The Paradox of the Generous Bonuses and the Lazy Workers
WORKPLACE BEHAVIOR IS ONE OF THE MOST IMPORTANT SUBJECTS STUDIED in economics in general and behavioral economics in particular. I myself have devoted a great deal of attention to this topic in recent years.
One elementary reason this subject is of such importance is the fact that the single most expensive factor of production in just about every commercial enterprise is human labor. The amounts of money that corporations and organizations spend on human resources overshadow all their other expenditures. Proper planning of incentive structures at workplaces (one of the main subjects studied in human resources economics) can pay dividends in significant savings of unnecessary expenses but, perhaps even more importantly, can also lead directly to increased production and profits.
A classic case of the bottom-line miracles that a company can bring about if it uses the right incentives happened at Continental Airlines in the years 1992–1997, as detailed in a paper by Marc Knez and Duncan Simester.1 In the early 1990s, Continental Airlines suffered a financial crisis so severe that the company registered a $125 million loss in 1992. Internal company audits identified late departures and landings as the main cause of the enormous losses, which continued to balloon out of control, swelling to a $199 million loss in 1993 and a staggering $619 million in 1994. This amount of money hemorrhaging could not, of course, continue indefinitely, and Continental Airlines came perilously close to outright bankruptcy.
Continental executives correctly reasoned that they had to change the incentives they were offering employees if the company were to survive. Ensuring that airline flights adhere to tight time schedules is a “production process” that is only as reliable as its weakest link. A delay in even one of many preflight inspections and preparations that is required prior to every take-off is sufficient to cause significant lateness in getting the plane off the ground. After a long series of meetings analyzing the issue, management decided it would make one last, desperate attempt to save the airline from extinction by implementing what they code-named the Go Forward plan.
One of the central elements of Go Forward was the promise of a $65 bonus to every employee following every month in which the company was ranked among the top five airlines in on-time departures and landings. The plan’s effect was dramatic and immediate. Within one year, in 1995, Continental went from a $619 million loss to $224 million in profits.
Profits continued to soar, reaching $385 million in 1997. Interestingly, the monetary incentive aspect of Go Forward, which played a crucial role in saving the company, was a collective incentive, not a personal one. Teamwork was being rewarded, not outstanding individual effort.
So, here is the question that we’ll ponder in this chapter: what did the Go Forward initiative get right? As a business manager, how can you emulate this program to get better results?
Teams of workers in places of employment are a very interesting microcosm of social interaction, involving both rationality and emotions. Teams constitute an important element in economic and organizational efforts virtually everywhere around the world. A 1995 survey conducted by Paul Osterman revealed that more than 54 percent of organizations in the United States base their workforce activities on teams. In commercial companies that number was even higher, at 66 percent.
Advanced understanding of the behavior of teams in workplaces requires extensive use of mathematical models and game theory, because without first understanding how we should expect team members to behave under assumptions of rationality and self-interest, we cannot expect to understand how psychological and emotional phenomena contribute to a team’s success or failure. I have personally conducted several research efforts in recent years studying team behavior using game theory, more specifically a subfield of game theory called contract theory.
A contract between two or more individuals can be viewed as a game, because each contract defines rules of interaction (which parallel strategies in games) and furthermore specifies the payoffs to each party to the contract as a result of the actions undertaken by the parties within the framework of the contract (parallel to the payoffs in a game, which are determined by the strategies used by the players). This is how game theory enables us to answer questions about contract planning and contract negotiations. Using game theory we can ascertain what is the best contract from the perspective of contracting party A that will also be acceptable to contracting party B.
Research into this subject has expanded extensively in recent years. It now embraces many different research methods, both theoretical and empirical. Data on the subject have been collected based both on real-life observations and laboratory experiments. Some of the observations have been quite surprising, because they contradict our most basic intuitions regarding teamwork.
One surprising result reveals that the relationship between monetary incentives and work motivation is far from being as straightforward as is usually imagined. A paper I published in 2009 showed that, in the context of workplace teams, personal monetary incentives can have the effect of reducing work motivation.2 This has nothing to do with psychological effects, nor is it related to the distinction between internal motivation and monetary motivation (which was discussed in a previous chapter, as detailed in a research paper by Gneezy and Rustichini). It can occur in teams even when each individual is selfishly concerned solely with personal gain. I will present a small and simple model of teamwork to try to illustrate this paradoxical phenomenon and its practical implications.
Presenting all the details of this model will take several pages and will include logical arguments that some readers may find challenging. Readers who wish to do so may skip this presentation, as the rest of the chapter is self-contained.
