Economic Sector Choices of Mexican Migrants to the U.S.: Evidence from the 2011 EMIF Border Survey

Transcription

1 Economic Sector Choices of Mexican Migrants to the U.S.: Evidence from the 2011 EMIF Border Survey André Rossi de Oliveira Finance and Economics Department Utah Valley University Orem, UT Ph: (801) Rossitza B. Wooster Department of Economics Portland State University Portland, OR Ph: (503) Draft: March 2013 Abstract: This paper uses a random utility model to investigate the economic sector choices of Mexican labor migrants who intended to cross the U.S. border in We use individual level data from the 2011 EMIF Norte Border Survey to identify migrants according to prior work experience and intended sector of work in an effort to determine what demographic and socioeconomic characteristics explain economic sector mobility. Our empirical analysis focuses two separate but related questions. First, we estimate a probit model with sample selection to identify migrant characteristics that explain differences between industry of employment at place of origin and the intended sector of work at their destination. Second, we estimate a multinomial probit of the choice of sector to further explore migrant characteristics that explain occupational selection. Our results show that sector mobility is significantly more likely for migrants who are documented and those with higher educational attainment, specifically, spoken English skills. At the same time, the probability that sector of work coincides with intended sector of work in the US is significantly higher for migrants who are male, married, traveling in larger groups, have family in the US and earned a higher wage prior to migration. The results from our multinomial analysis further clarify these patterns. We find that work sector prior to migration is more likely to match intended sector in the agriculture, construction, transportation and trade industries and significantly less likely to match in the services sector relative to other occupational categories. Keywords: international migration, self selection, individual level data JEL Classification Codes: F22, O15, D33 Contact Author 1

2 After four decades that brought 12 million current immigrants most of whom came illegally the net migration flow from Mexico to the United States has stopped and may have reversed The standstill appears to be the result of many factors, including the weakened U.S. job and housing construction markets, heightened border enforcement, a rise in deportations, the growing dangers associated with illegal border crossings, the long term decline in Mexico s birth rates and broader economic conditions in Mexico. It is possible that the Mexican immigration wave will resume as the U.S. economy recovers I. Introduction (Passel et al., 2012) The importance of the Latino vote for the outcome of the 2012 presidential election, coupled with the slow recovery of employment in the U.S. following the economic crisis, have made once again elevated the hotly contested comprehensive immigration reform debate to the top of the public policy agenda (Nakamura, 2013). Mexican US immigration patterns, in particular, have received a significant amount of attention for two main reasons. First, Mexican immigrants comprise approximately 30% of all current US immigrants and just over half of them are undocumented (Passel et al., 2012). Second, Mexican immigrants tend to have demographic and socioeconomic characteristics that differ significantly not only from those for the native born population, but from other immigrants as well (Borjas and Katz, 2007). Related to the latter, an important point of debate in the literature has been whether Mexican migrant selection is negative (i.e. U.S. attracts workers from the low end of the Mexican distribution of skills) or positive (migrants are on average drawn from the high skilled end). In recent years, significant contributions to this debate have come from studies of migrant selection based on U.S. and Mexican microdata samples which identify the demographic and socioeconomic factors influencing the decision to migrate as well as the effects of migration on the sending and receiving countries (Peri, 2012). 1 For example, evidence of positive selection has been documented by Chiquiar and Hanson (2005) using US and Mexican census data and, to some extent, by 1 The more established (and voluminous) literature on the determinants of international migration based on macrodata serves as a foundation for the newer microdata research on the characteristics of migrants and dates back to Sjaastad (1962) (Bertoli et al, 2012). Some important recent studies using macrodata on bilateral migration flows to study migration decisions include Grogger and Hanson (2008), Belot and Hatton (2008), and Ortega and Peri (2009), among others. 2

3 McKenzie and Rapoport (2010) and Orrenius and Zavodny (2005) using data from the Mexican Population Project (MPP). In contrast, Kaestner and Malamud (2010) use data from the Mexican Family Life Survey (MxFLS) to document significant negative selection of Mexican immigrants on age, education and wages suggesting that both male and female migrants are more likely to be younger, with fewer years of education and from rural places of origin in Mexico. In addition, Ambrosini and Peri (2012) also use the MxFLS data and find that undocumented migration by less educated Mexicans (less than 12 years of schooling) partly explains negative selection as it increases the cost of migration for those at the high end of the skill distribution. Overall, the discrepancy in findings of positive and negative selection in recent literate are primarily due to an under count of unskilled migrants in US sources underscoring the importance of micro level datasets that document immigrant characteristics irrespective of whether they are documented or not (Moraga, 2011). One dimension that has received little attention in the migrant selection literature is how the demographic and socioeconomic characteristics of migrants influence occupational and economic sector choices. A notable exception is Patel and Vella (2012) who use U.S. census data to study occupational choices by recent immigrants form all countries. The authors identify network effects as an important and positive determinant of both the choice of occupation and the earnings migrants receive in that line of work. While we are not aware of any studies focusing strictly on the economic sector choice of Mexican immigrants, historical patterns suggest that Mexico to U.S. migrants have been selected from the middle to lower ranges of the socioeconomic hierarchy, from rural areas and agricultural jobs in Mexico, and for low skill jobs in the United States (Bustamonte et al., 1998). The objective of this paper is to examine the economic sector choices of Mexican migrants and identify the demographic and socioeconomic factors that increase the probability of migrants selecting into specific industries. We make three distinct contributions to existing literature. First, we use 2011 data from the 2011 EMIF Norte Border Survey of Migration (EMIF) to document demographic and 3

