Estimating Income Effects Using a Sample of Lottery Players

点击翻译按钮获取中文摘要

9812057 Imbens A previous project (NSF grant number SBR9423018) collected data on people who played the Massachusetts lottery, including those who won small onetime prizes and people who won large prizes paid out over twenty years. This project uses these data to compare people who played and won large amounts of money in the lottery, "winners", with people who played the lottery but did not win large amounts of money "losers"; the points for comparison include their labor supply, consumption and savings behavior, and the educational achievement of, and expenditures for, their children. The critical reason for collecting lottery data was to exploit the randomization in the lottery to reconstruct a randomized experiment with an infusion of income as the treatment condition. For three reasons, namely the response rate in the survey was 50%, the probability of winning the lottery depends on the number of tickets bought (a variable which is only imperfectly observed), and the winners were drawn from a different population than the losers, there are pretreatment differences between the various groups. The investigators pursue a variety of methods to deal with adjustments to make the subsequent comparisons more credible. Preliminary results suggest that adjusting for pretreatment differences leads to credible estimates of the causal effect of income on labor supply.