Leading by example with and without exclusion power in voluntary contribution experiments

Abstract

We examine the effects of leading by example in voluntary contribution experiments. Leadership is implemented by letting one group member contribute to the public good before followers do. Such leadership increases contributions in comparison to the standard voluntary contribution mechanism, especially so when it goes along with authority in the form of granting the leader exclusion power. Whether leadership is fixed or rotating among group members has no significant influence on contributions. Only a minority of groups succeeds in endogenously installing a leader, even though groups with leaders are much more efficient than groups without a leader.

Introduction

Leading by setting an example is an important means to influence the behavior of others when (often incomplete) contractual relationships or hierarchical authority are absent. In fact, many well-known historical persons have relied on this principle of voluntary leadership to promote the well-functioning of groups or society, think of Albert Schweitzer (“Example is leadership.”) or Mahatma Gandhi (“We must be the change we wish to see in the world.”).

In this paper we provide a comprehensive experimental study on three aspects of leading by example, which we will elaborate in detail below. First we examine the effects of leadership and the influence of leader's strength on behavior in voluntary contribution games. Second, we check whether it makes a difference for overall contributions if one single group member is always the leader or if all group members rotate in the leader's role. Third, we address whether groups want to install a leader by voting and how the outcome of the vote is related to contribution behavior.

Only recently (an early exception is von Stackelberg, 1934) the issue of leadership has received attention in the economics literature. From a theoretical perspective, Hermalin (1998) presents a model of leadership in the presence of asymmetric information between leaders and followers. He shows that a leader can induce others to follow suit either by example or by sacrifice. Arce (2001) and Foss (2001) identify leadership in their models as a means to achieve efficient outcomes in social dilemma games. From an empirical perspective, the effects of leading by example have received particular attention with respect to charitable fundraising. If well-known persons donate to a certain project and this is publicly announced, others often tend to follow (Vesterlund, 2003).¹ A similar effect arises when at the start of a public fundraising campaign a notable fraction of the goal is pledged as seed money (List and Lucking-Reiley, 2002). Several experimental papers have investigated the effects of voluntarily leading by example. Employing a full information setting, Moxnes and Van der Heijden (2003), Van der Heijden and Moxnes (2003), and Gächter and Renner (2004) have implemented leadership in the laboratory by letting a leader decide and announce his contribution before the other group members contribute (simultaneously). These studies find that leaders' and followers' contributions are highly correlated even in one shot games (Gächter and Renner, 2004), and that average contributions with leadership are (in most cases significantly) higher than without leadership.

Comparing a full information setting with a situation of asymmetric information – where a leader has private information about the returns from contributing – Potters et al. (in press) show that asymmetric information plays an important role in the success of leadership because it allows the leader to signal the returns from contributing to the follower. These findings of Potters et al. (in press) contrast somewhat with results from Meidinger and Villeval (2002) who report an experimental test of Hermalin's (1998) model of leadership. Meidinger and Villeval (2002) explain the success of leadership by reciprocity on the side of followers, but attribute only little influence to signaling. Parts of these differences as far as the experiments with asymmetric information are concerned might be due to Meidinger and Villeval (2002) relying on a partners design with repeated play in which reciprocity is more important than in the strangers design of Potters et al. (in press). Yet, as far as their full information treatments are concerned, both Potters et al. (in press) and Meidinger and Villeval (2002) report that followers condition their contribution on the contribution of their leaders, meaning that even in the absence of private information leadership can change behavior.

In our paper, we will rely on a full information setting and will concentrate on the following three aspects of leading by example. First, we reexamine the effects of leading by example. An important, but hitherto neglected, aspect of leading by example relates to the leader's formal authority. All previous studies have implemented leadership by letting the leader contribute before the followers, thereby restricting the leader's feasible options to setting an example and to reacting on followers' behavior in previous periods. However, leaders may have formal authority allowing them to discipline misbehaving followers, as it is typical in hierarchically structured organizations or as it is also possible in clubs where club members can use the club goods without rivalry in consumption, but where the president or the board of the club have exclusion power. Hence, we implement a treatment where leading by example is strengthened by the leader's power to exclude one follower from the group in the ensuing period.² Installing a leader with exclusion power seems akin to the opportunity of the leader of a work group to dismiss or suspend a member from the group.

Punishment through exclusion has costs not only for the excluded member but also for the group itself, because it reduces the number of potentially contributing members. Consequently, the possible efficiency losses from exclusion are endogenously borne by the group itself.³ Note that the possibility to exclude someone from contributing to and consuming the public good is at odds with the usual assumptions of public goods games. We therefore speak of voluntary contribution experiments in the following, but rely on the usual public goods scenario as our workhorse, except for the exclusion possibility (which will always apply to the next period, as will be explained in Section 2).

The second aspect of leading by example that we examine is the way in which a leader is appointed. Previous experiments have concentrated on the case where one single group member carries the burden of leading by example for the entire experiment. However, leading by example often rotates among the members of a given group or institution. For instance, in many European universities there is a tradition that the president of the university is elected in a rotating order from different schools. Similarly, professional organizations – like the American Economic Association – elect their presidents (who are supposed to lead by setting a good example) for a limited time. Given the various ways of determining leaders, we compare a treatment in which leadership is granted to a single group member to a treatment in which leadership rotates among all group members in a predetermined and publicly known order.

The third aspect we are interested in is whether groups actually want some member to lead by example. We examine this issue by letting groups vote on whether they want to have a leader or not. Allowing for an endogenous determination of leadership may provide insights into why some groups are more efficient than others and which group members are more likely elected as a leader. This aspect of our paper is related to the literature on endogenous institutional choice.

Early experimental studies of endogenously changing the institutional structure in common-pool-resource dilemmas have been carried out by Messick et al. (1983), Samuelson et al. (1984), and Samuelson and Messick (1986). Participants in these experiments could delegate their decision on how much to extract from a common pool resource via a collective vote to a leader. When the common pool was near depletion, subjects had a stronger tendency to delegate their decision to a leader. Vyrastekova and Van Soest (2003) have explored in a common pool experiment whether allowing resource users to vote on a natural resource management's incentive structure enhances the efficiency of resource use. In fact, compared to games in which appropriate incentives are imposed exogenously, behavior is more cooperative if a majority has voted for that structure. However, the quorum for implementing the appropriate incentives is failed in more than half of the cases what leads to less efficiency. Recently, Potters et al. (2005) have shown in a setting with asymmetric information that pairs of one leader and one follower are remarkably often successful in the endogenous sequencing of contributions. When given an option to vote on whether the leader's contribution shall be disclosed to the follower, more than 80% of leaders and 99% of followers support such an endogenous sequencing.

Our results show that leading by example has a positive impact on contribution levels. Moreover, strong leaders with exclusion power induce substantially higher contributions than leaders without such formal power. The way of appointing a leader – either a single group member or all group members in turn – has no significant influence on efficiency. When given the opportunity to endogenously select a leader, only about 40% of groups succeed in appointing a leader, even though groups with a leader outperform groups without a leader by far.

The paper proceeds as follows: Section 2 introduces the voluntary contribution game. Section 3 is devoted to the experimental design. The experimental results are presented in Section 4. A concluding discussion is offered in Section 5.