Do High-Powered Incentives Harm Performance? First Aid From the Labor Market

2015-09-09

2019-11-30

The number of tables correctly solved (productivity).

Outcome is not constructed.

There are no secondary outcomes

Subjects had to find the two numbers that add to 10 in a table containing 12 numbers. They had to solve as many tables as possible in a four-minute window with a maximum of 20 tables. There was a counter of correctly solved tables and a chronometer at the top of the page. The only button available on the screen was a ``Next" button, which submitted an answer and led to the next table. There was no penalty for incorrect answers. Subjects perform the task under two payment conditions a high and a low-payment condition. The stakes in the High-Payment condition were extremely large for this subject population.

Treatments: (1) Baseline. In this treatment, subjects do not have any background information about the task and no instance for practicing it beyond the unpaid round. This informational setting replicates that in the standard laboratory design studying the harmful effects of high-powered incentives in the economics literature. (2) Selection treatment. In the labor market, workers self select into their preferred jobs. To explore whether this sorting affects the negative response to high-powered incentives, in this treatment, the recruitment flyer, as well as the email sent to interested subjects, offered details about the task. As in the Baseline, subjects in this treatment had no opportunity to practice the task before the payment conditions. (3) High-Powered Practice treatment. In the labor market, workers can train to reap the benefits of high-powered incentives. To study whether training can ameliorate the negative response to large-stakes incentives, in this treatment, participants from the Baseline and Selection treatments were invited to repeat the same study after having the opportunity to practice the task online for approximately six days. The online platform was available 24 hours a day, and subjects could log in at any time, from any computer. (4) Practice treatment. Even though awareness of the compensation scheme is important for real-world practice, one might wonder what is the pure effect of becoming more skillful at the task. To explore whether practice has an effect on the response to powered incentives without considering the motivating effects of rewards, a new group of subjects was offered the opportunity to rehearse the task in the laboratory up to 20 minutes, after the task was introduced, but before the payment conditions and thus payment information was released.

(1) Payment conditions are randomly assigned at the subject level (within-subject design at the payment condition level).
(2) Baseline and Practice treatments were randomized using a random number generator after all subjects in a given leg were recruited. Subjects in the Selection treatment self-selected into the study (see "intervention" above). Subjects in the High-Powered treatment were a subsample of subjects in the Baseline and Selection treatment. In the first leg of the experiment only subjects negatively affected by high-powered incentives were invited to repeat the study. In the third leg, all subjects in the Baseline and Selection treatment were invited.

Treatments were randomized at the subject's level without clustering.

No

Two hundred and seventeen students from mainly three universities in Chile.

217

Around 30 to 100