Mestres, 2010; Duval and Wolff, 2010), while others suggest a positive one (e.g. Bollard et al., 2011). A common issue that these studies share is that they are not able to sufficiently control for characteristics of the household left behind. If these unobserved household characteristics are correlated with both the education level of migrants and 1
their remittance behavior, then not controlling for these would lead to biased estimates. We contribute to the literature by analyzing the effect of migrants education on their remittance behavior controlling for unobserved characteristics of the household at the origin country. Our empirical analysis is based on household survey data from five Sub- Saharan African countries collected in 2009, which contain information on the education and remittance behavior of all migrants having left the household. Using a fixed-effects approach that is based on variation in education and remittances of migrants within the same household, we do not find evidence that the level of education is a determinant on the decision to send remittances. It is, however, an important factor to determine the amount of remittances sent. Conditional on sending remittances, highly educated migrants send a significantly higher amount of remittances than migrants with lower levels of education. The structure of the paper is as follows: in the next section, we discuss related literature. In section 3, we describe the data and sample used. Section 4 outlines the empirical methodology. In Section 5, we show the estimation results and section 6 concludes. 2 Related Literature The vast majority of studies that investigate whether the brain drain is associated with a larger flow of remittances is based on macro data, estimating the relationship between highskilled emigration rates and remittances inflows at the country level. Faini (2007) proposes a simple theoretical model of the relationship between the share of high-skilled migrants leaving the household and the amount of remittances sent home. The model predicts that with an increase in the share of high-skilled migrants per-capita remittances increase due to migrants higher wages ( wage effect ), but decrease because high-skilled migrants are more likely to reunite with close family members in the host country ( reunification effect ), leaving the overall effect ambiguous. In his empirical analysis based on data for OECD destination countries, he finds some support for a negative relationship between migrants education and remittances, though the respective coefficient is not statistically significant. Similar results are found by Niimi et al. (2010), who take into account the endogeneity of migration and migrant s education using instrumental variables, such as the distance between home and host countries and the labor market participation rate in the home country. By restricting his analysis to 76 low- and middle-income sending countries for which this information is available, Adams (2009) further controls for the level of poverty in the home country. His results, however, still suggest that countries that export a larger share of high skilled migrants receive less per capita remittances than those exporting a larger share of unskilled migrants. An exception to this is the study by Docquier et al. (2012), who use bilateral remittances data and find an ambiguous relationship between the level of remittances and education. 2
The main disadvantage of using macrodata to estimate the effect of high-skilled emigration on remittances is that these studies are only able to identify whether countries that send a larger share of educated migrants receive larger or smaller remittance flows, and not whether more educated individuals send a higher amount of remittances (Bollard et al. 2011). Microdata allow associating the education level of the migrant to their remitting behavior in terms of the likelihood of sending remittances and amount sent. Empirical evidence using microdata is still inconclusive. Some studies find that the effect is negative. For example, Dustmann and Mestres (2010) use the German Socio- Economic panel and find a negative relationship between years of schooling and remittances