Bracha, Anat; Fershtman, Chaim Working Paper Competitive incentives: Working harder or working smarter? Working Papers, No. 12-12 Provided in Cooperation with: Federal Reserve Bank of Boston Suggested Citation: Bracha, Anat; Fershtman, Chaim (2012) : Competitive incentives: Working harder or working smarter?, Working Papers, No. 12-12, Federal Reserve Bank of Boston, Boston, MA This Version is available at: http://hdl.handle.net/10419/96428 Standard-Nutzungsbedingungen: Terms of use: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Documents in EconStor may be saved and copied for your Zwecken und zum Privatgebrauch gespeichert und kopiert werden. personal and scholarly purposes. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle You are not to copy documents for public or commercial Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich purposes, to exhibit the documents publicly, to make them machen, vertreiben oder anderweitig nutzen. publicly available on the internet, or to distribute or otherwise use the documents in public. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, If the documents have been made available under an Open gelten abweichend von diesen Nutzungsbedingungen die in der dort Content Licence (especially Creative Commons Licences), you genannten Lizenz gewährten Nutzungsrechte. may exercise further usage rights as specified in the indicated licence. No. 12-12 Competitive Incentives: Working Harder or Working Smarter? Anat Bracha and Chaim Fershtman Abstract: Almost all jobs require a combination of cognitive effort and labor effort. This paper focuses on the effect that competitive incentive schemes have on the chosen combination of these two types of efforts. We use an experimental approach to show that competitive incentives may induce agents to work harder but not necessarily smarter. This effect was stronger for women. JEL Classifications: J33, D03 Anat Bracha is an economist at the Federal Reserve Bank of Boston. Her e-mail address is [email protected]. Chaim Fershtman is a professor of economics at Tel Aviv University, and a research fellow at the Centre for Economic Policy Research and at the Tinbergen Institute. His e-mail address is [email protected]. The work presented in this paper was supported by Grant No. 0610114192 from the Israel Science Foundation. The authors thank Lynn Connell-Price for her excellent research assistance. This paper presents preliminary analysis and results intended to stimulate discussion and critical comment. The views expressed herein are those of the authors and do not indicate concurrence by the Federal Reserve Bank of Boston, or by the principals of the Board of Governors, or the Federal Reserve System. This paper, which may be revised, is available on the web site of the Federal Reserve Bank of Boston at http://www.bostonfed.org /economic/wp/index.htm. This version: October 11, 2012 1. Introduction Competitive incentives are commonly used by firms and modern organizations to motivate their workers. While competitive incentives have proven to be effective in inducing more effort, the focus of this paper is whether such incentives induce individuals to work smarter, or just to work harder. Almost all jobs require a combination of cognitive effort and labor effort. Research and development (R&D), running a company, building a house, teaching a class, or working on a factory floor requires both cognitive and labor efforts. The tradeoff between these two types of effort exists whenever agents need to think about how to perform a task or to choose a method of solving a problem before they actually implement it. For instance, consider the task of trying to find the highest value of a function: people may try to analyze the function (cognitive effort), they may try to check it for many parameter values (labor effort), or they may attempt a combination of the two methods. It is the combination of cognitive and labor effort that determines whether people work harder or smarter. The general intuition holds that providing competitive incentives motivates individuals to exert more effort.1 But once we distinguish between cognitive and labor efforts an additional question arises: what is the effect of different incentive schemes on the combination of the two types of efforts? If different incentive schemes do indeed influence workers’ chosen combination of effort this could be important for firms, as different firms may wish to elicit different combinations of effort. Firms that specialize in R&D and innovation, for instance, may want workers to exert more cognitive effort—to work smarter—while some manufacturing firms may wish to motivate workers’ labor effort—to work harder. The main application of our study is that the optimal incentive scheme also depends on the type of effort firms wish to elicit. To answer the question of whether incentive schemes affect the chosen combination of effort, we consider and examine players’ effort allocation under two types of incentives: (i) a simple pay-for-performance incentive scheme (hereinafter PFP), in which agents are paid according to their own performance, and (ii) tournament incentives 1For a survey see Lazear (2000). 