The moonlighting game

Abstract

We introduce the moonlighting game. Player A can take money from or pass money to player B, who can either return money or punish player A. Thus, our game allows to study both positively and negatively reciprocal behaviour. One-shot experiments were conducted with and without the possibility of making non-binding contracts beforehand. We find that retribution is much more compelling than reciprocity. Although contracts are not binding they increase trust, but we do not find evidence that they also encourage reciprocity.

Introduction

Consider the following situation: an illegal moonlighter has been engaged for some piece of work. He has access to a till containing money in order to buy materials, and is supposed to be paid for the work after he has finished. Since the whole activity is forbidden, neither the moonlighter’s performance nor the principal’s payments can be legally enforced. In this situation, the moonlighter has several options: he may take money out of the till and disappear. He may not work at all. Or he may work at an arbitrary activity level, where more effort causes higher costs for him, but also a higher principal’s surplus. After the work has been done (or not), the principal has several options: she may pay him the amount agreed upon, but she might as well pay less or nothing. However, if she was betrayed by the moonlighter, she can do nothing against him but go to court and sue him for damages, but since both parties violated the law, both will bear negative consequences.

It is clear that according to the orthodox assumption of pure rationality no exchange of work and money will take place, the moonlighter will rather take as much as possible out of the till. He knows that the principal will neither pay him nor go to court, for both options are costly for the principal. Hence, the moonlighter can neither expect to be rewarded for his effort nor to be penalised for embezzlement.

Looking at the little example, we see that the game considered here is characterised by four features: first, the moonlighting activity improves the situation of both players, and thus, at least if the view is reduced to the two parties immediately involved, increases efficiency. Second, an agreement is settled to gather the surplus. Though it is not binding, it establishes a social relationship where the two parties declare a common interest. Third, the mutual improvement requires trust and reciprocity. The employer must feel an obligation to pay the bill after the work has been done, even if there is no legal means to force him to do so. Fourth, the moonlighter faces the employer’s unspoken threat of retribution, the fear of being punished for betrayal, despite the knowledge that this is not rational.

We designed an experiment integrating all features mentioned above: The principal (player B) proposes a non-binding contract to the moonlighter (player A) specifying the actions to choose in the following play. Player A can accept or reject the proposal. The contract will not affect further play in any way, thus it is cheap talk in the game theoretic sense. After having accepted or rejected the contract, player A decides upon either taking money from the second mover, which represents the embezzling action, or passing money to her. The passed amount is tripled by the experimenter, standing for the surplus gained by the moonlighter’s activity. The second mover now can either return money, i.e. pay the moonlighter, or specify a fine that she imposes on him, which is also costly for her. This action represents the option that the principal claims for damages in court, leading to prosecution of both parties.

A central feature of our moonlighting situation is that both kind and unkind moves are feasible for both players. It is an open question how the co-existence of both opportunities affects human behaviour with respect to trust, reciprocity³ and retribution. In the existing experimental studies, however, either positive or negative reciprocation is examined. Fehr et al. (1993) analyse (positive) reciprocity in a labour market context. They observe that experimental firms systematically overpay workers, compared to the competitive equilibrium wage, to induce an increased effort. In the investment game by Berg et al. (1995), the first moving player can also (but only) pass money to the second mover, which is tripled by the experimenter. The second mover can voluntarily return money. The results clearly refute the hypothesis of subgame perfect rationality and support the impact of reciprocal fairness.⁴ In contrast, Van Huyck et al. (1995) find support for strategic behaviour and reject fairness and trust hypotheses. In their peasant–dictator game the peasant decides about an investment which reduces his current credit but results in a multiple future taxable income. The dictator imposes a tax on that income. In one treatment the dictator chooses the tax rate before, in a second condition after the peasant decides on his investment. Strategic considerations imply a positive investment in the first treatment, while in the second one the peasant would not invest anything. The authors report that their experimental data are highly correlated with these strategic predictions.

The role of retribution is focused mainly by the experimental literature on the ultimatum game⁵. In these experiments, low offers are typically punished by rejections. Our moonlighting game involves a more incremental punishment: more punishment is more costly, and, contrary to the ultimatum game, it gets the more expensive to punish player A down to a given payoff level the more aggressive the first mover’s demand has been. Fehr et al. (1997) use a related punishment facility in a third stage which they add to their gift exchange game. They find that the threat of (non-rational) punishment increases workers’ reciprocity, and by that high levels of co-operation can be achieved.

In many real life situations legally binding agreements are either impossible or their transaction costs are prohibitive, but there is at least a facility to agree upon a non-binding arrangement. Does the opportunity to conclude contracts effect reciprocal behaviour even if the contracts are not binding? There are only few studies which analyse the impact of non-binding contracts. Irlenbusch (in press) studies a five stage goods exchange game and finds that with non-binding contracts a considerably high goods exchange activity takes place, even more than in a control treatment in which some players have to adhere to the contract. However, in absence of a control group without contracts an isolated effect of non-binding agreements cannot be inferred from these results. To gain insight into this question, we conducted our experiment in two treatments: in one condition we left out the pre-stage in which the contract is made.