ISSN (Print) - 0012-9976 | ISSN (Online) - 2349-8846

A+| A| A-

Public Goods Experiments and Social Preferences

Part 4 of the survey offers an overview of public goods experiments, an area where experimentation has been particularly fruitful.


4 Public Goods Experiments and Social Preferences

Part 4 of the survey offers an overview of public goods experiments, an area where experimentation has been particularly fruitful.

Economic & Political Weekly

august 27, 2011 vol xlvi no 35

4.1 Self-interest and Competition

In the animal world, the only motivation behind any action is self-preservation. Self-preservation, again, with a few exceptions, can be broken down to actions designed to find food, actions designed to prevent one from becoming the food of others, and actions related to procreation. The human species, on the other hand, is much more evolved and therefore more complex. A consequent, and logical, question that has pervaded enquiry is what constitutes the fundamental motivation that determines human decisions and action.

Various disciplines have approached this question in various ways. Economics, particularly of the neoclassical variety, has always accorded priority to self-interest. Self-interest is not only seen to be the fundamental motivation behind economic decisions and actions, it is even perceived to be perfectly compatible with the given competitive economic environment. In other words, the competitive market economy is mostly able to deliver efficient outcomes primarily because human beings act in a self-interested manner. So much has been the primacy attached to self-interest in economics that rationality in human beings, which is considered to be a desirable faculty, is defined in terms of it. A person is considered to be acting in a rational manner only if he or she undertakes such decisions and actions that, given the objective situation the person is in and the information set at his or her disposal, seek to maximally serve his or her self-interest.

The philosophical understanding about the consequences of this view of rationality is that the self-interestedness that lies at the root of this rationality will always trigger competitive behaviour in human beings, and such overall competition will allow market mechanisms to deliver efficient outcomes. Subsequent theoretical research, however, has shown that under some circumstances the opposite may be true. There may be situations that require humans to cooperate, rather than compete, but since selfinterested rationality drives a human being away from cooperation, overall defection is the final outcome. It is easily demonstrated that in this final outcome all decision-makers are worse off compared to a situation where they may have been had they cooperated. But given the human motivational structure discussed above, human beings are unable to move away from overall defection to overall cooperation. So we have a situation where a superior outcome is in sight but given the way the self-interested human mind works, such solutions remain elusive.

Experimental economics literature has made significant contributions in this regard. Many experimental games designed to test the tendency to cooperate against the possibility of defection conclude that even when theory predicts defection by self- interested individuals, experimental results show that subjects do cooperate to a significant extent. Attempts to explain such decisions often rest on some presumption of altruistic (pure or impure) or other-regarding behaviour. It is now well recognised that all human beings possess some motivational tendency to act in an altruistic or other-regarding fashion and such motivations can also affect economic decisions. A case in point is the issue of provision of public goods. Since such goods are non-excludable and non-rival in consumption, any attempt to elicit users’ willingness to pay for such goods will inevitably fail. The free rider problem will ensure that consumers do not reveal what value they place on a good being so provided because each individual will believe that others will pay and the good will be created; and once the good is created, non-excludability will ensure he or she cannot be prevented from consuming it and will face no resistance from other users because non-rivalness will ensure that the consumption of this good by one has no adverse effect on others.

Experimental work involving a game using a voluntary contribution mechanism (VCM) has routinely thrown up quite the opposite result. It is observed that self-interested individuals do contribute substantially to the provision of a public good. Usually some form of altruism is invoked to ensure the existence of a cooperative equilibrium, which in turn is justified by the presence of trust, belief, common history and such other-regarding behaviour. One can alternatively pose the question whether self-interested individuals, who fully understand how cooperation in a game form can yield an outcome better than overall defection, may “learn” from past mistakes and choose an overall strategy of cooperation that really reflects the need of individuals to reach a better outcome than the one they would get if they simply defected. Further, by this these subjects reach a higher level of rationality that rejects narrow self-interestedness and aspires towards a higher or wider form of rationality, where experienced individuals consciously choose cooperation as a superior strategy for long-term gain.

The above point can only be understood by conducting these games under various treatment conditions. We will discuss specific forms of a public goods game and show how, under various treatments, overall cooperation does occur when the level of cooperation improves and when it does not. Through this we hope to offer the reader an understanding about human motivations.

4.2 Public Goods Experiments

A pure public good is characterised by non-rivalness and nonexcludability in consumption. Non-rivalness in consumption implies that a number of consumers can consume the goods or services at the same time, while non-excludability means that no one can be prevented from using the benefits of the goods or services once it has been produced. In other words, non-excludability implies that no one can be prevented from enjoying the benefits of the goods or services even if he or she has not contributed to the provision of the goods. This is what makes the provision of public goods susceptible to freeriding.

The incentive to freeride in the provision of public goods has led to the belief that decentralised or voluntary mechanisms cannot be relied on to provide public goods. That is, it will either lead to low provision or no provision at all. However, there is much empirical evidence to show that people do not always freeride and community

62 members can come together to obtain privately provided public goods. Andreoni (2008) writes that in 2005 charitable giving in the US was over $269 billion or 1.9% of total personal income. Of this, 70% came from individuals. There is also ample evidence from India – Wade (1987), Baviskar (1994), D’Silva and Nagnath (2002), Moose (2003) and others have shown how community members have donated private resources to obtain public goods or have evolved stable rules to maintain and regenerate both public goods and common property resources.

4.3 Voluntary Contribution Mechanism

The idea here is to design a game to see if people voluntarily contribute to ensure the provision of public goods. First, to avoid any effect of differential income on contributions, all subjects are provided with an equal initial endowment, from which they make a contribution. Second, in a laboratory setting, it is difficult to actually create a public good out of whatever contributions are obtained. To simulate the effect of private benefits from a public good, the amount contributed is multiplied by a factor greater than one (signifying that the benefit from a public good is often many times the original capital investment) and divided equally among the subjects.

Thus, the simplest type of public goods experiment involves a VCM. In a VCM there are n subjects and each is given an initial endowment of H. From this endowment, they are requested to contribute any amount, F”F” H̓, towards a common fund. The sum S thus collected is multiplied by a factor P! and equally divided among the Q subjects. The return to any individual L is given by

Ri = e – ci + mȈMLSj + ci) / n L

Where, H±FL= amount spent on private investment, and Ȉji Sj =

amount contributed by all participants other than i.