after controlling for return intentions and family members who are living in the country of origin (spouse and children). Similar to the previous study, Duval and Wolff (2010) find that highly educated children in Albania are less likely to send money to their parents and Bouoiyour and Miftah (2015) argue that a high level of education has no significant effect on the remitting behavior of the migrants. A positive relationship between education and remittances is found by Bollard et al. (2011) using household survey data from 11 OECD countries. The study finds mixed evidence at the extensive margin (the decision to remit), but a strong positive relationship at the intensive margin (the decision on how much to remit). For the overall effect, the study finds that when compared to migrants without a degree, migrants with a university degree send $300 more per year. 3 The Data The data used for this article come from the Migration and Remittances Households Surveys conducted by the World Bank in six Sub-Saharan African countries: Burkina Faso, Kenya, Nigeria, Senegal, Uganda, and South Africa. The surveys, which are part of the Africa Migration Project in 2009, are standardized across countries 1. South Africa is excluded from this analysis because it is regarded as a migrant-receiving country, thus, the questionnaire differs from the other countries. The survey collects comprehensive information on the household at the origin country, as well as on the characteristics of all household members of those still living in the household and of former household members who migrated. With respect to the latter, the database includes information on demographic characteristics, migration motives, and remittance patterns of each individual. In each country, about 2,000 households were interviewed. In contrast to other data sources, we are able to observe remittances at the individual 1 The surveys are nationally representative for Nigeria, Senegal, and Uganda. In the case of Burkina Faso and Kenya, they are representative for the 10 largest provinces and the top 17 districts with the highest concentration of migrants, respectively. 3
level sent through formal or informal channels. The households report if the migrants sent remittances back home in the last 12 months before the interview was conducted, as well as the amount sent in local currency. For comparison purposes, we convert all financial values to U.S. dollars using the average exchange rate for 2009 for each country. For the purpose of our identification strategy, we restrict the sample to households that report having two or more individuals who were former household members and migrated before the interview was conducted. Our final database consists of 6,648 migrants. With respect to our variable of interest, the surveys report the migrant s education level in different categories across countries, we re-classify them into four groups. i) Migrants without formal education: where we include individuals with no education, or who did not complete primary schooling. ii) Primary: which refers to individuals who completed primary education. iii) Secondary: which includes individuals who have completed general middle education or vocational trainings. iv) Tertiary: which includes individuals who have a university degree or postgraduate studies. Moreover, our database allows us to include different characteristics of the migrant that influence the probability of sending remittances and the amount transferred. We control for demographic characteristics such as age, marital status, gender and years spent at the destination country. In order to capture non-linearities between the control variables and the outcome, we include the squared term of the migrant s age and of the years spent at the destination country. As we do not have information on the income level of the migrant, we include a vector of dummies that indicate the labor force status of the migrant. Another important determinant for the remitting behavior is the reason to migrate. For instance, we would expect that individuals who migrated to pursue a degree are less likely to send remittances. We include dummy variables indicating if the migration decision