410 E.-S. Park J. of Economic Behavior Org. 43 2000 405–421
two experimenters. Each session lasted about one-and-a-half hours. Average earnings per subject were 11.27.
In Experiment I, the Decomposed Game experiment was conducted to assess individual value orientations. Subjects were given instructions and a numbered envelope. A copy of
instruction for Experiment I can be found in Appendix B. Subjects were told that their task is to choose between Option A and Option B: each Option lists the points heshe will
receive or pay himherself and the points that the other will receive or pay. At no point did the subjects know what the other subjects were choosing. For each of the 24 decision
problems an experimenter provided each subject with a decision form. The 24 decision problems were presented in a random order from the list in Appendix A. Subjects made
their decisions by circling an option in a statement like “Choose between Option A=14.5, 3.9 and Option B=13, −7.5”. Subjects were told that each point is worth 0.10 in payoff
for example, 14.5 points would earn 1.45. In each round of decision problem they were instructed to put their decision form in the numbered envelope provided after they have
chosen their options. Subjects were told that after all of the 24 decision problems, the second experimenter will collect, shuffle the envelopes, and randomly pair one subject’s form with
that of another subject in this experiment. They were told that the second experimenter calculates total earnings for each subject and they will be paid after the completion of
Experiment II.
In Experiment II, Andreoni’s framing experiment was replicated. Instructions that are nearly identical to the one used in Andreoni’s experiment Andreoni, 1995, pp. 14–19
were used. The subjects were given instructions and a packet of 10 ‘Investment Decision Form’, which subjects use to record their decisions to allocate 60 tokens between Individual
Exchange Account and Group Exchange Account. The subjects played the game for 10 rounds, and are randomly and anonymously assigned to new groups each round. In each
round of the experiment, one experimenter collects an investment decision form from each subject. The experimenter, using a computer, randomly assigns subjects to groups of five,
calculates payoffs, and prepares an ‘earnings report’ for each subject. The earnings report tells subjects their investment decision, the group’s investment decision in the public good,
and their monetary payoffs. After the second experiment has ended subjects were paid privately their earnings, in cash, from the two experiments.
4
4. Experimental results
Of the 200 subjects 131 were classified as individualistic 65.5: 63 in the positive frame experiment and 68 in the negative frame. Sixty-four subjects were classified as cooperative
32: 34 in the positive frame and 30 in the negative frame. One subject in the negative frame was classified as competitive. Four subjects 3 in the positive frame and 1 in the
4
In order to obtain an independent data on value orientations of the subjects, the decomposed game is played before the public goods game. In the 10-round public goods game, the total contribution of a group to the public
good is revealed to each subject of the group in each round. Thus, if the public goods game is played before the decomposed game, subjects’ decisions in the decomposed game may be affected by their experiencespayoffs in
the public goods game.
E.-S. Park J. of Economic Behavior Org. 43 2000 405–421 411
Table 3 Mean percent of tokens contributed to the public good
Condition Orientation
1 2
3 4
5 6
7 8
9 10
All Positive 1
Ind 12 49.1
42.2 35.5
37.1 32.4
33.6 22.1
16.2 18.4
15.7 30.23
Coop 7 62.3
56.2 52.6
43.1 42.5
39.1 26.5
34.4 27.7
26.5 41.09
All 20 54.2
49.3 43.2
41.9 34.3
