In this Spotlight feature we focus of Johannes Haushofer, PhD, Assistant Professor of Psychology and Public Affairs in the Department of Psychology at Princeton University. He founded the Busara Center for Behavioral Economics in Nairobi, Kenya, a research facility for behavioral economics studies. Dr. Haushofer’s SOBC research investigates how stress may influence health behaviors by disrupting aspects of self-regulation.

 

Q: Your SOBC research explores the self-regulatory facets of self-efficacy, executive control, and temporal discounting. The SOBC network members may have a somewhat clearer sense of what self-efficacy and executive control are than they do about temporal discounting. Therefore, can you explain a bit about what temporal discounting is, why it might be important for behavior change, and how you measure it?

 

Temporal discounting involves a decrease in subjective value as a reward or cost is delayed into future. It often is studied as it applies to benefits, such as monetary rewards, but it is also relevant for costs. In this latter case it may involve postponing an undesirable task into the future. Health behaviors are often tasks that you have to perform such as getting your kids vaccinated or chlorinating your water. The cost tomorrow can seem less daunting than the cost today, and so one might decide to do something tomorrow because the cost is subjectively lower. There are two incarnations of the concept of temporal discounting: impatience and present bias. Impatience is often characterized parametrically with an exponentially decreasing discounting curve. You can imagine a constant percentage interest rate being applied every period in which case people may prefer to receive a reward sooner rather than later if they thought there was a constant probability of losing the outcome for every period into the future. The other incarnation is present bias, which is particularly interesting to policy makers and scientists because it implies dynamic inconsistencies. When you are present-biased, you attach particular value to whatever happens now. That is often characterized with hyperbolic discounting, which creates time-inconsistent behavior. It means that you make a plan today, and you don’t execute that plan in the future. This kind of temporal discounting applies to behaviors like medical regimen adherence where you may have very good intentions to take your medications this evening and every day after that, but when the evening rolls around, the cost is too annoying to incur it, so you postpone it. Even more than impulsivity and impatience, I think present bias is the concept related to temporal discounting that is probably most fruitful for trying to understand how health behaviors can change.

 

Q: You conducted this very impressive large-scale project with over 1,000 participants in Kenya. Can you tell a bit about the challenges you overcame in adapting the tasks for this population and collecting data from such a large number of people?

 We were able to collect a large amount of data very efficiently in our behavioral economics-style lab in Kenya. This lab is situated in the Busara Center for Behavioral Economics, which we established in Kenya during Phase 1 of SOBC. It is a non-profit, non-governmental organization (NGO) that allows researchers around the world to conduct behavioral science. We have a large subject pool from the Nairobi informal settlements and other surrounding areas where we can recruit quickly through text messages and phone calls to invite people to participate in studies. We run sessions in groups so that up to 40 people can participate in a study at the same time. For a study that takes two hours, we can run four sessions a day, thereby collecting data from up to 160 people in a day. Additionally, we have three lab spaces so that we can run sessions in parallel, thereby increasing the numbers even further.

 

The tasks and interventions often have to be adapted to make them have “bite” in the local context. A good example of this is the Trier Social Stress Test. This task is designed to induce stress by asking people to speak in front of a panel of judges and a group of their peers, but it turned out that Kenyan participants generally did not have any problem doing this, so we had to work on the instructions about the way the panel of actors interacted with participants to make sure that the task could generate stress. Additionally, it was important that all task interfaces be quite simple. Literacy is often quite low in our study population, and even in literate people, there was often very little prior exposure to research measures. Very rarely do people even fill in a questionnaire in that local context. Therefore, we used a touch screen monitor to make the interfaces more intuitive, and we administered task instructions orally. We also built in comprehensive explanations so the instructions phase often took even longer than the actual experiment. We included “understanding questions” to test people’s comprehension of the task. In this way they were incentivized in the sense that they could not continue on to the actual task until they correctly answered all the understanding questions. We can then use those questions during analysis to distinguish between people who understood well from others who did not. Finally, we used standard cross-cultural techniques to adapt scales in which scales are translated forward and backward, and we performed cognitive debriefing to determine the local acceptability of the adapted measures.

 

Q: Some of the task-based and self-reported measures of self-regulation were more strongly correlated than others. For example, in your presentation at the Steering Committee meeting, you showed that greater self-efficacy was associated with lower executive function for some measures but not others. What were some of the important points you learned from the psychometric validation phase of the project about self-regulation in this population?

The big headline finding was that—despite some exceptions—there was actually not much of a correlation between the scales and the tasks that were designed to measure similar constructs. That is a little frustrating, of course, because there is not a gold-standard measure for the constructs we are studying, and thus we are trying to approach the truth of the construct by using multiple tasks and scales. If they do not correlate with each other, it puts one in the uncomfortable position of having to choose measures without a clear sense of which type of measure is closer to the truth. However, there’s an a priori argument that is useful in this regard: responses that are incentivized may be more meaningful than ones that are not incentivized. For this reason, I have an easier time believing the behavioral economics tasks than the questionnaires. But then there are cases where there are not yet valid and reliable behavioral economics tasks available, such as for the construct of self-efficacy. We devised a decent version of such a task, but it has not yet existed for long enough to stand the test of time, and so I still put more trust in the self-efficacy questionnaire for the time being given that it has been around for a long time and has good psychometric properties. It is also worth pointing out that for self-efficacy, as with measures of various aspects of mental health, the subjective markers are the ones that should count the most.

 

Q: An important part of the project involves understanding how stress influences those self-regulation targets. Can you describe a bit about the ways that you measure stress experimentally?

We measure stress in two ways. One is with self-report scales, including a visual analog scale where we simply ask people how stressed they feel, and we also use the Positive and Negative Affect Schedule (PANAS). Second, we measure the stress hormone cortisol by collecting saliva samples several times at regular intervals during the course of the two-hour experiment. The data from this phase of the project is forthcoming.

 

Q: You have an extensive background in economics, neurobiology, and psychology. Given this diversity of expertise, are there important lessons you wish that psychologists would learn from economists or vice versa?

That’s an interesting question. One thing psychologists can learn from economists is to be careful about replicability and effect sizes. That principle largely comes down to having appropriate sample sizes. For my taste, sample sizes in psychology are still too small for the sizes of the effects that are often being studied. I think we are moving in the right direction, but it will be good to accelerate the progress. In the other direction regarding lessons that economists can learn, psychologists are especially good at getting into people’s heads and thinking very carefully about cognitive processes, whereas economists have traditionally taken a more “black-box” approach to behavior and decision-making. However, that approach is changing very rapidly in economics, and we now see economists starting to open up that black box. Nevertheless, that is an area where psychologists still have an edge because they have been doing it for so much longer.