ESSAY MIN 2 FULL PAGES / INSTRUCTIONS ARE ALSO ATTACHED.

ContentServer.pdf

Home >Business & Finance homework help >ESSAY MIN 2 FULL PAGES / INSTRUCTIONS ARE ALSO ATTACHED.

L E A D E R S H I P A N D M O T I V A T I O N F O R P U B L I C G O O D S C O N T R I B U T I O N S

Bryan C. McCannon*

ABSTRACT

Results from a leader–follower public goods game are presented. An individual, when randomly selected to make a contribution knowing others will observe the

selection, gives more than in the simultaneous-move public goods game. Follow-

ers adopt a quasi-matching strategy where they systematically donate less than

the leader, but contribute more when the leader does and contribute less when

the leader free rides. The net result is increased provision of a public good when

contributions are sequential. The results highlight that psychological preferences,

rather than solely social preferences, can explain behavior.

I INTRODUCTION

A classic dilemma in public economics is how to discourage free riding and

encourage contributions to public goods. A significant amount of experimen-

tal research has investigated this (Ledyard, 1995; Chaudhuri, 2011) dating

back to the early studies of Marwell and Ames (1981) and Isaac et al. (1985).

In the standard Public Goods Game considered, individual contributions

improve the outcome for others, but have a negative net effect on personal

benefit; i.e., the marginal return to the individual for each dollar contributed

is less than one dollar.

To rationalize the contributions that are made, these works typically refer-

ence social preferences, or rather, ‘other-regarding’ preferences (Fehr and Fis-

chbacher, 2002). It is argued that there exists a component to each person’s

utility function that incorporates the well-being of the other players, or rather,

there is a ‘warm glow’ benefit (Andreoni, 1995). Examples of social prefer-

ences are altruism, inequality aversion, egalitarianism, and fairness (see

Eswarah and Kotwal (2004) for an example). Having social preferences pro-

vides a trade-off between personal gain and the benefit of others, which can

explain the lack of complete free riding.

An alternative argument promoted here is that psychological preferences

are also important to explaining behavior. In psychological game theory, first

introduced by Geanakoplos et al. (1989) and extended to extensive-form

*West Virginia University

games by Battigalli and Dufwenberg (2009), the preference ordering over the

set of possible outcomes of a game can be made dependent on other players’

beliefs regarding one’s strategy selections, rather first-order beliefs, or a

player’s beliefs on other players’ beliefs regarding one’s strategy, second-order

beliefs, or so on. An example of a psychological preference is guilt aversion. 1

The important distinction between social preferences and psychological prefer-

ences is that in the former, an individual cares about the outcome of the

game, while in the latter, a player is concerned with how others view the strat-

egy taken.

In standard public goods games, one is, for the most part, unable to differ-

entiate social and psychological preferences. This is due to the fact that

choices are made not only anonymously but also simultaneously. 2 In practice,

though, public goods contributions are frequently made sequentially. A fund-

raising campaign, for example, may first rely on a large initial donor before

marketing the project to other targeted donors. Sequential public goods con-

tributions allow for leadership decisions to influence outcomes.

If psychological preferences are not important, then an individual will sim-

ply trade-off personal gain and the outcome obtained by others. In fact, a the-

oretical model is presented where if only social preferences matter, the leader

contributes less than the followers. If a first-mover acknowledges that the fol-

lowers have strong social preferences, then his lack of contribution will

increase the marginal utility of giving by the followers. Thus, a leader with

strong social preferences will be motivated to give less than if contributions

were simultaneous. Alternatively, if psychological preferences are important

drivers, then being in a leadership position, where others will observe and

respond to your giving, encourages greater contributions.

To appreciate the role of leadership and psychological preferences on deci-

sions, experiments of the Public Goods Game are contributed. In the design

used, subjects are put into groups of four and are each endowed with five

experimental dollars. One is selected at random to make his contribution first.

The randomly selected leader’s contribution is made public to the other mem-

bers of the group before they make their contributions. The total amount

given is tripled and evenly shared. In a second treatment, the four in the

group each make a contribution without knowing the offer of any other per-

son.

Results show that leaders make larger contributions to the public good than

in situations where there is no sequencing of decisions. Furthermore, followers

adopt a quasi-matching strategy. When the leader makes a larger contribu-

tion, the followers do so as well, but when the leader free rides, so too do the

1 Huang and Wu (1994) discussed how psychological game theory can act through social

norms to reduce deviant, criminal behavior. Also, see Colman (2003) for an early important discussion. McCannon (2017) used it to capture the role of contingent cooperation prefer- ences and social norms in explaining cooperation in Prisoner Dilemma games.

2 An important exception is the experimental design of Dufwenberg et al. (2011) were they

use different framing techniques to explore psychological preferences. They do not attempt to differentiate it from social preference setups.

LEADERSHIP AND PUBLIC GOODS 69

followers. The net effect, though, is that sequential public goods contributions

outpace simultaneous giving. The results, then, provide support for psycholog-

ical preferences also being an important component of behavior.

Additionally, choices made if selected as the leader are collected from each

subject, along with contributions offered if a follower, knowing the leader’s

giving level. This allows for an investigation of switching behavior of an indi-

vidual and how this responds to leadership. Not only do subjects pull back

offers when not acting as the leader, but the magnitude of their retrenchment

depends on the contribution of the leader. This is further strong evidence that

psychological preferences are important drivers of outcomes.

Sequential contributions to public goods were first investigated theoretically

by Varian (1994). He illustrates that there exist environments where less is

contributed than the simultaneous-move game. Sequential contributions,

though, allow for emotions, based on beliefs about play of others, to be

expressed. For example, Battigalli and Dufwenberg (2007) formulate guilt, by

letting other players down and corresponding guilt aversion behavior, in psy-

chological games. Dufwenberg et al. (2011) applied this theory to public

goods contributions. They conduct experiments on the Public Goods Game

investigating the accuracy of psychological game theory at explaining play.

Their focus is on the framing of the game to the subjects and its effect on

players’ beliefs. Strong evidence is presented that changes in the presentation

of the game adjust subjects’ beliefs and, consequently, contributions.

This is not the first experimental investigation of sequential public goods

contributions. Levati et al. (2007) considered sequential public goods experi-

ments with one player selected at random to be the leader. The focus is on

asymmetric endowments and incomplete information. The aggregate outcomes

they find match those identified here in that leaders make larger contributions

and followers’ donations adhere to those of the leader. They do not investi-

gate the determinants of individual-level behavior, as is done here. G€uth et al.

(2007) introduced the ability of a leader to exclude a player in the next round

of a repeated game and show that this increases contributions further.

G€achter et al. (2012) investigated how the attributes of the individual explain

the heterogeneity of leader contributions in such environments. Andreoni

et al. (2002) considered a public goods situation where players have ideal

aggregate contribution levels. Sequential contributions, then, predict free rid-

ing by the first-mover. They show, indeed, contributions are less by the leader

than the follower and total giving is below simultaneous-move contributions.

Relatedly, G€achter et al. (2010) compared and contrasted a similar two-player

public good contribution, but focus on asymmetric returns to donations. They

too find that aggregate contributions are lower in the sequential-move treat-

ment when the player who values the public good more moves first. Alterna-

tively, Haigner and Wakolbinger (2010), Rivas and Sutter (2011), and Arbak

and Villeval (2013) considered public goods contributions when subjects can

choose to donate first before other members. Voluntary leadership leads to

higher contributions than exogenously selected leadership. This matches the

findings of Potters et al. (2005). They consider two-player voluntary

70 BRYAN C. McCANNON

contribution games where only one player is informed of the value of the pub-

lic good. Allowing a voting round to determine whether the informed must

make her contribution first reveals that, frequently, the players unanimously

agree to sequence the contributions. This leads to greater giving than treat-

ments with exogenous timing. They do not allow for differing sizes of contri-

butions. In a follow-up project, Potters et al. (2007) differentiated signaling by

the informed leader from the possibility of social preferences. They fail to find

any difference between giving in the sequential and simultaneous-move game

without the asymmetric information.

The results presented here contribute to this literature by focusing on the

individual-level decision-making. This is possible due to the experimental

design employed. By randomly selecting the leader after each subject has

made his or her choice of contribution as a leader, the amount offered can be

compared to what was actually given if, instead, chosen to be a follower. 3 The

two-step decision process allows for a study of both leadership choices, along

with shirking and matching decisions as a follower. Hence, rather than focus

on endogeneity of leadership, asymmetric information/endowments/returns, or

exclusion, the work here is able to investigate, in depth, the utility function

driving behavior.

The results of the laboratory experiment conform to field study observa-

tions. Beekman et al. (2014) showed that public good giving in rural Nigeria

is less in areas that experience more corruption by political leaders. Thus, fol-

lower’s matching strategy depends on their beliefs about the actions of the

leaders. In a field study in Bolivia, Jack and Recalde (2015) considered

resource giving to local schools. Public good contributions increase when

democratically elected politicians lead by example. They argue that the leaders

are motivated by the public’s view of their behavior and that followers are

more likely to follow when the leader contributes. The work presented here

provides a complementary investigation by studying the form of these prefer-

ences.

The work also contributes to the growing literature on leadership in eco-

nomics. Weber et al. (2001) reported on experiments where a randomly

selected member of the group attempted to facilitate coordination. Komai

et al. (2011) presented results from an experiment where the payoffs take the

scenario of a public good nature or one of coordination. Treatments vary by

whether all subjects or just a leader is informed of which scenario is active.

