Selectivity, Transferability of Skills and Labor Market Outcomes. of Recent Immigrants in the United States. Karla J Diaz Hadzisadikovic

Save this PDF as:
 WORD  PNG  TXT  JPG

Size: px
Start display at page:

Download "Selectivity, Transferability of Skills and Labor Market Outcomes. of Recent Immigrants in the United States. Karla J Diaz Hadzisadikovic"

Transcription

1 Selectivity, Transferability of Skills and Labor Market Outcomes of Recent Immigrants in the United States Karla J Diaz Hadzisadikovic Submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy under the Executive Committee of the Graduate School of Arts and Sciences COLUMBIA UNIVERSITY 2012

2 2012 Karla J Diaz Hadzisadikovic All Rights Reserved

3 ABSTRACT Selectivity, Transferability of Skills and Labor Market Outcomes of Recent Immigrants in the United States Karla J Diaz Hadzisadikovic This dissertation analyzes how immigrants individual and home country characteristics affect and determine their labor market participation, returns to education and wages in both their country of origin (before migration occurs) and in the United States. The dissertation also estimates the extent to which immigrant skills are transferable to the American labor market. The research is carried out using data from the New Immigrant Survey, released in 2003, supplemented with the Mexican Household Income and Expenditure National Survey and other data sources. The New Immigrant Survey has two specific sets of questions that other surveys do not have. First, it has a detailed set of questions about the socioeconomic experience of the migrants before they left their home country. Secondly, the survey asked participants about their immigrant visa categories, allowing a separation of immigrants into economic migrants (those who came through employment preferences or other categories directly linked to economic objectives) and non-economic migrants (refugees, migrants entering through family preferences, etc.., whose direct migration motive was political, family-related, and not directly economic in nature). The dissertation reveals that there exist significant differences between

4 economic and non-economic migrants in the determinants of labor market participation, wages, migration selectivity and transferability of skills. Substantial differences are observed between men and women. Immigrants have low labor participation rates in their country of origin although they are highly educated compared to non-migrants at home. Those from English-speaking and high GDP countries are less likely to work but earn higher wages. Non-economic migrant men are more likely to be employed (than economic migrant men), but their wages do not differ. Among women, both economic and non-economic migrants are just as likely to be employed but noneconomic migrant women earn less. Using Mexico as a case study to examine the selectivity of legal immigrants, it was found that documented migrants were less likely to have been employed before migration to the US, but their level of education and wages were significantly higher than those of non-migrants. The individual characteristics of these two groups affect their employment and wage determinants differently. Some of the literature has emphasized how immigrants may be positively selected because of greater motivation, willingness to undertake risks, etc. But, on this basis, most of the existing analysis of the determinants of immigrant wages in the U.S. suffers from omitted variable bias because there are no data for these unobserved characteristics or skills (motivation, persistence, etc.) and they are ignored in the statistical analysis. In an examination of immigrant wages in the U.S., this dissertation used wages earned abroad as a measure of the unobserved skills of the migrants and their level of transferability. An analysis of the determinants of immigrant wages was then carried out, examining rates of

5 return to education, experience, differences in wages between economic and noneconomic migrants, etc. The level of transferability of skills was found to be higher for economic migrants than for non-economic migrants.

6 TABLE OF CONTENTS List of figures List of tables Acknowledgments...v... vi... xi Chapter 1. Introduction... 1 I. The Debate on Immigrant Selectivity and Assimilation... 2 II. Structure of the Dissertation... 6 Chapter 2. Migrant Selectivity and the Labor Market Outcomes of U.S. Immigrants in their Home Countries I. Literature Review A. Quality of Immigrants Education B. Quality of Immigrants Earnings and Assimilation C. Self-Selection D. Immigrant Labor Market Performance in the Source Country and the Selectivity of Migrants II. Data Description and Data Analysis A. Summary Statistics B. Labor Force Participation in Country of Origin Full Sample Results Results by Gender C. Returns to Education and Wage Determinants in the Country of Origin Full Sample Results Results by Gender i

7 D. Full Information Maximum Likelihood Heckman Selection Methodology Maximum Likelihood Heckman Selection Model Estimating the Full Information Maximum Likelihood function Full Sample Results Results by Gender III. Conclusions Chapter 3. Labor market outcomes of immigrants and non-immigrants from Mexico A Oaxaca Decomposition I. Introduction II. Data Description A. Mexico s Household Income and Expenditure National Survey B. Summary Statistics III. Immigrant and Non-migrant Labor Market Experience A. Labor Market Participation B. Rates of Return to Education and Wage Determinants IV. Blinder-Oaxaca Decomposition Methodology V. Decomposition of Observables and Unobservables A. Labor Market Participation B. Wages in Mexico VI. Conclusions Chapter 4. Transferability of Skills, Income Growth and Labor Market Outcomes of Recent Immigrants in the United States I. Introduction ii

8 II. Review of the Literature III. Summary Statistics IV. Labor market participation, return to education and transferable skills of immigrants in the United States A. Probability of Employment Full Sample Results Results by Gender B. Returns to Education Ordinary Least Squares Full Sample Results Results by Gender C. Heckman Selection Model Maximum Likelihood Heckman Selection Model Estimating the Full Information Maximum Likelihood function D. Heckman s Full Information Likelihood Maximization Estimation Results Full Sample Results Results by Gender E. Transferable Skills and Income Growth Theoretical Background Migration Model Methodology and Results a. Full Sample Results b. Results by Gender V. Conclusions Chapter 5. Conclusions and Implications iii

9 A. Shortcomings of the analysis B. Future research C. Policy implications References Figures Tables Appendix A Additional Tables iv

10 LIST OF FIGURES Figure 1. Probability of Employment Figure 2. Experience and the Probability of Employment Abroad Figure 3. Gender Gap in Labor Market Participation and Wages Figure 4. Number of Mexican immigrants that have become Legal Permanent Residents by decade Figure 5. Employment in the United States by Visa Category Figure 6. Transferability of Skills By Country Figure 7. Transferability of Skills By Country and Gender Figure 8. Wages in the USA and Wages Abroad By Education Figure 9. Wages in the USA and Wages Abroad By Education and Gender Figure 10. Wages in the USA and Wages Abroad By Language Spoken Figure 11. Wages in the USA and Wages Abroad By Language Spoken and Gender Figure 12. Income Growth and Income Level Abroad By Education Figure 13. Income Growth and Income Level Abroad By Education and Gender Figure 14. Income Growth and Income Abroad By Language Spoken Figure 15. Income Growth and Income Abroad By Language Spoken and Gender v

11 LIST OF TABLES Table 1. Summary Statistics by Employment Abroad Table 2. Summary Statistics by Employment Abroad and Gender Table 3. Probability of Employment in Country of Origin (Probit) Table 4. Men s Probability of Employment in Country of Origin (Probit) Table 5. Women s Probability of Employment in Country of Origin (Probit) Table 6. Rate of Return to Education of New Immigrants in Country of Origin (OLS) Table 7. Rate of Return to Education of New Immigrant Men in Country of Origin (OLS) Table 8. Rate of Return to Education of New Immigrant Women in Country of Origin (OLS) Table 9. Rate of Return to Education of New Immigrants in Country of Origin (Heckman) Table 10. Rate of Return to Education of New Immigrant Men in Country of Origin (Heckman) Table 11. Rate of Return to Education of New Immigrant Women in Country of Origin (Heckman) Table 12. Summary Statistics for Immigrant and Non-Migrant Mexicans Table 13. Summary Statistics for Immigrants and Non-Migrants from Mexico by Gender Table 14. Employment Probability in Mexico of Immigrants and Non-Migrants vi

12 Table 15. Employment Probability in Mexico of Immigrant and Non-Migrant Men Table 16. Employment Probability in Mexico of Immigrant and Non-Migrant Women Table 17. Weekly Wages in Mexico of Immigrants and Non-Migrants Table 18. Weekly Wages in Mexico of Immigrant and Non-Migrant Men Table 19. Weekly Wages of Immigrant and Non-Migrant Women Table 20. Oaxaca Decomposition of Employment Status in Mexico Full Sample Table 21. Oaxaca Decomposition of Employment Status in Mexico - Men Sample Table 22. Oaxaca Decomposition of Employment Status in Mexico Women Sample Table 23. Oaxaca Decomposition of Weekly Wages in Mexico Full Sample Table 24. Oaxaca Decomposition of Weekly Wages in Mexico Men Sample Table 25. Oaxaca Decomposition of Weekly Wages in Mexico Women Sample Table 26. Summary Statistics by Employment in the United States Table 27. Summary Statistics by Employment in the United States and by Gender Table 28. Probability of Employment in the United States (Probit Regression) 218 Table 29. Men s Probability of Employment in the United States (Probit Regression) vii

13 Table 30. Women s Probability of Employment in the United States (Probit Regression) Table 31. New Immigrants Return to Education and Transferable Skills (Dependent Variable: Wages in the United States) Table 32. New Immigrant Men s Return to Education and Transferable Skills (Dependent Variable: Wages in the United States) Table 33. New immigrant Women s Return to Education and Transferable Skills (Dependent Variable: Wages in the United States) Table 34. New Immigrants Return to Education and Transferable Skills Heckman Estimation (Dependent Variable: Log of Weekly Wages in the United States) Table 35. New Immigrant Men s Return to Education and Transferable Skills Heckman Estimation (Dependent Variable: Log of Weekly Wages in the United States) Table 36. New Immigrant Women s Return to Education and Transferable Skills Heckman Estimation (Dependent Variable: Log of Weekly Wages in the United States) Table 37. New Immigrants Return to Education and Transferable Skills Heckman Estimation by Immigrant Type (Dependent Variable: Log of Weekly Wages in the United States) Table A1. Summary Statistics by Country of Origin Table A2. Top 15 Languages Spoken at Home Table A3. Probability of Employment in Country of Origin (Probit) Table A4. Men s Probability of Employment in Country of Origin (Probit) viii

14 Table A5. Women s Probability of Employment in Country of Origin (Probit). 252 Table A6. Return to Education of New Immigrants in Country of Origin (Dependent Variable: Log of Weekly Wages Abroad) Table A7. Return to Education of New Immigrant Men in Country of Origin (Dependent Variable: Log of Weekly Wages Abroad) Table A8. Return to Education of New Immigrant Women in Country of Origin (Dependent Variable: Log of Weekly Wages Abroad) Table A9. (Heckman) Return to Education of New Immigrants in Country of Origin (Dependent Variable: Log of Weekly Wages Abroad) Table A10. (Heckman) Return to Education of New Immigrant Men in Country of Origin (Dependent Variable: Log of Weekly Wages Abroad) Table A11. (Heckman) Return to Education of New Immigrant Women in Country of Origin (Dependent Variable: Log of Weekly Wages Abroad) Table A12. Probability of Employment in the United States (Education as a Continuous Variable) (Probit) Table A13. Men s Probability of Employment in the United States (Education as a Continuous Variable) (Probit) Table A14. Women s Probability of Employment in the United States (Education as a Continuous Variable) (Probit) Table A15. New Immigrants Return to Education and Transferable Skills (Dependent Variable: Wages in the United States) (Education as a Continuous Variable) ix

15 Table A16. New Immigrant Men s Return to Education and Transferable Skills (Dependent Variable: Wages in the United States) (Education as a Continuous Variable) Table A17. New Immigrant Women s Return to Education and Transferable Skills (Dependent Variable: Wages in the United States) (Education as a Continuous Variable) Table A18. New Immigrants Return to Education and Transferable Skills Heckman Estimation (Dependent Variable: Log of Weekly Wages in the United States) (Education as a Continuous Variable) Table A19. New Immigrant Men s Return to Education and Transferable Skills Heckman Estimation (Dependent Variable: Log of Weekly Wages in the United States) (Education as a Continuous Variable) Table A20. New Immigrant Women s Return to Education and Transferable Skills Heckman Estimation (Dependent Variable: Log of Weekly Wages in the United States) (Education as a Continuous Variable) x

16 ACKNOWLEDGMENTS I have received a great deal of support in every step that has taken me to this point, the completion of my PhD. I am eternally grateful to my dear advisor Francisco Rivera-Batiz, whose support, advice and encouragement has been invaluable. I am also very grateful to Charlie Becker, for being my mentor since before I was even accepted into a doctoral program. I would like to thank the members of my dissertation committee for taking the time to read about a subject I m passionate about, to give me excellent feedback and to suggest extensions to my work. It is the faith that my family has always had in my capabilities that encouraged me to pursue this path. A mi papá y mi hermana, Erika, gracias por siempre creer en mí. A mamá, gracias por tu apoyo y por ser mi modelo a seguir. And last, but certainly not least, I want to thank my #1 fan, my best friend, my husband, Alan, for his love, support and encouragement, and for making my dream our reality. xi

17 To Frida and Julia. xii

18 1 CHAPTER 1. INTRODUCTION The proportion of skills that immigrants can transfer to the United States (US) or another destination country affects their initial income. When immigrants come to the US or to any other country, they bring with them sets of skills that were acquired in a different social, cultural and economic context. These skills derive from, and include aspects of education, language, and past experiences, in addition to their culture, customs and networks. Since these factors may vary by individual and by country of origin, it is rare that the skill sets immigrants bring are fully transferrable or can be equally compensated for in the US labor market context, especially in the case of adult immigrants. As Jasso, Rosenzweig, and Smith (2002) have noted, the degree of transferability of skills and the extent to which they are rewarded in the United States is therefore an essential question in analyzing the economic status of immigrants. The purpose of this dissertation is to measure the skills that recently-arrived adult immigrants bring into the US, their transferability, and their labor market impact. The issue of how immigrants adapt to their home country labor market and society is one that has been debated for centuries. Social scientists, and economists in particular, have indeed studied for a long time the issue of how immigrants adjust or assimilate to their host country society. The degree of assimilation of immigrants is also perennially one of the main political issues in the US and other countries. Fears that immigrants will become an underclass and a burden for native-born workers permeate the views of some while others argue

19 2 that immigrants fully assimilate to their host countries over time, achieving substantial progress and making substantial contributions to the host countries. This dissertation examines the issues of immigrant selectivity and assimilation in the United States using a unique data set recently available. The New Immigrant Survey offers a unique opportunity for exploring the skill transferability of immigrants, because it is the only publicly-available and representative dataset that provides documented immigrants wage information before and after immigration to the US. This nationally representative sample of 8,573 respondents was selected from administrative records that the US government collects on new immigrants through the US Citizenship and Naturalization Service (USCIS). The data on the first and only available cohort to this date was collected between May and November The New Immigrant survey questionnaire asks immigrants a battery of questions regarding their situation before entering the US, back in their country of origin. This allows a more careful study of immigrant selectivity and assimilation. The research seeks to contribute to the long history of research in economics examining these topics, summarized next. I. The Debate on Immigrant Selectivity and Assimilation In the economics literature, the early seminal work of Chiswick (1978) led to the idea that immigrants are positively self-selected, that is, immigrants have abilities, skills and/or a drive to succeed that is higher than those of the average person in their country of origin. Immigrants, therefore, tend to be positively

20 3 selected from their counterparts in home countries because in order to compensate for the substantial costs of migration only those who have the strongest drive and motivation and the expectations of great rewards will actually undertake the migration process. As a result, their economic progress in the host country is substantial and, after a period of time, they are able to catchup with the native-born in terms of employment and earnings. Chiswick examined empirically various immigrant groups in the US, finding indeed that immigrant earnings grew over time, gradually approaching those of native-born workers with the same characteristics. But the view that immigrants are positively-selected has been questioned by Borjas (1985, 1995). He presents the hypothesis (referred in the literature as the Roy model) that, if those with greater skills or abilities are rewarded more highly compared to the less-skilled in the origin area when compared to the destination region, this will generate less incentives for those at the top of the skills or ability distribution to emigrate compared to those at the bottom of the distribution, causing a negative selectivity of migrants. Therefore, the emigrant contingent will be positively or negatively selected depending on the relative inequality of the distribution of income at home and abroad. Borjas also criticized Chiswick s empirical results. He argued that Chiswick s results were biased because he did not correct for changes in the unobserved characteristics of migrants. In his view the selectivity of migrants to the US has declined over time because of increased numbers and shifts among the source countries. If the unobserved characteristics of earlier (older) immigrant cohorts are lower than those of more recent (younger) cohorts, the earnings of older cohorts will be

21 4 growing at a faster pace than those of the younger immigrant cohorts not because they have been a longer period of time in the US and have adjusted very well to their host country but rather because they have other unobserved characteristics that lead to higher wages. Revising Chiswick s empirical work, Borjas (1995) found that there existed large differences in the economic progress of US immigrant cohorts and that the immigrants earnings disadvantage relative to natives had grown over time. Recent immigrants in the late 1980s cohort earned between 19 to 27 percent less than natives. Similar immigrants who arrived between 1960 and 1964 earned only 2 to 8 percent less than natives. The work of Chiswick, Borjas and others (such as Chiswick, 1999; Duleep & Regets, 1997a; Duleep & Regets, 1997b; Duleep & Regets, 2002; LaLonde & Topel, 1992; and Rivera-Batiz, 2007) on immigrant selectivity has a serious shortcoming. The issue of whether immigrants are positively-selected or negatively-selected or whether their characteristics have changed over time-- requires a comparison between immigrants and non-immigrants in the country of origin, information that most researchers could not directly observe. Instead, the selectivity of migrants has been typically measured indirectly, on the basis of observable characteristics such as level of education and wages earned in the receiving country, because commonly used data in the context of immigration research, such as the US Census and the Current Population Survey, do not provide information on the performance of immigrants (education, income) in their country of origin. As noted earlier, to measure the adult immigrants selectivity, most of the literature uses indirect, destination-area variables, such as the wages in the host country. However, these indices do not measure selectivity

22 5 separately from the initial transfer of skills. In order to disentangle these two measures, it is necessary to include wages earned by immigrants abroad before they came to the US as a measure of skill transferability, separate from selectivity on other characteristics. Some studies have sought to introduce overall country of origin characteristics (average income per capita or education in the source country) in analyses of the economic progress of immigrants, to measure possible differences in the selectivity of migrants, but this ignores variations in the selectivity of migrants within a country. Analyses of immigrant selectivity have also sought to compare immigrants in the US with their counterparts in the home country (those who stayed at home) combining Census and other data in the source countries with US Census and other data on immigrants (see, for example, Chiquiar and Hanson, 2005). But these studies also lack data on the performance of immigrants in the home country (their wages, income, etc.) and cannot therefore provide a baseline for the comparison of their relative economic performance before and after migration. Some recent studies have available individual information on the wages of Mexican workers and can identify those who have later migrated to the US. These studies, therefore, can compare migrant and non-migrant characteristics in Mexico. Fernandez-Huertas Moraga (2011) used the Encuesta Nacional de Empleo Trimestral (ENET). The ENET follows the household for five consecutive quarters and reports if a household member moved to the US. The Encuesta Nacional de la Dinámica Demográfica (ENADID) is another Mexican survey that collects data on migrants as long as there is at least one member of the household

23 6 that remains in Mexico. McKenzie and Rapoport (2010) use this data to examine the labor market outcomes of migrants and non-migrants in Mexico. Finally, Kaestner and Malamud (2010) used the Mexican Family Life Survey (MxFLS) to examine the issue of Mexican migrant selectivity. The MxFLS is a longitudinal and representative survey of Mexicans, which follows Mexican households over time. Note, however, that this recent literature only considers Mexican migrants to the United States and can only examine the limited issue of Mexican selectivity, comparing migrants and non-migrants in Mexico. The question of US immigrant selectivity for countries other than Mexico and the question of how migrants assimilate in the US labor market are left open. II. Structure of the Dissertation This dissertation examines the issues of immigrant selectivity and assimilation in the United States using the New Immigrant Survey. The nationally representative sample of 8,573 respondents, all recent immigrants to the US, was asked a battery of questions regarding their social and economic situation before entering the US, back in their country of origin. The dissertation uses this information, in addition to the data the survey provides on the economic performance of immigrants in the US, to provide a more careful and complete analysis of the issues of immigrant selection and assimilation relative to the previous literature. The dissertation is divided into three chapters, in addition to the introduction and conclusion. The first chapter (Chapter 2) is an analysis of the

24 7 relative labor market performance of migrants in their home countries. Research is carried out to determine the determinants of employment and wage differences among US immigrants before they moved to the United States, as reported by the New Immigrant Survey. Most of the immigrants surveyed were (1) able to come to the US through marriage or other family relationships to a US citizen or in some cases to another permanent resident, and (2) those seeking asylum. The visa categories they fall under can be considered of non-economic nature. Those persons who migrate to the US (1) on the basis of employment preferences, (2) undocumented, or (3) through a lottery (diversity immigrants) used visas that can be considered economic-motivation visas. This first chapter analyzes the labor market participation and earnings of immigrants in their country of origin and studies the determinants of differences in the probability of employment and in earnings among the various groups of immigrants in their country of origin. The empirical results presented in this chapter confirm that men and women, as well as those with economic and non-economic visas have different experiences in the labor market abroad and therefore different incentives to migrate. The first chapter of this dissertation shows that there are important characteristics that affect the labor market outcomes of immigrants abroad and this may consequently affect their labor market outcomes in the US. Previous research has been ignoring these characteristics and therefore possibly producing biased results in their analysis. Chapter 4 introduces these characteristics in the analysis of the labor market outcomes of immigrants in the US to correct for this deficiency.

25 8 According to the New Immigrant Survey data, on average, immigrants who have recently become legal permanent residents in the US have a level of education that is at least 50% higher than the average population in their country of origin. But how else do they differ from the general population in their country of origin? Do these differences persist after one controls for other observable characteristics such as gender, experience or marital status? And how do these differences affect their labor market experiences in their country of origin? In chapter 3, the dissertation will focus on only one country, Mexico, to answer these questions. Choosing Mexico as the country for this analysis has its advantages. First, Mexico is the largest net exporter of labor to the US. In 2000, Mexicans accounted for about 30% of all immigrants in the US (Congressional Budget Office, 2004). Second, Mexicans are also the largest group represented in the New Immigrant Survey. Third, given that Mexicans are the largest group of immigrants in the US and that the US is the main destination for Mexican immigrants, there already exists a vast amount of research in the area of immigrants labor market outcomes, assimilation and self-selectivity focusing on Mexican immigrants. Chapter 3 compares the labor market outcomes of Mexican immigrants in the US with those of non-migrant Mexicans. In particular two outcomes are examined: labor market participation and wages in Mexico. Two different data sets are used for the analysis: the New Immigrant Survey for the immigrant sample and the Household Income and Expenditure National Survey from Mexico for the non-migrant sample. The evidence presented in this chapter suggests the conclusion that Mexican legal immigrants are a positively selected

26 9 group among Mexican workers. Although they suffer from higher rates of unemployment before they move to the US, they have higher levels of education and earn higher wages than non-immigrant Mexicans. These results agree with those of Chiquiar and Hanson (2005). Chapter 4 focuses on the issue of assimilation, selectivity, and the determinants of the labor market performance of immigrants in the US. It analyzes three aspects of the labor market experience of immigrants in the US. First, the determinants of labor force participation are examined. Second, wage equations are estimated to examine the rate of return to education and the determinants of wages in the US for those who were employed ignoring wages from abroad, but including other information from previous employment. Third, use is made of the information about wages of immigrants abroad to estimate transferability of skills and income growth of recent immigrants. Education, English-language skills and drive or motivation are the three sets of skills that one would assume are important in the determination of employment and wages for immigrants in the US. Nonetheless, the analysis of Chapter 4 showed that education and English language skills were not always strong determinants, but that, although small in magnitude, unobserved skills such as drive or ability (as measured by higher wages abroad) were always statistically significant, particularly for migrants with economic visas. The effect of these three skills on employment and wages differed by gender and affected the selectivity of immigrants from the general population in their country of origin. It was also found that, in addition to these skills, country characteristics,

27 10 race and other characteristics can explain the wages of immigrants in the US but they were not a good measure or proxy for unobserved drive and ability. Chapter 5 summarizes the main results of the dissertation. It also presents some lines for future research as well as the policy implications of the analysis.

28 11 CHAPTER 2. MIGRANT SELECTIVITY AND THE LABOR MARKET OUTCOMES OF U.S. IMMIGRANTS IN THEIR HOME COUNTRIES The relative economic situation of immigrants before they come to the US is important because this may influence how they perform in the US. This chapter studies the economic performance of US migrants in their country of origin and how their success in the labor market is connected to observed and unobserved individual characteristics, country characteristics and the selectivity of the migrant group. More specifically, it examines whether migrants with noneconomic visas have greater or lower success in their home country than those with economic-based visas. A comprehensive study on the characteristics and labor market outcomes of migrants before immigration has occurred (back in their home country) has not been conducted in the literature. What we know about migrants before immigration is limited to internal migrants (in the case of the US), migrants whose cost of migration is arguably very low (in the case of Germany 1 ), or to Mexican migrants. The New Immigrant Survey provides an opportunity to examine the home-country labor market performance of US immigrants. The rest of this chapter is organized as follows: Section 2 presents a brief literature on the labor market performance of migrants before leaving their place of origin; Section 3 presents the summary statistics of the New Immigrant Survey. The next 3 sections discuss the results from the labor participation 1 In 1990, virtually all barriers to immigration to from East to West Germany were eliminated; the migrants spoke the same language spoken in West Germany, they were of the same race, and they

29 12 analysis, Mincerian wage analysis and Heckman selection wage analysis. Section 7 concludes. I. Literature Review In the economics literature, one of the most controversial issues regarding immigrants is the extent to which, after controlling for age, gender, and other relevant characteristics, immigrants positively or negatively self-select 2 themselves from their country of origin. There is also the issue of whether the selectivity of migrants has been declining over time, an issue that is often described as whether the quality or skills of immigrants has been declining over time. Borjas (1994) argues that since the 1980 s, immigrants have become less skilled than US natives on average, while Davis and Weinstein (2002), for example, argue that the flow of immigrant labor skilled or unskilled has been more or less equivalent to the existing factor shares in the American economy. The most common, simplest and possibly the most adequate measure of the quality of immigrants entering the US is their distribution of education (in terms of number of years and/or degrees attained) and of their earnings in the American labor market. Economists use education to measure the quality of immigrants because education has been shown to predict wages (Hause, 1972; Mincer, 1974; Shultz, 1961; Welch, 1974) and it is linked to productivity (Chevalier, Harmon, Walker, & Zhu, 2003; Gintis, 1971; Griliches & Mason, 1972; Lazear, 1977). Still, education obtained in the country of origin might not be as 2 A migrant is said to be positively self-selected when he or she has above-average observable or unobservable characteristics (relative to nonmigrants), such as education or motivation.

30 13 relevant to productivity in the US, because the quality of education varies by country of origin and language of instruction. Years of schooling is still the single most powerful explanation for productivity in the US and around the world, regardless of any direct comparison between the education received by natives of the US and the education received by immigrants in their native countries and the comparison among education received by immigrants from different countries is possible. Differences in earnings have also been used to determine migrant selectivity. However, it is more difficult to compare earnings than to compare education between natives and immigrants. One of the reasons for this is that the self-selection determinants of labor force participation in the US are probably different for natives, legal immigrants and undocumented immigrants, which has implications for earnings comparisons. Additionally, earnings or income comparisons between natives and immigrants are usually limited to wages, ignoring earnings from self-employment. This presents biases in the results since immigrants are more likely to be self-employed (Borjas, 1986; Yuengert, 1995), and therefore, the analyses ignore capital accumulation for many immigrants. If self-employed workers were to be included in the analyses, the gap that exists between natives and immigrants in earnings could potentially be reduced by 14% on average (Lofstrom, 2002). Besides the selection into self-employment, another source of bias from assimilation analyses is attrition caused by emigrants. Depending on the characteristics of the immigrants, it is not possible to predict the direction of the bias. Assimilation or rates of wage convergence

31 14 could be underestimated (or overestimated) if the proportion of emigrants is defined as successes (or failures ). The existing literature on immigrants that touches on the immigrants skills and its transferability can be represented by one of three categories as follows: the quality of the immigrants in terms of their education, the quality of the immigrants in terms of their wages, and the self-selectivity characteristics of the immigrants. A. Quality of Immigrants Education The quality of the cohorts of immigrants that arrive in the US can be measured in terms of the amount of human capital they bring with them (years of education) or in terms of the income they earn in the American labor market. A recurring problem in the literature is that the standard measure used to describe the quality of immigrants, focuses on their level of education in their country of origin, and the quality of education in one country cannot be directly compared across countries. It is highly unlikely that, for example, eight years of education in the US is equivalent to eight years of education in a developing country in Africa (lower quality) or in Finland (possibly higher quality). Regardless of this limitation, level of education is still the best predictor of labor market productivity in the US and around the world. A second problem is that using the average level of education for immigrants causes a potential source of bias, because it has been shown that immigrants usually come from all areas of the distribution, and that the concentration varies by gender, as well (Chiquiar & Hanson, 2005).

32 15 Borjas (1995) wrote that the number of years of education of immigrants has fallen relative to natives, but the rapid increase of years of education of natives cannot solely explain this phenomenon. He attributed this change to the observed decline in the proportion of immigrants arriving from Europe. Older cohorts of immigrants are being replaced by an immigrant population that has a lower average number of years of education (Abowd & Freeman, 1991; Funkhouser &Trejo, 1993), and derive mostly from Asia, Mexico and other Latin American countries. Fix and Passel (1994) also attributed this change to the acceleration of the growth of undocumented immigrants. B. Quality of Immigrants Earnings and Assimilation An alternative measure of the quality of immigrants and their rate of assimilation is the measure of relative earnings of the immigrant worker compared to the native US worker. This measure is also flawed because the selfselection forces that drive an immigrant to leave his or her country differ between legal and undocumented immigrants as well as between American mainstream participants in the US labor market. Another source of bias when measuring the quality of immigrants is that from the earnings perspective, these studies usually limit the income source to wages and ignore self-employed income, even though self-employment is arguably a better measure of assimilation into a foreign country than wages alone. A third source of bias is attrition, since the available data is not able to identify those immigrants who have returned to their country of origin, nor their characteristics.

