“Siglo XX, Cambalache”
CONTEXTO HISTÓRICO 1999: Siglo XX “Cambalache”
Above, we saw that the governing institution in Fix guarantees efficiency by design (if it is implemented), but the lacking equality reduces its implementa-
0 5 10 15 20 Average Contribution 0 5 10 15 20 Period VCM FIX RE SIM BUN
Figure 3.1. Development of Average Contributions over Time
Notes: In all treatments (except for Vcm) contributions exhibit an increasing time-trend. The
trend is significant for Re (Spearman’s rho r = 0.18, p < 0.01) and Bun (r = 0.13, p = 0.02). In treatments Fix and Sim they fall short of being significant (Fix: r=0.08, p=0.16; Sim: r = 0.09,
p= 0.12). For Vcm, the trend is significantly negative (r = 0.47, p < 0.01).
tion rate. In Re, the forced equity garners high support among all types, but fails to ensure efficiency. In essence, the singular institutions are only able to address a singular matter of the efficiency-equity tradeoff. In the next step, we analyze whether the joint availability of both institutions will be able to handle this tradeoff successfully.
3.4.2.1 TreatmentSim
First, we look at the behavior in treatment Sim, where both institutions are avail- able and can be implemented separately. Here the voting process can have one of four outcomes: (1) if both institutions are supported by all three players, both will be installed and efficiency and equity will be ensured simultaneously. (2) if only the governing institution receives unanimous support, the ensuing VCM is governed by the same institution as in treatment Fix; (3) if only redistribution
3.4 Results | 61 0 .2 .4 .6 .8 1
Share of Groups with Institution
0 5 10 15 20 Period FIX RE SIM BUN
Figure 3.2. Share of Groups with Institution over Time
Notes: In the treatment with single institutions the share of successful formations increases sig-
nificantly over time (Fix: r = 0.654, p < 0.01; Re: r = 0.45, p = 0.05). The trend is similar but weaker in other treatments (Sim: r = 0.403, p = 0.058; Bun: r = 0.34, p = 0.14 ).
is adopted, the institution of treatment Re is present; (4) if both institutions are rejected, the regular VCM is played.
When faced with the two institutions simultaneously, the players vote such that at least one institution is installed in 74% of all cases. The share of groups with an institution is thus higher than in treatment Fix, but slightly lower than in treatment Re. Only the governing institution is implemented in 31%, while solely redistribution is established in only 8% of all cases and both institutions are installed in 35% of all cases. Especially the high prevalence of both insti- tutions being installed hints either at a high degree of inequality aversion by the high-type subjects or at the fear of a reciprocal exchange of votes among the different types (rejection of one institution inducing rejection of the other institution). Turning to voting behavior, we see that each type of player votes more frequently for the institution that offers the larger benefits to himself. High types vote in favor of the governing institution in 94% of all instances, but they support redistribution in only 68% of all cases. For low types the picture is just the opposite: 73% vote for the governing institution with fixed contributions
0 .2 .4 .6 .8 1
Share of Affirmative Votes
0 5 10 15 20
Period
high types low types
(a) Governing Institution
0 .2 .4 .6 .8 1
Share of Affirmative Votes
0 5 10 15 20
Period
high types low types
(b) Redistribution
Figure 3.3. Voting behavior on the two institutions by subject type in treatment Sim Notes: Left displays voting on the governing institution, right displays voting on redistribution.
Support of high types for the governing institution increases significantly over time (r = 0.227,
p< 0.01).
and 88% vote for redistribution.12Overall, the implementation rates are similar to the other treatments, and the contribution level of 16 tokens is only slightly higher (vs. Fix MWU p = 0.17; vs. Re MWU p = 0.87). If an institution is imple- mented, the contributions reach 19.2 tokens, which is very close to the efficient level of 20.
Result 3: If governance and redistribution are separately available,
a) the governing institution receives more support from high type players, b) the redistribution rule receives more support from low type players, c) and average contributions are not significantly higher compared to treatments with only a single institution.
Overall, we find that the simultaneous presence of both institutions is not able to increase contributions significantly above the level that was already achieved by a single institution. The inverse voting patterns of high and low types highlight their conflict to settle on a common institutional setup. We will now check whether this conflict is eliminated by the bundling procedure in treat- ment Bun.
12
If examined using a Wilcoxon matched-pairs signed-ranks test that uses the difference in average voting behavior of single group as independent observation, the differences in voting behavior across types are significant at the 5 % level (p = 0.012 for voting on governance, p = 0.014 for voting on redistribution).
3.4 Results | 63
3.4.2.2 TreatmentBun
Recall that the only difference between treatments Bun and Sim is the more restrictive choice set in Bun, because the decision on the formal redistribution rule is linked to the decision on the governing institution. Thus, subjects either have to support both institutions or none. In principle, both institutions could also be implemented in Sim. Yet, our data show that bundling is better at elimi- nating the tradeoff between equality and efficiency. The bundled institutions are installed in 87% of all instances. The share of successful institution formation is larger than in treatment Sim, even if the difference falls short of significance (MWU p=0.109). The high implementation rate is the result of nearly identical voting behavior by both types. High types vote in support in 93% of all cases, compared to 98% of low types. Already after a few periods, the support is nearly unanimous across both types.13The high rates of support come along with high contribution rates. Average contribution reaches 18.4 tokens, which is signifi- cantly more than in treatment Sim (MWU p=0.048). The average contribution rate being close to the efficient maximum indicates that the bundling approach is able to overcome the conflicting interest between heterogenous types (that was still present in treatment Sim, as seen by the frequent rejections). By bundling the governing institution and the redistribution rule, i.e., by combining equity and efficiency, the negative effects of heterogeneity and social preferences on institution formation are eliminated.
Result 4: The details of the implementation approach make a difference:
a) Support and implementation rates for the institutions are higher in treatment Bun than in treatment Sim.
b) Contribution rates are significantly higher in treatment Bun than in treatment Sim.