Risk aversion and HIV/AIDS: Evidence from Senegalese female sex workers

HIV/AIDS is the second cause of mortality globally and there are 5000 new infections each day. Globally, sex workers are 13 times more at risk of HIV than the general population and in Senegal they have an HIV prevalence 16.5 times greater. Therefore, it is urgent to encourage behaviour change, which requires a better understanding of the reasons why sex workers engage in risky behaviours. We provide new evidence of the role of risk preferences on sexual behaviours, health behaviours and health outcomes of 600 female sex workers in Senegal in July and August 2017. We measure risk aversion of sex workers using an incentivised Gneezy and Potters task in addition to specific risk-taking scales in four domains (in general, finance, health and sex). Understanding of the experimental task was high despite low literacy level of participants. Using ordinary least squares, we find that risk aversion is an important predictor of sex workers’ sexual behaviours. We find that sex workers with higher level of risk aversion have less sex acts with clients, have less clients at risk of HIV, are more likely to engage in protected sex acts and as a result earn less money per sex act. Furthermore, we find that sex workers exhibiting higher level of risk aversion are less likely to be infected with sexually transmitted infections. Results highlight that some associations between risk preferences and sexual and health behaviours are domain specific. To conclude, our results confirm the role of risk preferences in the spread of HIV/AIDS epidemic and suggest the importance of collecting information on self-reported risk aversion to identify individuals who are at a greater risk of HIV/AIDS. Finally, our results provide some rationale in using lottery-based financial incentives to prevent sexually transmitted infections and HIV/AIDS among high-risk populations.

For instance, let's imagine that you decide to put CFAF 1,000 in this business. Your earnings will amount to CFAF 2,000 if you lose -if you draw the white ball with the black cross (3000 − 1000). Your earnings will amount to CFAF 4,500 if you win -if you draw the white ball (2, 000 + 2, 500). In other words, you are sure to keep the amount of money that you won't invest in the small business but you may increase your earnings if you invest in this business. The table below presents the possible gains.
We will first start by doing a training round in order to make sure that you well understood this task. You will then randomly draw a ball among four balls places in a black bag. Two of these balls are white with a black cross and two of them are just white. It is thus as likely that you draw a white ball with a cross or a white ball.
QUESTION: How much money are you willing to invest in this task? INTERVIEWER: Once the respondent made her choice, please ask her what are the amounts she will earn with this choice. This in order to be sure she well understood the decision she took.
QUESTION: On a scale going from 0 to 100, in your view, what is the probability that you draw a white ball? For those who do not say 50, INTERVIEWER: Remind the participant that she is going to draw a ball in the bag and that the probability of winning the amount does not depend on her expertise in running a business but that there is a one in two chance that she wins and a chance out of two that she loses.
READ: "You will now randomly draw a ball in this black bag in order to know what is the amount you would have won if this task is drawn and if this was not a training round. You have drawn out: White ball with a black cross on it/ White ball INTERVIEWER : Ask the respondent which amount she would have earned if this was the real task and report the amount.
We will now proceed with the real task. QUESTION:How much money are you willing to invest in this task? QUESTION: On a scale going from 0 to 100, in your view, what is the probability that you draw a white ball?
READ: "You will now randomly draw a ball in this black bag in order to know what is the amount you would have won if this task is. You have drawn out: White ball with a black cross on it/ White ball " INTERVIEWER: Ask the respondent which amount she thinks she has earned and report the amount. INTERVIEWER: Note and announce how much the respondent won. INTERVIEWER: Ask the supervisor to come and proceed with the payment.

Appendix 2 -List experiment
The principle of the list experiment is to allocate respondents randomly to two different groups: a "control" and a "treatment" group. Individuals allocated to the "control" group are presented with a number of non-sensitive statements. They are not asked to say whether they agree on each of the statements but only with how many of them they agree on. The same statements are presented to the "treated" group; the difference is that a sensitive statement is added to the series of non-sensitive statements. Assuming that the two groups have a similar opinion of the non-sensitive statements, one can deduce the share of individuals in the "treated" group who agreed with the sensitive item by comparing the average number of agreed statements in each group (see Glynn, 2013;Holbrook and Krosnick, 2010;Kuklinski et al., 1997).
In the survey the "control" group was presented with the following question: I [the interviewer] will read three statements. I will then ask you with how many of these statements you agree on. You should not tell me which specific statement you agree on but the number of statements you agree on. I will give you three marbles and you have to hold them in your right hand. Keep both of your hands on your back side. For each of the statements, if you agree on it, please transfer one marble from your right hand to your left hand behind you.
If you do not agree on it, please do not transfer any marble. At the end, I would like to know the total number of statements you agreed on. This number should correspond to the number of marbles you have in your left hand. I will now read the statements.
1. It is safer to bring a client home than going in a hotel.
2. I prefer that the client pays me before the intercourse.
3. Monday is the day I have the greatest number of clients.
Participants in the "treatment" group were presented the same statements plus the sensitive item that relates to condom use.

I used a condom during my last sexual intercourse with a client.
We can investigate the relation between condom use and respondents' risk preferences using a simple linear regression with interaction terms: where Y i is the number of statements the respondent agreed with. T i takes value 1 if the individual was in the treatment group. RP i is a characteristics of individual i that may be correlated with condom use. The p-value of the coefficient α indicates if the condom use depends on the individual's risk preferences.
Appendix 3 -Pairwise correlations for FSWs who report not being willing to take any risk   (3) and (4)). Each reported coefficient estimate is based on a seperate OLS regression in Panels 1 and 1b. SRRP stands for self-reported risk preferences. Higher SRRP mean greater risk aversion. Columns (3) and (4) refer to the two last paid sex intercourses. Column (11) comes from medical records of registered sex workers. Differences in the number of observations in columns (5), (6) and (9) are due to missing information. Registration status information (Column (7)) is available for active FSWs only. In column (2), the reported coefficients refer to the interaction term RP i × T i , see Appendix 2.  (3) and (4)). Each reported coefficient estimate is based on a seperate OLS regression. p<0.01, p<0.05, p<0.1. SRRP stands for self-reported risk preferences. CRRA stands for constant relative risk aversion. Higher CRRA and lower SRRP mean greater risk aversion. Columns (3) and (4) refer to the two last paid sex intercourses.