1 in which pairs of participants compete for a prize. Our claim is that, due to competitive pressure, competitive incentives will lead participants to exert less cognitive effort but more labor effort, relative to PFP. That is, we expect that under a competitive incentive scheme participants will work harder but not smarter. This conjecture is related to the psychological literature identifying several mechanisms that result in ―choking under pressure.‖ This literature suggests that pressure in various forms such as large stakes, performing in front of an audience, and competition may lower performance in various types of tasks (see, for example, Baumeister 1984; Baumeister and Showers 1986; Beilock, Kulp, Holt, and Carr 2004). In economics, Ariely, Gneezy, Loewenstein, and Mazar (2009) recently demonstrated the choking under pressure effect and showed that excessively high rewards have a detrimental effect on performance. However, choking under pressure does not necessarily occur in all circumstances: for instance, Beilock, Kulp, Holt, and Carr (2004) show that pressure adversely affects performance in solving novel but not heavily practiced math problems,2 whereas solving novel math problems requires high working memory—cognitive effort—while the practiced questions do not. To test our claim that under competition people will work harder, not smarter, we designed a simple lab experiment with two computerized tasks, a ―sequences‖ task where participants were asked to solve numeric sequences that require cognitive effort and a ―filing‖ task that is a simple number categorizing task mainly requiring manual dexterity. Participants in this study could engage in either task and were free to switch between the two during the entire duration of the experiment. In examining performance across the two incentive schemes we focus on whether the competitive incentives affected time allocation between the sequences task and the filing task as well as the players’ success rate in solving sequences, compared with PFP.3 Our main results are that under competitive incentives participants devote less time to the sequences task and have a lower success rate than when they are provided with PFP incentives. In other words, the tournament incentive did indeed lead participants to work harder but not smarter. 2To correctly answer a practiced problem one only needed to retrieve the solution from memory. 3Success rate is the percentage of sequences solved correctly over the number of sequences attempted. 2 However, our results are gender sensitive. Under the PFP incentives the performance of women is lower than the performance of men, as women attempted to solve less sequences and devoted more of the allotted time to the simpler task of categorizing numbers. This is despite the fact that in the PFP treatment men and women had the same success rate in solving sequences. Analyzing the effect of competitive incentives by gender, we find that relative to the PFP treatment, competitive incentives induce both men and women to spend less time on the sequences task and more on the routine filing task. However, the negative effect of the competitive incentives on the success rate is entirely a female effect. This paper adds to the experimental investigation of tournament relative to pay-for- performance incentives (see, for example, Bull, Schotter, and Weigelt 1987) and the recent literature on gender differences in response to competition (for example, Gneezy, Niederle, and Rustichini 2003; for a survey see Croson and Gneezy 2009 and Bertrand 2011). However, previous studies either use nonreal effort settings where effort is a number chosen by a participant in the study, or a real-effort task which requires one type of effort. Hence, this paper adds to the literature by considering a new aspect of competitive incentives—its effect on effort allocation—and points at the differential effect that competition has on effort allocation across gender. This paper also adds to the recent literature on the psychological foundation of incentives, which provides an important critical view of the traditional incentive theory (for a survey see Fehr and Falk 2002). The main claim in this literature is that considering monetary incentives alone is too narrow, empirically questionable, and limits our understanding of incentives. Nonpecuniary motives such as reciprocity, the desire for social status, and fairness concerns are powerful drivers of human motivation. Our results extend this literature by focusing on the combination of cognitive and labor effort. 