The Costs and Benefits of Longitudinal Data: Three Applications from the Mexican Family Life Survey
Longitudinal surveys have revolutionized empirical research and our understanding of the dynamic processes that affect the economic prosperity, health and well-being of the population. This dissertation explores and provides evidence, through three empirical applications, on the costs and benefits of designing, implementing and using data from a new, innovative longitudinal survey, the Mexican Family Life Survey (MxFLS). The survey, which is representative of the Mexican population living in Mexico in 2002, is designed to follow movers within Mexico and also those who move to the United States. This design lies at the center of the contributions of my research to the scientific literature.
Attrition is the Achilles heel of longitudinal surveys. The first essay of the dissertation focuses on the cost of attrition for scientific knowledge. Following the same individual through time allows a researcher to trace the evolution of a respondent's behaviors and outcomes in a dynamic framework; however, if attrition is selected on unobserved characteristics, the advantage of using panel data could be severely hindered. Exploring different methods to adjust for attrition, this essay provides evidence of limitations of standard post-survey adjustments strategies that are the standard in the literature. These approaches, exploit only baseline characteristics of the respondents and, conditional on those characteristics, treat attriters as missing at random. I provide evidence that this assumption is substantively important and rejected in the MxFLS in spite of the fact that attrition in that survey is low relative to other nationally-representative surveys conducted in the United States and abroad.
The second essay in this dissertation exploits the fact that MxFLS follows movers within Mexico and those who move across the Mexico-US border to provide new insights into the mechanisms that underlie the selectivity of migrants within Mexico, how they differ from migrants who move from Mexico to the U.S. and how those who return contrast with the migrants who remain in the U.S. more permanently. The results provide evidence that human capital is predictive of migration within Mexico and to the United States, but that there is little indication that the decision to stay in the United States is highly correlated with education. In contrast, having relatives in the United States is not only a powerful predictor of migration to the United States, but it is also predictive of successful economic assimilation.
The third essay exploits a different dimension of the longitudinal survey in order to address an important question regarding the impact of unanticipated crime and violence on population well-being. To wit, the essay rigorously examines the impact of the recent surge in violent crime in Mexico on the labor market outcomes, migration, and wealth of the Mexican population. The timing of the last two waves of the MxFLS paired with the panel nature of the survey, allows the comparison of outcomes of the same individual in periods of low and high violence, which removes the potentially endogenous time-invariant unobserved heterogeneity between respondents. Moreover, due to the fact that the MxFLS was designed to follow migrant respondents, this study is able to directly test whether there is a systematic migratory response to crime. The results from this analysis find that crime predicts migration and it negatively affects the labor outcomes of self-employed individuals. In addition, the negative effects on the labor outcomes have translated into reductions in per capita expenditure at the household level, which suggests that the recent wave of violence in Mexico may have long-term consequences on the wealth and well-being of Mexican households.
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 3.0 United States License.
Rights for Collection: Duke Dissertations