Voluntary contributions to public good. What does experimental evidence say?

Voluntary contributions to public good. What doesexperimental evidence say?

In recent years, a substantial number of empirical

studies has been conducted in order to analyse one of the

central issues in public economics – voluntary contributions

to a public good. Since the definition of a public good (PG)

suggests that PG is both non-excludable and non-rival, each

individual has no incentive to contribute to the provision of

the PG (Stiglitz, 1988). According to standard economy theory,

individual is rational and driven by self-interest, hence in

the context of voluntary contributions to the PG, individual

has an incentive to rely on others and choose his best

response to free-ride (or, in other words, contribute zero).

However, the empirical studies (Andreoni, 1995; Fischbacker et

al., 2001) provide evidence that in reality people are less

selfish and tend to contribute to the PG more than theory

predicts. This essay aims to review three empirical studies

and summarise their main findings. The paper will be organised

as follows: the first part will summarise the James Andreoni’s

(1995) paper Cooperation in Public-Goods Experiments: Kindness or Confusion?;

the second section will deal with Fischbacher et al (2001)

1

research Are People Conditionally Cooperative? Evidence from a Public Goods

Experiment; the third part will review The Crowding-out Effect of

Governmental Transfers on Private Charitable Contributions: Cross-Sectional

Evidence by Abrams and Schmitz (1984); the final part will draw

main conclusions and findings.

James Andreoni (1995) in his research uses a reiterated

laboratory experiment in order to distinguish two prevailing

hypotheses in public-goods experiments – kindness and

confusion. Kindness is usually related with people’s concern

not only about the maximisation of their own earnings but also

about others’ welfare. In the meantime, confusion, as

explained, can arise due to subjects’ inability to understand

the instructions or due to low monetary payoffs. In order to

test both hypotheses, Andreoni (1995) conducted an experiment

where 120 undergraduate economics students were placed into

groups of 5. Each individual was endowed with 60 tokens budget

which expires after each round. As in the standard experiment,

subject has to decide how many tokens should be invested in a

PG and how many – in private consumption. As a result,

individuals face a prisoner’s dilemma; marginal returns are

2

set in a way so that the dominant strategy is to free-ride and

invest zero into the PG.

Andreoni (1995) in his methodological design aims to

remove social and cultural factors that might potentially lead

individuals to kind behaviour and establish three conditions.

(1) Regular condition – individuals are paid according to

their experimental earnings (most cooperative); (2) Rank

condition – subjects’ monetary payments depend on their

experimental earnings rank in comparison to other subjects in

the group (least cooperative and less incentives for kind

behaviour); (3) RegRank condition – same treatment as in the

Regular condition except individuals have an access to

information about the ranks.

In addition, given the three conditions, Andreoni (1995, p.

895) notes that “the difference in cooperation between RegRank

and Rank will provide a measure of the minimum amount of

cooperation that would be attributable to kindness” whereas

“the cooperation seen in the Rank condition will provide a

measure of the minimum amount of cooperation that is

attributable to confusion”.

3

In terms of results, Andreoni (1995) finds that there is

a substantial decline in contribution to the PG in all three

treatment groups (Figure 3). As to be expected, individuals in

Regular condition contributes more to the PG than those in

Rank condition. Additionally, the differences all three

conditions are statistically significant.

Figure 3. Source: Andreoni (1995)

Figure 4 allows to analyse motivations of the individuals.

As it is seen, in round 1 confusion accounts for 81.3 percent

of cooperation, and, due to experience and learning, by round

10 it decreases to 13.6 percent. In the meantime, kindness

increases and reaches its peak by round 5 (60 percent), and

then falls to 50 percent by round 10. The patterns of kindness

and confusion reveal that from round 1 to 6 the total amount

of cooperation is rather stable, however, confusion is

declining and kindness is increasing (Andreoni, 1995). From

round 6 to the final round, confusion is relatively stable

4

whereas kindness falls (from 60 percent to 50 percent)

(Andreoni, 1995). This implies that since the game is

reiterated, those players, who were confused at the beginning,

figure out their dominant strategy (to contribute zero);

however, they attempt to cooperate due to kindness but

eventually still choose to free-ride.

