پاداش های نسبی در جبران خسارت مبتنی بر تیم

How to design compensation schemes to motivate team members appears to be one of the most challenging problems in the economic analysis of labour provision. We shed light on this issue by experimentally investigating team-based compensations with and without bonuses awarded to the highest contributors in teams. A purely team-based compensation scheme induces agents to voluntarily cooperate while introducing an additional relative reward increases effort and efficiency only when the bonus is substantial. In this case, however, the data suggests that tournament competition crowds out voluntary cooperation within a team.

Teamwork is increasingly seen as an appropriate structure to organise various labour environments (Beyerlein, 2000, Mueller et al., 2000, Prat, 2002, Zwick, 2004 and Van Hootegem et al., 2005). The suitable provision of incentives for teams, however, appears to be one of the most challenging tasks in labour economics (Main et al., 1993, Dematteo et al., 1998 and Hamilton et al., 2003). In general wage contracts of teams are not conditioned on individual contributions (as it is the case in piece rate contracts) because contributions can neither be (easily) disentangled nor verified before a court. This is one of the reasons why team members are often rewarded according to the output of the team as a whole. Another purpose is to encourage cooperative behaviour in the sense that individual team members strive for the best outcome of the whole team. From a strategic point of view, however, such a scheme provides considerable free-riding incentives, which might lead to inefficient effort levels (see, for example, Alchian and Demsetz, 1972, Newhouse, 1973, Holmström, 1982 and Hansen, 1997). An intuitive approach to reduce the severity of free-riding incentives in teams is to promise relative rewards to the best individual performers in the team ( Heneman and Von Hippel, 1995). In practice relative rewards within teams often take the form of bonuses 2 (for example, for the “employee of the month”) or promotions which imply higher salaries. One reason why relative rewards are so popular stems from the fact that output needs only to be measured relatively which constitutes a big advantage when absolutely measuring individual outputs is prohibitively costly. Even if the individual contributions of team members are not verifiable to a third party, the employer can commit himself to a relative incentive scheme. The commitment is credible as long as the individual (relative) outputs are observable (for example by the team leader) and the act of actually transferring the ex-ante promised relative rewards is verifiable (for example, before a court). 3 Relative reward schemes that induce competition among team members for positions in a ranking are also referred to as tournaments ( Lazear and Rosen, 1981). Why might an additional tournament structure alleviate the free-rider problem in teams? The idea is that the additional competition for rewards between team members induces them to exert more effort which would align the individual interests with the responsibility for the common good. We aim at contributing to the understanding of how compensation schemes affect human behaviour by providing a clear-cut comparison of a hybrid pay scheme to its components. We concentrate on the strategic and behavioural aspects of team-based compensation schemes with and without a relative reward by investigating them in a game-theoretic model and in an abstract non-real effort experiment. By doing so we provide evidence that additional individual rewards in teams are likely to crowd out voluntary cooperation. To the best of our knowledge a crowding out effect for the highly realistic compensation package composed of tournament incentives and team incentives has not been documented before in the literature. Laboratory experiments have the decisive advantage that one can compare behaviour under different clear-cut conditions without distorting effects which are normally present when collecting data from the field. Thus, experiments appear to be a valuable complementary source of insight (in addition to theory and field studies) in order to gain a more complete picture regarding the behavioural consequences of reward schemes (Falk and Fehr, 2003). In addition to a purely team-based compensation treatment we consider two treatments in which the team incentive is supplemented by a relative reward — a low and a high bonus, respectively. As a benchmark we also look at a pure bonus situation without a team component. A team-based compensation with a relative reward is modelled such that it is self-funded by letting the agents finance the additional bonus (in equal shares). This guarantees that the wage costs for the employer are the same under the different remuneration schemes (provided that the total amount of effort is the same). Our results indicate that average effort and efficiency increase significantly only if the bonus is high. A low bonus does not have a noticeable influence on effort or efficiency. Moreover, only in the absence of a relative reward (i.e., only with the