Household Recombination, Retrospective Evaluation, and Educational Mobility over 40 Years

Analysis of the long term economic mobility using longitudinal household survey data requires an understanding of patterns of household recombination--that is the processes by which households divide and fuse over time. In this paper we examine, in particular, educational mobility at the household and village level in Matlab Bangladesh over a 40-year period. Using a novel resampling procedure that relies on longitudinally collected demographic surveillance data, we correct for biases that arise from household recombination that occurred subsequent to a baseline census but prior to the collection of the first round of detailed economic data. Our results suggest that household advantages persisted over three generations but that conditional on household advantage, village educational differentials persist only over the first 20 years. We also show that approximate corrections for these biases using more limited data are reasonably effective. #4819