«IZA DP No. 3901 A Behavioral Account of the Labor Market: The Role of Fairness Concerns Ernst Fehr Lorenz Goette Christian Zehnder December 2008 ...»
One interesting aspect in these papers is that the selfish players do not merely raise their effort levels when they can acquire a good reputation, but they literally mimic the effort patterns of the fair-minded players in the non-final periods. This means that they provide low effort in response to low wages and high effort in response to high wages, thus contributing to the steepness of the wage effort relationship. As a consequence, employers face a strong incentive to pay non-competitive job rents because this enables them to elicit high effort from both fair and from selfish workers.
In the meantime, a series of additional papers has confirmed the robustness of the finding that reputation formation in relationships that evolved endogenously greatly fortifies the positive impact of fairness on performance. These papers provide evidence that reputation effects can be sufficiently strong to sustain very high levels of efficiency, even under rather adverse conditions.12 Fehr et al (2009)and Brown & Zehnder (2007) show that reputation formation in long-term relations leads to stable trades, even when fairness alone cannot prevent a market collapse. Fehr & Zehnder (2006) confirm this finding in a setup where asymmetric information implies that output is only an imprecise indicator for effort. Finally, Brown et al.
(2008) illustrate the effectiveness of the interaction between reputational incentives and fairness in a market with excess demand for labor and no threat of unemployment.
Evidence from the field on the role of gift exchange in repeated interactions is provided by a recent study (Bellemare & Shearer 2007) which reports observations from a field experiment in a tree-planting firm. Bellemare and Shearer implement a onetime wage gift and examine the workers' response. They show that effort increases significantly. The setup in this study is somewhat complex, as the tree planters are paid a piece rate and the income change induced by the wage gift may interact with these explicit incentives (see Goette et al (2004)) for a discussion of these motives). However, one may argue that it is even more remarkable to find a gift effect on effort in a piece rate environment where workers' marginal cost of effort provision is already presumably quite high. The study also provides evidence on the interaction between fairness and repeated game effects because they have workers in their sample who did not return to their employer during the next tree-planting season. They show that the effect of the wage gift on effort is significantly positive for both types of workers but for those who returned the next season the effect is stronger. This pattern confirms the complementarity between fairness effects and repeated game effects that is predicted by theory and observed in the laboratory.
Other evidence on gift exchange in repeated interactions is more indirect and circumstantial.
In these cases, firms changed the employment conditions in ongoing employment relations in ways that plausibly led to a negative effect on workers' fairness perceptions. It is interesting to study these episodes, because they can be interpreted as a permanent change in the firm's For a much more detailed discussion of these papers see Fehr et al (2009).
policy towards its workers. Krueger & Mas (2004) examine the quality of Bridgestone/Firestone tires manufactured in different plants and years. The plant in Decatur, IL, experienced serious labor strife after the company not only announced lower wages for new hires and unfavorable changes in the schedule for shift rotations, but also threatened to fire existing staff and hire replacement workers (and later did so.). The results show clearly that tires manufactured during the labor strife at Decartur were of significantly lower quality compared to the same type of tires manufactured at different plants in the same years.
In a similar vein, Mas (2008) presents evidence showing that a labor dispute at Caterpillar, a large manufacturing company producing construction equipment, tractors, and other vehicles, had a similarly negative impact on production quality. Negotiations between the union and management broke down after Caterpillar refused to accept a contract that the same union had closed with John Deere, a firm similar to Caterpillar. Relative to comparable Caterpillar equipment produced outside the U.S, the equipment produced in the U.S. during the labor strife shows a lower resale value. Since effort is an important determinant of quality in this business, this indicates that effort was lower during labor strife.
No studies exist that examine how the repeated nature of an employment relationship affects the effectiveness of gift exchange in labor markets. Maréchal & Thöni (2007), however, conduct an experiment in a related field study that allows them to tap into a similar business context. They show that sales representatives who visit stores to sell pharmaceuticals can increase sales by giving the store manager a gift – consisting of 6 samples of a product – at the beginning of their visits. Giving a gift strongly increases sales during the representative's visit from approximately CHF 60 in the condition without gift to about CHF 270 in the gift condition. Interestingly, the effect is only present if the sales representative had already visited the store; gifts on first visits lead to no change in sales. This suggests that the gift taps into an ongoing relationship between the sales representative and the store manager where the effects of fairness are predicted to be largest.
There is also evidence that employers' actions that are considered unfair trigger stronger responses than those that are considered fair. Mas (2006) examines the outcomes of finaloffer arbitration cases that involved police departments in New Jersey, in which the police officers' union and the city were unable to negotiate a new contract. In this case, it is rather clear which of the outcomes the workers find fairest, as they would not end up in final-offer arbitration if they did not disagree with the offer the employer made. The study documents a large and significant decline in many indicators of police performance (e.g., number of crimes cleared, probability of incarceration, etc.) if the arbitrator's decision favors the employers.
This decline in effort is highly sensitive to the size of the loss relative to the expected outcome from final-offer arbitration. If the arbitrator rules in favor of the union, there is a small positive effect on performance, which is almost completely insensitive to the size of the gain.
Lee & Rupp (2007) examine the impact of wage cuts for airline pilots on flight delays.