4 socioeconomic characteristics of migrants in the post net zero migration period. Second, we identify migrants according to prior work experience and intended sector of work in an effort to determine what demographic and socioeconomic characteristics explain economic sector mobility. Here we estimate a probit model with sample selection and identify migrant characteristics that explain differences between industry of employment at place of origin and the intended sector of work at their destination. Third, we provide a sector specific analysis of migrant choices based on a multinomial probit of the choice of sector. Our results show that sector mobility is significantly more likely for migrants who are documented, those with higher educational attainment, and those with spoken English skills. At the same time, the probability that sector of work coincides with intended sector of work in the US is significantly higher for migrants who are male, married, traveling in larger groups, have family in the US and had relatively higher earnings in Mexico prior to migration. The results from our multinomial analysis further clarify these patterns. We find that work sector prior to migration is more likely to match intended sector for workers who indicated agriculture, construction, transportation and trade industries as their destination relative to our baseline sector of other. In contrast, workers indicating services as their choice of economic sector at their destination were significantly less likely to originate from service occupations in Mexico to our other baseline category. The rest of this paper is structured as follows. Section II gives a brief description of the data set, whereas section III provides an overview of demographic and socioeconomic characteristics Mexican migrants and discusses patterns evident in the univariate analysis. Section IV outlines our estimation framework. Section V presents and discusses the empirical results. Section VI concludes with policy implications and directions for future research. 4

5 II. Brief description of data set The source of our data is the Encuesta sobre Migración en la Frontera Norte de México (EMIF Norte, from now on), a survey that has been conducted since 1993 (continuously since 1999) by several Mexican government agencies 2 together with El Colegio de la Frontera Norte (COLEF), a Mexican higher education institution whose teaching and research focus on social aspects of United States Mexico border issues. This survey gathers information about Mexican nationals who are moving from Mexico to the US or from the interior of Mexico to its Northern border. Individuals interviewed for this survey must be over 12 years of age and, depending on their answers to interview questions, are assigned to one of four different migratory flows that encompass the northern border of Mexico: Flow originating in the South: Individuals not born in or resident of the US, non residents of the border town of the interview, and whose stay at the border town is due to one of the following: work; change of residence; in transit towards the US (with or without proper documentation), or; study, tourism, or visit with family or friends with no return date to their place of origin. Flow originating in the Northern border of Mexico: Individuals not born in the US, nonresidents of the border town of the interview, who are going from the border region to the interior of Mexico and whose stay at the border town is due to one of the following: work or change of residence, regardless of the duration of their stay; study, tourism or visit with family or friends with duration of at least one month. Flow originating in the US: Individuals not born in the US, non residents of the border town of the interview, who are (voluntarily) coming from the US, and whose stay in the US was due to one of the following: work or change of residence, regardless of the duration of 2 Secretaría de Gobernación (through the Consejo Nacional de Población and the Instituto Nacional de Migración), Secreataría del Trabajo y Previsión Social, and Secretaría de Relaciones Exteriores (since 2004). 5

6 their stay; study, tourism or visit with family or friends with duration of at least one month 3. Flow of migrants returned to Mexico by US officials: Individuals handed over by US officials to Mexican officials at one of the designated points along the border. Even though most of the individuals in this category are Mexicans, this flow also includes nationals of other countries like El Salvador, Guatemala and other countries in Central America. The survey does not reach several categories of migrants: Those who enter or depart from the US by air using airports located in cities that are not in the sample; those who travel by sea; and those who come from the south and cross into the US by car. It doesn t capture information about local migration either, which refers to migrants who live on one side of the border and work on the other. Observation units are not individuals, but rather what the survey calls displacements. In order to better understand this idea, we need to provide a brief description of the sampling process, which has geographic and temporal dimensions. Let s start with a depiction of the geographic dimension. There are three sampling regions. The east region consists of border towns located to the east of Ciudad Acuña, state of Coahuila, namely Ciudad Acuña itself, Piedras Negras, Nuevo Laredo, Reynosa, and Matamoros. Towns that are part of the center region are Altar, Nogales, Agua Prieta, and Ciudad Juárez, all located between Ciudad Acuña and Nogales, state of Sonora. The last region includes two border towns to the west of Nogales, namely Tijuana and Mexicali, and is called the west region. Within each of the cities mentioned above, sampling zones are identified. They can be bus stations (or bus line terminals), airports, border crossing bridges, or security or customs checkpoints on the Mexican side. And within each sampling zone, sampling points are selected in such a way that a given individual would only go there once during her visit to the city. Examples of points like that are boarding platforms or 3 Individuals who have spent some time in the US and some time in a border town are assigned to either the Flow originating in the Northern border of Mexico or the Flow originating in the US depending on the duration of their stay (the longer of the two determines the outcome). 6