Imagine yourself as the owner of a software company with two employees: one, named Mr. D, responsible for development, and the other named Mr. M, responsible for marketing. The software produced by the company can be profitable only if both development and marketing are conducted successfully. Each employee can choose one of two possible behaviors: investing a lot of effort at his job or investing very little effort. An employee who invests a lot of effort is guaranteed to succeed at his task, but if he invests little effort his chances of success are only 50 percent.
Software production and marketing are conducted in two consecutive stages: first Mr. D develops the software and then Mr. M markets it. The two employees are in different situations not only because the bulk of their work occurs at different time periods but also because Mr. M can see whether or not Mr. D is investing a lot or little effort, while Mr. D has no way of knowing whether Mr. M will work hard or not. You, as the owner, have the responsibility of fashioning an incentive system for your employees that will maximize the expectations that they will work hard. Unfortunately you have no way to monitor the true extent of the effort that each employee is investing in the job. The only thing you can measure with certainty is whether or not your company is seeing a profit (i.e., whether both development and marketing have been conducted successfully or at least one of them has failed). The incentive system you are considering offers both employees a bonus only if the project is successful and the company sees a profit.
To complete the model we need one more important detail: the extent to which an employee suffers from investing a great deal of effort at work. After all, if the employees enjoy putting in extra effort, there is no need to give them an incentive by offering a bonus. We will suppose that the suffering engendered by the effort required is equivalent to $1,000. This, however, does not mean that $1,000 is a sufficient amount to offer as compensation to each employee. The reason is simple: if $1,000 is the offered bonus, to be paid only if the project succeeds, then Mr. D will have no incentive to make an effort, because by making such an effort he will “lose” $1,000 worth of hard work while risking getting nothing if Mr. M does not contribute his efforts to making the project a success. If Mr. M does not invest in making an effort, the probability that the project will succeed will fall to 50 percent, in which case Mr. D will get compensation for the $1,000 worth of work he put into the project with a probability of only 50 percent.
Keep in mind that you, the owner of the company, have no way to know after the project has been completed which of the employees, Mr. D or Mr. M, has invested effort—both, one of them, or neither of them may have done so. The only thing you know is whether or not it has succeeded. How much, then, should you offer each employee as a bonus?
Consider the following sample bonus structure: Mr. D is offered a bonus of $1,400 and Mr. M is offered a bonus of $2,010 (with the bonuses paid only if the project succeeds). Let’s try to work out the rational considerations that the employees will take into account, assuming that they care only about their own personal welfare. Start with Mr. M, and suppose that Mr. M has observed Mr. D investing a lot of effort in the development phase of the project. At that point, if Mr. M decides not to make a strong marketing effort then the project will succeed with 50 percent probability. That means that he will receive $2,010 with 50 percent probability, which is equivalent to getting $1,005 in certainty (in economic terms this is the called the “expected payoff”). If, in contrast, he decides to make a strong marketing effort, he will suffer $1,000 worth of hard work but receive a $2,010 bonus with 100 percent certainty. On balance, he receives a total of $1,010, which is a higher sum than the expected payoff of $1,005 that he gets without putting in an effort. In this case Mr. M is better off investing effort in marketing (even though the payoff difference between making an effort and not doing so is only $5).
If Mr. M observes Mr. D not investing any effort in development, then he clearly has no incentive at all to make an effort himself, because under that condition the probability that he will receive a bonus drops to 50 percent even before he gets started. If he also makes no serious effort, the probability that the project will succeed now drops again, to 25 percent. Thus Mr. M’s expected payoff is only $502.50 (equal to one-fourth of $2,010). An investment of effort on his part is worth only $5 to him (an expected bonus of $1,005 minus the cost of $1,000 in the hard work he put in).
Next, let’s analyze the considerations of Mr. D, whose role in the software development/marketing process comes before that of Mr. M. If Mr. D invests a lot of effort in the development phase, he knows that Mr. M will also work hard (using the reasoning described in the previous two paragraphs), and the project will therefore definitely succeed. In that case, Mr. D will pay a cost of $1,000 worth of hard work but receive in compensation $1,400 in a guaranteed bonus. On balance he will receive a net payoff of $400.
Suppose instead that Mr. D decides not to invest any extra effort in development. Then Mr. M certainly will not invest any extra effort in marketing (as explained above). In that case the probability of the success of the project plummets to 25 percent (which is the probability that the project succeeds if neither employee invests any hard work into it). That means that Mr. D will receive the $1,400 bonus with a probability of 25 percent, translating into an expected payoff of $350. The conclusion of our analysis is that both Mr. D and Mr. M will decide to invest a good deal of hard work in the project, with the two of them earning a combined bonus of $3,410.
