Data and Descriptive Statistics

does not result from the selectivity of return migrants, but from the endogeneity of accumulated migration experience.

V. Data and Descriptive Statistics

The Mexican Migration Project MMP is a collaborative research project based at the Princeton University and the University of Guadalajara. 7 The data collected by the MMP present researchers with the unique opportunity to observe earnings for particular individuals along with detailed migration histories. Each year, the MMP selects a group of Mexican communities and surveys a random sample of the households in each location. The MMP survey collects demographic and economic data on households and individuals, with a particular emphasis on migration experi- ence. The survey also requests a detailed, self- reported life history from household heads recording some economic, demographic, and migration variables for every year in their lives. These life histories record whether an individual migrated in a given year, how many months an individual spent in the United States, and what documents, if any, were used to migrate. Our sample consists of male household heads aged 18–65 who were surveyed during the years 1987–2009, and who were in Mexico at the time of the interview. The MMP asks each household head to report his or her current occupation, and their income in Mexico. Although the income variable is reported at different rates weekly, biweekly, monthly, we convert all income measures to monthly values throughout the paper. In order to interpret the income variable as a measure of earn- ings we only consider those individuals who are currently employed, and who are not self- employed and do not own a business. After imposing these restrictions, and dropping those individuals with missing values for important regressors, we also trim the data by dropping individuals in the top and bottom 1 percent of the monthly earnings distribution, and those return migrants in the top 1 percent of the migra- tion experience distribution where migration experience is measured in years. After making these restrictions, our full sample consists of 6,200 men. For the descriptive statistics and throughout the paper, we defl ate earnings using CPI indices for Mexico and the United States 2000 base year taken from the IMF’s International Financial Statistics series. The appendix provides a more thorough description of the data and the sample selection criteria. We present summary statistics of log earnings and important characteristics of migrants and nonmigrants in Table 1. The fi rst pair of columns displays statistics for the full sample of individuals interviewed while in Mexico. This includes non- migrants and return migrants, but excludes individuals who have migrated but have not returned to Mexico. The average age for the full Mexican sample is about 40, and the average level of education is about 6.8 years. Most individuals are married 89 percent. A substantial fraction of individuals have some experience migrating to the United States 28.0 percent. The average log of real monthly earnings is about 8, while the average real monthly earnings in levels is roughly 3,596 pesos per month. The second pair of columns reports summary statistics for nonmigrants, while the 7. See the MMP website at mmp.opr.princeton.edu last pair of columns reports summary statistics for return migrants. Return migrants tend to be less educated, with an average number of years of education 5.54 that is more than a year and a half lower than the average education level of nonmigrants 7.29. This pattern is consistent the idea that return migrants are negatively selected compared to nonmigrants. The average earnings of return migrants is about 5 percent lower than the average earnings of nonmigrants. Table 1 also presents some summary statistics related to the migration experience of return migrants. About 91 percent of return migrants have some experience as an undocumented migrant, while only 20 percent have any documented migration experi- ence. The migrants in our sample are thus predominantly engaging in undocumented migration. The average return migrant has accumulated about 2.63 years of experience in the United States, with about 2.18 years of undocumented migration experience, and about 0.45 years of documented experience. The model presented in Section II suggests that if skill is negatively correlated with the Mexico- US wage gap but unrelated to other model parameters, then un- der quite general conditions there should be a negative relationship between skill and the accumulated migration experience of return migrants. Such a negative correlation would induce a downward bias in the OLS estimate of the return to a year of migration experience. To test this implication of the theory, we examine the relationship between education an observable component of skill and the accumulated migration experience of return migrants in Table 2. We are implicitly assuming that the pattern of correlation between observables and migration expe- rience is informative about the correlation between unobservables and migration experience. In Column 1, we regress migration experience on education and other controls. The point estimate suggests a weak negative relationship between educa- Table 1 Summary statistics Entire Sample Nonmigrants Return Migrants Mean Standard Deviation Mean Standard Deviation Mean Standard Deviation Age 40.44 11.53 40.30 11.60 40.80 11.32 Education 6.81 4.77 7.29 4.92 5.54 4.10 Married 0.89 0.31 0.87 0.33 0.93 0.25 Log earnings 8.00 0.59 8.02 0.58 7.97 0.62 Ever migrant 0.28 0.45 Ever undocumented 0.91 0.28 Ever documented 0.20 0.40 USExp 2.63 3.18 Undocumented experience 2.18 2.69 Documented experience 0.45 1.60 Observations 6,200 4,527 1,673 tion and migration experience. A one- year increase in education is associated with a decline in accumulated migration experience of –0.024 years. This relationship is not statistically signifi cant. To explore this a bit further, we add a quadratic term in education as a regressor in Column 2. Neither education coeffi cient is precisely estimated, but the point estimates suggest a nonmonotonic relationship, with migration experience rising with education up to about fi ve years, and there- after declining. For very low levels of education it thus appears that accumulated migration experience is positively associated with migration experience. However, we emphasize that this appears to hold only at the lowest levels of education. In Column 3 of Table 2, we restrict the sample to those with at least three years of ed- ucation, a group containing about 75 percent of the migrants in our main sample. In this subpopulation, there is a statistically signifi cant, negative relationship between education and accumulated migration experience. An additional year of education here is associated with a reduction in accumulated migration experience of about 0.0562 years. The prediction of a negative relationship between migration experience and edu- cation weakly holds in the entire sample, and holds with much greater strength for the mass of individuals with three or more years of education. At the very least, we certainly do not fi nd clear evidence of a positive relationship between education and migration experience. Even the negative correlations that we fi nd are rather small in magnitude, suggesting that the endogeneity of accumulated migration experience probably explains little of any correlation between migration experience and Mexican income. Table 2 Migration Experience and Education 1 2 3 Age 0.202 0.210 0.224 0.055 0.055 0.069 Age2 –0.226 –0.229 –0.243 0.068 0.068 0.089 Educ –0.024 0.082 –0.056 0.032 0.102 0.031 Educ 2 –0.007 0.005 Married –0.352 –0.360 –0.817 0.619 0.609 0.759 Sample Complete Complete Educ ≥ 3 Observations 1,673 1,673 1,262 R 2 0.128 0.130 0.165 Note: Stars signify the following: signifi cant at the 0.01 level, signifi cant at the 0.05 level, signifi cant at the 0.1 level. Standard errors are reported in parentheses. All specifi cations include MMP community ef- fects which subsume time effects.

VI. Empirical Results