An experimental analysis of moral self-regulation

ABSTRACT

This article examines the validity of the moral self-regulation hypothesis in a laboratory setting. The experiment is comprised of a public good game preceded or followed by a matrix task. The data show that the recall of an immoral action (cheating in the matrix task) motivates the individual to do morally right thing (contributing to group account) and the recall of a moral action (contributing to group account) motivates the individual to act out self-interest (cheating in the matrix task). Both moral licensing and moral cleansing hypotheses are confirmed by the results of the experiment. Additionally, the findings indicate that the subjects who had been given a chance to cheat ‘at first’ allocated more funds to the group account; and the subjects who had been given a chance to voluntarily contribute ‘at first’ cheated more in the matrix task.

KEYWORDS:

• Moral self-regulation
• moral cleansing
• moral licensing
• public good game
• matrix task

JEL CLASSIFICATION:

• C90
• C91
• D03

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

¹ Andreoni (1990) uses a simple application of the public goods model in order to question the validity of the conventional view of charitable giving. Battigalli and Dufwenberg (2007) develop a general theory of quilt aversion in order to explain the economic behaviour of the agents who are affected by guilt. Charness and Dufwenberg (2006) investigate the influence of communication in a one-shot principal-agent game designed to capture the essence of hidden action. Fehr and Fischbacher (2003) identify conditions in order to show the interaction between selfish and strongly reciprocal individuals.

² Additionally, two benchmark experiments are implemented in order to control the results.

³ Past studies generally focus on drawing inferences on the distribution of lying behaviour (e.g. Ploner and Regner 2013; Meub et al. 2016).

⁴ Past research mostly uses a dictator game (e.g. Brañas-Garza et al. 2013; Clot, Grolleau, and Ibanez 2013).

⁵ The benchmark experiment in treatment 1 is named as control 1. The subjects who participated in control 1 only played the public good game.

⁶ Part 1 and part 2 are not within the same domain in this study. However, the past research regarding the moral licensing indicates that this is not a problem (e.g. Miller and Effron 2010; Tiefenbeck et al. 2013).

⁷ The benchmark experiment in treatment 2 is named as control 2. The subjects who participated in control 2 only took the matrix task.

⁸ 1.5 TL per correct matrix (this amount was equivalent to about 0.5 USD at the time of the experiment).

⁹ Because the same number is written into the last cell of the sample matrix, it is possible to identify each individual.

¹⁰ All groups in the study are mixed gender.

¹¹ Show-up fee is 5 TL.

¹² Wilcoxon test gives a significant result, as well (z = 6.25, p < 0.001).

¹³ The number of the subjects who cheated in the matrix task/treatment 1 was 81.

¹⁴ Additionally, a probit regression model is created to check the results. Charitable giving is used as the binary dependent variable (if the amount of money donated smaller than 15, the dependent variable takes the value of 0, and 1 otherwise. The result confirms the relationship between the level of cheating and charitable giving (coefficient = 0.3, p = 0.00).

¹⁵ Wilcoxon test gives a significant result, as well (z = −6.45, p < 0.001).

¹⁶ Additionally, a probit regression model is created to check the results. Cheating is used as the binary dependent variable (if the number of the difference is equal or smaller than 3, the dependent variable takes tha value of 0, and 1 otherwise. The result confirms the relationship between charitable giving and cheating (coefficient = 0.2, p = 0.00).