7 waiting rooms at bus stations, and baggage claim areas at airports. Individuals (migrants and nonmigrants) are interviewed at different sampling points. The temporal dimension is easier to describe. Each survey year is divided into quarters. Within each quarter, there are several sampling days. Each sampling day consists of sampling hours. Individuals (migrants and non migrants) are interviewed at different sampling hours. A general displacement is defined as a pair (sampling point, sampling hour), and corresponds to an individual who was interviewed at a given place and at a given time. Each general displacement observation therefore corresponds to a filled out filter survey form. This filter form contains questions that are used to distinguish between general displacements and migratory displacements (i.e. those associated with individuals who are deemed migrants). Once a migrant is identified, she is asked a series of questions about socio demographic characteristics, work history, motivation for migrating, origin and destination. Information is also collected about the conditions under which the migratory movement takes place, and the risks of crossing the border. We list and describe the variables we elected to work with in the next section. III. Characteristics of Mexican Migrants in 2011: Evidence EMIF Norte Border Survey The EMIF Norte Border Survey provides detailed characteristics of individual migrants in each year. In this paper we work with data from 2011 only, the most recent sample available. Demographic variables of interest include age, gender, marital status, education and wages in the last place of employment. Additionally, the survey provides information on a migrant s place of origin, employment sector prior to migration, and whether he comes from rural or urban areas in Mexico. For each migrant, the survey also provides information on whether he traveled to the U.S. for work before, the number of 7

8 people he is travelling with, whether he has family members in the U.S., and the sector of employment he intends to look for work in once he has crossed into the U.S. Table 1 provides summary statics of the variables for the full sample and difference of means tests by gender (Male vs. Female) and by intention to cross the U.S. border or not (Cross vs. No Cross). We elaborate on these in more detail below. Age and Gender The average age of migrants in the sample is 32 years, which is consistent with historical estimates across sample (Bustamonte et al., 1998). The average age for women in our sample is 43, which is statistically significantly higher than the average age of 30 for men. While the high percentage of men in the full sample (84%) is also consistent with historical migration patterns for Mexicans, it is even higher among those who intend to cross the border (91%) relative to those who are migrating within Mexico for work. Marital Status Just over half of migrants in the full sample report that they are married or engaged (54%), while the remainder is single, divorced, or widowed. There is no significant difference in marital status between males and females, but migrants who intend to cross are significantly more likely to be married or engaged relative to migrants who state they do not intend to cross into the U.S. for work. Educational Attainment, English Language Proficiency and Literacy Figure 1 presents a decomposition of educational attainment by levels of education with numerical correspondence as follows: 0 = no education, 1 = primary education, 2 = junior high, 3 = high school, and 4 = post high school education and professional training. The 2011 EMIF data show that for over 80% of the sample, educational attainment was less than junior high, corresponding to an average of 1.74 in Table 1 or approximately 6.8 years of schooling. Table 1 also shows that females have a significantly higher educational attainment than males (1.87 vs. 1.71). Educational differences, however, are most drastic between those who intend to cross into the U.S. for work and those who do not (1.59 vs. 2.02). 8

9 In terms of language skills and literacy, only 13% of the men report they can speak English, compared to 11% for women. Table 1 also shows that people who indicate they intend to cross the border are significantly more likely to speak English relative to those who do not. Consistent with the educational attainment patterns, we also note that literacy rates, at 96% for the full sample, are significantly lower for migrants destined for the U.S. relative to those who are not, while no significant difference exists between males and females. Wages The average monthly wage in the sample was pesos, which is higher than the average minimum wage of 1378 pesos/month in Mexico in While the average wage for males is almost double that for females in the sample, it is noteworthy that average wages for those intending to cross into the US are significantly lower than those for migrants indicating they do not intend to cross. Places of Origin in Mexico and Rural Versus Urban Origins In 2011, the top ten migrant sending Mexican states were Chiapas, Oaxaca, Veracruz, Michoacán, Jalisco, Guanajuato, Warrior, Puebla, Tamaulipas, and Mexico (in that order). These states together accounted for over 72% of the migrant total. In contrast, the Baja California peninsula (Baja California including Baja California Sur) and the Yucatan peninsula (Campeche, Quintana, Yucatan) were place of origin for only 1.29% of the full sample. With respect to rural versus urban origins, it is interesting to note that while the full sample is almost evenly split between urban and rural places of origin, over 60% of migrants who indicate an intention to cross are from rural areas compared with 32% among those who state they do not intend to cross into the US. This suggests that worker migration within Mexico is significantly more urban as compared with the migrants who are destined for the US. Migration experience characteristics Regarding migration characteristics, 66% of respondents in the 2011 EMIF sample indicated they intend to cross into the U.S. for work, and a significantly higher percentage of these were male (72%) compared 9