They present results indicating that if the information is concentrated with the

3 Arbak and Villeval (2013) conducted a related analysis. In their design if two or more

volunteer to lead, one is selected at random and the others become followers. They extend their analysis to study an ordered variable of revision up, down, or no-change. They show that revision is driven by gender and leader contribution, as is shown here, but their results suffer from endogeneity and do not consider nonlinear effects. Also, G€achter et al. (2012) employed a similar design. They classify individuals based on whether they provide high, low, or intermediate contributions when a follower. They show that leader contributions are correlated with follower type and conclude that driving leader contributions is social con- cerns. It does not, necessarily, study switching behaviors, but rather correlation between type of follower and leadership choices.

LEADERSHIP AND PUBLIC GOODS 71

leader, so that followers only observe the leader’s choice, public good contri-

bution rates are higher. There was not an important difference between

leader–follower and simultaneous investments in the coordination scenario. The work is extended to include the salience of gender in Grossman et al.

(2015). Similarly, Sahin, Eckel, and Komai (2015) also experimentally contrast

behavior in public goods and coordination games. They differentiate leader-

ship as ‘leading by example’ from managerial cheap talk suggestions. They do

not, though, find important differences in the types of leadership in the public

goods treatments. Leadership is more effective for coordinating activities.

Ertac and Gurdal (2012) investigated leadership in risk-taking decisions for a

group. In a social psychology investigation of leadership in public goods, Van

Vugt and De Cremer (1999) had participants play one round of a threshold

public good game and then rate, on a 1–7 scale, on their preference for having each of six ‘types’ of leaders in such a setting. The subjects’ most-preferred

type of leader was ‘democratic’ who collects desired contribution levels from

all members of a group before making giving decisions. These are just refer-

ences to experimental work on leadership. For theoretical investigations to the

study of the economics of leadership, see Hermalin (1998), Rauch (2001),

Komai et al. (2007), Komai and Stegeman (2010), and Lazear (2012).

Finally, the results suggest an alternate explanation for using sequential

donations in fund-raising campaigns. Vesterlund (2003) and Andreoni (2006)

documented and discussed the practice of fund-raising announcing past con-

tributions. The theoretical environment they explore is a signaling model,

rather than the behavioral framework explored here. The announcement of a

substantial donation before soliciting additional contributions, along with sig-

naling the value of the project to the uninformed, may also tap into psycho-

logical features of one’s preferences that elicit greater donations from both

following and leading contributors. Relatedly, Andreoni and Petrie (2004)

showed that information and identification of donors can increase contribu-

tions. Romano and Yildrim (2001) considered the theoretical problem of char-

ities announcing donations as they accrue.

Section II describes briefly the theoretical framework, while Section III

describes the experiment design and procedures. Section IV presents the

results. An extension to the experimental methods, evaluating the main

results’ robustness, is done in Section V. A concluding discussion occurs in

Section VI.

II THEORY

The objective was to outline a straightforward theoretical framework of a

public goods game that incorporates psychological preferences and differenti-

ates predictions from outcomes with individuals who have selfish or social

preferences. The model presented is an extension of the environment devel-

oped by Dufwenberg et al. (2011).

Consider a public goods game with four players labeled 1, 2, 3, and 4. Let

P = {1, 2, 3, 4} be the set of players. Each player selects an action

72 BRYAN C. McCANNON

ai 2 Ai = {0, 1, 2, 3, 4, 5}, which represents her contribution. Let bij denote i’s ‘first-order belief’ about j’s choice (i 6¼ j; i, j 2 P), i.e., bij is the mean of a probability measure i has over the possible values over Aj. Let ciji denote i’s

‘second-order belief’ about bji.

If individuals are standard wealth maximizers, then their payoff function is

uWi ðaÞ ¼ 5 � ai þ 0:75ða1 þ a2 þ a3 þ a4Þ ð1Þ (i 2 P) where a = (a1, a2, a3, a4) is the contributions of the four players. As a consequence, in the Nash equilibrium of the simultaneous-move game ai* = 0 for all i, or rather, full free riding with no public good contributions arises.

Alternatively, suppose individuals have social preferences where they care

about not only their own outcome but also the earnings of others. The payoff

function can be expressed as

uSi ðaÞ ¼ 5 � ai þ 0:75ða1 þ a2 þ a3 þ a4Þ � riSðxÞ ð2Þ (i 2 P) where x is the difference with the average contribution of the other three players, x = �a � ai = (aj + ak + al)/3 � ai where j, k, l 6¼ i. The parameter ri = 0, then represents selfish preferences, while ri > 0 allows for social preferences to be influencing behavior. One can expect dS/dx > 0 for x ≥ 0 and S(0) = 0 so that S is the disutility from deviating from the contri- butions of others. It may also be reasonable to presume dS/dx ≤ 0 for x < 0 where a disutility is also experienced for exceeding others’ contributions. This

would arise from inequality averse individuals (Fehr and Schmidt, 1999) or

norm compliance preferences (Bicchieri, 2006; McCannon, 2017). Alterna-

tively, though, this disutility can be set to dS/dx = 0 for x < 0. With social preferences, nonzero contributions to the public good can arise.

As an illustration, suppose S(x) = max {0, x}. If the social preference com- ponent of the payoff function is sufficiently high (ri > ¼), the gap between other’s contributions and earnings and the player’s is harmful. Consequently,

it is in the best interest of the decision maker to match the contributions of

others. Any level of nonzero contribution, then, can be rationalized.

Consider a sequential public goods game. Suppose player 1 selects first, so

that the other three players know his contribution prior to making their

choices. It is straightforward to verify that the structure of the game (simulta-

neous vs. sequential contributions) has no effect on the outcome if players are

selfish (ri = 0) as ai* = 0 for all i 2 P.

Proposition 1: In the sequential public goods game with selfish preferences

(ri = 0 for all i), the unique Subgame Perfect Nash equilibrium entails com- plete free riding by both the leader and followers (ai* = 0 for all i), which matches the giving in the simultaneous-move public goods game with selfish

preferences.

The decision-making in the sequential public goods game when players have

social preferences differs from that in the normal-form game. Similar to the

LEADERSHIP AND PUBLIC GOODS 73

result derived in Varian (1994), the leader, in a public goods giving environ-

ment, contributes less than the followers.

To formalize this claim suppose, as a second illustration, that S(x) = x 2 .

As in a Stackelberg Game, the followers (players 2, 3, and 4) best response to

the known contribution of the leader (player 1). In the subgame perfect Nash

equilibrium, then, the leader’s choice not only affects his contribution to the

public good but also acknowledges that his contribution drives the decisions

of the followers (due to their social preferences). In fact, the more he contri-

butes, the less the followers do, while if he free rides, then the giving of the

followers increases. The following proposition provides the Subgame Perfect

Nash equilibrium with social preferences.

Proposition 2: In the sequential public goods game with social preferences

(where the social preferences are not too low, i.e., ri > r for all i for a thresh- old r > 0), the Subgame Perfect Nash equilibrium entails a (weakly) lower contribution by the leader, than the followers.

Thus, one would expect to see nonzero contributions if players have social

preferences and in the sequential game leaders would give less than the fol-

lowers.

An alternative payoff function, explored here, is that psychological prefer-

ences, rather than social preferences, explain contributions to public goods.

The primary distinction between the two is that with social preferences,

players care about the earnings of the other players (modeled here as a disuti-

lity from contributing less than the others). With psychological preferences,

players care about what others think of their actions. If a player believes that

others expect a greater contribution out of him, then a greater disutility is

experienced when free riding than when a player believes that others expect

less from him. It is necessary to compare and contrast the anticipated contri-

butions between players with selfish, social, and psychological preferences in

both the normal-form and extensive-form public goods game to delineate the

value of the competing mechanisms.

In Dufwenberg et al. (2011) a simultaneous-move public goods game with

psychological preferences, which they refer to as a guilt-averse utility function,

is formulated,

uGi ða; cÞ ¼ 5 � ai þ 0:75ða1 þ a2 þ a3 þ a4Þ � ci � maxf0; ðciji þ ciki þ ciliÞ=3 � aig ð3Þ

where ci = (ciji, ciki, cili), i 6¼ j 6¼ k 6¼ l; i, j, k, l 2 P, is the third-order beliefs of i. The parameter ci ≥ 0 measures the degree of guilt aversion, as it captures how important is deviating from the average of what i believes the others

expect out of him.

While both social preferences and psychological preferences generate a disu-

tility from being different than other three players, the important contrast

between the two is that social preferences model the difference in outcomes

74 BRYAN C. McCANNON

(contributions and earnings in the public goods game), while psychological

preferences model how a player’s giving differs from his expectations of what

others believe he will play. It is this important distinction that matters for the

mechanism’s operation.

In the simultaneous-move game, again, nonzero contributions can be

rationalized. For example, suppose ci > ¼. The best response from (3) is to select ai = (ciji + ciki + cili)/3. Thus, if player i believes others expect him to contribute generously, then he will do so. Consequently, numerous equilibria

outcomes arise at varying levels of expectations, which are consistent with

play in equilibrium.

Consider the extension to sequential contributions. As the leader, player 10s psychological preferences drive him to experience guilt if he does not ‘live up

to’ the expectations of the followers (or, more specifically, his beliefs of others’

expectations of his behavior). Hence, let

uLða; c1Þ ¼ 5 � a1 þ 0:75ða1 þ a2 þ a3 þ a4Þ � k1fðc1j1 þ c1k1 þ c1l1Þ=3 � a1g ð4Þ

be the utility function of one with psychological preferences of ‘leadership

expectations.’ Thus, k1 = 0 represents standard, selfish preferences, while k1 > 0 is a player with psychological preferences for being a good leader. Fur- thermore, suppose followers experience guilt about contributing less than the

leader. Hence, let

uFi ðaÞ ¼ 5 � ai þ 0:75ða1 þ a2 þ a3 þ a4Þ � uiFða1 � aiÞ ð5Þ be the payoff to a follower, i 2 {2, 3, 4}. Thus, φi ≥ 0 measures the degree to which free riding off the leader creates disutility to a follower, or rather, the

importance placed on being a ‘good follower.’ Again, φi = 0 represents the standard, selfish preferences, while φi > 0 allows for other-regarding prefer- ences.