Now suppose that you, as the owner of the company, decide to offer the employees significantly higher bonuses, either out of empathy for the work they will have to do or because you believe that doing so will increase their incentives to work hard for your company. The new bonus offers are $1,900 to Mr. D and $4,020 to Mr. M for the successful completion of the project. Just as before, if Mr. M observes Mr. D investing extra effort, then it is worthwhile for him to do the same (because he will now receive a net payoff of $3,020 if he works hard and an expected payoff of $2,010 if he does not).
But consider what will happen if Mr. M observes Mr. D being lazy during the development phase. In that case Mr. M has a 50 percent probability of receiving his bonus if he works hard and only a 25 percent probability if he is also lazy in his marketing efforts. The first possibility grants Mr. M $1,010 (after taking into account the loss of $1,000 he experiences because of the extra work he puts in) and the second possibility yields him an expected payoff of only $1,005 (which is $4,020 divided by four). The conclusion is that in this case Mr. M always has an incentive to work hard in marketing, whether or not Mr. D has worked hard in the development. We see here that the attractive bonus offered to Mr. M does indeed boost his incentive to invest effort in his work. The stakes are simply too high for him, and he would prefer to exert effort even if Mr. D decides to shirk.
But the fact that Mr. M will be incentivized to work hard no matter what Mr. D chooses to do now changes Mr. D’s incentives. Previously, we concluded that Mr. D would have an incentive to invest extra effort on the development because he understood that if he were instead to be lazy, then Mr. M would observe that and definitely choose not to invest any extra effort himself. But under the new conditions Mr. M will choose to invest extra effort no matter what Mr. D does. What, then, is the best choice of action for Mr. D? If he works hard, the project will definitely succeed and he will receive a $1,900 bonus—but taking into account the price of $1,000 of hard work he will have to put in, the net payoff to Mr. D is only $900. Alternatively, should he choose not to work hard, the project will have a 50 percent probability of succeeding (since Mr. M will work hard no matter what), which translates into an expected bonus of $950—a higher payoff than what he receives if he invests a lot of effort into the project. In other words, Mr. D’s incentive to work hard has been reduced because he knows that Mr. M will do so in any event. Increasing the bonuses has actually worsened the situation. Previously both employees had an incentive to invest a lot of effort in their work; now only one of them is incentivized, even though both of them are rational and seek to maximize their material welfare.
This paradox arises from the mutual influences the employees have on one another. One employee’s incentives influence the other employee’s incentives. Promising the second employee too high a bonus reduces the indirect threat against the first employee, namely that if the first employee avoids hard work, then the second employee will also refrain from investing heavily, thus harming the interests of both of them.
Planning incentive structures can be a difficult task, one that must be conducted carefully. Our intuitions can easily lead us astray, with significantly negative consequences. I have termed this phenomenon, in which increasing bonuses offered to all the employees paradoxically reduces their incentives to work hard, “incentive reversal.” Even though the explanation may seem quite specific and technical, the emotional logic suggests itself often in reality. My research indicates that incentive reversal is quite a general phenomenon that can occur in nearly every organizational structure and any sized workforce.
I recently conducted a large lab experiment, along with several colleagues at the Max Planck Institute in Germany, which elicited a clear and strong instance of incentive reversal.3 The explanation I presented above for the existence of incentive reversal was based on the fact that the compensation an employee received for investing extra effort in his job was higher when the other employee also invested extra effort. In the example of the employees of the software company this characteristic followed from the development and marketing process, in which the weakest link in the chain determined the probability that the project would succeed. In other words, the employees complement one another: the success of the project depends on the success of both the development stage and the marketing stage. What would happen if the employees did not complement each other as described above, and instead were substitutes for each other (which would be the case if, for example, both of them were developers and the success of the project depended on only one of them making a successful effort)? The incentive reversal paradox would not happen in that case. This conclusion follows from both the mathematical model and the data observed in the lab experiment.
Next, consider what would happen if neither employee could observe the amount of effort that the other employee invested in the project. What then would be the best contract to offer them? In an article that I published in 2004 I mathematically proved that in that case the best thing to do is to create a certain discrepancy between the employees’ bonuses, even if they are entirely identical in their roles and capabilities.4 This is because an employee who is offered a lower bonus has added incentive to invest extra effort because he is then certain that the other employee (who stands to receive a larger bonus) will also invest extra effort.
