It is by now fairly well established in the behavioral science literature that people do not respond to monetary incentives in a purely linear fashion. For example, Uri Gneezy and Aldo Rustichini have found that tiny financial rewards for correctly solving word problems can hurt performance (relative to a situation in which there are no financial rewards or moderate financial rewards). On the flip side, Dan Ariely, Uri Gneezy, George Loewenstein, and Nina Mazar have also demonstrated that extremely large financial rewards for solving math problems can also impair cognitive performance (relative to a condition with moderate incentives for accuracy).
However, less is known about how differences in performance incentives between people influence motivation and effort. This question is important from an organizational perspective, where wages differ between employees. In some cases, these differences can be quite salient, due to workplace gossip or the public posting of salaries (for example, at public universities). While much research has demonstrated that perceived pay inequity predicts job satisfaction and organizational citizenship, the influence of differences in pay on motivation and effort has not received the attention it deserves.
Leslie John, George Loewenstein, and I recently addressed this question in a paper forthcoming at Organizational Behavior and Human Decision Processes. In particular, we were interested in understanding when wage differences drive people to cheat. In two lab experiments, we gave participants a quiz that they later graded themselves. They reported their performance and threw away their quiz booklet; we assessed their true performance by fishing their quiz booklet out of the trash, grading it, and matching it to their self-reported score. To (gently) encourage cheating, we needed a quiz in which people could convince themselves that they deserve credit for incorrect responses. We decided to use questions in which the correct answer was not obvious but was very likely to be considered as a possible response, and thus tempt some participants to conclude that they should get credit since they “almost” put the correct response. In particular, we used questions that would be familiar to Family Feud fans (for example, “Name a sport that requires a net”). In a pre-test, we asked participants to answer these questions, and then in the experiment, we asked participants to guess the most common response among their pre-tested peers (the most common answer to the “net” question was “tennis”).
In our first experiment, we randomly assigned participants to earn either 5 cents or 25 cents for each correct guess. Critically, we also manipulated whether or not information about alternative wages was public. Thus, for example, participants who earned 5 cents per correct guess either did or did not know that some other participants were earning 25 cents per correct guess. When alternative wages were not publicly known, participants who earned 25 cents per correct guess (“25-centers”) were slightly more likely to cheat than participants who earned 5 cents per correct guess (“5-centers”), which is consistent with standard economic theory. However, when alternative wages were public information, 5-centers were significantly more likely to cheat than 25-centers. Although 5-centers had 5 times less economic motivation to cheat than 25-centers, awareness that they could have earned more motivated them to cheat.
In our second experiment, we attempted to gain some insight into the underlying psychological process. 5-centers who are aware that they could have earned more may simply be upset that they did not earn the higher wage (aversive counterfactuals). Alternatively, 5-centers who are aware that they could have earned more may be upset that others are earning the higher wage that they themselves could have earned (aversive counterfactuals plus aversive social comparisons). We teased apart these different explanations by randomly and publicly assigning everyone in the same room to the same wage (meaning 5-centers suffer only from aversive counterfactuals), or by randomly and publicly assigning people in the same room to different wages (meaning 5-centers suffer from both aversive counterfactuals and aversive social comparisons). We found that aversive counterfactuals alone were not sufficient to encourage cheating among 5-centers. Instead, only the combination of aversive counterfactuals and aversive social comparisons caused 5-centers to be significantly more likely to cheat than 25-centers.
Ian Larkin, Lamar Pierce, and Francesca Gino have previously proposed that aversive social comparisons can help to explain why performance-based pay is observed less frequently than standard economic theory would predict. Our work provides empirical support for this notion. When wage differences are public, the unethical consequences of upward social comparisons may, in some cases, outweigh the benefits of better performance among the highly compensated.