33 16 To measure the rate of wage convergence of immigrants relative to natives, it may therefore be more accurate to study the same immigrant cohort prospectively, and not different cohorts over the same time period, in order to avoid cohort variations (bias). Further factors that are important to consider are the immigrant s level of education, year of immigration and age of arrival in the new country. Reliable data on immigrants for purposes of analyzing wage convergence and selectivity is also limited. The original work of Chiswick (1978) used the 1970 US Census data to analyze the economic adaptation of immigrants. His methodology included a dichotomous variable that identified the immigration or native status of workers and the numbers of years that the immigrant had been in the US. Chiswick concluded that the rate of wage growth was greater for immigrants than for natives and that the negative effect found on the immigrant status dummy variable should be interpreted as the percentage wage differential between immigrants and natives at the time of entry. Chiswick explained his results as follows: when immigrants arrive in the new country, they lack country-specific skills, but as they assimilate into the new culture, their wages increase and surpass natives wages because immigrants are positively selected from their home countries and they choose to work harder and longer than natives. Borjas (1985, 1995), however, believed that Chiswick s results were biased in determining the self-selectivity issue of immigrants. If the unobserved characteristics of earlier immigrant cohorts differed from those of more recent cohorts, the results would possibly be biased upwards (or downwards) if the earlier cohort were positively (or negatively) selected. Using censual data, Borjas

34 17 (1985, 1995) observed that the difference between immigrants and natives wages had been decreasing to the point whereby the difference was now negative. Borjas (1995) attributed this difference to the lower level of skills of new immigrants of 1980 and 1990, who, according to him were more likely to be high school dropouts than their counterparts in Still, in this 1995 article, Borjas found that the quality of immigrants had declined over the past few decades. Borjas (1995) found that there existed large differences between cohorts and that the immigrants disadvantage over that of the natives had grown over time. Using the age-education deflator, recent immigrants in the late 1980s cohort earned between 19 to 27 percent less than natives. Similar immigrants who arrived between 1960 and 1964 earned only 2 to 8 percent less than natives. Although Borjas s (1995) observations may be valid, under different methodologies, other scholars (Chiswick, 1986; Duleep & Regets, 2002; Duleep & Regets, 1997a; Duleep & Regets, 1997b; La Londe & Topel, 1992) have shown that Borjas misinterpreted the data and concluded that there was no clear evidence that the quality of more recent immigrant cohorts had indeed declined. Rather than interpreting the lower earnings as lower quality in the cohorts, Borjas should have interpreted these as a decline in transferable skills due to the shift in the composition of country of origin of newer immigrants. Chiswick s (1986) and Borjas (1985, 1995) results still face the problem of attrition biases, because about a third of all immigrants return home. Censual data is limited in the sense that it is not possible for the econometrician to identify individuals from one census to the other (it is not

35 18 longitudinal), and it does not allow for the identification of immigrants who return to their country or those who immigrate multiple times. The magnitude of the bias is not negligible since there is a high likelihood that immigrants return to their countries and econometricians cannot identify which types of immigrant return home. The rate of wage convergence is underestimated (or overestimated) if the proportion of immigrants who return to their native countries is denoted as successes (or failures ). In the 1980 s, the average level of real wages was relatively stable, although there were significant changes in the structure of relative wages. The group where this change was more noticeable was identified as the highly educated workers (Bound & Johnson, 1992). Katz and Murphy (1992) attributed this change to the growth in the demand for more educated workers. LaLonde and Topel (1992) decomposed the changes in relative wages of immigrants into changes in skill and changes in wage structure. In the absence of assimilation, immigrants in the same category of age and education (or skill) would have the same growth that natives experienced during the same period. The difference between the growth experienced by natives and the growth experienced by immigrants was interpreted as evidence of assimilation by immigrants. Because income inequality increased in the US by skill level across immigrants and natives (LaLonde & Topel, 1992), less educated immigrants found themselves at an increased disadvantage in the labor market after Immigrants who arrived between 1965 and 1969 with less than 10 years of education were earning 45% less than natives, and their predicted wage differential without assimilation for 1980 was 50%. The observed differential was

36 19 actually -47%, so there was negative relative growth in their wages. Since the predicted wage differential was -50%, there was actual growth, but it was very small (3%). Using the census data, Borjas (1991) reported that immigrants ages 25 to 44 who arrived in 1980, were earning 26% less than natives compared to the 1% positive difference between those who arrived before 1940 and the natives (Schultz, 1998). However, data from the Current Population Surveys (CPS) from 1979, 1983, 1986 and 1989 painted a different picture for immigrants. The relative average educational attainment of immigrants improved for males 18 to 61 years of age; there was an increasing share of immigrants coming from Europe and Asia in the 1980s; the proportion of less educated immigrants continued to increase, but this trend only applied to immigrants from Latin American countries other than Mexico (Funkhouser & Trejo, 1993). Since the CPS data offered better income information, Funkhouser and Trejo s (1993) results should be considered more reliable than Borjas (1985, 1995) results from the Census data. Funkhouser and Trejo (1993) attributed the increase of relatively highly educated European and Asian immigrants in the 1980s to the increase in income inequality that occurred in the US (Murphy & Welch, 1992), which benefited the educated workers disproportionately, and therefore increased the supply of educated immigrants. This increase in income inequality cannot fully explain the decrease of relative wages of immigrants during that period and thereafter. Duleep and Regets (2002) argued that three errors are commonly made in this area of research: equating earnings to human capital, assuming that earnings growth

37 20 rates are constant across entry cohorts in cross-sectional models and assuming that quality is measured by differences in earnings at time of entry. Their study found that there is a strong inverse relationship between entry earnings and earnings growth. The hypothesis was as follows: if the proportion of skill transferred was used for learning rather than for earning wages, then it would be less used for productivity (earning) and the gap between the proportion of skills transferred to learning and the proportion transferred to earnings would grow wider. However, skills not valued (through higher wages) in the labor market are not necessarily wasted since they can be used to produce new (human) capital. The negative correlation between initial earnings and income growth found in this study suggested that lower initial earnings within education and age groups are also associated with faster growth. Older, more educated immigrants experience less growth in earnings in the first 10-year period since migration. However, from among the younger population, those who experience faster growth rates when they entered with relative low wages were the highly educated population. These results agreed with LaLonde and Topel (1992) which correlated the 1970 earnings of immigrants with earnings growth rates. They also agreed with Schoeni, McCarthy, and Vernez (1996) who found that East Asian immigrants had initial low wages, but a fast wage growth and that European immigrants had high initial wages, but a slow wage growth. C. Self-Selection

38 21 One of the main topics in labor economics is self-selection due to the underlying assumption that rational individuals make optimizing decisions in job participation, education, etc. based on their tastes, skills and expectations about the future. Therefore, effects are not considered exogenous causal relationships, but rather they are endogenous outcomes of an optimized decision. The Roy model was the first attempt in economics to address this issue based on the factors that Roy (1951) identified to affect the choices that workers make. These factors were the distribution of the skills of the population, the correlation among their skills, the technologies available to apply these skills and the demand for different outputs. Roy used a two index model (two occupations) and he assumed that the population had skills for both occupations, but could only work in one of them. Workers self-select themselves into the occupation that will give them the highest expected earnings. The distribution of income observed, thus, is a truncated distribution of potential earnings of individuals who have selected themselves into that particular activity. In his 1987 reputable contribution, Self-Selection and the Earnings of Immigrants, Borjas applied the Roy model to the issue of self-selection in migration, and presented it in a more formal manner (Roy s presentation of the model was done in a narrative manner, the way it used to be done more than half a century ago). Borjas (1987) findings and conclusions have been challenged intensely (Chiswick, 1999; Duleep & Regets, 2002; Chiquiar & Hanson, 2005; Jasso & Rosenzweig, 1988; Jasso & Rosenzweig, 1990). Borjas (1987) results showed that although the proxies he used in order to measure political and economic conditions (number of assassinations, income

39 22 inequality, political competitive system and recent loss of freedom) could weaken the link between the theory and the empirical results, there was a negative relationship between income inequality and the wage differential between natives and immigrants. Assimilation was measured in a 10-year interval. Borjas (1987) showed that immigrants from wealthy countries tended to assimilate faster than immigrants from poorer countries. However, coming from a country in Africa or Asia had a positive impact, while coming from a country in South or North America had a negative impact, compared to those coming from Europe. In terms of change of cohort quality, the specifications were slightly different because the exogenous variables reflect changes over time. The change in level of GNP per capita had a positive impact on the change of cohort quality. In terms of the determinants of emigration rates, Borjas (1987) found that the distance between the US and the source country had a negative impact on the emigration rate, while high GDP levels in the source country had a small negative impact in emigration. He also found that countries with a greater income inequality had lower emigration rates, which he interpreted as negative selection. Borjas (1987) final conclusion asserted that if the source country had greater income inequality, immigrants were negatively selected, and that immigrants from developed countries performed well in the US, but that the same could not be said for immigrants from developing countries. The major contribution of this paper was the clear mathematical representation of the Roy model in the context of immigration. However, there has been no consensus on his empirical results. Borjas (1987) results could have

40 23 alternative interpretations. The negative sign associated with income inequality regressions does not necessarily imply that the immigrants come from the lower tail of the income distribution. This result can be interpreted, following Roy s model s predictions, as a less positive selectivity. It is also not clear that Borjas (1987) measure of income inequality actually measured selectivity. It is possible that it only measured the degree of selectivity or another source of inequality. After controlling for additional variables, the negative coefficient of this variable lost significance and in some cases it became positive. There are more alternatives to the human capital model of immigration developed by Sjaastad (1962) that have appeared in the literature to address the issue of selectivity and assimilation. Some of these include asymmetric information, temporary migration, and non-economic determinants of migration. The asymmetric information model in migration was introduced by Katz and Stark (1984, 1987). When employers in the destination do not have information on the ability of the immigrant (which is known only to the immigrant himself), the employer pays according to the average immigrant productivity (expected productivity), therefore, it is more costly for high ability workers to migrate since their opportunity cost will be higher while at the same time their relative wage would be lower than for the low ability worker. If employers are not able to detect true levels of ability, high ability workers would not have any incentive to migrate, since the flow of low ability workers would increase, and therefore the average (expected) wage would decrease. On the other hand, if employers are able to develop a system to identify ability, then the

41 24 shorter the time it takes to adjust wages according to observed ability the higher the incentive for high ability workers to immigrate. The second alternative model drops the assumption that the worker remains in the host country indefinitely or at least for a significant number of years. This is the case of return immigrants (voluntarily or involuntarily). If immigrants were to invest in host-country-specific skills, there would be a loss after departure, and if the immigrant were to invest in source-country specific skills, he or she might run the risk of having depreciated skills when he or she returns to the source country. Their options, therefore, seem to be limited to the investment in transferable skills only or in no skills at all. Therefore this group of immigrants is expected to have lower level of skills than those who plan to immigrate permanently. Looking at undocumented immigrants, this prediction has been empirically supported by Rivera-Batiz (1999), and Kossoudji and Cobb- Clark (1998), among others. People are driven by incentives, but not all incentives are economic incentives. Therefore, the determinants of migration are not limited solely to the extent of earnings or other economic opportunities. Mincer (1978), in his study on family migration, identified what he called tied-movers (move to join family members) and refugees (or ideological immigrants) as groups of immigrants whose incentive to immigrate was not purely economic. His study and others done by Chiswick (1978, 1979, and 1982) concluded that tied movers are less favorably selected and tend to have lower employment rates and earnings. This disadvantage, however, seems to decrease overtime, as the groups assimilate to their new country of residence.

42 25 Following a different approach, Chiquiar and Hanson (2005) tested Borjas (1987) negative selection hypothesis using censual data from Mexico and the US. To test for selectivity, Chiquiar and Hanson (2005) compared the income distribution of Mexican residents and Mexican immigrants in the US. Their preliminary results showed that men are concentrated in the third and fourth quintile and that women are concentrated in the first two quintiles, that is, men are positively selected on observables and women are negatively selected on observables. The distribution of wages for immigrants differs from that of residents of Mexico possibly due to the difference in the distribution of skills and to differences in prices paid for the same skills. To correct this issue, the authors examined the wage distribution for both samples under a common set of skill prices and considered the rate of labor force participation, following the techniques developed by DiNardo, Fortin, and Lemieux (1996). From the 1990 data, the actual density and the counterfactual density were very close to each other; however, they still presented some insights in the subject of selectivity. It appeared that men who were more likely to immigrate came from the middle of the distribution, although between the medium- and high- wage earners, the former were more likely to migrate. For women, there was even less support for negative selection. The positive difference between the densities was more pronounced for females than for males. It is important to remember that the results shown here were based on observed characteristics (education). If there was a high correlation between the observables and the unobservables, such as access to migration networks or

43 26 motivation, then these results are potentially reliable. If, on the other hand, the correlation was low, these estimates would not be able to provide information on self-selectivity. Another issue was that of attrition. Censuses undercount undocumented immigrants. A sufficiently high proportion of undocumented immigrants are likely to alter these results. Another weakness of the research by Chiquiar and Hanson (2005) was the data used for their study. In order to estimate the wage density of immigrants (counterfactual), they used the characteristics of Mexican immigrants observed in the US Census, since this data was not provided in the Mexican Census. US Census data undercounts immigrants, especially undocumented immigrants, who are more likely to come from Mexico. Fernandez-Huertas Moraga (2011) used the Encuesta Nacional de Empleo Trimestral (ENET) data from Mexico, which is similar to the CPS of the US. Using the data from ENET, Fernandez-Huertas Moraga (2011) observed immigrants before they left Mexico and estimated the counterfactual density using actual observations. Fernandez-Huertas Moraga concluded that negative selection exists among Mexican immigrants. He attributed the difference between his results and those of Chiquiar and Hanson (2005) to the different data used for the study. D. Immigrant Labor Market Performance in the Source Country and the Selectivity of Migrants The existing literature that examines the characteristics and labor market outcomes of migrants before they move has been limited to migrants within one

44 27 country since this type of information (wages, employment before migration and other outcomes) is likely to be available in easily accessible and available data sources, such as the US Census and IPUMS in the US and the German Socio- Economic Panel in Germany. For the US, Lansing and Morgan (1967) found that there is no difference between migrants and non-migrants, however, Wertheimer (1970) and Masters (1972) did find a difference among black individuals and that the gain from migration to the North was larger for blacks than from whites. Therefore, Greenwood (1975) suggested taking this type of characteristic into consideration when analyzing migrant characteristics. Distance has an inverse relationship with immigration probability as information about the destination decreases and hence uncertainty. That is, the farther the destination is from the home the lower the information transference factor (Greenwood, 1975). In the US, the personal characteristics that seem to affect the probability of migration are age, education and race. Age has an indirect relationship with the probability of migration. Young individuals have fewer ties with their place of origin, while older individuals seem to appreciate tenure and family ties more than their young counterparts (Gallaway, 1969). In addition, older individuals have less time to realize the gains of migration making migration less profitable, while young individuals who can profit from migration, benefit the most moving sooner rather than later since the initial gains from migration are discounted at a lower rate than later gains (Becker, 1964). This may explain why we can see even in international studies that immigrants are younger than the average individual from their country of origin.

45 28 In terms of education, for short-distance migration, there is no real difference between the low educated and the highly educated, but the correlation between education and migration increases as distance increases (Folger & Nam, 1967; Hamilton, 1965), probably because information and opportunities increase with education (Schwartz, 1973). Although unemployed individuals are not more likely to migrate, nonwhites are more likely to be unemployed when they migrate than white individuals, but this is more related to their level of education and occupation than their race. Nonwhites tended to be in occupations that were less likely to offer job transfers than were whites, therefore, it was observed that whites were more likely to be employed when they moved than nonwhites (Greenwood, 1975). Among black migrants from the South to the North, Tolnay (1998) showed that black migrants were less likely to be from rural areas in the South and had more education than the average black stayer, but compared to the blacks in the north, they had lower levels of education. The gaps between stayers and migrants and blacks from the north and migrants have been getting smaller in the period covered of 1880 to The German Socio-Economic Panel, a longitudinal study of West German adults beginning in 1984, started to collect data on East Germans in 1990 after the reunification of the Federal Republic of Germany and the German Democratic Republic in that year, and they followed those that migrated to the West as well as the ones who stayed in East Germany. Since migrants were followed if they moved the West, there is pre- and post-immigration information on East Germans regarding employment and wages, among other information,

46 29 allowing Burda and Hunt (2001), Hunt (2001, 2002, 2006), and Zaiceva (2010), among others to examine the characteristics, performance and selection of migrants in Germany. Hunt (2001) and Burda and Hunt (2001) found that migrants have more education than the stayers of East German and that the young and highly skilled are more likely to migrate, confirming the findings of Gallaway (1969) and Schwartz (1973) from studies on American internal migrants. Although Zaiceva (2010) looks specifically at migrant women, who if married are most likely tied movers or non-economic migrants, she does not examine how non-economic migrants (those who moved to join their husbands in the West) compare to economic migrants before migration. Identifying the characteristics and labor market outcomes of international migrants before they move is more challenging, since the data collected in the receiving countries is limited to their performance in that particular country. A notable exception is data on Mexican immigrants to the US. It appears that both countries, the US and Mexico, have made a commendable effort to collect and make available data 3 on Mexican migrants who intend to move or have moved to the US, being that Mexicans form the largest immigrant group in the US and the US is the number one destination of Mexican migrants. The Mexican Migration Project (MMP) was created in the US and samples undocumented Mexican migrants over time and non-migrants living in Mexico (it ignores heads of households living in the US permanently). Besides finding that immigrants come from the middle of the education distribution, Orrenius 3 These data sets will be mentioned and briefly explained throughout this review of the literature.

47 30 and Zavodny (2005) found that low-skilled immigrants were less likely to migrate when there was greater border enforcement. Similarly to Orrenius and Zavodny (2005), Chiquiar and Hanson (2005) found that Mexican immigrants come from the middle of the education and skill distribution of Mexicans using the Mexican Census and the US Census. However, they did not have employment or wage information for Mexican immigrants while they were in Mexico, so they used predicted wages instead of real wages to draw their conclusions. An improved study was conducted by Fernandez-Huertas Moraga (2011) who used the Encuesta Nacional de Empleo Trimestral (ENET) which is structured similarly to the Current Population Survey of the US. The ENET follows the household for five consecutive quarters and reports if a household member moved to the US, making it possible to examine the labor market characteristics and selectivity of recent Mexican migrants. 4 He found that Mexican migrants earned less than non-migrants, and that migrants come from the middle and lower part of the education distribution. The Encuesta Nacional de la Dinámica Demográfica (ENADID) is another survey that collects data on migrants as long as there is at least one member of the household that remains in Mexico. McKenzie and Rapoport (2010) use this data and their results agree with Fernandez-Huertas Moragas (2011) that immigrants come from the bottom of the education distribution. However, they add that immigrants who come from communities that do not have high rates of migration have higher levels of education than non-migrants. 4 If the entire family moves to the United States, the household is dropped from the survey sample.

48 31 There exists another data set that overcomes all the weaknesses of the previous data sets. The Mexican Family Life Survey (MxFLS) is a longitudinal representative survey of Mexicans, which will follow Mexican households for a long period of time, regardless if the household decides to leave Mexico. The survey is a collaboration between Mexican and American institutes and researchers. As previous research has shown, migrants in this data set are more heavily represented among young adults than among older ones. Kaestner and Malamud (2010) use this survey to measure self-selection between Mexican migrants and non-migrants and they found that migrants are just as likely as non-migrants to be employed. They also find that the skills of the immigrants that we observe in the US depend on the returns to skills in their country of origin (Mexico, in this particular case) and on the cost of migration. Their results may not describe all Mexican immigrants in the US since the data is limited to recent immigrants only. II. Data Description and Data Analysis The New Immigrant Survey offers a fresh opportunity for exploring the skill transferability of immigrants, because it is the only publicly-available and representative dataset that provides documented immigrants wage information before and after immigration to the US. 5 The nationally representative sample of 8,573 respondents was selected from administrative records that the US government collects on new immigrants through the US Citizenship and 5 Although an estimated 8% of the sample entered the United States with tourist visas, bordercrossing cards or without inspection, all respondents were currently Legal Permanent Residents.

49 32 Naturalization Service (USCIS). 6 The data on the first and only available cohort to this date was collected between May and November The data set consists of two samples, adults and children. Table A1 presents the summary statistics by country of origin. Individuals were surveyed in their language of preference in order to maximize the response rate and the quality of the data. The designers of the survey estimated that less than 50% would prefer to be interviewed in English. The survey was fully translated into Spanish, and all instruments were translated into Chinese, Korean, Polish, Russian, Tagalog and Vietnamese. Only a set of key concepts was translated into Arabic, Croatian, Farsi, French, Gujarati, Hindi, Serbian, Ukrainian and Urdu. Bilingual interviewers conducted the interviews in Spanish, Chinese, Korean, Polish, Russian, Tagalog, Vietnamese, Amharic, French and Haitian Creole. All other interviews that required translation were conducted with an interpreter. The instruments of the survey included the topics of health, schooling, marriage and family, skills, languages, labor force participation, earnings, travel, and religion, among others. Table A2 contains a list of the 15 languages most commonly spoken by the adults in this sample, which covers 61% of the sample. A. Summary Statistics 6 12,500 immigrants were initially selected for the sample, but only 8,573 completed the interviews (a 68.6% response rate).

50 33 The weighted sample of the New Immigrant Survey consists of 8,488 observations. Out of those, 37.7% were employed for pay at some point in their country of origin. The average weekly wage was $2,051 adjusted for PPP prices. Table 1 reports the summary statistics for the sample by employment status. An immigrant is considered to be employed abroad if he or she reported wages in one or two jobs in his or her country of origin. The higher wage of the two reported was used for the analysis. Employed immigrants had, on average, two years of education more than unemployed immigrants, but they had less experience. Experience was calculated in the traditional manner, but because the individuals were asked about any job ever held abroad, the interpretation for experience abroad, if no jobs were ever held, should be years out of school. The survey asks for current marriage status but does not ask directly what the marriage status was at the time of employment abroad. Using information about the number of marriages, the length of each marriage, the year the marriage took place, and the periods of employment, a marital status abroad variable was constructed for all individuals. However, this information did not allow clear identification of those who were living together but not married, so most of them may had been identified as single rather than living together. Employed immigrants were more likely to be single at the time of employment while unemployed immigrants were more likely to be married before they immigrated to the US. Of the married sample, about 40% were employed compared to 36% employed among the single sample.

51 34 There is no direct information in the survey as to whether the individual was a student in his or her country of origin right before he or she immigrated. Using information about the year and country in which the immigrant obtained his or her highest degree and the year he or she arrived in the US, a student variable was constructed that was equal to one if the year of the degree outside of the US was the same as the year he or she arrived in the US. If the year or country of degree were missing, then the age of the individual plus the number of years of education abroad were used to estimate in which year (minimum, if there were no breaks during his or her schooling) he or she should have finished schooling and compared this to the year they left their country of origin. Among the employed 11.2% were students compared to 20.7% among the unemployed. 17.1% of all students were employed. Using information about children s living or death status, period of employment or year of immigration for the unemployed, year of birth for biological children, year of adoption for adopted children, and year of marriage for step children, the number of children that immigrants had while they were working or before they immigrated for the unemployed was computed. It was also calculated how many of those children were six years of age or younger. Employed individuals had more children in total and also more children under the age of six. Immigrants, who were ever employed, left their country of origin at a later age than those who were not employed, 33.5 and respectively. Employed individuals were more likely to live in a country with higher average schooling, but with very similar average GDPs per capita.

52 35 When a foreigner applies for a legal permanent residence, they are given different types of temporary immigrant visas, depending on the path used to immigration. The paths could be a relationship (spouses, parents, children of) with US citizens or Legal Permanent Residents (LPRs), refugees or asylum seekers, and whether they were in diversity, legalization and employment categories. This variable was called visa category in the survey. Diversity visas was the term used to describe those visas obtained in a lottery and legalization was the term used to describe those immigrants that had entered without inspection or had stayed in the country after their non-immigrant visas expired and then adjusted their status to LPRs. Employed immigrants were more likely to become LPR in the US through marriage to a US citizen, being a parent of a US Citizen or through legalization. Children of US citizens and spouses of LPRs were less likely to be employed than anyone else. Diversity immigrants and employment visa holders were the ones most likely to have been employed in their country of origin. Based on the country of the last job abroad or country of origin and the year they left the country, Table A1 also shows the real GDP per worker from the Penn World Tables, and the county s average schooling levels of the population 25 years of age or older from Barro and Lee (1993), if available. In Table 2, the survey s summary statistics are presented for women and men by employment status separately. Out of the 4,793 women, 33.6% were employed, compared to 43% or 1,590 out of 3,695 men. Weekly wages were higher for men than for women, and men worked about 2 weeks more per year than women. Employed women had years of experience and employed men

53 36 had about years. Among the unemployed, women had been out of school on average for 18 years and men had been out of school for years. Employed women were slightly more likely to be single than married, while unemployed women were more likely to be married. Among men, those employed were more likely to be married and those unemployed were more likely to be single. About 36% of both single men and women were employed. Only 32.7% of married women were employed, compared to 52.6% married men. The percentage of students employed was about the same among men and women, however among those employed and also unemployed, a higher percentage of men were students than among women. On average, unemployed women had more children than employed women and unemployed men, but employed men had the most children. Employed men also had the highest number of children under the age of six, followed by employed women. Unemployed men were the youngest when they left their country of origin, at 29 years of age. Unemployed women left earlier than employed women. Employed men were the oldest when they left, at almost 34 years of age. Unemployed women were least likely to speak English. 37% of those who spoke English were employed compared to 41.9% of men who spoke English. 41% of employed women had a partner who was also employed, compared to 48.5% among employed men. Only 9.7% of unemployed men had a partner who was employed. The highest wages earned by spouses were from those whose wives worked.

54 37 In addition to spouses of US citizens (the most likely category for all), employed women were more likely to have had an employment preference when they immigrated while unemployed women were more likely to be parents of US citizens. Employed men were more likely to have had an employment preference or be a diversity immigrant, while unemployed men were more likely be legalized. Among women, diversity immigrants and employment preference were more likely to be employed. Among men, those with family preferences and diversity immigrants were most likely to have been employed. B. Labor Force Participation in Country of Origin Suppose that the reserve wages of an individual depends on individual and their country of origin characteristics. If the reserve wage is below the offered wage, the individual decides not to work, and we don t observe wages for them. (1) * Y if Y 0 and * Y 2 1 Where Y missing if Y * 0, 1 * Y1 is the reserve wage of the individual and * Y 2 is the actual wages for those who work. Since we do not observe the reserve price, but only whether the individual works or not, I use a probit estimation for binary responses (works and does not work) to determine labor market participation. x' (2) F( x) Pr( y 1 x) ( t) dt ( x' ) x 1 exp 2 1 ( 2 x) 2 dx where ( x' ) is the cumulative density function for the standard normal and β is the parameter of interest. The marginal effect of the individual characteristic k

55 38 (Xk), where k= 1, 2, 3,, K, on the likelihood of employment is Φ(XB)*βk where βk is the estimated parameter associated with Xk. The marginal effects are usually estimated at the mean of the X variable. 1. Full Sample Results In this analysis, unemployment status was defined when the immigrant reported in the survey he had no wages from any job while living abroad. Table 3 reports the probit regression results for the full sample of immigrants. All regressions controlled for year when they left the job and were weighted by the survey weight provided by the Survey. The marginal effects are reported next to the estimated beta coefficients. 7 The results in the first column controlled for gender, education, experience, marital status and for whether the respondent reports speaking English. Women were 7% less likely to be employed than men, and the average change in probability of employment increased by 2.2% for each year of education. The marginal effect of experience was positive and statistically significant. Single individuals and those who are living together, but not married were less likely to be employed than those who were married. Speaking English did not have an effect on employment. The selectivity among immigrants from different countries, independent of their individual characteristics, can be assessed with country-level characteristics, which affect which type of person exits their country to migrate to 7 Marginal effects were calculated using the margeff command in Stata. It calculates the average effects for all the points in the distribution and reports the average.

56 39 the US. We expect the characteristics of immigrants to differ by country. Therefore, column (2) includes country of last job or country of origin fixed effects. The effect for women compared to men and experience remain virtually unchanged, while the effect of education decreased the magnitude by almost one third, implying that education has different effects on employment for immigrants from different countries. Speaking English (self-reported) is now statistically significant and it decreased the probability of employment by 2.8%. The ability of immigrants to speak English, holding the country of origin constant, may be affected by their employment status in the country of origin. For example, if an individual plans to immigrate to the US because of the lack of employment in their country of origin, then he or she may invest in learning English and therefore the results would show a negative effect on employment. This negative effect cannot account for the possible differences between Anglospeaking and non-anglo-speaking countries. In column (3), more characteristics are added that affect the decision to enter the labor market, but do not affect wages directly. Students were 3.6% less likely to have been employed and the older the person was to leave the country the more likely he or she was to have been employed, holding experience and education constant. Although the number of children under the age of six did not affect labor market participation, individuals with more children were more likely to have been employed. Controlling for marital status, immigrants whose spouses worked were 18.4% more likely to have been employed, than those with unemployed spouses.

57 40 Speaking English was no longer statistically significant after controlling for student status and other selection characteristics. A possible explanation for this may be that there is a positive correlation between speaking English and being a student, but there is a stronger and negative correlation between the age when the person left the country and speaking English. So English may have been a proxy for young students in column (2) which did not control for this characteristic. Because this data set provides information on the type of visa used for immigration, we can also observe whether there is a difference in employment abroad between migrants with economic-based and non-economic-based visas. Although the true motivation for migration cannot be assumed based solely on the visa category, there is still an implication that the initial motivation for migration is either economic or non-economic. Refugees and those who had a family relation to a US citizen or to a legal permanent resident are said to have migrated with non-economic-based visas. Migrants with economic-based visas were those who migrated illegally, on an employment preference visa or with a diversity visa. For simplicity, I will call immigrants with economic-based and non-economic-based visas economic-visa and non-economic-visa migrants, respectively. 8 I did not find a difference between these two groups for the whole sample. Taking into account that the probability of employment depends on the economic conditions of the country, in column (4), the unemployment rate before 8 The dummy variable in the tables is called Non-economic Migrants

58 41 migration, country s average schooling and log of GDP per capita are added. The average marginal effects of most of the variables decreased in column (4) and some by more than half their size, which shows that the probability of employment of immigrants is not only affected by microeconomic conditions, but by macroeconomic conditions as well. Results showed the unemployment rate was positive and statistically significant. Immigrants were more likely to have been employed in countries with higher rates of unemployment. Average schooling and log of GPD per capita were also statistically significant, but negative. This implies that holding GDP per capita and years of education constant, individuals that come from countries with lower levels of education are more likely to have been employed. Negative log of GDP per capita implies that immigrants who come from richer countries are less likely to have been employed before they immigrated to the US. How can this be explained? Note that the cost of immigration includes indirect costs such as lost wages. Not having a job decreases this type of indirect costs if we assume that his or her wages would have continued to be zero had he or she not migrated. If costs are positively related to the wealth of the country, where individuals from richer countries earn higher wages than individuals with the same characteristics but from poorer countries then employed individuals in wealthier countries have higher indirect costs than those employed in a poorer country. Therefore, holding individual observable characteristics constant, individuals who are employed in countries with higher GDP per capita are possibly less likely to immigrate because of their high indirect costs than their unemployed individuals

59 42 counterparts. Therefore one can observe that immigrants that came to the US from wealthier countries were less likely to have been employed before they immigrated. Experience abroad cannot be appreciated by looking at the average marginal effect of experience and experience squared, since the change in the marginal effect changes as experience changes. Figure 1 shows the relationship between employment and experience. The marginal effect was positive but decreased with each additional year of experience. The marginal effect became negative at about 14 years of experience. With each additional year, the probability of employment became more and more negative. In the Appendix, Table A3 presents the country effect on the likelihood of employment. The controls for each column are those variables shown in Table 3. Independent of all other characteristics, we can still observe differences among immigrants from different countries, in terms of employment abroad. Holding constant English language, country-level characteristics and immigrant type (column 4), immigrants from all countries, except for Nigeria, Russia, and United Kingdom, were less likely to be employed than those that came from Canada. The least likely group was from Mexico, followed by El Salvador and Haiti. 2. Results by Gender It is generally accepted in the labor economics literature that the determinants of labor force participation differ for men and women. Therefore, the next step in the research was to separate the immigrant sample by gender and

60 43 analyze their determinants. The probit regression results are presented in Table 4 for men and Table 5 for women. In Table 4, column (1), the only controls are experience, education, marital status and whether the person spoke English. The effect of education is positive and statistically significant. Education increased the probability of employment by 1.7%. Immigrant men living together abroad or who were single were less likely to be employed than those who were married. Speaking English had no effect on employment for men, even when country fixed effects were included in column (2). The effect of education decreased, but it was still positive and statistically significant. Column (3) includes immigrant type and other variables believed to affect labor market participation even if these characteristics are not directly compensated through wages. Non-economic-visa migrants were more likely to be employed than economic-visa migrants. Holding everything else constant, an economic-visa migrant will pursue better opportunities if he is unemployed while for a non-economic-visa migrant, employment status would be irrelevant or not as important. Adding number of children, number of children under the age of six, spouse employment status, student status and age at the time the immigrant left the country increased the marginal effect of living together from to and of single men from to Being a student had no effect on the probability of employment. Number of children had a positive effect on men employment while number of children under the age of 6 has a negative effect. Having a wife or a partner in the labor market increased the probability of employment, which is an

61 44 indication that couples share observed and unobserved characteristics, such as motivation. How much the spouse made, however, did not have an effect on their participation. Age at the time they left country of origin had a positive and significant effect on employment. Column (4) adds country-level characteristics. Adding the average level of education reduced the marginal effect of education from to and it was no longer statistically significant, while experience and experience squared remained unchanged. Age at the time they left the country is no longer statistically significant, either. The effect of number of children decreased in size by half; the effect of young children decreased by 0.01 points. If the effect of family structure differs by country of origin then these selection variables were capturing country effects in column (2). Although speaking English did not affect the probability of employment, individuals from English-speaking countries were less likely to be employed. Unemployment rate was positive and statistically significant. Holding the log of the GDP constant, men that lived in a country with lower levels of education were more likely to have been employed than those that lived in a country with higher levels of education. There is a high correlation between the age when the immigrant left the country and their level of experience since both of them were calculated based on age at the time of arrival in the US. Experience was measured as age when they arrived in the US minus years of education abroad minus six unless the year the individual left his or her last job is available. If the year when the individual left their last job was known, that year was used to measure working experience since being employed abroad was defined by having had a job abroad at any time and

62 45 not having had a job immediately before leaving their country of origin. About 40% of those employed had a job within the last 3 years before immigration. Age when they left the home country is the age when the individual left the country of origin for the very first time, whether it was to come to the US and stay, come to the US and go back or move to another country. This variable arguably measures how much country-specific knowledge the individual has accumulated that could help them succeed at finding a job in their country. In column (3), age when the individual left the country and experience abroad are no longer statistically significant. In regressions not reported here, the age variable was removed to test whether the effect of experience was being dissolved by it, but found that experience remained statistically insignificant. In the Appendix, Table A4 presents the country effect on the likelihood of employment. The controls for each column are the corresponding columns in Table 4. After controlling for English language, country-level characteristics and immigrant type (column 4), and compared to Canada, only immigrant men from Nigeria, Russia and the United Kingdom were not less likely to have been employed than men from Canada. Men from the rest of the identified countries were less likely to have been employed. The most disadvantaged groups came from Mexico, El Salvador and Haiti. The relationship between experience and employment is better represented in Figure 2. The effect of experience fell faster for men than for women and it eventually became negative at about 14 years of experience. Table 5 reports the results for women and their labor force participation.