Virtually all the pay cuts they examine were consensual, i.e., the pilot union agreed to take a cut, as many of the airlines were in, or on the verge of, bankruptcy. There is essentially no effect on delays when pay cuts are consensual. This is in line with the survey evidence in Bewley (1999), showing that workers are willing to accept pay cuts when they feel they are justified, e.g., by the looming bankruptcy of the employer. However, there is one pay cut in Lee and Rupp (2007) that was the result of arbitration, and opposed by the pilots – a 26% pay cut at Alaska Airlines. This instance of a pay cut is thus comparable to Mas (2006), and the records show that it lead to a massive increase in flight delays at Alaska Airlines over the next several months, in line with the predictions from the fairness model.
To summarize, the field evidence on fairness manipulations in ongoing relationships shows that firms may incur extremely high costs if they treat workers in ways that are perceived as unfair. The evidence from Lee and Rupp (2007) also shows that, as indicated by survey evidence, that no adverse consequences follow from pay cuts that employees do not perceive as unfair (Bewley, 1999). Less is known, however, about the impact of treating workers in a way that is clearly perceived to be fair. While the results in Maréchal and Thöni (2007) are suggestive of positive effects from fair treatment, this remains to be documented in a labor market setting. In particular, long-term studies with explicit randomization or credible exogenous changes in compensation policies, such as adoption of a set of policies when a firm is bought by another firm, are needed.
4.3 Internal Labor Markets In a pioneering book, Doeringer & Piore (1971) assert that there is a sharp distinction between internal and external labor market arrangements. In particular, workers seem insulated from outside labor market conditions once they are employed in firms. They argue that these arrangements are difficult to explain from the viewpoint of a neoclassical model: "[W]e doubt that any of the major strands of conventional research will prove capable of assimilating the internal labor market into conventional theory in a useful and meaningful way." As we argue below, fairness preferences have interesting new implications for how firms set wages over time, giving rise to two of the most important features of internal labor markets.
The evidence on fairness perceptions suggests a shift in workers' perceived entitlement as they enter a firm (Kahneman et al 1986). When forming judgments about fairness, new workers compare the firm's offer to what they could otherwise earn in the labor market, evidence, while the evidence suggests that incumbent workers compare a proposed change in the employment relationship to the status quo.13 A second important regularity is that loss aversion appears to have a strong effect on fairness judgments. For example, a small decrease in the wage does much more damage to fairness judgments than a small increase in the wage does to boost fairness perceptions (Kahneman et al 1986). It is not clear, a priori, whether loss aversion in fairness judgments applies to the nominal or the real wage. The survey scenarios in Kahneman et al. (1986) hold the real wage cut constant, showing that over and above the loss in the real wage, individuals consider nominal wage cuts particularly unfair. Shafir et al (1997) also show that nominal wage cuts are genuinely perceived as more unfair and (Fehr & A similar effect can be observed in the fairness judgments of price changes. For example, Bolton et al. (2003) find that, in repeated transactions, the price that a firm charged last was the relevant reference price, much more so than the price the competitors were offering.
Tyran 2008; Fehr & Tyran 2001) document that money illusion prevents a quick fall in prices after a negative monetary shock in an experimental price setting game. As discussed earlier, Goette and Huffman (2007) present evidence that the salience of a nominal wage cut forms the fairness judgment. They show that if nominal wages are not cut, individuals care strongly about real wage changes.
These features give rise to three specific predictions in the theoretical framework we discussed earlier.14 The first prediction the model makes is that entry-level wages and the wages of incumbent workers respond differently to changes in labor market conditions. Entrylevel wages should strongly depend on labor market conditions. If the labor market is tight, workers can find alternative employment at relatively high wages. Thus, a high wage is needed to elicit high effort. When unemployment is high, workers' outside offers will be worse, and they will be willing to exert effort for a lower wage. As a consequence, the firm's optimal entry-level wage is lower when the labor market is slack. For incumbent workers, however, the reference outcome is the contract that was in place the last period, not the worker's outside options. This in itself makes the incumbent workers' wages relatively
independent of labor market conditions. The model also predicts cohort effects in wages:
because last year's contract becomes the reference outcome for this year, keeping the same contract is viewed as fair. Thus, if a worker started out with a high entry-level wage, this wage will become the reference wage for the next period, influencing future wage outcomes.
The third prediction is related to loss aversion: if workers' fairness judgments are more strongly affected when their situation worsens, then firms should be reluctant to cut wages.
The fairness model is silent as to whether real or nominal wages are the relevant measuring stick for fairness judgments. However the evidence in Kahneman et al. (1986) and Goette and Huffman (2007) suggests that nominal wage cuts in particular are considered unfair.
This section draws heavily on Benjamin (2005), in which proofs of all the statements can be found. See also Cabrales et al. (2008) for a theory of how social preferences affect labor markets. Their paper focuses on how concerns for fairness can provide incentives to segregate workers.
The evidence is generally supportive of these predictions. Several studies document that job changers' wages are more cyclical than those of job stayers. Recent studies include (Devereux 2001; Devereux & Hart 2006; Haefke et al 2006; Solon et al 1994). In all studies, wages of individuals entering firms are far more sensitive to business cycle variations. It should be noted that the fairness model does not predict that the incumbents' wages will never change.