The Economics of Women Entrepreneurship: Intrahousehold Dynamics

We propose to partner with an NGO that has successfully developed and implemented a Street Business School (SBS) program in Uganda for more than ten years. The program consists of several training modules that take place over the course of six months in a local training center. In the different modules, women learn how to identify a market niche according to their talents, how to find funding for starting a business, hear the stories of successful alumni of the program, and learn basic skills on how to run their businesses. The mentoring visits and office hours provide both technical knowledge taught in the modules as well as emotional support to increase perseverance and grit. After six months all women in the two treatment groups are invited to an official graduation ceremony where they receive a diploma by the NGO. Women in the control group do not receive any type of training but get invited to participate in two networking events. The first one takes place simultaneously to the first day of training for the treatment groups, the second on the day of graduation. As the trainings take place in a fixed group of women on a regular basis, providing control women with an opportunity to network with their peers controls for treatment effects due to networking rather than the content of the programe modules.

2018-09-01

2020-06-30

1. Financial information and decision-making within the household:
(a) Difference between man and woman's estimate of man's earnings.
(b) Differencce between man and woman's estimate of woman's earnings.
(c) Difference between man and woman's estimate of man's contributions to household expenses (including rent).
(d) Differencce between man and woman's estimate of woman's contributions to household expenses (including rent).
(e) Difference between man and woman's estimate of other household members' earnings.
(f) Likelihood of the spouses having any joint account at a bank, SACCO, etc.
(g) Who decides on use of savings.
(h) Who decides on use of man's earnings.
(i) Who decides on use of woman's earnings.
(j) Who decides on woman working.

2. Spousal psychometric measures:
(a) Locus of control - internality: aggregate score of questions 3, 11, 12, and 14 of section question G1.
(b) Locus of control - powerful others: aggregate score of questions 2, 7, 8, 10, and 13 of section question G1.
(c) Locus of control - chance scale: aggregate score of questions 1, 4, 5, 6, and 9 of section question G1.
(d) Self-efficacy: aggregate score of section question G2.
(e) Grit: aggregate score of section question G3.
(f) Income aspirations: percentage increase of monthly income in 10 years relative to current monthly income (G4).
(g) Education aspirations: level difference of children's highest level of schooling relative to own level of education (G5).
(h) Social status aspirations: level difference of social status in 10 years relative to social status today (G6).

3. Gender attitudes
(a) Gender attitudes index, own beliefs: aggregate score of section I7.
(b) Gender attitudes index, beliefs about what others think: aggregate score of section I8.
(c) Wife attitudes index, own beliefs: aggregate score of section I9.
(d) Wife attitudes index, beliefs about what others think: aggregate score of section I10.
(e) Intimate partner violence index: aggregate score of section I11.
(f) Sex refusal index: aggregate score of section I12.
(g) Controlling behavior index, own beliefs: aggregate score of section I12a.
(h) Controlling behavior index, beliefs about what others think: aggregate score of section I12b.
(i) Abuse index: aggregate score of section I13.
(j) Ever been abused.
(k) Abused in last 12 months.
(l) Ever abused spouse.
(m) Abused spouse in last 12 months.

4. Game outcomes
(a) Likelihood of choosing dedicated consumption over cash in the private decision.
(b) Likelihood of switching the choice for dedicated consumption vs cash in the public and bargained decisions.
(c) Willingness to pay for joint resources (amount contributed to the common pot) with returns of 150%, 200%, and 300%.
(d) Difference between contribution to the joint pot under public versus private decision.
(e) Difference between contribution to the joint pot under bargained versus private decision.
(f) Difference between contribution to the joint pot under bargained versus public decision.

5. Bargaining process
(a) Time spent bargaining.
(b) Likelihood of agreeing right away with no discussion.
(c) Likelihood of disagreeing or trying to convince one another to change decisions.
(d) Likelihood that the woman convinces her partner to take her preferred decision or insists that he takes her preferred decision.
(e) Likelihood that the woman convinces her partner to take her preferred decision or insists that he takes her preferred decision.
(f) Likelihood that the couple dicsusses gender, household roles, or earnings during negotiation.
(g) Likelihood that the woman speaks first.
(h) Likelihood that the woman speaks at any time during the negotiation.
(i) Likelihood that the couple cannot reach a consensus (e.g., they agree to disagree).

