Large Stakes and Big Mistakes
Duke University, Fuqua School of Business
University of California San Diego, Rady School of Management
Carnegie Mellon University, Social and Decision Sciences
University of Toronto, Rotman School of Management
Aug 20, 2008
Workers in a wide variety of jobs are paid based on performance, which is commonly seen as
enhancing effort and productivity relative to non-contingent pay schemes. However,
psychological research suggests that excessive rewards can in some cases result in a decline in
performance. To test whether very high monetary rewards can decrease performance, we
conducted a set of experiments in the US and India in which subjects worked on different tasks
and received performance-contingent payments that varied in amount from small to very large
relative to their typical levels of pay. With some important exceptions, very high reward levels
had a detrimental effect on performance.
Acknowledgements. The authors are grateful for the help of the faculty and students at Narayanan College
in Madurai, who carried out the experiment in India. Faculty: Dr. R. Srinivasan; Prof. A. Narasimhamurthy; Dr. K.
Ramasamy; Dr. M. Jayakuma. Students: J. Moses Gnanakkan; P. Kalaignar; M. Ramesh; G. Selvakumar; K.
Prabakara Doss. Other thanks go to Christopher Simeone, Mark Porter, and Jackie Zires for their help with running
the experiments at MIT and the University of Chicago, Ricardo Paxson for designing the set of numbers for the
matrices in the adding task and advising on the analyses, as well as Kristina Shampanier, On Amir, Uri Simonsohn,
Avi Goldfarb, Peter Perkins, and Lowell Taylor for their invaluable insight into stats. The authors are also indebted
to the editor, and the referees for their helpful comments.
Payment-based performance is commonplace across many jobs in the marketplace. Many, if not
most upper-management, sales force personnel, and workers in a wide variety of other jobs are
rewarded for their effort based on observed measures of performance. The intuitive logic for
performance-based compensation is to motivate individuals to increase their effort, and hence
their output, and indeed there is some evidence that payment for performance can increase
performance (Lazear, 2000).
The expectation that increasing performance-contingent incentives will improve
performance rests on two subsidiary assumptions: (1) that increasing performance-contingent
incentives will lead to greater motivation and effort, and (2) that this increase in motivation and
effort will result in improved performance.
The first assumption, that transitory performance-based increases in pay will produce
increased motivation and effort, is generally accepted, although there are some notable
exceptions. Gneezy and Rustichini (2000a), for example, have documented situations, both in
laboratory and field experiments, in which people who were not paid at all exerted greater effort
than those who were paid a small amount (see also Gneezy and Rustichini, 2000b; Frey and
Jegen, 2001; Heyman and Ariely, 2004). These results show that in some situations paying a
small amount in comparison to paying nothing seems to change the perceived nature of the task,
which, if the amount of pay is not substantial, may result in a decline of motivation and effort.
Another situation in which effort may not respond in the expected fashion to a change in
transitory wages is when workers have an earnings target that they apply narrowly. For example,
Camerer and colleagues (1997) found that New York City cab drivers quit early on days when
their hourly earnings were high and worked longer hours when their earnings were low. The
authors speculated that the cab drivers may have had a daily earnings target beyond which their
motivation to continue working dropped off.
Although there appear to be exceptions to the generality of the positive relationship
between pay and effort, our focus in this paper is on the second assumption - that an increase in
motivation and effort will result in improved performance. The experiments we report address
the question of whether increased effort necessarily leads to improved performance. Providing
subjects with different levels of incentives, including incentives that were very high relative to
their normal income, we examine whether, across a variety of different tasks, an increase in
contingent pay leads to an improvement or decline in performance. We find that in some cases,
and in fact most of the cases we examined, very high incentives result in a decrease in
performance. These results provide a counterexample to the assumption that an increase in
motivation and effort will always result in improved performance.
2. PRIOR RESEARCH ON THE CONNECTION BETWEEN EFFORT AND
Unlike the relationship between pay and motivation/effort, the relationship between
motivation/effort and performance has not attracted much attention from economists, perhaps
because the belief that increased motivation improves performance is so deeply held. However,
research by psychologists has documented situations in which increased motivation and effort
can result in a decrement in performance - a phenomenon known as "choking under pressure"
Although conventional economics assumes a positive relationship between effort and
performance, there are a wide range of psychological mechanisms that could produce the
opposite relationship. These include: increased arousal, shifting mental processes from
"automatic" to "controlled," narrowing of attention, and pre-occupation with the reward itself.
