«IZA DP No. 4262 PAPER Gift Exchange and Workers’ Fairness Concerns: When Equality Is Unfair Johannes Abeler DISCUSSION Steffen Altmann Sebastian ...»
ﬁnd that co-workers’ wages do not matter much for agents’ decisions. However, their design diﬀers from ours in several important points. While Charness and Kuhn focus on heterogeneity in productivity, we look at the eﬀect of actual output diﬀerences between agents. Furthermore, we allow for richer comparisons between the agents, as in their design agents are not aware of the magnitude and direction of the productivity diﬀerences. The diﬀerent results underline the importance of information for determining the reference group: Charness and Kuhn’s results rather apply to groups of workers that are loosely related and know little about each other, while our focus is on close co-workers who have a good understanding about their peers’ abilities and eﬀorts.
Regarding compensation practice in ﬁrms, our ﬁndings highlight the importance of taking the concerns for co-workers’ wages into account. However, doing so by paying equal wages to a group of agents may actually do more harm than good.
As soon as agents diﬀer in their performance, equal wages which seem to be a fair institution at ﬁrst sight might be considered very unfair. While the discouraging eﬀect of equal wages on hard-working agents has long been informally discussed (e.g., Milgrom and Roberts 1992, p. 418f) this paper provides controlled evidence in favor of this intuition. Moreover, it suggests that it is the violation of the norm of equity that causes the discouragement and low performance. Our results should not be interpreted as arguments against wage equality in general. They rather point to limits of equal wages.4 Wage equality is potentially a good choice in occupations where, e.g., due to technological reasons, workers’ performance diﬀers only slightly or where performance diﬀerences are due to random inﬂuences. In addition, the transparency of co-workers’ work eﬀorts and wages might have an inﬂuence on the optimal choice of the pay scheme.
The remainder of this paper is structured as follows. In the next section we describe the experimental design and discuss theoretical predictions. In Section 3 we present and discuss our results and Section 4 concludes.
Independent of equity-equality trade-oﬀs, equal wages might be beneﬁcial for the principal because they could increase peer monitoring (Knez and Simester 2001) and lower transaction costs since contracts do not have to be negotiated with every worker individually (e.g., Prendergast 1999).
2 Experimental Setup
2.1 Design and Procedures In the experiment, one principal is matched with two agents. The subjects play a two-stage game. In the ﬁrst stage, agents decide simultaneously and independently how much eﬀort they want to provide. Exerting eﬀort is costly for the agents. Eﬀort choices range from 1 to 10 and are associated with a convex cost function displayed in Table 1. The principal reaps the beneﬁts of production: every unit of eﬀort increases his payoﬀ by 10.
In the second stage, after observing the eﬀort decisions of his agents, the principal decides on wages for the two agents. The wages have to be between 0 and 100. Neither eﬀorts nor wages are contractible. The only diﬀerence between treatments is the mode of payment. In one treatment the principal can only choose one wage w that is paid to each of the agents (equal wage treatment or EWT). In the other treatment he can discriminate between the two agents by choosing wages w1 and w2 for agent 1 and 2, respectively (individual wage treatment or IWT). The EWT is thus a special case of the IWT. At the end of each period, the two agents and the principal are informed about eﬀorts, wage(s), and the resulting payoﬀs for all three players. The payoﬀ functions for the players are summarized in Table 2.
This game is played for twelve periods. We implemented a stranger design to
from confounding reputation eﬀects, i.e., at the beginning of each period principals and agents were rematched anonymously and randomly within a matching group. A matching group consisted of three principals and six agents. The subjects kept their roles throughout the entire experiment. After the last period, subjects answered a short post-experimental questionnaire. The experiment was conducted in a labor market framing, i.e., principals were called “employers” and agents were called “employees”.5 Our setup is related to the gift-exchange game introduced by Fehr et al. (1993) but diﬀers in two important aspects. First, in our experiment agents move ﬁrst while in Fehr et al.’s setup the principal moves ﬁrst. Our move order allows the principal to base his wage decision on the actually exerted eﬀort. More importantly, a principal in our experiment is matched with two agents instead of one. This is an essential prerequisite to analyze the interaction between gift exchange and payment modes. It allows us to study the impact of relative wages on the perceived fairness of the wage scheme and agents’ behavior.