63 46 The results from women varied significantly from men. Controlling only for experience, marital status and speaks English (column 1), the effect of education on employment for women was almost five times larger than for men. Marital status was not statistically significant, nor was the ability to speak English. In the second column, the effect of education was reduced from 8.4% to 1.8%, after country fixed-effects were included. Single men and men who lived in a marriage-like relationship were less likely to be employed, while women who lived in marriage-like relationship were just as likely to be employed, but single women were more likely to be employed than married women. Speaking English remained indistinguishable from zero, similar to the men sample. Additional characteristics were added in column (3) without changing the results too radically. The results showed that the likelihood of employment in the home country for women did not vary by student status, immigrant type, or number of children. Those whose spouses were employed abroad were more likely to have been employed themselves. Age when they left the country had a positive effect on employment. After controlling for country-level characteristics (column 4), and unlike men, age when they left their country had a positive and statistically significant effect on women employment. For women, the unemployment rate was not statistically significant, while average years of schooling and log of GDP per capita remained negative and statistically significant. Table A5 in the Appendix reports the country effect on the likelihood of employment. The controls for each column are the corresponding columns in Table 5. Immigrant women from Nigeria and Russia were just as likely to be

64 47 employed as those from Canada. The rest of the women from the identified countries were less likely to be employed in their country of origin than immigrant women from Canada. Women from Mexico, El Salvador and Guatemala were the least likely to be employed. C. Returns to Education and Wage Determinants in the Country of Origin Wages are determined by education, experience, experience squared, marital status and other sometimes observed or unobserved individual characteristics. The linear relationship between the log of wages and these characteristics is traditionally referred to as the Mincerian wage equation, and it is estimated using Ordinary Least Squares. The rate of return to education is the estimated coefficient on the education variable. (3) Yic 0 1X ic 2Zc ic is the Mincerian wage equation and Yic is the log of weekly wages (in the PPP prices) of immigrant i in his or her country of origin c; X ic is a vector of the individual characteristics of immigrant i, including education; Z is a vector of country-level characteristics; and ic is the individual error term. c 1. Full Sample Results Table 6, reports the OLS regression results based on equation (3). Except for marital status, all results are of the expected sign and magnitude. Women earned 58% less than men abroad. Without country fixed-effects, the rate of

65 48 return to education was 10.1% and income increased with experience but at a decreasing rate. None of the marital status categories were statistically significant. Individuals who worked more weeks per year earned 1.2% more on average per week. Immigrants that reported speaking English earned 58% more than those who do not. Some characteristics may be compensated for differently in different countries, therefore I controlled for country of employment in the second column. Although the differences in pay between men and women did not change after controlling for country of employment, the rate of return to education decreased to 7.2%, while experience decreased to 6.6%. Speaking English could have signaled in the first specification that those who worked in English-speaking countries earn more than those in non-english speaking countries, however, after controlling for country of employment, the English variable remained statistically significant, and the effect decreased by more than half its size. The rate of return to education is 7.2%, which was smaller than when country fixed-effects was not included in the specification. To estimate the selectivity among immigrants from different countries, holding constant all other characteristics, in addition to the country dummies, column (3) adds the country of origin s log of GDP, average schooling, unemployment rate and whether the country is an English-speaking country. The estimates in column (3) remained almost the same when I added these four variables, except for the rate of return to education. Unemployment rate and whether one of the official languages is English had no effect on wages abroad. Holding everything else constant, and keeping in mind that these wages have been adjusted by PPP, individuals working in richer countries earned more than

66 49 those with lower GDPs. There is a negative relationship between earnings and country s average level of schooling. Since more than 95% of the sample had more education than the average adult, the larger the gap between the individual s years of education and the country s average schooling, the more the individual earned. One of the most important issues in immigration research is the quality or skills of immigrants in the US. Highly skilled immigrants, as measured by their level of education and special skills, who do not have a family connection in the US, can obtain an employment visa to immigrate to the US. This group of immigrants, together with those with diversity and legalization 9 visas, was mainly economically motivated to immigrate. Immigrants that had family-related visas, such as marriage, were not necessarily economic migrants, or at least not primarily. Non-economic-visa migrants earned less in their country of employment showing that this group is less motivated or of lower abilities than economic-visa migrants, holding everything else constant. These results are shown in column (4). In the Appendix, Table A6 reports the results for each country using the same specifications as in Table 6. The excluded country is Canada. Of the identified countries, nobody earned more than immigrants from Canada, while some earned comparably. The biggest deviations were observed from Mexico, Haiti and the Ukraine. That is, holding everything else constant, immigrants from 9 These groups of immigrants do not have family connections to immigrate to the United States either.

67 50 Mexico, Haiti and Ukraine earn the least, implying that they may be negatively selected (or the least positively selected) among all immigrants to the US. 2. Results by Gender Next, the wages for men and women are analyzed separately and the results are reported in Table 7 and Table 8, respectively. As shown, the return to education was greater for women than for men. Women, unconditional on country of employment, had rates of returns of 12.8%, while men had a return rate of 8%. Experience, as expected, was positive and the effect increased at a decreasing rate. The effect was about one-third larger for men than for women. In contrast to what is expected among working men and women in the US, single men earn more than married men, while single women earn 30% less than married women. Effort, as measured by weeks worked per year, was positive and statistically significant for both genders, but the effect was greater for men than for women. Unconditional on country of employment, men who spoke English earned almost 30% more than men who did not, while women who spoke English earned almost 90% more if they spoke English, compared to women who did not. Conditional on country of employment, the return to education for men decreased to 6.1% and for women to 8.8% (column 2), while experience remains unchanged for women but decreased for men. After including country fixed effects, the difference in pay between married and single women disappeared while the difference for men increased even more. Speaking English for men had no effect on wages, but women earned about 46% more if they did. Just as done

68 51 in the full sample analysis, column 3 includes country-level characteristics in addition to the country fixed-effects. Controlling for unemployment rate, average schooling, GDP per capita and for whether the country is an English-speaking country did not affect the previously discussed independent variables. For both men and women, the unemployment rate and working in an English-speaking country did not affect wages, but the log of the GDP had a positive and significant effect on wages for both men and women, implying that holding everything else constant (and keeping in mind that wages are expressed in PPP prices), individuals from wealthier countries tend to earn more than those in poorer countries. Average schooling had a negative and significant effect on men s and women s wages due to the fact that most of the individuals in the sample had a higher level of education than the average individual in their country of employment. Lastly, column (4), adds a dummy variable for non-economic-visa migrants. Among men, there was no distinction in earnings between economicvisa and non-economic-visa immigrants. However, among women, those who migrated for family reunification purposes were earning less in their country of origin than those who migrated to the US for economic reasons. Economic-visa migrant women, therefore, show more motivation abroad than non-economicvisa migrant women, while this difference is not obvious in the men sample. Economic-visa migrant women who work abroad may be a more selective sample than men working women are more likely to have had a non-economic visa and there are only half as many women with economic visas as men. Table A7 in the Appendix shows the country effects for the male sample, where the excluded country is Canada. Only immigrant men from Ethiopia

69 52 earned more abroad than Canadian immigrant men. Men from all the other countries earned either less than or comparable wages to the Canadians. Immigrant men from China and Mexico earned the least when abroad, compared to Canadians. Table A8 of the Appendix, shows the country dummy variables for the female sample. After controlling for country level characteristics and visa category (column 4), we can observe that none of the immigrant women earned more than the women from Canada. Immigrant women from Mexico, Haiti and Ukraine earned the least compared to Canadian immigrant women. That is, compared to immigrants from Canada, wome from Mexico, Haiti and the Ukraine are either negatively or less-positively self-selected among all immigrants. D. Full Information Maximum Likelihood Heckman Selection Methodology The analysis of the wages of immigrants before immigration may suffer from selection bias because not all immigrants in the survey were employed. Therefore the Mincerian equation discussed above is potentially biased by evaluating the rate of return to education and other determinants based only on those who participate in the labor market. This is a more important issue for women, who may have a greater choice as to whether they would work or stay at home, based on their marital status and number of children.

70 53 The following analysis uses the (full information) maximum likelihood (FIML) Heckman selection model, which is more efficient and consistent than the two-step Heckit. Consider again the wage equation (3) from the previous section. If there exists a choice of participating in the labor market, then it would suffer from selfselection bias: where (3) Ln( Wic ) 0 1X ic 2Zc ic LnW ic is the log of wages for individual i in country c; X ic is a set of individual characteristics such as gender and education for individual i in country c, Z c is a set of country level characteristics such as log of GDP and average level of education for country c, and ic is the individuals error term. The Heckman method requires that there should be at least one independent variable that affects the selection equation (participation in the labor market), but has no effect for the outcome equation (the wage equation) for identification. In addition to the traditional exclusionary variables used for wages (number of children and number of children under the age of six), the equation includes age of emigration, spouse employment status and wages in the selection equation. If the covariance of the errors in the selection and outcome equations equals zero, then one can use the Mincerian equation to estimate the determinants of the immigrants wages abroad. This statistic is called rho (ρ) in the results tables.

71 54 1. Maximum Likelihood Heckman Selection Model Wages of immigrants in their country of origin can be modeled by two latent dependent variables models: * (4) Y1 Z 1 * (5) Y2 X 2 Equation (4) is the selection equation and equation (5) is the outcome equation. X and Z are sets of individual characteristics, not necessarily exclusive of each other, but with at least one explanatory variable in equation (4) that does not affect equation (5). The errors are assumed to be jointly normally distributed 1 0 ~ n, 2 0 where 1 2 and (1,1 ) Since the outcome of the participation equation (4) is only observed in binary form, the variance of 1is not observed, however we are only interested in the sign of * Y 1, so the variance is therefore arbitrarily set to 1. We only observe * Y if Y 0, which indicates market labor participation, * 2 1 therefore the observed equation becomes (6) * Y Y 2 if * Y 1 0 and * Y missing if Y 0, 1 and the probability of participating in the labor market is pr( z 1) ( ' i w i ). Therefore the expected value of y (individuals wages), conditional on participation ( z 1) and on individual characteristics x ) is ( i

72 55 (7) ' E y z 1, x ) x E z 1) ( i i i ( 1 i i ' xi E( 1 i 2i wi ) ' ' ( wi ) xi since ' ( w ) i ' ' ( wi ) E( 1 i 2i wi ). ' ( w ) i ' ( wi ) is the Inverse Mills ratio and it is always positive, therefore the sign of ' ( w ) i the bias depends on the covariance, where is the correlation between 1and 2, and is the variance of 1. If 0, this would indicate that equations (4) and (5) are independent of each other and that there is no selection bias. 2. Estimating the Full Information Maximum Likelihood function Let D=1 if Y is not missing, that is if the immigrant is working, and D=0 otherwise. An individual can have 2 possible outcomes, pr ( D 0) or pr D 1, y ) 10, where ( 2i ' (8a) pr ( D 0) pr( z ) ( z i 1i i ' 1 ( ' ) and ' (8b) pr D 1, y ) f, ) ( 2i z i ) ( 1i z i 2i f ' z i 1 (, 1 i 2i ) d 1i 10 Notation adopted from Bierens (2007).

73 56 1 ' ) ( ) ( z i i i i i d f f 1 ' ) ( ) ( z i i i i i d f f and using ' 2 2 i i i x y 2 ' 2 ' ' ) )( / ( 2 ) ( exp 2 1 i i i i i x y z x y 11 The maximum likelihood function is estimated by multiplying both probabilities, equations 6a and 6b: (9) ), ( L ) 1, ( 0)* ( i i i y y pr y pr n i D i i i i i D i x y z x y z 1 2 ' 2 ' 2 2 ' ' 1 ) )( / ( 2 1 exp 2 1 * )) ( (1 Since the logarithm of the likelihood function is easier to maximize and the maximum of the function is the same as the maximum of the logarithm of the function, it is therefore more common to estimate the logarithm of the likelihood function. 11 ) ( 2i f ' ) ( exp 2 1 i x i y 1 ' ) ( z i i i i d f ) ) (( ' 2 1 i i i z pr 2 ' 2 2 ' 2 ' ) )( / ( 1 ) )( / ( ) ( i i i i i i i x y z x y pr 2 ' 2 ' 1 ) )( / ( i i i x y z

74 57 To simplify the equation and following Green s notation, 12 let q (10) i ' zi 1/ / 2 / 1 then 1 1 ln atanh atanh exp(2 ) 1 ( ) exp(2 ) 1 (11) log L log ( qi ) d 0 d 1 log (1/ 2)( y i ' x ) i 2 log [ ( y i ' x ) q i i 2 1 ] which is the equation of interest (the FIML Heckman Selection equation) and the equation to be estimated. For technical reasons 13, Stata, the statistical tool, estimates atanh 14 and not, but the null hypothesis H0: ρ=0 is equivalent to H0: atanh ρ=0. 3. Full Sample Results The results of the Heckman analysis for the full sample are presented in Table 9. Unconditional on country of employment, women abroad earned 61% less than men. This was just slightly more than without the bias correction. The 12 Green s Lecture Notes 13 It provides more numerical stability during the maximization to the function. 14 atanh ( 1/ 1 2)log 1

75 58 rate of return to education was 9.3% and experience increased at a decreasing rate, as expected. As shown in the OLS regressions, marital status did not have an effect on wages. Weeks per year worked, or effort was not statistically significant as it was in the OLS regressions. Speaking English, regardless of country of employment, was positive and statistically significant and had a larger effect when corrected for selection bias. The selection variables age of emigration, number of children and spouse employment status were statistically significant. However, athrho, which is the inverse hyperbolic tangent of ρ, was not statistically significant. Recall that ρ measures the covariance between the error terms of the selection and outcome equations. This means that there is no selection bias in the employment of immigrants while in their country of origin and that the OLS estimates are unbiased as well. To control for differences in countries of employment, column (2) presents the results that included country fixed effects. Doing so, reduces the rate of return to education and the return on experience, but the men and women s wage differential remains almost unchanged. The difference in earnings due to the ability to speak English was larger in the Heckman equation (31%) than in the OLS regression (24%). In column (3), I controlled for country-level characteristics. The interpretation for level of schooling being negative is the same as for OLS. Holding years of education constant, those who lived in a country with higher average years of schooling earned less than those who lived in countries with lower average years of schooling. Of the selection variables excluded from the wage equation only spouse employment remained statistically significant. Rho, the covariance between the error terms, was not different from

76 59 zero, therefore there was no selection bias in terms of employment for the whole sample of immigrants. Using visa category to classify the immigrants initial motivation for moving as either economic and non-economic immigrants, column (4) controls for visa type as a proxy for immigrants drive. 15 Non-economic immigrants earned less than economic immigrants, as shown in the OLS regressions, although the estimated effect with the Heckman methodology is larger in absolute value. Adding this variable did not change previous results significantly either. Except for spouse employment, all selection variables remained statistically insignificant, as did ρ. The conclusion drawn from these results is that there is no selectivity bias and that the OLS results should be very similar to the Heckman FIML. Although the results were not identical, they were very similar, especially the estimation of the return to education. The results from the probit regressions and from the selection section of the Heckman regressions are not expected to be similar because unlike, the two step heckit, the FIML estimates the selection and the outcome equations simultaneously. Among the identified countries of employment, immigrants from China, Cuba, Dominican Republic, Nigeria, United Kingdom and Vietnam do not earn significantly different wages than Canadian immigrants; immigrants from Ethiopia earned more than immigrants from Canada in their countries of origin; and immigrants from the rest of the identified countries earned significantly less 15 Driven individuals outcomes could differ from those who are not both in the country of origin and in the United States.

77 60 than immigrants from Canada. These results are presented in the appendix, in Table A9. 4. Results by Gender Tables 10 and 11 present the estimated coefficients for the Heckman FIML separately for men and women to see if there was selectivity in at least one of the groups. In Table 10, the results for the male sample are shown. In column (1), which did not control for country fixed effects, all selection variables except for age of emigration and spouse s wages were statistically significant. Athrho (atanh (ρ)) was statistically significant, which indicates selectivity among immigrant men. The return to education was 7.6% and wages increased with experience at a decreasing rate, as expected. Single men earned more than married men abroad. Numbers of weeks worked per year, which is a measure for effort was positive and statistically significant in the OLS results. However, after controlling for selectivity, this variable was no longer significant. Speaking English seemed to have a larger and more positive effect on wages among men than the OLS results previously showed. Column (2), adds country fixed effects to the specification. This affected the results similarly to the OLS results. The return to education is lower, and it is still lower than the one estimated by OLS, but not by much. The effect of marital status remained unchanged and the difference between men who spoke English and those who did not disappeared once the country fixed effects were added to

78 61 the specifications implying that men in countries where English is spoken earned more than those who worked in countries where English is not. Column (3) presents results controlled for country-level characteristics. The effect of the country s wealth (log of GDP) was the same as in the OLS specification, but the average schooling effect was larger in magnitude in the Heckman equation than in the OLS equation. Unemployment remained insignificant. Both dummies speaking English and English-speaking country were not statistically significant. All selection variables, except for spouse s wages, were statistically significant. Although age of emigration was expected to have a positive effect, among the men sample, and adjusted for bias, this variable was negative and statistically significant. However, the size of the effect was very small (-0.002). In column (3), ρ was no longer statistically significant. Adding immigrant type to the equation (column 4) did not change the basic conclusion that there is no selectivity bias for men working abroad. In Table A10, in the appendix, I the estimated country fixed effects for immigrant men are presented. Holding everything else constant, Ethiopian men who eventually migrated to the US earned more on average than Canadian men who immigrate to the US. Men from China, Jamaica, India, Mexico, and Peru used to earn less in their country of origin than Canadian men did in Canada. Men who worked in one of the rest of the identified countries did earn similarly to Canadian men. The results for female immigrants are presented in Table 11. Although only the selection variables age of emigration and spouse s employment are statistically significant, there is no self-selection bias among immigrant women

79 62 working abroad. The results were very similar to the OLS results in Table 8, which was to be expected when ρ is not statistically significant. The rate of return to education abroad was 12.2%. Experience was positive and statistically significant and it increased at a decreasing rate. Single women earned less than married women in their country of origin. The number of weeks per year had no effect on wages. Speaking English, without controlling for country of employment, increased their weekly wages significantly, but part of this estimate is possibly due to the differences in pay between English-speaking and non-english speaking countries, since this variable went from to when I included country fixed-effects in the second column. The log of GDP and average schooling was statistically significant as well (column 3). Following the same pattern as before, in column (4) I included visa type to the specification. Just as shown in the OLS estimates, this variable was negative and statistically significant for women, that is, economic-visa migrant women earned more, and therefore seem more motivated, than non-economicvisa migrant women abroad. Since economic-visa migrant women were more likely to be single than non-economic women, and based on their visa they were not migrating to join a spouse in the US, these women faced higher costs of migration than men, and therefore showed more selectivity than men. This is consistent with the human capital theory that higher costs of immigration lead to more positive selection. The results for the country dummies for women s wages abroad are presented in Table A11. The results, as expected from the lack of selection bias were similar to the OLS estimates.

80 63 III. Conclusions This chapter has analyzed the labor market participation and wages of US immigrants in their country of origin and the selectivity among immigrants. The empirical results presented in this chapter confirm that men and women, as well as economic and non-economic-visa migrants have different experiences in the labor market abroad and therefore different incentives to migrate. Immigrants have low levels of labor participation abroad, but high levels of education. Although women earn significantly less than men (57% less), women are only 7% less likely to work. Their return to education is higher than that of men and the effect of education on employment is also higher for women than for men. In fact, men s education effect on employment disappeared when country-level characteristics were introduced in the specification. On the other hand, the effect of experience on employment and wages is higher for men than for women. This result should be expected if women take time off to raise a family sometime in their careers, while men usually do not. Another expected result is the positive effect that spouse s employment has in one s employment, for both men and women. Number of children, usually a determinant of employment for women, has no effect on women s employment abroad, but it has a positive significant effect for men. Men are less likely to work with the number of children under the age of six, holding everything else constant. Immigrants who have more desirable characteristics arguably those that come from English-speaking countries, because of the knowledge of the language, and those that come from countries with higher GDP and average schooling are

81 64 less likely to work. I offered the explanation that employed individuals in high GDP countries had higher costs of migration and therefore the unemployed were more likely to immigrate. Another possibility had to do with the definition of employment itself. Employed immigrants, as I have defined it, are those immigrants that have had a paying job or had some income before leaving their country of origin. Some individuals that fall under the category of unemployed, decided to not participate in the labor market, such as homemakers or some disabled individuals. As a result, there appear to be a higher rate of unemployment among immigrants that the average rate of unemployment in their country of origin. Additionally, less developed countries (lower GDP) are more likely to have an informal economic sector allowing more individuals to be employed (and report wages) in these countries than in more developed economies. Meanwhile, among those that work, immigrants earn more in countries with higher GDP; and women who speak English earn significantly more than women who do not. It was also concluded that non-economic-visa migrant men are marginally more likely to be employed than economic-visa migrant men, even though one can assume that economic-visa migrants are more motivated individuals than non-economic-visa migrants, ceteris paribus. At the same time, among employed men, wages do not differ by immigration type. For women, the reverse is true. Although economic-visa migrant women are just as likely to be employed as noneconomic-visa migrant women, among the employed females, non-economic-visa migrants earn less than economic-visa migrants. A motivated individual is more likely to look for employment opportunities and migrate if he finds himself

82 65 unemployed, and therefore we are more likely to observe unemployed economicvisa migrants than non-economic-visa migrants; or, as in the case among women, motivation is a trait, although unobserved by the researcher, that is compensated positively in the labor market and therefore, holding everything else constant, economic-visa migrant women are compensated better in the labor market than non-economic-visa migrant women. This chapter also explored the possibility of immigrant selection in the labor market. Using the full information maximum likelihood Heckman selection method, the self-selectivity of immigrants into employment was examined for the full sample, as well as separately for men and women. In addition to the traditional variables used for selection criteria such as number of children, number of children under the age of six and spouse s employment status, age when the immigrant left the country of origin was added, since there is a direct relationship between age at the time of departure and employment status. Although women usually have a choice to either work for pay in the labor market or work in the household, it appears from this analysis that this recent group of immigrants, men and women, do not show any differences as regards this issue of selectivity. Therefore, one can conclude that the OLS estimates discussed here are unlikely to suffer from selectivity bias. Holding all available individual characteristics constant, this chapter also showed that selectivity among immigrants does vary by country level characteristics, which remained, for the most part, statistically significant. Since there is no information on those who remain in their country of origin in this data set, I can only conclude about the selectivity of immigrants, relative to each other,

83 66 and not whether they are positively or negatively selected from their country s general population. This chapter has also shown that there are important characteristics that affect the labor market outcomes of immigrants abroad, and this may consequently decide which immigrants select themselves to leave their country of origin and choose to come to the US, and affect their labor market outcomes in the US. Previous research has been ignoring these characteristics and therefore possibly producing biased results. Later, chapter 4 will introduce these characteristics in the analysis of the labor market outcomes of immigrants in the US to correct for this deficiency.

84 67 CHAPTER 3. LABOR MARKET OUTCOMES OF IMMIGRANTS AND NON-IMMIGRANTS FROM MEXICO A OAXACA DECOMPOSITION I. Introduction On average, immigrants who have recently become legal permanent residents have a level of education that is at least 50% higher than the average population in their country of origin 16. But how else do they differ from the general population in their country of origin? Do these differences persist after one controls for other observable characteristics such as gender, experience or marital status? And how do these differences affect their labor market experiences in their country of origin? We know from the previous chapter that 90% of the sample has at least 6 years of education and more than half have at least 12 years of education, and at the same time, education is not a major determinant of employment. In fact, the influence of education on employment is zero for men, and for women it is smaller than that of spouse s employment status or their marital status. Employment rates for these immigrants range from 15% for immigrants from El Salvador, to 64% for immigrants from Canada and Russia. For those immigrants that were employed before immigrating to the US, their return to education is positive and around 5 to 7 percent. The previous chapter also estimated that immigrant women abroad earn 59% less on average than immigrant men, holding everything else constant. To 16 Based on NIS and Barro-Lee data sets. This is the case for at least those countries where data on average schooling is available. See Table A1

85 68 put this in perspective, women in the US earn about 24% less than men, on average. 17 Based on the Global Gender Gap Report (2008) created by World Economic Forum, only about a third of the countries represented in the NIS data experience greater gender wage inequality. Figure 3 presents the Global Gender Gap Report data in graphic form for those countries represented in the NIS data set. The upper limit of each bar represents the inequality index on employment participation while the lower limit represents the inequality index on wages. The inequality index reflects the ratio of women to men, therefore a score of 1 implies total equality. This chapter will focus on only one country, Mexico, to answer the questions put forward in the introductory paragraph. Choosing Mexico as the country for this analysis has its advantages. First, Mexico is the largest net exporter of labor to the US. In 2000, Mexicans accounted for about 30% of all immigrants in the US (Congressional Budget Office, 2004). Forty-four percent of Mexican immigrants are legal immigrants (Passel, 2005a; Passel & Cohn, 2008). The proportion of Mexicans becoming legal permanent residents reached a third of all permanent residents in the 1990s (see Figure 4). Second, Mexicans are also the largest group represented in the New Immigrant Survey. Although the majority of Mexicans that have arrived in the US in the past few decades have been undocumented (Passel, 2005b), there is still a large proportion of Mexicans that have immigrated through the proper channels. The NIS data shows us that Mexican immigrants who became permanent 17 See, for example, Blau and Kahn (2000), Cooke et al. (2009)

86 69 residents in 2003 are more likely to be married to a US citizen than to have been undocumented or had an employment preference. Third, given that Mexicans are the largest group of immigrants in the US and that the US is the main destination for Mexican immigrants, there already exists a vast amount of research in the area of immigrants labor market outcomes, assimilation and self-selectivity focusing on Mexican immigrants. 18 It is therefore hoped that the results presented in this chapter would then add to this substantial literature. Previous research comparing Mexican immigrants and non-migrants have used census data from Mexico or the US, but neither of these sources specifically identifies the legal status in the US of the individual (Chiquiar & Hanson, 2005; Fernandez-Huerta Moraga, 2011); Chiquiar and Hanson (2005) found positive selection using the US census 19 while Fernandez-Huerta Moraga (2011) found negative selection using ENET (similar to the Current Population Survey of the United States). Others have used the Mexican Migration Project (MMP) 20 and have found positive selection, however their analysis was limited to undocumented migrants (McKenzie & Rapoport, 2007; Orrenius & Zavodny, 2005). 18 See for example, Kossoudji and Ranney (1984), Massey (1987), Donato and Massey (1993), Borjas (1996), Hanson and Spilimbergo (1999), Rivera-Batiz (1999), Kochhar (2005), McKenzie and Rapoport (2007), Orrenius and Zavodny (2005), Hanson and Chiquiar (2005), and Fernandez-Huerta Moraga (2011). 19 U. S. Census tends to undercount immigrants, particularly undocumented immigrants (Hanson, 2006). 20 MMP, a survey limited to about 50 communities in Mexico that has a large proportion of their population migrating to the United States, is not representative of the Mexican population. It only covers rural communities in a number of states in Mexico.

87 70 The research in this chapter of the dissertation attempts to answer the following questions: How immigrants from Mexico differ from Mexicans who do not immigrate in terms of labor market participation? How do they differ in terms of their rate of return to education? Are their wages determined differently? In addition, I will also address the question of whether their experiences and performances in the labor market differ due to observable or unobservable characteristics. The rest of this chapter is organized as follows: in the next section, the data sets used in this analysis are discussed and the summary statistics of the data sets presented. Section 3 reports the results for the first three questions addressed in this chapter (labor market participation, return to education, wage determinants). Section 4 describes the methodology that is used to identify the sources of differences between immigrants and non-migrants from Mexico, the Oaxaca decomposition. Section 5 discusses the results of the Oaxaca decomposition analysis, and section 6 concludes the chapter. II. Data Description A. Mexico s Household Income and Expenditure National Survey The data for this analysis came from two sources. The first is NIS data for immigrants from Mexico and the Encuesta Nacional de Ingresos y Gasto de los

88 71 Hogares (ENIGH) 21 or Household Income and Expenditure National Survey, for non-migrant Mexicans. NIS data has been described in the previous section. ENIGH is a survey that is conducted in Mexico every two, three, four or five years and since 1956 has been conducted under different names and by different public entities. Currently and since 1983, the ENIGH has been conducted by the Instituto Nacional de Estadística y Geografía (INEGI) 22, an independent and autonomous institution ( Since then, the survey s methodology has been made more uniform and the objectives have been expanded. The purpose of ENIGH is to obtain a statistical picture of income and expenses of Mexican families and therefore be able to shape public policies and make economic and political decisions. In addition to income and expenses, the survey provides information on occupation, socio-demographic characteristics, housing characteristics and other possessions. The non-migrant data consist of 19,267 observations extracted from the INEGI survey data from 2000 and In 98% of the cases, the head of household was a male, if the couple is married. In order to have a more representative distribution of men and women, only one member of each household was randomly selected, which resulted in a data set of 19,267 observations, out of 149,433 original observations. In the resulting data set, 53% of the sample were female and 65% were employed. After randomly selecting one observation per household (there were a total of 27,275 households), only those individuals who were 15 years of age or older were kept for the analysis. There is 21 Source: INEGI website: 22 The institute used to be named Instituto Nacional de Estadística, Geografía e Informática. INEGI changed its name in 1983 but kept the same acronym.

89 72 no way to know from this data set if any of these individuals planned to migrate in the future or if they have any type of relationship with someone in the US that would facilitate migration for them legally. The INEGI sample asks about the income earned in each of the past 6 months. The income information was from the last month. If last month s income is missing, the income from the other months that were available is used instead. The sample was taken from two different years, 2000 and 2002, so in order to make them comparable to each other and to the NIS data, I adjusted for inflation and adjusted all wages to 2010 prices using data INPC series published by the Mexican central bank. 23 The New Immigrant Survey data set includes 1,487 observations from immigrants who were originally from Mexico. Out of these 1,487 Mexicans, 177 were employed between 1995 and 2004 and another 512 left Mexico after 1995, but were unemployed. The income reported from working in Mexico was collected in Mexican currency and then transformed in PPP prices. I used the information in Mexican pesos and used the INPC series data to adjust all wages to 2010 prices. B. Summary Statistics Table 12 presents the summary statistics for the INEGI and NIS data on Mexicans, divided by employment status. Out of the 19,267 non-migrant individuals, 65% were employed and they earned an average of $1, pesos 23 Banxico is the Mexican cental bank ( Indice de Precios al Consumidor (INPC) series (the Mexican CPI) is published monthly. When the month information was available in the ENIGH data, the corresponding month was employed, otherwise, the average over the whole year was used.

90 73 per week 2425, while only 24% of Mexican immigrants were employed before leaving Mexico. For those immigrants that were employed, they earned almost 25% more than non-migrant Mexicans. There is a larger proportion of women in the immigrant sample (65%) than in the non-migrant sample (53%). Another major difference between these two groups is the level of education. Nonmigrants had an average of 6 years of schooling, compared to 8 years among immigrant individuals. The estimated experience for non-migrants was larger than for immigrants - 29 and 21 years, respectively. While non-migrants were earning less than migrants, they worked 45 hours per week compared to 43 hours worked by migrants. Immigrant Mexican individuals were more likely to be married or single than non-migrants. They also had more than twice the total number of children and children under the age of six, on average. In Table 13, I present the summary statistics for these characteristics separately for men and women. As shown, non-migrant men are the group most likely to be employed (84%) while immigrant women were the group least likely to have been employed in Mexico (20.20%), and non-migrant women were more likely to be employed than immigrant men. Although on average immigrants were less likely to be employed among men and women, those who work earn more than non-migrants. Immigrant men earned 36% more ($2,359 Mexican pesos per week) than non-migrant men ($1,734 Mexican pesos per week). Immigrant women earned 44% more ($1,594 Mexican pesos per week) than non- 24 I dropped all the observations where the year of employment or the year they left Mexico was before 1995 to better match the labor market environment in Mexico between immigrant and non-immigrants. 25 In 2010 prices.

91 74 migrant men ($1,110 Mexican pesos per week). This, however, could be due to the difference in their observable characteristics, since immigrant men and women have higher levels of education than non-migrants and the difference in education between employed and unemployed immigrants is more pronounced than between employed and unemployed non-migrants. For example, the difference in education for non-migrant employed and unemployed men is less than half a year and it is negative, while the difference between employed and unemployed immigrants is almost 3 years of education. On average, immigrant men have 9 years of education compared to only 6 yeas among non-migrant Mexican men. Non-migrant and immigrant women have 6 and 7 years of education, respectively. Immigrant women who are employed have almost 11 years of education compared to only 6.5 years of education among employed nonmigrants. Non-migrant men worked more than 2 hours longer per week than immigrant men, while immigrant women worked less than an hour longer than non-migrant women. Immigrant women were most likely married, while immigrant men were likely single. Non-migrant women were least likely to be single, while immigrant men were least likely to be married. Immigrant women had the highest number of children in total, but the lowest number of children under the age of six. Non-migrant women had the least number of children and non-migrant men had the highest number of children under the age of six. However, the average number of children is never more than 2 and the average number of children under the age of six is never more than one, for all groups.

92 75 III. Immigrant and Non-migrant Labor Market Experience There are two labor market outcomes that I was interested in investigating. The first one was participation. Specifically, what characteristics matter for immigrant and non-migrants in the determination of labor market participation. If immigrants and non-migrants vary in unobservable characteristics, one can expect the determinants of employment to vary between these two groups. The methodology employed for this analysis was the same as that used in the employment market participation analysis done in the previous chapter for all immigrants in their country of origin, using a probit estimation. The second outcome was performance in terms of wages. In the previous chapter, I established that there was no statistical bias in the estimation of wages, and therefore I chose to present the results from the traditional (Mincer) wage equation. A. Labor Market Participation Results for the probit estimation are presented in Table 14. Columns 1 and 2 represent the coefficients and marginal effects, respectively, of the probit equation for non-migrants (INEGI data). Columns 3 and 4 represent the estimated coefficients and marginal effects for the immigrant set (NIS data). For the full sample, the signs of the estimated coefficients and marginal effects for immigrants and non-migrants are the same, except for the number of children and divorced dummy variable, which were not statistically significant.

93 76 The female dummy was negative for both, as expected, but not statistically significant for immigrants. The effect of education on the likelihood of employment was similar, positive but small, for both samples. The effect of experience, however, was larger for the non-migrant sample than for the immigrant sample. Although immigrant women were not less likely to be employed than men (the effect is small and negative, but not statistically significant), holding everything else constant, non-migrant women were 35% less likely to be employed than men. Marital status seemed to have a bigger effect for non-migrants than from immigrants. Only widowed immigrants were more likely to be employed than married immigrants, while unmarried non-migrants were more likely to be employed than married non-migrants. Holding marital status and total number of children constant, non-migrants were more likely to be employed and have children under the age of six, while young children does not have an effect among immigrants. There is a clear difference between men and women s labor market participation rates and their characteristics; therefore, in Tables 15 and 16, I presented the results for men and women separately. The first significant difference was on the effect of education on employment. The increase in the probability of employment with each additional year of education was faster for immigrant men and non-migrant women. The size of the effect of education on the probability of employment for non-migrant men may be small (but positive) due to the fact that employed non-migrant men have on average less education than their unemployed counterparts.