Now from equation (i) we get ˜RL˜ci =+ m/n for n m.

This implies that the marginal return to individual i from an extra unit of contribution to the common fund is negative. Therefore the dominant strategy for any individual would be to not contribute to the common fund. By symmetry, this should be true for all individuals and the value of total contributions should be 0 or very low. Here, it may be interesting to note the similarity of the structure of pay-off and outcome of the VCM with those of the prisoner’s dilemma game. This is why the VCM is sometimes considered as an n-person prisoner’s dilemma game. Thus, the conventional game theoretic prediction for a public goods experiment with the VCM was low contribution or no contribution by the subjects. But experimental results have come up with evidence in contradiction to the theoretical predictions. (See Ledyard 1995 for an excellent survey of some important public goods experiments.)

Marwell and Ames (1979) tried to experimentally examine the degree of freeriding in public goods games and determine the effects of varying experimental conditions like group size, distribution of resources, divisibility of public goods, etc, on contribution to a public good. Two hundred and fifty-six high school students were divided into 64 groups of four students each. Eight such groups were subjected to eight different treatments. The subjects could invest their endowments either in a private good with a fixed return or in a public good whose return depended on

august 27, 2011 vol xlvi no 35

the group contribution. The subjects were provided with a payoff for a large group of 80 with unequal benefits and resources. The peculiar feature of the pay-off structure was that at a moderately high level of total group contribution (7,999 tokens, at the rate of 1 cent per token) the marginal return was about 6 cents, while for all other levels the return was less than 1 cent. This implied multiple Nash equlibria, one at which freeriding was the optimal strategy and the other where all contributed partially. The important finding of the experiment was that nearly 57% of the available resources were invested in public goods. This was clear evidence of the lack of complete freeriding.

In another experiment by Isaac, McCue and Plott (1985), the earning of each period was determined by the level of the public goods provided in that period minus the individual’s contribution in that period. The level of the public goods was estimated by dividing the total contribution by the unit cost of the project. An important feature of the experimental design was the introduction of repetition. The result of the experiment showed that in the initial rounds the subjects contributed about 50% of their endowment but by the fifth round contributions dropped to 9% of their endowment. Kim and Walker (1984) used a similar design and came across similar results. In the first period of their experiment, the average contribution was 41% of the group pay-off and it declined to 11% by the third period.

Isaac, Walker and Thomas (1984) set up a computerised lab experiment to estimate the effect of different factors on freeriding. The different factors examined were group size, repetition, marginal pay-off and the experience of subjects. They chose two group sizes, n=4 and n=10. The number of periods in each treatment was 10 and the different marginal pay-offs were 1.2, 3 and 7.5. The average contribution across all the treatments was 42% while the average of all first period contributions was 51%. These were very close to results observed in previous experiments. The other results of the experiment were that contribution increased with group size for the low marginal pay-off of 3 but no perceptible effect of group size was observed for the high marginal pay-off of 7.5. Similar effects were observed for repetition except that repetition with the lower marginal pay-off decreased contribution. The effect of increasing the marginal pay-off from 3 to 7.5 was quite strong as it increased contribution in all the cases. Finally, it was observed that inexperienced subjects contributed more than experienced subjects. The last observation seemed to suggest that there was some form of learning during the course of the experiment.

From the selected experiments that we have surveyed in this section we find that there is little or no evidence of complete freeriding exhibited by experimental subjects. In one-shot experiments or in the initial rounds of repeated experiments, average contributions have ranged between 40% and 60%. However, it is also to be observed that the relatively high contributions in the initial rounds of repeated experiments start declining, approaching the Nash equilibrium.

4.4 What Drives Giving in Public Goods Experiments?

Different explanations have been extended to understand positive contributions in public good games even when the dominant strategy is “no contribution”. Some of the important explanations

Economic Political Weekly

august 27, 2011 vol xlvi no 35

are pure altruism, impure altruism, inequality aversion, reciprocity and conditional cooperation. According to Nagel (1970), altruism is “not abject self-sacrifice, but merely a willingness to act in the consideration of interests of other persons, without the need of ulterior motives”. The point to be noted in this definition is that there is an element of other-regarding behaviour without an ulterior motive. This implies that an increase in others’ well-being directly increases one’s own utility. Such other-regarding behaviour may be termed pure altruism. Most of us give alms to the poor (who may even be strangers) and surely we do not expect anything in return. However, the act of pure altruism also implies that people’s desire to contribute for public goods will decline when the contribution by others is high. Theoretically it has been suggested that in large groups altruists will not contribute (Sugden 1982; Andreoni 1988). But empirical evidence shows that the crowding out of private contributions by government donations is only partial (Andreoni 1993). Moreover, in reality, we observe significant individual donations to big charities such as the Red Cross in the US or the Prime Minister’s Relief Fund in India.

Then what could be the reason behind people contributing to big charities? Here it may be interesting to note that Andreoni and Miller (2002) showed that the concept of pure altruism was quite consistent with the theory of self-interest. Self-interest has been interpreted as an individual choice conforming to an underlying preference ordering that is complete, reflexive and transitive. Thus, a utility function can describe individual behaviour. However, this does not tell us anything about the variables to be included in the utility function. So, while the inconsistency between theoretical predictions and experimental outcomes can be explained by the hypothesis of self-interest, such self-interest does not follow only maximising monetary pay-offs. Becker (1974) observed that apparent charitable behaviour might be driven by the desire to avoid social scorn or receive social acclaim. Andreoni (1988, 1990) extended the theory to impure altruism or “warm glow”. He termed social and psychological factors such as social pressure, self-respect, self-esteem, and so on as “warm glow”. This implied that besides enjoying an increase in the well-being of others, the act of giving provided internal pleasure to the donor. The theory of impure altruism fares better in explaining the partial crowding out of private donations (donations by individuals or private agencies) by public donations (by the state) as individuals do not conceive private donations to be a perfect substitute for public donations.