10 to female (38%). In addition, 13% of the sample report to have crossed before and this ratio is significantly higher for both males and those who report they intend to cross. Finally, the full sample average number of people traveling is 4.83, and the average is statistically higher for migrants who intend to cross compared to those who do not. Occupation experience prior to migration and the choice of economic sector to seek work in Sector of employment information based on 3 digit NAICS codes is tabulated in Table 2. In terms of frequency, three industries stand out as disproportionately more likely to generate migrants relative to others: agriculture, construction and manufacturing (of both durable and non durable goods). Together, these three economic sectors account for over 65% of migrants in the sample and over 70% of those with prior work experience. The frequencies of those who intend to cross and those who do not are more evenly split when it comes to migrants with no prior work experience as well as those with work experience in the trade, transportation and utilities sector. In contrast, the services sectors tend to display a distinctly lower frequency of migrants who are destined for the U.S. to seek work. In terms of demographic characteristics across sectors, migrants originating in agriculture, construction and manufacturing tend to be younger and mostly male. Interestingly, agricultural workers are the least educated category in the sample (even relative to migrants without prior work experience) and earn the lowest average wages relative to other groups with prior work experience. Of significant interest in the present analysis are migrant characteristics that explain the choice of intended sector of work. Figure 2 presents the mean educational attainment level by sector, revealing agriculture as a destination sector for the least educated and commerce (which includes the information and financial service and the education and health services sectors (see Table 2)) as the destination for the most educated. Figure 3 shows that prior earnings mirror the educational attainment pattern in Figure 2, with the exception of the family and friends category, where migrants more frequently report zero wages in a prior employment experience yet have higher educational attainment than, say, 10

11 agriculture, on average. Figure 4 illustrates the importance of rural origin for the choice of sector. Most notably, agriculture and services are choices of destination sector for migrants from rural places of origin, while construction, manufacturing and commerce are more likely to be associated with migrants from urban places of origin. Finally, Figure 5 shows the patterns of sector choice for two important characteristics of migrants who indicate they intend to cross into the US: whether they have family in the US and whether they are documented. On average, migrants in agriculture, manufacturing and commerce are both more likely to have family in the US and more likely to be documented, while construction and services have the lowest rates of documentation and family. Interestingly, migrants coming to the US to visit family and friends are mostly documented and, as previous tables showed, mostly female and older. Finally, Table 3 presents information on the correspondence between a migrant s prior work experience and the choice of economic sector for the subsample of migrants who indicated they intended to cross into the US. The overall pattern that emerges is that there is not a one to one match between prior work experience and the sector of intended work after crossing the U.S. border. This gives us an opportunity to explore sector mobility among Mexican migrants to the US by evaluating what factors explain the probability that prior work experience matches the economic sector of intended work. Additionally, the patterns discussed above point to the possibility that different demographic factors emerge as more important in choosing some alternatives over others, something we investigate in more detail below. IV. Model and econometric specification As stated above, the goal of our investigation is twofold. First, we wish to evaluate what migrant characteristics help explain the probability of crossing into the U.S. for migrants in the EMIF 2011 sample. Conditional on selecting to cross, we wish to also evaluate the characteristics that help explain 11

12 sector mobility, i.e. whether sector choices of intended work match prior employment activity or not. Second, we investigate what migrant characteristics are most significantly associated with different sector choices. The model we use to study the first problem is probit with sample selection (Van de Ven and Van Praag, 1981). Let y 1 be an indicator variable equal to 1 if the potential migrant indicates he intends to cross the border into the US and 0 otherwise. Then the participation (or selection) equation is given by * 1 if y1 0 y1 * 0 if y1 0, where * y 1 is a latent variable that depends linearly on a set of regressors. It can be interpreted as the difference between the utility of crossing and that of not crossing, so that an individual crosses the border if the net utility is greater than zero. The outcome equation is given by y y if y 0 * * * y1 if 0 where y 2 is the observed variable and * y 2 is the latent variable. The latter is defined as an indicator variable equal to 1 when there is a match between the sector in which the individual worked prior to being interviewed and the sector of intended work after crossing the U.S. border, and zero otherwise. Notice that the model specifies that y 2 is observed when any meaningful value when * y1 0. * y1 0 and that it does not need to take on As mentioned above, * y 1 is given by a linear model. More precisely, we posit y x, * T where 1 is a vector of parameters to be estimated, 1 is a random error distributed according to the standard normal, and x 1 is a vector of regressors that consists of the following variables: AGE, AGE SQUARED, GENDER, MARITAL STATUS, EDUCATIONAL ATTAINMENT, SPEAK ENGLISH, LITERACY, LOG 12