94 77 Experience increased the probability of employment faster for nonmigrant women and immigrant men than for the other two groups, even though on average the experience between the unemployed and employed for these two groups was very similar. Single non-migrant men earned less than their married counterparts. Among women, all marital status dummies showed positive coefficients and marginal effects meaning that married women, immigrant and non-migrant, were less likely to be employed than all other women. Holding marital status constant, non-migrant men were more likely to be employed with the number of young children, but less likely with the total number of children. Number of children made no difference to the probability of employment for immigrant men. Among women, I observed the opposite. Number of children and number of young children had no effect on the probability of employment for non-migrant women, while the probability of employment for immigrant women increased with the total number of children, but decreased with the number of children under the age of six. B. Rates of Return to Education and Wage Determinants The results of the wage equation for the full sample are presented in Table 17 by immigration status. As shown, immigrant and non-migrant women earned 37% and 39% less than men in Mexico, respectively. The return to education was only slightly higher for non-migrants (11.8%) than for immigrants (11.1%). However, the return on experience for non-migrants was more than twice the return for immigrants. There could be differences in how immigrants and non-

95 78 migrants accumulate experience in Mexico that may be observable to the employer, but not to the researcher that may explain why non-migrants are compensated more adequately for their experience than immigrants. Singles and those living together non-migrants earned less than their married counterparts; while separated, divorced and widowed non-migrants earned more than married non-migrants. Among immigrants, those who were divorced earned more than married immigrants. A measure of work intensity, hours per week, had no effect on immigrants wages, but it had a positive and significant effect on non-migrants. It was possible that for employers there was a visible difference in the effort that immigrants and non-migrants put into their work and therefore they rewarded experience and those who worked more efficiently differently for immigrants and non-migrants. Next, I attempt to provide a more in depth look at the wage determinants for immigrants and non-migrants by estimating separate equations for men and women. The results are presented in Tables 18 and 19, respectively. Although women usually had higher returns to education than men, this was only true for non-migrants. Immigrant men and women had the same return to education (11%). However, among men, non-migrants had lower returns to education then immigrants, while the opposite was true for women. Non-migrant women seemed to find jobs that rewarded their education more successfully than immigrant women, possibly leading the latter to select themselves into migration. Because the opposite was true for men, men who do not move did not seem to face the same job opportunity experiences as immigrant men did before they emigrated.

96 79 The differences in the return to experience persisted when I separated the sample by gender. Non-migrant men seemed to be rewarded more highly for their experience than non-migrant women. For immigrants, the story was different: experience had no effect on wages. There was a negative relationship between experiences and wages among immigrant men, but it was not statistically significant. For women, the relationship was positive, but was not statistically significant either. Non-migrants who were not married, but living together or were separated or single earned less than married non-migrants. Immigrant divorced men earned more than married immigrant men. Hours per week had a positive effect on the log of wages of non-migrants, but no effect on log of wages of immigrants. Non-migrant women who were separated, divorced, widowed or single earned more than non-migrant married women. Marital status did not affect the log of wages among immigrant women. Similar to men, hours per week had a positive effect on wages for non-migrant women, but no effect for immigrant women. IV. Blinder-Oaxaca Decomposition Methodology From the previous section, it is evident that immigrants and non-migrants experiences are different and that their observable characteristics are varied as well. Could the difference in labor market participation and wages be explained solely by the difference in their observable characteristics? In order to answer this question, I will introduce to the discussion the Oaxaca decomposition in this section and apply it to the results above in the following section.

97 80 The Blinder-Oaxaca Decomposition ( Oaxaca decomposition ) allows the researcher to identify factors that contribute to the differences between the outcomes, such as the wages, of two groups. The Oaxaca decomposition decomposes the difference in wages between these two groups into differences in observable and unobservable characteristics. The differences in observable characteristics are labeled as differences in Endowments and the differences in unobservable characteristics are labeled as differences in the estimated Coefficients (Blinder, 1973; Oaxaca, 1973). The Oaxaca decomposition was initially developed to measure discrimination (unobservable differences) in wages against women and racial minority groups, but the methodology has largely been used in other areas where unobservable characteristics are suspected to explain the difference between two groups or two periods of time (Gang et al., 2002; O Donnell, et al., 2008, Rivera-Batiz, 1999; among others). Let group m be Mexican individuals that will become immigrants in the US, and let group n consist of Mexicans who do not migrate. If immigrants and non-migrants had the same unobservable and/or observable characteristics, then the previous results would have been identical for both groups. However, if these two groups are different, then one could decompose these differences into endowments and coefficients. Consider the following equations: (1) (2) where is log of weekly wages in Mexico; is a set of individual characteristics; and and are the individual errors for immigrants and non-migrants,

98 81 respectively. If all characteristics are observed and are compensated equally, then = and OLS would be sufficient for this analysis. However, when individuals self-select themselves into one group or the other, as in this case the decision to immigrate or to stay in Mexico, it is because, presumably, these individuals are different from each other. These differences may be due to observable or unobservable characteristics. The difference between wages earned by non-migrants and immigrants can be expressed as follows: (3) where and are the mean of each characteristic and equals zero since and. The Oaxaca decomposition also allows us to estimate how much of each difference is due to the differences in the s and differences in the s. Let be the difference between the wages, then (4) (5) (6) The magnitude and interpretation of the decomposition differs depending on which term is added in equation (5), because these terms become the weights used for the interpretation of the results. This two-fold decomposition can be augmented in a three-fold decomposition that allows for an interaction term between the difference in endowments (x) and difference in coefficients (β)

99 82 (Daymont & Andrisani, 1984; Jones & Kelley, 1984; Winsborough & Dickinson, 1971). The interaction term accounts for the fact that differences in coefficients and endowments occur at the same time and cannot be completely separated. To conduct the decomposition from the view point of the immigrant group (how would the wages look like if immigrants had the same characteristics as nonmigrants), I added and subtracted,, and to equation (4), which yields (7) where the first term is the difference explained by endowments, the second term is the unexplained difference in wages (the difference on the return to or prices of the individual characteristics of each group) and the third term is the interaction of these differences. A positive interaction term indicates that non-migrants are paid higher for a characteristic if the mean of the said characteristic is higher for non-migrants than for immigrants. Yun (2000) showed the same procedure can be modified and used for nonlinear specifications, such as in probit and logit regressions. Suppose that we have a nonlinear function (8) for all G=n, m

100 83 where Y is the binary dependent variable and X is a set of independent variables. 26 Then the mean difference of the outcome variable between immigrants and non-migrants can be decomposed as follows: (9) [ ] Yun (2000) suggests that we use two approximations to identify the contribution of each characteristic and coefficient to the difference in the outcome between these two groups, and evaluate the value of the function using the means of the characteristics and the first order Taylor s expansion to linearize the characteristics and coefficients effects around and, respectively. These steps lead to the final equation 27 : (10) [ ] where 26 Following Yun (2000), we are also assuming that F is the mapping of a linear combination of X to Y and F(.) is a once-differentiable function. 27 Yun (2000) shows a step-by-step how to get from equation (9) to equation (10)

101 84 V. Decomposition of Observables and Unobservables A. Labor Market Participation Categorical variables in the Oaxaca decomposition suffer from an identification problem because the estimated coefficients do not vary with the choice of the reference group. Yun (2005) suggests normalizing the categorical variable so that the constant and all the coefficients of the dummy variables can be identified. I applied this suggestion to the year, race and marital status variables. The results showed there was a positive difference between the likelihood of employment of the non-migrant and immigrant Mexicans, 65% and 26% respectively. Based on the Oaxaca decomposition, 13% of the difference in the outcome was due to endowments or observable characteristics. The main difference in employment was due to unobservable characteristics. The difference in coefficients accounted for 67% of the difference in employment participation. Twenty percent was due to the interaction between the coefficients and the endowments. Whenever non-migrants had larger observable characteristics than immigrants, it increased their probability of employment by more than it does for immigrants. The major contributions to the endowment effect were education and experience. The difference in education was negative since the average education of immigrants was larger than non-migrants average education. The difference in experience was positive, which implies that the difference in employment was

102 85 also due to the positive difference in experience between non-migrants and immigrants. The main contributors to the differences in employment due to these coefficients were the female and single dummy variables. The difference due to gender (negative, as expected) was offset by the single dummy and the other controls, which were mostly positive. The interaction of the differences between endowments and coefficients accounted for about 20% of the differences in probabilities of employment, which was also part of the unexplained difference. All three parts of the decomposition were statistically significant. For men, the largest contribution to the difference in endowments was experience. Non-migrant workers had more experience and were therefore more likely to be employed. The large positive difference in experience offset the negative difference in education endowments for men, another significant contribution to the difference in employment. Coefficients accounted for 72% of the difference in employment for men. Although the proportion of the difference was statistically significant, none of the elements of the decomposition were. The negative sign of the difference due to the interaction between endowments and coefficients means that the probability of employment for men increased more slowly for non-migrants than for immigrants when non-migrants had better observable characteristics. The difference, however, was small (3%) and not statistically significant. These results are in Table 23. In Table 24, I present the results for women. The differences in employment shown among the women can be attributed to both, endowments and coefficients. Since the interaction term was negative, one can conclude that non-migrant women are more likely to be employed when their mean

103 86 characteristics are lower than migrant women s. Sixty percent of the difference in employment can be attributed to differences in endowments, while 80% (-40% from the interaction) can be attributed to differences in the coefficients. All three parts of the decomposition were statistically significant. In the endowments portion, almost all controls, with the exception of experience, were statistically significant. The largest contributor was marital status. The difference in education was negative because immigrant women had more education than non-migrant women. Being single was the main contributor to the difference in the coefficients. Since this was positive, the probability of single non-migrant women being employed compared to married non-migrant women was higher than the probability between single and married immigrant women. B. Wages in Mexico In the following three tables I present the results for the decomposition of non-migrants and immigrants wages for the full sample and for men and women separately. Table 22 presents the results for the full sample. The differences between the log of wages of immigrants and of non-migrants was negative and statistically significant, however, the differences attributed to endowments, coefficients and their interaction did not appear to be statistically significant. Nevertheless, some individual characteristics were statistically significant for each group. Endowment differences account for about 25% of the difference in wages. Among the observable characteristics, education accounts for 56% of the

104 87 difference 28. On average, immigrants had higher levels of education than nonmigrants; therefore the difference due to endowments was negative. There were larger proportions of immigrant women working than non-migrant women, therefore the contribution to endowment difference was positive, since the coefficient was also negative. The same interpretation can be applied to the difference in wages due to the single dummy: There was a higher proportion of single immigrants than non-migrants and they earned less compared to married non-immigrants, therefore the contribution to the difference was positive. The negative difference in wages due to differences in coefficients accounted for 48% of the difference in log of wages, implying that immigrants are paid more highly for the same characteristics than non-migrants. However, only the difference in the coefficients of single, hours per week and the intercept were statistically significant. The difference in intercept was the difference in the average of the conditional wages between non-migrants and immigrants. Since immigrants earned more than non-migrants, the difference was negative. The difference in wages due to the interaction between the coefficients and endowments accounted for 28% of the total difference. Only the single dummy was statistically significant and negative. The larger proportion of singles among the immigrants yielded a larger penalty on their wages compared to married immigrants. Another large contributor to the difference due to the interaction was experience. Non-migrants were compensated better for their increasing 28 The contribution by education to the difference due to endowments is negative, while the differences due to the Female and Single dummy variables are positive, offsetting the difference, and therefore it is possible for the contribution of education to be larger than the overall contribution of the endowments.

105 88 experience than immigrants. The difference, though, was not statistically significant. I report the results for the Oaxaca decomposition of wages for men in Table 26. The difference in men s wages was negative and statistically significant. Following the same method as before, the difference was divided in three categories; and although the largest contributor to the difference seemed to be the endowments, the difference was not statistically significant. The difference due to the coefficients accounted for 64% of the differences in wages and it was statistically significant. Finally, the difference in wages due to the interaction between coefficients and endowments was -78%. This implies that non-migrants are paid more highly for a characteristic if the mean of the said characteristic is lower for non-migrants than for immigrants. The difference due to this component, however, was not statistically significant. The main contributors to the difference due to differences in endowments were education and marital status, which, with the exception of the single dummy, had negative signs. The main contributor to the difference due to the coefficients was experience. The return to experience was higher for non-migrant men than for immigrant men. Marital status was also the largest contributor to the difference due to the interaction between coefficients and endowments. The results for the decomposition of women s wages are presented in Table 27. The negative difference in wages for women, as it was the case for men, was significant. But in contrast with men, women s differences in wages were mainly due to the interaction between endowments and coefficients. Twelve percent of the difference in wages was due to the difference in endowments, and

106 89 the main contributor was the difference in education. Only 7% of the difference was due solely to any difference in coefficients. While the proportion of the interaction was negative for men, it was large and positive for women. This implies that non-migrant women are compensated more for their observable characteristics when on average they are better or larger than those of immigrant women. Although non-migrant women work less hours per week on average than immigrant women, their coefficient is larger, therefore non-immigrant women seem to be compensated better for their effort than immigrant women, in this case. None of the three components of the decomposition appeared to be statistically significant, however. VI. Conclusions The purpose of this chapter was to analyze how immigrants from Mexico performed in the labor market in Mexico compared to Mexicans who did not migrate. In particular, two outcomes were examined: labor market participation and wages in Mexico using two data sets for the analysis - the New Immigrant Survey for the immigrant sample and the Household Income and Expenditure National Survey from Mexico for the non-migrant sample. To examine employment outcomes, the probability of employment for both groups was estimated and the analysis was carried out separately for men and women. In general, immigrants are less likely to be employed than nonmigrants, and with the exception of education, observable characteristics affect each group s probability of employment differently. While education and experience are the most significant determinants of employment for immigrants,

107 90 the probability of employment among non-migrants is affected by gender, marital status and number of children, in addition to education and experience. Non-immigrant men, who are the most likely to be employed, are least affected by education in terms of the probability of employment, which can be explained by their low average levels of education. On the other hand, immigrant women, who are the least likely to be employed and who have more education than non-migrants, seem to encounter more difficulties in employment. In fact, education has a small effect on their employment. These results appear to be consistent with those of Lozano and Lopez (2007), who argue that Mexican immigrant women are positively self-selected. If labor market discrimination is stronger in Mexico than in the US, then high-ability women may find a strong incentive to move north. Furthermore, if education has such a small impact on their employment outcome, as found in this chapter, then they may be driven to find other opportunities abroad, particularly if they have the means and relationships necessary to do so. The analysis in this chapter has also shown that immigrants earn significantly more than non-migrants in Mexico, before their migration to the US. Although the rate of return to education for both groups in the full sample is the same, the determinants of their wages vary between the groups. Experience and number of weeks per year are only significant for non-migrants, which indicate that employers compensate non-migrants based on effort and knowledge acquired in the job, as well as for education, while immigrants may be earning more due to better quality of education or by having different occupations or working in different industries. Immigrant men and women have similar rates of

108 91 return to education and this appears to be the main observable characteristic for the determination of their wages. Moreover, non-migrant men have lower returns to education than immigrant men while non-migrant women have higher returns to education than immigrant women. Immigrant women, who are highly educated but mainly unemployed, and experiencing a lack of opportunities in Mexico, are underemployed and therefore have a lower return to education than non-migrant women. The fact that they earn more than non-migrants does not contradict their underemployment status, because their earning potential may be even greater. Mexican women may find that their skills are better compensated for in the US. The final section in this chapter showed that the large differences in labor market participation and wages between immigrants, who are currently legal permanent residents in the US, and non-migrants cannot be solely explained by their differences in observable characteristics. The application of the Oaxaca decomposition strategy showed that 13% of the employment gap and 25% of the wage gap between non-migrants and immigrants can be explained by their different endowments. The large proportion of the employment and wage gaps between non-migrants and immigrants that cannot be explained by their endowments suggests that immigrants suffer different treatment in the labor market, which is due to differences in unobservable characteristics. By separating the samples by gender, even more differences were uncovered. Although both immigrant men and women have higher education and wages than non-migrants, their characteristics are compensated for differently in the labor market. Only 12% of the differences in the wages of

109 92 women are explained by differences in observable characteristics, while 7% of the difference is explained by the differences in the coefficients, plus 81% of the difference is due to the interaction (recall that the interaction term is part of the unexplained difference between the two groups). A positive interaction term tells us that the characteristics in which immigrant women are better (more experience on average, for example), are compensated for at a better rate than non-immigrants (larger coefficients). The negative difference between immigrant and non-migrant men can be explained by differences in their characteristics, for the most part. But the interaction term, which is part of the unexplained difference, is large and negative. It is important to remember that the sample of immigrants used in this study are migrants who for the most part went through the immigration process to the US legally, and therefore they do not represent the whole population of Mexican immigrants that live in the US since a large portion of them, especially recent immigrants, entered the country illegally or over-stayed their nonimmigrant visas. An immigrant who has the resources to wait for their immigration papers and the networks to support them as they go through the process of immigration may demonstrate that he or she is already in a better position (financially or otherwise), possibly due to better education and opportunities, than the immigrant that sees no other choice than to cross the border illegally. Therefore, it is not surprising to see that the recent legal permanent residents from Mexico have more years of education and earned higher wages than their counterparts in Mexico. In terms of employment, though, they lag behind non-migrant Mexicans. This may be due to the lack of

110 93 opportunities that they encounter based on their education and their reservation wages. This is especially true for immigrant women who have lower returns to education and a very low probability of employment even though the wage premium paid over non-immigrant women is larger than that between immigrant and non-migrant men. The negative differences found in the wage decomposition between nonimmigrants and immigrants leads to the conclusion that documented immigrants from Mexico are positively self-selected in observable as well as in unobservable characteristics, which are captured by the differences in endowments and coefficients, respectively. Although they suffer from higher rates of unemployment, immigrants have higher levels of education and earn higher wages than non-immigrant Mexicans. These results are stronger for women and tend to agree with those of Chiquiar and Hanson (2005) who provided evidence that women from Mexico, more so than men, are positively selected. However, Chiquiar and Hanson s results were limited to observable characteristics. With the use of the Oaxaca decomposition strategy, the results in this chapter suggest that legal immigrant Mexicans are also positively selected in terms of unobservable characteristics, such as ability.

111 94 CHAPTER 4. TRANSFERABILITY OF SKILLS, INCOME GROWTH AND LABOR MARKET OUTCOMES OF RECENT IMMIGRANTS IN THE UNITED STATES I. Introduction This chapter analyzes three aspects of the labor market experience of immigrants in the US. First, the determinants of labor force participation are explored. Second, the chapter examines the factors affecting wages, including an analysis of rates of return to education. Third, the New Immigrants sample s data on wages abroad (before their move to the US) is used to estimate transferability of skills and income growth of the immigrants in the sample. I test the hypothesis that immigrants wages from abroad explain wages in the US independent of all other characteristics. I expect to find that in addition to measureable observable individual and country-level characteristics, there is a set of skills that individuals possess that can be measured through wages earned abroad, when we assume that wages measure unobservable skills that affects the individuals productivity in the labor market, in addition to quantifiable skills such as education and experience. The chapter carries out research on a set of topics rarely explored in the literature. For instance, analysis is carried out on how visa/immigration categories affect immigrant labor market outcomes. Since there are different channels through which someone can (legally or illegally) immigrate to the US, which includes family ties, employment, protection for refugees, etc., not all immigrants in the US are seeking economic improvement, that is better

112 95 employment opportunities or higher wages. A better life can be attained by moving to a safer country, by reuniting with family, by supporting the spouse with household work while he or she goes to a better-paying job in the US, or by improving their quality of education, etc. Indeed, those immigrants that arrive with family-related visa categories are less likely to be employed than those with other types of visas such as employment preference, refugees and legalization (See Figure 5). The chapter will examine the extent to which visa categories matter in determining the economic progress of immigrants. I test whether immigrants with economic and non-economic visas perform differently in labor market. More specifically, I test whether the level of transferable skills is different for these two groups. I expect to show that noneconomic visa holders have higher levels of transferable skills, independent of other observable characteristics. Previous research has shown that immigrant income growth depends on the level of skill transferability. The quality of skills such as education and English language skills and their transferability depends not only on the individual but also on the country of origin. Those who come from countries where English is the official language are expected be better able to use their skills in the US than those who cannot communicate in English. A higher percentage of wages in the US is explained by wages abroad for immigrants from the United Kingdom, Canada and India, than for immigrants from Haiti Effect of wages abroad on wages in the United States of Guatemalan immigrants (with no additional controls) disappears when wages are estimated by gender.

113 96 Figures 6 and 7 show the relationship between wages in the US as explained by wages abroad and their country of origin. 30 An immigration policy reform seems to be inevitable in the near future. Whether the US will change the family reunification and employment preferences policies will depend on what outcomes they desire. Immigrants are expected to be productive and contribute to the US socially as well as economically. In general, we expect immigrants to have a positive impact on their communities. Should the US attract immigrants based on their skills? Is the current migration policy responsible for the current distribution of immigrant skills in the US or are other selection criteria at play? While it is not the purpose of this chapter to answer these questions directly, I show that there exist significant differences between immigrants with economic and non-economic visas, in terms of transferable skills, employment status, wages earned in the US and country of origin characteristics. This chapter also shows that although observed skills are important determinants of labor market outcomes of recent legal immigrants, unobserved skills that cannot be measured through education and experience, are also important determinants of initial wages and income growth. Additionally, I show that observed skills such as education and English language do not necessarily capture the effect of unobserved characteristics, when these are left out of the equation. 30 Figure 6 is the estimation for the full sample. Figure 7 divides the sample by gender.

114 97 II. Review of the Literature Transferable skills are those skills that immigrants have acquired through education and experience in the country of origin, and are typically compensated for in the labor market of their country of origin, but due to differences between the US and the country of origin, these skills are not fully compensated for or utilized by the immigrant in the American labor market. In the immigration literature there have been different approaches used to measure the skill transferability of immigrants in the US. Duleep and Regets (1996) use initial wages and income growth to show that the decline in initial wages of immigrants, relative to natives, is not due to the lower quality of immigrants, but rather lower transferable skills. They found the newer cohorts experience faster income growth than older cohorts, and they argued that since positive income growth could not be attributed to lower immigrant skills, that the cause of lower initial wages must be lower transferable skills. Chiswick and Miller (2009) measured the level of skill transferability of immigrants by identifying whether immigrants were over or undereducated in their positions relative to the mean level of education required for the current held job. They found that the more experiences abroad they possessed the more likely they were to be mismatched in the skill requirements of the job. Chiswick and Miller attributed this result to the employer s imperfect information on the value of skills acquired in foreign countries. Over education and under education are found to be affected by the country of origin and the immigrants English-

115 98 language skills. Haley and Taengnoi (2011) used the effect of pre-immigration education and experience on wages (including interactions with the country of origin) to measure skill transferability, and focused only on high-skilled workers. They found each year of education abroad increased the earnings of immigrants from Japan and English-speaking developed countries by 10%. They also found post-immigration education had a larger effect on immigrants from developing countries than from English-speaking developed countries. III. Summary Statistics The New Immigrant Survey offers a unique opportunity to explore the skill transferability of immigrants because in addition to wages in the US, it provides documented immigrants wage information before immigration, which allows us to measure skills transferability separately from the selectivity measured by the other characteristics, such as education and language skills. Table 26 shows the weighted summary statistics of the full adult sample by employment status in the US. As observed, 54% of the sample did not report earning any wages in the US. Those who were employed earned an average of $725 per week and worked 46 weeks per year. Immigrants not employed in the US used to earn significantly more than those employed in the US back in their home countries, before their move to the US, but they used to work about the same amount of weeks per year. The average earnings abroad was more than double the average weekly earnings in the US, but this was due to a few (less than 5%) outliers, that did not appear to be mistakes in the data. Forty percent of those employed in the US were

116 99 employed abroad, while only 35% of those not currently working were employed abroad. The 18.5% of the sample who attended school in the US had an average of 4.5 years of schooling there. There is a negative relationship between years of schooling in the US and years of schooling abroad, which suggested that education in the US is a substitute for education abroad. Immigrants employed in the US had an average of 11.7 years of education abroad and.08 years of education obtained in the US. Those who had less than 6 years of education were less likely to be employed and the unconditional probability of employment increased as the years of education increases. Approximately 66.2% of those with 20 or more years of education were employed compared to only 20.2% of those with no education. Among the employed, more than 50% had more than 12 years of education, compared to 41% among the unemployed. There were significantly more women among the unemployed than the employed in the US, 71% versus 44%. Even though the sample consisted of individuals who had recently become legal permanent residents, most of them had spent at least some time in the US. Those who are employed had an average of 3 more years in the US than those who are unemployed. Almost 10.4% of the sample were students, and were distributed almost equally among the employed and unemployed. Employed immigrants were more likely to be married or single, and the same can be said among the unemployed. About 54% of those married and 59% of singles were employed. The 70% of those living together (2.4% of the sample) were employed. Employed and unemployed immigrants had about the

117 100 same number of children under the age of six, but employed immigrants had fewer children in total than unemployed immigrants. Because English would be considered one of the most easily transferable skills to the US, it was one of the most important skills considered in the analysis below. The English variable was decomposed into four categories: a native English speaker is one who comes from a country where English is an official language and the individual listed English as one of the languages he or she speaks; a non-native English speaker is one who reported speaking English in the NIS survey, but English was not an official language in his or her country of origin; an individual who did not speak English is one who did not list English as one of the languages he or she speaks; 31 the forth category was for individuals who did not speak English but who had worked or lived in an English-speaking country to control for the possibility of having had previous experience in an English-speaking country, or that there at least there was minimal knowledge of the English language, but not enough to be listed in the survey. Only 2.5% of the sample fell into this last category. About 46.5% of those who were employed were English-speakers compared to only 31.6% among the unemployed. Native and non-native English speakers were more likely to be employed than those who did not speak the language. There were no significant differences between the unemployed and employed as far as other country variables were concerned. The rates of unemployment, average schooling and log of GDP per capita were similar for both groups. 31 NIS survey allows respondents to list up to five languages. Given that English is considered a world language and the survey most likely took place in the United States, I assume that if English is not a top five language for a respondent, then the respondent does not know English.

118 101 Employed immigrants, as well as unemployed immigrants were more likely to have obtained legal permanent status through marriage to a US citizen or to a legal permanent resident. About 49% of those employed had some relationship in the US (such as a spouse, parent, child or another relative) compared to 69% among the unemployed. About 74% of immigrants with employment preferences or those who were refugees were employed, while 78% of those who were legalized were employed. The group least likely to be employed was the parents of US citizens followed by fourth family preferences and spouses of legal permanent residents. While most of the sample were principal immigrants, under some visa categories, such as employment preference, the principal immigrant can bring his or her spouse, therefore it was not necessarily expected that all immigrants with employment preference visas would be employed. 32 About 9.5% of those employed and 13.6% of those unemployed were spouse of principals. A green card adjustment occurs when the temporary or visiting US immigrant decides to apply for the permanent residence status, therefore these individuals may have more experience or assimilation time in the US than those who apply while outside the US, making them possibly more likely to be employed in the US - 73% of those employed had a green card adjustment compared to only 39% among the unemployed. In separating the sample further by gender, more differences among these groups were evident (see Table 27). That is, only 42% of women were employed compared to almost 70% of men. Those women who were employed earned less 32 In addition, once a person obtains the legal permanent resident status, the individual is does not need to remain employed with the sponsor to remain legally in the country.

119 102 than men by more than $340 per week, even though men only worked one more week per year than women, on average. Women who were employed in the US used to earn more than $445 more abroad than women who are unemployed in the US but were employed abroad. The main difference in wages abroad and in the US comes from unemployed men. Employed men used to earn $1,109 on average more abroad than what they were earning now, while unemployed men used to earn almost $5,250 per week abroad and had no income currently. Again, this figure includes very large outliers that do not appear to be mistakes in the data. Almost 37% of employed women and only 32% of unemployed women used to be employed. In contrast, unemployed men were more likely to have been employed abroad than those who were employed in the US. Employed men had the most years of education abroad, followed by employed women. Unemployed men had the most years of education in the US, while employed women had the least amount of years of education in the US. Approximately 51% of employed women and only 40% of unemployed women had 13 or more years of education. Only 11.3% of women and more than 47% of men who did not have any education were employed. In general, less than half of men and women who had no education or 1 to 5 years of education were employed. In contrast almost 74% of men who had 20 years of education or more were employed. Employed men and women had spent more time in the US than the unemployed. There was a longer lag between their last job abroad (or graduation from school) and getting a job in the US for women than for men. A larger percentage of employed women than unemployed women were students, while

120 103 the opposite was true for men. About 9.5% of employed men compared to 10.2% of unemployed men were also students. Single men and women were more likely to be employed than unemployed. In contrast only married men were more likely to be employed, but not married women. The groups most likely to be employed were men and women who were living together, but were not married. Unemployed women had the most number of children and the most number of children under the age of six. Employed women had the least number of children, but unemployed men had the least number of children under the age of six. Most of the individuals in the sample did not speak English, but this group was also less likely to be employed, regardless of gender. Those who did speak English, but were not native speakers were the most likely to employed (53.6% of women and 77.8% of men). Individuals who did not speak English but worked in an English-speaking country were the group least likely to be employed, regardless of gender. There was not a large difference between men and women regarding log of GDP per capita and their country of origin s average schooling level. Employed men came from countries that had lower unemployment rates than unemployed men, but employed women came from countries with higher unemployment rates than unemployed women. IV. Labor market participation, return to education and transferable skills of immigrants in the United States A. Probability of Employment

121 104 Suppose that the reservation wage of an individual depends on individual and their country of origin characteristics. If the reserve wage is below the offered wage, the individual decides not to work, and we don t observe wages for them. (1) * Y if Y 0 and * Y 2 1 Where Y missing if Y * 0, 1 * Y1 is the reserve wage of the individual and * Y 2 is the actual wages for those who work. Since I did not observe the reserve price, but only whether the individual worked or not, I used a probit estimation for binary responses (works and does not work) to determine labor market participation. x' (2) F( x) Pr( y 1 x) ( t) dt ( x' ) x 1 exp 2 1 ( 2 x) 2 dx where ( x' ) is the cumulative density function for the standard normal and β is the parameter of interest. The marginal effect of the individual characteristic k (Xk), where k= 1, 2, 3,, K, on the likelihood of employment is Φ(XB)*βk where βk is the estimated parameter associated with Xk. The marginal effects are usually estimated at the mean of the X variable. An individual is considered to be unemployed if they did not report wages in the US during the interview. This group may include those individuals who are not looking for employment because they are homemakers, disabled or students. I include a dummy variable for student status and retired individuals are excluded from the analysis. An individual was considered employed if he or she reported earning positive wages for a job performed. The sample was limited to

122 105 individuals whose American job took place in the US 33 and who was between 18 and 65 years of age. 1. Full Sample Results In Table 28, probit results for the probability of employment in the US are presented for the full sample. The full sample consists of 7,890 observations, but the weighted amount of observations is actually 7,869. Column 1 in Table 28 shows the probit results and the marginal effects that control for country of origin (aggregated by region), race, education, experience, gender, years in the United States, years of unemployment 34 and whether the individual had a job abroad. Only those individuals who had more than 12 years of education were more likely to be employed than those without education. 35 Females were 22% less likely to be employed. Experience had a positive effect on employment, as well as years in the US. Each year of unemployment decreased the probability of employment even more and those who never had a job abroad were also less likely to be employed. Even though region of origin was controlled for, there were still differences in employment between black and white immigrants. Black immigrants were 9% less likely to be employed than white immigrants. All other 33 Survey asks the respondent where each job took place, even when asking specifically about jobs in the USA. In 3% of the cases, the answer was not the United States. 34 If the individual did not work abroad, then this variable measured the time since leaving school. This variable could be highly correlated to experience, but only for those individuals who were unemployed abroad and who were currently unemployed in the United States. However, both variables, experience and years of unemployment, were always statistically significant in these regressions. 35 Table A12 I show the results of using continuous variables for education (years of education in the United States and years of education in country of origin) instead of categorical variable, with very similar results. Years of education in the United States is only marginally statistically significant in the first column.

123 106 races did not experience a difference in employment compared to white immigrants. Only immigrants from Western Europe, excluding the United Kingdom were more likely to be employed in the US than developed Englishspeaking countries. Immigrants from all other regions did not show any significant differences. In column (2) a dummy was added to differentiate whether the individual had been an immigrant before, and some selection variables were also added. These selection variables could influence the decision of the individual to participate in the labor market, although they are not expected to have a significant effect on wages because they are not directly related to productivity or perceived productivity. In addition to the number of children and number of children under age six, which are the most commonly used selection variables in the labor market participation literature, some immigration variables were included in the equation. These immigrant data identify whether the individual s initial motivation for migration were family reunification/safety or better economic returns. A non-economic-visa migrant was one who migrated through a family preference or refugee visa, while an economic-visa migrant was one who migrated through employment preference or diversity visa or arrived in the US illegally. 36 Spouses of principals may be tied-movers and green card adjustments are required for those who apply for permanent residence once they are already 36 As mentioned in Chapter 2, although true motivation cannot be fully assumed, the visa category implies some motivation, at least initially. Immigrants with employment preference or diversity visas and those who were illegally in the US were mainly economically motivated to immigrate. Immigrants that had family-related visas, such as marriage, were not necessarily economic migrants, or at least not primarily.

