One of the topics of discussion within behavioral economics is the motivation of effort. Although the typical economic model usually assumes monetary incentives, they do not preclude the existence of other factors, like preferences for being first or psychological and economical benefits from reputation. The growing experimental literature on this area will help to determine under which circumstances the different kinds of incentives work better. It is within this context that Dellavigna and Pope (2108) Dellavigna, S., and Pope, D. 2018. What Motivates Effort? Evidence and Expert Forecasts. Review of Economic Studies 85, 1029–1069. conduct an extensive experiment, and compare their results with previous theoretical and experimental literature, as well as with the prediction by experts.
The authors use the Amazon Mechanical Turk platform that allows them to use a big sample of experimental subjects at a very low cost (the task is performed on-line, takes little time and is cheap). The sample consists of almost 10,000 subjects (about 550 per treatment) and matches the U.S. population except for a somewhat over-representation of high education groups and younger individuals. The task of the experiment consists basically on pushing buttons; a repetitive, boring routine that potentially requires some incentive to do it fast. In all treatments the subjects receive $1 for their participation and, then, different motivations in 18 different treatments.
These are the results of the different treatments:
Treatment 1: No incentives.
Treatment 2: One additional cent for every 100 points.
Treatment 3: 10 additional cents for every 100 points.
These three treatments are the benchmark to calibrate a theoretical model of incentives that includes both monetary and psychological incentives. They will also be used to make comparisons with the rest of the treatments and to be presented to the experts so that they can forecast the results in the rest of the treatments. The average number of points in Treatment 1 was 1521. Treatment 2 got 33% more, and Treatment 3, an additional 9%.
Treatment 4: 4 additional cents for every 100 points.
Treatment 5: One additional cent for every 1000 points.
The result in Treatment 4 was reasonably forecasted by the experts and responded to the theoretical expectation. However, Treatment 5, with a very low reward, could cause a crowd-out effect, according to the literature. When the incentive is too low, it is perceived as an insult and may induce an effort even lower that in the case of no incentives. The actual result was 24% higher than Treatment 1 (no incentive), much higher than the experts’ forecast, based in the literature, and close the theoretical expected level with no crowd-out effects.
The next three treatments deal with social preferences, studied in the literature of behavioral economics.
Treatment 6: One cent will be given to the Red Cross for every 100 points.
Treatment 7: 10 cents will be given to the Red Cross for every 100 points.
Treatment 8: A bonus of 40 cents in appreciation for performing the task.
In no one of these three treatments the received money depends on the effort. The first two measure the altruism of the subject, and the third measures their responsiveness to gift exchange. Treatments 6 and 7 showed a level of effort higher than Treatment 1, but, contrary to expected behavioral literature and to expert opinion, there was almost no difference between them. Also, the increase was 24-25% higher than in Treatment 1, but, also contrary to expectations, lower than in Treatment 2, indicating small weight on social preferences. The unexpected, unconditional bonus of 40 cents in Treatment 8 had the smallest effect of all treatments compared to the no-incentive Treatment 1, with just a 5% increase in performance.
Treatment 9: One additional cent for every 100 points to be paid in two weeks.
Treatment 10: One additional cent for every 100 points to be paid in four weeks.
According to the literature on behavioral economics, hyperbolic discount fits better the individuals’ time preferences. However, the experimental results of these two treatments, with respective increases in performance of 31,7% and 29,5%, compared to the immediate reward of Treatment 2, are more compatible with the standard exponential discount most used in the theoretical literature. Experts missed this, too. (Here the reader can see how the two types of discount are calculated, and their importance to economic mechanism design.)
The next three treatments try to measure the differences in effort when the incentives are framed in terms of gains versus when they are framed as gains. According to the behavioral literature, the effect should be higher when the reference is the payoff without a loss relative to the case when the reference is the payoff before the gain.
Treatment 11: A bonus of 40 cents if the performance reaches 2000 points.
Treatment 12: A loss of 40 cents if the performance does not reach 2000 points.
Treatment 13: A bonus of 80 cents if the performance reaches 2000 points.
The result is an increase of 40-41% on the level of effort. However, and contrary to the behavioral literature, the difference between the two treatments is not statistically significant. Also, according to the literature in behavioral economics a bonus twice as big is necessary to get the same incentives with the gains framing as with the loss framing. Treatment 13 measures that, and finds that the effect is much bigger than expected (almost a 44% increase). Experts gave forecast according to this literature and also miscalculated the levels of effort.
Treatment 14: A 1% chance of winning one extra dollar for every 100 points.
Treatment 15: A 50% chance of winning 2 extra cents for every 100 points.
Notice that the expected incentive of these two treatments are the same, and coincide with Treatment 2. As expected within the most standard theory of expected utility, the probabilistic payoffs reduce the incentive relative to the fixed equivalent (24-29% versus 33%). However, the results do not support the standard behavioral hypothesis that small probabilities are over-weighted, as Treatment 14 shows a lower level of effort than both treatments 15 and 2.
Treatment 16: The subject is informed that many participants scored more than 2000 points.
Treatment 17: The subject is informed that they will be revealed their relative positions.
Treatment 18: The subject is asked to try hard to help in the experiment.
The last three treatments are about psychological motivations, without additional monetary payments. These treatments provided the lowest incentives (21-15-14% increase) except for Treatment 8. Still, they are cost-effective as they increase output for no additional cost.
This is how the authors summarize their findings:
“We find that (1) monetary incentives work largely as expected, including a very low piece rate treatment which does not crowd out effort; (2) the evidence is partly consistent with standard behavioral models, including warm glow, though we do not find evidence of probability weighting; (3) the psychological motivators are effective, but less so than incentives. We then compare the results to forecasts by 208 academic experts. On average, the experts anticipate several key features, like the effectiveness of psychological motivators. A sizeable share of experts, however, expects crowd-out, probability weighting, and pure altruism, counterfactually. As a further comparison, we present a meta-analysis of similar treatments in the literature. Overall, predictions based on the literature are correlated with, but underperform, the expert forecasts.”