All participants started the experiment with an initial endowment of 400 points that also served as their show-up fee. Points earned were converted at an exchange rate of 0.01 Euro/point. The experiment was conducted at the BonnEconLab at the University of Bonn in April 2005 using z-Tree (Fischbacher 2007). For each treatment, we ran four sessions with a total of 8 matching groups (144 participants). The experiment lasted approximately 70 minutes. On average subjects earned 8.30 Euro.
2.2 Behavioral Predictions
Eﬃciency is determined by agents’ eﬀort choices. It is maximized if both agents exert the highest possible eﬀort of 10. However, if all players are rational and selﬁsh the principal will not pay anything to the agents since wage payments only reduce his monetary payoﬀ. Anticipating this, both agents will provide the minimal eﬀort of one in the ﬁrst stage. The ﬁnite repetition of the game in randomly rematched groups does not change this prediction. This subgame perfect equilibrium is the same for both payment modes. If all players were selﬁsh we should therefore expect no An English translation of the instructions is available from the authors upon request.
diﬀerence between treatments.
By contrast, in laboratory experiments studying labor relations with incomplete contracts, one typically observes that eﬀorts and wages exceed the smallest possible value. Moreover, wages and eﬀorts are positively correlated (e.g., Fehr and G¨chter a 2000). These ﬁndings illustrate the potential of reciprocal gift exchange in enforcing incomplete contracts, as postulated in Akerlof and Yellen’s fair wage-eﬀort hypothesis (Akerlof and Yellen 1990). A fundamental prerequisite for the functioning of giftexchange relations is that workers perceive their wage as fair. The fairness of a wage payment, however, may not only be evaluated in absolute terms, but also relative to the wages of other members in a worker’s reference group.6 This is not important for the special case of bilateral gift-exchange relationships where only one agent interacts with one principal (e.g., Fehr et al. 1997). However, horizontal fairness considerations potentially play a crucial role in our setup where workers can compare to co-workers.
How do the behavioral predictions depend on which horizontal fairness principle is most important? If agents in the experiment care foremost about wage equality, the EWT—which guarantees equal wages by design—should lead to eﬃcient gift exchange between ﬁrms and workers. Additionally, we should expect no behavioral diﬀerences between treatments since ﬁrms in the IWT can pay their workers equal wages, too.
Given that ﬁrms in the IWT recognize workers’ desire for equal treatment, they will decide to do so. Thus, the wage-eﬀort relationship and average eﬀort levels should not diﬀer across treatments. If some ﬁrms nevertheless wage discriminate between workers, the IWT should lead to less eﬃcient outcomes than the EWT.
By contrast, if workers consider equity to be more important than equality, we should expect diﬀerences in behavior between treatments. The equity principle demands that a person who exerts a higher eﬀort than his co-worker should receive a higher wage and payoﬀ. Our experimental treatments diﬀer in the extent to which the equity principle can be fulﬁlled by principals. Under the equal wage institution, the equity norm is violated whenever agents diﬀer in their performance. Since both Potentially many variables inﬂuence a worker’s fairness perception of his wage, e.g., the unemployment rate, unemployment beneﬁts, the prevailing market wage, etc. (see Akerlof 1982, Akerlof and Yellen 1990). These factors are ruled out by our experimental design, allowing us to isolate the inﬂuence of co-workers’ wages on fairness perception and eﬀort provision.
workers receive the same wage but have to bear the cost of eﬀort provision, the worker who exerts more eﬀort receives a lower monetary payoﬀ. Under individual wages, principals’ behavior determines endogenously whether the equity norm is violated or not. By diﬀerentiating wages in accordance to eﬀort diﬀerences, principals can adhere to the norm. If we assume that at least some principals do so, we expect to see less norm violations in the IWT than in the EWT.
What are the behavioral consequences of such diﬀerences in norm fulﬁllment?