was made to search for a job, to pursue education, to reunite with family members, or other reasons (e.g. related to conflict or weather conditions). As we restrict the sample to households with at least two migrants, the relationship from the migrant to the current household head in the origin country may be an important determinant for the remitting behavior. We, therefore, include dummy variables indicating if the migrant is the partner, child, sibling, or other relative of the household head. Finally, the data includes information on the destination country of the migrant, we use this information to condition on destination fixed effects and alternatively, we build dummy variables indicating if the individual migrated locally, or to another destination in Africa, America, Europe, and Asia or Oceania. Table 1 presents descriptive statistics for the full sample and compares the characteristics of migrants who remitted in the last 12 months before the interview with those who did not remit. The data highlights some of the characteristics of Sub-Saharan African migrants: 66% of the migrants in the sample are men and 46% send remittances. There is a large 4
number of local migrants who moved from rural to urban areas (63%), and the average time spent in the migration location is 6.8 years. The main reason to migrate is work related (42%) and most of the migrants are either full-time employed or self-employed. Furthermore, most of the migrants are either low-skilled or medium skilled, as only about 8% has completed tertiary education. About half of the migrants in the sample remit, sending on average $803 annually. The data show that there are significant differences between migrants who send or not remittances. For example, those individuals who send remittances are more likely to be married, male, and older compared to non-remitters. On average, 46% of the remitters are full-time employees, compared with 20% of non-remitters. Remitters are less likely to be inactive and more likely to be close relatives to the household head in the origin country. In terms of destination, remitters are more likely to be located in America or Europe. To gain some first insight into the relationship between migrants education and the amount of remittances sent to the household left behind, Figure 2 shows the underlying distribution of the logarithm of remittances (conditional on sending remittances) by skill group 2. In contrast to low-skilled individuals, the distribution of remittances sent by medium-skilled migrants is shifted to the right, and even more for high-skilled migrants. This provides some first descriptive evidence of a positive relationship between migrants education and the amount of remittances sent to the households left behind. 4 Methodology Our main objective is to provide evidence that the education level of the migrant has an impact on the amount of remittances sent back home. We estimate the following regression: R ijd = α 0 + µ E i + θ X i + δ j + δ d + ɛ ijd (1) where R ijd measures remittances sent by migrant i, currently living in destination d, to household j. E i is a vector of education dummies, X i a vector of demographic characteristics of the migrants as described above, and δ j and δ d are household and destination fixed effects, respectively. α 0, β, θ are parameter estimates and ɛ ijd the error term. We use three alternative measures of remittances in equation 1. First, in order to capture the overall effect of migrants education we use the logarithm of the amount remitted unconditional of remitting. We add a factor of one to the total amount of 2 We define skill groups in the following way: low-skilled include individuals with no formal and primary education, medium-skilled refer to individuals who completed secondary education, and high-skilled those who completed tertiary education. 5