That article received considerable attention. Some claimed that the theoretical advantage of discriminating between employees in the bonuses offered to them would be canceled by the anger against such blatant discrimination that some employees would feel. But two groups of experimenters who independently tested the theory in lab experiments came to the conclusion that a certain amount of moderate discrimination between employees, as described above, does indeed lead to increased work incentives.5,6 We naturally recoil from inequality, but when it serves our interests, we tend to accept it, even when we are the party who gets the lesser end of the deal.
In the rest of this chapter, we will no longer assume that individuals selfishly care only about their own material benefits and instead consider more realistic work environments in which self-interest, psychology, and emotions all interact. In such environments peer effects play significant roles, adding emotional and social incentives over and above monetary incentives that often improve team performances. Peer effects can cause an employee to avoid investing too much effort in her work if she knows or believes that most of her coworkers are themselves shirking from putting in extra effort, but they can also spur her to work much harder if she believes that the others are doing the same. Following are three examples of this, taken from the economics literature.
Italy, more than any other European Union country, is characterized by extreme cultural gaps between different geographical areas, especially the north of the country versus the south. In particular, the sharp contrasts in the work ethics of the north and the south have been a vexing subject for many Italian politicians. Two Italian researchers, Andrea Ichino and Giovanni Maggi, studied a database containing information on the behaviors of thousands of employees at one of Italy’s largest banks.7 The collected data on each employee included detailed information on the number of times the employee came to work late or failed to show up entirely, promotions to higher ranks in the bank hierarchy, and transfers from one branch to another. Using this information it was possible to identify bank employees who had moved from bank branches in the north to branches in the south and vice versa.
Ichino and Maggi discovered that bank employees who moved, for example, from Milan in the north to Naples in the south, exhibited extreme changes in their work behavior. Once in Naples, they were frequently late to the office and missed significantly more days of work. Since only sick days are officially considered acceptable reasons for missing days of work at the bank, one might surmise that the move from one city to another led to deteriorations in the health of transferred employees, but the researchers also discovered that employees transferred from Naples to Milan also exhibited different patterns of behavior, which in this case was expressed in lower rates of tardiness at work and fewer missed work days.
Further analysis of the database proved in a convincing manner that the only reasonable explanation for these changes in behavior patterns was the phenomenon of peer effects. Employees transferred from Milan to Naples very quickly learned that their colleagues in Naples had a weaker work ethic than the one they had been used to in Milan. This reduced the internal incentives they had to maintain the high standards of work ethic that they had been used to in Milan. In contrast, employees transferred from Naples to Milan learned (although apparently not as quickly as those moving in the opposite direction) that they were now in a very different work environment, in which their colleagues invested more time and energy at their jobs. This, of course, was an uncomfortable position for them, but it still created an incentive for them to adopt the work ethic of those around them.
In follow-up work, Ichino, this time in collaboration with Armin Falk of the University of Bonn, studied peer effects in a laboratory setting.8 In an early version of Ichino and Falk’s experiment, subjects (students) were employed in preparing invitation letters to a (fictitious) conference that was to take place in Zurich. Subjects were hired to work for fixed time periods at a set wage ($20 per hour). Their task was to fold invitation letters, insert them into envelopes, seal the envelopes, and stamp them.
The subjects of the experiment were divided into two groups, with each group in a separate room. The experimenter visited the first room several times over the course of the experiment, each time placing on a prominent table in the center of the room a thick parcel of envelopes ready to be stuffed with more invitations. In contrast, he entered the second room a relatively small number of times, bringing into the room only a small stack of envelopes on each visit.
The results of the experiment indicated that in the first room, where subjects frequently saw large batches of envelopes placed in front of them, the students made more of an effort to work hard than in the other room. As in the other experiment, Ichino and Falk explained the difference in the observed behaviors in the two rooms as due to peer effects. The subjects in the first group were given the impression that their peers were working hard, stuffing the envelopes at a rapid pace. A potential sense of shame if their accomplishments were to fall too far from the standards of their peers spurred these students to work particularly hard. Interestingly, the peer effect appeared here despite the fact that the subjects did not know one another.
In the other group, the effect went the other way. With only small numbers of new envelopes being introduced into the room, the subjects received the impression that they were managing to stuff more envelopes than the others. In other words, subjects who worked too hard started feeling like they were “suckers.” This is an unpleasant feeling, and they therefore slowed down their pace of work to avoid experiencing it. This experiment showed convincingly that it is not only monetary incentives that have effects on worker behavior—social incentives also play a significant role.