In the sequential public goods game, if psychological preferences of leader-

ship expectations and being a good follower are not important to an individ-

ual (k1 ≤ 0.25, φi = 0), then the best response for the individual is to free ride, ai = 0. Followers care primarily about personal wealth and do not con-

tribute. The leader, who does not put much weight on defying expectations,

fails to provide, even partially, the public good.

Alternatively, suppose that others’ expectations regarding the leader’s role

is important to the leader. Consequently, substantial leadership contributions

arise, exceeding those driven solely by self-interest or social preferences, or

that would arise in the simultaneous-move game. If there are no expectations

on the followers’ behavior, then their decision reverts back to the level driven

solely by social preferences and/or utility from personal, financial wealth.

4 One can add a guilt aversion term to the payoff function of the leader and followers as

well without changing the main arguments presented. Also, one could replace F(a1 � ai) in equation (5) with F(a1 � {(ciji + ciki + cili)/3}) where the followers’ disutility is driven by psy- chological preferences rather than social preferences. The theoretical framework and experi- mental design, though, are unable to distinguish between the two setups.

LEADERSHIP AND PUBLIC GOODS 75

Instead, if individuals care about following the leader, then nonzero contribu-

tions occur.

Proposition 3 provides a description of the equilibria under the assumption

that F(zi) = zi 2 where zi = a1 � ai. The functional form is used to derive a

closed-form solution.

Proposition 3: In the sequential public goods game with psychological prefer-

ences, the leader contributes more to the public good than the followers.

One can model a number of theoretical frameworks to extend dynamic

psychological game theory to sequential public goods contributions. The

model provided, though, does illustrate how the theory can be applied to

better understand the preferences of individuals. Testable predictions

arise.

1 In the simultaneous-move public goods game, zero contributions are

expected out of selfish individuals, while positive contributions can be

explained by either social preferences or psychological preferences.

2 In the sequential public goods game, behavior does not change if indivi-

duals have selfish preferences.

3 In the sequential public goods game, leaders contribute less than followers

if individuals have social preferences.

4 In the sequential public goods game, leaders contribute more than fol-

lowers if individuals have psychological preferences.

Thus, the theoretical framework provides contrasting predictions based on

the particular utility function describing players. Experimental data can be

used to test these predictions to identify whether behavior is consistent with

the theoretical framework provided.

III EXPERIMENTAL METHODS

To address this issue experiments were conducted with undergraduate students

at a small, private university in upstate New York. Subjects were recruited

from general education classes. Additionally, individuals were recruited from

classes within the business school. 5 An online reservation manager was used

to schedule the sessions. The recruitment strategy targeted students in classes

taken by underclassmen, along with classes taken by upperclassmen, in both

the general education courses and those within the business school.

A total of nine experimental sessions were conducted in November 2012

and April 2013. The number of participants in each session ranged from 12 to

17. There were a total of 147 experimental subjects and no subject engaged in

the experiment twice. Each session lasted approximately 1 hour. Subjects com-

pleted a background information questionnaire and engaged in the experi-

ment.

5 Economics is within the school of business and the faculty are joined with those in

finance into one department.

76 BRYAN C. McCANNON

Two treatments were considered. In treatment 1, the subjects played the

Public Goods Game. In this game, the subjects were randomly selected into

groups of four. Each person in the group is endowed with five ‘experimental

dollars’ (hereafter E$) and chose how much to contribute to a ‘common pot’.

The subjects were informed that the pot tripled and then was evenly shared

among the four group members. In treatment 2, the subjects played the Lea-

dership Game. In it, again, the subjects were randomly selected into groups of

four endowed with five E$. First, subjects reported how much they would like

to contribute to the common pot if selected to go first. One in the group was

then selected at random. The members of the group were informed of this

person’s selection (but not identity) and asked how much they would like to

contribute. Therefore, for each subject in each round, two binding choices are

made.

In the first two sessions, the subjects engaged in treatment 1 and played five

rounds of the game. In the next three sessions, subjects again engaged in treat-

ment 1 and played six rounds. In the final four sessions, the experimental

volunteers first played two rounds of treatment 1 and then played four rounds

of treatment 2. In each round of each treatment, new random groupings were

made to mitigate reputation, retaliation, and history-dependent play. Subjects

made their initial decision not knowing who was in their group (not even the

subjects’ identification codes). Subjects were informed of their earnings from

one round before making their selections in the next. The purpose of having

subjects in the last four sessions conduct both treatments is to compare results

in treatment 1 to those obtained by individuals in previous sessions to verify

consistent behavior across the sample.

It was explained that the more experimental dollars they were able to earn

in the rounds, the more real dollars they obtained. Specifically, the total num-

ber of experimental dollars earned by a subject in all rounds of play in a ses-

sion would be aggregated. The total experimental dollars earned, then, would

be converted into real dollars. The subjects were instructed that the amount

they earned would be determined not only by the choices they made but also

were going to be affected by the choices of others. They were also informed

that in previous, typical sessions (based on a pilot study) subjects earned on

average over $20, but the amounts ranged between $10 and $40. The payouts

provided a $10 ‘show-up’ payment. A scale was adopted where if the maxi-

mum payoff was achieved in each round $50 could be earned, down to a mini-

mum of just the show-up fee. Payments were rounded to $5 increments (or

rather, a step-scale was developed). The average monetary payment received

by a subject in the experiment was $22.50.

With this design, the marginal per capita return (MPCR) each subject

receives is 0.75 E$ for every 1 E$ contributed. Thus, subjects must forgo pri-

vate earnings to create a net positive benefit for the group. While an MPCR

of 0.75 is common (Carpenter, 2007; Reuben and Riedl, 2013; Nosenzo et al.,

2015), so too are lower ones such as 0.4. Utilizing a larger MPCR provides

additional incentive to contribute, which presumably moves more subjects

away from the lower bound, corner solution of zero contributions. With

LEADERSHIP AND PUBLIC GOODS 77

interior points being selected, introducing a leader–follower framework allows me to investigate the extensive margin of giving.

The procedure used in each session was the following. After providing writ-

ten, informed consent, the rules of the game were presented. PowerPoint slides

and printed instructions were given. Individuals were also given the opportu-

nity to ask questions. In treatment 1, individuals filled out a short form asking

how much they would like to contribute. The paper forms were completed

and random groupings were done at the podium in front of the subjects. The

selections were scored and earnings were posted before conducting the second

round. In treatment 2 individuals again filled out a short form choosing how

much they would like to contribute if selected to be first. Again, random

groupings were done in front of the subjects. On the spreadsheet projecting at

the front of the room, each subject could see the first-mover’s contribution. 6

Then, after observing the leader’s selection, each subject again filled out a

form writing down how much they would like to contribute. Outcomes were

then provided on the spreadsheet before playing the next round.

The value to providing feedback is that subjects can form immediate and

accurate expectations regarding others’ behavior and the consequences of their

decisions. Random repairing eliminates reputation-based play. Having contri-

butions be public, but anonymous, replicates the realistic conditions that pub-

lic goods giving typically incorporates (e.g., charitable giving).

In sessions with a number of subjects not divisible by four, responses were

selected at random to complete a four-person group and score it. Thus, for

example, in a session with 15 subjects, 12 are put into three groupings. Hence,

this leaves three remainders. One of the 12 would be selected at random to

provide the fourth contribution to the remaining three subjects. 7

At the end of each experimental session, one round of the Dictator Game

was played. In it, subjects were informed they were to be paired with one

other person, one of the two would be endowed with five E$ and could

choose how much they want to give to the other, endowed with nothing. As

in the baseline Public Goods Game, the amount given would be tripled. The

Dictator Game is commonly used as a control for altruistic giving motivations

(Cox, 2004; List, 2007).

With regard to the background information solicited, common demographic

control variables were collected. They include gender, nationality, state of resi-

dence, major, and year in school. Also, given that the experiments occurred in

November of 2012 and April 2013, each subject was asked whether or not s/

he had voted in the recent election. Previous work has shown that this corre-

lates with ‘other-regarding’ behavior (McCannon, 2014). Table 1 provides

descriptive statistics of the sample.

The subject pool is dominated by USA citizens from the state of New

York. There are more males than females, which is a result of oversampling

6 The leader’s contribution was posted next to the subject’s confidential ID to observe their

leader’s giving level. 7 Variation in the number of participants in each session is an artifact of students signing

up for a session, but not showing up for it!

78 BRYAN C. McCANNON

from business classes. The sample contains both underclassmen and upper-

classmen. In the Dictator Game, the average subject donated 30.8% of his/her

endowment, which is in line with previous work (List, 2007).

IV RESULTS

First, Table 2 provides a summary of the choices made during the nine ses-

sions.

Recall, sessions 1–5 engaged in treatment 1 and sessions 6–9 participated in treatment 2. Two rounds of treatment 1 were conducted in sessions 6–9 to compare behavior across the cohorts. Thus, the two samples behave similarly.

The contributions in treatment 1 are in line with previous findings (Led-

yard, 1995). In the standard, simultaneous-move Public Goods Game, subjects

contribute 50% of their endowment to the common fund. This is not consis-

tent with selfish, wealth-maximizing behavior. In fact, in <20% of the obser- vations was complete free riding witnessed.

With regard to the Leadership Game, striking results arise. If randomly

selected to be the leader, subjects were willing to contribute, on average, 50%

more than in the simultaneous-move game. A substantially higher proportion

of the sample was willing to give their entire endowment and many fewer were

willing to free ride. Interestingly, while still slightly higher than treatment 1

giving, when instead selected to be the follower in the sequential contribution

game, subjects were willing to contribute much less (28% less) than if the lea-

der in the game. This suggests that individuals are motivated by others’ beliefs

about their behaviors, or rather, they have strong psychological preferences.

