EXPERIMENTAL DESIGN

Table 1 summarizes the magnitudes of the experimental parameters used in the seven treatments conducted in this research. In the titles of the treatments, the notation “y” denotes group public good, “G” denotes societal public good, “LT” represents low tax, and “HT” stands for high tax. Each treatment was run twenty rounds, and in the beginning of each round every subject was endowed with an income (w) of 100 points. Twelve subjects were recruited for one session, and two sessions were conducted for each treatment. Subjects in the same session were divided into groups of four, and they were randomly re-matched when each new round started.

Denote n the size of the group and N the number of participants of each session. Hence, n = 4 and N = 12 were used in the experiment. All subjects were undergraduate students at National Chengchi University in Taiwan. They participated in the experiments voluntarily and none of them had ever participated in any public goods experiments.

[Table 1 about here]

There are two types of public goods in each treatment. One is the group public good (y), which only benefits the four members in the same group. The other is the societal public good (G), which yields benefits to all twelve participants in the same session. In the no-tax treatment

subjects pay no tax, and each of them decides that how many points of his (her) income he (she) would like to contribute to the group public good y and the societal public good G. Besides the no-tax treatment, there are three high-tax treatments (the only-y-HT, the only-G-HT, and the G&y-HT treatments) and three low-tax treatments (the only-y-LT, the only-G-LT, and the G&y-LT treatments). In the beginning of every round of each high-tax treatment, every subject has to pay a tax of 30 point, or 30 percent of his (her) income. In the only-y-HT treatment, all the tax is only allocated to the group public good y; in the only-G-HT treatment, the tax is only used to provide the societal public good G; and in the G&y-HT treatment, half of the tax (15 points) paid by each subject is used to provide y and the other half is used to provide G. The amount of the tax in each round of the three low-tax treatments is 16 points, or 16 percent of each subject’s income. In the only-y-LT treatment the tax is only used to fund the group public good y, and in the only-G-LT treatment the tax is only used to fund the societal public good G.

In the G&y-LT treatment, half of the tax (8 points) is used to fund y and the other half is used to fund G. After paying the tax, each subject then decides that how much of the rest of his (her) income (70 points in the high-tax treatments and 84 points in the low-tax treatments) he (she) would like to contribute to the public goods y and G.

According to the above setting, a subject’s contribution to the group public good y includes two parts: one is the tax that he (she) pays to provide y, and the other is his (her) voluntary (or private) contribution to y. Call the sum of these two parts the total contribution to y by a subject.

in the same session.

In the experiment the marginal per capita return of the group public good (MPCRy) is set at 0.4, and the marginal per capita return of the societal public good (MPCRG) is 0.2. That is, each point in the group public good y yields a return of 0.4 points to each subject in the same group, while each point in the societal public good G yields a return of 0.2 points to all subjects who participate in the same session. MPCRG is set lower than MPCRy because the societal public good G is farther from subjects than the group public good y in either location or emotion or both. Although MPCRG is lower than MPCRy, since the societal public good benefits more people, it has a higher APCR (average per capita return, which is defined as n×MPCR/N) than the group public good. Notice that MPCRy must meet the condition 1/n < MPCRy < 1, and similarly MPCRG must satisfy 1/n < MPCRG < 1. This is because if MPCRy  1 or MPCRG  1, then everyone will have an incentive to allocate all of his (her) income to public good y or G.

By contrast, if MPCRy ≤ 1/n, then no one will have an incentive to contribute to the group public good y. Likewise, if MPCRG ≤ 1/N, then will be no voluntary contribution to the societal public good G.

The no-tax treatment serves as the baseline treatment. Comparing the voluntary contributions to y and G in each high- and low- tax treatments with the voluntary contributions to y and G in the no-tax treatment helps us understand how the introduction of a tax crowds out voluntary contributions to the two public goods. Furthermore, a comparison between the voluntary contributions to y and G in the only-y-HT, only-G-HT, and G&y-HT treatments helps us exploring how the introduction of a high level of tax (30 percent of income) crowds out voluntary contributions to y and G when the tax is used to provide either y or G or is split equally between y and G. A similar comparison can be also made between the only-y-LT, only-G-LT,

and G&y-LT treatments. Finally, comparing the only-y-HT treatment with the only-y-LT treatment, the only-G-HT treatment with the only-G-LT treatment, and the G&y-HT treatment with the G&y-LT treatment tells us how the crowding-out effects are associated with the high and low levels of tax when the tax is used to provide only y, only G, or both.

The experimental procedure was run as follows. Recall that subjects were each endowed with a 100-point income in the beginning of every round. This income was deposited into each subject’s private account. After paying the tax (except in the no-tax treatment), each subject decides that how many points of his (her) remaining income he (she) would like to contribute to y and G. Only neutral terms were used in the experiment. Hence, in the experiment the group public good y was called the group account, the societal public good G was called the social account, the tax was called the automatic deposit, and the voluntary contribution was called the additional deposit. The amount of points remains in a subject’s private account is his (her) earning from the private account. The earning from the group public good y is MPCRy times the total contributions to y by the four members in the same group, and the earning from the societal public good G is MPCRG times the total contributions made by all twelve participants in the same session. A subject’s payoff in each round is the sum of the earnings from the private account, the public good y, and the public good G. At the end of each round, each subject received an earnings report, which indicated the amounts of tax that he (she) paid to provide the public goods y and G, his (her) voluntary contributions to y and G, the amount of taxes paid by

same earnings report.

Subjects were given written instructions in Chinese during the experiment. The experimenter read the instructions aloud, performed the calculations of two examples on the whiteboard, and answered any questions raised by the subjects. Subjects played there trial rounds first and then they were required to answer six questions related to the experiment. The experiment would not proceed unless all subjects had answered all questions correctly. Each session lasted about 80 minutes. The average payoff for all participants is NT$572.42 (with a standard deviation of NT$85.88, a maximum of NT$816.6, and a minimum of NT$385).¹

3. EXPERIMENTAL RESULTS