remittances and then take the log, in order to allow for non-positive remittances. Second, we create a binary variable that indicates whether the migrant remits or not (extensive margin). Third, we use the logarithm of the total amount of remittances conditional on remitting (intensive margin). By including household fixed effects, we are able to control for unobserved characteristics of the family at the origin country that may be correlated with the education level of the migrant and the decision to send remittances. For example, migrants that come from families with a higher socioeconomic status may have a higher education level and a lower need of sending remittances. Therefore, not controlling for the characteristics of the household in the origin country may lead to biased estimates. It is worth mentioning that we take into account the correlation of migrants that come from the same households by clustering the standard errors at the origin household level. Although the database we use provides a rich set of covariates and allows us to include different control variables for the migrant and the household at the origin country, there are clear limitations. First, we do not have information on the migrant s income level, which is a key determinant on decision of sending remittances and the amount of remittances sent. Instead, we control for the labor force status of the migrant, as both measures are highly correlated. We expect that highly educated migrants have a higher probability of being employed at the destination country. Second, we limit our sample to households that have more than one migrant. This is a concern if individuals that come from one-migrant households differ systematically from those coming from multiple-migrant households. 5 Results We report the results for regressions of three remittance measures on education. In Table 2 we report the overall effect using the log of the total amount of remittance for the unconditioned sample. Table 3 reports the results at the extensive margin using a binary variable indicating if the migrant sent remittances in the past year. Table 4 reports the results for the intensive margin, where we use the log of total remittances conditional on remitting. The regressions include different sets of fixed effects: column I reports the regression results including country of origin fixed effects. Column II, III, and V include household fixed effects, column IV includes both, household and destination fixed effects. In Table 2 we can observe that the effect of education on remittances is not statistically significant for the overall sample. This effect is driven by the decision to send remittances and the amount transferred, so it is relevant to look at both effects separately. Concerning the likelihood of sending remittances, Table 3 shows that the level of education of the migrant is not a relevant determinant. High-skilled migrants seem to have a lower probability of sending remittances but this effect is not statistically significant. If we do 6
not control for the employment status of the migrant the coefficient is positive. With respect to the amount of remittances sent, Table 4 shows that in contrast with migrants with no formal education, highly-educated migrants send more remittances. The coefficient is positive and highly significant across specifications, but when conditioning on household and destination fixed effects the tertiary educated coefficient is reduced considerably. This could be due to unobserved factors that determine remittance patterns and the education level of the migrant. In general, our results suggest that the level of education of the migrant has no impact at the extensive level, but a large positive impact at the intensive level. In addition, the results show that some of the migrant characteristics have a significant effect on both, the likelihood of remitting and the amount sent. Gender is an important determinant at the extensive and intensive margins. Male migrants are more likely to send remittances and to send higher amounts. As well, the variable age is positive in both cases and the quadratic term negative, which implies that remittances increase at a decreasing rate. The martial status of the migrant only affects the probability of sending remittances but has no significant impact on the amount sent. The migrant s destination is also relevant, in contrast to local migrants, international migrants have a higher probability of transferring money and their transfers are larger. Variables related to the migration