purely team-based compensation) significant over-contributions (compared to the equilibrium effort level) are observed which decline over time. Interestingly, in the team-based compensation scheme with the high bonus average effort is quite stable over rounds with almost non-existing over-contributions right from the beginning. Why do over-contributions vanish in the presence of a high relative reward? One reason could be that pure tournament incentives induce effort levels below the equilibrium prediction. If this is true over-contributions and under-contributions would simply cancel out in the hybrid compensation scheme. However, in our pure bonus benchmark treatment actually observed effort levels fit the equilibrium prediction quite well — if at all, we only find evidence in the direction of over-contributions. Thus, our data suggest that the competition induced by an additional individual relative reward seems to destroy the cooperative attitude of team members and, in this sense, crowds out motivation to exert effort above individually rational levels. From a strategic, behavioural point of view it seems that a pure team reward enables team members to see the necessity for cooperation more clearly. The promise of an individual relative reward tempts team members to focus on obtaining the bonus and, thereby, seems to distract them from potential gains of voluntary cooperation. Some studies conducted in natural environments already suggest that one should be careful when combining group compensation and individual rewards. Rosenbaum et al. (1980), for example, find that groups performed relatively poorly when hybrid rewards (i.e., group-based rewards augmented by individual rewards) rather than pure group rewards were given for performance on highly independent tasks. They concluded that even a small proportion of individual rewards undermines performance of a group task. A study by Wageman (1995) points into a similar direction. She argues that the introduction of group level rewards in addition to individual pay weakens worker's sense of individual responsibility, without providing strong enough collective motivation. It has also been observed that additional extrinsic incentives (provided for example by rewards or punishment) do not necessarily lead to an increase but can even reduce effort. The detailed mechanisms that drive such crowding out effects, as they are termed by Frey (1997), still constitute a largely unresolved puzzle. Evidence in favour of crowding out effects has been provided by Gneezy and Rustichini, 2000a and Gneezy and Rustichini, 2000b. By conducting imaginative field experiments they show, for example, that students who individually collect donations for charity programs perform better when they receive no monetary reward at all compared to those who are paid a small bonus proportional to the amount they collect. To some extent these observations can be explained by an approach proposed by psychologists, according to which an activity can have a motivation of its own, called intrinsic motivation. They argue that intrinsic motivation may be diminished by extrinsic incentives such as performance contingent monetary rewards. 4 Our setting is designed to deliberately eliminate the effect of intrinsic motivation which might emerge from an (enjoyable) real effort task. By representing effort through the choice of abstract numbers we concentrate on an additional, important aspect of motivation crowding out, which is constituted by the purely strategic aspect of crowding out of voluntary cooperation, without allowing for intrinsic motivation. Fehr and Rockenbach (2003) show that the performance of agents in an abstract, non-real effort task is considerably reduced if a principal chooses incentives in the form of a punishment scheme. An analogous but weaker effect of performance crowding out is reported in Fehr and Gächter (2002). They find that even the promise of a performance contingent reward of a fixed size may undermine voluntary effort contributions. 5 Gneezy and Rustichini (2000a), Gneezy (2005), and Pokorny (in press) report evidence that extrinsic incentives influence effort in a non-monotonic way: While relatively small monetary incentives bear the potential to crowd out motivation, relatively large incentives increase effort at least as long as the resulting wages are below a certain reference level. All these studies look at work environments for individuals without considering team settings. In line with other experimental studies on team compensation schemes (for example, Nalbantian and Schotter, 1997, Croson, 2001 and Sutter, 2006) we focus on a strategic setting similar to a voluntary contribution mechanism for public goods (for overviews see Davis and Holt, 1993, Ledyard, 1995 and Camerer, 2003). Similar to public goods situations our team members equally share the group output. However, there are also some differences: In our team compensation scheme it appears to be more suitable to implement a convex and increasing cost function for effort which leads to an interior equilibrium solution (normally supply functions with constant marginal costs are considered in public