«IZA DP No. 3901 A Behavioral Account of the Labor Market: The Role of Fairness Concerns Ernst Fehr Lorenz Goette Christian Zehnder December 2008 ...»
2.1 Preferences for Fairness The prototypical game to test whether individuals not only care about their own payoffs, but also about their payoffs relative to those of others, is the ultimatum game, first proposed by Güth et al (1982). In the ultimatum game, a proposer can make an offer of how to divide, say, $10 to a second party, the responder. The responder can accept or reject the offer. If he accepts, the parties receive the split the proposer offers. If the responder rejects, both parties receive nothing. A robust result in these games is that offers of less than 20 percent to the respondent are often rejected, while offers of 40 percent or more are usually accepted (Camerer 2003; Fehr & Schmidt 2002; Roth 1995). As a consequence, many proposers make offers that are close to the equal split. These results challenge the selfishness hypothesis, because the game is played only once, meaning a strictly selfish responder should always accept any positive amount offered. A plausible interpretation for the rejection of low offers – which is typically also supported by responders’ verbal accounts of their behavior – is that subjects perceive them as unfair. The responder incurs a psychological cost if he or she only receives a small share when the proposer could have chosen an equitable split. If this psychological cost is large enough, the responder may prefer forgoing his share in order to avoid the unfair outcome.
Indications of social preferences are not limited to rejections of low offers in the ultimatum game. There exist many other (anonymously played) one-shot games in which observed behaviors cannot be reconciled with the selfishness assumption. In dictator games – a modified version of the ultimatum game in which the responder cannot reject the offer – many people transfer positive amounts although they could earn more money by keeping everything for themselves (Kahneman et al 1986). In public good games many people deviate from the dominant strategy of complete free-riding and voluntarily contribute to the public good (Ledyard 1995) and – if given the possibility – take on additional costs to punish noncontributors (Fehr & Gachter 2000). Many trustors in trust games place resources at the disposal of a trustee in the (often correct) expectation that the trustee returns the favor and pays back (Berg et al 1995). And in gift-exchange games (which will be discussed in detail in section 4) subjects in the role of workers are frequently willing to provide higher effort than contractually enforceable if their experimental employer provides them with a generous wage rent (Fehr et al 1993). These examples not only illustrate that fairness matters in various contexts, but they also show that fairness manifests itself in different forms. Besides the retaliatory or punishing behaviors observed in the ultimatum game; a substantial share of subjects also exhibit a tendency towards altruistic choices and frequently display positively reciprocal behaviors, i.e., they respond to generous acts with generosity. However, there is an emerging consensus among experimentalists that nonselfish punishment motives are often considerably stronger than nonselfish motives to behave in altruistic or positively reciprocal manners – an asymmetry that may cause asymmetric effects of wage cuts and wage increases on workers’ behavior (Offerman 2002).
The evidence for the existence of social preferences is also not confined to student subject pools but holds in subject pools that are representative for whole countries such as Germany (Fehr et al 2003; Dohmen et al 2009), the U.S. (Naef et al 2008), and the Netherlands (Bellemare & Kröger 2007; Bellemare et al 2008); it holds under stake levels of up to three months incomes (Cameron 1999; Fehr et al 2002; Slonim & Roth 1998) and across many different cultures (Oosterbeek et al 2004; Roth et al 1991). However, it is also important to stress that the evidence shows considerable individual heterogeneity in the strength of social preferences, i.e., a significant share of individuals also exhibits fairly selfish behaviors, a fact that turns out to be important in the way social preferences affect labor market phenomena.