The increased arousal account is embodied in the "Yerkes-Dodson law" (Yerkes and
Dodson, 1908), which posits that there is an optimal level of arousal for executing tasks, and that
departures from this level in either direction can lead to a decrement in performance. The effect
was first demonstrated by Yerkes and Dodson with rats that were placed in a cage and forced to
repeatedly choose between exploring either the left or right of two passages. On each trial, the
experimenters randomly hung a white card in one passage and a black card in the other. While
exploring the passage with the white card resulted in a reward, exploring the passage with the
black card always resulted in a shock. For some rats the shock was always small, for some
medium, and for a third group it was strong. The main finding was that the rats learned to avoid
the shocks most quickly when the shocks were at an intermediate level of intensity. Related
results have been obtained in human motor performance (Neiss, 1988), when arousal was
increased by stimulant drugs, muscle tension, or electric shocks. Since arousal is tightly linked to
motivation and performance, the "Yerkes-Dodson law" implies that increases in motivation
beyond an optimal level can, in some situations, produce supra-optimal levels of arousal and
hence decrements in performance.
Another possible mechanism for the negative effects of increased incentives is that
increased incentives can cause people, involuntarily, to consciously think about the task, shifting
control of behavior from "automatic" to "controlled" mental processes even though it is well
documented that controlled processes are less effective for tasks that are highly practiced and
automated (Langer and Imber, 1979; Camerer et al., 2005). Sports provide a prototypical
example of such over-learned, automatic tasks. Thinking about how one is swinging the golf
club or bat, or about how to get the basketball into the net can have perverse effects on
performance. In fact, there are some studies of choking under pressure in sports, including one
Australian study, which found that free-throw shooting performance among elite Australian
basketball players was worse during games than during training (Dandy et al., 2001). This
mechanism can also help to explain why the presence of an audience or competition, which tends
to increase the motivation to perform well, can have detrimental effects (see also Zajonc, 1965).
A third mechanism by which increased motivation is likely to have a negative effect on
performance involves the focus of attention. Increased motivation tends to narrow individuals'
focus of attention on a variety of dimensions (Easterbrook, 1959), including the breadth of the
solution set people consider. This can be detrimental for tasks that involve insight or creativity,
since both require a kind of open-minded thinking that enables one to draw unusual connections
between elements. McGraw and McCullers (1978) provided support for this mechanism by
showing that the introduction of monetary rewards for tasks that involved problem-solving had
detrimental effects on performance. In addition to the narrowing of attention, large incentives
can simply occupy the mind and attention of the laborer with thoughts about her future in case
she would get the reward and her regrets if she would not, distracting her from the task at hand.
In summary, psychological research has identified several mechanisms that can produce
choking under pressure, suggesting that there are diverse factors that can create the type of
pressure that produces choking. The sources most relevant for this type of pressure seem to be
the presence of an audience (passive onlookers), competition (i.e. presence of others involved in
the same activity; "coaction effect"), personal traits such as competitiveness, and ego-relevant
threats like the belief that a task is diagnostic of something, such as intelligence, that one cares
about (see Baumeister and Showers, 1986, for an in-depth discussion of these effects; McGraw,
1978; Zajonc, 1965; also Frey and Jegen, 2001).
For economics, however, the most interesting determinant of performance pressure is the
level of performance-contingent monetary incentives - an important variable for most economic
behavior and labor (i.e. situations driven by external motivation). Accordingly, experiments in
economics have examined the effects of the magnitude of monetary incentives on decisions. For
example, Slonim and Roth (1998) report the results of an experiment with repeated ultimatum
games conducted in the Slovak Republic in which financial incentives were varied by a factor of
25. Their findings suggest that increasing incentives has only a small effect on behavior of
inexperienced participants, but a larger effect as participants gain experience with the game:
Experienced participants rejected offers less frequently and made lower offers as the stakes
increased (although see Cameron, 1999, who found no difference in behavior when stakes were
changed in the ultimatum game).
In a review of the experimental literature on the effect of incentives, Camerer and
Hogarth (1999) report 74 studies in which the level of pay was varied in different kinds of tasks
(e.g., bargaining, trading, choosing, problem solving). They reach an inconclusive view: the
majority of studies do not find any effect on performance, but many studies did observe a
decrease in variance, arguably because people put more effort into the task. A few papers
reported an increase in performance, and even fewer reported a decrease. Decreases were
typically found in prediction tasks or tasks in which simple intuition or habit provides an optimal
answer and thinking harder makes things worse (see e.g., Arkes, Dawes, and Christensen, 1986;
Ashton, 1990; and Hogarth et al., 1991). Most of the studies surveyed in Camerer and Hogarth
(1999) look only at the difference between no pay and low pay, which is different from our goal
in the current paper. The studies that do use high pay and compare it to low pay generally do not
show an effect. However, the high pay in these papers is still far from the level of pay we use,
and the tasks tend to be quite different.