Models of inequality aversion (Fehr and Schmidt 1999; Charness and Rabin 2002) try to show that subjects behave pro-socially towards those who are worse off but punish those who are better off. Charness and Rabin, in a dictator game, asked subjects to choose between an equal outcome and an unequal but socially efficient outcome. The result showed that most subjects chose the unequal, socially efficient outcome. However, the experiment failed to clearly differentiate whether the subjects were more interested in social efficiency or inequality aversion. People are sometimes also concerned with the intention of the actions of others (Rabin 1993; Falk and Fischbacher 2006). According to Rabin, people’s own choice depended on not only the choice made by others but also on the choice that they could have made but did not. This implied that people were generous towards those who have been kind to them but were willing to punish those who have been unkind to them. This has been termed reciprocity in the literature. While altruism refers to unconditional kindness, reciprocity refers to non-selfish behaviour determined by the earlier actions of others. That is, when an individual decides to contribute to the common fund of a public goods experiment, he or she will first judge the behaviour of others to be generous and adjust his or her action accordingly. As we shall see in a later section, the implication of reciprocity in public goods experiments is that subjects exhibit higher cooperation when the option of punishing free riders is incorporated in an experiment.

The concept of conditionality or conditional cooperation is close to that of reciprocity. This means that there is a positive correlation between the contribution made by a subject and his or her perception of others’ contributions. Fischbacher et al (2001) found in a laboratory public good game that 50% of the subjects contributed more when others also did so. Frey and Meier (2004) conducted a field experiment with students of Zurich University to examine conditional cooperation. The subjects were to contribute to two social funds after being randomly informed either that many (64%) others had contributed or that few (46%) others had contributed. The information regarding high contribution by others was based on the behaviour of the students in the earlier semester, while the low contribution rate was the average contribution over the last 10 years. The result showed that the subjects who were informed about high contributions gave more compared to those who were informed that contributions had been low.

4.5 Contributions: Kindness or Confusion?

We have seen in the public goods experiments discussed so far that participants contribute significantly higher than that predicted by theory. Andreoni (1995) questioned whether this giving was out of kindness and altruism or sheer confusion and error. He designed an experiment in which he tried to control for kindness, which would leave confusion as the only explanation of cooperative behaviour. The experiment had three treatments. The first was a regular VCM. The second was also a VCM but the subjects were paid according to their ranks based on the experimental earnings. Since the highest rank would be obtained by the highest free rider, there was no incentive to cooperate in this treatment. To make the results of regular and rank treatments comparable, he conceived a third treatment that paid the subjects their experimental earnings but their ranks were also communicated to them. The information about their ranks would remind them about the incentive structure of the game. The difference between the results of the second and third treatments would provide an estimate of the number of students who understood the incentives but contributed out of kindness. The result of the second experiment would provide an estimate of the number of students who were confused. The outcome of the experiment revealed a significant presence of kindness and confusion. About 75% of the subjects were found to be cooperative, of whom half were found to be confused.

Houser and Kurzban (2002) addressed the same issue as Andreoni. However, in their experiment the incentives to kindness were controlled by making the human subjects participate

64 in a VCM with computers. The human subjects were informed that the contributions made by the computers were independent of the contributions made by them. Consequently, the human players had no opportunity to benefit either themselves or other subjects through cooperation. The result of the experiment showed that 50% of the contributions were out of confusion and that confusion was responsible for higher contributions in the earlier rounds. Ferraro and Vossler (2010) focused on the causes and consequences of confusion. They carried out an elaborate project consisting of four experiments (including two repeated ones) combined with micro-econometric models. Following Houser and Kurzban (2002), they also included one treatment in each experiment where the human subjects had to play with virtual players (computers). The important findings of the experiments were that confused subjects used experimental parameters and the behaviour of other subjects (especially in repeated experiments) as cues. Confusion often mimicked other-regarding behaviour and thus distorted estimates of altruism. They also found that modifying standard instructions could substantially reduce confusion, leading to changes in the distribution of contributions and splitting of endowments between private and group exchanges.

4.6 Factors Affecting Contributions in Public Good Experiments

There are many factors such as group size, marginal pay-offs, repetition, communication among subjects, options of punishing free riders, and so on that may have a direct effect on the level of contributions in a public good experiment. Among these factors, here we concentrate on two important ones, communication and punishment. Yamagishi (1986, 1988), Ostrom, Walker and Gardner (1992), Fehr and Gaechter (2000, 2002) and others have conducted public goods experiments which showed that subjects exhibited a higher level of cooperation when they were given the option of punishing free riders even at some cost to themselves. In the experiment by Ostrom et al, the subjects were provided with the opportunity of punishing free riders by paying a fee. The fee would be borne by individuals but the benefits of higher contributions would go to the group as a whole. Thus, the Nash equilibrium would be no punishment but the outcome of the experiment showed that there were significant levels of punishment.

In the experiment by Fehr and Gaechter (2000), the subjects were assigned to two groups, strangers and partners. In the stranger group, the subjects were selected at random after each round. There were four subjects in each group and the groups played 10 rounds of a VCM with punishment and 10 rounds without punishment. In the rounds without punishment the groups played the normal VCM but in the rounds with punishment there was an additional stage in which they could punish low contributors in their group. This experiment was motivated by Hirshleifer and Rasmusen (1989) in which they showed that given the opportunity of ostracising non-cooperators, rational egoists could maintain cooperation for (T-1) periods in a T period prisoner’s dilemma game. Here punishment was in the form of a reduction of 10% of the points of the punished subject. But punishment was not costless and this implied that if the subjects were only interested in their own pay-offs, there would be no punishment.

august 27, 2011 vol xlvi no 35

The outcomes of the no-punishment VCM reflected the results of earlier experiments with significant contributions in the earlier rounds followed by a decay or decline in later rounds. The opportunity to punish had a significant impact on contributions as the low contributors faced a credible threat of punishment. Even though punishment was costly, contributions improved among both groups and no perceptible decline was observed over rounds. In the stranger treatment with punishment, no stable behavioural stability or regularity was observed in individual contributions. But the partner group contributions tended to converge to the socially optimal level. The experiment explained why social norms were upheld in societies where there was the opportunity to punish those who did not obey norms, even at a cost to the punisher.

