**Principal Investigator(s):** **Hunt Alcott** Microsoft Research Email: allcott@gmail.com Home page: https://sites.google.com/site/allcott/home **Benjamin Lockwood** University of Pennsylvania Email: ben.lockwood@wharton.upenn.edu Home page: https://benlockwood.com/ **Dmitry Taubinsky** University of California, Berkeley Email: dmitry.taubinsky@berkeley.edu Home page: https://www.dmitrytaubinsky.com/ **Sample size**: 3013 **Field period**: 10/02/2018-05/05/2020 **Abstract** People have long debated whether state-run lotteries exploit the poor or are a win-win that generates enjoyment and government revenues. We study socially optimal lottery design in an optimal taxation framework with biased consumers and estimate sufficient statistics for optimal policy. Lottery sales respond more to changes in jackpot expected values than to changes in price or lower prizes, consistent with a specific type of probability weighting. In our survey, bias proxies such as innumeracy decline with income and explain 43 percent of lottery spending. In our model, current multi-state lottery designs increase welfare but may harm heavy-spending low-income people. **Hypotheses** Lottery expenditures will vary with income. Proxies for behavioral biases will be correlated with lottery spending, conditional on respondent demographics and measures of preferences. (See working paper for details.) **Experimental Manipulations** Please see working paper, section 4. **Outcomes** Measures of lottery expenditures; measures of proxies for behavioral biases including self-control, overconfidence, and numeracy. **Summary of Results** Please see working paper for full details. Key findings: lottery expenditures decline across the income distribution, and are predicted by bias proxies for self-control, financial illiteracy, and statistical mistakes, but not by overconfidence or predicted life satisfaction. **References** https://www.nber.org/papers/w28975