goods which result in a corner solution as an equilibrium). Additionally, in our setting output in teams is determined not only by effort but also by a stochastic influence, which represents luck or (subjective) measurement uncertainties of individual contributions. One general finding in the public goods literature is that in repeated settings individuals cooperate to a high extent above the equilibrium especially in the beginning of an interaction while cooperation diminishes over time. This effect was initially observed in linear public goods games (for example Isaac et al., 1984) in which the equilibrium effort (no contribution) and the efficient effort (full contribution) are both corner solutions. The two most prominent arguments for the decay in contributions are: (i) participants learn to free-ride over rounds and (ii) participants play history-dependent strategies which lead both conditional cooperators and free riders to decrease their contributions over time. Although some studies report hints that the learning explanation cannot be completely rejected, a majority provides evidence in support of a more pronounced effect of the “strategies” explanation. 6 Muller, Sefton, Steinberg, and Vesterlund (2006), for example, report a cleverly designed experiment: They analyse two public goods games which are consecutively played by the same participants by eliciting complete strategies. This two-stage game is repeated five times always with a new matching such that no participant is ever matched with another participant more than once. They find that on average, stage-two contributions are 45% lower than stage-one contributions while the stage-one contributions decrease only by on average 7% in each repetition. In line with this evidence in favour of the “strategies” explanation, initially high contribution levels and a subsequent decay are also the rule in non-linear public goods games with an interior equilibrium as we have in our setting ( Keser, 1996, Sefton and Steinberg, 1996 and Nalbantian and Schotter, 1997). A second category of experimental work which is related to our study is constituted by experimental studies on tournament incentive schemes (see for example Bull et al., 1987, Van Dijk et al., 2001, Harbring and Irlenbusch, 2003, Orrison et al., 2004 and Harbring and Irlenbusch, 2005). Among other findings in general they confirm one of the central results from tournament theory: Effort increases in the size of the bonuses. None of these studies, however, deals with the integration of individual and group incentives. The three experimental studies which are most akin to our analysis also investigate individual incentives as a way to increase effort among members of work teams. Dickinson and Isaac (1998) model the use of individual monetary rewards in team production with workers of different abilities by conducting conventional public goods games with heterogeneously endowed agents. They introduce bonuses for high absolute contribution levels and bonuses for high relative (to endowment/ability) contribution levels. Absolute and also relative rewards significantly increase contributions to the public goods compared to a situation without any bonus. Furthermore, they find that the increase is larger if relative rather than absolute bonuses are paid. Dickinson (2001) extends the former study by additionally analyzing negative incentives in the form of penalties. Bonuses (carrots) and penalties (sticks) are awarded based on absolute or relative (to endowment/ability) contributions. He finds that not only bonuses but also penalties increase contributions which are highest if the relative to endowment measure is used. If the absolute measure is used penalties seem to induce higher contributions than bonuses. Sutter (2006) investigates team compensation schemes in the form of threshold public goods games. In his innovative extension team members compete for higher remuneration in the form of high contributors “earning” the role of an early proposer in an alternating-offer bargaining process. Hereby the tournament prizes can be endogenously determined by the team members which nicely resembles the situation of self-managed teams. All three studies mentioned constitute interesting starting points in the experimental analysis of the combination of individual and group incentives. Our approach is different as we compare the impact of two standard compensation schemes (team compensation and relative rewards) to their combined effect by integrating the labour tournament model suggested by Lazear and Rosen (1981) into an adapted voluntary contribution mechanism. The paper is structured as follows: In Section 2 we introduce a simple model of team-based compensation with a bonus and derive a game theoretic equilibrium benchmark. In Section 3 our hypotheses are presented. These are followed by the experimental design and procedures in Section 4. The experimental results are summarised in Section 5 while Section 6 contains an additional comparison with a pure bonus compensation scheme. Section 7 concludes by discussing practical implications and pointing at interesting extensions and further research.