reason are only relevant for the decision to send remittance. However, once we control for the current employment status of the migrant, the effect is not significant anymore. The employment status of the migrant in the host country is an essential determinant in the decision to send or not remittances. In contrast to full time employees, immigrants who work part-time, are self-employed, or inactive, have a lower probability to send remittances. Correspondingly, at the intensive margin, inactive migrants send significantly less remittances when compared with full-time employed migrants. Results show that there are no significant differences between close relatives in terms of the likelihood of sending remittances. When compared to adult children of the household head (at the origin country), the partner, and siblings do not differ significantly in the probability of sending remittances. Yet, at the intensive margin when conditioning on those who remit, we can observe that the partner of the household head sends more remittances than the children. As expected, more distant relatives have a lower probability of sending remittances and they transfer significantly less than the children of the household. 6 Conclusion This paper investigates the relationship between remittances and migrants education using microdata from five countries in Sub-Saharan Africa. To the best of our knowledge, 7
this is the first econometric study that estimates the impact of education on remittances using microdata from the migrant and their origin households for this region. We focus on households with two or more migrants, which allow us to include household of origin fixed effects in the estimations. By doing this, we control for unobserved characteristics of the household left behind that may influence the education level of the migrant and their remittance behavior. Our results suggest that when combining the extensive and intensive margins there is no significant effect of education on remittances. However, once we separate these effects, we observe that education has no impact on the propensity of sending remittances, but that highly educated migrants send significantly higher amounts of money to their origin household. This could be explained by the fact that high-skilled migrants have access to better jobs and earn more money at the destination country. The brain drain is associated with negative externalities on the source country. One of these concerns is that high-skilled migrants send fewer remittances to their household of origin, which is not supported by the empirical evidence. If remittances increase with the level of education, some of the negative externalities of the brain drain on the source country may be counterbalanced. 8
References Adams, R. H. (2009). The determinants of international remittances in developing countries. World Development, 37 (1), 93 103. Bollard, A., McKenzie, D., Morten, M. and Rapoport, H. (2011). Remittances and the brain drain revisited: The microdata show that more educated migrants remit more. The World Bank Economic Review, 25 (1), 132 156. Bouoiyour, J. and Miftah, A. (2015). Why do migrants remit? testing hypotheses for the case of morocco. IZA Journal of Migration, 4 (1), 1 20. Brücker, H., Capuano, S. and Marfouk, A. (2013). Education, gender and international migration: insights from a panel-dataset 1980-2010. http://www.iab.de/en/ daten/iab-brain-drain-data.aspx, accessed: 2015-11-16. Docquier, F., Rapoport, H. and Salomone, S. (2012). Remittances, migrants education and immigration policy: Theory and evidence from bilateral data. Regional Science and Urban Economics, 42 (5), 817 828. Dustmann, C. and Mestres, J. (2010). Remittances and temporary migration. Journal of Development Economics, 92 (1), 62 70. Duval, L. and Wolff, F.-C. (2010). Remittances matter: Longitudinal evidence from albania. Post-Communist Economies, 22 (1), 73 97. Faini, R. (2007). Remittances and the brain drain: Do more skilled migrants remit more? The World Bank Economic Review, 21 (2), 177 191. Niimi, Y., Ozden, C. and Schiff, M. (2010). Remittances and the brain drain: skilled migrants do remit less. Annals of Economics and Statistics/Annales d Économie et de Statistique, pp. 123 141. OECD-UNDESA (2013). World migration in figures. http://www.oecd.org/els/mig/ World-Migration-in-Figures.pdf, accessed: 2016-01-04. 9