Various theories have been suggested for modeling this type of social preferences (Benjamin 2005; Bolton & Ockenfels 2000; Charness & Rabin 2002; Falk & Fischbacher 2006; Fehr & Schmidt 1999; Rabin 1993). While they differ in how they model the details of preferences for fairness and reciprocity, they share the important common feature that individuals with such preferences are willing to pay to reduce ostensibly unfair outcomes or to punish unfair behaviors. Some authors have argued that rejection in the ultimatum game can be reduced through sufficient experience with the game (Binmore et al 1995; Roth & Erev 1995). In light of the procedures used in the experiments, we view such explanations as somewhat farfetched, as the consequences of rejecting an offer in the one-shot version of the ultimatum game are straightforward and easily understandable. More recently, neuroscientific studies have corroborated the view that fairness concerns and the punishment of unfair behavior are genuine expressions of preferences. A recent paper (De Quervain et al 2004) finds that reward related neural circuits are activated if subjects decide to punish unfair behavior. These neural activations even occur if subjects have to pay to punish. Thus, punishing unfair behavior is, in terms of neural circuitry, no different from spending money for buying a positively valued good. Likewise, other authors (Tabibnia & Lieberman 2007) found stronger activation of reward related brain circuitry if subjects receive an ultimatum game offer of $5 when the stake size is $10 than when they receive a $5 offer from a stake size of $30. Thus, keeping absolute economic payoffs constant, the reward circuitry in the human brain apparently treats a fair outcome as a reward in itself. It is also interesting that diminishing the activity in the brain region most important for humans’ ability for rational decision-making – the prefrontal cortex – limits subjects' ability to reject unfair offers (Knoch et al 2007; Knoch et al 2006);
this stands in clear contrast to arguments suggesting that subjects who reject unfair offers do not understand the nature of one-shot interactions (Binmore et al. 1995). More specifically, experimentally induced reductions in the neural activation in the right dorsolateral prefrontal cortex – a brain region that is crucially involved in cognitive control of mental operations and prepotent impulses – substantially reduces rejection rates in the ultimatum game. Thus, those brain regions crucial for rational choices are also needed for making fair choices, consistent with the notion that fair choices can be perfectly rational.
2.2 The Reference Frame of Fairness Judgments The evidence we reviewed so far establishes that many individuals are willing to incur costs in order to avoid unfair outcomes or punish unfair behaviors. While the simplicity of the ultimatum game obviously renders the 50/50 split a fair outcome, at least in Western cultures5, many economic transactions may be more complicated. For example, individuals may have no choice but to make unfair offers, or changes in the economic environment may give one party a better bargaining position. How do such factors affect fairness judgments and what is their relevance for the labor market?
Falk et al (2003) provide evidence on how the economic environment can affect fairness judgments by conducting a series of ultimatum games in which they restrict the set of possible offers the proposer can make to the respondent. In some treatments, the proposer can only make unfair offers, while the proposer has a choice between unfair and fair offers in others.
The evidence indicates that responders are more likely to reject the unfair offer when it was intentional, i.e., when the proposer had a choice between a fair and an unfair offer, thus favoring models of fairness in which not only the outcome matters, but also how the outcome came about (Dufwenberg & Kirchsteiger 2004; Falk & Fischbacher 2006).
In some remote small scale societies (e.g. among the Machiguenga, a small tribe in the Peruvian Amazon) such fairness norms seem to be absent (Henrich et al. 2001). Likewise, chimpanzees also don’t reject low offers in the ultimatum game (Jensen et al. 2007).
Evidence from surveys indicates that individuals feel strongly about the perceived fairness of wage changes. Kahneman et al (1986) show that individuals' fairness judgments about whether a pay cut was fair depend strongly on the reason why pay was cut: workers consider it highly unfair if a firm wants to cut pay simply because labor market conditions deteriorated, while cutting pay to save a firm from bankruptcy is more acceptable. Evidence from surveys with personnel managers and compensation officers shows that they also anticipate these motives in their employees (Agell & Lundborg 1995; Agell & Lundborg 2003; Bewley 1999).
In the context of labor markets, an important and old question is whether individuals solely take the real buying power of their wages into account, or whether they also care about nominal wages. Dating back at least to Keynes (1936), some economists have assumed that individuals are not fully aware of changes in the price level, and thus also care about the nominal wage. Goette & Huffman (2007) conduct a study with students about to enter the labor market. In their sample, fairness judgments are solely a function of the real wage, with one exception: Nominal wage cuts trigger very strong negative reactions. They argue that the salience of a nominal wage cut elicits strong negative emotions that sway the students' judgment. This evidence suggests that fairness judgments are primarily affected by real wages, except in the case of nominal wage cuts.