124 107 living (legally and temporarily) in the US. It is possible that non-economic-visa migrants are less likely to be employed if the main motive to migrate is not economic gain, and the same could be assumed for tied movers. With the added variables, the marginal effect of years in the US decreased by more than half, but years of unemployment stayed unchanged. Black individuals were still less likely to be employed but the effect decreased from 9.4% to 5.8% (in absolute value). Those immigrants who were immigrants in their last country of residence were less likely to be employed in the US, holding employment status abroad constant. Although immigrants who had moved to more than one country may signal that they are motivated to move in search of their best possible return, they may also have invested in country-specific skills in more than one of one country that may or may not be transferable or relevant in the US and therefore are more likely to find themselves unemployed initially than those who arrive directly from their country of origin. As predicted, non-economic-visa migrants as well as spouses of principals were less likely to be employed, while those who adjusted their immigration status after they arrived in the US were more likely to be employed. Total number of children did not affect the probability of employment, while the number of children under the age of six did negatively. 37 While country of origin characteristics were not expected to have an effect on employment in the US directly, the log of GDP per capita, average schooling and unemployment rates may capture information that the country fixed effects 37 The name of the dummy variable in the tables is Non-economic migrant

125 108 may not, since some of them vary by year of immigration. Therefore I included these variables in the third column. The higher the GDP per capita from their country of origin, the less likely they were to be employed, however, the higher the unemployment rate, the more likely they were to be employed in the US. It appears then that those who were experiencing high unemployment rates or lower wages abroad were more likely to migrate for economic reasons and therefore may be more likely to be employed in the US. The addition of country-level variables affected the coefficient on the race variable as well. Black immigrants were now 8.7% less likely to be employed (from 5.8% in column 2). In the previous specification, race may have been capturing country level characteristics such as higher unemployment or lower GDP per capita among black immigrants and therefore the marginal effect was smaller than in column 3. In column (4), controls were added for whether individuals spoke English or not and for their student status. Individuals who were not native speakers or did not speak English were not less likely to be employed than those who were native English speakers, holding everything else constant. However, being a student had a negative effect on employment. Including these variables in the specification did not change previous results. Finally, in column (5) marital status was added to the specification. Widowed immigrants were statistically less likely to be employed and divorced

126 109 immigrants were more likely to be employed than married immigrants. 38 The remainder of the variables, including number of children, remained virtually unchanged. 2. Results by Gender Men. Table 29 presents the probit results and marginal effects for the male sample only. Among men, education did not make a difference in the probability of employment for all categories compared to those without education. 39 Experience and years in the US had positive effects on the probability of employment, while years of unemployment and being unemployed abroad had negative effects on employment. Only black men were less likely to be employed than white males. Region of origin has no effect on employment. Among men, non-economic-visa migrants and husbands of principals were less likely to be employed, while men who had their immigration status adjusted were more likely to be employed. Number of children, but not the number of children under the age of six, increased the probability of employment. Country-level characteristics, added in column (3), did not have an 38 76% of widowed immigrants immigrated with Parent of U.S. citizens visas, the least likely visa category to be employed as well. 39 In Table A13, continuous variables for education were used (years of education in country of origin and years of education in the United States) instead of the categorical variable for education. In the male sample, years of education in the country of origin was only statistically significant in the first column, but years of education in the United States have a significant and negative effect on the probability of employment that does not disappear with the inclusion of the student dummy. Controlling for years of education in the United States in Table 29 did not change the results and years of education in the United States did not appear to be statistically significant.

127 110 effect on the probability of employment, and neither did being an immigrant abroad. In column (4) controls were added for student status and English skills. Students were less likely to be employed, while English skills had no effect for men. In column (5) marital status was added. Widows and single men were less likely to be employed than married men. After controlling for English skills, student status, country-level characteristics and marital status, it was men who had children under the age of six who were less likely to work while the total number of children has no effect on employment. Women. In contrast to men, women s education did have a positive effect on employment in the US, as shown in Table 30. Women with 1 to 6 years of education were 13% more likely to be employed than those without education, and the probability increased by years of education up to 34% for those with 20 years of education or more. After controlling for other characteristics, such as the selection variables, English ability, student status and country-level indicators, education of less than 12 years became statistically insignificant but having 13 years of education or more still had a positive effect on employment. Similarly to the male sample, experience and years in the US had a positive effect on employment. However, having been unemployed or an immigrant abroad had no effect on employment for women. In contrast to men, race had no effect on women s employment status in the US, and this did not change after the inclusion of country-level characteristics. Women who emigrated from Western Europe were more likely to be employed than women who come from Englishspeaking developed countries.

128 111 The selection variables in column (2), confirm that, as it was the case among men, women with family preferences or refugee visas (non-economic-visa migrants) and tied movers (wives of principals) were less likely to be employed in the US, as it seems that employment was not the main motivation for migration for this group of women. The probability of employment decreased with the number of children under the age of six for women, a variable that remained negative and statistically significant for women regardless of the additional variables I included in the specification. In contrast to the male sample, country-level characteristics were all statistically significant for women. The likelihood of employment decreased as the level of GDP per capita increased, while it increased as average years of schooling and the unemployment rate increased. Recall that using the same sample, women were less likely to work abroad as the average years of schooling increased, and now the reverse in terms of employment in the US is observed. The larger the gap there was between the woman s level of education and the average education experienced in her country of origin, the more likely she was to be employed, due to possibly receiving a premium in her over-education, but this premium may not be present in the US, and therefore, on average she is less likely to be employed in the US, holding employment abroad constant. After country-level characteristics were added, spouse of principal lost its significance, which may be because the distribution of the average years of schooling was very different for those who were spouses of principals and those who were not and being a spouse of a principal was not as important in the

129 112 determination of employment as the average schooling of the women s country of origin. Similar to men, English skills had no effect on the probability of employment, but marital status did (column 5). Divorced, separated, those women not married but living together and single women were more likely to be employed than married women. The effect of number of children under the age of six did not change when marital status was added to the specification. It is evident from these results that a complete picture of all the determinants that affect immigrants in the labor market cannot be fully appreciated by looking at the survey sample as a whole. The results for immigrant men and women showed that their labor market participation was different in terms of some aspects of their education, experiences, unemployment situation abroad, immigration situation and country of origin. For men, experience appeared more important, while education appeared more important for determining women s employment. Number of children under the age of six affected both men and women negatively, but total number of children only affects men (positively), but only if other characteristics, such as marital status, are ignored. And while single men were less likely to be employed, single women were more likely to be employed. B. Returns to Education Ordinary Least Squares Wages are determined by education, experience, experience squared and other sometimes observed or unobserved individual characteristics. The linear

130 113 relationship between the log of wages and these characteristics is traditionally referred to as the Mincerian wage equation, and it is estimated using Ordinary Least Squares. The rate of return to education is the estimated coefficient on the education variable. (3) Yi 0 1X i 2Zc i is the Mincerian wage equation and Yi is the log of weekly wages in the United States of immigrant i from country of origin c; X i is a vector of the individual characteristics of immigrant i, including education; Z is a vector of country-level characteristics; and i is the individual error term. For this analysis, the sample was limited to individuals who reported positive wages in the US, whose job was in the US and who were 18 to 65 years of age at the time of the survey. Wages in the US, the dependent variable, are weekly wages reported to the interviewer earned in their most recent job in the US. Weekly wages were calculated based on the amount earned, the frequency of the paychecks and the number of weeks that they worked per year. In addition to the standard controls for all wage equations (education, experience, experience squared, gender and race) and for wage equations for immigrants (years in the US, years in the US squared, country of origin), I also included whether the individual was a native or non-native English speaker or did not speak English; country-level characteristics; immigrant type and other visa-related variables; and whether their highest degree was obtained in the US. The analysis was also able to separate how many years of education were c

131 114 obtained in the US and how many years of education were obtained in their home country. Those results are presented in the appendix (Tables A15, A16, and A17). 1. Full Sample Results Columns (a) in Table 31 present the results for the OLS regressions. Column (1) only controls for education, gender, experience, years in the US, number of weeks per year worked, race and country of origin. The results for the country of origin are presented in Table A18, in the Appendix. Having less than 12 years of education did not affect wages more than not having any education at all; however, having 13 to 19 years of education (post-secondary education) increased wages by 23% and having more than 20 years of education increased wages by 52%. 40 Females earned almost 40% less than men. As expected, experience was positive, but although experience-squared was negative, it was virtually zero. Years in the United States, a measure of assimilation, also increased wages at a decreasing rate. Weeks per year, a measure of individual s drive, was positive and statistically significant. Compared to whites and holding everything else constant, Hispanics earned more than white immigrants, Asians earned similar wages, and blacks earned about 20% less than whites. Immigrants from all regions earned 40 In Table A15 (in the Appendix) instead of using total education, I separate education in years abroad and years in the United States. Both variables are positive and statistically significant, but the return is higher for years of education abroad than in the United States. Having a degree from the United States (which is positive and statistically significant) changes years of education in the United States to be only marginally significant, probably due to the fact that most likely immigrants who do not have a degree from the USA probably also do not have any or very few years of education in the USA, so these two variables are highly correlated.

132 115 less than immigrants from developed English-speaking countries, and the largest difference was observed between Latin America and the Caribbean and Englishspeaking developed countries. In column (2a), the English variable was added, which shows that those who did not speak English earned less than native English speakers, even after controlling for country of origin, but those who were not native English speakers or who had worked in an English-speaking country, but did not speak English, did not earn differently from native English speakers. Earning a degree in the US, another sign of assimilation, had no effect on wages. The difference observed between immigrants from English-speaking developed countries and Western Europe, Asia and Africa appeared to be due to the differences in English skills. Once controlled for English skills, the difference disappeared, and only those immigrants from Eastern Europe and Latin America and the Caribbean earned less than immigrants from English-speaking developed countries. In column (3a) country-level characteristics were added, which did not have an effect on wages. Other variables associated with country characteristics, such as English ability and race, may have changed slightly. For example, it was previously estimated that immigrants that did not speak English earned 22% less than native English speakers, but after controlling for country characteristics, the difference was 17% instead of 22%. It was also previously estimated that black immigrants earned 23% less, but now the difference was closer to 26%. The sign and significance level of the regions of origin remain unchanged. In column (4a), visa-related variables were added. These indirectly measure the motivation (economic or non-economic) to migrate to the US. These

133 116 variables were discussed in a previous section of this chapter. Non-economic-visa migrants earned about 24% less than economic-visa migrants and among immigrants who worked, spouses of principals earned about 10% less than principals. Those immigrants that applied for immigration after entering the country earned 16% more than those who applied for their immigrant visas while still living outside the US. Including these variables eliminated the effect of years in the US, a measure of assimilation, on wages. Since Green Card adjustment may also be considered a measure of assimilation, this change was not very surprising. Immigrants from Asia and Oceania are no longer differentiated from those from English-speaking developed countries. This may be because most immigrants from Asia are considered economic-visa migrants and without controlling for immigrant type the region dummy was acting as a proxy for this. 2. Results by Gender The results for men and women are presented in Table 32 and Table 33, respectively. Holding everything else constant, men and women who had at least 13 years of education earned significantly more than their counterparts without education. Women who had at least 20 years of education (about 4% of the sample) earned 51% more than women without education, compared to men with the same education (about 5.3% of the sample) who earned 62% more than men without education. When education is introduced as a continuous variable, as is shown in Tables A16 and A17, one can see that both education abroad and

134 117 education in the US affected women s wages positively, but only education abroad was statistically significant for men. The rate of return to education abroad was very similar for men and women. Having a degree from the US, after controlling for education abroad and in the US had no effect for females, but it had a positive effect for males. Women s experience (accrued from abroad and in the US) was not compensated for in the American labor market. Men s experience also becomes indistinguishable from zero when immigrant type and other visa-related variables were included in the specification. Years in the US, a measure of assimilation, had a positive and significant effect on women and men wages, but the effect was larger for men. However men s assimilation grew at a decreasing rate, while women s assimilation seemed to have a linear relationship with wages. Race has a different effect on wages for men and women. Black women earned about 24% less than white women while Asian women do not earn differently from white women. Conversely, although Hispanic immigrant men earned more than white immigrants, black men did not earn more or less than white immigrant men. Hispanic immigrant men earned approximately 39% more than white immigrant men, a difference that only increased when country-level characteristics and visa category controls were included in the specification. Having a degree from the US (added in column (2a)) had no effect on women s or men s wages. Men who were not native English speakers or who did not speak English earned 22% and 34% less than men who were native speakers, respectively. This was not the case for women, who were not penalized for not speaking English or not being native English speakers.

135 118 Before controlling for English skills, men from Asia and Oceania and Africa earned 46% and 37% less than immigrants from English-speaking developed countries, but after including English skills, we see a difference due to this skill and the regional difference are no longer observed. The difference between immigrant women from Eastern Europe and English-speaking developed countries was only marginally significant, and disappeared when English skills were added to the specification. Country-level characteristics, added in columns (3a) showed that men earned less as their average years of schooling increased, while women earned more. The GDP per capita and unemployment rate had no effect on men s or women s wages. Controlling for these characteristics changed the estimated effect of English skills for men. Non- native English speakers did not earn differently from native English speakers, but those that did not speak English skill earned less (24%) than native English speakers. No changes were observed in the women sample. In columns (4a) immigrant type and other visa-related characteristics were added to the specification. Non-economic-visa migrant men and women earned 27% and 19% less than economic-visa migrants. Being a spouse of a principal had no effect on wages, but those with Green Card adjustments earned 13% more than those who applied for their migrant visas while abroad. Since the estimated premium paid to migrant men with more than 13 years of education increased in column 4, it seems evident that, at least for the male sample, the non-economic-visa migrant dummy was capturing higher unobserved skills in addition to motivation (for migration). Immigrant black men earned less

136 119 (although the difference is only marginally significant) and Hispanic men earned more than immigrant white men, but this was after controlling for region of origin. Immigrant men from Latin America and the Caribbean earned the least, compared to immigrants from English speaking developed countries. Men from Eastern Europe and Asia and Oceania also earned less than immigrants from English-speaking developed countries, but the difference in wages was not as large. Among women, only black immigrant women earned less than immigrant white women. The difference between immigrant women from Latin America and the Caribbean and English-speaking developed countries was not as large as that observed among men, and the difference was only marginally significant. After estimating wage equations with Ordinary Least Squares, one should be concerned about having obtained biased results, especially since the current analysis was limited to those individuals whose wages one could observe and individuals whose reserved wage were not evident were excluded from the sample. Given the importance of this issue, the following section presents the sample section bias methods used in this chapter. C. Heckman Selection Model The underlying reasons that individuals have for entering the labor market, such achieving a reserve price for their time, are usually not observable to the econometrician. What can be observed are wages obtained when an individual works; otherwise, having missing information renders valid calculations difficult. In the case of immigrants, it is also believed that an

137 120 individual decides to immigrate if he or she anticipates that his or her wages in the foreign country will exceed his or her wages in the home country plus all the direct costs (such as travel) and indirect costs (such as lost wages) associated with migration. The analysis of the transferability of skills of immigrants suffers from selection bias because even though we can observe wages earned before migration, not all immigrants, either in their country of origin or in the US decide to work or are able to find a job. In order to address this issue, I used the (full information) maximum likelihood (FIML) Heckman selection model instead of the two-step Heckman selection model to better understand this issue, since most of the explanatory variables in the selection equation are shared in the wage equation. The FIML also offers more efficient and consistent results over the twostep method, is easier to use and more popular, since both equations share many of the same explanatory variables (Nawata, 1993) and the FIML estimates all parameters simultaneously. Consider the traditional wage equation (4), which, if estimated, will suffer from self-selection bias: u a (4) LnWic LnWic X ic Cc ic where u LnW ic and a LnW ic are log of wages in the United States and abroad for individual i from country c, respectively. such as gender and education for individual i in country c, X ic is a set of individual characteristics C c is a set of country level characteristics such as log of GDP and average level of education for country c, and ic is the individuals error term.

138 121 This equation suffers from sample selection bias since we only observe US wages if the immigrant decided to work or if he or she is able to find a job. In order to address this issue using the Heckman selection method, one needs to find an explanatory variable that affects the selection equation without directly affecting the wage equation. For immigrants in the US, I noticed that individuals that have spent more time in the US tended to be more likely to be employed than those that had just arrived in the country. I also tested the possibility that experience in the US (added together over the number of times they have immigrated to the US) had an effect on employment. This variable, however, did not have a lot of predictive power in the context of assessing the probability of employment. One should thus be cautious in assuming that years in the US only affected the probability to participate in the job market and that it had no effect on wages. In my analysis, I saw that the variable for years in the US lost its predictive power in the wage equation when total experience or experience in the US was added to the equation. Therefore it is safe to assume that years since migration affected wages through the probability of employment. 1. Maximum Likelihood Heckman Selection Model Wages of immigrants in the United States can be modeled by two latent dependent variables models: * (5) Y1 Z 1 * (6) Y2 X 2

139 122 Equation (5) is the selection equation and equation (6) is the outcome equation. X and Z are sets of individual characteristics, not necessarily exclusive of each other, but with at least one explanatory variable in equation (5) that does not affect equation (6). The errors are assumed to be jointly normally distributed 1 0 ~ n, 2 0 where 1 2 and (1,1 ) Since the outcome of the participation equation (5) is only observed in binary form, the variance of 1is not observed, however we are only interested in the sign of * Y 1, so the variance is therefore arbitrarily set to 1. From this, we only observe * Y if Y 0, which indicates market labor * 2 1 participation, therefore the observed equation becomes (7) * Y Y 2 if * Y 1 0 and Y missing if Y * 0, 1 and the probability of participating in the labor market is pr( z 1) ( ' i w i ). Therefore the expected value of y (individuals wages), conditional on participation ( z 1) and on individual characteristics x ) ( i is (8) ' E y z 1, x ) x E z 1) ( i i i ( 1 i i x E( w i ' 1i 2i i ) ' ' ( wi ) xi since ' ( w ) i ' ' ( wi ) E( 1 i 2i wi ). ' ( w ) i

140 123 ' ( wi ) is the Inverse Mills ratio and it is always positive, therefore the sign of the ' ( w ) i bias depends on the covariance, where is the correlation between 1and 2, and is the variance of 1. If 0, this would indicate that equations (5) and (6) are independent of each other and that there is no selection bias. 2. Estimating the Full Information Maximum Likelihood function Let D=1 if Y is not missing, that is if the immigrant is working, and D=0 otherwise. An individual can have 2 possible outcomes, pr ( D 0) or pr D 1, y ) 41, where ( 2i ' (9a) pr ( D 0) pr( z ) ( z i 1i i ' 1 ( ' ) and ' (9b) pr D 1, y ) f, ) ( 2i z i ) ( 1i z i 2i f ' z i 1 (, 1 i 2i ) d 1i f ' z i 1 ( 1 i 2i ) f ( 2i ) d 1i f ( 2 i ) f ( 1 i 2i ) ' z i 1 d 1i and using y ' x 2i 2i i 41 Notation adopted from Bierens (2007).

141 124 2 ' 2 ' ' ) )( / ( 2 ) ( exp 2 1 i i i i i x y z x y 42 The maximum likelihood function is estimated by multiplying both probabilities, equations 9a and 9b: (10) ), ( L ) 1, ( 0)* ( i i i y y pr y pr n i D i i i i i D i x y z x y z 1 2 ' 2 ' 2 2 ' ' 1 ) )( / ( 2 1 exp 2 1 * )) ( (1 Since the logarithm of the likelihood function is easier to maximize and the maximum of the function is the same as the maximum of the logarithm of the function, it is therefore more common to estimate the logarithm of the likelihood function. To simplify the equation and following Green s notation, 43 let (11) i i z q ' / 1 / 42 ) ( 2i f ' ) ( exp 2 1 i x i y 1 ' ) ( z i i i i d f ) ) (( ' 2 1 i i i z pr 2 ' 2 2 ' 2 ' ) )( / ( 1 ) )( / ( ) ( i i i i i i i x y z x y pr 2 ' 2 ' 1 ) )( / ( i i i x y z 43 Green s Lecture Notes

142 125 2 / ln atanh atanh exp(2 ) 1 ( ) exp(2 ) 1 then (12) log L log ( qi ) d 0 d 1 log (1/ 2)( y i ' x ) i 2 log [ ( y i ' x ) q i i 2 1 ] which is the equation of interest (the FIML Heckman Selection equation) and the equation to be estimated. For technical reasons 44, Stata, the statistical tool, estimates atanh 45 and not, but the null hypothesis H0: ρ=0 is equivalent to H0: atanh ρ=0. D. Heckman s Full Information Likelihood Maximization Estimation Results As discussed earlier, there is a need to reexamine the OLS results previously presented and consider the possibility of sample selection bias. This bias may arise, for both men and women, from a credible possibility that the population that works and the population that does not differ in their unobservable characteristics, and therefore violates the OLS assumption that the error is uncorrelated with any independent variables and its expected mean equals zero. The previous section explained the Heckman FIML model that estimates the likelihood of participation in the labor market while simultaneously 44 It provides more numerical stability during the maximization to the function. 45 atanh ( 1/ 1 2)log 1

143 126 estimating the wages for those who do participate. This procedure requires that there should be at least one variable in the estimation that explains selection into the labor market that does not explain level of wages. I followed the literature and used number of children, number of children under the age of six and marital status as the excluded selection variables. In the following section, I introduce the Heckman Full Information Maximum Likelihood model and then present the results for the full sample and for men and women separately. 1. Full Sample Results I present the Heckman FIML estimation results for the full sample in Table 34. At the bottom of the table, the statistics sigma, rho and lambda are reported as well. As previously discussed, rho (ρ) is the statistic of interest in the Heckman model. ρ is the correlation between the errors of the selection and the wage equations. If ρ=0, then there is no correlation and there is no selection bias and the results would be equivalent to the OLS results. If, then the Heckman selection model is superior to the OLS model. The estimated marginal effects from the Heckman equation, reported in Table 34, were very similar in size and equal in sign and statistical significance to the OLS results presented in Table 31. The larger differences could be observed in the education category. Although those immigrants who had 12 years of education or less did not earn any more than those with no education, those who had more than 12 years earned more than previously reported. The difference,

144 127 however, was small. The estimated premium for college education was 30% instead of the estimated 25% estimated with OLS. The premium for 20 years of education changed from 56% to 61%. After controlling for English skills, country level characteristics and migrant type, having a degree from the US, conditional on having a degree, was estimated to be positive and statistically significant, which OLS estimated as non-significant. The coefficients of the selection variables employed are also reported on the bottom of the table. Almost all selection controls used were statistically significant. Individuals not married but cohabitating, separated, divorced and single were less likely to be employed in the wage equation estimation than married immigrants. Number of children had no effect, but children under the age of six had a negative effect on participation. The statistic rho was not significant, which implies that there was no selection bias in the OLS estimation since the error terms of the selection and the wage equations were not correlated. However, controlling for participation due to marital status and number of children did change some of the OLS results. Therefore, although there was no selection bias there may have been an omitted variable bias related to education that the selection variables may have been capturing indirectly Another possibility for the difference in the estimated effect of education on wages using the Heckman method may be due to the definition of the education variable. In table A18, I use education as a continuous variable and divide the years of education in years of education abroad and in the United States. The results presented in this table are not very different from the ones presented in Table A15.

145 Results by Gender The results of the Heckman estimation for men and women are presented in Tables 35 and 36, respectively. Sigma, lambda and rho are reported in the bottom of the tables. The Wald-test rejected the value ρ=0 for the female sample, but not for the male sample. The estimated bias for women was estimated to be between and Women who chose to work earned approximately 15% less than if a random sample from the general immigrant population decided to work. In the male sample, where selection does not appear to be a significant issue, the estimated marginal effects did not vary significantly from the estimated coefficients derived using OLS. Of the selection variables used, and estimated simultaneously with the wage equation (Heckman Full Information Model), marital status was not significant, but the total number of children increased participation in the labor force, while the number of children under age six, holding total number of children constant, decreased participation. For the female sample, where there was selection bias, the premium paid for 20 or more years of education over those without education was smaller than previously estimated, but the premium paid to those with 13 to 19 years of education remained unchanged. 47 Experience was not statistically significant, but years in the US was significant. However, the effect disappeared once migrant type was controlled for. 47 In Table A20, I present the results for years of education abroad and in the United States for women. The OLS seems to have underestimated the effect of education in the United States, but no education abroad. All other results are not altered by changing the definition for education.

146 129 Obtaining a degree in the US and English skills were not statistically significant, as seen before in the OLS estimation. On the other hand, although OLS had estimated average years of schooling of country of origin to be positive and statistically significant, adjusting for selection bias showed that the relationship between average years of schooling and women s wages in the US was negative, but not distinguishable from zero. The OLS estimated that immigrant women from Latin America and the Caribbean earned less than women from English-speaking developed countries, but the difference was only marginally significant. Adjusting for selection bias showed that the difference was not distinguishable from zero, although the relationship was still negative. The selection variables, marital status and number of children, were statistically significant. Women who described themselves as cohabitating, separated, divorces and single were more likely to select themselves into the labor market than married immigrant women, while women with children under the age of six were less likely to do so, holding marital status and total number of children constant. E. Transferable Skills and Income Growth 1. Theoretical Background Migration Model 48 The return to immigration depends on wages abroad, wages in the US, direct costs of immigration, and indirect costs of immigration. If the return is 48 This model follows Chiswick (2000) migration model.

147 130 higher than the internal discount rate of individual, immigration occurs. Indirect costs include, but are not limited to, lost wages due to unemployment or underemployment, and investment in new skills; direct costs are the costs of adjustment with schooling, on-the-job training, 49 language, information. In the long run, transferable skills affect wages in the US negatively if the immigrant fails to assimilate or does not invest in American-specific skills. Return to immigration can be defined as (13) W r c u d W c i a where r is return to immigration, W u and Wa are wages in the United States and abroad, respectively, cd are direct costs associated with migration, which not only include transportation but also investment in training, schooling, language and information, among other costs, and ci are indirect costs or forgone wages. Assumption #1 There are two types of immigrants: one with a high level of transferable skills and one with a low level of transferable skills. Immigrants with a high level of skill transferability have lower indirect costs since they are more efficient at migrating since they are able to find better jobs in less time than those immigrants with low levels of skill transferability, who probably need to invest in language or other country-specific skills in order to obtain jobs that fits their overall skill set. 49 Based on Chiswick (2000)

148 131 Assume that W uh, wages of immigrants with high level of skill transferability in the United States, are 100k percentage higher than W ul, wages for immigrants with low level of skill transferability, that is W ( 1 k) W. Further assume that indirect costs are a proportion of wages abroad (t) and that uh ul proportion is lower for high skill transferability individuals, that is, c tw, i a where 0 t t 1. Also assume that direct costs and wages abroad for low and h l high levels of transferable skills are the same, other things, such as gender, education, industry, etc., being equal. Then I can rewrite Equation (13) for individuals with high and low levels of transferable skills: (14a) (14b) r r uh ul (1 k) Wul Wa c t W W c d lu d W l a t W a h a Individuals with high transferable skills have higher returns to immigration than immigrants with low transferable skills, making them more likely to migrate. Assumption #2 Now let us assume that immigrants with high transferable skills also have lower direct costs because they do not have to invest in additional schooling or language. Let be the relationship parameter between low and high direct costs: c ( 1 ) where 0. dh c dl (15) r h (1 k) W (1 ) c dl ul W h a t W a

149 132 Return to migration is higher for immigrants with highly transferable skills when direct costs are not proportional to wages, and therefore they are more likely to migrate than persons with low ranking transferable skills. Assumption #3 So far, I have assumed that the wages abroad are the same for low and high levels of transferability skills, other things held equal. Suppose now that these wages are not the same. Assume that wages for individuals with low transferable skills are 100ka percent higher than for those with high transferable skills, which may be the case for individuals who have skills that are valued in the USA but not in their country of origin, then W ah Wal and equation (13) 1 k ) ( a becomes (16) r h (1 k) W ul (1 ) c Wal (1 k ) dl t W h a a The return to immigration is higher for individuals with high transferable skills when these skills are not compensated for as well abroad, as in the US, making individuals with high transferable skills more likely to migrate. If instead, wages for individuals with high transferable skills are percent higher than for those with low transferable skills, when all other characteristics are equal, then W ( 1 k ) W and equation ah a al 100ka (17) r h (1 k) Wul (1 ka ) Wal (1 ) c t W dl h a

150 133 The return to immigration depends on the ratio of the premium paid over to individuals who have higher transferable skills. If the premium in the USA, k u, is higher than the premium paid abroad, k a, then individuals with higher transferable skills have a higher return to immigration. If, however, the premium paid abroad is higher, then individuals with lower transferable skills have a higher incentive to migrate. 2. Methodology and Results Growth of wages due to immigration is assumed to be determined by individual and country characteristics and affected by the proportion of penalized wages due to immigration. Duleep and Regets (2002) showed that the higher the level of transferable skills the lower the income growth that is experienced by immigrants. They argued that individuals whose transferable skills are high already start with a higher income and therefore their opportunity cost to invest in more country-specific skills is higher than for those individuals that arrive with lower transferable skills. Therefore immigrants with low transferable skills are more likely to invest in country-specific skills and experience faster wage growth than immigrants with high transferable skills. Since we know that the standard controls for the wage equation have the same sign and similar magnitudes in the US and abroad, we expect there to be a positive relationship between wages earned abroad and wages earned in the US by the same individual. The correlation, however, differs by the skills of the immigrant. For example, the linear relationship between wages abroad and wages in the US for immigrants with less than 13 years of education is very flat,

151 134 while the line for those with more than 12 years of education is steeper (see Figure 8). Interestingly, if one were to divide the sample further into men and women, we can see that the slope for men with less than 13 years of education is very similar to the slope for women with 13 years of education or more, while the slope for women who have less than 13 years of education is actually negative (see Figure 9). We can also observe differences in the slope by whether they speak English or not. Figure 10 shows that there was a higher correlation between wages of those who were native English speakers compared to those who spoke English, but were not from an English-speaking country. Those who did not speak English had a flatter slope. Dividing the sample further by gender, one could see that there was actually a negative relationship between wages for men and women who did not speak English. Women who were native English speakers and men who spoke English (non-native) had very similar slopes. Men who were native English speakers were the ones with the steeper positive slope (Figure 11). Graphically, one can see there was a negative relationship between income growth and wages earned abroad. In Figure 12, showing the relationship between income growth and transferable skills divided by level of education, 50 although the slope is clearly negative, there was no difference between those with more than 12 years of education and those with less education. The difference between men and women was not that large either, but women experienced higher income growth than men (Figure 13). 50 Income growth is defined as log of wages in USA log of wages abroad. Transferable skills is defined by β*log of wages abroad, where β is the estimated coefficient in Table 30, column 4b.

152 135 The slope for native English speakers in Figure 14 was flatter than for the other two groups, implying that in order for them to experience the same kind of growth as those who did not speak English or were not native English speakers, they had to have a larger amount of transferable skills or a higher level of wages earned abroad. Between men and women, women who were native English speakers had to earn considerably more (have more transferable skills) abroad in order to have the same income growth as men who did not speak English. The relationship between growth and transferable skills for both men and women was negative for all language skills groups (Figure 15). To study income growth and skills of recent immigrants, consider the equation (18) ln Y lny lny X C, usa abroad abroad where ln Yusa is the log of wages in the United States; Yabroad ln is the log of wages abroad; lny lny usa abroad is how income growth is traditionally defined; 0 1 is the initial devaluation of wages due to migration. Weekly wages abroad are wages reported to the interviewer earned in the last 2 jobs in their country of origin and they are converted to PPP prices to make them comparable. Individuals who reported having a job, but were paid in kind were considered to have had earned zero wages, and therefore considered unemployed. It was only a very small percentage of the sample that fell in this category. Wages reported from abroad may have been earned many years before immigrating or many years before the interview was conducted. I control for the year when the immigrant left his or her last job abroad and how many years passed between the last year from abroad

153 136 and the current job in the US. PPP prices address not only the different cost of living across countries, but across years within the same country. (19) ln Y (ln Y lny X C usa abroad abroad) is initial income growth and this equation is equivalent to (18). The difference lny usa (ln Y lny ) was expected to be positive, since income or quality abroad abroad of life improvement is the most important incentive for individuals to migrate, although there are non-pecuniary reasons for migration as well. In my sample, about 80% of the observations showed a positive difference between these two wages. X is a vector of individual characteristics and C is a vector of country level characteristics. The terms in equations (18) and (19) can be arranged: (20) lny usa lny abroad X, where 1 is the proportion of transferable skills. 0 implied that there are no initial devaluations of wages abroad and therefore the skills were 100% transferable ( 1). These additional transferable skills, isolated from observable skills such as education and experience, reflect unobservable skills, such as industry-specific skills, and ability. Higher wages abroad are an indication of higher skills that an American employer may not be able to observe due to imperfect transferability of skills. a. Full Sample Results

154 137 Table 34, column (1b) 51, shows that the proportion of transferable skills was positive and estimated to be about That is, a 10% increase of income abroad increased income in the US by 0.3%. Using equation 20, it can also be concluded that an increase of 10% in wages abroad decreased wage growth by 9.7%. Income growth also depends on level of education, gender, and total job experience, among other characteristics. The estimates and marginal effects were very similar to wage results from column (1a). Holding transferable skills constant, level of wages in the US was 31.2% higher for immigrants with 13 to 19 years of education, which was only slightly larger than the 30.2% estimated from the wage equation, not controlling for level of wages abroad. Immigrants with more than 20 years of education earned 61.3% more, holding transferable skills constant. Females with the same level of earning abroad earned 38.6% less than men in the United States. Experience and years in the US had an increasing relationship, but at a decreasing rate, on the level of wages in the US, holding wages abroad constant. Black immigrants started at a lower level of wages, while Hispanics started at a higher rate than white immigrants. Immigrants from Eastern Europe and Latin America and the Caribbean earned less than immigrant s from English-speaking developed countries. The difference was at least ten percentage points smaller than before controlling for wages abroad. In column (2), English proficiency and degree in the US are added as dummies. The effect of having more than 12 and 20 years of education decreased 51 Since I have established in the previous section that there is a selection bias in the wage equation, I will discuss the results from the Heckman specification, although the OLS results are presented in Table 31, next to the wage equation results (columns b).