This is confirmed when considering, in Table 2, how followers responded to

leaders’ actual level of giving. Followers consistently, on average, lagged

behind the leader with regard to the size of the contribution, but did adopt a

form of a matching strategy where when the leader gave more, the followers

tended to give more as well, but when the leader engaged in free riding so too

Table 1

Descriptive statistics

Variable Description Mean

Dictator amount given in the Dictator game 1.542

Vote = 1 if subject voted in November 2012 election 0.519 Male = 1 if subject is male 0.646 Business = 1 if subject is a business major 0.655 Freshman = 1 if subject is a freshman 0.392 Sophomore = 1 if subject is a sophomore 0.142 Junior = 1 if subject is a junior 0.196 Senior = 1 if subject is a senior 0.263 MBA = 1 if subject is an MBA student 0.007 USA = 1 if subject is a USA citizen 0.946 NY = 1 if subject is from New York state 0.726

LEADERSHIP AND PUBLIC GOODS 79

did the followers. Importantly, the follower’s contributions are made after

observing the leader’s giving and are, thus, not hypothetical decisions. This

result is consistent with the theoretical model of players motivated by other-

regarding preferences. 8

The experimental procedure implemented collected each subject’s intended

contribution if selected to be a leader, regardless of whether s/he was actually

selected to be the leader. In addition, what the leader contributed to the group

was made public, but not his/her identity, and each subject provided an actual

contribution given the known amount offered by the leader. Thus, for each

subject in each round, both the offer if selected to be the leader and if selected

to be the follower is provided. Thus, the final section of Table 2 provides

information on the switching behavior of each subject. When an individual

had offered to contribute five if selected to be the leader, how much that per-

son did contribute once the leader’s donation is shown. Consistently, those

who had initially been willing to give all of their endowment pulled back their

offer. As an extreme case, if an individual had been willing to give all five, but

the selected leader only gives one dollar, then no one followed through with

Table 2

Public goods contributions

Mean % = 5 % = 0 %match

Treatment 1

Contributions (sessions 1–5) 2.50 18.3% 20.9% Contributions (sessions 6–9) 2.51 21.6% 18.9% Treatment 2

If selected to be the leader 3.75 46.3% 10.3%

If selected to be a follower 2.71 27.7% 19.9%

If the leader gave 5, followers gave 3.77 63.1% 11.5% 63.1%

If the leader gave 4, followers gave 3.21 7.7% 9.6% 50.0%

If the leader gave 3, followers gave 2.14 2.4% 19.0% 52.4%

If the leader gave 2, followers gave 1.93 10.7% 17.9% 35.7%

If the leader gave 1, followers gave 1.50 6.3% 31.3% 31.3%

If the leader gave 0, followers gave 1.82 17.9% 35.7% 35.7%

If selected to be a follower, but had offered 5 if chosen to be leader

& leader gave 5 3.69 60.5% 14.8%

& leader gave 4 3.07 14.3% 14.3%

& leader gave 3 1.88 6.3% 31.3%

& leader gave 2 1.67 11.1% 33.3%

& leader gave 1 1.00 0.0% 60.0%

& leader gave 0 2.07 14.3% 28.6%

8 As discussed in footnote 5, the theoretical framework and experimental design is unable

to distinguish social preferences, where the follower cares about his contribution differing from the leader’s, and psychological preferences, where the follower cares about his assess- ment of others’ beliefs regarding deviations from the leader’s contribution. Thus, the broader ‘other-regarding preferences’ descriptor is used here. The sequential nature of the game allows for a differentiation between the theories in the leader’s preferences.

80 BRYAN C. McCANNON

their initial offer and 60% did not make a contribution. Even when an indivi-

dual would have been willing to give all five, and the leader actually gives five,

almost 40% of the subjects reduce their actual contribution. In the full data-

set, subjects reduced downward their offer in over 52.8% of the observations.

Thus, strong evidence is provided that a significant component of an indivi-

dual’s desire to contribute to public goods is not strictly social preferences in

that they just care about the well-being of others, but rather care about how

other’s perceive their actions and prefer to respond similarly to a leader.

Regarding wealth, the leaders, who tend to make greater contributions than

the followers, average a payoff of 9.78 E$, whereas the followers are able to

generate 11.04 E$.

To formalize these outcomes, econometric estimations are conducted with

the dependent variable being the actual amount given to the public good.

First, in the full dataset (treatments 1 and 2 combined), the contrast between

the simultaneous and sequential game can be analyzed. The indicator variable

Leadership is equal to one if the observation came from sequential experi-

ments (treatment 2). In these sessions, the dummy variable Selected equals

one if the subject in that round of the session was the one selected to be the

leader. Background characteristics and round fixed effects (to account for

potential deterioration in play) are included as controls. Table 3 presents the

results with standard errors clustered by round of play to account for the pos-

sibility of less variation in play by a subject across rounds than across subjects

within a round.

Observations that occur in the Leadership Game are associated with more

giving. Thus, the net effect is less free riding with sequential contributions.

Also, when the individual is selected to be the leader, a further increase in

contributions arises (with a marginal effect of approximately 81 E₵). Thus, leadership is important for public good provision. Even, though, when a lea-

der shirks the followers also free ride, the motivation to ‘look good’ to the

other subjects acts to increase the total amount of giving. The estimated effect

is that leaders give 1.16 E$ more (=0.805 + 0.357) and followers give 36 E₵ more (7.2% of their endowment) than in a standard Public Goods Game.

Many of the control variables are statistically significant. Even though an

individual’s choices were anonymous, gender and age are determinants of con-

tributions (Sell, 1987). As to be expected, altruistic giving, measured by Dicta-

tor Game sharing, is correlated with public goods giving.

The results in Table 3 are robust. First, standard errors clustered by round

of play are presented. If either heteroskedasticity-robust standard errors or

standard errors clustered by session are calculated, then the coefficients of the

main variables of interest remain highly significant. Furthermore, if either the

background control variables are dropped or session fixed effects are added,

the significance of the results persist. Similarly, if the sixth round of treatment

1, which was played only in Sessions 3–5 is excluded (so that all sessions of treatment 1 play five rounds) or if the two rounds of treatment 1 in Sessions

6–9 are dropped (which were used as a comparison to check the consistency of the two cohorts), then the main results continue to hold. Finally, the

LEADERSHIP AND PUBLIC GOODS 81

dependent variable takes discrete values between 0 and 5. The results pre-

sented treat it as an unbounded, continuous variable by using OLS to estimate

the relationship. If, alternatively, an Ordered Logit model or a Poisson Count

Data model is estimated, the sign and statistical significance of the main vari-

ables of interest remain.

To understand better behavior in the sequential game, play in only rounds

of the Leadership Game can be considered separately. Again, the dependent

variable is the actual contribution to the public good. Along with background

characteristics controls and round and session fixed effects, the publicly

known contribution of the leader, Leader, is included as a regressor. Again,

clustered standard errors are reported.

Thus, for each additional experimental dollar contributed by the leader an

additional 43 E₵ is given by the individual (8.5% of their endowment), while when selected to be the leader the marginal impact on the actual contribution

is estimated to be 87 E₵.9

Model 2 allows for a nonlinear effect of the leader’s contribution and,

again, confirms the previous findings. The positive impact of the leader’s giv-

ing is important when the leader is more generous.

Again, the results presented in Table 4 are quite robust. The same main

results persist if alternative standard errors are calculated (heteroskedastic-

robust or clustered by session). Also, the main results hold when a Poisson

Count Data model or an Ordered Logit is estimated.

The previous results confirm that leaders and followers behave differently,

even when the payoff structure remains unchanged. From the theoretical

Table 3

Full dataset results (Dependent variable = Contribution, N = 822)

OLS Ordered Logit Poisson

Selected 0.805*** 0.827*** 0.255***

(0.186) (0.261) (0.017)

Leadership 0.357** 0.397** 0.133**

(0.175) (0.161) (0.059)

Dictator 0.104*** 0.113** 0.037**

(0.030) (0.048) (0.016)

Round Controls YES YES YES

Background Controls YES YES YES

Adj R 2

0.05 0.01

AIC 3290.1 2879.1 3316.6

Background controls include gender, business major, year in school indicator variables, USA citizen, and voting. The standard errors reported in parentheses are clustered by round of play. ***1%, **5%, *10%.

9 A potentially interesting dimension to consider is gender discrepancies in play. If being a

female is interacted with either Selected or both Selected and Leader, the interaction terms are statistically insignificant (p = 0.91 in the former and p = 0.95 and 0.67 in the latter). Thus, gender differences in play do not arise in the treatment considered.

82 BRYAN C. McCANNON

model, behavior is consistent with players having psychological preferences,

rather than selfish or social preferences. What one wants to know, though, is

what it is about the preferences of individuals in the experimental sessions

that is driving the result. Could the results be explained by individuals simply

placing a premium on being a leader and, thus, raising their contributions in

isolation from the actions of others? Alternatively, are there psychological pre-

ferences at work where leaders lead based on their assessments of others’

expectations, and followers’ behavior is driven by their expectations about

others’, especially the leader’s, beliefs about them? In the former a constant

premium would be placed on leadership giving. In the latter, contributions

and, specifically, the reversion in choices from being a leader to being a fol-

lower, should depend on the behavior of the first-mover.

