Attrition is the Achilles heel of longitudinal surveys. This award funds research that will evaluate different methods of tracking and interviewing a group of individuals who are respondents who have attrited from the third wave of the Mexican Family Life Survey (MxFLS), an on-going population-representative longitudinal survey of Mexicans. A unique feature of the MxFLS is that Mexican migrants to the U.S. are followed; the survey includes data on these migrants. In this research project, the coPI will choose a sub-sample of these migrants who are no longer participating in the survey effort. She will locate these participants (using information from their original Mexican communities to locate them) and will interview them.
Information from the interviews will be combined with the survey data from other respondents who continued to participate in the survey. The result will be a uniquely rich analytical sample that fully represents the population of international migrants from Mexico to the US since 2002. The data will be publicly available. The co PI will use the data to test whether attrition can be predicted by personal characteristics, and , importantly, whether attrition can be explained by changes in the lives of respondents that occurred after the baseline survey interview. These results will allow her to evaluate the accuracy of standard post survey adjustments that are widely used to correct for the effects of attrition, including reweighting and multiple imputation methods. Because the impact of attrition on estimation and inference will be model-specific, the comparisons can be conducted in the context of widely-used empirical models of economic behavior.
Survey methods are widely used by governments to collect information to inform policy. Therefore the broader impacts of this research include possible improvements in the methods used to conduct and analyse these surveys. The research will identify cost-effective ways to find and interview the hardest to find respondents.
