Mobility and Dynamics of Competition

The intervention we employ is revealing information feedback about the performance of other group members in the four-group competitive environment. We compare this to a control where participants receive information about their own performance and competitive thresholds only if competing. This question is important because there are many Government, companies, and other entities evaluating the implementation of fully transparent wage policies as a way to promote competition and reduce gender and other wage inequities.

2023-09-18

2023-12-31

Performance in the real-effort task, efficient sorting based on tier levels, the rate of upward and downward mobility, persistence (sticking with a tier after losing), comfort zone (moving up a tier after winning), confidence based on rank-belief elicitation. This analysis is performed over 10 competitive choice periods to understand the dynamics and mobility of competition.

risk-aversion, facial-expression valence and emotions, demographics

Eckel-Grossman task, and the quality and strength of emotions

Subjects are randomly assigned to groups of four members and groups are randomly assigned to one of two feedback conditions. In the Baseline condition, subjects learn the outcome of the round if competing, their own score and the score they had to beat, earnings for the round and the payment scheme they selected. In the Tier-Score-Disclosure treatment, subjects receive information about the scores of the other group members (and tier-scores) regardless of the endogenously selected payment scheme along with the information provided in Baseline. In both treatments, subjects complete a forced piece-rate in round 1, followed by a forced tournament in round 2. Following these two rounds, subjects make endogenous competitive-tiered or piece-rate entries for the next 10 rounds with feedback following round 3.

Not available

Randomization done by a computer. Subjects randomly seated in the lab.

Experimental groups.

Yes

60 groups.

240 student subjects.

120 student subjects for each feedback condition for a total of 240 subjects.

We rely on the meta-analysis by Markowsky & Beblo (Journal of Economic Behavior and Organization, 2022) for sample size estimation based on power analysis. Cohen's d in the seminal Niederle & Vesterlund (The Quarterly Journal of Economics, 2007) paper is roughly 0.77 (73% and 35% competitive entry rates of men and women, respectively, 40 men and 40 women) and Markowsky & Beblo's meta-analysis suggests that informational feedback or advice reduces the gender gap as an intervention moderator by about 8 percentage points. Accounting for the reduced gap due to informational feedback reduces the effect size to roughly 0.61. Using a significance level of 5% and a power of 80% as the basis for power calculation, we obtain results that suggest a sample size of 88 subjects (44 men, 44 women) is necessary for detecting similar effects for each of the two feedback conditions. The novelty of our design introduces a secondary (less competitive) level of competition between the piece-rate and the tournament prize which could further reduce the effect size in competing at the highest tier. While we expect more efficient sorting in the treatment where we disclose tier scores relative to the baseline, we only expect small (0.3) to medium (0.5) effect sizes. We take a conservative approach by committing to a total sample size of 240 subjects (120 subjects per feedback treatment) which assumes the ability to detect minimum effect sizes of 0.364.