1. 1 Coordinating Collective Resistance through Communication and Repeated Interaction Timothy N. Cason, Purdue Univ. Vai-Lam Mui, Monash Univ.

2. 2 The Divide-and-Conquer CR Game AcquiesceChallenge Acquiesce12, 2, 212, 2, 1 Challenge12, 1, 20, 7, 7 Subordinate B Subord. A Leader Transgress against both AcquiesceChallenge Acquiesce6, 8, 86, 8, 7 Challenge6, 7, 80, 7, 7 Subordinate B Subord. A Not Transgress AcquiesceChallenge Acquiesce8, 2, 98, 2, 8 Challenge8, 1, 90, 7, 7 Subordinate B Subord. A Transgress against A Transgress against B (symmetric payoffs to matrix in subgame above)

3. 3 Transgression and Resistance What institutions and social mechanisms can constrain leaders who have incentives to exploit their power? –A recent theme in political economy and organizational economics –Coordination problem in collective resistance

4. 4 Transgression, Collective Resistance, and Communication Multiple equilibria The outcome of no transgression against any subordinate can not be supported as part of a SPNE with purely self- interested agents The “beneficiary” subordinate, who receives some surplus when transgression occurs against the other, has a dominant strategy to acquiesce Communication should not change the fact that no transgression cannot be supported as an equilibrium However, if some beneficiaries are altruistic punishers, then some joint resistance can occur in the divide-and- conquer (DAC) subgame This also implies that (cheap talk) communication might facilitate coordination against transgression (Cason and Mui (2006) found support for this hypothesis)

5. 5 What about Repeated Interaction? Is repetition also effective in facilitating collective resistance in the absence of communication? –Our previous results suggest that even finite repetition may help increase resistance and reduce transgression (type identification) –Due to multiple SPNE in the stage game, cooperation among subordinates is an equilibrium even for (short) finite repetition –Weingast (1997), citing the folk theorems, emphasizes infinite repetition Is repetition more effective than communication in facilitating collective resistance and deterring transgression?

6. 6 Experimental Design (468 Total subjects) Random Matching Long Horizon Finite Repetition (50 periods) Equivalent Horizon Finite Repetition (8 periods) Indefinite Repetition (7/8 prob of continuation) No Communication 8 Sessions (72 Subjects) 6 at Monash Univ., 2 at Purdue Univ. 6 Sessions (54 Subjects) 2 at Monash Univ., 4 at Purdue Univ. 6 Sessions (54 Subjects) 2 at Monash Univ., 4 at Purdue Univ. 6 Sessions (54 Subjects) 2 at Monash Univ., 4 at Purdue Univ. Ex Post Communication 8 Sessions (72 Subjects) 6 at Monash Univ., 2 at Purdue Univ. 6 Sessions (54 Subjects) 2 at Monash Univ., 4 at Purdue Univ. 6 Sessions (54 Subjects) 2 at Monash Univ., 4 at Purdue Univ. 6 Sessions (54 Subjects) 2 at Monash Univ., 4 at Purdue Univ.

7. 7 Results Summary: The Big Picture Repetition reduces the transgression rate, but communication (even with random matching) is at least as effective as the best form of repetition Communication, but not repetition, significantly increases the subordinates’ coordinated resistance Type identification of the other subordinate appears to be better facilitated through (restrictive) communication than repeated play

8. 8 No Transgression Rates Communication Random Matching, No Comm. Baseline

9. 9 Rates of No Transgression Treatment Averages Dropping Periods 1-20 Random Matching Long Horizon Finite Repetition Equivalent Horizon Finite Repetition Indefinite Repetition No Communication 7.5%32.6%20.2%12.1% Ex Post Communication 24.7%55.7%39.9%46.8% Note: Red arrows denote significant differences at p-value<0.05; blue arrows denote significant differences at p-value<0.10 (one-tailed Mann-Whitney tests)

10. 10 No Transgression Rates: Summary Repeated play—particularly repeated play over a long horizon—reduces the rate that leaders transgress. Cheap talk, in the form of binary signals of intentions ex post, is just as effective as the best type of repeated play (Long Horizon) in reducing the transgression rate (Mann- Whitney U=20, n=8, m=6; ns). Holding the matching protocol constant, adding cheap talk always reduces the transgression rate, although this effect is only marginally significant in the finite repetition treatments. The No Transgression rate exceeds 50 percent in the treatment with a long but finite horizon and cheap talk, compared to less than 10 percent in the baseline random matching treatment with no communication.

11. 11 Successful Joint Resistance Rates

12. 12 Random Matching Long Horizon Finite Repetition Equivalent Horizon Finite Repetition Indefinite Repetition No Communication 7.1%14.0%8.4%9.5% Ex Post Communication 14.8%22.4%16.0%26.9% Successful Joint Resistance Rate Treatment Averages Dropping Periods 1-20 Note: Red arrows denote significant differences at p-value<0.05; blue arrows denote significant differences at p-value<0.10 (one-tailed Mann-Whitney tests)

13. 13 Successful Joint Resistance Rates: Summary Repeated play does not increase the rate of successful joint resistance to DAC transgression in the No Communication condition. Within the communication condition, only indefinite repetition increases the joint resistance to DAC rate, compared to communication/random matching. Communication increases the rate of successful joint resistance, compared to no communication and random matching baseline, for all matching treatments (random effects probit model; highest p-value<0.01).

14. 14 Successful Joint Resistance Rate Treatment Averages Dropping Periods 1-20 Random Matching Long Horizon Finite Repetition Equivalent Horizon Finite Repetition Indefinite Repetition No Communication 7.1%14.0%8.4%9.5% Ex Post Communication 14.8%22.4%16.0%26.9% Note: Red arrows denote significant differences at p-value<0.05; blue arrows denote significant differences at p-value<0.10 (one-tailed Mann-Whitney tests)