34.1 24.5
27.1 23.2
20.1 35.19
Positive 1 Ind 10
56.0 47.5
42.5 37.7
29.5 32.5
27.5 22.6
19.5 15.0
33.03 Coop 10
57.2 52.5
40.5 43.5
35.7 37.5
33.5 32.0
33.5 36.5
40.02 All 20
56.6 50.0
41.5 40.6
32.6 35.0
30.5 27.3
26.5 25.8
36.64 Positive 3
Ind 15 49.1
47.6 42.3
37.6 33.4
32.5 29.0
21.4 17.5
14.3 32.47
Coop 4 54.6
47.5 41.4
45.5 42.5
36.5 35.0
27.5 22.5
17.5 37.05
All 20 47.8
46.8 40.1
35.3 37.7
27.6 30.1
21.3 18.6
15.0 32.03
Positive 4 Ind 10
46.5 40.8
41.5 36.7
29.4 34.5
27.6 24.5
15.7 11.8
30.90 Coop 9
54.2 53.4
48.5 46.5
40.8 36.2
31.7 26.5
33.5 29.1
40.04 All 20
51.2 47.9
42.6 39.3
33.1 33.5
28.1 24.2
22.9 19.0
34.18 Positive 5
Ind 16 53.6
47.4 45.6
32.4 26.4
34.3 23.6
28.5 22.4
15.2 32.94
Coop 4 56.2
52.4 43.2
42.4 39.7
45.8 38.5
32.2 34.5
29.5 41.44
All 20 54.1
48.4 45.1
34.4 29.1
36.6 26.6
29.2 24.8
18.1 34.64
Average Ind 63
50.9 45.4
41.7 36.1
30.2 33.5
25.9 22.9
18.9 14.5
32.00 Coop 34
57.0 52.9
45.5 44.3
39.7 38.3
32.3 30.5
31.1 29.4
40.10 All 100
52.8 48.5
42.5 38.3
33.4 33.4
28.0 25.8
23.2 19.6
34.55 Negative 1
Ind 14 23.5
27.6 22.4
17.6 14.5
7.5 2.5
4.5 5.5
2.5 12.81
Coop 6 57.6
52.4 46.5
40.3 37.5
32.5 26.5
21.5 17.5
13.2 34.55
All 20 33.7
35.0 29.6
24.4 21.4
15.0 9.7
9.6 9.1
5.7 19.32
Negative 2 Ind 16
29.8 22.4
17.6 15.2
7.0 11.5
8.5 6.5
2.5 2.1
12.31 Coop 4
52.4 47.9
42.4 38.5
32.8 32.5
21.5 22.5
12.5 7.5
31.05 All 20
34.3 27.5
22.6 19.9
12.2 15.7
11.1 9.7
4.5 3.2
16.07 Negative 3
Ind 12 27.4
22.6 14.2
11.5 7.8
11.4 14.5
11.2 5.4
0.0 12.60
Coop 7 52.5
46.2 42.4
37.6 31.4
36.8 27.6
22.5 19.2
10.4 32.66
All 20 34.8
29.7 23.4
20.1 16.1
19.7 18.4
14.6 10.0
3.6 19.07
Negative 4 Ind 16
22.6 14.5
13.5 12.5
11.5 10.5
9.5 6.5
5.5 2.5
10.91 Coop 4
48.5 43.2
36.5 31.8
27.4 26.2
21.4 24.3
14.5 10.2
28.40 All 20
27.8 20.2
18.1 16.4
14.7 13.6
11.9 10.1
7.3 4.0
14.41 Negative 5
Ind 10 11.5
16.5 10.6
14.5 11.6
9.5 3.2
6.2 2.5
0.0 8.61
Coop 9 52.4
49.2 45.3
41.4 36.1
28.9 32.5
26.2 18.9
15.3 34.62
All 20 32.4
32.2 28.3
26.4 24.0
19.6 16.4
14.2 12.3
6.9 21.27
Average Ind 68
23.7 20.8
16.0 14.3
10.4 10.1
7.8 6.9
4.3 1.6
11.59 Coop 30
52.9 48.2
43.3 38.6
33.7 31.6
27.2 23.7
17.3 12.0
32.85 All 100
32.6 28.9
24.4 21.4
17.7 16.7
13.5 11.6
8.6 4.7
18.03 Difference
Ind 27.2
24.6 25.7
21.8 19.8
23.4 18.1
16.0 14.6
12.9 20.41
Coop 4.1
4.7 2.2
5.7 6.0
6.7 5.1
6.8 13.8
17.4 7.25
All 20.2
19.6 18.1
16.9 15.7
16.7 14.5
14.2 14.6
14.9 16.52
412 E.-S. Park J. of Economic Behavior Org. 43 2000 405–421
Fig. 2. Mean contribution percentage per condition.
negative frame appeared to choose randomly in Experiment I.
5
This distribution is similar to Offerman et al. 1996.
6
Table 3 lists the percent of the endowment subjects contribute to the public good each round see also Fig. 2. The first thing to note is that the results from the positive-frame
condition are very similar to the results of Andreoni’s positive frame experiment. Contri- bution rate starts at 52.8 and decays to 19.6 by round 10. Overall cooperation is about
35. Comparing this with the negative-frame, there is a considerable difference between the two conditions. Under negative frame, contribution rate starts at 32.6, which is about
35 below of the positive frame counterpart. Contribution rate declines to only 4.7 by round 10. On average, the subjects in the negative frame condition contribute about only
half as much as the positive frame subjects. The current experiment with a bigger number of subjects and more experimental sessions confirms Andreoni’s framing effect.
The statistical significance of the difference is tested using Mann–Whiteny rank-sum nonparametric test. This test is conducted using the mean contribution levels per round
as observations. The test statistic is z=3.189. That is, there is a significant difference in contribution rates under the two frame conditions.