Another important factor that has a positive impact on contributions is the opportunity for communication among subjects. Subjects are able to form a perception about what others are doing through communication. But non-binding communication being non-enforceable, it has been theoretically considered “cheap talk” (Farrel and Rabin 1996). But there is experimental evidence to show that pre-play communication, even when non-binding, increases contributions to the common fund. Isaac and Walker (1988) found that the mean contribution was higher and the rate of decline slower with pre-play communication. It was also observed that when communication was preceded by periods of non-communication, contributions rose very slowly and when communication preceded rounds of non-communication, the rate of decline in contributions was very slow. Brosig et al (2003) suggested that the increase in contributions owing to face-to-face communication might be caused by cues from facial expressions, body language, tone of voice and removal of anonymity.

Bochet et al (2006) conducted an experiment with free-form communication, removing anonymity and eliminating vocal and visual cues. The experiment also tested the effectiveness of combining opportunities for communication and punishment. The different treatments designed were a baseline treatment, a treatment with monetary punishment (called the reduction treatment), face-to-face communication treatment, face-to-face communication treatment with reduction, chat room treatment, chat room treatment with reduction, numerical communication treatment and numerical communication treatment with reduction. The important findings of the experiment were that face-to-face communication was most effective, that without a mechanism to ensure commitments, numerical communication lacked effectiveness, and that monetary punishment without communication improved contributions but not as much as in the case of verbal communication.

4.7 Strategies and Learning in Public Goods Experiments

Why does the average contribution decline with repetition? Two answers are generally proposed to explain this phenomenon. The first hypothesis relates to a process of learning by the subjects about the incentive structure of the experiment and the corresponding choice. As we saw in Isaac, Walker and Thomas (1984), inexperienced subjects contribute more than experienced subjects. Palfrey and Prisbrey (1997) explained that normally the subjects would make low contributions unless there was some

Economic Political Weekly

august 27, 2011 vol xlvi no 35

element of altruism or error or both on their part. The second hypothesis is that subjects behave strategically, contributing higher amounts in the initial rounds with the expectation that others will also contribute high. But when they find that others have not made a high contribution, they reduce their contributions. This has been termed reciprocity or conditional cooperation.

Here we discuss a much cited paper by Andreoni (1988) in which he designed a lab experiment to isolate the effects of learning from the experimental outcome so that the difference could be attributed to strategic play. Repetition permitted the subjects to develop multi-period strategies that allowed for some strategic behaviour even when it was common knowledge that “no contribution” was the dominant strategy. That is, repetition provided the subjects with the opportunity to signal future moves to each other and, according to Andreoni, this formed the basis of the strategies hypothesis. The public goods experiment was conducted under two conditions, strangers and fixed partner groups. In the first, 20 subjects were randomly assigned to four groups (five subjects in each group). They were informed that they would play the game for 10 rounds but the group members would change randomly after each period. Thus, there was little possibility of developing strategic play among the subjects because every round in the stranger composition was an end round. In the second condition, 15 other students were assigned to three groups of five fixed partners each. They too played 10 rounds but the subjects had no knowledge who the other partners were. Here it may be noted that the fixed partner composition gave the subjects scope to play strategically. In a fixed partner group, contributions were expected to be higher as a partner could continue to contribute high even after knowing that “no contribution” was the dominant strategy in a one-shot game. To isolate the learning hypothesis, the session was conducted twice; that is, a “restart” was included. This was done with the understanding that if learning was responsible for the decline in contributions, both the strangers and partners would start the second session with low contributions. This session lasted for only three periods.

The results of the experiment showed that strangers contributed more than partners and the difference in their contributions grew in the later rounds. The percentage of free riders was more among partners than strangers in the first session but the partners started at high contributions in the “restart” session while the strangers were unaffected. The outcome of the experiment showed that the strangers, who had no opportunity to play strategically, contributed more. This led to a rejection of the strategies hypothesis. On the other hand, the fact that the subjects contributed high even after an understanding of free rider incentives, falsified the learning hypothesis. Thus, both the hypotheses were contradicted in this experiment. In his concluding section, Andreoni suggested that one needed to consider non-standard theories, looking beyond the characteristics of equilibrium, into how subjects make their decisions. The partners and the strangers might have had different decisionmaking processes with the partners updating their prior beliefs on the likelihood of cooperation at a faster rate than the strangers. The alternative suggested was that giving was consistent with social norms on participation in a social dilemma. The decline might have been a reflection of the groups’ attempt to enforce norms through punishing low contributors.

4.8 Theories and Experiments on Charitable Fundraising

Prosocial behaviour or simply the act of giving to a public cause should depend on the price of giving. This implies that there will be a positive relation between prosocial giving and subsidisation of the cost of giving. In charities, two types of incentives are provided – giving rebates on donations and the incentive-providing agency making matching contributions. The first incentive can take the form of a rebate on income tax and the second can be a matching grant provided by the government. To what extent is individual giving influenced by the price of giving? It is shown in Andreoni (2008) that if t is the tax rate, then the price of giving a dollar is (1-t). If the policy of the government is to have a positive impact on donation by reducing the tax rate, the required elasticity is įg/į(1-t). (1-t)/g -1; that is, the policy is effective if giving, g, is price elastic. It was generally estimated (Clotfelter 1985; Steinberg 1990; Andreoni 2006a) from cross-sectional surveys that income elasticity was below 1 (between 0.4 and 0.8) and price elasticity was below -1 (between -1.1 and -1.3); that is, it satisfied the price elasticity condition. But Randolph (1995), using tax returns from 1979 to 1989, came out with different results. According to his estimate, the long-run price elasticity was -0.51 but the short-run elasticity was -1.55. This implied that givers shifted their giving from years of low marginal tax rates to years with high marginal tax rates.

However, Auten, Sieg and Clotfelter (2002) came up with a result consistent with the earlier view. They used a similar panel of taxpayers but used the technique of having restrictions on the covariance matrices of income and prices by assuming the permanent income hypothesis. Their analysis returned a permanent price elasticity of -1.26. According to Andreoni, the sensitivity of the elasticities to the estimation technique and identification strategy has left the literature unsettled. Another method of changing the relative price of giving is through matching individual contributions. This instrument has also been found to have a positive impact on charitable contributions (Meier 2005) and it has also been found that matching is a more powerful instrument than rebate (Eckel and Grossman 2003, 2005).