To address this question, the variable Reversion is calculated. It is the dif-

ference between the amount an individual in a round is willing to contribute

if selected to be the leader and the amount the same subject in the same round

is willing to give if, instead, chosen to be the follower (now knowing what the

leader actually contributes). A positive value for Reversion, then, denotes a

subject who reduces his/her contribution when following, rather than leading,

within the group. 10

In the subject pool, 50.7% of the subjects reduce their

contribution when selected to be the follower and 31.0% maintain the same

level of giving (i.e., Reversion = 0). The new dependent variable is measured

in both level and as a percentage of the initial, leadership offer. 11

As Leader’s

Contribution was shown to have a nonlinear effect on actual contributions of

followers, the models reported in Table 5 also allow for such a relationship.

Again, session and round fixed effects, background controls, and clustered

standard errors are included and reported.

The results in Table 5 provide strong evidence that the behavior of indivi-

duals is driven by the choices of others. Thus, the results are consistent with

individuals who respond to their beliefs of others’ expectations of their play

when both the leader and the follower. Psychological preferences are an

important driver of public goods contributions.

Again, the results in Table 5 are robust. The significance and sign of Lea-

der’s Contribution and Leader’s Contributionr 2 hold if heteroskedastic-robust

10 With the method employed, followers provide a response after observing the leader’s

contribution (but not his/her identity). As this is a binding decision, the reversion calculates the difference between what a subject actually offers if the leader and the amount the subject does contribute when the follower, rather than hypothetical decisions.

11 The percentage calculation obviously has difficulties when the initial offer is zero. A cod-

ing strategy is needed to deal with the zero offers. If the leader and follower offers are both zero, then the percentage change is recorded as a zero, as there was no any change in the offering. As the minimum value for nonzero leadership contributions is �4 (initially offer 1 and increase it to 5), then a move from offering 0 to 5 (the largest possible change) is recorded as �5. Consequently, a change from 0 to 4 is recorded as �4.8, 0 to 3 as �4.6, 0 to 2 as �4.4, and 0 to 1 as �4.2. The 0.2 steps are, then, a linear interpolation of the values. As the results in the second column (levels) of Table 5 mimic those of the first column (rates), this coding strategy does not seem to have an impact on the results. Alternatively, if the dataset is restricted to only those observations where Reversion > 0, the same qualitative results arise.

LEADERSHIP AND PUBLIC GOODS 83

standard errors or standard errors clustered by session are calculated. Further-

more, suppose, instead, an indicator variable equal to one if and only if

Reversion > 0 is considered. The main results of Table 5 continue to hold if either a binary logit or probit is estimated. Additionally, as Arbak and Ville-

val (2013), an ordered variable is created equaling 0 if the following offer is

greater, 1 if equal, and 2 if less than the offer if a leader is considered. The

results hold when ordered logit and ordered probit are estimated. Finally, fol-

lowing G€achter et al. (2012), if leadership offer is used as the dependent vari-

able and following offer is a control, actual leader giving has a separate,

statistically significant effect showing that it is not only just the characteristics

of the subject, as illustrated by G€achter et al. (2012), but also the response to

behaviors of others that determines contributions.

Table 4

Leadership game results (Dependent variable = Contribution, N = 292)

OLS Ordered Logit Poisson OLS

Selected 0.868*** 0.916*** 0.267*** 0.872***

(0.288) (0.220) (0.057) (0.303)

Leader’s Contribution 0.427*** 0.551*** 0.175*** �0.065 (0.030) (0.103) (0.034) (0.059)

Leader’s Contribution 2

0.091**

(0.044)

Dictator 0.019** 0.036 0.004 0.026

(0.050) (0.082) (0.022) (0.047)

Round Controls YES YES YES YES

Background Controls YES YES YES YES

Adj R 2

0.26 0.06 0.27

AIC 1118.2 926.5 1142.5 1115.7

Table 5

Reversion in offers (OLS, N = 292)

Reversion (total) Reversion (%)

Leader’s Contribution 0.590*** (0.195) 0.381*** (0.071)

Leader’s Contribution 2 �0.130*** (0.048) �0.071*** (0.004)

Dictator �0.023** (0.051) 0.020 (0.034) Round Controls YES YES

Background Controls YES YES

Adj R 2

0.26 0.06 0.27

AIC 1118.2 926.5 1142.5 1115.7

84 BRYAN C. McCANNON

Figure 1 graphically illustrates the estimated reversion (y-axis) for differing

levels of leader contributions (x-axis). Mean values for the background con-

trols are used.

Thus, when the leader makes a generous contribution, subjects pull back

their offers, but only modestly. Additionally, if the leader free rides, subjects

also do not reduce their offers much. This is consistent with individuals ‘pick-

ing up the slack’ for the leader. It is also consistent, from the theoretical

model, to individuals having social preferences where the second-movers

increase their contributions to reach a desirable level of group output. For

intermediate contributions, though, reversion is large.

V ROBUSTNESS

As with any experimental investigation, numerous decisions need to be made

regarding design. Consequently, while each can be justified, one can always be

concerned that the results are specific to the particular design feature and are

not reproducible in similar environments. Also, the extension of the design

allows for additional analysis which provides for the opportunity to enrich the

analysis. Therefore, I present robustness results utilizing a complementary

design.

Design

The laboratory sessions summarized previously allow individuals to see, pro-

jected on a screen at the front of the room, both the leadership offer by the

first-mover in their group, but also the leadership offers of individuals in other

groups. Thus, the design allows for a more rich social interaction and reason-

able belief formation. Furthermore, subjects observed the choices of the

0.5

1.5

2.5

0 1 2 3 4 5

Re ve

rs io

Leader's contribution

Figure 1. Relationship between leader’s contribution and reversion.

LEADERSHIP AND PUBLIC GOODS 85

individuals in all groups before making their choices in the next round. This

allows for a social learning effect to arise. Public goods contributions are a

social activity, but one would like to identify whether the findings are driven

by the high levels of information, along with potential adaptation and learn-

ing across the rounds of play, or by the direct structure of the game.

Therefore, I conducted three additional sessions (N = 47). Subjects engaged in both the leader–follower version and the standard Public Goods Game as described previously. An important difference is that in these additional ses-

sions, the decision-making was done privately on an online program. Subjects

did not know who they were grouped with and did not receive feedback on

what others in the group chose to give. Importantly, subjects were uninformed

of the activities in other groups. If the same results arise in this low informa-

tion treatment, then I can be assured that it is the structure of the giving that

is driving the results.

Relatedly, the main results utilize a design where individuals play multiple

rounds of the game. Along with the opportunity to learn from others, this

design allows for individuals to adapt and learn time. Therefore, in the three

additional sessions, the subjects played one-shot games. There was not any

repetition.

Additionally, as discussed in Section III, public goods games can be para-

meterized by the marginal per capita return. The MPCR is the proportion of

each unit contributed that a subject receives from the public good. To present

a dilemma for the subject, putting into conflict the well-being of others and

personal wealth, MPCR are fractions between 0 and 1. In the design pre-

sented, the MPCR is 0.75. While this is rather common in the literature

(Nosenzo et al., 2015), the selection of an MPCR of 0.4 is also popular. To

illustrate, it is common to give each subject 20 ‘tokens’ to begin the game and

to be put into groups of five. Each token contributed to the group fund dou-

bles (multiplier M = 2), but then is evenly shared (N = 5). The MPCR of M/ N = 2/5 = 0.4 results in a rather low private return to contributions. There- fore, the three additional sessions use this design, employing both a multiplier

of 0.4 and 0.75 in two different one-shot games both with and without a lea-

der.

Finally, in a laboratory experiment employing strategic decision-making,

choices can be made (and recorded) using either the direct-response method

or the strategy method (Brandts and Charness, 2011). In the direct-response

method, only the actual choices made are recorded. In the strategy method,

subjects provide up front the full contingency plan. The design developed uti-

lizes both. Each individual reveals how much they would contribute if selected

to be a leader. Choices are not restricted to the actual amount those selected

to be the leader gave. The value of using the strategy method for recording

the leadership choice is to be able to evaluate the revision in behavior. For

followers, though, the direct-response method is used where subjects only

responded to the actual contribution of the person selected randomly to be

the leader. In the three additional sessions considered here, the strategy

method is employed for the follower’s decision. That is, each subject

86 BRYAN C. McCANNON

completed a table providing the amount they would contribute if the person

selected to go first gives X, where X ranges from 20 down to 0. Subjects knew

that these choices are binding. The value of introducing the strategy method

to the follower’s play is that the full contingency plan can be evaluated.

Therefore, if the main results persist when relaxing the elicitation method,

the online vs. offline nature of the social interaction, and an important para-

meter of the payoff function, then one can be assured that the results are

robust.

Results

The primary results presented previously constitute three main findings: (i)

subjects give more when in a leadership position than in the simultaneous-

move game, (ii) followers engage in a quasi-matching strategy where they

shirk relative to the leader but increase their contribution when the leader

does so, and (iii) subjects when not selected to be the leader revise downwards

their willingness to contribute and that the magnitude of this reversion is

related to the leader’s giving level. Minor findings also arose: (iv) average con-

tributions are higher in the sequential Public Goods Game and (v) when the

leader’s contribution is extremely low followers respond by contributing

slightly more and revising downward their contributions less. I now turn to

the results of the three additional sessions to establish if these findings con-

tinue to arise in the new setting.

The first observation is that individuals contribute more as the leader than

they give in the simultaneous-move Public Good Game, (i). In Low Treat-

ment, 67.4% of subjects give more if selected to be the leader than they are

willing to contribute in the simultaneous-move game. For these individuals,

the average increase is 1.35 tokens. In High Treatment, 76.1% offered to con-

tribute more than without sequencing of contributions, with an average

increase of 2.54 tokens. Hence, a supermajority of the subjects offer more

and, when they do, the increase is substantial. Thus, overall, (i) arises in the

new treatments.