Our primary goal in the studies reported herein is to test, in experiments that satisfy
standard experimental economics criteria, whether increasing monetary incentives beyond some
threshold may result in lower performance. A second major goal that distinguishes our work
from previous contributions is to examine the generality of any detrimental effect of incentives.
Among the six tasks in the first experiment, therefore, we included some that drew primarily on
motor skills, some that drew primarily on memory, and some that drew primarily on creativity.
Based on the literature showing detrimental effects of incentives on motor skills and creativity,
we anticipated that the high monetary rewards might interfere with tasks that draw primarily on
these skills, but not with those involving primarily memory. As will be seen, however, no such
differences emerged; the highest levels of monetary rewards produced lower performance on all
tasks in the first experiment. To examine this issue further in the second experiment, we included
a task that required only physical effort. Such a task should not be subject to any of the
mechanisms leading to choking under pressure as identified in the psychology literature. In this
case the predicted differences between tasks did emerge. Finally, our third experiment extends
the scope of investigation from financial to social incentives.
3. EXPERIMENT 1
Eighty-seven residents of a rural town in India were recruited to participate in the experiment,
which took place late in 2002.1 Subjects were recruited by word of mouth in the village. The
researchers first collected names of people interested in participating, and then contacted
interested individuals to schedule experimental sessions. The sample consisted of 26.4 percent
females and 73.6 percent males. The majority of participants (90.8 percent) were Hindu, 5.7
percent were Christian, and 3.4 percent were Muslim. The standard of living of our participants
can be best described by their level of education and type of possessions. Participants in this
experiment had, on average, 5.6 years of education, and 26.4 percent had no formal education.
Approximately half of the participants reported that they owned a TV (M = 49.4 percent), and
about half owned a bicycle for transportation (M = 51.7 percent). None owned a car, and only
6.9 percent had a telephone in their house.
The experiment was conducted with one participant at a time. Participants were randomly
assigned to one of three treatments in which they faced incentives (on all games) that were either
relatively small, moderate, or very large. In each treatment, participants played six different
games in a random order and were promised payments for each game if they reached certain
performance levels. The magnitude of the payment in each game depended on the treatment
(low, mid or high incentive magnitude) and whether they reached either of two specified
performance levels which we labeled "good" and "very good." In each game, participants
received full payment (i.e. 4, 40 or 400 Indian Rupee Rs depending on the treatment) if they
1 The experiment was conducted by local research assistants from Narayanan College at Madurai, India, who were
naive to the hypotheses.
reached the "very good" performance level, half of that if they reached the "good" performance
level, and nothing if they failed to reach the "good" performance level (these two performance
levels, as well as the games themselves, were selected based on pre-testing with MIT students).
The maximum possible payment for any one game in the high incentive treatment (Rs
400) was relatively close to the all-India average monthly per capita consumer expenditure
(MPCE) in rural areas, which was Rs 495 (Rangachari, 2003).2 Thus, in the unlikely event that a
participant in the high payment treatment achieved "very good" performances on all six games,
she would earn an amount approximately equal to half of the mean yearly consumer expenditure
in the village. These stakes are effectively much larger than those that are typically offered in
The Games - The six games fell into three broad categories based on whether they
required primarily creativity, memory, or motor skills.
The game that was used as a creativity task was "Packing Quarters." In this game
participants were asked to fit 9 metal pieces of quarter circles into a black wooden frame within a
given time. It is easy to fit 8 pieces, but, to fit all 9, the pieces have to be packed in a particular
way. The good performance level was defined by a completion of the game within 240 seconds.
The very good performance level was defined by a completion of the game within 120 seconds.
Participants had only 1 trial to reach these goals.
The memory tasks included two games: "Simon" and "Recall last 3-digits." "Simon" is
an electronic game that requires memory and repetitions. The game flashes a sequence of colored
lights accompanied by the light-specific sounds, and the goal is to repeat the sequence by
pushing the corresponding light-buttons in the same order. The good performance level was
2 The conversion is based on the average exchange rate in December 2002 of Indian Rupee Rs 47.93 = US $1 (see
Federal Reserve Statistical Release, 2003).