Experimental Design
We will recruit a random sample of individuals outside betting shops in Dar es Salaam (Tanzania) and we will randomly split them into six groups.
Group A1 will receive an endowment to place a bet on the outcome of an English Premier League game to be played the same evening. The bet will be placed with the research team (not with a betting company) and it will be made clear to respondents that the exercise is conducted for research purposes. The odds associated with each possible outcome of the match (Team 1 wins, Draw, or Team 2 wins) will be obtained from a major online betting company and shown to respondents. In all respects, this bet will be akin to the ones betters are used to place in the betting shop. Crucially, after placing the bet, participants will be asked a question to elicit their subjective probabilities of winning the bet and of each outcome occurring, as well as their certainty equivalent (as outlined below).
Group A2 will face the same conditions as Group A1 and will be asked the same questions to elicit subjective probabilities and certainty equivalents. Instead of choosing which outcome to bet on, however, subjects in this group will be given a ticket for an already-placed bet on one of the three game outcomes.
Groups A3 and A4 will face the same conditions as Groups A1 and A2, respectively. However, in addition to the game odds normally provided by betting companies, subjects in these two groups will be given the probabilities that each game outcome will occur implied by the odds (with the objective to test whether this affects the subjective probabilities assigned to the three outcomes, the perceived probability of winning, and the certainty equivalent of the bet).
Group B1 will receive the same endowment as Group A1 and will be asked to bet on the outcome of a neutral lottery framed as a simple draw from an urn containing 100 balls of three different colors in known proportions. They will then be asked to bet on which color will come out upon randomly drawing a ball from the urn. Each color will be associated with a different reward. Crucially, the number of balls and the associated rewards will be set to exactly replicate the odds presented to participants in Group A1. As to the participants in the first group, we will ask Group B a question to elicit their subjective probabilities of winning the bet and of each outcome occurring.
Group B2 will face the same conditions as Group B1 and will be asked the same questions to elicit subjective probabilities and certainty equivalents. Instead of choosing which color to bet on, however, subjects in this group will be given a ticket for an already-placed bet on one of the three colors.
By comparing subjective probabilities with actual odds in the framed vs unframed lottery, we will be able to test whether potential misalignments are affected by the framing. This will be a direct test of our core hypothesis. We will, however, take several additional steps to test competing mechanisms.
Most importantly, we will test whether people’s tolerance to risk differs when a lottery is framed as a sports bet. In order to do this, we will need a measure of people’s risk preferences, which we will elicit as follows. After placing the bets but before the outcome of the lottery is revealed, subjects in all groups will be shown a list of potential buy-back prices and asked to choose for each price whether they would prefer to sell the lottery ticket or keep it and await the outcome of the match or urn draw. This will allow us to elicit a certainty equivalent and, therefore, it will provide us with an indication of their risk tolerance. To ensure truthful revelation of preferences, we will incentivize this question by telling respondents that after they make their choices, a random price will be drawn. If they stated that they would sell the ticket at that price, they will receive this amount and "hand in" their lottery ticket.
The experiment will be conducted on tablets and it will be accompanied by a short survey to gather key socio-demographic information about the respondents and additional data to explore important dimensions of heterogeneity and investigate specific mechanisms.
The outcomes of the bets will be revealed to respondents in Group A the same evening when the relevant football match will take place. Respondents in Group B will find out the outcome of the lottery immediately after the end of the interview. Both groups, however, will be informed from the beginning of the experiment that the winnings from the lotteries will be paid by the end of the day through mobile money. At the end of the interview, all respondents will receive a show-up fee.