Additionally, the final observation was the overall contributions are higher

in the sequential Public Goods Game. In Low Treatment, along with higher

leader giving, the average contribution of a follower is higher than the simul-

taneous-move Public Goods Game level of giving when the leader contributes

18, 19, or 20 tokens. In High Treatment, the average follower’s level of giving

is greater than the simultaneous-move game for leader contributions of 15

through 20 tokens. Therefore, sequential giving generates a higher overall level

of contributions. 12

This is evidence in support of (v).

Regarding the second result, (ii), the strategy method employed provides

me with the amount a follower is willing to contribute for every level of giving

12 As full elicitations are made without real-time scoring in the laboratory, the actual aver-

age level of contribution made in the sequential Public Goods Game is determined by the randomization result. Therefore, the sample averages of the actual decisions made is pre- sented.

LEADERSHIP AND PUBLIC GOODS 87

by the leader. Therefore, the average contribution across the subjects can be

derived for each level of giving. Figure 2 depicts the follower’s decision-mak-

ing.

As one can see, the same pattern emerges. The x-axis is the leader’s con-

tribution, while the y-axis is the average amount contributed by a follower.

The dashed line would then be the behavior of a follower who matched the

leader’s giving. The gray line presents giving in High Treatment and the

black line depicts giving in Low Treatment. As the leader’s contribution

reduces, so too does the follower’s. So long as the leader does not free ride

too much, the giving is less. To verify this relationship, the choices of the

subjects are pooled (N = 965). The amount the leader contributes is regressed on the offer to contribute when in the follower role. Rather, the

following model is estimated,

Leader’s Contribution ¼ di þ a � Follower’s Contribution þ u ð6Þ With a subject fixed effect included, the parameter a captures the relation-

ship depicted in Figure 2. For Low Treatment, a = 0.861 (SE = 0.033; p < 0.001), and in High Treatment a = 0.842 (SE = 0.033; p < 0.001). There- fore, individual-level behavior mimics the group-averaged behavior depicted

in Figure 2. The relationship between the amount the leader gives and the

amount an individual is willing to follow with is upward sloping and flat (at

least flatter than the 45-degree line). Therefore, the results are consistent with

(ii).

Furthermore, for very low giving levels by the leader, the followers main-

tain higher contributions consistent with (v). In a pooled OLS model, the esti-

mated crossing point where subjects are willing to give more than the leader is

11.5 for Low Treatment and 10.9 for High Treatment.

Interestingly, behavior with a high MPCR level and a low MPCR level are

nearly identical. The correlation coefficients between each of these 21 choices

0 2 4 6 8

10 12 14 16 18 20

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Fo llo

w er

's c

on tr

ib ut

io n

Leader's contribution

MPCR = 0.4 MPCR = 0.75

Figure 2. Average public goods contributions by followers. [Colour figure can be viewed at

wileyonlinelibrary.com]

88 BRYAN C. McCANNON

are very large (exceeding 0.8 in 18 of the choices never falling below 0.67).

Again, a fixed effects model estimated on the pooled data can be used to iden-

tify the relationship between behavior across the subjects,

Follower’s ContributionðHighÞ ¼ di þ b � Follower’s ContributionðLowÞ þ u ð7Þ The estimated value is b = 0.859 (SE = 0.018; p < 0.001). Therefore, deci-

sion-making is nearly identical, when in the role of the follower, in the two

treatments. This suggests that it is not the private gain that is influencing the

second-movers’ choices, but rather the giving level of the subject selected to

contribute first. These new findings are a consequence of the alternative elici-

tation method employed.

Turning to (iii), the difference between the amount the subject is willing to

contribute if selected to be the leader and how much s/he is willing to con-

tribute if instead selected to be the follower, Revision, is considered in the

pooled dataset. As the main results suggest there is a nonlinear relationship

(Table 5), squared values of the explanatory variables are considered. Table 6

presents the results.

In both treatments, the amount a subject revises his/her contribution

depends on both the amount the leader contributes and the amount s/he was

willing to contribute if selected to be the leader. Holding fixed what the sub-

ject was willing to do, when the person in the leadership position contributes

more, the revision reduces. As revision is the difference between the choice in

the first-mover and the second-mover position, a reduction in revision indi-

cates the contribution when the follower is closer to the amount the subject

was willing to contribute if the leaders. Rather, followers respond positively

to the contributions of the leaders.

Table 6

Revisions in offers

Low treatment High treatment

Leader’s Contribution �0.361*** 0.931*** �0.400*** 1.115*** (0.096) (0.067) (0.110) (0.077)

Leader’s Contribution 2 �0.006 �0.058*** �0.005 �0.067***

(0.005) (0.004) (0.005) (0.005)

If Leader 0.199** 0.409***

(0.096) (0.122)

If Leader 2

0.014*** 0.009*

(0.004) (0.005)

Constant 2.593*** 1.301

(0.634) (0.841)

Adj R 2

0.435 0.172 0.425 0.189

AIC 5593.5 6066.4 5543.4 6014.6

N 945 945 903 903

The standard errors are presented in parentheses. N = 945 in the first two columns and N = 902 in the second two. ***1%, **5%, *10%.

LEADERSHIP AND PUBLIC GOODS 89

Second, holding fixed the leader’s choice, the more an individual offers if

the first-mover, the greater is the revision. There is an important nonlinear

effect, where this effect is even stronger when the subject makes a more gener-

ous opening offer. This suggests that subjects respond adversely to the leader’s

limited contributions when they themselves were willing to contribute gener-

ously in the leadership position. Again, this is a new result made possible by

the alternative design eliciting the full contingency plan of follower’s behavior.

Finally, the second column for each treatment presents results consistent

with (iii). There is an inverted U-shape relationship between the amount a

subject revises his/her contribution and the leader’s contribution. This indi-

cates that the findings in Section IV are robust to the experimental methods,

but also suggests that the nonlinear relationship previously found can be dis-

aggregated into a response to the leader and the leader’s differentiation from

what the follower would have done (rather, a violation in the social norm).

Consequently, the five results identified previously extend to a new environ-

ment where the information availability, MPCR, and elicitation method are

different. Thus, the results are robust. Furthermore, new results, namely the

invariance of followers’ contributions to the MPCR and the differentiation of

leader’s actual giving and the leader’s break from the subject’s anticipation,

enrich our understanding of how psychological preferences are expressed in

public goods environments.

VI CONCLUSION

The work explores the role of preferences, formulated in psychological game

theory, in public goods contributions contrasting them with selfish preferences

and the more commonly studied social preferences. A simple theoretical model

is developed that predicts those with social preferences behave differently than

those with psychological preferences in sequential public goods games. The

experimental results are consistent with psychological preferences, where indi-

viduals respond to their assessment of others’ expectations regarding their

behavior.

Sequential giving is able to encourage greater contributions out of the lea-

der, as part of his/her utility is driven by the expectations regarding other’s

beliefs on his/her contribution. Motivated to ‘look good’ the leader increases

giving to a public good. This result can be thought of as complementing the

findings of Luccassen (2012) who showed that voluntary contributions to pub-

lic goods can be explained in part by one’s personality. Hence, being in a

leadership position complements the individual’s personality traits. Followers,

also driven in part by their psychological preferences, ‘avoid the guilt’ of free

riding and adopt a quasi-matching strategy. As the leader gives more (than in

the standard, simultaneous-move public goods game), the followers contribute

more increasing aggregate welfare. Thus, the results highlight not only the

value in proper institutional design but also highlight the importance of

understanding the nature of individual utility functions.

90 BRYAN C. McCANNON

The work builds on the well-established research on the optimal provision

of public goods dating back to the seminal contribution of Groves and Led-

yard (1977). The incorporation of behavioral economics insights into the free

rider problem, as initiated by Andreoni (1988, 1989, 1995), provides the

opportunity to explore a wider range of analysis into the problem. An appli-

cation of the important distinction between social preferences and psychologi-

cal preferences can be found in Hallsworth et al. (2014). They report on a

field experiment providing information on the effect of unpaid taxes on soci-

ety. Letters were sent to UK taxpayers encouraging them to pay. Notes point-

ing to being in the minority of nonpayers generated greater payment rates

than notes highlighting how others benefit from the public goods made possi-

ble with tax revenues. The former focuses on one’s psychological preferences,

while the latter relies only on the social preference of the taxpayer.

The framework explored is one of sequential giving. Alternatively, one

could investigate further the role of punishment (Fehr and G€achter, 1992,

2000), conditional, contingent giving (Fischbacher et al., 2001), communica-

tion (Isaac and Walker, 2000), repeated play (Fischbacher and G€achter,

2010), culture (G€achter and Hermann, 2009), and framing (Sonnemans et al.,

1998) in public goods donations. Psychological preferences can be expected to

affect behavior in these complementary environments. Thus, the role of psy-

chological preferences in optimal institutional design of public goods contribu-

tions, along with any other number of economic environments that rely on

nonselfish behavior, warrants further consideration.

The theoretical framework promotes psychological preferences as an impor-

tant ‘other-regarding’ preference that is expressed in sequential public goods

environments. These preferences are contrasted to social preferences focusing

utility on disparities in outcomes. One can be concerned that other competing

theories of behavior can also explain the results provided. Hence, future

research should consider additional economic environments in which to test

psychological game theory as a dominant factor explaining behavior.

Finally, it is worth investigating potential covariates with psychological

preferences. For example, Grossman et al. (2015) provide evidence that there

are important gender differences in leadership behavior when gender is

revealed. If the genders differ in the degree to which they are socialized to put

more weight on others’ assessments of their behavior, or if a norm that

‘women believe that others’ believe that women should sacrifice more (or con-

tribute less) to a group’s welfare’ is pervasive, for example, gender and the sal-

ience of gender as found by Grossman et al. (2015) would explain variation in

giving levels. This is just one potential example of using psychological game

theory to justify and identify differences in behavior and, hence, is a path for

future research.