155 138 to 26.1 and 54.7% in column 2b, respectively. Having a degree from the US had no effect on wages in the US. 52 Only those immigrants that did not speak English and did not work in an English-speaking country prior to immigration earned 21.5% less than English native speakers. The estimated transferable skill went down slightly from to and it was still statistically significant. Females earn 39.4% less than men with similar wages abroad. Country-level characteristics (log of GDP per capita, average schooling and unemployment rate) were not statistically significant, however, adding these variables to the specification slightly increased the effect of education on the initial level of wages of immigrants. Column (4) presents results controlled for visa type. As previously discussed, a non-economic-visa migrant was defined as one who immigrated with a family-related or refugee visa to the US, and an economic-visa migrant was one who used an employment preference, diversity or legalization visa instead. Most estimates remained unchanged, but not all. The effect of wages abroad decreased to 0.027, while years in the United States became insignificant. Not speaking English had a penalty of 14%, but the difference was only marginally significant. Degree in the US, which was previously not significant was now (only marginally) significant and the difference was estimated to be 9%. Immigrants from Eastern Europe and Latin America and the Caribbean were still the only groups that earned lower initial wages compared to English-speaking developed countries. 52 That is, unless education is defined as 2 continuous variables in terms of years of education in the United States and abroad. While the return to education abroad is about 4%, the return to education in the United States ranges from 0 to 2 % and having a degree from the United States increases wages by about 14% (see Table A18)

156 139 Non-economic-visa immigrants initially earned almost 24% less than economic-visa migrants; spouses of principals earned 10% less and those who adjusted their immigration status while in the United States earned 16% more. In the second part of Table 34, the estimated coefficients and marginal effects of the selection variables used for this Heckman specification are presented. Interestingly, the estimated marginal effects for marital status were negative and statistically significant (except for the widow category), which was the opposite of the estimated marginal effects found in the labor market participation analysis, where none of the categories were significant except for the widow category which was negative and statistically significant. The number of children under the age of six increased the likelihood of participating in the labor market for immigrants in the US, with total number of children held constant. The term ρ, the correlation of the errors of the selection equation and the wage growth equation, was not statistically significant and the Wald test confirmed that ρ =0. b. Results by Gender To get a clearer picture on how immigrant women s and immigrant men s experience in the US differed, the same Heckman specifications for men and women were estimated separately. Men s proportion of transferable skills was.023 which implies that a 10 percent increase in wages abroad increases wages in the US by 0.23% and that income growth is penalized by 9.77%, while women s proportion of transferable skills was

157 140 In column (1b) of Table 35, one can see that men with 13 to 19 years of education earned 27.3% more than men who do not have any education, holding wages abroad constant, while those with more than 20 years of education earned 55.5% more. The effect of experience on income growth was the same as the one on wages and the same can be said of years in the US. The effect of education decreased, and the difference between white and Hispanic immigrants increased (in favor of Hispanics), when the English variable was added to the specification. Immigrants who did not speak English earned 33% less than native English speakers, but those who did not speak English, but worked in a country whose official language was English, did not suffer a difference in pay compared to native English speakers. This may be a signal that these types of individuals already have some level of assimilation in another English speaking country and therefore are more prepared when arriving and working in the US than someone who does not speak English or is not a native speaker. Although having a degree from the U S is positively related to initial wages, the effect was only marginally significant. Including additional skills to the equation (English skills) weakened the effect of wages abroad, which was now not different from zero. Therefore only observable skills, such as education, experience and English skills, were compensated in the labor force for new immigrant men that have no transferable skills. This implies as well that income growth for men should be faster than income growth for immigrant women, since the coefficient on wages abroad for women was not zero. In column (3), country level characteristics were added. Log of GDP per capita had no effect on income growth, but average schooling had a significant

158 141 and negative effect. Immigrants from countries with lower average education had a faster income growth, holding income abroad constant. In column 4, migrant type characteristics were added to the specification. Non-economic-visa migrants had a slower income growth since, holding constant income earned abroad, they had lower earnings in the US. The estimated marginal effect of green card adjustment was positive and statistically significant. Immigrant men from almost every region, except Western Europe, earned significantly less than those immigrants from English-speaking developed countries. The largest gap can be seen in men from Latin America and the Caribbean followed by men from Eastern Europe. Of the selection variables, marital status had no influence on the decision to participate in the labor market for men. Without controlling for wages abroad, number of children was positive and statistically significant, and number of children under the age of six was negative, but only marginally statistically significant. Adding wage abroad to the equation to basically measure income growth while holding income earned abroad constant, showed that neither marital status, nor number of children had an influence on the participation of men in the labor market. Table 36 shows the Heckman results for the wage and the income growth equations for females. In the income growth equation, the log of wages abroad was added, which was positive and statistically significant for all specifications. For women, a 10% increase in wages abroad increased wages in the US by 0.32 to 0.39%. For the female sample, the Heckman specification showed there was a selectivity issue when analyzing wages and transferable skills (ρ 0).

159 142 Opposite to what was observed in the male sample, adding wage information to the Heckman specification decreased the effect of education for women with more than 12 years of education. Experience remained insignificant while years in the US was still positive and statistically significant. Although having been an immigrant abroad (country of job does not match country of origin or country of birth), had no effect on men, it had a positive effect on women s wages and income growth. This, as well as the larger portion of wages abroad transferred to the US, may be an indication that women are more positively self-selected (or less negatively self-selected) than men. While Hispanic men experienced faster income growth (holding wages abroad constant) than white immigrants, we don t see this effect for women. However, black immigrant women earned lower initial wages suffering therefore from slower initial income growth. For women, neither speaking English, nor having received their college degree, conditional on having a degree, increased their initial wages or initial wage growth. Average schooling in country of origin (column (3)), was positive and statistically significant. The estimated marginal effect decreased from to when wages abroad were included in the specification, indicating that country-level factors did capture some of the immigrants unobserved skills, but not all. Column (4) added the variables related to immigration. Although the results were similar to those for men, the difference between economic and noneconomic-visa migrant women was not as large as those for men (19.4% versus 26.3%). Controlling for immigration type, being the spouse/wife of the principal

160 143 had no effect on initial wages or initial income growth. These specifications also controlled for region of origin. Compared to English-speaking developed countries, only women from Latin America and the Caribbean earned significantly less. The factors that affected labor force participation for immigrant women were marital status most women, except for the widow category, were more likely to be in the labor force than married women. The total number of children was not important, but the number of children under the age of six discouraged them from working. Economic and non-economic-visa migrants. Previous sections demonstrated how wages in the US differ by immigrant type. It was also shown that wages abroad have no effect on men s wages in the US. A deeper analysis of the data, however, yields different results. In Table 37, the wages earned abroad by visa type, economic and noneconomic, were separated. 53 Without changing the rest of the Heckman specifications and the results yielded before, one can see that wages abroad have different effects for immigrants with these two types of visas. For the full sample, where previous results showed an effect of around 0.03, an increase of 10% in wages abroad increased wages in the US for economic-visa migrants by 0.4%, while the effect of wages abroad for non-economic-visa migrants was half the size and only marginally significant. For the male sample, where there was previously no effect of wages abroad on wages in the US, the new results show that although non-economic-visa 53 The dummy variables in the tables are Non-economic Migrants and Economic Migrants.

161 144 migrants are not transferring any skills, as measured by wages abroad, the effect of wages abroad on economic-visa migrants is The effect observed for women was even larger. Non-economic-visa migrant women did not have transfer skills (independent of their education and language skills), but holding education, language skills and other characteristics constant, an increase of 10% of wages abroad increased women s wages in the US by 0.6% if they were economic-visa migrants. V. Conclusions This chapter has described the characteristics and experiences in the US of new immigrants using past wage information from their country of origin, current wage information from the US, and individual and country characteristics. The analysis in the present and previous chapters showed that individual as well as country characteristics affect wages abroad and in the US. The effect of country characteristics abroad on individual wages and labor market participation determine the selection of the individuals who immigrate. In this chapter, I analyze the labor market outcomes in the US of those immigrants, which vary by country as well in the US. While it has been argued before that country-level characteristics alone cannot determine the selectivity of immigrants, the fact remains that even after controlling for immigrant s motivation and transferable skills, there still exists differences among immigrants by country of origin.

162 145 Wages from abroad were introduced as part of the analysis of the economic outcomes of immigrants in the US, and they were interpreted as being related to transferable skills not measured by education and language skills. Wages have two components - skills and a random factor. The difference between wages in the US and wages abroad, independent of observed skills (education, experience, English language skills), is the difference in how and which skills are being compensated independent of the random factor. 54 Additionally, wages abroad measure skills that the employer, but not the econometrician, may see and compensate in the labor market (in the US and abroad) when everything else is being controlled for. Therefore, wages abroad could also be thought of as unobserved characteristics, such as drive and motivation, and the estimated coefficient is the proportion transferred. If wages from abroad was a measure of only unobserved characteristics and not of transferable skills, then we would be assuming that unobservable characteristics are compensated equally abroad and in the US and the difference between the effect of unobservable characteristics on wages abroad and in the US will be equal to zero. I have shown that this difference is not zero. Employed and unemployed immigrants differed in many of their individual characteristics, but the main differences were observed between men and women. For example, Hispanic women earned less than white women, but Hispanic men earned more than white immigrant men. The bias of the OLS wage equation for men was positive, while the bias for women was negative. Men were 54 The random factor is ε which has an expected value of 0.

163 146 more likely to be employed, but those who were unemployed earned a much higher income abroad than those who were employed in the USA. The Heckman specification showed that although there is no selection bias in terms of employment for men, there was selection among women. As was the case for natives, immigrant women s employment also depended on their marital status, number of children and number of children under the age of six years. If the immigrant did not have family or a support network, one could easily see these characteristics were even stronger determinants for employment among recent immigrants than for assimilated immigrants or natives. There are three sets of skills that one would assume are important in the determination of employment and wages for immigrants in the US, whether these were valued abroad or not. These are education, English-language skills and drive or motivation, which by its nature, it not readily observed by the econometrician. 55 Nonetheless, in the present analysis it was found that education and English language skills were not always strong determinants, but that, although small in magnitude, unobserved skills were always statistically significant, particularly for economic-visa migrants. This implies that immigrants that are considered to be economically motivated to immigrate have a higher level of transferable skills. It was also found that, in addition to these skills, country characteristics, race and other characteristics can explain the wages of immigrants in the US, but they were not a good measure or proxy for unobserved drive and ability. 55 Experience and years in the United States do not appear to have effects as large as education, race or country of origin do.

164 147 Motivation to immigrate was captured by the visa category and showed that those individuals that migrated for mainly economic reasons were more likely to be employed and earned higher wages that those who might have been motivated by non-pecuniary reasons. Moreover, the effect of these three skills on employment and wages differed by gender and affected the selectivity of immigrants from the general population in their country of origin. Whereas education, English skills, and country characteristics did not seem to affect men s probability of employment, women s employment appeared to depend on their education and country characteristics (in addition to the country fixed effects). 56 High levels of education (12 years or more) earned both men and women higher wages, but English skills only benefited men, however, there were no differences in the wages of the native and non-native English speakers. Furthermore, it was the case for both men and women that wages from abroad only had an effect on economic-visa migrants US wages -unobserved skills of non-economic-visa migrants were either too low or non-transferable to the American labor market. Additionally, the effect of wages abroad on wages in the US was higher for women than for men. Said effect for women is almost twice the size of the effect for men. Lastly, in this chapter, I presented a migration model which predicts that immigrants with higher transferable skills are more likely to migrate if 56 In results not shown here, income abroad had a negative, very small and only marginally significant effect on men s employment. The rest of the results for the men sample were not affected by the omission of this variable in the discussion. Wages abroad was not statistically significant for the women sample.

165 148 individuals with lower and higher transferable skills earn the same abroad (holding everything else constant); higher transferable skills immigrants earn more in the US than lower transferable skills immigrants; direct costs are either equal or less for higher transferable skills; and indirect costs are proportional to wages and the proportion is lower for higher transferable skills. I have shown that immigrants with economic visas have higher transferable skills (see Table 37), and that the premium on wages abroad earned by economic migrants is higher in the US than abroad, therefore, immigrants with highly transferable skills have a higher return to migration, and therefore they would be more likely to migrate. 57 Without information about the general population in the country of origin on their wages and level of transferable skills, this conclusion cannot be extended to the general population. Since the prediction is that those with higher transferable skills are more likely to migrate, but we still observe that those with low transferable skills migrate to the US, we can further argue that the return to migration is not as important for non-economic visa immigrants as it is to economic-visa immigrants. In conclusion, this chapter adds to the literature on transferable skills of immigrants by using wages from abroad (a unique feature of the New Immigrant Survey) as a measure of transferable skills. I argue that wages abroad are a better measure of transferable skills than previous measures, because in addition to focusing on only one cohort, skills are not limited to education or experience, but 57 Table 9 shows that immigrants with non-economic visa earn 21% less abroad, therefore, the premium paid to economic visa immigrants abroad is 27% ((1/(1-.21)=1.27). Table 34 shows that immigrants with non-economic visa earn 23.8% less in the US, therefore, the premium paid to economic visa immigrants in the US is 31% (1/(1-.238)=1.31).

166 149 it could also include unobserved (transferable) skills, and income growth is due to migration and not to assimilation. While an effort was made to include a rich set of characteristics in the specifications, some results were not fully explained. Even after controlling for education, gender, race, ability or drive, an immigrant s economic motivation, language skills, country-specific characteristics, and other factors, additional factors may be influential. For example, immigrant men from Latin America and the Caribbean were found to earn significantly less than men from other regions of the world. Since education and language skills were held constant, it is possible these differences were due to the quality, and not the transferability, of these skills I explored the possibility that the difference lies on difference in industry or occupation, but the difference between regions remained large and statistically significant.

167 150 CHAPTER 5. CONCLUSIONS AND IMPLICATIONS About three percent of the world s population does not live in the country where they were born, and immigrants tend to move from developing countries to developed countries, mainly to the United States, Canada and Australia. Assimilation of immigrants is a central political issue for these receiving countries. Some fear that immigrants will become an underclass and a burden to the native-born workers, while others argue that immigrants are able to fully assimilate over time and make contributions to the host countries. These immigrants arrive with a set of skills acquired in a different social, cultural and economic context. The proportion of skills that immigrants are able to use to assimilate and to be productive in the US labor market has been the focus of this dissertation. In contrast to previous research which has measured immigrants skills and selectivity using only observable characteristics, such as education, and measured level of assimilation using wages in the United States, this dissertation examines the issue of selectivity and assimilation of immigrants using a unique data set recently available, the New Immigrant Survey, which gathered detailed information on the immigrants situation before entering the United States. This data set also offers a unique opportunity for exploring the skill transferability of immigrants, because it is the only publicly-available and representative dataset that provides wage information of immigrants before they immigrated to the United States. Furthermore, the data set asks participants a variety of questions regarding the visa they acquired to enter the country. This allowed us to separate

168 151 the sample into economic-visa migrants (who entered through employment preferences or other categories associated with economic objectives) and noneconomic-visa migrants (refugees, those who obtained visas obtained through family preferences, etc.). Our analysis found substantial differences in the labor market outcomes and transferability of skills of economic and non-economic-visa migrants. In three chapters, this dissertation examined the issues of selectivity and assimilation of immigrants in the United States, using the New Immigrants Survey. The first chapter analyzed the labor market participation and earnings of immigrants in their country of origin; then the second chapter compared the characteristics and labor market outcomes of immigrants and nonimmigrants from Mexico; and the last chapter focused on the issue of assimilation, selectivity, and the determinants of the labor market performance of immigrants in the US. Men and women, as well as economic and non-economic-visa migrants have different experiences in the labor market abroad and therefore different incentives to immigrate. This explains why we also observe differences in the labor market outcomes between men and women and economic and noneconomic-visa migrants in the US. Abroad, men s experience is better valued than women s, while women s education has a higher return than men s and has a higher effect on employment than does men s education. English language skills also benefit women in the work place but not men. Economic-visa migrant men were slightly more likely to be unemployed in their country of origin than noneconomic-visa migrants, confirming that non-economic-visa migrants move to the United States for reasons other than employment. Non-economic-visa

169 152 migrant women earned less than economic-visa migrants in their country of origin, attesting to the theory that economic-visa migrants are more driven individuals and this drive is a trait valued in the labor market for women. The persistent differences between immigrants and non-migrant Mexican men and women examined in the third chapter leads to the conclusion that immigrant Mexican women, and to some extent men, are positively self-selected in observable (such as education) as well as unobservable (such as ability) characteristics - the employment and wage gaps between immigrants and nonmigrants are mainly due to unobservable characteristics. The last chapter of the dissertation addresses a deficiency and possible bias in previous research by introducing characteristics about immigrants that has been previously omitted. This chapter focused on the issue of assimilation, selectivity, and the determinants of the labor market performance of immigrants in the US, and it showed that education, English language skills, and countrylevel characteristics, which are commonly used as proxies for motivation and ability to examine selectivity and assimilation in the economics of immigration literature do not capture unobserved ability and motivation. Ability, as measured by wages earned abroad, had a higher effect for women than for men. In fact, this effect was only observed among economic-visa migrant, but not among noneconomic-visa migrant men. A. Shortcomings of the analysis

170 153 Although the analysis carried out throughout this dissertation made use of a rich set of characteristics available through the New Immigrants Survey data, the Penn World Table and the Barro-Lee data set, there were some results that cannot be explained by differences in individual characteristics, country of origin, or motivation to immigrate. Men from Latin America and the Caribbean earn significantly less in the US than men from other regions of the world; and this gap cannot be explained by the characteristics included in the analysis (individual characteristics, country-level characteristics, wages from abroad, and country fixed effects). Therefore, this gap should be explained by other unobserved factors not captured by our measures for ability and motivation. I speculate that this gap may be explained not by the amount of transferable skills that these men bring with them, but by the quality of these skills. Additional data would be needed in order to implement an analysis in the quality of the skills from various regions of the world. Current research, including this dissertation (Chapter 3), on selectivity of immigrants compared to non-immigrants in their country of origin has been limited to Mexicans because of the amount of data available in both countries. Although the New Immigrant Survey provides a vast amount of information on immigrants before and after they immigrated, it is not possible to use it to compare immigrants to non-immigrants in their country of origin for many of the represented countries because the sample size for immigrants would be very small.

171 154 B. Future research The results of this dissertation raise additional questions that can be address in future research. A follow-up analysis on income growth and skill transferability to augment will be possible when the second wave of this panel data is released in the near future. Additionally, the most important distinctions revealed by the analysis in the three chapters were between men and women, leading us to conclude that the experiences, selection process and incentives for migration are different for these two groups. The underlying reason could be gender discrimination in the labor market or in other aspects of their societies. Future analysis can reveal if the level of gender discrimination abroad, relative to the US, is linked to the selectivity of immigrant women. Another question raised by the results in Chapter 4 was the wage gap between men from Latin American and the Caribbean and men from other regions. If individual and country characteristics, including ability and motivation, cannot explain this gap, the idea that recent immigrants are of lower quality because the majority comes from Latin America and the Caribbean, including Mexico, is still debatable. Additional data should be included to control for the quality of their characteristics and test whether the gap is due to differences in quality of education or due to discrimination or other factors. C. Policy implications Immigration policy in the US has historically given preference to family of American citizens, while limiting the number of immigrants based on skills, while

172 155 other countries, such as Canada and Australia, seek to attract highly skilled immigrants to support their economies. In order to immigrate to the United States for employment, one would either need to have a American citizen relative in the US, hope to win the diversity lottery, enter or stay illegally in the US, or prove that they possess special skills (or enough capital, in the case of entrepreneurs) for a vacancy that cannot be filled by someone in the US already. The process to obtain an employment visa is long and costly and only a specific number of visas are available each year, which may discourage employers from pursuing the right candidate and the individual may opt for applying to migrate to another country. If the current climate of the immigration debate is a fear that immigrants are becoming an underclass and that they will eventually become a burden for the natives of this country, it is important for policy makers to consider what type of immigrants the current immigration policies is attracting. Although previous research and this dissertation has shown that income growth is faster for those with lower transferable skills, which are also more likely to be non-economic visa immigrants, attracting immigrants with lower transferable skills will put downward pressure on initial wages on all immigrants, when employers start observing that the average skill set of immigrants is low, regardless of their actual individual skill level. Therefore, immigration policy could incentivize non-economic-visa migrants to invest in human capital by giving preferences not only based on the relationships to an American citizen but also based on their level of education and knowledge of the language. Whether the United States should increase the

173 156 number of employment preference visa or streamline its admission process will depend on the actual need for foreign workers to fulfill vacancies that current residents cannot. Offering cost- and time-effective opportunities to immigrants to learn English or to acquire additional education can accelerate assimilation and increase their return on investment.

174 157 REFERENCES Abowd, J. M. & Freeman, R. F. (Eds.). (1991). Immigration, trade and the labor market. Boston, MA: NBER Books, National Bureau of Economic Research. Antecol, H., Cobb-Clark, D. A., & Trejo, S. T. (2003). Immigration policy and the skills of immigrants to Australia, Canada and the United States. Journal of Human Resources, 38(1), Barro, R. & Lee, J. W. (1993). International comparisons of educational attainment. NBER Working Paper No. W4349. Becker, G. S. (1964). Human capital. New York: National Bureau of Economic Research. Bierens, H. J. (2007). Maximum likelihood estimation of Heckman s sample selection model [PDF document]. Retrieved from Lecture Notes Online Web site: HECKMAN.PDF Blau, F. D., & Kahn, L. M. (2000). Gender differences in pay. Journal of Economic Perspectives. 14(4), Blinder, A. S. (1973). Wage discrimination: Reduced form and structural variables. Journal of Human Resources, 8, Borjas, G. J. (1985). Assimilation, changes in cohort quality, and the earnings of immigrants. Journal of Labor Economics, 3(4), Borjas, G. J. (1986). The self-employment experience of immigrants. Journal of Human Resources, 21(4), Borjas, G. J. (1987). Self-selection and the earnings of immigrants. The American Economic Review, 77(4), Borjas, G. J. (1991). National origin and the skills of immigrants in the postwar period. Boston: National Bureau of Economic Research, Inc. Borjas, G. J. (1994). The economics of immigration. Journal of Economic Literature, 32(4),

175 158 Borjas, G. J. (1995). Assimilation and changes in cohort quality revisited: What happened to immigrant earnings in the 1980s? Journal of Labor Economics, 13(2), Borjas, G. J. (1996). The earnings of Mexican immigrants in the United States. Journal of Development Economics, Elsevier, 51(1), Borjas, G., & Katz, L. (2005). The evolution of the Mexican-born workforce in the United States. NBER Working Paper Bound, J., & Johnson, G. (1992). Changes in the structure of wages in the 1980's: An evaluation of alternative explanations. The American Economic Review, 82(3), Burda, M. C. & Hunt, J. (2001). From reunification to economic integration: productivity and the labor market in Eastern Germany. Brooking Papers on Economic Activity, 2, Card, D., & Freeman, R. B. (Eds.). (1993). Small differences that matter: Labor markets and income maintenance in Canada and the United States. Chicago: University of Chicago Press. Chevalier, A., Harmon, C. P., Walker, I., & Zhu, Y. (2003). Does education raise productivity or just reflect it? (CEPR Discussion Paper No ) Chiquiar, D., & Hanson, G. H. (2005). International migration, self-selection, and the distribution of wages: Evidence from Mexico and the United States. Journal of Political Economy, 113(2), Chiswick, B. R. (1978). The effect of Americanization on the earnings of foreignborn men. Journal of Political Economy, 86(5), Chiswick, B. R. (1979). The economic progress of immigrants: Some apparently universal patterns. In W. Fellner (Ed.), Contemporary Economic Problems (pp ). Washington, DC: American Enterprise Institute. Chiswick, B. R. (1982). The employment of immigrants in the United States. Washington, DC: American Enterprise Institute. Chiswick, B. R. (1986). Human capital and the labor market adjustment of immigrants: Testing alternative hypothesis. In O. Stark (Ed.), Research in human capital and development (Vol. 4, pp. 1-26). Greenwich, CT: JAI Press. Chiswick, B. R. (1999). Are immigrants favorably self-selected? The American Economic Review, 89(2),

176 159 Chiswick, B. R. (2000). Are immigrants favorably self-selected? An economics analysis (pp ). In C. D. Brettell & J. F. Hollifeld (Eds.), Migration theory: Talking across the disciplines, New York, NY: Routledge. Chiswick, B. R., & DebBurman, N. (2004). Educational attainment: Analysis by immigrant generation. Economics of Education Review, 23(4), Chiswick, B. R., & Miller, P. W. (2009). The international transferability of immigrants human capital. Economics of Education Review, 28(2), Congressional Budget Office. (2004). A description of the immigrant population. Congress of the United States. Cooke, T., Boyle, P., Couch, K., & Feijten, P. (2009). A longitudinal analysis of family migration and the gender gap in earnings in the United States and Great Britain. Demography, 46(1), Davis, D. R., & Weinstein, D. E. (2002). Technological superiority and the losses from migration. National Bureau of Economic Research Working Paper No Daymont, T. N., & Andrisani, P. J. (1984). Job preferences, college major and the gender gap in earnings. Journal of Human Resources, 19(3), DiNardo, J., Fortin, N. M., & Lemieux, T. (1996). Labor market institutions and the distribution of wages, : A semiparametric approach. Econometrica, 64(5), Donato, K. M. & Massey, D. S. (1993). Effect of the Immigration Reform and Control Act on the wages of Mexican immigrants. Social Science Quarterly, 74, Duleep, H. O., & Regets, M. C. (1996). Earnings convergence: Does it matter where immigrants come from or why? The Canadian Journal of Economics, 29(1), S130-S134. Duleep, H. O., & Regets, M. C. (1997a). Measuring immigrant wage growth using matched CPS files. Demography, 34(2), Duleep, H. O., & Regets, M. C. (1997b). The decline in immigrant entry earnings: Less transferable skills or lower ability? Quarterly Review of Economics and Finance, 37(Special Issue on Immigration), Duleep, H. O., & Regets, M. C. (2002). The elusive concept of immigrant quality: Evidence from (IZA Discussion Paper No. 631). Retrieved March 2008, from

177 160 Fernandez-Huertas Moraga, J. (2011). New evidence on emigrant selection. Review of Economics and Statistics, 93(1), Fix, M., & Passel, J. S. (1994). Immigration and immigrants: Setting the record straight. Washington, DC: The Urban Institute. Folger, J. K. & Nam, C. B. (1969). Education of the American population. A 1960 Census monograph. Prepared in cooperation with the Social Science Research Council. Washington, D. C.: USGPO for US Bureau of the Census, Funkhouser, E., & Trejo, S. J. (1993). The decline in immigrant labor market skills: Did it continue in the 1980's. Santa Barbara, CA: University of California Santa Barbara (Department of Economics). Gallaway, L. E. (1969). Age and labor mobility patterns. Southern Economic Journal, 36(2), Gang, I. N., Rivera-Batiz, F. L. & Yun, M. (2002). Economic strain, ethnic concentration and attitudes towards foreigners in the European Union. IZA Discussion Papers 578, Institute for the Study of Labor (IZA). Gintis, H. (1971). Education, technology and the characteristics of worker productivity. The American Economic Review, 61(2), Gonzalez, A. (2003). The education and wages of immigrant children: The impact of age at arrival. Economics of Education Review, 22(2), Green, W. Modeling consumer decision making and discrete choice behavior [PowerPoint slides]. Retrieved from Lecture Notes Online Web site: pages.stern.nyu.edu/~wgreene/econometrics/econometrics-i-23.ppt - Greenwood, M. J. (1975). Research on internal migration in the United States: A Survey. Journal of Economic Literature, 13(2), Greenwood, M. J., & McDowell, J. M. (1991). Differential economic opportunity, transferability of skills, and immigration to the United States and Canada. The Review of Economics and Statistics, 73(4), Griliches, Z, & Mason, M. (1972). Education, income and ability. Journal of Political Economy, 51(4), Haley, M. R. & Taengnoi, S. (2011). The skill transferability of high-skilled US immigrants. Applied Economics Letters, 18(7),

178 161 Hamilton, C. H. (1965). Practical and mathematical considerations in the formulation and selection of migration rates. Demography, 2, Hanson, G. H., & Spilimbergo, A. (1999). Illegal immigration, border enforcement, and relative wages: Evidence from apprehensions at the U.S. Mexico border. The American Economic Review, 89, Hause, J. C. (1972). Earnings profile: Ability and schooling. Journal of Political Economy, 80(3), S108-S138. Hausmann, R., Tyson, L., & Zahidi, S. (2008). The global gender gap report. World Economic Forum. Heckman, J. J. (1979). Sample selection bias as a specification error. Econometrica, 47, Hunt, J. (2001). Post-unification wage growth in East Germany. Review of Economics and Statistics, 83(1), Hunt, J. (2002). The transition in East Germany: When is a ten-point fall in the gender wage gap bad news? Journal of Labor Economics, 20(1), Hunt, J. (2006). Staunching emigration from East Germany: Age and the determinants of migration. Journal of the European Economic Association, MIT Press, 4(5), Jasso, G., & Rosenzweig, M. R. (1988). How well do U. S. immigrants do? Vintage effects, emigration selectivity and occupational mobility. Research in Population Economics, 6, Jasso, G., & Rosenzweig, M. R. (1990). Self-selection and the earnings of immigrants: Comment. American Economic Review, 80(1), Jasso, G., Rosenzweig, M. R., & Smith, J. P. (2002). The Earnings of U.S. Immigrants: World Skill Prices, Skill Transferability and Selectivity. Jones, F. L., & Kelley, J. (1984). Decomposing differences between groups: A cautionary note on measuring discrimination. Sociological Methods and Research, 12, Kaestner, R., & Malamud, O. (2010). Self-selection and international migration: New evidence from Mexico. NBER Working Papers Kao, G., & Tienda, M. (1995). Optimism and Achievement: The Educational Performance of Immigrant Youth. Social Science Quarterly, 76(1), 1-19.

179 162 Katz, L. F., & Murphy, K. M. (1992). Changes in relative wages, : Supply and demand factors. Quarterly Journal of Economics, 107(1), Katz, E., & Stark, O. (1984). Migration and asymmetric information: comment. The American Economic Review, 74(3), Katz, E., & Stark, O. (1987). International migration under asymmetric information. The Economic Journal, 97(387), Kochhar, R. (2005). Survey of Mexican migrants, part three: The economic transition to America. Washington, DC: Pew Hispanic Center. Kossoudji, S. A., & Cobb-Clark, D. A. (1998). Coming out of the shadows: learning about legal status and wages from the legalized population: Department of Economics and School of Social Work, University of Michigan. Kossoudji, S. A., & Ranney, S. I. (1984). Wages rates of temporary Mexican migrants to the U. S.: The role of legal status. Paper presented at the Econometric Society Annual Meetings. Dallas, TX. LaLonde, R. J., & Topel, R. H. (Eds.). (1992). The assimilation of immigrants in the U.S. labor market (67-92). In G.J. Borjas, & R. B. Freeman (Eds.), Immigration and the workforce. Chicago: University of Chicago. Lansing, J. B., & Morgan, J. N. (1967). The effect of geographical mobility on income. The Journal of Human Resources, 2(4), Lazear, E. (1977). Academic Achievement and Job Performance: A Note. The American Economic Review, 67(2), Lofstrom, M. (2002). Labor market assimilation and the self-employment decision of immigrant entrepreneurs. Journal of Population Economics, 15(1), Lozano, F., & Lopez, M. (2007). Immigrant selection and the gender wage gap of Mexican immigrants in the United States. Paper presented at the Western Economic Association 2007 Annual Meeting in July. Seattle, WA. Massey, D. S. (1987). Understanding Mexican migration to the United States. American Journal of Sociology, 92, Masters, S. H. (1972). Are black immigrants from the South to the Northern cities worse off than blacks already there? The Journal of Human Resources, 7(4),

180 163 McKenzie, D., & Rapoport, H. (2007). Network effects and the dynamics of migration and inequality: Theory and evidence from Mexico. Journal of Development Economics. Elsevier, 84(1) McKenzie, D., & Rapoport, H. (2010). Self-selection patterns in Mexico-U. S. migration: The role of migration networks. Review of Economics and Statistics, 92(4), Mincer, J. (1974). Schooling, experience, and earnings. New York: National Bureau of Economic Research. Mincer, J. (1978). Family migration decisions. Journal of Political Economy, 86(5), Murphy, K. M., & Welch, F. (1992). The structure of wages. The Quarterly Journal of Economics, 107(1), Oaxaca, R. (1973). Male-female wage differentials in urban labor markets. International Economic Review. 14(October), O Donnell, O., van Doorslaer, E., Wagstaff, A. & Lindelow, M. (2008). Analyzing health equity using household survey data. A guide to techniques and their implementation. Washington, DC: The World Bank. Orrenius, P. M. & Zavodny, M. (2005). Self-selection among undocumented immigrants from Mexico. Journal of Development Economics, Elsevier, 78(1), Passel, J. S. (2005a). Unauthorized migrants: Size and characteristics. Washington, D. C.: Pew Hispanic Center. Passel, J. S. (2005b). Estimates of the size and characteristics of the undocumented population. Washington, D. C.: Pew Hispanic Center. Passel, J. S. & Cohn, D. (2008). Trends in unauthorized immigration: undocumented inflow now trails legal flow. Washington, D. C.: Pew Hispanic Center. Rivera-Batiz, F. L. (1999). Undocumented workers in the labor market: An analysis of the earnings of legal and illegal Mexican immigrants in the United States. Journal of Population Economics, 12(1), Rivera- Batiz, F. L. (2007). How do immigrants from Latin America and the Caribbean fare in the U.S. labor market? The World Economy, 30(9),

181 164 Roy, A. D. (1951). Some thoughts on the distribution of earnings. Oxford Economics Papers, 3, Schoeni, R. F., McCarthy, K. M., & Vernez, G. (1996). Pursuing the American dream: Economic progress of immigrant men in California and the nation. Santa Monica, CA: Rand Corp. Schultz, P. T. (1998). Immigrant quality and assimilation: A review of the U.S. literature. Journal of Population Economics, 11(2), Schultz, T. W. (1961). Investment in human capital. The American Economic Review, 51(1), Schwartz, A. (1973). Interpreting the effect of distance on migration. The Journal of Political Economy, 81(5), Sjaastad, L. A. (1962). The costs and returns of human migration. The Journal of Political Economy, 70(5), Tolnay, S. E. (1998). Educational selection in the migration of southern blacks, Social Forces, 77(3), Welch, F. (1974). Relationships between income and schooling. Review of Research in Education, 2, Wertheimer, R. F. III. (1970). The monetary rewards of migration within the U. S. Washington, D.C.: The Urban Institute. Winsborough, H. H., & Dickinson, P. (1971). Components of negro-white income differences. Proceedings of the American Statistical Association. Social Statistics Section: 6-8. Yuengert, A. M. (1995). Testing the hypothesis of immigrant self-employment. Journal of Human Resources, 30(1), Yun, M. (2000). Selection bias and the decomposition of wage differentials. Departmental Working Papers , Rutgers University, Department of Economics. Yun, M. (2005). A simple solution to the identification problem in detailed wage decompositions. Economic Inquiry, 43, Zaiceva, A. (2010). East-West migration and gender: is there a differential effect for migrant women? Labour Economics, 17,

