Professional Identity, Competition, and Unethical Behavior: A Field Experiment in the Chinese Banking Sector

Description:
The study employs a field experiment embedded within a online survey distributed to employees of a large rural commercial bank in China. The design consists of two main interventions: a between-subjects manipulation of professional identity salience and a within-subjects manipulation of competitive pressure.
1. Professional Identity Priming (Between-Subjects):
We manipulate the salience of professional identity by varying the timing of a specific survey module. This module contains questions regarding the employee's perception of the financial profession, specifically focusing on the concepts of "pride" and "honor".
Treatment Group (Priming): Participants answer these identity-related questions before performing the behavioral tasks. This serves to activate their professional identity and moral standards during the decision-making process.
Control Group (Baseline): Participants answer these same questions at the very end of the survey, after the behavioral tasks are completed. Thus, their decisions in the tasks are made without the induced salience of professional honor.
2. Competitive Pressure (Within-Subjects):
Participants perform an hypothetical "Mind Cheating Game" (mentally selecting a letter from A to F and reporting it for a monetary reward). Each participant performs this task under two distinct incentive schemes to measure their propensity for dishonesty under different pressures:
Individual Scheme: The payoff depends solely on the participant's own reported outcome (Piece-rate).
Competitive Scheme: Participants are paired with another randomly selected participant. The individual reporting the higher outcome value wins double the reward, while the other receives zero (Tournament).
Randomization:
Participants are randomly assigned to the Identity Priming Treatment or Control group upon entering the survey.
The order of the Individual and Competitive schemes is randomized at the individual level to control for order and learning effects.

2025-12-15

2026-01-31

Dishonesty (Cheating Rate):
Since the participant's "true" mental choice is unobservable, dishonesty is measured statistically at the aggregate level. In the Mind Cheating Game, a fair player should report each of the 6 letters with a probability of ~16.7%.
We define the primary outcome as the "High Payoff Reporting Rate" (binary variable equal to 1 if the reported letter corresponds to the highest or second-highest monetary reward, 0 otherwise). We compare the distribution of reported letters against the uniform distribution to infer the magnitude of cheating.

1. Heterogeneity by Job Role:
We investigate how the treatment effects vary across three distinct job roles: Branch Managers, Relationship Managers (RMs), and Tellers. We hypothesize that RMs, who face high-powered sales incentives daily, may react differently to the competitive scheme compared to Tellers, who operate under strict compliance rules.
2. Interaction Effects:
The interaction term between the "Professional Identity Prime" and the "Competitive Scheme" dummy. We measure whether the prime reduces the marginal increase in cheating caused by competition.

The experiment utilizes a Mind Cheating Game adapted for a online survey platform (https://www.wjx.cn/). The design employs a 2×2 mixed structure (Between-subjects: Identity Prime vs. Control; Within-subjects: Individual vs. Competitive Schemes).
1. The Mind Cheating Game Protocol:
In each decision round, participants are presented with a set of 6 letters (A, B, C, D, E, F). The task follows a strict three-step sequence to ensure validity:
Mental Selection: Participants are asked to mentally select one letter from the set.
Confirmation: They must click a button to confirm they have made their mental selection.
Payoff Revelation & Reporting: Only after the mental selection is confirmed, a payoff table is displayed showing the hypothetical monetary value associated with each letter. Participants then report the letter they selected.
Crucially, the mapping of letters to monetary values is revealed only after the mental selection, ensuring that the decision to lie (by reporting a high-value letter different from the mental choice) is a reaction to the specific moral trade-off presented, unconfounded by prior selection bias.
2. Treatment Blocks and Randomization:
Each participant completes two distinct blocks of the game, with the order of blocks randomized to control for order and learning effects.
Individual Block (5 Rounds): Participants play 5 rounds where the payoff depends solely on their own reported letter.
Competitive Block (5 Rounds): Participants play 5 rounds under a tournament scheme. They are informed they are paired with another randomly selected participant. If their reported total value over the 5 rounds exceeds their partner's, their reward is doubled; otherwise, it is zero.
3. Payoff Structure and Balance:
To ensure strict comparability between the Individual and Competitive conditions, the mapping of letters to payoff amounts is varied across the 5 rounds but is kept structurally identical between the two treatment blocks. We generated a balanced pool of letter-to-payoff mappings. In each round, the displayed mapping is distinct. This design rules out the possibility that observed differences in honesty are driven by random variations in specific letter values rather than the competitive nature of the task.
4. Hypothetical Incentive Design:
Due to strict regulatory constraints and administrative barriers within the banking sector, we employ a hypothetical vignette design for the monetary incentives. Participants are instructed to make decisions as if the monetary consequences were real. While traditional economic experiments often use real stakes, a growing body of methodological literature (e.g., in management and behavioral ethics) suggests that hypothetical measures of dishonesty are highly correlated with incentivized behavior, particularly in field settings where professional reputation is salient. We will provide auxiliary analysis and references in the final paper to validate the external validity of this design choice.

This experiment is embedded within an larger employee survey of the partner bank. The survey also includes psychometric scales and management questions intended for a separate correlational study regarding job performance.

Participants are randomly assigned to the Identity Priming Treatment or Control group upon entering the survey.
The order of the Individual and Competitive schemes is randomized at the individual level to control for order and learning effects.

Individual

No

0

Branch Managers: 500; elationship Managers:1200; Tellers: 1000

A random half of participants are assigned to the Professional Identity Priming condition, with the remaining half assigned to the control condition.

Based on the total number of available employees in the partner bank.