Why do people engage in risky behaviour, like drink-driving, or risky sexual behaviour? In theory, if people are rational, they weigh up the costs and benefits of each action, and undertake actions only when the benefits of the action outweigh the costs. People who engage in risky behaviour look (to a rational observer) like they are engaging in behaviour where the costs (including an assessment of the risk) likely outweigh the benefits. Why do they do it?
I see two potential explanations here. First, perhaps this is a case of behaviour that is boundedly rational. The people engaging in the risky behaviour may not have accurate information about the costs of the behaviour, and underestimate those costs. Making their decision based on the benefits and (underestimated) costs makes them more likely to engage in the risky behaviour. Second, perhaps these people are quasi-rational, and one of the characteristics of quasi-rational decision-making (as I discuss in my ECONS102 class) is a tendency to heavily discount the future. In cases where the benefits of a risky activity occur now, but the costs are faced at some point in the future (and so, heavily discounted), a quasi-rational decision-maker may be more likely to engage in the risky behaviour.
Either of those explanations may account for risky behaviour like drink-driving, or risky sexual behaviour. If people are unaware of, or discount the value of, the full costs of drink-driving or risky sexual behaviour, they may be more likely to engage in those activities. Also, the costs of drink-driving (such as harm to themselves, or others, or property, or the risk of penalties if they are caught) occur in the future, as do the costs of risky sexual behaviour (such as the risk of a sexually-transmitted disease, or unwanted pregnancy), while the benefits occur immediately.
Which explanation accounts for more of the activity? This is unclear. However, if most risky activity is explained by a lack of accurate information, then there is an obvious policy solution: provide accurate information about the risks and costs of the activity. If most risky activity is explained by discounting the future, then it would be more difficult to address the behaviour easily with a policy solution.
That brings me to this 2018 article by Pascaline Dupas (Stanford University), Elise Huillery (University of Paris-Dauphine), and Juliette Seban (Sciences Po), published in the Journal of Economic Behavior and Organization (ungated earlier version here). They report on a randomised experiment conducted with teenage schoolgirls in Cameroon, where they provided information about the risks of unprotected sex, and measured the effect on health outcomes and teen pregnancy rates. Specifically:
We use a field experiment conducted with teenage girls in 318 junior high schools in Cameroon to study, within one context, how the type of risk information being provided and the delivery method (teacher, outside professional or questionnaire) affect adolescents knowledge, perceived risks and behavior... We randomized HIV and sexual education interventions that differed in their delivery mechanism and intensity, as well as content, across schools. In each school, one eighth grade class was targeted for the study.
We consider four interventions. The first (In-Class Quiz) was completely “hands-off”, and not labeled as an educational intervention: students were simply asked to fill in an anonymous questionnaire with questions on HIV as well as on their own sexual behavior and that of their peers. The questionnaire took about one hour to go through, including the time to introduce it. The In-Class Quiz was a group-administered questionnaire and did not provide students with direct information, but required that students think actively about risk levels...
The other three interventions were clearly labeled as HIV education programs. Two of them consisted of general information on HIV prevention methods (abstinence, faithfulness and condom use) and the average HIV prevalence at the national level (the “basic message”). A third one mimicked the “sugar daddy risk information” first proposed in Dupas (2011) and included, on top of the basic message, detailed information on HIV prevalence disaggregated by gender and age group and a special module on cross-generational relationships, locally known as relationships with “sponsors”, and their contribution to the spread of HIV. The difference between the two “basic message” interventions is that one was delivered through regular school staff which received special training (Teacher Training), while the second one was delivered by an outside consultant who did a special visit to the school to deliver the message (Consultant). The intervention that included the sugar daddy module was also delivered by an outside consultant (Consultant +). Both interventions by consultants lasted approximately one hour.
Since classes were randomised to receive one of the four treatments, comparisons across the treatments (at the class level) provide an assessment of the impact of the intervention. The primary outcome was self-reported pregnancy measured 9-12 months after the treatment. Dupas et al. found that:
...all interventions were successful at reducing the incidence of teenage pregnancy during our follow-up period. The magnitude of the effects are relatively large, with an average drop of 2.9% points in the likelihood of having started childbearing at the time of the endline, off a mean in the control group of 9.5%, thus a 30% reduction.
So, score one for information as an intervention to reduce risky behaviour! Except that:
The most surprising results is that the most hands-off intervention, the In-Class Quiz, was successful, by itself, at reducing the incidence of unprotected sex and hence pregnancy in the following 12 months.
The in-class quiz provided no information, and only asked the participants to reflect on risk. So, what happened? Dupas et al. look into the mechanisms, finding that:
...interventions increased the likelihood that girls report adopting a clear, one-pronged strategy against HIV: abstinence...
Importantly, the mechanism through which the interventions helped girls adopt a clear and simple strategy against HIV differs between the In-Class Quiz and the education interventions. The In-Class Quiz led participants to revise upward their subjective beliefs about risk, while the other interventions improved knowledge without changing risk perceptions... In contrast, the education interventions did not change perceived risks (it did not make them even more pessimistic as the Quiz, but it did not bring them much closer to reality either), but it did affect girls’ knowledge about HIV transmission and prevention. In our context, these two mechanisms (change in subjective beliefs about risk and change in knowledge) turnout to be equally effective at changing girls’ plans and behaviors.
So, it wasn't just information about risk that mattered, but making the risks more salient as well. When people are provided with information about risk, it has to make the risks seem more important, if we want them to change their behaviour. However, there is reason to be sceptical about this single study. As anyone who has tried any information intervention can tell you, simply providing people with information will almost always have little to no effect on behaviour (although David Evans enumerates a number of counter-examples here). So, we should be cautious before we proclaim that we have the ultimate solution to risky behaviour (and, to be fair, the authors don't claim this), until this study has been replicated in other settings.
No comments:
Post a Comment