APPENDIX

Proof of Proposition 1: Using the utility function in (1) it follows that duWi /

dai = � ¼ < 0. In each subgame, then, a2* = a3* = a4* = 0.

LEADERSHIP AND PUBLIC GOODS 91

Consequently, a1* = 0, which constitutes the unique Subgame Perfect Nash equilibrium. Obviously, then, the outcome remains if the simultaneous-move

game is considered. h

Lemma 1: In the Nash equilibrium of the simultaneous-move public goods

game with social preferences, if S(x) = x 2 or if S(x) = max {0, �al � ai} and

ri < ¼ for all i, then each player contributes ai* = 0. If S(x) = max {0, �al � ai} and ri ≥ ¼ for all i, then any a* 2 [0, 5] where a1* = a2* = a3* = a4* = a* is a Nash equilibrium.

Proof of Lemma 1: Using the payoff function in (2) and the assumption that S

(x) = x 2 , duSi /dai = � ¼ + 2ri(�al � ai). First, if �al > ai, then an increase to

ai + e results in a payoff of 5 � ai � e + 0.75(a1 + a2 + a3 + a4 + e) � ri(�al � ai � e)2. In a symmetric Nash equilibrium (a1* = a2* = a3* = a4* = a*) this simplifies to 5 � a* � e + 0.75(3a* + a* + e) � r (a* � a* � e)2 = 5 + 2a* � 0.25e � rie2, which is decreasing in e. Second, if �al ≤ ai, then, again, the payoff is decreasing in e. Thus, the best response is ai = 0, so that the symmetric Nash equilibrium is a1* = a2* = a3* = a4* = 0.

Alternatively, using the payoff function in (2) and the assumption that

S(x) = max {0, �al � ai}, duSi /dai = � ¼ + r when �al > ai and duSi / dai = � ¼ when �al ≤ ai. First, if ri ≤ ¼ for all i, then duSi /dai ≤ 0 and the best response is ai = 0. Hence, if ri ≤ ¼ for all i, then the symmetric Nash equilibria entail a1* = a2* = a3* = a4* = 0. Second, if ri > ¼ and �al > ai, then a deviation to ai closer to �al is profitable. Thus, consider �al = ai � a+. If this is adopted by all players, this generates a payoff to each of

5 � a+ + 0.75(4a+) = 5 + 2a+. A deviation by a player to a+ � e results in 5 � a+ + e + 0.75(4a+ � e) = 5 + 2a+ � 0.25e, which is less for any value of e. Hence, if ri > ¼ for all i, then the symmetric Nash equilibria entail a1* = a2* = a3* = a4* = a* and any value of a* 2 [0, 5] can be rational- ized. h

Proof of Proposition 2: Suppose S(x) = x 2 . First, consider the decision-mak-

ing of players 2, 3, and 4 in the subgames. Let â1 denote the known contribu-

tion made by player 1. Consider player j, j 2 {2, 3, 4} where uSj = 5 � aj + 0.75(a1 + a2 + a3 + a4) � rj (�aj � aj)2 where �aj = (ai + ak + al)/3, j 6¼ i, k, l. The best response for player j is aj = �aj � 1/8rj. In the sym- metric equilibrium of the subgame, af* (the contribution of a follower) is 1/ 8rf � â1. Now, using backward induction, consider the contribution by player 1. The payoff is uS1 = 5 � a1 + 0.75(3/8r1 � 2a1) � r1(1/8r1 � 2a1)2 so that duS1/da1 = � 2.25 � 4ra1 < 0. Consequently, the unique Subgame Perfect Nash equilibrium has a1* = 0 and af* = 1/8rf. It follows from Lemma 1 that the leader contributes (weakly) less, while the followers contribute more,

than in the simultaneous-move game.

Alternatively, suppose S(x) = max {0, �al � ai}. Again, following back- wards induction, consider the decision-making of 2, 3, and 4 in the subgames.

92 BRYAN C. McCANNON

It follows that duSj /daj = � ¼ + rj when �al > ai and duSj /daj = � ¼ when �al ≤ ai. If rj > ¼, then du

S j /daj > 0 when �aj > aj and deviation of aj closer to

�aj is profitable. Thus, again, consider �al = ai � a+ for i 2 {2, 3, 4}. If this is adopted by players 2, 3, and 4, then this generates a payoff to each of

5 � a+ + 0.75(4a+ � h) = 5 + 2a+ � 0.75h where h is any difference between the leader’s contribution, a

+ � h, and the followers. A deviation by a player in {2, 3, 4} to a+ � e results in 5 � a+ + e + 0.75 (4a

+ � e � h) = 5 + 2a+ � 0.25e � 0.75h, which is less for any value of e. Thus, a

+ = �al must hold in equilibrium. As a result, �al = 3a

+ � h, so that combining the two, 2a

+ = h, or rather, a+ = h/2 (assuming h > 0; otherwise

a +

= 0). Now, consider the initial decision by the leader � player 1. His choice of h results in a utility of 5 � a+ + h + 0.75(3h/2 + a+ � h) � r1 max {0, h/2 � a+ + h}, which simplifies to 5 � 0.25a+ + 7h/4 � r1 max {0, 3h/ 2 � a+}. If 3h/2 < a+, then duS1/dh > 0 and h = a+ = 0. If 3h/2 ≥ a+, then duS1/da1 = 7/4 � 3r1/2, which is less than zero when r1 > 7/6. In this case h = 2a+/3 so that in equilibrium a1 = a

+ /3. Otherwise, if r1 ≤ 7/6, then

duS1/dh > 0 and h = a + . As a result, if r1 ≤ 7/6, then

a1* = a2* = a3* = a4* = 0, while if r1 > 7/6, then any a* 2 [0, 5] for play- ers 2, 3, and 4 with a1* = a*/3 is a Subgame Perfect Nash Equilibrium. h

Corollary 2: The Subgame Perfect Nash equilibrium of the sequential public

goods game with social preferences (and selfish preferences as well) generates

less than the socially optimal (aggregate utility maximizing) level of public

goods giving, which is a1 = a2 = a3 = a4 = 5.

Proof of Corollary 2: Let W = ΣuSi . Here W = 20 � (a1 + a2 + a3 + a4) + 3(0.75) (a1 + a2 + a3 + a4) � ri Σ S(xi). It follows immediately that for any sum (a1 + a2 + a3 + a4), W is improved if a1 = a2 = a3 = a4 � a. Hence, W = 20 + 0.75(a1 + a2 + a3 + a4) = 20 + 3a, which is maximized at a = 5. Thus, the socially optimal level of contribution for each player, whether ri = 0 or ri > 0, is not reached in the Subgame Perfect Nash equilibrium of Proposi- tion 1 (selfish preferences) and Proposition 2 (social preferences). h

ACKNOWLEDGEMENTS

I thank Greg DeAngelo, Claudio Detotto, Martin Dufwenberg, Phil Gross-

man, Joseph Guse, Mana Komai, Todd Palmer, and Mark Wilson for helpful

discussions and participants at Saint Bonaventure University, Business

Research Consortium, and Southern Economic Association. I also appreciate

the financial support from the Koch Foundation and the Clare College at

Saint Bonaventure University, especially David DiMattio, for encouraging

and financially supporting the research. I also thank Kim McCannon for

being a lovely assistant during the experimental sessions.

LEADERSHIP AND PUBLIC GOODS 93

References

ANDREONI, J. (1988). Why free ride? Strategies and learning in public goods experiments.

Journal of Public Economics, 37, 3, pp. 291–304. ANDREONI, J. (1989). Giving with impure altruism: applications to charity and Ricardian

equivalence. Journal of Political Economy, 97, 6, pp. 1447–58. ANDREONI, J. (1995). Warm-glow versus cold-prickle: the effects of positive and negative fram-

ing on cooperation in experiments. Quarterly Journal of Economics, 110, 1, pp. 1–21. ANDREONI, J. (2006). Leadership giving in charitable fund-raising. Journal of Public Eco-

nomics, 8, 1, pp. 1–22. ANDREONI, J. and PETRIE, R. (2004). Public goods experiments without confidentiality: a

glimpse into fund-raising. Journal of Public Economics, 88, 7, pp. 1605–23. ANDREONI, J., BROWN, P. M. and VESTERLUND, L. (2002). What makes and allocation fair?

Some experimental evidence. Games and Economic Behavior, 40, 1, pp. 1–24. ARBAK, E. and VILLEVAL, M.-C. (2013). Voluntary leadership: motivation and influence.

Social Choice and Welfare, 40, 3, pp. 635–62. BATTIGALLI, P. and DUFWENBERG, M. (2007). Guilt in games. American Economic Review, 97,

2, pp. 170–6. BATTIGALLI, P. and DUFWENBERG, M. (2009). Dynamic psychological game theory. Journal of

Economic Theory, 144, 1, pp. 1–35. BEEKMAN, G., BULTE, E. and NILLESEN, E. (2014). Corruption investments, and contributions

to public goods: experimental evidence from rural Liberia. Journal of Public Economics,

115, 1, pp. 37–47. BICCHIERI, C. (2006). The Grammar of Society: The Nature and Dynamics of Social Norms.

Cambridge: Cambridge University Press.