Monetary incentives in certain conditions may decrease the intrinsic motivation to contribute for a public cause. Titmuss (1970) explained that paying for blood undermined social value and reduced the willingness of voluntary blood donors. Intrinsic motivation is a well-established concept in psychology. Deci, Koestner and Ryan (1999) found from a meta-analysis of 128 psychological experiments that tangible rewards had a significant negative effect on intrinsic motivation whereas verbal rewards had a positive effect. Even though rewards were able to influence human behaviour, the negative impact was that they sometimes undermined self-regulation and consequently people took less responsibility for motivating themselves.

But do government grants to charities crowd out private donations? Studies by Kingma (1989), Okten and Weisbrod (2000), Hungerman (2005) and others found that the crowding out of private donations was negligible. But Payne (1998) argued that

66 the government officials who approved the grants to a charity were elected by the same people who made private contributions to the charity. Thus, both had an interest in giving to the charity. To take care of this endogeneity, she performed a two-stage least squares analysis and considered government transfers to individuals as government grants to charities. She estimated crowding out to be 50%. However, all the above analyses do not consider the impact of government grants on fundraising efforts. Andreoni and Payne (2003) considered a 14-year panel of charitable organisations and found that there was a significant reduction in fundraising efforts in response to government grants. This may be a possible explanation for the partial crowding out of private contributions.

The focus has recently shifted from the supply-side to the demandside of the charitable fund market; that is, from the donors to the fundraisers. Andreoni (1998) developed a positive theory of capital campaigns by charitable organisations. The implication of the theory was that a publicly announced seed contribution could significantly increase the total amount of giving to a capital campaign. Charitable giving for a threshold public good could have multiple equilibria. In the absence of initial seed money, the outcome could be a Nash equilibrium with zero contributions. But with an initial commitment of seed money decreasing the remaining money to be raised, the zero equilibrium could be eliminated. According to Andreoni, individuals wanted to give to fundraising but would value their gift more if the campaign were successful.

Andreoni (2006b) explored the role of leadership in giving to charitable funds. A single wealthy donor (or a group of donors) could make a substantially large contribution to signal the high quality of the charity to other donors. If the single wealthy donor, the leader, was successful in giving a credible signal to his or her followers and they gave more to the charity, it would be in the benefit of the leader. But this might give rise to a war of attrition over who would be the leader. Andreoni suggested that the person who had the lowest cost-benefit ratio in signalling would be the first mover. Karlan and List (2007) conducted a natural experiment with 50,000 prior donors to test the effectiveness of matching grants on charitable giving by measuring the comparative static effects of different rates of matching. The experiment’s results showed that a matching offer increased both the revenue per solicitation and the probability of individual giving. However, in contradiction to conventional belief, it was observed that higher matching ratios of $3:$1 and $2:$1 had little effect on the lower ratio of $1:$1.

A second instrument at the disposal of the fundraisers of threshold public goods is refunding contributions if the charity is not successful (Bagnoli and Lipman 1989). Bagnoli and McKee (1991) experimentally confirmed the prediction of Bagnoli and Lipman that refund mechanisms produced an efficient provision of public goods. List and Lucking-Reiley (2002) conducted a natural experiment33 to examine the comparative impacts of both seed money and refund mechanisms. Their experiment showed that contributions to a charitable fund were positively related to both the mechanisms, but the magnitude of the refund effect was smaller than the seed money effect. Seed money increased contributions five times while refunds increased contributions only by 21%.

Frey and Meier (2004) found evidence of conditional giving in a fundraising field experiment with students of University of

august 27, 2011 vol xlvi no 35

Zurich as its subjects. Meier (2005) also found that prosocial behaviour was positively related to a matching of people’s donations because matching funds reduced the relative price of giving.

4.9 Voluntary Contributions, Social Networks and Social Capital

Some of the important theories that have been extended to explain prosocial behaviour in public goods experiments are altruism, fairness, reciprocity, conditional cooperation, and so on. All these theories come under the concept of social preferences. Bowles (2005: 96) explains the essence of social preferences – “in choosing to act, individuals commonly take account not only of the consequences of their actions on themselves but for others as well. Moreover they often care not only about consequences but also about the intentions of other actors.” The root of social preferences lies in social capital. This has been defined and viewed from different perspectives by different authors. But the common characteristics of social capital that have been defined by most authors are cooperative norms, trusts and membership of associations or networks within a societal structure. The three authors most commonly referred to in the literature on social capital are Bourdieu (1983), Coleman (1988, 1990) and Putnam (1993, 2000). Similar ideas are also reflected in the writings of North (1990), where he uses the term “informal institutions”. North defined informal institutions as codes of conduct, norms of behaviour and conventions. Most of these unwritten codes of conduct and norms might stem from formal rules but people seldom took cognisance of these rules in their daily interactions. Moreover, North observed that a good institution (both formal and informal) promoted cooperation and reduced transaction costs. These characteristics were also prominently present in social capital.

Bourdieu was interested in explaining how privileged groups maintained their status or position of advantage by establishing and utilising connections with other privileged people. This enabled them to have better access to resources and power. He viewed social capital as a resource associated with group networks and its effective mobilisation depended on the size of the network. Coleman (1988) observed that social capital was created within the family or kinship and community networks. He highlighted the possibility that marginalised communities or members of the working class could also benefit by possessing social capital. He viewed social capital as an integral part of a societal structure and underlined that certain social structures and institutions were better suited to develop trust, reciprocity and cooperation than others. Putnam examined the importance of social capital and the quality of civic life in the development of a democratic society. He referred to social capital as social networks, norms and trust that improved the efficiency of society. Civic virtue worked best when supported by a network of relationships involving social reciprocity.

Dasgupta (2005) stated that trust was the key to cooperating and what scholars had meant by social capital was merely one of the means of creating trust. He further argued that social capital was a private good with both positive and negative externalities. Positive externalities could spread beyond immediate groups but negative externalities, like community cohesion developed on the basis of caste, religion or race, could take place at the cost of

Economic Political Weekly

august 27, 2011 vol xlvi no 35

outsiders. In the latter case, outsiders might be treated with hostility and suspicion. For the economy as a whole, considerable resources might have to be diverted from development purposes to resolve intercommunity conflicts. Thus, there was no need to attach positive values only to social capital. Dasgupta therefore considered social capital to mean interpersonal networks, nothing more.