Over the course of this project we will collect detailed business and household data from each woman in the main RCT three times: prior to the start of the program in each location, return after six months to measure immediate effects from the treatments, and return again 2 years after the baseline data collection, i.e. one and a half years after the completion of the treatments. We survey spouses and conduct the lab in the field experiments two to three months prior to the endline survey in each location. We will not obtain survey data for women that have not been part of the main experimental sample but who's spouses have been sampled in addition. Data on these women will be collected during the lab in the field experiment and through the spouses only.

The surveys for participating women will consist of five modules. The first is a module on household characteristics and the socio-economic background of women. This will ensure balance among the control and treatment groups and enable us to measure changes in these characteristics as a result of the treatments. The second is a module on household consumption decisions including information on the overall contributions of household members to household income as well as expenditures of various consumption categories. The third is a module on business outcomes consisting of established measures on sales and profits, business practices, investment decisions, and expectations about future business growth, profits, and variability. Fourth, we will collect detailed data on psychometric indicators similar to the aspirations measures established by Bernard et al. (2014) as well as on further measures that build on Lajaaj and Marcours (2017), and administer a brief quiz on technical business skills based on what the women should have learned during SBS. Finally, to measure spillover effects within locations we will obtain detailed network data between participants.

The spousal survey consists of four modules. The first resembles the initial module in the women's survey and concerns household characteristics, the socioeconomic status of the spouse, and beliefs about relative contributions to household expenses of various household members. The second addresses the income generating activities being pursued by the spouse, including any business that the spouse runs as well as savings behavior. The third resembles the psychometric indicators that we also collect from the women. Finally, we have a section on gender attitudes including information on self-reported conflict within the household as well as beliefs about the prevalence of abusive behavior among peer households.

The lab in the field games consist of a series of four choices that each member of a given couple makes in different information environments. The first choice is a hypothetical one where the respondent chooses between cash, a voucher of an equivalent amount for food, or a voucher of an equivalent amount for clothes. In the second choice, the participant has to decide how much of an endowment to keep for themself versus contribute to a common pot with their spouse, where money in the joint pot gets multiplied by 1.5. The third and fourth choices are the same as the second choice except that the pot gets multiplied by 2 and 3, respectively. In the first round, each member of the couple makes decisions privately, without their spouse's knowledge. In the second round each member makes the same set of choices with the knowledge that their decisions will be shared with their spouse. In the final round, the couple is allowed to discuss and bargain over the decision. We further collect qualitative data about the bargaining process in the third round. At the end of the game, we select one game to be paid out to the couple, excluding the games with vouchers that are only hypothetical choices. To preserve the privacy of decisions made in the first round, we add a random amount to the payout and couples are not informed which decision is being paid out.

Randomization through private lottery with respondent after baseline survey.

Household female representative.

No

Our sampling frame consists of four of the five total locations where we worked for the initial intervention. In these four locations, we identify all married or partnered women in our original sample and interview their husbands/partners. We estimate that this will yield approximately 650-700 couples. In addition, we randomly select an additional group of households to expand the sample by approximately 30%. These households consist of women who did not self-select into the initial study and their spouses. In total, our sample will consist of just over 900 households.

We will collect survey data for approximately 650-700 women from the main experimental sample, and around 900 spouses. We will collect experimental lab data for around 900 women and 900 spouses.

We have approximately 200-250 households in the pure control group of couples whose women did not sign up for our program in the first place, 200-250 households in the control group of women that demonstrated interest in the program but were randomized out, and 450-550 households in the treatment group.

If we assume normally distributed data with mean zero and standard deviation one, our power calculations estimate that we will be able to detect direct effects of around 0.2 standard deviations.