Group size and matching protocol in contests

Abstract

We investigate the effects of partner and random stranger matching protocols in Tullock contests between two and between three players. In the groups of three players, we find no difference in the level or the dispersion of bids between matching protocols. For the two-player case, however, a partner matching results both in a lower level and a lower dispersion of bidsindicating tacit collusion. We conclude that unless one is interested in two-player contests or unless a particular matching protocol is required for the specific research, using a partner matching is preferred since it allows more flexibility in budget, logistics and data analysis. Furthermore, adding to the literature on group size, we find no difference in absolute bid levels due to group size across matching protocols. But the overbidding rate, defined as the bid over the Nash equilibrium level scaled with the Nash bid, is lower for the two-player groups.

Overbidding and heterogeneous behavior in contest experiments:A meta-comment on cross-cultural differences

“…Second, we test whether our extended analysis can replicate the findings of Sheremeta (2013) and Chowdhury and Moffatt (2017) with a particular focus on the null result related to learning in Tullock contests, which contradicts a substantial body of literature (Davis and Reilly, 1998;Fonseca, 2009;Sheremeta, 2010;Mago et al, 2016;Baik et al, 2021).…”

Identity, Communication, and Conflict: An Experiment

“…The matching protocol employed was random stranger, in which each subject was randomly matched with another subject in every round. We employed such a matching protocol because tacit collusion may arise in two-player experimental contests with a partner matching protocol (Baik et al, 2022). Stranger matching avoids any multiplying effect on tacit collusion when communication is possible.…”

The Lifecycle of Affirmative Action Policies and its Effect on Effort and Sabotage Behavior

“…Before the first round of Part 3, subjects were informed that they will interact with the same partner as in Part 2. Hence, our contests used a partner matching protocol (à la Baik et al, 2022) for studying repeated interaction situations that occur frequently in contests outside the laboratory, such as in the work-place settings.…”