2. Experimental Design To capture the two different types of effort we introduced two tasks, each of which emphasized either cognitive or labor effort. Subjects could freely choose to engage in either solving sequences (―sequences‖ task)—finding a missing number in a sequence of four numbers—or classifying a random number into an ―odd‖ or ―even‖ category (―filing‖ task) by pressing the appropriate button on the computer screen. The sequences 3 task requires cognitive effort in the form of abstract thinking, while filing numbers mainly requires labor effort. Both tasks were available during the experiment, and engaging in each of the two tasks was done by simply clicking on the section of the screen with the desired task (see figure 1). Figure 1: Sequences Task and Filing Task Source: Authors’ calculations. 2.1 Treatments To analyze the effect of incentives we use a between-subject design with two treatments: Pay for Performance and head-to-head tournament. In our experimental design the goal was to earn money, and the complementarity between the tasks was achieved using the compensation scheme described below. Pay-for-Performance (PFP): subjects were paid $2 per net correctly solved sequences, 3 cents per net correctly filed numbers, and a 1 cent extra reward for the product obtained by multiplying the net sequences by the net filed numbers.4 The net number of correctly solved sequences is the number of correctly solved sequences minus half the number of incorrectly solved sequences. Penalizing incorrectly solved sequences was designed to prevent guessing. The net filed numbers equals the correctly filed numbers minus the incorrectly filed numbers. Penalizing incorrectly filed numbers was designed to prevent random clicking. The extra reward introduces a complementary term, as a greater number 4This compensation is different from a piece rate since there is a multiplicative term in their incentives. 4 of net correctly solved sequences (filed numbers) increases the marginal return to successful filing (sequence). Tournament: in this treatment, subjects were randomly paired using a randomly generated subject identification (ID) number. The pairs were announced before the beginning of the task and by subject ID, such that the identity of one’s opponent was not revealed. The winner was determined according to the accumulated number of points for each of the opponents in a pair. The point schedule was exactly the same as under the PFP compensation scheme—2 points per net sequence, 0.03 of a point per net number filed and an extra 0.01 of a point for the product of net sequences times the net numbers filed. The winner’s prize was $60, and the loser received the minimum guarantee of $10, such that the expected earning was $35, similar to the average earning under PFP. At the end of the study, after completing the time devoted to the task, the accumulated number of points for each participant was announced (by the randomly generated subject ID), and the earnings were determined and announced. 2.2 Procedure The sessions were conducted at the Harvard Decision Science Laboratory at the Harvard Kennedy School. A total of 134 Harvard students participated in the study, 74 in the PFP treatment and 60 in the tournament treatment. In each session, participants sat at individual computer stations and read the instructions on their individual screens. Once all were done with the instructions, they were given a code to proceed such that all started working on the task at the same time. Under both conditions, they were given 10 minutes to work on the two tasks. In the tournament treatment, once all were done with the instructions, and before giving the code to proceed, the experimenter announced the pairs by subject ID. 3. Competitive Incentives Economic intuition suggests that competitive incentives may induce individuals to 5 exert more effort.5 But individuals may also react to the pressure created by the competitive incentives. Competitive pressure may have different effects on cognitive and labor effort. Competitive pressure may affect cognitive and labor efforts differently. Our main hypothesis is that competitive pressure mainly affects cognitive effort by making it more difficult or more costly to perform cognitive tasks. Hypothesis 1: (i) Competitive incentives induce individuals to reduce their cognitive effort while increasing their labor effort. (ii) Competitive incentives reduce the effectiveness of cognitive effort which results in a lower success rate. In our experiment individuals were given ten minutes to engage in the two tasks. This time limit may have created an additional type of pressure—time pressure. We expect that the time pressure, if it exists, will be stronger at the end of the given time frame and that time pressure will be stronger in the tournament treatment, as earning an additional point may be the decisive factor for winning the tournament. Hypothesis 2: Time pressure will be effective in the tournament treatment resulting in lower success rates at the end of the tournament relative to those at the beginning of the tournament.6 In the tournament treatment, time pressure will affect performance and result in lower success rates at the end