Agents who value equitable treatment should suﬀer from norm violations, feel dissatisﬁed and subsequently try to restore equity by adjusting their behavior. Equity theory proposes several possible reactions of agents after norm violations, such as altering own or others’ eﬀorts or payoﬀs, changing one’s reference group or quitting the relationship (see Adams 1965). The virtue of our experimental design is that we can clearly identify agents’ reactions, because the only variable that an agent can change after experiencing a norm violation is his work eﬀort. An agent who faces a disadvantageous norm violation (i.e., relative underpayment) should lower his eﬀort in the following period. An agent who experiences an advantageous norm violation (i.e., relative overpayment) should increase his eﬀort. Note that a norm violation always includes one agent facing a disadvantageous violation and one agent facing an advantageous violation. Dissatisfaction and the resulting strength of reactions, however, is likely to depend on the direction of the norm violation. Previous evidence suggests that the decrease of eﬀort after a disadvantageous norm violation will be stronger than the increase of eﬀort after an advantageous violation (Loewenstein et al. 1989, Mowday 1991, Th¨ni and G¨chter 2008). Consequently, a violation of the equity norm should o a lead to an overall decrease of eﬀorts in the subsequent period.
If workers care about equitable payment in the sense of the postulated equity norm, aggregate eﬀort in the EWT should thus be lower compared to the IWT since we expect to observe less norm violations in the latter.
In this section we present the results of the experiment and discuss possible explanations for the observed behavior. We ﬁrst analyze eﬃciency implications of the two payment schemes by comparing the eﬀort choices of agents. We then demonstrate that the diﬀerence in agents’ performance obtains even though monetary incentives— implied by principals’ wage setting—should lead to similar eﬀort choices in both treatments. Subsequently, we show that workers’ behavior seems to be strongly aﬀected by the equity principle, which is more frequently violated in the EWT. Finally, we report the results of an additional control experiment. They demonstrate that the higher eﬃciency of the IWT is not driven by the fact that principals can set two wages instead of one (as in the EWT) but by the fact that principals set wages that are in line with the equity principle.
3.1 Eﬀort Choices and Eﬃciency
Figure 1 shows the development of average eﬀorts over time. Under equal wages, eﬀorts are lower already in the ﬁrst period (Mann-Whitney test: p = 0.03)7 and decrease over time. Eﬀorts under individual wages stay constant (Wilcoxon test for periods 1–6 against 7–12: IWT, p = 0.56; EWT, p 0.01). This results in a strong overall treatment diﬀerence: average eﬀorts are almost twice as high in the IWT compared to the EWT (8.21 vs. 4.40; Mann-Whitney test: p 0.01). The treatment diﬀerence is also present when individual matching groups are considered: the highest average eﬀort of an EWT matching group (5.88) is still lower than the lowest average eﬀort of an IWT matching group (7.47).
The diﬀerence in agents’ behavior can also be seen in the histogram of eﬀort choices (Figure 2). In the individual wage treatment agents choose the maximum eﬀort of 10 in 49% of the cases; 84% of the choices are higher than 6. Under equal wages, agents choose an eﬀort higher than 6 in only 26% of all cases. The eﬀort decisions are more spread out in the EWT, the minimal eﬀort of 1 being the modal choice with 24% of The comparison of ﬁrst period eﬀort choices is based on individual observations. Unless otherwise noted, all other tests use matching group averages as independent observations. Reported p-values are always two-sided.
the choices. Since higher eﬀorts increase production and since the marginal product of eﬀort always exceeds its marginal cost, the diﬀerences in eﬀort provision directly translate into diﬀerences in eﬃciency.
Result 1: The two payment modes exhibit strong diﬀerences with respect to the performance they elicit: agents who are paid equal wages exert signiﬁcantly lower eﬀorts than agents who are paid individually. This results in much higher eﬃciency under individual wages.
Both, the agents and the principals beneﬁt from the increase in eﬃciency. The average proﬁt per period of a principal is 56 in the EWT compared to 100 in the IWT (Mann-Whitney test: p 0.01), while an agent on average earns 10 under equal wages vs. 17 under individual wages (Mann-Whitney test: p 0.01).
3.2 Wage Setting and Monetary Incentives