The reference dependence of fairness judgments and the associated framing effects have also been stressed by Kahneman et al. (1986), who show that wage cuts – which are perceived as losses – are viewed more unfairly than identical reductions in real pay that are perceived as an elimination of a gain. They gave, for example, respondents a scenario of a firm that paid a ten percent annual bonus over a longer period, and then abolished it. The vast majority of respondents considered this as fair, even though it effectively cut the workers' incomes by 10 percent. Other respondents were given a scenario where the workers' base wage was cut by 10 percent. In this case, the majority of the respondents stated that the action is unfair.
Taken together the experimental evidence overwhelmingly suggests that reference-dependent social preferences are an important component of the human motivational repertoire. This does, of course not mean that such preferences always play a role. In fact, the social preference models mentioned above make clear that in certain competitive environments such preferences may play little role (Bolton & Ockenfels 2000; Falk & Fischbacher 2006; Fehr & Schmidt 1999). Therefore, it is important to describe the relevant labor market environment in some detail.
3 The Economic Environment: Labor Markets and Employment Relationships Many microeconomic textbooks still treat labor markets as a special case of neoclassical trade theory. Instead of trading physical products, the employer and the worker simply exchange well defined units of labor at a price market competition determines. The literature on the theory of the firm has long recognized that this simple view is not accurate, as it neglects many important aspects of real life employment relationships (Coase 1937; Simon 1951;
Williamson 1975). Most occupations involve multidimensional and complex tasks that are hard to describe in a perfectly fashioned complete contract. The problem is that when the parties conclude the contract, they cannot foresee all the relevant contingencies that may become important in the course of their relationship. As a consequence, the typical employment relationship is governed by an incomplete contract that does not perfectly specify important aspects of the collaboration between the employer and the worker. Contractual incompleteness makes it hard for third parties like a court to evaluate whether the parties have met their obligations, therefore severely limiting the judicial enforceability of the contract. As a result, the employer faces a fundamental problem: if duties and obligations are only vaguely specified, how can the employer motivate the worker to provide more than minimal effort?
3.1 Explicit versus implicit incentives If the worker's output is easy to measure, pay for performance may be a simple solution to the motivation problem. If the worker's earnings depend on his output, it is in his interest to exert non-minimal effort, even though nobody can legally force him to do so (Grossman & Hart 1983; Harris & Raviv 1979; Ross 1973). However, agency theory also provides important arguments for why firms may want to refrain from explicit performance pay. One problem with explicit incentives is the so called "multi-tasking problem" (Baker 1992; Holmstrom & Milgrom 1991). Output is complex in many jobs, and only some of its many dimensions are objectively measurable, while others are not. In these cases, the provision of explicit incentives may cause distorted outcomes because workers will allocate all their effort toward those activities the incentive scheme rewards. Incentive pay is thus not effective if workers should devote time and effort to activities which contribute to the non-measurable dimensions of output. There is laboratory (Fehr & Schmidt 2004) and field evidence (Prendergast 1999) for the relevance of the multi-tasking problem. Another related problem is that performance pay may create severe intertemporal problems. The issue is that the employer will always be tempted to adjust the standard for a performance evaluation based on the worker's past performance. If the worker anticipates that there is a tendency for the standard to increase after a period of good performance, the incentives to provide high effort are diluted – a phenomenon that has been called "the ratchet effect" (Carmichael & MacLeod 2000; Freixas et al 1985; Gibbons 1987; Kanemoto & MacLeod 1992; Laffont & Tirole 1988; Lazear 1986).
In order to avoid the ratchet effect, the employer may want to commit to fixed piece rates. But even if such a commitment is credible, the limited ability to react to innovations and progress may reduce the attractiveness of explicit performance pay considerably for the employer.