Figure 4. Source: Andreoni (1985)

The final results reveal 75 percent of individuals are

cooperative, and that both kindness and confusion are equally

significant. Although subjects understand the incentives and

5

their dominant strategy, they still tend to make cooperative

decisions implying that individuals are less selfish than

rational economic theory predicts. Therefore, there is

something unique in human nature that in essence contradicts

static economic theory.

Another study by Fischblacker et al. (2001, p. 397)

carried out a one-shot public goods experiment and found that

“a third of the subjects can be classified as free riders,

whereas 50 % are conditional cooperators”. Before moving to an

investigation of the experimental design and results, it is

important to define “conditional cooperators.” According to

Fischblacker et al. (2001, p. 397), conditional cooperators

are those “who are willing to contribute the more to a public

good the more others contribute.” In essence, people’s

strategy whether to contribute to the PG or not depends on how

other individuals behave. In other words, assumption of

conditional cooperators tests whether there is a correlation

between other subjects’ willingness to be cooperative and

his/her own behaviour.

In order to directly elicit individuals’ preferences for

conditional cooperation, Fischblacker et al. (2001) apply

6

strategy method. Authors conducted a computerised lab

experiment in two sessions where 44 non-economics

undergraduate degree students took part in. These individuals

formed a total of eleven groups of four subjects, and since

the game was repeated only once, the decisions of all

individuals involved are independent observations

(Fischblacker et al., 2001). As in the standard PG experiment,

subjects (in this case – 4) are endowed with a certain amount

of tokens (in this case – 20) and have to decide what fraction

of tokens should be invested into a PG and how many should be

kept for their own consumption. The simplified payoff function

set by authors shows that if individual contributes to the PG,

he yields 0.4 tokens. As a result, subject, as in the standard

game theory, is facing a prisoner’s dilemma where his dominant

strategy is to free-ride.

Once authors verified that all subjects understood the

principles of the experiment, two types of contribution

decisions were asked to make. Fischblacker et al. (2001) asked

subjects to make both decisions: an “unconditional

contribution” and a “contribution table”. The “unconditional

contribution” is the standard decision where individual,

7

without knowing others’ choices, has to decide how many tokens

should be allocated to the PG; whereas the “contribution

table” (which later elicits a contribution schedule) is a

strategy method where individuals “have to indicate for each

of the 21 possible average contribution levels of the other

group members (rounded to integers) how much they are willing

to contribute to the public good” (Fischblacker et al., 2001,

p. 399).

Additionally, authors encouraged individuals to take

both types of decisions seriously by accommodating a random

mechanism that controls which of the two decisions will become

the one determining the actual payoffs (Fischblacker et al.,

2001). As a result, for each individual the probability of

having a payoff-relevant decision in a group of 4 subjects is

¼. Authors create an experiment in a way so that subjects

would be able to learn the rounded average contribution of the

other individuals and make the decision accordingly. The

rational theory predicts that provision of information related

to others’ average contributions will lead the subject to

free-ride (i.e. contribute zero).

8

In terms of results, authors found that even though the

game was played only once, not all the participants choose to

free-ride. As it is seen in Figure 1, the bold line represents

an increasing trend in the average contribution of other group

members. Therefore, according to Fischblacker et al. (2001, p.

401), on average, “subjects display conditional cooperation”.

Figure 1. Source: Fischblacker et al (2001)

However, the patterns on individual level reveal more

interesting results. In essence, the data on individual level

displays that subjects are heterogeneous in their preferences

and behaviour. To illustrate that, Fischblacker et al. (2001)

analyse data and establish three different categories (Figure 2)

9

Figure 2. Source: Fischblacker et al. (2001)

Conditional cooperation (22 subjects) that represents the

largest fraction of all treatment groups (50 % of all

participants). However, it is important to note that decisions

are different, and the data reveals a bias in the selfish

direction (see for example 19, 24, 39) (Fischblacker et al.,

2001). These decisions lie at or below the diagonal. Free-

riders (13 subjects) represent 1/3 of all individuals in the

experiment who are driven by self-interest irrespective of how

others behave. “Hump-shaped” contributions (6 subjects) are

10

those subjects who tend to contribute up to 10 tokens of the

other group members, however, beyond this level their

investments into the PG fall (Fischblacker et al., 2001).