Figures Source: Authors analysis based on data from Brücker et al. (2013). Figure 1: Evolution of High-Skilled Migration by Region Source: Authors analysis based on data described in the text. Figure 2: Kernel Density Distribution by Skill Group 10
Tables Table 1: Descriptive Statistics of the Sample All migrants Non-remitters Remitters mean s.d. mean s.d. mean s.d. t-test. Probability of remitting 0.464 0.499 Total remittances 372.931 1434.202 803.125 2021.105 Male 0.664 0.472 0.597 0.491 0.741 0.438 12.53 Age 31.166 9.694 28.715 9.330 33.994 9.330 23.01 Married 0.556 0.497 0.452 0.498 0.676 0.468 18.83 Years since emigration 6.868 6.692 6.208 6.603 7.629 6.713 8.68 Education No formal education 0.276 0.447 0.266 0.442 0.287 0.452 1.89 Primary 0.170 0.376 0.192 0.394 0.144 0.351 5.23 Secondary 0.479 0.500 0.474 0.499 0.485 0.500 0.89 Tertiary 0.075 0.263 0.067 0.250 0.084 0.277 2.55 Migration reason Education 0.192 0.394 0.274 0.446 0.097 0.296 18.71 Job 0.417 0.493 0.343 0.475 0.502 0.500 13.28 Family 0.365 0.481 0.348 0.476 0.384 0.486 3.06 Other 0.026 0.160 0.035 0.183 0.016 0.126 4.75 Labor Force Status Full time employed 0.319 0.466 0.200 0.400 0.457 0.498 23.26 Part time employed 0.059 0.235 0.053 0.224 0.065 0.246 2.03 Self employed 0.351 0.477 0.301 0.459 0.408 0.491 9.10 Not in labor force 0.272 0.445 0.445 0.497 0.071 0.257 37.72 Relationship to head Son/daughter 0.627 0.484 0.608 0.488 0.649 0.477 3.46 Partner/head 0.027 0.163 0.017 0.128 0.040 0.196 5.81 Sibling 0.196 0.397 0.204 0.403 0.188 0.391 1.64 Other relative 0.097 0.296 0.104 0.306 0.089 0.284 2.16 Destination Africa 0.190 0.392 0.184 0.388 0.197 0.398 1.26 America 0.062 0.241 0.049 0.216 0.076 0.265 4.57 Asia or Oceania 0.009 0.095 0.008 0.090 0.010 0.100 0.82 Europe 0.108 0.310 0.067 0.250 0.155 0.362 11.65 Within country 0.631 0.483 0.691 0.462 0.562 0.496 10.97 Observations 6648 3561 3087 Significance levels: 5%, 1%,.1%. t-test: remitters vs non-remitters. 11
Table 2: Pooled OLS: Determinants of Remittances (Full-Sample) I II III IV V Education (Ref: No Education) Primary 0.166 0.196 0.096 0.201 0.076 (0.119) (0.175) (0.167) (0.165) (0.168) Secondary 0.110 0.078 0.065 0.166 0.143 (0.119) (0.169) (0.162) (0.163) (0.168) Tertiary 0.391 0.312 0.143 0.026 0.583 (0.220) (0.282) (0.266) (0.269) (0.276) Reason (Ref: Work Related) Education 0.076 0.068 0.182 0.238 1.048 (0.140) (0.174) (0.159) (0.158) (0.157) Family 0.005 0.319 0.099 0.134 0.325 (0.105) (0.145) (0.139) (0.137) (0.145) Other 0.954 0.973 0.766 0.754 1.035 (0.182) (0.284) (0.268) (0.284) (0.298) Status (Ref: Full time employed) Part time employed 0.618 0.958 1.075 1.022 (0.230) (0.258) (0.244) (0.246) Self employed 0.961 0.857 0.694 0.679 (0.121) (0.149) (0.140) (0.141) Not in labor force 2.623 2.338 2.159 2.170 (0.126) (0.160) (0.152) (0.149) Relationship to head (Ref: Child) Partner 0.573 0.533 0.632 0.626 0.863 (0.287) (0.349) (0.319) (0.314) (0.320) Sibling 0.569 0.306 0.362 0.310 0.212 (0.114) (0.192) (0.174) (0.171) (0.181) Other relative 0.305 0.696 0.745 0.666 0.751 (0.133) (0.192) (0.183) (0.184) (0.194) Male 0.335 0.181 0.261 0.250 0.571 (0.083) (0.096) (0.092) (0.092) (0.097) Age 0.091 0.115 0.091 0.081 0.153 (0.020) (0.027) (0.026) (0.025) (0.027) Age 2 0.001 0.001 0.001 0.001 0.001 (0.000) (0.000) (0.000) (0.000) (0.000) Married 0.446 0.400 0.329 0.310 0.334 (0.092) (0.115) (0.111) (0.109) (0.117) Years since emigration 0.048 0.029 0.032 0.038 0.056 (0.016) (0.019) (0.018) (0.019) (0.019) Years since emigration 2 0.002 0.001 0.001 0.001 0.001 (0.001) (0.001) (0.001) (0.001) (0.001) Destination (Ref: Local migrant) Africa 0.628 (0.175) America 1.777 (0.278) Asia or Oceania 0.378 (0.654) Europe 2.195 (0.212) Constant 0.393 0.304 0.318 0.022 2.612 (0.375) (0.517) (0.487) (0.475) (0.472) Household FE no yes yes yes yes Destination FE no no no yes yes Origin FE yes no no no no Observations 6648 6648 6648 6648 6648 Adj-R2 0.306 0.508 0.543 0.555 0.514 Note: Standard errors in parentheses (clustered at the household level). Significance levels: 10%, 5%, 1%. 12