The economic importance of social capital, as commonly highlighted in the literature, is that it reduces the transaction costs of monitoring contracts, and negotiating and enforcing contracts and state regulations. Owing to repeated interactions among the members of a group or community, trust and cooperative norms are established. This trust, along with norms of reciprocity, helps to foster collective actions. Social capital arising out of frequent and repeated interactions among members also facilitates access to private information about each other, which is not accessible by outsiders. Through peer monitoring and social sanctions, the problem of freeriding can also be avoided.

According to Grief (1994), societal organisation in developing countries is collectivist while it is individualist in the developed western world. The social structure in the former is segregated, in the sense that individuals interact mostly with members of a specific group to which they belong and feel involved in the lives of other members of the group. This endogenous partition of society into small groups makes individual members dependent on groups. This also facilitates in-group communication and economic and social collective punishment. Thus, contract enforcement is achieved through informal social and economic institutions where groups can influence their members to comply with established norms through economic, social and moral sanctions. On the other hand, an individualist society is more integrated and has a vertical social structure. The dependence of an individual on any particular group is weakened and as a result the ability of any group to use economic, social and moral sanctions against a member is weak. This has led to the development of societal organisations based on legal, political and (second party) economic organisation for enforcement and coordination.

Cardenas and Carpenter (2005) reported that in a public goods game in Peru, at least 81% of the participants contributed to the common fund and the group contribution was naturally quite high. The authors referred to a prisoner’s dilemma experiment in which Chinese subjects cooperated twice more frequently than their American counterparts. Similarly, in a public goods experiment it was found that initial contributions were higher in Vietnam than in Thailand. This, according to the authors, might be because of the collectivist culture promoted by the governments of China and Vietnam compared to the individualism promoted in the US and Thailand.

In the Indian context, such an effect of the institutional environment on contributions was observed by Mitra and Gupta (2009). The experiment was a non-computerised, single-period VCM. There were four treatments and the experiment was conducted in three sessions. The subjects in the three sessions were from different sociocultural backgrounds. In the first session, the subjects were 30 first-semester students from four different engineering streams of a technical college in Kolkata. They were mostly from urban Kolkata, exposed to a western system of education. The second session had 30 members from a village community-based organisation (CBO), who had a common history of participating in social programmes and sharing local common pool resources (CPRs), and had resided together for long. The subjects of the third session were from six different CBOs, located in different villages. Even though they had a similar sociocultural background to the subjects of the second session, they had never worked together as a group. The four treatments were designed to find the effect of group size and removal of anonymity, allowing pre-play communication and contributing for a real-life project, on contribution. An important finding of the experiment was that the CBO members in the second group were the highest contributors in all the treatments, with an average contribution rate of above 85%. The difference in contributions between the subjects in the first and third sessions was statistically not significant. Even though the subjects in the third group had a similar sociocultural background, the bond between the members of different CBOs was weak. This might have been the cause of lower contributions in this session. The lower contribution in the case of subjects in the first session could be attributed to their urban background, exposure to a western individualistic culture and that they had very little time to interact with each other in the group. However, it should also be mentioned that in no treatment was the average contribution below 50%.

4.10 Findings So Far

We list some of the important findings from the experiments on public goods surveyed so far. They are,

  • (i) Public goods experiments have shown that the problem of freeriding is not as pervasive as predicted in the theoretical literature. There is little evidence of complete freeriding and subjects were found to contribute between 40% and 60% in one-shot games or in the initial rounds of repeated games.
  • (ii) The cooperative behaviour exhibited by the participants has been explained in terms of social preferences like altruism, inequality aversion and reciprocity. But some experiments have also found confusion on the part of the participants to understand the game structure and incentives to be an important cause of higher contributions.
  • (iii) Contributions start declining or decaying in the later rounds of repeated games. This decay in contribution still remains an unexplained phenomenon. There are basically two hypotheses that have been put forward to explain it. The first is that the subjects take some time to learn the incentive structure of the experiment and it is only after they have played a few rounds that they realise that zero contribution is the dominant strategy. The second states that subjects strategically contribute higher amounts in the initial rounds expecting to motivate others to contribute high. But when they find that the others have not been as generous as expected, they too start contributing low. As we have seen in Section 4.7 above, Andreoni (1988) found both the hypotheses were contradicted in an experiment conducted by him. However, findings by Andreoni (1995), Houser and Kurzban (2002), Ferraro and Vossler (2010) and others that confusion accounted for nearly 50% of the initial contributions lend partial support to the learning hypothesis.

    (iv) Some factors such as an increasing marginal pay-off, the opportunity for communication and the option of punishing free

    68 riders have been identified as having the potential to improve contributions or even reverse the decay. These findings are important as they suggest possible environments that can promote pro-social behaviour in obtaining a public good.

  • (v) There has been considerable research on charitable fundraising. The focus has recently shifted from the donors to the fundraisers. Since giving to charitable funds was found to be dependent on the relative price of giving, fundraisers now try to attract more donations by providing incentives such as rebates and matching grants. Experiments have confirmed the effectiveness of these incentives.
  • (vi) Finally, it has been revealed that the sociocultural backgrounds of the subjects have a strong impact on the outcome of experiments. These are factors such as norms, habits and trust, which the subjects bring with them into the laboratory. It has been observed that subjects coming from collectivist societal structures with a strong sense of trust and bonding within the community contribute more compared to those who come from individualist societies. The finding that non-economic variables such as trust matter for societies is interesting and has spurred a good amount of experimental research. The next section reviews some of these findings.
  • 4.11 Trust and Reciprocity
    4.11.1 Trust and Economic Organisation

    K Arrow (1974) stressed the ubiquity of trust in almost every economic transaction, pointing to a positive association between a higher level of trust and higher rates of investment and economic growth. Trust is seen as an essential social lubricant which facilitates cooperation in society and contributes to the maintenance of social order at both the micro and societal levels (Fukuyama 1995; Putnam 1993, 2000; Luhmann 1988). Trust may solve social dilemmas that occur mainly due to the lack of a mechanism to monitor, to enforce contracts or commitments and to sanction cheaters, and where the role of an external authority may be inefficient or very costly. It indirectly promotes social exchange, helping society to attain Pareto optimal outcomes. Thus, it can be viewed as a facilitator of spontaneous cooperation and informal social exchange.