BRANDTS, J. and CHARNESS, G. (2011). The strategy versus direct-response method: a first sur-

vey of experimental comparisons. Experimental Economics, 14, 3, pp. 375–98. CARPENTER, J. P. (2007). Punishing free-riders: how group size affects mutual monitoring and

the provision of public goods. Games and Economic Behavior, 60, 1, pp. 31–51. CHAUDHURI, A. (2011). Sustaining cooperation in laboratory public goods experiments: a

selective survey of the literature. Experimental Economics, 14, 1, pp. 47–83. COLMAN, A. M. (2003). Cooperation, psychological game theory, and limitations of rational-

ity in social interactions. Behavioral and Brain Sciences, 26, 1, pp. 139–98. COX, J. C. (2004). How to identify trust and reciprocity. Games and Economic Behavior, 46, 2,

pp. 260–81. DUFWENBERG, M., G€ACHTER, S. and HENNIG-SCHMIDT, H. (2011). The framing of games and

the psychology of play. Games and Economic Behavior, 73, 2, pp. 459–78. ERTAC, S. and GURDAL, M. Y. (2012). Deciding to decide: gender, leadership and risk-taking

in groups. Journal of Economic Behavior & Organization, 83, 1, pp. 24–30. ESWARAH, M. and KOTWAL, A. (2004). A theory of gender differences in parental altruism.

Canadian Journal of Economics, 37, 4, pp. 918–50. FEHR, E. and FISCHBACHER, U. (2002). Why social preferences matter – the impact of non-selfish

motives on competition, cooperation and incentives. Economic Journal, 112, 478, pp. C1–33. FEHR, E. and G€ACHTER, S. (1992). Altruistic punishment in humans. Nature, 415, 1, pp. 137–40. FEHR, E. and G€ACHTER, S. (2000). Cooperation and punishment in public goods experiments.

Nature, 90, 4, pp. 980–94. FEHR, E. and SCHMIDT, K. M. (1999). A theory of fairness, competition, and cooperation.

Quarterly Journal of Economics, 114, 3, pp. 817–68. FISCHBACHER, U. and G€ACHTER, S. (2010). Social preference, beliefs, and the dynamics of free

riding in public goods experiments. American Economic Review, 100, 1, pp. 541–56. FISCHBACHER, U., G€ACHTER, S. and FEHR, E. (2001). Are people conditionally cooperative?

Evidence from a public goods experiment. Economic Letters, 71, 3, pp. 397–404. G€ACHTER, S. and HERMANN, B. (2009). The limits of self-governance when cooperators get

punished: experimental evidence from urban and rural Russia. European Economic

Review, 55, 2, pp. 193–210.

94 BRYAN C. McCANNON

G€ACHTER, S., NOSENZO, D., RENNER, E. and SEFTON, M. (2010). Sequential vs simultaneous

contributions to public goods: experimental evidence. Journal of Public Economics, 94,

7, pp. 515–22. G€ACHTER, S., NOSENZO, D., RENNER, E. and SEFTON, M. (2012). Who makes a good leader?

Cooperativeness, optimism, and leading-by-example. Economic Inquiry, 50, 4, pp. 953–67. GEANAKOPLOS, J., PEARCE, D. and STACCHETTI, E. (1989). Psychological games and sequential

rationality. Games and Economic Behavior, 1, 1, pp. 60–79. GROSSMAN, P. J., KOMAI, M. and JENSEN, J. E. (2015). Leadership and gender in groups: an

experiment. Canadian Journal of Economics, 48, 1, pp. 368–88. GROVES, T. and LEDYARD, J. O. (1977). Optimal allocation of public goods: a solution to the

“free rider” problem. Econometrica, 45, 4, pp. 783–810. G€UTH, W., VITTORIA LEVATI, M., SUTTER, M. and van der HEIJDEN, E. (2007). Leading by

example with and without exclusion power in voluntary contribution experiments. Jour-

nal of Public Economics, 91, 5–6, pp. 1023–42. HAIGNER, S. D. and WAKOLBINGER, F. (2010). To lead or not to lead: endogenous sequencing

in public goods games. Economics Letters, 108, 1, pp. 93–5. HALLSWORTH, M., LIST, J. A., METCALFE, R.D. and VLAEV, I. (2014). The behavioralist as tax

collector: using natural field experiments to enhance tax compliance. National Bureau

of Economic Research, No. w20007.

HERMALIN, B. E. (1998). Toward an economic theory of leadership: leading by example.

American Economic Review, 88, 5, pp. 1188–206. HUANG, P. H. and WU, H.-M. (1994). More order without more law: a theory of social

norms and organizational cultures. Journal of Law, Economics & Organization, 10, 2,

pp. 390–406. ISAAC, R. M. and WALKER, J. M. (2000). Communication and free-riding behavior: a volun-

tary contribution mechanism. Economic Inquiry, 26, 4, pp. 585–608. ISAAC, R. M., MCCUE, K. F. and PLOTT, C. R. (1985). Public goods provision in an experi-

mental environment. Journal of Public Economics, 26, 1, pp. 54–74. JACK, B. K. and RECALDE, M. P. (2015). Leadership and voluntary provision of public goods:

field evidence from bolivia. Journal of Public Economics, 122, 1, pp. 80–93. KOMAI, M. and STEGEMAN, M. (2010). Leadership based on asymmetric information. RAND

Journal of Economics, 41, 1, pp. 35–63. KOMAI, M., STEGEMAN, M. and HERMALIN, B. E. (2007). Leadership and information. Ameri-

can Economic Review, 97, 3, pp. 944–7. KOMAI, M., GROSSMAN, P. J. and DETERS, T. (2011). Leadership and information in a single-

shot collective action game: an experimental study. Managerial and Decision Economics,

32, 1, pp. 119–34. LAZEAR, E. P. (2012). Leadership: a personnel economics approach. Labour Economics, 19, 1,

pp. 92–101. LEDYARD, J. O. (1995). Public goods: a survey of experimental research. In A. E. Roth and

J. H. Kagel (eds), Handbook of Experimental Economics. Princeton: Princeton Univer-

sity Press, pp. 111–81. LEVATI, M. V., SUTTER, M. and van der HEIJDEN, E. (2007). Leading by example in public

goods experiment with heterogeneity and incomplete information. Journal of Conflict

Resolution, 51, 5, pp. 793–818. LIST, J. A. (2007). On the interpretation of giving in dictator games. Journal of Political

Economy, 115, 5, pp. 482–93. LUCCASSEN III, R. A. (2012). Individual differences in contributions and crowding-out of a

public good. Scottish Journal of Political Economy, 59, 4, pp. 419–41. MARWELL, G. and AMES, R. E. (1981). Economists free ride, does anybody else?. Journal of

Public Economics, 15, 3, pp. 295–310. MCCANNON, B. C. (2014). Do economists play well with others? Experimental evidence on

the relationship between economics education and pro-social behavior. American Econo-

mist, 59, 1, pp. 1–7.

LEADERSHIP AND PUBLIC GOODS 95

MCCANNON, B. C. (2017). Social norms and cooperation. International Journal of Economic

Theory. Forthcoming.

NOSENZO, D., QUERCIA, S. and SEFTON, M. (2015). Cooperation in small groups: the effect of

group size. Experimental Economics, 18, 1, pp. 4–14. POTTERS, J., SEFTON, M. and VESTERLUND, L. (2005). After you – endogenous sequencing in

voluntary contribution games. Journal of Public Economics, 89, 8, pp. 1399–419. POTTERS, J., SEFTON, M. and VESTERLUND, L. (2007). Leading-by-example and signaling in vol-

untary contribution games: an experimental study. Economic Theory, 33, 1, pp. 169–82. RAUCH, J. E. (2001). Leadership selection, internal promotion, and bureaucratic corruption

in less developed polities. Canadian Journal of Economics, 34, 1, pp. 240–58. REUBEN, E. and RIEDL, A. (2013). Enforcement of contribution norms in public goods games

with heterogeneous populations. Games and Economic Behavior, 77, 1, pp. 122–37. RIVAS, M. F. and SUTTER, M. (2011). The benefits to voluntary leadership in experimental

public goods games. Economics Letters, 112, 2, pp. 176–8. ROMANO, R. and YILDRIM, H. (2001). Why charities announce donations: a positive perspec-

tive. Journal of Public Economics, 81, 3, pp. 423–47. SAHIN, S. G., ECKEL, C. and KOMAI, M. (2015). An experimental study of leadership institu-

tions in collective action games. Journal of the Economic Science Association, 1, 1, pp.

100–113. SELL, J. (1987). Gender, strategies, and contributions to public goods. Social Psychology

Quarterly, 60, 3, pp. 252–65. SONNEMANS, J., SCHRAM, A. and OFFERMAN, T. (1998). Public goods provision and public

bad prevention: the effect of framing. Journal of Economic Behavior and Organization,

34, 1, pp. 143–61. VAN VUGT, M. and DE CREMER, D. (1999). Leadership in social dilemmas: the effects of

group identification on collective actions to provide public goods. Journal of Personality

and Social Psychology, 76, 4, pp. 587–99. VARIAN, H. R. (1994). Sequential contributions to public goods. Journal of Public Economics,

53, 1, pp. 165–86. VESTERLUND, L. (2003). The informational value of sequential fundraising. Journal of Public

Economics, 87, 3–4, pp. 627–57. WEBER, R., CAMERER, C., ROTTENSTREICH, Y. and KNEZ, M. (2001). The illusion of leader-

ship: misattribution of cause in coordination games. Organizational Science, 12, 5, pp.

582–98.

SUPPORTING INFORMATION

Additional Supporting Information may be found in the online version of this

article:

Table S1. Merged dataset descriptive statistics.

Table S2. Leadership game results (OLS, dependent variable = Contribution,

N = 292). Table S3. Leadership game results (N = 292). Table S4. Leadership game results (N = 292). Table S5. Merged dataset results (OLS, dependent variable = Contribution,

N = 822). Table S6. Merged dataset results (dependent variable = Contribution;

N = 822). Table S7. Merged dataset results (dependent variable = Contribution;

N = 786).

Date of receipt of final manuscript: 25 August 2017

96 BRYAN C. McCANNON

Copyright of Scottish Journal of Political Economy is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.