“Rewards for weight loss don’t work; would penalties for failure do better?”
Tyler is referencing an NBER Working Paper with the following abstract:
Obesity rates in the U.S. have doubled since 1980. Given the medical, social, and financial costs of obesity, a large percentage of Americans are attempting to lose weight at any given time but the vast majority of weight loss attempts fail. Researchers continue to search for safe and effective methods of weight loss, and this paper examines one promising method – offering financial rewards for weight loss. This paper studies data on 2,407 employees in 17 worksites who participated in a year-long worksite health promotion program that offered financial rewards for weight loss. The intervention varied by employer, in some cases offering steady quarterly rewards for weight loss and in other cases requiring participants to post a bond that would be refunded at year’s end conditional on achieving certain weight loss goals. Still others received no financial incentives at all and serve as a control group. We examine the basic patterns of enrollment, attrition, and weight loss in these three groups. Weight loss is modest. After one year, it averages 1.4 pounds for those paid steady quarterly rewards and 3.6 pounds for those who posted a refundable bond, under the assumption that dropouts experienced no weight loss. Year-end attrition is as high as 76.4%, far higher than that for interventions designed and implemented by researchers.
I was very pleased to see this, since the framing of incentives is something that is right up my alley (and most likely in my dissertation). You see, people in general exhibit what behavioral economists call loss aversion. Simply put, loss aversion means that people dislike things that their brains categorize as losses more than they like things that their brains categorize as gains. Since I study incentives, my question is whether people are willing to work harder to avoid a loss than they are to acquire a gain. Loss aversion would suggest that the answer is yes, and furthermore that a manager could change how hard his employees work simply by changing the framing of the incentive rather than changing the underlying incentive itself.
Consider the following experimental setup:
- Subjects are brought into the laboratory and told that their task is to complete multiplication problems, let’s say 3 numbers multiplied by 3 numbers. (The multiplication task was chosen because the speed at which it can be done is dependent on effort rather than just ability and the problem is irritating enough that there is a perceived cost to doing the problems. Subjects are also given some form of distraction, such as a newspaper or something else to read, so that the opportunity cost of their time is not zero. In other words, I don’t want them to feel like their only options are “multiply” or “stare at wall”…though I suppose that’s how a lot of children are conceived…)
- The control group is paid an amount of money (let’s say 50 cents) for each multiplication problem completed within a given amount of time, let’s say 10 minutes. (To count as completed, the actual answer doesn’t have to be correct, but it has to be clear that the subject actually tried to do the problem and didn’t just write down gibberish. I might go back and look at differences in accuracy at some point, since that could be interesting.) The time period has to be long enough that subjects would presumably get bored with the task and have to force themselves to keep going.
- The experimental group is told that the expectation is that they finish 20 problems in 10 minutes (these numbers would obviously have to be calibrated in some fashion), so they are given $10. They are then told that they will have to give back 50 cents for each problem up to 20 that they don’t complete. For example, the subject would have to give back $1.50 if 17 problems were completed, $1 if 18 were completed, etc. If the subject completes more than 20 problems, she gets 50 cents for each problem over 20 in addition to keeping the original $10.
- Note that the control and experimental groups are identical in terms of dollar payouts, it’s only the framing of the payouts that have changed.
Loss aversion would suggest that that the people in the experimental group would work harder, since they view completing the task as avoiding a loss rather than achieving a gain. This would also suggest that you would get some bunching in performance around the 20 problem mark for the experimental group, since the subjects might feel like they “outran the bear” once they finish 20 problems and thus can stop.
Given that behavioral economists like to try and keep in mind that people are people and not economic robots, it is also worth thinking about how doing the task makes the subjects feel. My hypothesis is that, while the experimental group may work harder that one time, they will report having disliked the task more and be more reluctant to do it again. I am also curious as to what their performance would be on the subsequent task.
In case you’re curious, I plan on running this experiment soon and will keep you posted. If you have any suggestions, feel free to send them along!