    Recent studies in experimental economics investigate conditions for the emergence and sustainability of trust and cooperation in the light of various social situations, individual preferences and motives, opportunities to communicate, reciprocate and punish, patterns of interactions (one-shot vs repeated), past experience, culture, incentive structure, and the like. The prisoner’s dilemma game has been considered the foundation for the analysis of trust and cooperation as it appropriately represents the conflicting incentives of the players in any social dilemma. In the standard simultaneous prisoner’s dilemma game, given the lack of communication and coordination between players, individual incentives to defect outweigh the collective benefit that could be obtained from mutual cooperation. To maximise his or her material pay-off, every player ends up defecting, obtaining a socially deficient outcome. Here the inability to trust others to not preferring individual gain over collective gain appears to be the core problem.

    august 27, 2011 vol xlvi no 35

    One of the pioneers of studying the prisoner’s dilemma experimentally, Deutsch (1960) stated that individuals must develop mutual trust if they are to cooperate with one another. He argued that an individual’s orientation to his or her partner was an important determinant of development of trust. Deutsch said that mutual trust positively affected cooperation when individuals were positively oriented to each other’s welfare. He also found that individuals with a cooperative orientation made cooperative choices that resulted in mutual gain, whereas those with a competitive orientation made non-cooperative choices, resulting in mutual loss. These findings established the relationship between trust and cooperation, although the exact nature of this relationship was not very clear. Deutsch, with his prisoner’s dilemma experiments, initially theorised that mutual trust facilitated cooperation, later arguing that mutual trust developed as a result of orientation (cooperative or competitive), bringing circularity to his arguments. He added to this circularity by arguing that a cooperative choice by subjects could be interpreted as a “trusting choice”, so cooperation could be an indicator of trust. In short, trust was seen both as a feature of a relationship as well as a feature of a particular behaviour.

    4.11.2 Measuring Trust

    To get rid of this circularity, attitudinal measures of trust other than behaviour in prisoner’s dilemma settings were required. Rotter (1967) was the pioneer in conducting surveys to measure trust and correlating the data with outcomes of prisoner’s dilemma experiments. The trust scale he used included social objects (such as parents, teachers and physicians) against which subjects were supposed to express their trust. Questions were asked to measure a broader conception of trust or general optimism regarding society and filler items used to disguise the purpose of the scale. This interpersonal trust scale was additive in nature and a higher score showed trust in a great variety of social objects. With the help of this interpersonal trust scale, he tried investigating the extent to which trust was a characteristic of personality and thus would play a role in personal interactions. After extensive work on measurement of trust, validity of scales and the link between the results of the trust scale with other factors viewed as possible predictors of levels of trust (social class, age, Machiavellianism,34 belief in human benevolence) for two decades, he showed an overall positive correlation between (low) high trust and (un)trustworthy behaviour. High-trusters were more trustworthy and found it difficult to lie. They were more likely to contribute for public goods through volunteering. They were also observed to have the capacity to differentiate people as potential trustees or cheaters on the basis of fairly specific cues. The study also showed that high and low trusters had a diverse approach towards strangers, with the former more likely to trust strangers than the latter. The result was consistent with the experimental results on the trusting behaviour of citizens of the US and Japan by Toshio Yamagishi and Midori Yamagishi (1994).

    Thornton and Kline (1982) and T Yamagishi (1986, 1994) also presented scales to measure variations in trust among individuals known as the belief in human benevolence and trust scales respectively. A reasonable correlation between high trust measured by

    Economic Political Weekly

    august 27, 2011 vol xlvi no 35

    such attitudinal measures and behaviour in various experimental settings helps determine causation between trust and cooperative behaviour. However, survey researchers in several fields are still investigating what exactly the items included in trust scales actually measure (Levi and Stocker 2000).

    4.11.3 Trust and Games

    Instead of relying on a trust scale, the design of the standard prisoner’s dilemma game was modified by providing an exit option to subjects with the aim of overcoming the confounding relationship between cooperation and trust. It was predicted that the choice to play while there was an exit option could be considered as a behavioural indicator of trust whereas the choice to cooperate would simply be a behavioural indication of cooperation (Orbell, Schwarz-Shea and Simmons 1984). Although experimental results of this modified prisoner’s dilemma game were more relevant for an analysis of trust and cooperation, the investigation was more stimulated by the “trust game”. Under the trust game, the first mover (sender) of a subject pair is given a certain amount of money and asked whether he or she would like to send any amount to an anonymous partner (receiver). Any positive amount sent is tripled before it reaches the receiver, who then decides whether to return a part of the tripled amount to the sender. In this sequential game, with complete information, common knowledge of rationality and under anonymous conditions, the sender predicts that the receiver in an attempt to maximise his or her monetary pay-off may not share the tripled amount. Thus, the sub-game perfect Nash equilibrium is not to send and receive anything.

    Empirical evidence from experimental economics, however, strongly contradicts the standard perfect Nash equilibrium. Berg et al (1995) found that about 93% of proposers deviated from the theoretical sub-game Nash equilibrium and sent around 50% of their initial endowment to their partners. In sending money, the players trusted that the partners would return some money to them. Reciprocating, the partners returned sizeable amounts of the increased endowment. Thus, the amount sent could be treated as trust35 exhibited by the sender while the amount returned represented the degree of trustworthiness. The average amount returned stood at 46% of the total. More than 35% of subjects returned more than the amount sent. Glaeser et al (2000), in a similar experiment, found that an average 83% of the endowment was sent by the first player and 46% of the doubled amount was returned by the recipient. Barr (2003) also supported the high correlation between trust and trustworthiness by conducting the trust game in anonymity in 24 Zimbabwean villages with 141 pairs.