182 FIGURES 165

183 166 Figure 1. Probability of Employment.01 Probability of Positive Outcome Years of Experience Abroad Experience Abroad

184 167 Figure 2. Experience and the Probability of Employment Abroad.01 Probability of Positive Outcome Experience Abroad Years of Experience Abroad Females Males

185 Figure 3. Gender Gap in Labor Market Participation and Wages 168 Gender Gap Index 100 India Guatemala Mexico El Salvador Korea Nigeria Dominican Republic Cuba Philippines Peru Ukraine Jamaica Poland Ethiopia Russia Colombia United States China Canada United Kingdom Vietnam Lower end is the Wage Gap. Upper end is the Participation Gap Source: The Global Gender Gap Report By Ricardo Hausmann, Laura D'Andrea Tyson, Saadia Zahidi

186 Figure 4. Number of Mexican immigrants that have become Legal Permanent Residents by decade Number of Immigrants from Mexico Year of Arrival Immigrants from Mexico Percentage from Mexico Source: Department of Homeland Security

187 Figure 5. Employment in the United States by Visa Category percent Employed Unemployed Legalization Refugee/Asylee/Parolee Diversity Immigrants Family Fourth Preference Employment Preferences Child of U.S. Citizen Spouse of Legal Permanent Resident Parent of U.S. Citizen Spouse of U.S. Citizen Other

188 Figure 6. Transferability of Skills By Country 171 Jamaica Cuba Vietnam Nigeria Poland Haiti* India*** United Kingdom*** Guatemala** Canada** China Colombia* Korea PeruEl Salvador Ukraine Philippines Russia Mexico* Ethiopia Dominican Republic Country of Origin * significant at 10%; ** significant at the 5% level; *** significant at the 1% level

189 172 Men India*** United Kingdom*** Canada** Colombia** Guatemala China Ethiopia Russia Mexico** El Salvador Korea Ukraine Peru Poland Philippines Dominican Cuba Republic Nigeria Vietnam Jamaica Haiti* Country of Origin Women United Kingdom** India*** Canada China Philippines Ethiopia Peru Nigeria Ukraine Colombia Korea Mexico Jamaica Russia Cuba Guatemala El Salvador Dominican Republic Poland*** Vietnam Haiti* Country of Origin Figure 7. Transferability of Skills By Country and Gender * significant at 10%; ** significant at the 5% level; *** significant at the 1% level

190 Figure 8. Wages in the USA and Wages Abroad By Education Weekly (log) wages Abroad 12 years of education or less More than 12 years of education

191 Weekly (log) wages Abroad Men with 12 years of education or less Men with more than 12 years of education Women with 12 years of education or less Women with more than 12 years of education Figure 9. Wages in the USA and Wages Abroad By Education and Gender

192 Figure 10. Wages in the USA and Wages Abroad By Language Spoken Weekly (log) wages Abroad Speaks English - Native Speaks English - Non Native Does Not Speak English

193 Figure 11. Wages in the USA and Wages Abroad By Language Spoken and Gender Weekly (log) wages Abroad Speaks English - Native (Men) Speaks English - Non Native (Men) Does Not Speak English (Men) Speaks English - Native (Women) Speaks English - Non Native (Women) Does Not Speak English (Women)

194 Figure 12. Income Growth and Income Level Abroad By Education Transferable Skills 12 years of education or less 13 years of education or more Income growth

195 Figure 13. Income Growth and Income Level Abroad By Education and Gender Transferable Skills 12 years of education or less - Men 13 years of education or more - Men 12 years of education or less - Women 13 years of education or more - Women

196 Figure 14. Income Growth and Income Abroad By Language Spoken Transferable Skills Speaks English - Native Speaks English - Non Native Does not Speak English

197 Figure 15. Income Growth and Income Abroad By Language Spoken and Gender Transferable Skills Speaks English (Native) - Men Speaks English (Non Native) - Men Does not speak English - Men Speaks English (Native) - Women Speaks English (Non Native) - Women Does not speak English - Women

198 TABLES 181

199 Table 1. Summary Statistics by Employment Abroad 182

200 Table 2. Summary Statistics by Employment Abroad and Gender 183

201 Table2. Summary Statistics by Employment Abroad and Gender 184

202 Table 3. Probability of Employment in Country of Origin (Probit) 185

203 Table 3. Probability of Employment in Country of Origin (Probit) (Continued) 186

204 Table 4. Men s Probability of Employment in Country of Origin (Probit) 187

205 Table 4. Men s Probability of Employment in Country of Origin (Probit) (Continued) 188

206 Table 5. Women s Probability of Employment in Country of Origin (Probit) 189

207 Table 5. Women s Probability of Employment in Country of Origin (Probit) (Continued) 190

208 Table 6. Rate of Return to Education of New Immigrants in Country of Origin (OLS) 191

209 Table 7. Rate of Return to Education of New Immigrant Men in Country of Origin (OLS) 192

210 Table 8. Rate of Return to Education of New Immigrant Women in Country of Origin (OLS) 193

211 Table 9. Rate of Return to Education of New Immigrants in Country of Origin (Heckman) 194

212 Table 9. Rate of Return to Education of New Immigrants in Country of Origin (Heckman) (Continued) 195

213 Table 10. Rate of Return to Education of New Immigrant Men in Country of Origin (Heckman) 196

214 Table 10. Rate of Return to Education of New Immigrant Men in Country of Origin (Heckman) (Continued) 197

215 Table 11. Rate of Return to Education of New Immigrant Women in Country of Origin (Heckman) 198

216 Table 11. Rate of Return to Education of New Immigrant Women in Country of Origin (Heckman) (Continued) 199

217 Table 12. Summary Statistics for Immigrant and Non-Migrant Mexicans 200

218 Table 13. Summary Statistics for Immigrants and Non-Migrants from Mexico by Gender 201

219 202 Table 14. Employment Probability in Mexico of Immigrants and Non-Migrants Non- Migrants Immigrants Marginal Marginal Coefficients Effects Coefficients Effects Years of Education (0.003)*** (0.001)*** (0.019)*** (0.005)*** Experience (0.003)*** (0.001)*** (0.018)*** (0.005)*** Experience - Squared (0.000)*** (0.000)*** (0.000)*** (0.000)*** Female (0.023)*** (0.005)*** (0.163) (0.044) Marital Status Living together (0.035)*** (0.010)*** Separated (0.062)*** (0.017)*** Divorced (0.122)*** (0.034)*** (0.725) (0.194) Widowed (0.056)*** (0.015)*** (0.603)** (0.159)** Single (0.033)** (0.009)** (0.208) (0.055) Number of Children (0.015) (0.004) (0.048) (0.013) Number of Children Under Six (0.024)*** (0.007)*** (0.152) (0.041) Constant (0.058)*** (0.387)*** Observations 17, Pseudo R-Squared Log-Likelihood Chi-Squared p-value (0.000) (0.000) Number Employed 11, Employed in Mexico 67% 35% Note: Dependent variable is employment status in Mexico. Standard errors are in parenthesis. Additional controls include year effects. Left out dummy variable for marital status is married. * significant at 10%; ** significant at the 5% level; *** significant at the 1% level.

220 Table 15. Employment Probability in Mexico of Immigrant and Non-Migrant Men 203

221 204 Table 16. Employment Probability in Mexico of Immigrant and Non-Migrant Women Non- Migrants Immigrants Marginal Marginal Coefficients Effects Coefficients Effects Years of Education (0.004)*** (0.001)*** (0.025)** (0.005)** Experience (0.003)*** (0.001)*** (0.025)* (0.005)* Experience - Squared (0.000)*** (0.000)*** (0.001)** (0.000)** Marital Status Living together (0.048)** (0.016)** Separated (0.066)*** (0.021)*** Divorced (0.131)*** (0.045)*** Widowed (0.061)*** (0.020)*** (0.571)*** (0.138)*** Single (0.041)*** (0.014)*** (0.283)* (0.063)* Number of Children (0.019) (0.006) (0.055)** (0.012)** Number of Children Under Six (0.031) (0.011) (0.271)** (0.059)** Constant (0.072)*** (0.494)*** Observations 9, Pseudo R-Squared Log-Likelihood Chi-Squared p-value (0.000) (0.000) Number Employed 4, Employed in Mexico 50% 31% Note: Dependent variable is employment status in Mexico. Standard errors are in parenthesis. Additional controls include year effects. Left out dummy variable for marital status is married. * significant at 10%; ** significant at the 5% level; *** significant at the 1% level.

222 Table 17. Weekly Wages in Mexico of Immigrants and Non-Migrants 205

223 Table 18. Weekly Wages in Mexico of Immigrant and Non-Migrant Men 206

224 Table 19. Weekly Wages of Immigrant and Non-Migrant Women 207

225 Table 20. Oaxaca Decomposition of Employment Status in Mexico Full Sample 208

226 Table 21. Oaxaca Decomposition of Employment Status in Mexico - Men Sample 209

227 210 Table 22. Oaxaca Decomposition of Employment Status in Mexico Women Sample Overall Endowments Coefficients Interaction Group 1 (Non-Immigrants) (0.006)*** Group 2 (Immigrants) (0.036)*** Difference (0.037)*** Difference Explained by (0.040)*** (0.034)*** (0.038)*** Proportion 60% 80% -40% Years of Education (0.006)** (0.031) (0.004) Experience (0.017) (0.071) (0.011) Experience - Squared (0.015)* (0.056) (0.009) Marital Status Living together (0.002)*** (0.000) (0.003)*** Separated (0.001)*** (0.000) (0.000)*** Divorced (0.000)*** (0.000) (0.000)*** Widowed (0.002)*** (0.000)** (0.001)*** Single (0.012)*** (0.040)*** (0.014)*** Number of Children (0.005)* (0.012) (0.003) Number of Children Under Six (0.005)* (0.006)** (0.003)* Constant (0.085)*** Observations 9,827 9,827 9,827 Number Employed 4,659 Employed in Mexico 48% Note: Dependent variable is employment status in Mexico. Standard errors are in parenthesis. Additional controls include year effects. Left out dummy variable for marital status is married. In the Blinder-Oaxaca decompostion, the group of comparison is Non-immigrants. * significant at 10%; ** significant at the 5% level; *** significant at the 1% level.

228 Table 23. Oaxaca Decomposition of Weekly Wages in Mexico Full Sample 211

229 Table 24. Oaxaca Decomposition of Weekly Wages in Mexico Men Sample 212

230 Table 25. Oaxaca Decomposition of Weekly Wages in Mexico Women Sample 213

231 Table 26. Summary Statistics by Employment in the United States 214

232 Table 26. Summary Statistics by Employment in the United States (Continued) 215

233 Table 27. Summary Statistics by Employment in the United States and by Gender 216

234 Table 27. Summary Statistics by Employment in the United States and By Gender (Continued) 217

235 Table 28. Probability of Employment in the United States (Probit Regression) 218

236 Table 28. Probability of Employment in the United States (Probit Regression) (Continued) 219

237 Table 28. Probability of Employment in the United States (Probit Regression) (Continued) 220

238 Table 29. Men s Probability of Employment in the United States (Probit Regression) 221

239 Table 29. Men s Probability of Employment in the United States (Probit Regression) (Continued) 222

240 Table 29. Men s Probability of Employment in the United States (Probit Regression) (Continued) 223

241 Table 30. Women s Probability of Employment in the United States (Probit Regression) 224

242 Table 30. Women s Probability of Employment in the United States (Probit Regression) (Continued) 225

243 Table 30. Women s Probability of Employment in the United States (Probit Regression) (Continued) 226

244 Table 31. New Immigrants Return to Education and Transferable Skills (Dependent Variable: Wages in the United States) 227

245 Table 31. New Immigrants Return to Education and Transferable Skills (Dependent Variable: Wages in the United States) (Continued) 228

246 Table 31. New Immigrants Return to Education and Transferable Skills (Dependent Variable: Wages in the United States) (Continued) 229

247 Table 32. New Immigrant Men s Return to Education and Transferable Skills (Dependent Variable: Wages in the United States) 230

248 Table 32. New immigrant immigrant Men s Return to Education and Transferable Skills (Dependent Variable: Wages in the United States) (Continued) 231

249 Table 33. New immigrant Women s Return to Education and Transferable Skills (Dependent Variable: Wages in the United States) 232

250 Table 33. New immigrant Women s Return to Education and Transferable Skills (Dependent Variable: Wages in the United States) (Continued) 233

251 Table 33. New immigrant Women s Return to Education and Transferable Skills (Dependent Variable: Wages in the United States) (Continued) 234

252 Table 34. New Immigrants Return to Education and Transferable Skills Heckman Estimation (Dependent Variable: Log of Weekly Wages in the United States) 235

253 Table 34. New Immigrants Return to Education and Transferable Skills Heckman Estimation (Dependent Variable: Log of Weekly Wages in the United States) (Continued) 236

254 Table 34. New Immigrants Return to Education and Transferable Skills Heckman Estimation (Dependent Variable: Log of Weekly Wages in the United States) (Continued) 237

255 Table 35. New Immigrant Men s Return to Education and Transferable Skills Heckman Estimation (Dependent Variable: Log of Weekly Wages in the United States) 238

256 Table 35. New Immigrant Men s Return to Education and Transferable Skills Heckman Estimation (Dependent Variable: Log of Weekly Wages in the United States) (Continued) 239

257 Table 35. New Immigrant Men s Return to Education and Transferable Skills Heckman Estimation (Dependent Variable: Log of Weekly Wages in the United States) (Continued) 240

258 Table 36. New Immigrant Women s Return to Education and Transferable Skills Heckman Estimation (Dependent Variable: Log of Weekly Wages in the United States) 241

259 Table 36. New Immigrant Women s Return to Education and Transferable Skills Heckman Estimation (Dependent Variable: Log of Weekly Wages in the United States) (Continued) 242

260 Table 36. New Immigrant Women s Return to Education and Transferable Skills Heckman Estimation (Dependent Variable: Log of Weekly Wages in the United States) (Continued) 243

261 Full Sample Men Women Table 37. New Immigrants Return to Education and Transferable Skills Heckman Estimation by Immigrant Type (Dependent Variable: Log of Weekly Wages in the United States) 244

262 APPENDIX A ADDITIONAL TABLES 245

263 246 Table A1. Summary Statistics by Country of Origin All Countries Canada China Colombia Cuba Dominican Republic El Salvador Number of Observations Female 56.4% 59.6% 59.7% 64.1% 47.6% 56.8% 51.2% Log of wages abroad (0.048) (0.225) (0.154) (0.439) (0.292) (0.155) (0.146) Wages abroad 73, , (62, ) (1, ) ( ) ( ) (91.142) (48.0) (21.640) Log of wages abroad - no outliers (0.046) (0.225) (0.126) (0.439) (0.292) (0.155) (0.146) Wages abroad - no outliers 2, , ( ) (1, ) ( ) ( ) (91.142) (48.0) (21.640) Log of wages in USA (0.015) (0.121) (0.074) (0.082) (0.109) (0.047) (0.042) Wages in USA (47.617) (67.389) ( ) (26.757) (80.073) (13.560) (50.121) Employed abroad 38.1% 63.8% 57.4% 32.6% 19.1% 36.4% 15.2% Employed in USA 54.0% 58.2% 43.2% 58.2% 65.3% 52.5% 75.1% Age (0.172) (1.201) (0.851) (1.373) (0.967) (1.208) (0.501) Years of education in USA (0.031) (0.297) (0.096) (0.118) (0.123) (0.059) (0.169) Years of education abroad (0.062) (0.442) (0.216) (0.391) (0.321) (0.383) (0.225) Country's Average Schooling n/a 4.44 n/a (0.060) (0.276) (0.036) (0.039) (0.031) Student abroad 17.2% 21.8% 16.6% 14.6% 8.8% 11.7% 22.5% Student in USA 10.1% 11.3% 8.3% 14.1% 8.6% 3.6% 8.5% English Speaks English (Native) 8.3% 71.7% 0.4% 0.0% 0.6% 0.0% 0.9% Speaks English (Not native) 23.3% 7.1% 11.2% 20.0% 6.6% 4.7% 11.7% Does not speak English 46.6% 2.7% 75.8% 63.6% 74.2% 52.5% 52.6% Immigration Variables Years in USA (0.090) (1.041) (0.256) (0.567) (0.423) (0.474) (0.278) Age left country of origin (0.195) (1.069) (0.881) (1.386) (1.060) (1.261) (0.514) Years since last job abroad (0.183) (0.725) (0.783) (1.490) (1.036) (1.490) (0.656) Visa Category Spouse of US Citizen 34.2% 52.9% 23.3% 56.8% 6.6% 22.1% 11.0% Spouse of LPR 2.4% 0.0% 2.3% 0.7% 1.4% 6.9% 1.1% Parent of U.S. Citizen 11.9% 3.6% 23.7% 19.5% 2.0% 17.3% 2.7% Child of U.S. Citizen 3.4% 2.5% 3.1% 4.1% 0.8% 6.4% 1.5% Family Fourth Preference 6.4% 3.2% 15.7% 5.0% 0.0% 7.3% 0.8% Employment Preferences 9.6% 27.5% 18.4% 4.0% 0.7% 0.4% 0.9% Diversity Immigrants 8.1% 2.8% 0.0% 0.0% 2.6% 0.0% 0.0% Refugee/Asylee/Parolee 6.6% 0.0% 3.0% 1.1% 67.7% 0.0% 0.0% Legalization 8.0% 0.0% 0.0% 0.6% 0.7% 0.6% 77.4% Other 9.4% 7.6% 10.5% 8.3% 17.5% 39.0% 4.6% Green Card adjustment 57.4% 73.4% 38.7% 55.3% 73.0% 16.9% 91.2% Spouse of principal 11.4% 14.1% 18.5% 4.5% 13.3% 1.6% 0.8% Note: Standard deviations are in parenthesis. Some percentages do not add to 100 because of missing information. Countries left out of this table are those countries that were grouped together by region.

264 247 Table A1. Summary Statistics by Country of Origin (Continued) Ethiopia Guatemala Haiti India Jamaica Korea Mexico Number of Observations Female 45.1% 50.9% 57.6% 57.7% 50.7% 62.7% 60.1% Log of wages abroad (0.126) (0.159) (0.273) (0.097) (0.212) (0.267) (0.205) Wages abroad , , ( ) (32.897) (62.327) (5, ) ( ) (1, ) (97.051) Log of wages abroad - no outliers (0.126) (0.159) (0.273) (0.097) (0.212) (0.267) (0.205) Wages abroad - no outliers , , ( ) (32.897) (62.327) (5, ) ( ) (1, ) (97.051) Log of wages in USA (0.114) (0.059) (0.114) (0.063) (0.218) (0.093) (0.038) Wages in USA ( ) (24.382) ( ) (153.0) (44.143) (58.360) ( ) Employed abroad 46.3% 22.4% 19.9% 36.0% 42.1% 53.0% 21.8% Employed in USA 48.8% 77.5% 40.4% 46.6% 49.6% 50.4% 52.6% Age (1.121) (0.915) (1.580) (0.658) (1.138) (1.323) (0.448) Years of education in USA (0.10) (0.230) (0.318) (0.042) (0.269) (0.222) (0.103) Years of education abroad (0.323) (0.366) (0.497) (0.206) (0.353) (0.378) (0.155) Country's Average Schooling n/a (0.042) (0.054) (0.018) (0.019) (0.048) (0.024) Student abroad 8.1% 11.8% 18.4% 12.0% 23.6% 24.9% 21.1% Student in USA 11.3% 5.2% 11.3% 6.2% 15.1% 3.3% 8.9% English Speaks English (Native) 0.0% 0.0% 1.5% 44.3% 58.0% 0.0% 2.3% Speaks English (Not native) 15.0% 15.4% 21.3% 0.1% 0.0% 20.1% 23.3% Does not speak English 42.7% 55.0% 26.9% 26.0% 0.0% 74.4% 59.6% Immigration Variables Years in USA (0.480) (0.471) (0.630) (0.207) (0.699) (0.463) (0.262) Age left country of origin (1.045) (0.986) (1.628) (0.707) (1.332) (1.345) (0.511) Years since last job abroad (1.231) (1.054) (1.984) (0.760) (1.057) (1.397) (0.598) Visa Category Spouse of US Citizen 9.8% 16.5% 22.1% 16.6% 41.0% 40.1% 46.8% Spouse of LPR 0.0% 2.5% 3.4% 1.0% 1.6% 0.8% 8.7% Parent of U.S. Citizen 4.8% 4.9% 21.2% 13.8% 9.3% 10.1% 18.4% Child of U.S. Citizen 2.2% 2.5% 5.7% 0.2% 13.3% 1.3% 5.4% Family Fourth Preference 1.0% 1.9% 2.8% 21.0% 2.2% 5.4% 2.4% Employment Preferences 2.0% 1.5% 1.5% 36.7% 4.4% 36.6% 2.4% Diversity Immigrants 61.9% 0.0% 0.0% 0.4% 0.0% 0.0% 0.0% Refugee/Asylee/Parolee 17.3% 4.2% 20.8% 3.3% 0.0% 0.0% 0.1% Legalization 0.0% 59.5% 0.0% 0.0% 0.0% 0.0% 6.9% Other 1.0% 6.4% 22.6% 7.0% 28.4% 5.7% 9.0% Green Card adjustment 26.6% 89.5% 46.3% 47.2% 47.2% 73.4% 76.4% Spouse of principal 15.3% 1.0% 4.0% 28.4% 2.7% 24.3% 2.3%

265 248 Table A1. Summary Statistics by Country of Origin (Continued) Nigeria Peru Philippines Poland Russia Ukraine United Kingdom Vietnam Number of Observations Female 49.4% 65.9% 66.6% 48.6% 63.8% 52.1% 23.8% 66.0% Log of wages abroad (0.236) (0.291) (0.099) (0.296) (0.323) (0.535) (0.145) (0.258) Wages abroad , , , , (57.467) ( ) (920.0) (3, ) (2, ) (3, ) ( ) (67.938) Log of wages abroad - no outliers (0.236) (0.276) (0.099) (0.295) (0.323) (0.497) (0.145) (0.249) Wages abroad - no outliers , , , , (57.467) ( ) (920.0) (3, ) (2, ) (3, ) ( ) (68.512) Log of wages in USA (0.189) (0.099) (0.065) (0.076) (0.091) (0.094) (0.113) (0.081) Wages in USA 1, , , ( ) (37.086) ( ) (77.757) (71.534) (31.563) ( ) (21.817) Employed abroad 43.2% 36.4% 46.3% 59.3% 64.4% 53.2% 61.3% 32.4% Employed in USA 46.3% 69.6% 51.7% 56.9% 58.0% 55.1% 66.3% 27.3% Age (1.327) (1.471) (0.818) (1.382) (1.521) (1.358) (1.423) (0.920) Years of education in USA (0.222) (0.354) (0.064) (0.051) (0.262) (0.130) (0.165) (0.091) Years of education abroad (0.431) (0.505) (0.186) (0.254) (0.360) (0.221) (0.376) (0.311) Country's Average Schooling n/a n/a n/a 7.92 n/a (0.028) (0.037) (0.019) (0.240) Student abroad 19.4% 18.5% 6.4% 15.6% 14.8% 16.9% 16.1% 4.7% Student in USA 24.4% 12.0% 4.7% 10.4% 15.4% 12.2% 2.1% 8.6% English Speaks English (Native) 55.8% 2.1% 0.0% 0.0% 0.0% 0.0% 82.7% 0.0% Speaks English (Not native) 0.0% 26.1% 44.7% 14.4% 26.6% 14.8% 5.3% 9.4% Does not speak English 5.2% 51.4% 33.2% 68.5% 60.0% 70.1% 5.2% 71.9% Immigration Variables Years in USA (0.773) (0.654) (0.273) (0.592) (0.406) (0.339) (0.484) (0.213) Age left country of origin (1.507) (1.604) (0.872) (1.257) (1.609) (1.518) (1.708) (0.956) Years since last job abroad (0.939) (1.280) (0.815) (0.873) (0.725) (0.968) (1.251) (1.157) Visa Category Spouse of US Citizen 24.6% 54.5% 29.3% 20.9% 32.9% 16.3% 53.0% 32.8% Spouse of LPR 1.4% 0.0% 1.0% 0.0% 0.0% 0.0% 0.0% 0.5% Parent of U.S. Citizen 14.7% 9.7% 17.7% 9.2% 6.0% 6.1% 6.7% 11.1% Child of U.S. Citizen 3.8% 3.0% 5.4% 2.2% 2.2% 0.0% 0.0% 2.0% Family Fourth Preference 2.2% 4.2% 9.2% 5.5% 0.0% 0.0% 7.8% 35.8% Employment Preferences 2.6% 4.0% 19.4% 6.3% 4.1% 1.4% 26.4% 0.3% Diversity Immigrants 39.5% 10.9% 0.0% 40.3% 16.4% 33.7% 5.5% 0.0% Refugee/Asylee/Parolee 7.6% 2.9% 0.0% 0.0% 37.6% 42.4% 0.0% 5.4% Legalization 1.4% 0.8% 0.3% 0.0% 0.0% 0.0% 0.0% 0.0% Other 2.1% 10.0% 17.7% 15.6% 0.8% 0.0% 0.6% 12.1% Green Card adjustment 37.3% 68.3% 30.2% 36.2% 83.0% 63.0% 58.9% 35.8% Spouse of principal 14.8% 6.2% 12.2% 19.9% 16.0% 21.6% 16.0% 18.1%

266 249 Table A2. Top 15 Languages Spoken at Home Total Observations 5252 Spanish 1,919 English 720 Other Spoken in India 607 Chinese 482 Russian 266 Tagalog 250 Polish 151 Vietnamese 147 Arabic 146 Hindi 141 Korean 128 Amharic 87 French 85 Creole 67 Portuguese 56

267 Table A3. Probability of Employment in Country of Origin (Probit) 250

268 Table A4. Men s Probability of Employment in Country of Origin (Probit) 251

269 Table A5. Women s Probability of Employment in Country of Origin (Probit) 252

270 Table A6. Return to Education of New Immigrants in Country of Origin (Dependent Variable: Log of Weekly Wages Abroad) 253

271 Table A7. Return to Education of New Immigrant Men in Country of Origin (Dependent Variable: Log of Weekly Wages Abroad) 254

272 Table A8. Return to Education of New Immigrant Women in Country of Origin (Dependent Variable: Log of Weekly Wages Abroad) 255

273 Table A9. (Heckman) Return to Education of New Immigrants in Country of Origin (Dependent Variable: Log of Weekly Wages Abroad) 256

274 Table A10. (Heckman) Return to Education of New Immigrant Men in Country of Origin (Dependent Variable: Log of Weekly Wages Abroad) 257

275 Table A11. (Heckman) Return to Education of New Immigrant Women in Country of Origin (Dependent Variable: Log of Weekly Wages Abroad) 258

276 Table A12. Probability of Employment in the United States (Education as a Continuous Variable) (Probit) 259

277 Table A12. Probability of Employment in the United States (Education as a Continuous Variable) (Probit) (Continued) 260

278 Table A13. Men s Probability of Employment in the United States (Education as a Continuous Variable) (Probit) 261

279 Table A13. Men s Probability of Employment in the United States (Education as a Continuous Variable) (Probit) (Continued) 262

280 Table A14. Women s Probability of Employment in the United States (Education as a Continuous Variable) (Probit) 263

281 Table A14. Women s Probability of Employment in the United States (Education as a Continuous Variable) (Probit) (Continued) 264

282 Table A15. New Immigrants Return to Education and Transferable Skills (Dependent Variable: Wages in the United States) (Education as a Continuous Variable) 265

283 Table A16. New Immigrant Men s Return to Education and Transferable Skills (Dependent Variable: Wages in the United States) (Education as a Continuous Variable) 266

284 Table A17. New Immigrant Women s Return to Education and Transferable Skills (Dependent Variable: Wages in the United States) (Education as a Continuous Variable) 267

285 Table A18. New Immigrants Return to Education and Transferable Skills Heckman Estimation (Dependent Variable: Log of Weekly Wages in the United States) (Education as a Continuous Variable) 268

286 Table A18. New Immigrants Return to Education and Transferable Skills Heckman Estimation (Dependent Variable: Log of Weekly Wages in the United States) (Education as a Continuous Variable) (Continued) 269

287 Table A19. New Immigrant Men s Return to Education and Transferable Skills Heckman Estimation (Dependent Variable: Log of Weekly Wages in the United States) (Education as a Continuous Variable) 270

288 Table A19. New Immigrant Men s Return to Education and Transferable Skills Heckman Estimation (Dependent Variable: Log of Weekly Wages in the United States) (Education as a Continuous Variable) (Continued) 271

289 Table A20. New Immigrant Women s Return to Education and Transferable Skills Heckman Estimation (Dependent Variable: Log of Weekly Wages in the United States) (Education as a Continuous Variable) 272

290 Table A20. New Immigrant Women s Return to Education and Transferable Skills Heckman Estimation (Dependent Variable: Log of Weekly Wages in the United States) (Education as a Continuous Variable) (Continued) 273

262 Index. D demand shocks, 146n demographic variables, 103tn

262 Index. D demand shocks, 146n demographic variables, 103tn Index A Africa, 152, 167, 173 age Filipino characteristics, 85 household heads, 59 Mexican migrants, 39, 40 Philippines migrant households, 94t 95t nonmigrant households, 96t 97t premigration income effects,

More information

The Determinants and the Selection. of Mexico-US Migrations

The Determinants and the Selection. of Mexico-US Migrations The Determinants and the Selection of Mexico-US Migrations J. William Ambrosini (UC, Davis) Giovanni Peri, (UC, Davis and NBER) This draft March 2011 Abstract Using data from the Mexican Family Life Survey

More information

The Causes of Wage Differentials between Immigrant and Native Physicians

The Causes of Wage Differentials between Immigrant and Native Physicians The Causes of Wage Differentials between Immigrant and Native Physicians I. Introduction Current projections, as indicated by the 2000 Census, suggest that racial and ethnic minorities will outnumber non-hispanic

More information

Understanding Different Migrant Selection Patterns in Rural and Urban Mexico by Jesús Fernández-Huertas Moraga * Documento de Trabajo

Understanding Different Migrant Selection Patterns in Rural and Urban Mexico by Jesús Fernández-Huertas Moraga * Documento de Trabajo Understanding Different Migrant Selection Patterns in Rural and Urban Mexico by Jesús Fernández-Huertas Moraga * Documento de Trabajo 2013-02 January 2013 ** FEDEA and IAE, CSIC. Los Documentos de Trabajo

More information

Labor Market Dropouts and Trends in the Wages of Black and White Men

Labor Market Dropouts and Trends in the Wages of Black and White Men Industrial & Labor Relations Review Volume 56 Number 4 Article 5 2003 Labor Market Dropouts and Trends in the Wages of Black and White Men Chinhui Juhn University of Houston Recommended Citation Juhn,

More information

Immigrant Employment and Earnings Growth in Canada and the U.S.: Evidence from Longitudinal data

Immigrant Employment and Earnings Growth in Canada and the U.S.: Evidence from Longitudinal data Immigrant Employment and Earnings Growth in Canada and the U.S.: Evidence from Longitudinal data Neeraj Kaushal, Columbia University Yao Lu, Columbia University Nicole Denier, McGill University Julia Wang,

More information

NBER WORKING PAPER SERIES INTERNATIONAL MIGRATION, SELF-SELECTION, AND THE DISTRIBUTION OF WAGES: EVIDENCE FROM MEXICO AND THE UNITED STATES

NBER WORKING PAPER SERIES INTERNATIONAL MIGRATION, SELF-SELECTION, AND THE DISTRIBUTION OF WAGES: EVIDENCE FROM MEXICO AND THE UNITED STATES NBER WORKING PAPER SERIES INTERNATIONAL MIGRATION, SELF-SELECTION, AND THE DISTRIBUTION OF WAGES: EVIDENCE FROM MEXICO AND THE UNITED STATES Daniel Chiquiar Gordon H. Hanson Working Paper 9242 http://www.nber.org/papers/w9242

More information

THREE ESSAYS ON THE BLACK WHITE WAGE GAP

THREE ESSAYS ON THE BLACK WHITE WAGE GAP University of Kentucky UKnowledge University of Kentucky Doctoral Dissertations Graduate School 2009 THREE ESSAYS ON THE BLACK WHITE WAGE GAP Nola Ogunro University of Kentucky, nogun2@uky.edu Click here

More information

CROSS-COUNTRY VARIATION IN THE IMPACT OF INTERNATIONAL MIGRATION: CANADA, MEXICO, AND THE UNITED STATES

CROSS-COUNTRY VARIATION IN THE IMPACT OF INTERNATIONAL MIGRATION: CANADA, MEXICO, AND THE UNITED STATES CROSS-COUNTRY VARIATION IN THE IMPACT OF INTERNATIONAL MIGRATION: CANADA, MEXICO, AND THE UNITED STATES Abdurrahman Aydemir Statistics Canada George J. Borjas Harvard University Abstract Using data drawn

More information

Prospects for Immigrant-Native Wealth Assimilation: Evidence from Financial Market Participation. Una Okonkwo Osili 1 Anna Paulson 2

Prospects for Immigrant-Native Wealth Assimilation: Evidence from Financial Market Participation. Una Okonkwo Osili 1 Anna Paulson 2 Prospects for Immigrant-Native Wealth Assimilation: Evidence from Financial Market Participation Una Okonkwo Osili 1 Anna Paulson 2 1 Contact Information: Department of Economics, Indiana University Purdue

More information

WHO MIGRATES? SELECTIVITY IN MIGRATION

WHO MIGRATES? SELECTIVITY IN MIGRATION WHO MIGRATES? SELECTIVITY IN MIGRATION Mariola Pytliková CERGE-EI and VŠB-Technical University Ostrava, CReAM, IZA, CCP and CELSI Info about lectures: https://home.cerge-ei.cz/pytlikova/laborspring16/

More information

Educational Attainment: Analysis by Immigrant Generation

Educational Attainment: Analysis by Immigrant Generation DISCUSSION PAPER SERIES IZA DP No. 731 Educational Attainment: Analysis by Immigrant Generation Barry R. Chiswick Noyna DebBurman February 2003 Forschungsinstitut zur Zukunft der Arbeit Institute for the

More information

School Performance of the Children of Immigrants in Canada,

School Performance of the Children of Immigrants in Canada, School Performance of the Children of Immigrants in Canada, 1994-98 by Christopher Worswick * No. 178 11F0019MIE No. 178 ISSN: 1205-9153 ISBN: 0-662-31229-5 Department of Economics, Carleton University

More information

Canadian Labour Market and Skills Researcher Network

Canadian Labour Market and Skills Researcher Network Canadian Labour Market and Skills Researcher Network Working Paper No. 69 Immigrant Earnings Growth: Selection Bias or Real Progress? Garnett Picot Statistics Canada Patrizio Piraino Statistics Canada

More information

DETERMINANTS OF INTERNAL MIGRATION IN PAKISTAN

DETERMINANTS OF INTERNAL MIGRATION IN PAKISTAN The Journal of Commerce Vol.5, No.3 pp.32-42 DETERMINANTS OF INTERNAL MIGRATION IN PAKISTAN Nisar Ahmad *, Ayesha Akram! and Haroon Hussain # Abstract The migration is a dynamic process and it effects

More information

Education, Credentials and Immigrant Earnings*

Education, Credentials and Immigrant Earnings* Education, Credentials and Immigrant Earnings* Ana Ferrer Department of Economics University of British Columbia and W. Craig Riddell Department of Economics University of British Columbia August 2004

More information

ESSAYS ON MEXICAN MIGRATION. by Heriberto Gonzalez Lozano B.A., Universidad Autonóma de Nuevo León, 2005 M.A., University of Pittsburgh, 2011

ESSAYS ON MEXICAN MIGRATION. by Heriberto Gonzalez Lozano B.A., Universidad Autonóma de Nuevo León, 2005 M.A., University of Pittsburgh, 2011 ESSAYS ON MEXICAN MIGRATION by Heriberto Gonzalez Lozano B.A., Universidad Autonóma de Nuevo León, 2005 M.A., University of Pittsburgh, 2011 Submitted to the Graduate Faculty of the Dietrich School of

More information

Immigrants and the Receipt of Unemployment Insurance Benefits

Immigrants and the Receipt of Unemployment Insurance Benefits Comments Welcome Immigrants and the Receipt of Unemployment Insurance Benefits Wei Chi University of Minnesota wchi@csom.umn.edu and Brian P. McCall University of Minnesota bmccall@csom.umn.edu July 2002

More information

Volume Author/Editor: David Card and Richard B. Freeman. Volume URL:

Volume Author/Editor: David Card and Richard B. Freeman. Volume URL: This PDF is a selection from an out-of-print volume from the National Bureau of Economic Research Volume Title: Small Differences That Matter: Labor Markets and Income Maintenance in Canada and the United

More information

NBER WORKING PAPER SERIES SELF-SELECTION AND THE EARNINGS OF IMMIGRANTS. George J. Borjas. Working Paper No. 2248

NBER WORKING PAPER SERIES SELF-SELECTION AND THE EARNINGS OF IMMIGRANTS. George J. Borjas. Working Paper No. 2248 NBER WORKING PAPER SERIES SELF-SELECTION AND THE EARNINGS OF IMMIGRANTS George J. Borjas Working Paper No. 2248 NATIONAL BUREAU OF ECONOMIC RESEARCH 1050 Massachusetts Avenue Cambridge, MA 02138 May 1987

More information

The Occupational Selection of Emigrants

The Occupational Selection of Emigrants The Occupational Selection of Emigrants Miguel Flores, Alexander Patt, Jens Ruhose, and Simon Wiederhold PRELIMINARY VERSION Please do not cite without permission. Abstract The current literature that

More information

The Economic Status of Asian Americans Before and After the Civil Rights Act

The Economic Status of Asian Americans Before and After the Civil Rights Act D I S C U S S I O N P A P E R S E R I E S IZA DP No. 6639 The Economic Status of Asian Americans Before and After the Civil Rights Act Harriet Orcutt Duleep Seth Sanders June 2012 Forschungsinstitut zur

More information

What drives the language proficiency of immigrants? Immigrants differ in their language proficiency along a range of characteristics

What drives the language proficiency of immigrants? Immigrants differ in their language proficiency along a range of characteristics Ingo E. Isphording IZA, Germany What drives the language proficiency of immigrants? Immigrants differ in their language proficiency along a range of characteristics Keywords: immigrants, language proficiency,

More information

Immigrant Earnings Growth: Selection Bias or Real Progress?