Abrams and Schmitz (1984) used a cross-sectional data

from 1979 itemized tax returns in the US in order to

investigate the effect of state-level social-welfare transfers

on private-charitable giving. It is usually assumed that

charity is a normal good implying that with an increase in

income (positive income elasticity), people tend to

demand/give for charity more whereas negative price elasticity

in demand indicates as the price of giving goes up, the

quantity of charitable services demanded decreases (Abrams &

Schmitz, 1984). Additionally, some empirical studies found

that government’s expenditure on social welfare might

potentially affect private charitable contributions. More

precisely, private donors perceive government spending as aid,

thus the incentives as well as utility of private spending on

charities decrease.

Authors run log-linear regression (Figure 3) where GIVEij is

the average charitable contribution in income class i in state

j; Yij is the average taxable income in class i in state j; Pij

11

stands for one minus the marginal tax rate applicable to class

i in state j; Tj represents state j’s governmental social

welfare transfers; POVERTYj is a measure of charitable need in

state j. In order to avoid biases in their estimations, Abrams

& Schmitz (1984) use a religion variable to control for

people’s preferences and tastes assuming that some people in

certain states can be more altruistic than others.

Figure 3. Source: Abrams & Schmitz (1984)

In terms of results (Figure 4) in the simplest model 1st

Abrams and Schmitz (1984) find that controlling only for

prices and income reveal expected results. More concretely,

with an increase in income people tend to demand more charity

(coefficient of 0.611) whereas with increase in charity’s

prices – people demand less (coefficient of -1.239). A

variable for POVERTY (positive coefficients) reveal that an

increase in poverty would result in more private contributions

to charity implying that people take into account need of the

residents. Finally, control for state’s social welfare

12

transfers (negative coefficients) indicates that an increase

in government’s spending on social welfare crowds-out private

spending and reduces community’s incentives to donate.

Figure 4. Source: Abrams & Schmitz (1984)

Taking all arguments into consideration, this essay aimed

to review three empirical studies that analysed voluntary

contributions to the provision of the PG. In general, all

three empirical papers displayed results that contradict

standard economic theory where voluntary contributions to the

PG would lead to undersupply due to individuals’ selfishness.

Andreoni’s (1985) attempt to separate kindness and confusion

13

reveals that 75 percent of individuals are cooperative, and

that both kindness and confusion generate cooperative

decisions in public-goods game. Fischblacker et al. (2001)

conducted a one-shot public goods game aiming to elicit

individuals’ preferences for cooperation and found that half

of the subjects are conditionally cooperative whereas one-

third are free-riders. Abrams and Schmitz (1984) used a cross-

sectional analysis and proved that with increasing income

people tend to contribute more to charities and are sensitive

to rising poverty, however, an increase in governmental

spending on social welfare reduces people’s contributions

implying that governments ‘aid’ is perceived as substitute for

private contributions.

14

REFERENCES

Abrams, B. A., & Schmitz, M. D. (1984). The crowding-out

effect of governmental transfers on private charitable

contributions: cross-section evidence. National Tax Journal, 563-

568.

15

Andreoni, J. (1995). Cooperation in public-goods experiments:

kindness or confusion?. The American Economic Review, 891-904.

Fischbacher, U., Gächter, S., & Fehr, E. (2001). Are people

conditionally cooperative? Evidence from a public goods

experiment. Economics Letters, 71(3), 397-404.

Stiglitz, J. (1988). Economics of the public sector (2nd ed.). New

York: W.W. Norton.

1985 words in total

16