Experimental Design
The objectives of this project are to identify to what extent preferences for risk reduction, intrahousehold dynamics, and community-level norms underly the key empirical results from our initial RCT. These results are
1. Women quickly reinvest profits in additional businesses rather than growing their main business.
2. More intensive mentoring does not improve program outcomes, and in some cases worsens them. In addition, women qualitatively express a preference for less intensive, opt-in mentoring.
3. The program leads to reductions in intimate partner violence.
To identify the role of the potential mechanisms that we have identified (risk preferences, intra-household dynamics, and community norms), we propose to combine survey data with a series of lab in the field games designed to mimic real world decision-making.
Specifically, we plan to supplement our analysis with games that elicit risk preferences and a short survey module that clarifies the relationship between a woman’s businesses. In the games, we will first perform a direct, incentivized risk preference elicitation (Charness et al (2013)). We will then present each woman with a series of investment portfolios with different risk profiles. Some portfolios will contain investments that all pay off in similar states of the world while others will contain investments that pay off in different states of the world, thereby lowering risk. Observing which portfolios the women choose will allow us to better understand each woman’s preference for risk reduction in her investment decisions.
We plan to assess the role that intrahousehold factors play in women’s investment decisions using a spousal survey and a series of games between couples. Our sample will consist of three groups of spouses: those who are married to women who participated in the program, those who are married to women who showed interest in enrolling in the program (women in the control group), and those who are married to women who did not express interest in enrolling in the program. We will be asking men about their own earnings, their contributions to household expenses, their beliefs about their wives’ earnings and contributions to household expenses, and a similar set of psychometric
measures to what we use in the women’s survey. By combining these surveys with analogous responses from the women in our sample, we will be able to understand what information is public between couples versus private to individuals. The spousal survey will also include questions from the Demographic and Health Survey regarding the number of times men have engaged in different forms of intimate partner violence (IPV), their beliefs about the acceptability of IPV in different situations, and a version of the World Health Organization’s Domestic Violence module to measure men’s own experience of IPV.
We will invite women and their spouses to participate in a lab in the field experiment to collect experimental data on couples’ cooperative behavior and bargaining mechanisms. These additional measures will allow us to elicit the importance of the ability to hide independent income from spouses for household welfare and link real-world data to experimental evidence from a controlled lab setting. Like Ashraf (2009) and Iverson et al (2006), couples will receive a sum of money that they can choose to deposit or consume. We will vary the degree of observability of individual endowments, the possibility to communicate and discuss allocations, and the weight of each participant’s decisions to obtain measures of women’s bargaining power within the household and to observe what happens when we allocate greater bargaining power to the woman.
We propose to build upon our existing data in Uganda by conducting further interviews with the women in our sample. In total, we have we have conducted in-person, SMS, and phone surveys with 940 women since 2016. We have readily available, detailed data on women’s businesses, investment history, and their business practices, as well as information on their households and psychometrics. We will complement these data by conducting surveys with all women in our original sample to obtain up-to-date information on businesses and investment behavior and to elicit beliefs about the existence of witches in their location, their effectiveness, price, and services, second-order beliefs about whether other members of their community believe in and use the services of witches, and women’s past experiences of being bewitched. Additionally, we will elicit beliefs related to zero-sum mentality such as beliefs about the impact of others’ success for themselves, second-order beliefs, and beliefs over potential retaliations for being successful. In Uganda, anecdotal evidence suggests that witchcraft may be feared as a tool for retaliation. To understand the market for witchcraft in detail, we will conduct in-depth interviews with individuals that run witchcraft businesses in our sample locations.
We plan to complement the survey data with a series of four games with both the women in our sample and others in their community. The first two games will be a standard dictator game and ultimatum game, designed to measure altruism and inequality aversion (Forsythe et al (1994), Güth et al (1982)).
The third will be a game designed to capture both the existence of a zero-sum mentality and individuals’ perceptions of others’ zero-sum mentality. To do so, we will modify the classic joy of destruction game (Abbik and Sadrieh (2009)). Two players will each privately receive an endowment and learn that in the game, they and their opponent will simultaneously get to decide how much of the others’ endowment to destroy. However, prior to having the endowments revealed to the other player, each will have the option to pay to hide part or all of their endowment. If they do so, then only the part of the endowment that is not hidden will go into play during the remainder of the joy of destruction game. Participants’ willingness to pay to hide their endowment will reveal their beliefs about others’ zero-sum mentality, and final decisions in the subsequent joy of destruction game will reveal the actual prevalence of zero-sum mentality.
The fourth game will further modify the joy of destruction to model real-world decision making. In the fourth game, we will give the woman an opportunity to invest her endowment in an account with a high expected return but where the resulting return will be visible to her opponent, or an account with a lower expected return that can remain private to the individual. In turn, the opponent will be given the option to pay to destroy part of the observable return. We will vary the price of destruction for the opponent as well as the magnitude of the difference in the high value and low value investments.