7
We now look at contribution rates of the subjects with different value orientations in each of the two treatments positive and negative frames. Under positive frame, the contribution
rates of cooperators are consistently higher than those of individualists throughout the
5
Following Offerman et al., 1996, p. 827, random players are those subjects whose consistency measures are below 33. A choice is consistent if the alternative closest to the observed motivational vector is selected. For
example, if the observed motivational vector is 15, 0 and a subject chooses between Option A=−10.60, −10.60 and B=−7.50, −13,00, then B would constitute the consistent vector. p. 823 The overall consistency level for
the experiment is about 90.
6
In Offerman et al. 1996 65 of the subjects were individualistic, 27 were cooperative, 4 were competitive, 1 were aggressive, and 3 were random players.
7
Even though subjects were randomly and anonymously regrouped after each round, group contributions are not completely independent. If we look at data only from round 1 to guarantee independence, the result still holds
at a 5-level, z=2.090.
E.-S. Park J. of Economic Behavior Org. 43 2000 405–421 413
Fig. 3. Contribution percentage per orientation under positive frame.
Fig. 4. Contribution percentage per orientation under negative frame.
10 rounds, even though, on average, the difference between the two contribution rates of cooperators and individualists is only 7.9 see Fig. 3. The difference is statistically
significant z=2.771.
8
Under negative frame, on average, there is a considerable difference 21 in contribution rates between individualists and cooperators see Fig. 4. Again,
Mann–Whitney test confirms the difference is significant z=3.598.
9
We now turn to the framing effects on subjects with different value orientations. First, for those subjects classified as individualists under the positive frame, contribution rate
begins at 50.9 and declines to 14.5 by the end of the experiment. Overall contribution rate is 32. Comparing this with individualistic subjects in the negative frame, there is a
8
Based on the first round data only, z=3.480.
9
Based on the first round data only, z=4.822.
414 E.-S. Park J. of Economic Behavior Org. 43 2000 405–421
Fig. 5. Individualists contribution percentage per condition.
Fig. 6. Cooperators contribution percentage per condition.
considerable difference between the two conditions. Under the negative frame, contribution rate of individualists starts at 23.7, which is not even 50 of the positive frame counterpart
see Fig. 5. On average, the individualists in the negative frame condition contribute only about 36 as much as the positive frame individualists. Again, Mann–Whitney test confirms
the difference is significant z=3.589.
10
On the other hand, for those subjects classified as cooperators under the positive frame, contribution rate begins at 57 and decays to
29.4 by round 10. Overall contribution rate is 40.1. Comparing this with cooperative subjects in the negative frame, the difference between the two conditions is not as striking
as individualists. Under the negative frame, contribution rate of cooperators starts at 52.9 and decays to 12. Even though there is a considerable difference by round 10 29.6 versus
12, on average, the cooperators in the negative frame condition contribute about 82
10
Based on the first round data only, z=4.822.
E.-S. Park J. of Economic Behavior Org. 43 2000 405–421 415
Table 4 Summary of contribution percentage analyses
Contribution percentage Difference at 5-level
Overall Positive frame vs. negative frame
Significant z=3.189 Per condition
Positive frame Individualists vs. cooperators
Significant z=2.771 Negative frame
Individualists vs. cooperators Significant z=3.598
Per orientation Individualists
Positive frame vs. negative frame Significant z=3.589
Cooperators Positive frame vs. negative frame
Insignificant z=1.227
as much as the positive frame cooperators see Fig. 6. Mann–Whitney test confirms the insignificant difference z=1.227.
11
Table 4 summarizes the analyses thus far. We now look at the effect of framing on the propensity of subjects to free ride. Table 5
lists the percent of subjects choosing the strategy of free riding contributing zero token to the public good for each round.
12
For all subjects the positive frame starts with 16 of subjects free riding, rising to 45 by round 10, with a 10-round average of 33.9. On
the other hand, the negative frame starts with 35 of the subjects free riding, rising to 79 by round 10, with a 10-round average of 55.1. Even though, percent of subjects
free riding under the negative frame is considerably less than the counterpart of Andreoni’s experiment, on average, percent of negative frame subjects free ride is higher than that of
positive frame. The difference between conditions is about 21, while the difference in Andreoni’s experiments is nearly 30. The Mann–Whitney test statistic is z=3.182, which
is highly significant: the negative frame of the public goods problem significantly increases the percent of subjects contributing zero.
Analyses per condition and per value orientation using the data on percentage of free-riders confirm qualitatively the same effects as those using the data on contribution rates. They
are summarized in Table 6. In summary, when the incentives of the experiments are identical under the two framing
conditions, this experiment confirms Andreoni’s significant effect of framing on overall contribution rates. Its effect is most eminent on the subjects with individualistic value
orientation, but not on the subjects with cooperative value orientation.
5. Discussion
