Reciprocity and resistance to comprehensive reform

For a comprehensive survey on resistance to reform see also Heinemann (2004).

McKelvey and Ordershook also ran treatments in which any majority of participants could agree on passing or failing bills by signing an agreement card (with unrestricted discussion beforehand). In these treatments they mainly observe fair outcomes in three-player games and support for the competitive solution in five-player games.

We do so in order to abstract from additional reasons affecting coalition formation such as overrepresentation of own preference intensities (see, e.g., Myerson and Satterthwaite 1983; Casella 2005; Jackson and Sonnenschein 2007; Engelmann and Grimm 2012).

A copy of translated instructions can be found in the working paper version of this article at http://​www.​twi-kreuzlingen.​ch/​uploads/​tx_​cal/​media/​TWI-RPS-051-Fischbacher-Schudy-2010-05.​pdf.

For a risk-neutral subject, the probability of reward p _r is sufficiently high if p _r >1/3 because voting for another’s bill costs two points whereas reward yields six additional points.

The proof of this proposition can be found in Appendix A.

Proof of Proposition 2 can be found in Appendix B. Note also that due to the random matching procedure, there is no incentive for individual reputation building across periods, which might induce any additional motives for supporting monetarily unfavorable reforms.

The number of observations in Table 3 is calculated as follows: In OpenBallotCR reform failure occurred in 158 out of 252 cases. In SecretBallotCR reform failure occurred in 97 out of 204 cases. Thus in total there exist 255 situations in the CR treatments in which two subjects can cast a vote for an unfavorable bill which yields (2×255)=510 observations. For the NoCR treatments the number of observations can be easily derived from the number of subjects (51+54=105). Two-thirds of 105 subjects make a decision in 12 periods which yields a total of 540 observations. Note further that we cluster on the matching groups in order to control for players’ experience.

In order not to crowd the table unnecessarily we do not here include the regressions results for the third bill.

This share has to be considered as a lower bound for efficiency concerns, because not accepting subsequent bills after one’s own bill failed can be caused by negative reciprocity too.

To control for possible end-game effects, we also ran the regressions in Table 5 including a last period dummy. For both, reward and trust, the last period dummy is negative, but statistically insignificant. The coefficient of transparency is robust. For reward, the time trend variable is still negative but statistically insignificant. The coefficient of the interaction term (Period-1) × Comprehensive Reform Treatment is robust.

We cannot infer whether the latter increase is due to positive signaling or because members preferring the third bill expect the second bill to be passed anyway and therefore have nothing to lose by voting for the second bill.

To control for possible end-game effects, we also ran the regressions in Table 9 including a last period dummy. For reward (Open and Secret Ballot) and trust (Secret Ballot), the regression results are robust and the last period dummy is statistically insignificant. For trust in Open Ballot, the last period dummy is negative and statistically significant at the ten percent level.

We thank an anonymous referee for pointing this out.

See, for instance, Geng et al. (2011), who investigate how two different types of electoral campaigns (self-descriptions of personality and promises regarding prospective in-office behavior) affect choices by elected representatives. The authors find supporting evidence for the guilt aversion hypothesis (Charness and Dufwenberg 2006). When elections were promise-based elected candidates transferred more money to recipients than candidates selected by a random draw (although promises did not differ). Also, promises and beliefs on voter expectations were positively correlated but correlations between dictators’ second-order beliefs and their choices were weaker than predicted. Further, results from Weiss and Wolff (2013) cast doubt on the robustness of the finding that a voting mechanism may create or strengthen an entitlement effect in political-power holders.

See, e.g., Leibbrandt and Sääksvuori (2012).