of the tournament relative to the beginning of the tournament. To test these hypotheses we examine the time allocation between the two tasks and the success rate in solving sequences. The comparison is provided in Table 1.7 5 See Lazear and Rosen (1981) for a theoretical argument. Empirically, Ehrenberg and Bognanno (1990) demonstrate the positive effect of tournament incentives on effort in golf tournaments, and Kremer, Miguel, and Thornton(2009) find that for girls in Kenya, competing forscholarships has a positive effect on their scholastic achievement. Experimentally, the results of studies such as Gneezy, Niederle, and Rustichini (2003) and Niederle and Vesterlund (2007), although aimed at gender differences, show that competitive incentives increase performance among men relative to performance with under a piece rate reward scheme, and that even among women, performance is at least as good under tournament incentives as under piece rate rewards. 6 Our hypothesis is only with respect to the success rate and not to the allocation of time between the two tasks, as the allocation of time at the end is affected by the decisions made at the beginning of the ten- minute period. 7 In calculating the various averages, we first calculate the particular measure (such as success rate) for each individual and then average across individuals. Under PFP incentives, the average number of sequences attempted was 11 with a standard deviation of 6.69; under tournament, the average number of sequences attempted was 9.6 with a standard deviation of 6.78. There were two outliers, one in each condition, who attempted over 30 sequences in 10 minutes (32 sequences under PFP and 33 sequences under Tournament.) 30 attempts is more than two standard deviations from the mean. Therefore, in our analysis and in the statistics presented above we exclude these two outliers. 6 Table 1: Comparing the PFP and the Tournament Treatments PFP Tournament P-value Total Performance 33.95 32.99 n.s. Attempted Sequences 10.71 9.2 P=0.083 Correctly Solved Sequences 8.54 7.29 n.s. Success Rate 0.79 0.72 P=0.047 Time on Sequences 381.06 330.88 P=0.025 Time per Attempted Sequence 48.67 50.79 n.s. Time per Correctly Solved Sequence 63.42 70.74 n.s. Net Filing Rate 1.02 1.06 n.s. Note: Success rate, time per attempted sequence, and time per correctly solved sequence are based on individuals who actually solved a sequence. P-values are based on one-sided T-tests. 3.1 The Effect of Competitive Incentives The effect of tournament incentives on time allocation is best seen in figure 2B, which compares the minute-by-minute percentage of time devoted to solving sequences in the PFP and the tournament treatments. Figure 2B strikingly illustrates that the effect of tournament incentives is neither due to a single episode nor due to a particular stage of the task. Rather, the effect of tournament incentives on time allocation stems from different time allocation decisions made throughout the entire 10 minutes of the study. Figure 2: Allocation of Time Source: Authors’ calculations. Note: N=73 under PFP, N=59 under tournament. The bars in panel A represent the standard error of the mean. 7 To test the second part of Hypothesis 1 we calculate the success rate in solving sequences for each participant and then average these results across all individuals (see figure 3A). We find that the success rate is lower under competitive incentives (78.6 percent under PFP while only 72 percent under tournament). Labor effort can be captured by examining the average net filing rate which is the average speed of net filing across participants (see figure 3B). A significantly lower success rate coupled with a higher net filing rate (although the latter is not statistically significant) indeed demonstrates that competitive incentives induce participants to work harder but not smarter. Figure 3: Sequence Success Rate Source: Authors’ calculations. Note: Panel A: N= 73 under PFP, N=57 under Tournament; Panel B: N= 69 under PFP, N=56 under Tournament. Bars represent the standard error of the mean. Observation 1 (Tournament and cognitive effort): (i) Under competitive incentives participants devoted less time to the cognitive task than under the PFP incentives (330 second under competition versus 381 under PFP; p=0.025). (ii) The success rate in solving sequences is lower under tournament incentives—78.6 percent under PFP while only 72 percent under tournament incentives (p=0.047). Clearly the two parts of Observation 1 may be interdependent: if participants are aware of the fact that under tournament they have a lower success rate, then their rational reaction would be to reduce the time they spend on solving sequences. The effect of competitive incentives on the success rate may be due to an effect on the effort expended, an effect on ability, or both. We may take the conservative economic interpretation of an effect on effort and treat ability as exogenous. However, under competitive pressure it is 8