This paper experimentally investigates the incentive effects of a compensation package often observed in business organisations: team-based compensation with an additional relative reward for the highest contributor in the team. From a theoretical point of view an additional bonus within a team-based compensation scheme should increase output since the individual incentives can be separated into two additive components: the team incentive component and the tournament incentive component. If the bonus is high in the experiment we indeed observe a significant increase in effort which is not the case if the bonus is low. In the pure team compensation scheme and when the bonus is low we observe high over-contributions (i.e., voluntary cooperation above the individually rational level) which decline over rounds. Surprisingly, this standard result known from public goods experiments disappears if the bonus is high. In the team-based compensation scheme with the high bonus average effort is quite stable, although right from the beginning on a very low over-contribution level. It seems that an additional, substantial bonus awarded to only one of the team members induces higher effort but crowds out voluntary cooperation. Given these observations designers of compensation packages face a trade-off when introducing a tournament structure into a team setting. Relative rewards are likely to have counteracting effects. On the one hand, they increase output because of higher individual incentives. On the other hand, there are drawbacks from distracting team members to act cooperatively. The net result from these two effects is not obvious. We have seen that when the bonus is small crowding-out of voluntary cooperation is not a big issue but individual incentives remain on a relatively low level. With a high relative reward individual incentives clearly increase but might come with the cost of a diminished cooperative attitude. In this sense, our results advise some caution for the implementation of tournament incentives into team-based compensation. Why exactly does a high relative bonus reduce voluntary cooperation? Evidence from previous public goods experiments suggests that participants over-contribute (above the individually rational level) for strategic reasons. By contributing highly in the beginning they seem to aim at creating a cooperative atmosphere in the group to convince others to contribute highly as well. Naturally such an investment in creating cooperation pays off more in earlier rounds than in later ones. Competition introduced by a bonus for one of the team members seems to destroy the cooperative attitude within the team. One reason might be that relative rewards do not only create winners but leave also losers who could considerably be discouraged from contributing to the team's task (Müller and Schotter, 2005). Another reason might be that participants perceive a pure team situation with equal payoff—sharing differently than a team situation in which members compete for a bonus. Such a framing effect induced by different compensation schemes (Irlenbusch and Sliwka, 2006) might enable team members to focus more on the potential gains from voluntary cooperation under the pure team incentive scheme than under a pronounced tournament structure. Under the latter they might primarily see their team mates as competitors which reduces the scope for cooperation. To analyse these and other explanations for the detailed mechanisms behind the observed crowding-out effect in teams appears to be a promising field for further research. Additional studies are also needed to investigate whether the observed crowding-out effect can be mitigated and whether the decline of cooperation can be damped. It could well be that alternative procedures for awarding the bonus can preserve the cooperative attitude in the team, for example, by increasing team members' participation through majority voting. Additionally, in our study we abstract from peer effects which certainly play a significant role in teams (Falk and Ichino, 2003). It would be quite interesting to analyse whether peer pressure could help to sustain cooperation in a team when a relative reward is introduced. Another interesting question emerges from the observed decreasing trend in voluntary cooperation. Is there a way to stop or reduce this decline? The study by Isaac, Walker, and Williams (1994) provides some hints that the decline becomes flatter the longer the time period during which the team members interact. This would suggest that the life-span of a team should not be too short. The observed crowding-out of voluntary cooperation by a relative reward appears to be even more problematic when team members have the possibility to sabotage each other. Sabotage is indeed known to be exerted in tournaments in order to improve the own position in a ranking for rewards (Harbring and Irlenbusch, 2005). We consider our study as a starting point for investigating crowding out of voluntary cooperation in teams through compensation schemes which introduce individual competition. By bringing up the above questions we hope that our observations do not only suggest practical implications for providing incentives in teams but also point at several interesting issues for further research.