13 WAGE, RURAL ORIGIN, CROSSED BEFORE, NUMBER OF PEOPLE, FAMILY IN THE US, DOCUMENTED, REASON and SECTOR PREVIOUS EMPLOYMENT. We discussed the meaning of all these variables in the previous section. The other latent variable is defined by y * 2 T 1, if x T 0, if x22 2 0, where 2 is a vector of parameters to be estimated and 2 is a random error, also distributed according to the standard normal, whose correlation to 1 is a possibly non zero number. The vector of exogenous variables x 2 differs from x 1, and is comprised of the following variables: AGE, AGE SQUARED, GENDER, MARITAL STATUS, EDUCATIONAL ATTAINMENT, SPEAK ENGLISH, LITERACY, LOG WAGE, RURAL ORIGIN, CROSSED BEFORE, NUMBER OF PEOPLE, FAMILY IN THE US, DOCUMENTED. We discussed the meaning of all these variables in the previous section. The model is estimated by maximum likelihood. In order to investigate the second problem, namely the identification of migrant characteristics most significantly associated with different sector choices, we use a discrete choice model derived from utility maximizing behavior, known in the literature as the random utility model. The decision maker (migrant) n obtains utility U nj when he chooses alternative j from a set of J alternatives. In our model, the alternatives are five economic sectors, namely agriculture, construction, industry, commerce, and services, plus a sector we call other that includes every other possible sector named by the interviewee. The migrant chooses the alternative that gives him the greatest utility. Therefore, he chooses alternative i if and only if Uni Unj for all j i. Utility is decomposed as Unj Vnj nj, where V nj, called representative utility, is known by the researcher up to some parameters, and nj is unknown to the researcher and captures factors that affect utility but are not included in representative utility is linear in parameters and given by V nj. In our model, T Vnj j x n, where x n is a vector of observed 13

14 variables that do not vary across alternatives (sectors), and j is a vector of coefficients that do vary over alternatives. To ensure model identification, one of the j s must be set (normalized) to zero. The alternative chosen for normalization is called the base alternative, and in our case it is the other sector. The exogenous variables that enter x n are AGE, AGE SQUARED, GENDER, MARITAL STATUS, EDUCATIONAL ATTAINMENT, SPEAK ENGLISH, LITERACY, LOG WAGE, RURAL ORIGIN, CROSSED BEFORE, NUMBER OF PEOPLE, FAMILY IN THE US, DOCUMENTED, and SECTOR MATCH. The probability that migrant n chooses alternative i is calculated as Pr P Pr V V j i V V j i. ni ni ni nj nj nj ni ni nj The way the random variable nj is modeled determines the type of discrete choice model to be employed. In the multinomial logit model, the nj is assumed to be independently and identically distributed extreme value, and the resulting choice probabilities are given by the logit formula P ni Vni e V. nj e j Another popular option is the multinomial probit model, which assumes that errors are joint normally distributed. With full covariance matrix, any pattern of correlation and heteroskedasticity can be accommodated. This is a significant advantage over multinomial logit, where independence of the unobserved parts of utility leads to restrictive substitution patterns. More precisely, the multinomial logit model possesses the independence of irrelevant alternatives property, according to which the ratio of the choice probabilities for any two alternatives i and k does not depend on any alternatives other than i and k. In many settings, this is not an accurate representation of reality. The main limitation of multinomial logit is its reliance on the normal distribution. In some situations, unobserved factors may not be normally distributed. In addition, the probit choice 14

15 probabilities are given by integrals that don t have closed forms and have to be evaluated numerically though simulation. In what follows, we present the results of our multinomial probit estimations only, even though we ran the multinomial logit model as well. The results were very similar, and the logit estimation outcomes can be obtained from the authors upon request. V. Results and discussion The results of our probit with sample selection for sector mobility conditional on crossing into the US are in Table 4. Age in its quadratic specification significantly affects both the probability of crossing and sector mobility. Younger individuals in the sample are both more likely to cross and less likely to be mobile while the reverse is true for older individuals. While gender does not significantly affect the probability of crossing, being male significantly increases the likelihood of choosing the sector of previous employment suggesting that females in our sample have greater sector mobility relative to males. Educational attainment also has interesting and at the same time intuitive impact. Individuals with higher educational attainment are significantly more likely to cross and to have higher sector mobility. The latter is also supported by the estimated effect of the language variable suggesting that people with spoken English skills are more likely to have higher sector mobility. Interestingly, the same is true for illiterate migrants but their very small percentage (approximately 5%) may explain the weak significance of this variable. In terms of rural versus urban origin, we find that being from a rural area significantly decreases the probability of crossing while having no statistically significant impact on labor mobility. As can be expected, having crossed into the US before increases the probability of crossing again which most likely reflects cyclical labor migration patterns. More importantly, having family in the US both increases the probability of crossing as well as makes it more likely that migrants will chose the sector of their 15