    4.11.4 Strategic-Cognitive View of Social Orientation

    The social motivations outlined by McClintock (1972), which were further used to categorise subjects, enjoy a great deal of importance in the experimental work on cooperation and trust. Some subjects were motivated by individual goal maximisation (egoist), others tried to maximise the gap between their own and other’s pay-off (competitive), and another category attempted to minimise the gap between their own and other’s pay-off (cooperative). These motivations also reflected, among others, differences in expectations that could vary according to social context or situations. Thus, trust was viewed in this context as a positive expectation of the level of cooperativeness of others. Kuhlman, Camac and Cunha (1986) called this approach the strategiccognitive view of social orientation. It was the expectation about the behaviour of others (that others will cooperate/compete) that allowed subjects to be classified as cooperative or competitive. Competitors were less likely to expect others to cooperate and tended to see them as competitive. But cooperators viewed others occasionally as cooperative and occasionally as competitive (but not uniformly cooperative). This also meant that cooperators, unlike competitors, were more likely to differentiate between situations in which it “paid” to cooperate and those in which it did not. Even egoists were ready to cooperate if they understood the long-term consequences of non-cooperative behaviour and at the same time trusted others (Yamagishi 1995).

    New Variants of Trust Games

    Along with these motivations and types of subjects, experimentalists have developed new variants of the trust game to examine the factors that facilitate or inhibit the emergence of trust under different social conditions. We now discuss the structures of the games conducted and the results obtained. Yamagishi (2001) found high trusters (based on a six-item general trust scale) were more accurate in estimating their partners’ behaviour than both low and medium trusters in the prisoner’s dilemma game. Here, subjects met each other in a group discussion that had no connection with the experiment. Yamagishi, Kikuchi and Kosugi (1999) found that high trusters were more sensitive than low trusters to information revealing a lack of trustworthiness in others. Similarly, since high trusters could make more accurate predictions, they cooperated more. Yamagishi, Cook and Watabe (1998) conducted two experiments with college students in the US and Japan with the hypothesis that social uncertainty would promote commitment and that high trusters would form committed relations less frequently than low trusters. The hypothesis was proved with experimental data. Low trusters in both cultures were less willing to take the risk of locating a more profitable opportunity in an uncertain environment. They were more concerned about avoiding the possibility of exploitation. They opined that high trusters were not gullible but appeared to be expert readers of social signals, knowing when it paid to cooperate and when it did not.

    Orbell and Dawes (1993) showed that high trusters were more likely to interact with strangers, at least in the first encounter. They enjoyed more profitable interactions on confirmation of their expectations about trustworthiness than low trusters. The risk-taking or low-caution attitude of high trusters was further supported by Wright, Maggied and Palmer (1975), who gave telephone invitations to high and low trusters (based on their rank on a trust scale) for a final experiment and examined the pattern of questions asked by the subjects. While high trusters asked for details about the experiments, low trusters asked suspicious questions (regarding the nature of the study or means by which the caller got their contact numbers). The exercise concluded that high trusters not only asked significantly fewer questions than low trusters but also significantly fewer suspicious questions. This behaviour by low trusters could be considered an indicator of their risk-averse attitude. Parks and Hulbert (1995) studied this aspect with the help of a public good game and a resource dilemma game. They found that in the presence of fear (of being cheated) high trusters would cooperate more frequently than low trusters, while in the absence of fear, both cooperated at the same level.

    With the aim of backing biologically and culturally derived hypotheses that reciprocity supports cooperation, experimentalists control for all alternative explanations of cooperation in trust games, such as the effects of repeated-game reputation, punishment threats and the availability of information. The results of these experiments confirm that reciprocity exists as a basic element of human behaviour and that this accounts for the trust extended to anonymous counterparts (Berg, Dickhaut and McCabe 1995). However, a delay of reciprocity affects cooperation adversely. Komorita, Hilty and Parks (1991) conducted two treatments of the prisoner’s dilemma game, one with an immediate reciprocation opportunity and another where the opportunity to reciprocate was postponed by one round. They observed that the mean proportion of cooperative choices was greater when reciprocation was immediately possible than when it was postponed. The delay in reciprocation might have been perceived as a lower willingness by the partner to be cooperative.

    The experiments surveyed so far mainly demonstrate the emergence of trust and cooperation in response to structural variations in the game setting. It would be interesting to study its emergence in various social contexts where the potential determinants are external or unalterable to the experimenter, such as the social identities of the subjects, size of the group to which the individual belongs, the degree of communication, the existence of time pressure, and third-party effects. In-group favouritism fosters trust for those on the inside a circle and distrust for outsiders. In experiments, De Cremer and Van Vugt (1999) found social identification had a positive effect on promoting cooperation in a social dilemma. They stated that selfish individuals could be encouraged to cooperate by increasing the salience of their group membership. Common group membership increased mutual trust in the group, which could increase the level of cooperation. Inter-group trust and cooperation is a complex phenomenon yet to be explored fully. The size of a group also matters. If the group size is sufficiently large, strategic considerations become less relevant, resulting in eroding the significance of trust in cooperation (Sato 1988). On the other hand, if in-group biases are created, the group size can augment the effect of trust on cooperative behaviour.

    Communication plays an important role in promoting trust and thus cooperation. Granberg, Stevens and Katz (1975), in repeated rounds of the prisoner’s dilemma game, showed that cooperation decreased over time in the absence of communication. Mori (1996) studied the effect of pre-experimental communication on general trust (towards human beings in general; strangers were a part of group) and specific trust (specific to a particular person or interaction). The results revealed both types of trust to be contingent on communication and the opportunity to communicate strongly affected specific trust towards a partner with whom communication had occurred. Caldwell (1976) and Dawes, McTavish and Shaklee (1977) also showed the significance of pre-game communication, especially face-to-face interaction, as a facilitator of cooperation.

    august 27, 2011 vol xlvi no 35

    To read the full text Login

    Get instant access

    New 3 Month Subscription
    to Digital Archives at

    ₹826for India

    $50for overseas users


    (-) Hide

    EPW looks forward to your comments. Please note that comments are moderated as per our comments policy. They may take some time to appear. A comment, if suitable, may be selected for publication in the Letters pages of EPW.

    Back to Top