The third and last example presented in this chapter is, like the bank employees example above, based on data collected in real-life situations, not lab experiments. Two Berkeley researchers, A. Mas and E. Moretti, studied peer effects among checkout counter workers at a large American supermarket.9 Many work actions conducted by checkout counter workers are routinely recorded in computer databases (such as the start and end times of scanning each customer’s checkout items, and the number and types of items that a customer has brought to checkout). Using these figures, Mas and Moretti were able to estimate the efficiency of each worker in terms of number of items scanned during a given period of time. They discovered that when one worker ended a shift and was replaced by a different worker, the shift handover affected all the workers who were close enough to observe it. If the worker beginning a new shift was more efficient than the worker that he or she had replaced, nearby workers picked up the pace of their work. But if the new worker was slower, the others also slowed their pace accordingly.
These three research studies show that teamwork effort is significantly affected by the work environment. Each individual’s motivation to invest in harder work is increased if those around him are working hard. In the mathematical model presented above, this occurred because we assumed that the success of a project required that both employees put in extra effort, but this is not necessary in more general settings: psychologically, workers want to avoid feeling exploited on the one hand, while on the other hand they did not want to be perceived as exploiting their coworkers.
There are many other studies indicating that psychological and social elements are critically important to getting team incentives to work properly between peers. When interactions between employees at different levels of corporate hierarchies are studied, particularly those between supervisors and teams working under them, it turns out that the psychological and social elements can have even more decisive effects.
Unfortunately, the business world so far seems to have ignored a good deal of what has been learned in research studies focused on teamwork in particular and on incentives in organizations in general. Human resources directors and organizational consultants continue to stress monetary incentives (mainly bonuses) based solely on individual accomplishments, without considering team-based incentives. They do so not necessarily because they believe that team-based incentives may reduce worker efficiency, but for the simple reason that using standard and familiar methods based on purely individual incentives protects them from criticism if problems arise.
Since individual incentives require detailed monitoring of each individual employee’s effort, which is usually not feasible, human resources departments tend to use evaluation metrics that are often almost entirely irrelevant. This prompts employees to invest in optimizing their human resources evaluation scores instead of the job they were hired to do. Determining bonuses on the basis of such weak metrics distorts the system of incentives and furthermore engenders feelings of frustration, unfairness, and envy among employees, often leading to conflicts within teams instead of harmonious teamwork.
One of the most common evaluation metrics used by human resources departments is based on horizontal peer reviews, in which employees are asked to evaluate their colleagues’ work efforts and at times even rank their colleagues. In competitive work environments, in which the value of team cooperation wilts in the intense heat of each employee’s burning ambition to stand out in personal achievements, it should be self-evident that the trustworthiness of such reviews is extremely limited. Under this system employees have incentives to belittle the accomplishments of coworkers who threaten their positions while praising others who can later reward them. The reviews are subjective in any case, so this can be done without the employee being accused of lying.
A much more efficient incentive system is one that rewards successful teamwork while adding an additional but small bonus to the most industrious worker in the team. This harnesses together both a sense of group responsibility and a drive for personal achievement. In the description of the successful Continental Airlines incentive system that opened this chapter, it was not the offer of a gross benefit of $65 to each employee that was the key element to the project’s success, but rather it was the responsibility that the employees felt toward each other, with each employee wishing to avoid blame for denying the bonus from his or her coworkers.
Several weeks ago I was speaking with my mother on the telephone when I had to apologize for cutting the conversation short because I had promised to take several friends of my son to a basketball game in which they were playing. My son was ill that particular day and therefore could not join us. “Make sure everyone has their seatbelts on,” insisted my mother. “You are responsible for children that are not yours.” Later in the car I pondered what she meant. I had never heard her say anything similar when I was alone with my son in the car. She could not possibly be less concerned about her grandson than about his friends. Perhaps she trusted that my paternal protection instincts would naturally kick in when I was alone with my son and wanted to make sure I would follow those same instincts when his friends were in the car without him?
Eventually I understood that my mother’s admonishment stemmed from a moral imperative that requires a person to act especially carefully when he is responsible for children that are not his. A less extreme version of this same imperative is doubtless familiar to most of us. Imagine that it is your turn to take your child and your neighbor’s child in your car to their kindergarten when you discover that there is only one child seat in the car. Which child will you place in the child seat? Less drastically, consider the fact that we often avoid asking a friend to lend us an item out of fear that we will somehow damage it before we can return it. And if it does get damaged for some reason, we then feel ten times worse than if the same damage were to occur to an item we own.
Exactly the same moral mechanism comes into play when collective rewards are at stake, which is why they are so powerful. An employee working as part of a team is less concerned about losing her own bonus as a result of laziness on her part than about how it will look if her laziness costs her friends the bonus that they are entitled to.