16 previous employment, suggesting lower sector mobility due to networking effects and perhaps family tradition. Finally, migrants who are not documented are both more likely to cross into the US and more likely to specify the same economic sector as their previous employment. This together with the results on family (see Figure 5) suggests that network effects (as proxied by family in the US) may be inversely related to sector mobility. The results in Table 4 also suggest that sample selection is not statistically a concern in our estimation. The rho for correlation between the residuals of the two probit regressions is not statistically significant at conventional levels. Thus we are justified in estimating the multinomial probit results presented in Table 5 without further sample selection correction. Focusing on each sector relative to the capture all category of Other we find the following results. Agriculture Relative to the other sector alternative, migrants who chose agriculture as the sector of their intended employment in the US are on average older, with lower educational attainment, less likely to speak English, earning lower average monthly wages in previous employment. Additional characteristics that make selection of agriculture more likely over other include being from a rural area, having crossed before (suggesting cyclical migrants), having family in the US, being documented and having lower sector mobility. Taken together, these are consistent with more established histories of and migration patterns in agriculture for Mexican migrants. Construction Relative to the other sector alternative migrants who chose construction as the sector of their intended employment in the US are on average older and significantly more likely to be male. While educational attainment in this alternative is not significant, literacy does seem to increase the likelihood that migrants will choose construction over the base alternative. Additional characteristics that make selection of construction more likely over other sectors include being from an urban area, having 16

17 crosses before (suggesting cyclical migrants), and traveling with fewer people toward the United States. Additionally, and similar to agriculture, we find that being documented and having lower sector mobility significantly increase the probability of choosing construction over the base alternative. Manufacturing Relative to the base alternative, and similar to the construction sector, migrants who chose the manufacturing sector as their sector of intended employment in the US are on average older but, unlike construction, being male significantly decreases the probability of choosing manufacturing over the base alternative. Unlike agriculture and construction, however, higher educational attainment and being able to speak English significantly increase the probability of choosing manufacturing over the base alternative. Migrants choosing manufacturing commanded higher than average wages in their previous employment, are more likely to be documented and to have crossed before. Additional characteristics that make selection of manufacturing more likely over other sectors include having crossed before (suggesting cyclical migrants) and traveling with fewer people to the US destination, as well as being documented and having lower sector mobility. Trade Transportation and Utilities (Commerce) Relative to the base alternative, migrants who chose commerce as their sector of intended employment are once again more likely to be older, but significantly less likely to be male relative to migrants choosing the base alternative. Similar to manufacturing, higher educational attainment, being able to speak English and having higher than average wages in their previous employment are all factors that significantly increase the probability of choosing commerce over the base alternative. They are also more likely to be documented and to have crossed before (suggesting cyclical migrants). Having family in the US and having lower sector mobility increases the likelihood of selecting commerce over the base alternative as well. 17

18 Services Relative to the base alternative, and similar to the other sectors above, older migrants are more likely to choose the services sector as their sector of intended employment in the US compared to the other sectors and, like manufacturing and commerce, being male significantly decreases the probability of choosing services over the base alternative. Unlike all the other sectors above, being married significantly increases the probability of choosing services over other sectors. Higher educational attainment and being able to speak English also significantly increase the probability of choosing services over the base alternative. Like construction, manufacturing and commerce, migrants choosing services commanded higher than average wages in their previous employment, are more likely to be documented, to have crossed before, and to be traveling with fewer people to the US destination. Unlike the other sector choices discussed above, two important differences of people choosing the services sector over the base alternative is that they are less likely to have family in the US and more likely to have higher sector mobility. We also computed the marginal effects of the regressors on choice probabilities. The effects we present in table 6 are averaged over all individuals, and should be interpreted as follows. For a continuous variable, the coefficient gives the partial derivative of the probability of choosing a given sector with respect to a change in that variable. For an indicator or dummy variable, it gives the difference between the probability of choosing a given sector when that variable takes on variable one and the probability when it takes on value 0. For instance, the coefficient of the variable DOCUMENTED is equal to for construction. This means that the probability a documented migrant chooses to work in the US construction sector is percentage points lower than that of a non documented migrant, other things being equal. Three variables worth discussing in more detail are RURAL ORIGIN, FAMILY IN THE US, and EMPLOYMENT SECTOR MATCH. Notice first that the marginal effects of RURAL ORIGIN are not 18

