Introduction Manajemen | Fakultas Ekonomi Universitas Maritim Raja Ali Haji 768.full

I. Introduction

This paper investigates the relationship between migration experience and earnings among return migrants in the Mexican labor market. Our analysis com- plements a growing literature assessing the effects of out- migration on the economies of migrant- sending countries. Two main concerns dominate research on this topic: the consequences of skilled migration the “brain drain” and the effects of remittances 1 . However, if migration is temporary and migrants eventually return home, then out- migration might have an additional effect by upgrading the skills and raising the earn- ings of return migrants. We develop a theoretical model of return migration and skill upgrading to guide the interpretation of our empirical results. In particular, the model provides conditions under which OLS regression estimates provide a lower bound on the average causal effect of migration experience for migrants. We further explore the nature of skill upgrading by considering several mechanisms that might mediate this relationship including occupation- specifi c learning, knowledge spillovers in large urban areas, the acquisition of English skills, and legal status during migration. We use data from the Mexican Migration Project MMP and the Population and Dwelling Count 1995 to document the relationship between past U.S. migration experience and the labor market earnings of return migrants in Mexico. Our baseline specifi cation suggests that each additional year of migration experience is associated with an increase in earnings of approximately 2.2 percent, everything else equal. We fi nd that the return to migration experience is at least twice as high as the return to age in the Mexican labor market at nearly every point in the life cycle. A natural concern is that the observed relationship between earnings and migra- tion experience refl ects the self- selection of high- ability or high- skill individuals into return migration, or the endogeneity of length- of- stay in the United States. To assess the likelihood and consequences of such biases, we develop a model of temporary mi- gration with heterogeneity in ability or skill. Under plausible assumptions, the model suggests that our empirical estimates may understate the true effect of migration ex- perience on earnings. Furthermore, the model shows that extensive margin selection where return migrants are placed in the skill distribution does not bias OLS results once a dummy for any migration experience is included. Rather, the OLS estimates suffer from bias only if unobserved components of skill are correlated with accumu- lated migration experience within the group of return migrants. Instrumental variables techniques and fi xed effects methods are two common strat- egies for addressing this kind of endogeneity. In an earlier working paper, we used macroeconomic shocks in the United States to instrument for accumulated migration experience, and estimated a very large return around 9 percent to migration experi- ence. However, the instruments were weak in many specifi cations and quite sensitive to the inclusion or exclusion of age controls. Furthermore, one might be concerned that U.S. macro shocks are correlated with Mexican macro shocks, and thus cannot be credibly excluded from the earnings equation. In the absence of credible instruments, one could address this question with panel data. Unfortunately, existing Mexican panel 1. See Docquier and Rapoport 2009 for a review of recent literature on the positive and negative conse- quences of skilled migration for human capital accumulation in sending countries. Rapoport and Docquier 2006 provide a survey of recent work on the economics of remittances. data sets do not contain suffi cient information on return migrants to be useful. 2 While panel data with earnings observations before and after migration are not available, we can observe wage data from the United States for the migrants in our sample to con- struct a control for ability and unobserved skill. When we add this control to our basic specifi cation, our estimates are very close to the baseline. We argue that this exercise provides evidence against the proposition that the relationship between migration ex- perience and earnings is being driven by correlation between migration experience and permanent unobserved skill. Another contribution of the paper is to compare results on the returns to migra- tion experience across both the MMP and the Population and Dwelling Count Survey PDC. While the MMP provides superior information on migration histories, the MMP only includes data on Mexican income not wages. However, the PDC reports hours data that allow us to construct wage rates. When using comparable regressors and samples, we fi nd very similar results in both the MMP and the PDC. Furthermore, the analysis using the PDC wage data reveals that our results primarily refl ect a rela- tionship between migration experience and wages, not experience and hours worked. The paper also contributes to the literature on migration and human capital accu- mulation by suggesting and empirically testing some possible explanations for the ex- istence of a return to migration experience. We fi nd evidence that occupation- specifi c job experience accounts for much of this return. The return to migration experience is largest for migrants who worked in occupations in the United States that match their current occupation in Mexico. Indeed, the return to a year of this kind of job- relevant migration experience is estimated to be a little less than 4.8 percent in the whole sample, and as high as 8.7 percent when restricting the sample to unskilled manufacturing workers. It is noteworthy that our basic estimate of the return to a year of job- relevant migration experience is nearly as large as our estimate of the return to education in our baseline specifi cations. Another mechanism that could explain our results is the effect of exposure to urban labor markets. As documented by Glaeser and Mare 2001 and others, it could be the case that experience in large urban labor markets is more valuable than experience elsewhere because of greater knowledge spillovers that might occur in cities. The rich migration information available in the MMP allows us to test whether the return to migration experience is actually a return to urban labor market experience in the United States. However, we do not fi nd evidence that urban migration experience is more valuable than nonurban experience. We also test whether the return to migration experience is related to the acquisition of English skills. Data limitations prevent us from drawing fi rm conclusions here, but controlling for English ability does not substantially reduce the return to migration experience. This suggests that English language acquisition does not explain much of the relationship between migration experience and earnings. However, we do fi nd that the return to migration experience is higher for those with some English skills. This is consistent with the idea that individuals who can more easily communicate while abroad might be better able to absorb skills while working. 2. For example, we attempted to use the Mexican Family Life Survey’s 2002 and 2005 waves to explore this question. This yielded a very small sample of individuals with nonmissing Mexican income data in 2002 and 2005 that accumulated migration experience between the two waves. Just as English skills might allow migrants to interact more freely in U.S. labor mar- kets, one’s legal status while a migrant might also shape the opportunities that one has to accumulate skills while abroad. We fi nd evidence of a greater return to documented migration experience, and a greater propensity for documented migrants to acquire job- relevant work experience appears to drive this result.

II. Literature on Return Migration and Skill Upgrading