Abstract
I will investigate whether individuals’ decisions to participate in a lottery raffle (where, out of all tickets bought, n are selected to win a prize) are influenced by how many other people play. Using machine learning techniques, I aim to identify the types of people who perceive lottery participation as a game of strategic substitutes (as classical gambling theory predicts, where fewer players mean better odds) or as strategic complements (as suggested by the occurrence of mega-lotteries and social interactions, where more players make the activity more appealing), and whether this depends on the prize size.
I will collaborate with a commercial lottery in Ghana that runs weekly raffles and jackpots on different radio and TV stations. Individuals who have previously bought a ticket from a given station on a specific weekday will be randomized into a treatment and a control group. First, everyone will be asked to provide their prior beliefs about the number of tickets sold in the particular raffle in which they participated. The treatment group will then receive accurate information and the control group will not receive any information. After this, all participants will be asked to estimate how many people they believe will participate in the next upcoming lottery draw from a given station. My main outcome variable will be lottery participation after the survey, split by whether the person initially underestimated or overestimated the true number of tickets sold, and further by demographic information such as past winning history, play frequency, and prize size/lottery type.