19 statistically significant for the services and other sectors. The coefficient for agriculture tells us that a migrant from a rural area is approximately 8 percentage points more likely to choose the agriculture sector than someone from an urban area. The coefficients for construction, industry, and commerce, however, are all negative. The probability a migrant from a rural area chooses to work in the US construction, industry, or commerce sectors is respectively 8, 0.6, and 0.5 percentage points lower than that of someone coming from an urban area. The marginal effects of FAMILY IN THE US are all statistically significant. They are positive for all sectors but services. Other things being equal, the probability that a migrant chooses to work in the agriculture sector, for instance, is 3.7 percentage points higher when the migrant has a family member living in the US than when he has none. The probability he chooses to work in the services sector, on the other hand, is 11.3 percentage points lower, other things being equal. Finally, the marginal effects of EMPLOYMENT SECTOR MATCH are also all significant. Notice the relatively high coefficient of this effect for agriculture. Other things being equal, someone who worked in the Mexican agriculture sector prior to being interviewed has a probability of seeking work in the US agriculture sector that is 48.8 percentage points higher than someone who worked in a different sector. The coefficient of services is also high in absolute value, meaning that someone who worked in the services sector before migrating has a probability of seeking work in the US services sector that is 39.1 percentage points lower than someone who worked in a different sector in Mexico. The interpretation of the remaining effects is analogous and left to the reader. VI. Conclusion and future research In this paper we use the 2011 EMIF data which has previously not been used to evaluate what migrant characteristics help explain the probability of crossing into the U.S. for employment. Conditional on selecting to cross, we evaluate the characteristics that help explain sector mobility, and then separately 19

20 investigate what migrant characteristics are most significantly associated with different sector choices. Key similarities and differences in the choice agriculture, construction, manufacturing, commerce, and service over the base category of other sectors can be summarized as follows. Migrants who are older, have crossed into the US before and are documented are consistently more likely to choose among the above 5 sectors over other. Some important differences are that higher wages in employment prior to migration significantly reduce the probability of choosing agriculture while an important difference for people choosing the Services sector over base alternative is that they are less likely to have family in the US and more likely to have higher sector mobility. In future research, the importance of destination characteristics may be a useful way to further investigate occupational choices of Mexican migrants. 20

21 References Ambrosini, Will and Giovanni Peri (2012) The Determinants and the Selection of Mexico US Migrations the World Economy, Wiley Blackwell, vol. 35(2), pp Belot, M. V., and T. J. Hatton (2008) Immigrant Selection in the OECD," The Australian National University Centre for Economic Policy Research Discussion Paper Series, 571. Borjas, G. J. (1987) Self selection and the Earnings of Immigrants, American Economic Review, 77, pp Borjas, George J. and Lawrence F. Katz The Evolution of the Mexican Born Workforce in the U.S. Labor Market, in Mexican Immigration to the United States, edited by George J. Borjas, University of Chicago Press, 2007, pp Bustamonte, Jorge, Guillermina Jasso, J. Edward Taylor, and Paz Trigueros Legarreta (1998) Characteristics of Migrants: Mexicans in the United States, in Migration between Mexico and the United States: Binational Study. Mexico City and Washington D.C.: Mexican Ministry of foreign affairs and U.S. Commission on Immigration Reform. Chiquiar, D., and G. H. Hanson (2005) International Migration, Self Selection, and the Distribution of Wages: Evidence from Mexico and the United States," Journal of Political Economy, 113(2), pp Giovanni Peri (2012) Rationalizing U.S. Immigration Policy: Reforms for Simplicity, Fairness, and Economic Growth Discussion paper , The Hamilton Project, Washington D.C., May Grogger, J., and G. H. Hanson (2008) Income Maximization and the Selection and Sorting of International Migrants," NBER Working Paper Series, Kaestner, R. and Malamud, O. (2010), Self Selection and International Migration: New Evidence from Mexico, NBER Working Papers (Cambridge, MA: National Bureau of Economic Research, Inc). McKenzie, D. J., and H. Rapoport (2010) Self selection patterns in Mexico U.S. migration: The role of migration networks," The Review of Economics and Statistics, vol. 92(4), pp Moraga, Jesús Fernández Huertas (2011) "New Evidence on Emigrant Selection," The Review of Economics and Statistics, Vol. 93, No. 1, pp

22 Nakamura, David (2013, March 22) Immigration reform impasse deepens between labor, business, The Washington Post, Washington, DC: The Washington Post Company. Orrenius, P. M. and Zavodny, M. (2005), Self selection among Undocumented Immigrants from Mexico, Journal of Development Economics, 78, pp Ortega, F., and G. Peri (2009) The Causes and Effects of International Labor Mobility: Evidence from OECD Countries ," NBER Working Paper Series, Passel, J., D. Cohn, and A. Gonzalez Barrera, (2012) Net Migration from Mexico Falls to Zero and Perhaps Less, Pew Hispanic Center: Washington, DC. April. Patel, Krishna, and Francis Vella (2012) "Immigrant Networks and Their Implications for Occupational Choice and Wages," Review of Economics and Statistics, forthcoming (doi: /rest_a_00327). Van de Ven, Wynand P.M.M, and Van Praag, Bernard M.S. (1981) The Demand for Deductibles in Private Health Insurance, Journal of Econometrics, 17, pp