Immigrant Earnings Growth: Selection Bias or Real Progress? Catalogue no. 11F0019M No. 340 ISSN 1205-9153 ISBN 978-1-100-20222-8 Research Paper Analytical Studies Branch Research Paper Series Immigrant Earnings Growth: Selection Bias or Real Progress? by Garnett

More information

WORKING P A P E R. Immigrants and the Labor Market JAMES P. SMITH WR-321. November 2005

WORKING P A P E R. Immigrants and the Labor Market JAMES P. SMITH WR-321. November 2005 WORKING P A P E R Immigrants and the Labor Market JAMES P. SMITH WR-321 November 2005 This product is part of the RAND Labor and Population working paper series. RAND working papers are intended to share

More information

Can migration prospects reduce educational attainments? *

Can migration prospects reduce educational attainments? * Can migration prospects reduce educational attainments? * David McKenzie a and Hillel Rapoport b a Department of Economics, Stanford University, and World Bank Development Research Group b Department of

More information

Remittances and the Brain Drain: Evidence from Microdata for Sub-Saharan Africa

Remittances and the Brain Drain: Evidence from Microdata for Sub-Saharan Africa 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

More information

THE ENGLISH LANGUAGE FLUENCY AND OCCUPATIONAL SUCCESS OF ETHNIC MINORITY IMMIGRANT MEN LIVING IN ENGLISH METROPOLITAN AREAS

THE ENGLISH LANGUAGE FLUENCY AND OCCUPATIONAL SUCCESS OF ETHNIC MINORITY IMMIGRANT MEN LIVING IN ENGLISH METROPOLITAN AREAS THE ENGLISH LANGUAGE FLUENCY AND OCCUPATIONAL SUCCESS OF ETHNIC MINORITY IMMIGRANT MEN LIVING IN ENGLISH METROPOLITAN AREAS By Michael A. Shields * and Stephen Wheatley Price ** April 1999, revised August

More information

The Impact of Foreign Workers on the Labour Market of Cyprus

The Impact of Foreign Workers on the Labour Market of Cyprus Cyprus Economic Policy Review, Vol. 1, No. 2, pp. 37-49 (2007) 1450-4561 The Impact of Foreign Workers on the Labour Market of Cyprus Louis N. Christofides, Sofronis Clerides, Costas Hadjiyiannis and Michel

More information

Gender and Ethnicity in LAC Countries: The case of Bolivia and Guatemala

Gender and Ethnicity in LAC Countries: The case of Bolivia and Guatemala Gender and Ethnicity in LAC Countries: The case of Bolivia and Guatemala Carla Canelas (Paris School of Economics, France) Silvia Salazar (Paris School of Economics, France) Paper Prepared for the IARIW-IBGE

More information

The Decline in Earnings of Childhood Immigrants in the U.S.

The Decline in Earnings of Childhood Immigrants in the U.S. The Decline in Earnings of Childhood Immigrants in the U.S. Hugh Cassidy October 30, 2015 Abstract Recent empirical work documenting a declining trend in immigrant earnings relative to natives has focused

More information

Living in the Shadows or Government Dependents: Immigrants and Welfare in the United States

Living in the Shadows or Government Dependents: Immigrants and Welfare in the United States Living in the Shadows or Government Dependents: Immigrants and Welfare in the United States Charles Weber Harvard University May 2015 Abstract Are immigrants in the United States more likely to be enrolled

More information

What Happened to the Immigrant \ Native Wage Gap during the Crisis: Evidence from Ireland

What Happened to the Immigrant \ Native Wage Gap during the Crisis: Evidence from Ireland What Happened to the Immigrant \ Native Wage Gap during the Crisis: Evidence from Ireland Alan Barrett, Adele Bergin, Elish Kelly and Séamus McGuinness 14 June 2013 Dublin Structure Background on Ireland

More information

Higher Education and International Migration in Asia: Brain Circulation. Mark R. Rosenzweig. Yale University. December 2006

Higher Education and International Migration in Asia: Brain Circulation. Mark R. Rosenzweig. Yale University. December 2006 Higher Education and International Migration in Asia: Brain Circulation Mark R. Rosenzweig Yale University December 2006 Prepared for the Regional Bank Conference on Development Economics (RBCDE) - Beijing

More information

Do Highly Educated Immigrants Perform Differently in the Canadian and U.S. Labour Markets?

Do Highly Educated Immigrants Perform Differently in the Canadian and U.S. Labour Markets? Catalogue no. 11F0019M No. 329 ISSN 1205-9153 ISBN 978-1-100-17669-7 Research Paper Analytical Studies Branch Research Paper Series Do Highly Educated Immigrants Perform Differently in the Canadian and

More information

Modeling Immigrants Language Skills

Modeling Immigrants Language Skills DISCUSSION PAPER SERIES IZA DP No. 2974 Modeling Immigrants Language Skills Barry R. Chiswick Paul W. Miller August 2007 Forschungsinstitut zur Zukunft der Arbeit Institute for the Study of Labor Modeling

More information

New Evidence on Emigrant Selection

New Evidence on Emigrant Selection New Evidence on Emigrant Selection Jesús Fernández-Huertas Moraga April 10, 2008 Abstract This paper examines the extent to which Mexican emigrants to the United States are negatively selected, that is,

More information

Human Capital Outflows

Human Capital Outflows Policy Research Working Paper 8334 WPS8334 Human Capital Outflows Selection into Migration from the Northern Triangle Giselle Del Carmen Liliana D. Sousa Public Disclosure Authorized Public Disclosure

More information

NBER WORKING PAPER SERIES THE LABOR MARKET IMPACT OF HIGH-SKILL IMMIGRATION. George J. Borjas. Working Paper

NBER WORKING PAPER SERIES THE LABOR MARKET IMPACT OF HIGH-SKILL IMMIGRATION. George J. Borjas. Working Paper NBER WORKING PAPER SERIES THE LABOR MARKET IMPACT OF HIGH-SKILL IMMIGRATION George J. Borjas Working Paper 11217 http://www.nber.org/papers/w11217 NATIONAL BUREAU OF ECONOMIC RESEARCH 1050 Massachusetts

More information

Gender-Wage Discrimination by Marital Status in Canada: 2006 to 2016

Gender-Wage Discrimination by Marital Status in Canada: 2006 to 2016 University of Ottawa Gender-Wage Discrimination by Marital Status in Canada: 2006 to 2016 Major Paper submitted to the University of Ottawa Department of Economics in order to complete the requirements

More information

Remittances and Poverty. in Guatemala* Richard H. Adams, Jr. Development Research Group (DECRG) MSN MC World Bank.

Remittances and Poverty. in Guatemala* Richard H. Adams, Jr. Development Research Group (DECRG) MSN MC World Bank. Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Remittances and Poverty in Guatemala* Richard H. Adams, Jr. Development Research Group

More information

Immigration and the U.S. Economy

Immigration and the U.S. Economy Immigration and the U.S. Economy Pia M. Orrenius, Ph.D. Federal Reserve Bank of Dallas June 19, 2007 Mercatus Center, George Mason University Disclaimer: The views expressed herein are those of the presenter;

More information

Country of Origin and Immigrant Earnings, : A Human Capital Investment Perspective

Country of Origin and Immigrant Earnings, : A Human Capital Investment Perspective DISCUSSION PAPER SERIES IZA DP No. 8628 Country of Origin and Immigrant Earnings, 1960-2000: A Human Capital Investment Perspective Harriet Orcutt Duleep Xingfei Liu Mark Regets November 2014 Forschungsinstitut

More information

Entrepreneurship among California s Low-skilled Workers

Entrepreneurship among California s Low-skilled Workers Entrepreneurship among California s Low-skilled Workers April 2010 Magnus Lofstrom with research support from Qian Li and Jay Liao Summary Self-employment has grown significantly in California over the

More information

Household Inequality and Remittances in Rural Thailand: A Lifecycle Perspective

Household Inequality and Remittances in Rural Thailand: A Lifecycle Perspective Household Inequality and Remittances in Rural Thailand: A Lifecycle Perspective Richard Disney*, Andy McKay + & C. Rashaad Shabab + *Institute of Fiscal Studies, University of Sussex and University College,

More information

IMMIGRATION IN HIGH-SKILL LABOR MARKETS: THE IMPACT OF FOREIGN STUDENTS ON THE EARNINGS OF DOCTORATES. George J. Borjas Harvard University

IMMIGRATION IN HIGH-SKILL LABOR MARKETS: THE IMPACT OF FOREIGN STUDENTS ON THE EARNINGS OF DOCTORATES. George J. Borjas Harvard University IMMIGRATION IN HIGH-SKILL LABOR MARKETS: THE IMPACT OF FOREIGN STUDENTS ON THE EARNINGS OF DOCTORATES George J. Borjas Harvard University April 2004 1 IMMIGRATION IN HIGH-SKILL LABOR MARKETS: THE IMPACT

More information

Divergent Trends in Citizenship Rates among Immigrants in Canada and the United States

Divergent Trends in Citizenship Rates among Immigrants in Canada and the United States Catalogue no. 11F0019M No. 338 ISSN 1205-9153 ISBN 978-1-100-19362-5 Research Paper Analytical Studies Branch Research Paper Series Divergent Trends in Citizenship Rates among Immigrants in Canada and

More information

Effects of Institutions on Migrant Wages in China and Indonesia

Effects of Institutions on Migrant Wages in China and Indonesia 15 The Effects of Institutions on Migrant Wages in China and Indonesia Paul Frijters, Xin Meng and Budy Resosudarmo Introduction According to Bell and Muhidin (2009) of the UN Development Programme (UNDP),

More information

Immigration, Wage Inequality and unobservable skills in the U.S. and the UK. First Draft: October 2008 This Draft March 2009

Immigration, Wage Inequality and unobservable skills in the U.S. and the UK. First Draft: October 2008 This Draft March 2009 Immigration, Wage Inequality and unobservable skills in the U.S. and the First Draft: October 2008 This Draft March 2009 Cinzia Rienzo * Royal Holloway, University of London CEP, London School of Economics

More information

Characteristics of the Ethnographic Sample of First- and Second-Generation Latin American Immigrants in the New York to Philadelphia Urban Corridor

Characteristics of the Ethnographic Sample of First- and Second-Generation Latin American Immigrants in the New York to Philadelphia Urban Corridor Table 2.1 Characteristics of the Ethnographic Sample of First- and Second-Generation Latin American Immigrants in the New York to Philadelphia Urban Corridor Characteristic Females Males Total Region of

More information

Wage Differentials in the 1990s: Is the Glass Half-full or Half-empty? Kevin M. Murphy. and. Finis Welch

Wage Differentials in the 1990s: Is the Glass Half-full or Half-empty? Kevin M. Murphy. and. Finis Welch Wage Differentials in the 1990s: Is the Glass Half-full or Half-empty? and Finis Welch Abstract: There are many wrinkles and complexities that have been brought to our attention by the huge volume of research

More information

I'll Marry You If You Get Me a Job: Marital Assimilation and Immigrant Employment Rates

I'll Marry You If You Get Me a Job: Marital Assimilation and Immigrant Employment Rates DISCUSSION PAPER SERIES IZA DP No. 3951 I'll Marry You If You Get Me a Job: Marital Assimilation and Immigrant Employment Rates Delia Furtado Nikolaos Theodoropoulos January 2009 Forschungsinstitut zur

More information

3.3 DETERMINANTS OF THE CULTURAL INTEGRATION OF IMMIGRANTS

3.3 DETERMINANTS OF THE CULTURAL INTEGRATION OF IMMIGRANTS 1 Duleep (2015) gives a general overview of economic assimilation. Two classic articles in the United States are Chiswick (1978) and Borjas (1987). Eckstein Weiss (2004) studies the integration of immigrants

More information

Returning to the Question of a Wage Premium for Returning Migrants

Returning to the Question of a Wage Premium for Returning Migrants DISCUSSION PAPER SERIES IZA DP No. 4736 Returning to the Question of a Wage Premium for Returning Migrants Alan Barrett Jean Goggin February 2010 Forschungsinstitut zur Zukunft der Arbeit Institute for

More information

Edward L. Glaeser Harvard University and NBER and. David C. Maré * New Zealand Department of Labour

Edward L. Glaeser Harvard University and NBER and. David C. Maré * New Zealand Department of Labour CITIES AND SKILLS by Edward L. Glaeser Harvard University and NBER and David C. Maré * New Zealand Department of Labour [Revised version is forthcoming in Journal of Labor Economics 19(2), April 2000]

More information

THE DECLINE IN WELFARE RECEIPT IN NEW YORK CITY: PUSH VS. PULL

THE DECLINE IN WELFARE RECEIPT IN NEW YORK CITY: PUSH VS. PULL THE DECLINE IN WELFARE RECEIPT IN NEW YORK CITY: PUSH VS. PULL Howard Chernick Hunter College and The Graduate Center, City University of New York and Cordelia Reimers Hunter College and The Graduate Center,

More information

Languages of work and earnings of immigrants in Canada outside. Quebec. By Jin Wang ( )

Languages of work and earnings of immigrants in Canada outside. Quebec. By Jin Wang ( ) Languages of work and earnings of immigrants in Canada outside Quebec By Jin Wang (7356764) Major paper presented to the Department of Economics of the University of Ottawa in partial fulfillment of the

More information

Immigrants and Gender Roles: Assimilation vs. Culture

Immigrants and Gender Roles: Assimilation vs. Culture DISCUSSION PAPER SERIES IZA DP No. 9534 Immigrants and Gender Roles: Assimilation vs. Culture Francine D. Blau November 2015 Forschungsinstitut zur Zukunft der Arbeit Institute for the Study of Labor Immigrants

More information

Rethinking the Area Approach: Immigrants and the Labor Market in California,

Rethinking the Area Approach: Immigrants and the Labor Market in California, Rethinking the Area Approach: Immigrants and the Labor Market in California, 1960-2005. Giovanni Peri, (University of California Davis, CESifo and NBER) October, 2009 Abstract A recent series of influential

More information

Wage Structure and Gender Earnings Differentials in China and. India*

Wage Structure and Gender Earnings Differentials in China and. India* Wage Structure and Gender Earnings Differentials in China and India* Jong-Wha Lee # Korea University Dainn Wie * National Graduate Institute for Policy Studies September 2015 * Lee: Economics Department,

More information

Social Networks and Their Impact on the Employment and Earnings of Mexican Immigrants. September 23, 2004

Social Networks and Their Impact on the Employment and Earnings of Mexican Immigrants. September 23, 2004 Social Networks and Their Impact on the Employment and Earnings of Mexican Immigrants Catalina Amuedo-Dorantes San Diego State University Department of Economics San Diego CA 918-4485 Ph: 619-594-1663

More information

Transferability of Human Capital and Immigrant Assimilation: An Analysis for Germany

Transferability of Human Capital and Immigrant Assimilation: An Analysis for Germany Transferability of Human Capital and Immigrant Assimilation: An Analysis for Germany Leilanie Basilio a,b,c Thomas K. Bauer b,c,d Anica Kramer b,c a Ruhr Graduate School in Economics b Ruhr-University

More information

The Impact of Having a Job at Migration on Settlement Decisions: Ethnic Enclaves as Job Search Networks

The Impact of Having a Job at Migration on Settlement Decisions: Ethnic Enclaves as Job Search Networks The Impact of Having a Job at Migration on Settlement Decisions: Ethnic Enclaves as Job Search Networks Lee Tucker Boston University This version: October 15, 2014 Abstract Observational evidence has shown

More information

Since the early 1990s, the technology-driven

Since the early 1990s, the technology-driven Ross Finnie and Ronald g Since the early 1990s, the technology-driven knowledge-based economy has captured the attention and affected the lives of virtually all Canadians. This phenomenon has been of particular

More information

GEORG-AUGUST-UNIVERSITÄT GÖTTINGEN

GEORG-AUGUST-UNIVERSITÄT GÖTTINGEN GEORG-AUGUST-UNIVERSITÄT GÖTTINGEN FACULTY OF ECONOMIC SCIENCES CHAIR OF MACROECONOMICS AND DEVELOPMENT Bachelor Seminar Economics of the very long run: Economics of Islam Summer semester 2017 Does Secular

More information

Evaluating Methods for Estimating Foreign-Born Immigration Using the American Community Survey

Evaluating Methods for Estimating Foreign-Born Immigration Using the American Community Survey Evaluating Methods for Estimating Foreign-Born Immigration Using the American Community Survey By C. Peter Borsella Eric B. Jensen Population Division U.S. Census Bureau Paper to be presented at the annual

More information

IMMIGRANTS IN THE ISRAELI HI- TECH INDUSTRY: COMPARISON TO NATIVES AND THE EFFECT OF TRAINING

IMMIGRANTS IN THE ISRAELI HI- TECH INDUSTRY: COMPARISON TO NATIVES AND THE EFFECT OF TRAINING B2v8:0f XML:ver::0: RLEC V024 : 2400 /0/0 :4 Prod:Type:com pp:2ðcol:fig::nilþ ED:SeemaA:P PAGN: SCAN: 2 IMMIGRANTS IN THE ISRAELI HI- TECH INDUSTRY: COMPARISON TO NATIVES AND THE EFFECT OF TRAINING Sarit

More information

Cons. Pros. Vanderbilt University, USA, CASE, Poland, and IZA, Germany. Keywords: immigration, wages, inequality, assimilation, integration

Cons. Pros. Vanderbilt University, USA, CASE, Poland, and IZA, Germany. Keywords: immigration, wages, inequality, assimilation, integration Kathryn H. Anderson Vanderbilt University, USA, CASE, Poland, and IZA, Germany Can immigrants ever earn as much as native workers? Immigrants initially earn less than natives; the wage gap falls over time,

More information

The Transmission of Women s Fertility, Human Capital and Work Orientation across Immigrant Generations

The Transmission of Women s Fertility, Human Capital and Work Orientation across Immigrant Generations DISCUSSION PAPER SERIES IZA DP No. 3732 The Transmission of Women s Fertility, Human Capital and Work Orientation across Immigrant Generations Francine D. Blau Lawrence M. Kahn Albert Yung-Hsu Liu Kerry

More information

Comparing Wage Gains from Small and Mass Scale Immigrant Legalization. Programs

Comparing Wage Gains from Small and Mass Scale Immigrant Legalization. Programs UNR Economics Working Paper Series Working Paper No. 16-001 Comparing Wage Gains from Small and Mass Scale Immigrant Legalization Programs Sankar Mukhopadhyay Department of Economics /0030 University of

More information

Peruvians in the United States

Peruvians in the United States Peruvians in the United States 1980 2008 Center for Latin American, Caribbean & Latino Studies Graduate Center City University of New York 365 Fifth Avenue Room 5419 New York, New York 10016 212-817-8438

More information

Gender Wage Gap and Discrimination in Developing Countries. Mo Zhou. Department of Agricultural Economics and Rural Sociology.

Gender Wage Gap and Discrimination in Developing Countries. Mo Zhou. Department of Agricultural Economics and Rural Sociology. Gender Wage Gap and Discrimination in Developing Countries Mo Zhou Department of Agricultural Economics and Rural Sociology Auburn University Phone: 3343292941 Email: mzz0021@auburn.edu Robert G. Nelson

More information

The Demography of the Labor Force in Emerging Markets

The Demography of the Labor Force in Emerging Markets The Demography of the Labor Force in Emerging Markets David Lam I. Introduction This paper discusses how demographic changes are affecting the labor force in emerging markets. As will be shown below, the

More information

NBER WORKING PAPER SERIES IMMIGRANT EMPLOYMENT AND EARNINGS GROWTH IN CANADA AND THE U.S.: EVIDENCE FROM LONGITUDINAL DATA

NBER WORKING PAPER SERIES IMMIGRANT EMPLOYMENT AND EARNINGS GROWTH IN CANADA AND THE U.S.: EVIDENCE FROM LONGITUDINAL DATA NBER WORKING PAPER SERIES IMMIGRANT EMPLOYMENT AND EARNINGS GROWTH IN CANADA AND THE U.S.: EVIDENCE FROM LONGITUDINAL DATA Neeraj Kaushal Yao Lu Nicole Denier Julia Shu-Huah Wang Stephen J. Trejo Working

More information

Government data show that since 2000 all of the net gain in the number of working-age (16 to 65) people

Government data show that since 2000 all of the net gain in the number of working-age (16 to 65) people CENTER FOR IMMIGRATION STUDIES June All Employment Growth Since Went to Immigrants of U.S.-born not working grew by 17 million By Steven A. Camarota and Karen Zeigler Government data show that since all

More information

Mexican-American Couples and Their Patterns of Dual Earning

Mexican-American Couples and Their Patterns of Dual Earning Mexican-American Couples and Their Patterns of Dual Earning Lori Reeder and Julie Park University of Maryland, College Park For presentation at the annual meeting of the Population Association of America,

More information

Long live your ancestors American dream:

Long live your ancestors American dream: Long live your ancestors American dream: The self-selection and multigenerational mobility of American immigrants Joakim Ruist* University of Gothenburg joakim.ruist@economics.gu.se April 2017 Abstract

More information

NBER WORKING PAPER SERIES THE EFFECT OF IMMIGRATION ON PRODUCTIVITY: EVIDENCE FROM US STATES. Giovanni Peri

NBER WORKING PAPER SERIES THE EFFECT OF IMMIGRATION ON PRODUCTIVITY: EVIDENCE FROM US STATES. Giovanni Peri NBER WKG PER SEES THE EFFE OF IMGRATION ON PRODUIVITY: EVEE FROM US STATES Giovanni Peri Working Paper 15507 http://www.nber.org/papers/w15507 NATION BUREAU OF ENOC RESECH 1050 Massachusetts Avenue Cambridge,

More information

Assimilation of foreigners in Germany

Assimilation of foreigners in Germany Clemson University TigerPrints All Dissertations Dissertations 12-2014 Assimilation of foreigners in Germany Jan Schiebler Clemson University, jschieb@g.clemson.edu Follow this and additional works at:

More information

Immigration is a contentious issue in the industrialized nations of the

Immigration is a contentious issue in the industrialized nations of the Journal of Economic Perspectives Volume 9, Number 2 Spring 1995 Pages 23 44 The Impact of Immigrants on Host Country Wages, Employment and Growth Rachel M. Friedberg and Jennifer Hunt Immigration is a

More information

Second-Generation Immigrants? The 2.5 Generation in the United States n

Second-Generation Immigrants? The 2.5 Generation in the United States n Second-Generation Immigrants? The 2.5 Generation in the United States n S. Karthick Ramakrishnan, Public Policy Institute of California Objective. This article takes issue with the way that second-generation

More information

The Economic and Social Outcomes of Children of Migrants in New Zealand

The Economic and Social Outcomes of Children of Migrants in New Zealand The Economic and Social Outcomes of Children of Migrants in New Zealand Julie Woolf Statistics New Zealand Julie.Woolf@stats.govt.nz, phone (04 931 4781) Abstract This paper uses General Social Survey

More information

Chapter 9. Labour Mobility. Introduction

Chapter 9. Labour Mobility. Introduction Chapter 9 Labour Mobility McGraw-Hill/Irwin Labor Economics, 4 th edition Copyright 2008 The McGraw-Hill Companies, Inc. All rights reserved. 9-2 Introduction Existing allocation of workers and firms is

More information

Discussion Paper Series

Discussion Paper Series Discussion Paper Series CPD 11/15 International Import Competition and the Decision to Migrate: Evidence from Mexico Kaveh Majlesi and Gaia Narciso Centre for Research and Analysis of Migration Department

More information

Longitudinal Analysis of Assimilation, Ethnic Capital and Immigrants Earnings: Evidence from a Hausman-Taylor Estimation

Longitudinal Analysis of Assimilation, Ethnic Capital and Immigrants Earnings: Evidence from a Hausman-Taylor Estimation Longitudinal Analysis of Assimilation, Ethnic Capital and Immigrants Earnings: Evidence from a Hausman-Taylor Estimation Xingang (Singa) Wang Economics Department, University of Auckland Abstract In this

More information

Cross-Country Intergenerational Status Mobility: Is There a Great Gatsby Curve?

Cross-Country Intergenerational Status Mobility: Is There a Great Gatsby Curve? Cross-Country Intergenerational Status Mobility: Is There a Great Gatsby Curve? John A. Bishop Haiyong Liu East Carolina University Juan Gabriel Rodríguez Universidad Complutense de Madrid Abstract Countries

More information

MEXICO-US IMMIGRATION: EFFECTS OF WAGES

MEXICO-US IMMIGRATION: EFFECTS OF WAGES MEXICO-US IMMIGRATION: EFFECTS OF WAGES AND BORDER ENFORCEMENT Rebecca Lessem November 28, 2017 Abstract In this paper, I study how relative wages and border enforcement affect immigration from Mexico

More information

Social Security Contributions and Return Migration among Older Mexican Immigrants

Social Security Contributions and Return Migration among Older Mexican Immigrants Social Security Contributions and Return Migration among Older Mexican Immigrants Emma Aguila University of Southern California (USC) Alma Vega University of Pennsylvania 17 th Annual Joint Meeting of

More information

Europe and the US: Preferences for Redistribution

Europe and the US: Preferences for Redistribution Europe and the US: Preferences for Redistribution Peter Haan J. W. Goethe Universität Summer term, 2010 Peter Haan (J. W. Goethe Universität) Europe and the US: Preferences for Redistribution Summer term,

More information

CCIS. Immigrants and Their Schooling. By James P. Smith Senior Economist - RAND

CCIS. Immigrants and Their Schooling. By James P. Smith Senior Economist - RAND The Center for Comparative Immigration Studies University of California, San Diego CCIS Immigrants and Their Schooling By James P. Smith Senior Economist - RAND Working Paper 108 October 2004 Oct 2004

More information

The effect of age at immigration on the earnings of immigrants: Estimates from a two-stage model

The effect of age at immigration on the earnings of immigrants: Estimates from a two-stage model The effect of age at immigration on the earnings of immigrants: Estimates from a two-stage model By Chang Dong Student No. 6586955 Major paper presented to the Department of Economics of the University

More information

Immigration and the Labour Market Outcomes of Natives in Developing Countries: A Case Study of South Africa

Immigration and the Labour Market Outcomes of Natives in Developing Countries: A Case Study of South Africa Immigration and the Labour Market Outcomes of Natives in Developing Countries: A Case Study of South Africa Nzinga H. Broussard Preliminary Please do not cite. Revised July 2012 Abstract According to the

More information

MACQUARIE ECONOMICS RESEARCH PAPERS. Do Migrants Succeed in the Australian Labour Market? Further Evidence on Job Quality

MACQUARIE ECONOMICS RESEARCH PAPERS. Do Migrants Succeed in the Australian Labour Market? Further Evidence on Job Quality DEPARTMENT OF ECONOMICS MACQUARIE ECONOMICS RESEARCH PAPERS Do Migrants Succeed in the Australian Labour Market? Further Evidence on Job Quality Stéphane Mahuteau and P.N. (Raja) Junankar Number 3/2007

More information

How Should Immigration Affect the Economy? A D A M M. Z A R E T S K Y

How Should Immigration Affect the Economy? A D A M M. Z A R E T S K Y The by A D A M M. Z A R E T S K Y T he number of immigrants entering the United States legally is greater today than it was at the turn of the century. In fact, after peaking in the early 1900s and registering

More information

Undocumented Immigration to California:

Undocumented Immigration to California: Undocumented Immigration to California: 1980-1993 Hans P. Johnson September 1996 Copyright 1996 Public Policy Institute of California, San Francisco, CA. All rights reserved. PPIC permits short sections

More information

The Effect of International Trade on Wages of Skilled and Unskilled Workers: Evidence from Brazil

The Effect of International Trade on Wages of Skilled and Unskilled Workers: Evidence from Brazil The Effect of International Trade on Wages of Skilled and Unskilled Workers: Evidence from Brazil Aris Bijleveld E-mail: 336250ab@student.eur.nl June, 2011 ERASMUS UNIVERSITY ROTTERDAM Erasmus School of

More information

A COMPARISON OF ARIZONA TO NATIONS OF COMPARABLE SIZE

A COMPARISON OF ARIZONA TO NATIONS OF COMPARABLE SIZE A COMPARISON OF ARIZONA TO NATIONS OF COMPARABLE SIZE A Report from the Office of the University Economist July 2009 Dennis Hoffman, Ph.D. Professor of Economics, University Economist, and Director, L.

More information

NBER Volume on International Differences in Entrepreneurship

NBER Volume on International Differences in Entrepreneurship The International Asian Business Success Story: A Comparison of Chinese, Indian and Other Asian Businesses in the United States, Canada and United Kingdom NBER Volume on International Differences in Entrepreneurship

More information

Economic Gains from Educational Reform by US States

Economic Gains from Educational Reform by US States Economic Gains from Educational Reform by US States Eric A. Hanushek Hoover Institution, Stanford University, CESifo, Institute of Labor Economics (IZA), and National Bureau of Economic Research Jens Ruhose

More information

The Effect of Niche Occupations on Standard of Living: A Closer Look at Chinese, Filipino, and Asian Indian Immigrants

The Effect of Niche Occupations on Standard of Living: A Closer Look at Chinese, Filipino, and Asian Indian Immigrants The Park Place Economist Volume 19 Issue 1 Article 11 2011 The Effect of Niche Occupations on Standard of Living: A Closer Look at Chinese, Filipino, and Asian Indian Immigrants Paige Maynard '11 Illinois

More information