Table 3: Pooled OLS: Determinants of Remittances (Extensive Margin) I II III IV V Education (Ref: No Education) Primary 0.006 0.017 0.005 0.014 0.007 (0.023) (0.032) (0.031) (0.032) (0.032) Secondary 0.002 0.019 0.014 0.023 0.029 (0.022) (0.029) (0.029) (0.030) (0.031) Tertiary 0.004 0.001 0.012 0.024 0.072 (0.035) (0.046) (0.045) (0.046) (0.047) Reason (Ref: Work Related) Education 0.071 0.041 0.051 0.060 0.212 (0.023) (0.028) (0.027) (0.027) (0.028) Family 0.021 0.065 0.042 0.046 0.083 (0.019) (0.025) (0.024) (0.024) (0.026) Other 0.160 0.173 0.151 0.153 0.205 (0.039) (0.055) (0.054) (0.055) (0.059) Status (Ref: Full time employed) Part time employed 0.090 0.140 0.151 0.147 (0.037) (0.046) (0.046) (0.045) Self employed 0.131 0.122 0.109 0.106 (0.021) (0.025) (0.025) (0.025) Not in labor force 0.434 0.406 0.390 0.390 (0.022) (0.027) (0.027) (0.027) Relationship to head (Ref: Child) Partner 0.001 0.014 0.023 0.021 0.065 (0.045) (0.055) (0.053) (0.053) (0.055) Sibling 0.115 0.050 0.057 0.055 0.037 (0.020) (0.031) (0.030) (0.030) (0.032) Other relative 0.065 0.098 0.104 0.097 0.112 (0.024) (0.034) (0.033) (0.034) (0.035) Male 0.045 0.029 0.036 0.035 0.095 (0.015) (0.016) (0.016) (0.016) (0.017) Age 0.015 0.018 0.016 0.015 0.028 (0.004) (0.005) (0.005) (0.005) (0.005) Age 2 0.000 0.000 0.000 0.000 0.000 (0.000) (0.000) (0.000) (0.000) (0.000) Married 0.083 0.071 0.065 0.063 0.066 (0.017) (0.020) (0.020) (0.020) (0.021) Years since emigration 0.005 0.003 0.004 0.005 0.008 (0.003) (0.003) (0.003) (0.004) (0.004) Years since emigration 2 0.000 0.000 0.000 0.000 0.000 (0.000) (0.000) (0.000) (0.000) (0.000) Destination (Ref: Local migrant) Africa 0.080 (0.031) America 0.171 (0.042) Asia or Oceania 0.048 (0.106) Europe 0.204 (0.033) Constant 0.248 0.022 0.026 0.000 0.456 (0.072) (0.096) (0.094) (0.093) (0.091) Household FE no yes yes yes yes Destination FE no no no yes yes Origin FE yes no no no no Observations 6648 6648 6648 6648 6648 Adj-R2 0.248 0.488 0.499 0.504 0.459 Note: Standard errors in parentheses (clustered at the household level). Significance levels: 10%, 5%, 1%. 13
Table 4: Pooled OLS: Determinants of Remittances (Intensive Margin) I II III IV V Education (Ref: No Education) Primary 0.300 0.010 0.040 0.092 0.079 (0.103) (0.201) (0.187) (0.183) (0.183) Secondary 0.318 0.242 0.286 0.152 0.217 (0.107) (0.178) (0.164) (0.163) (0.169) Tertiary 1.025 0.611 0.579 0.404 0.476 (0.166) (0.250) (0.222) (0.229) (0.234) Reason (Ref: Work Related) Education 0.622 0.175 0.053 0.045 0.098 (0.137) (0.220) (0.191) (0.190) (0.189) Family 0.172 0.039 0.122 0.093 0.057 (0.086) (0.145) (0.128) (0.127) (0.127) Other 0.537 0.564 0.418 0.484 0.483 (0.203) (0.460) (0.389) (0.400) (0.405) Status (Ref: Full time employed) Part time employed 0.125 0.225 0.271 0.204 (0.155) (0.212) (0.197) (0.195) Self employed 0.371 0.218 0.078 0.078 (0.084) (0.134) (0.123) (0.122) Not in labor force 1.015 0.588 0.579 0.556 (0.145) (0.217) (0.197) (0.195) Relationship to head (Ref: Child) Partner 0.831 0.529 0.574 0.567 0.597 (0.153) (0.307) (0.270) (0.260) (0.259) Sibling 0.054 0.028 0.143 0.085 0.065 (0.086) (0.211) (0.180) (0.174) (0.171) Other relative 0.099 0.393 0.434 0.353 0.366 (0.105) (0.198) (0.164) (0.153) (0.154) Male 0.122 0.117 0.182 0.181 0.238 (0.070) (0.091) (0.084) (0.084) (0.086) Age 0.107 0.102 0.087 0.074 0.083 (0.019) (0.034) (0.031) (0.031) (0.032) Age 2 0.001 0.001 0.001 0.001 0.001 (0.000) (0.000) (0.000) (0.000) (0.000) Married 0.026 0.045 0.024 0.024 0.015 (0.073) (0.110) (0.098) (0.097) (0.098) Years since emigration 0.030 0.001 0.006 0.001 0.004 (0.011) (0.016) (0.015) (0.014) (0.014) Years since emigration 2 0.001 0.000 0.000 0.000 0.000 (0.000) (0.000) (0.000) (0.000) (0.000) Destination (Ref: Local migrant) Africa 0.315 (0.160) America 1.226 (0.190) Asia or Oceania 0.200 (0.504) Europe 1.644 (0.157) Constant 1.743 1.524 2.112 2.505 2.309 (0.362) (0.673) (0.622) (0.620) (0.641) Household FE no yes yes yes yes Destination FE no no no yes yes Origin FE yes no no no no Observations 3087 3087 3087 3087 3087 Adj-R2 0.359 0.609 0.669 0.684 0.681 Note: Standard errors in parentheses (clustered at the household level). Significance levels: 10%, 5%, 1%. 14