23 Figure1. Educational attainment by level for the Full sample 4.4% 4.8% 14.5% 40.1% 36.2% No Education Junior High Postgrad/Professional Education Primary Education High School Figure 2. Mean Educational Attainment by Sector of Intended Work Mean of Education Level agriculture construction industry commerce services other family and friends 23

24 Figure 3. Mean of Previous Earnings by Sector of Intended Work mean of wage 0 1,000 2,000 3,000 4,000 agriculture construction industry commerce services other family and friends Figure 4. Mean Rural Origin by Sector of Intended Work mean of rural agriculture construction industry commerce services other family and friends 24

25 Figure 5. Mean Documented Workers and Family in US by Sector of Intended Work agriculture construction industry commerce services other family and friends mean of crossdoc mean of family 25

26 Table 1. Demographic Profile Full Sample Difference in Means by Gender Difference in Means by Cross Mean Median Min Max Male Female Cross No Cross Age *** *** 35.2 Gender *** 0.69 Marital Status *** 0.50 Educational Attainment *** *** 2.02 Speak English *** *** 0.11 Literacy *** 0.97 Monthly Wage (pesos) , *** *** Rural Origin *** *** 0.32 Intend to Cross Border *** Crossed Before *** *** 0.03 Number of People *** *** 4.49 N

27 Table 2. Occupational Profile of Mexican Migrants prior to migration Employment Experience Cross (Freq) No Cross (Freq) Full Sample (Freq) Age (Mean) Gender (Mean) Education (Mean) Wage (Mean) No Experience 1,705 1,506 3, Agriculture & Mining 5, , Construction 2, , Manufacturing Trade, Trans. & Utilities , Info. & Fin. Serv Prof. & Bus. serv Education & Health Leisure & Hospitality Other services Number of obs. 11,322 5,772 17,094 17,094 17,094 17,094 17,094 27

28 Table 3. Reported Economic Sector or Activity by Respondents intending to Cross into the United States Destination Economic Sector or Activity Employment Family and Agriculture Construction Manufacturing Trade and Services Other Total Experience Friends Transportation No Experience ,611 Agriculture & 169 2,185 1, , ,083 Mining Construction , ,181 Manufacturing Trade, Trans. & Utilities Info. & Fin Services Prof. & Bus services Education & Health Leisure & Hospitality Other services Total 1,281 2,787 2, , ,897 28

29 Table 4. Probit for Economic Sector Mobility with Sample Selection for Immigration determinants Outcome equation: Employment Sector Selection equation: Cross vs. No Cross Match vs. Not Age *** ( ) *** ( ) Age(Squared) *** ( ) *** ( ) Gender *** ( ) ( ) Marital Status ( ) ( ) Educational Attainment *** ( ) *** ( ) Monthly Wage (pesos) *** ( ) ( ) Number of People *** ( ) ( ) Speak English *** ( ) ( ) Literacy 0.114* ( ) ( ) Rural Origin ( ) ** ( ) Crossed Before ( ) *** ( ) Family in the US *** ( ) *** ( ) Documented *** *** ( ) Constant *** ( ) ( ) *** ( ) Occupational Dummies No yes Reason for Migration Dummies No yes athrho ( ) rho ( ) LR test (rho = 0): chi2 stat (prob) 1.89 (0.1697) Number of Obs. 16,669 Censored Obs. 5,772 Uncensored Obs. 10,897 29

30 Table 5. Multinomial Probit Results for Sector Choice agriculture Construction Manufacturing Trade/Tran/Util services Age *** ( ) *** ( ) *** ( ) ** ( ) *** ( ) Age(Squared) *** ( ) *** ( ) *** ( ) ** ( ) *** ( ) Gender ( ) *** ( ) ** ( ) *** ( ) *** ( ) Marital Status ( ) ( ) ( ) ( ) ** ( ) Educational Attainment *** ( ) ( ) ** ( ) ** ( ) ** ( ) Speak English ( ) ( ) *** ( ) *** ( ) *** ( ) Literacy ( ) * ( ) ( ) ( ) ( ) Monthly Wage (pesos) ** ( ) *** ( ) *** ( ) *** ( ) *** ( ) Rural Origin *** ( ) ** ( ) ( ) ( ) ( ) Crossed Before *** ( ) *** ( ) *** ( ) ** ( ) *** ( ) Number of People ( ) *** ( ) ** ( ) ( ) * ( ) Family in the US ** ( ) ( ) ** ( ) * ( ) *** ( ) Documented *** ( ) ** ( ) *** ( ) *** ( ) ** ( ) Employment Sector Match *** ( ) *** (0.0834) *** ( ) *** ( ) *** ( ) Constant *** ( ) ( ) ( ) ( ) *** ( ) 30