Robustness Test Of ``moral Wiggle Room``

Abstract:

The experimental literature on fairness shows that the subjects have a preference

to help others even when there are no strategic motives for their action. Dana, Weber,

and Kuang s(DWK), (2007) find that the manipulation of the dictator game to relax

the one-to-one map between the dictators actions and outcome significantly decrease

fair behaviour. We design an experiment to study the robustness of their results and

the data of this within-subjects experiment confirms DWK s findings. Different from

DWK s results, we find 10% higher choice of self-interested outcome in the hidden

information treatment. In overall, 27% of the subjects deviate from their fair choice of

transparent treatment in the hidden-information treatment.

Introduction

The subjects in various experimental research show concern for others benefit. Dictator

game is the most precise game to show the altruistic behaviour in which dictator distribute an

endowment between herself and receiver in a one-shot decision. The receiver must accept the

distribution, and there is no strategic motivation for the dictator to allocate an endowment

to the receiver. Camerer (2003) shows that a majority of subjects have strong tendency to

give a positive amount to the receiver. The average amount of giving in dictator game is over 20%. Double-blind anonymity shows that dictator gives more than zero to an unknown

recipient (Hoffman et al. (1994)).

Experimental results on the altruistic behaviour in the dictator game depends on the context

that the game is being played. The seminal work in the contextual altruistic behaviour

by Hoffman et al. (1996) demonstrates that increasing the level of anonymity reduces the

dictators allocation of the endowment to a recipient. Also, the dictators giving declines

when there is a possibility to take from the recipient List (2007). This result shows that the

variation of dictator choice set changes the giving behaviour of the dictators. Introduction of

taking from partner s endowment decrease the level of generosity in the dictators decisions

Bardsley (2008).

In a significant work on contextual altruistic behaviour, Dana et al. (2007) (henceforth

DWK) design an experiment in which the dictators are given various excuses to choose the

selfish outcome. The participants are given an option of either staying ignorant about the

consequences of their decisions or click on a button to access information on how their choice

would influence their partners payoff. They reduce the level of transparency on the cause

of self-interested choices by introducing the excuses to dictators decisions. They show that

decreasing transparency lower the number of giving and create more selfish behaviour. In

the presence of the costless option, almost half of the subjects decide to stay ignorant about

the effect of their decision on their counterparts payoffs and choose the selfish outcome. The

number of fair choices decreases dramatically from the treatment where the subjects did not

have an option to stay ignorant.

Dana et al. (2007) argue that there might be other important motives for the altruistic

behaviour that cannot be captured by solely monetary payoffs. The situational pressure like

feeling compelled might be the main force to drive such fair choices, and in the absence of

transparency in consequence of the decisions, this apparent generosity is changed to selfish

behaviour. They report that ".. people feeling compelled to give due to situational factors,

while not valuing the corresponding outcomes."(p.77). These findings have a substantial effect on the interpretation of altruistic behaviour and the discussion on the motivation of

other-regarding choices. In the first part of this paper, we test the robustness of DWK s

results.

This experiment tests the robustness of DWK s results to the different pool of the subjects

and a change in experimental design. In a difference with DWK s experiment, we

conducted a within-subjects experiment that allows subject to make choices sequentially

across treatments. Within-treatment has a higher power to show the change in the choices

across the treatments. Since we track the subjects choices, we could see the decision in the

transparent treatment and if there is any change in the hidden-information treatment. Also,

the within-subject design enables us to collect more data in the experiment which makes the

statistical analysis more powerful.

We study the effect of uncertainty in the payoff matrix of the counterparts on the subjects

choices. The game that the subjects play in this experiment is the same as second treatment

of Dana et al. (2007). We have run this experiment at Tehran University in Iran. While

in DWK s experiment, the participants are paid, we do not have financial motives in this

experiment. The use of non-financial motivations is not considered as a major difference

between two experiment as the psychological motives in a classroom experiment is enough for

the effort exertion (Jalava et al. (2015)). Moreover, there is no payment in the hypothetical

choices for the dictator in DWK experiment.

Our results confirm the robustness of Dana et al. (2007) s result with more selfish choice

in the transparent treatment and lower level of revealing the game in the hidden information

treatment. Tracking dictator s decision across two treatments shows the change in their

decision from transparent treatment to the hidden information treatment. Approximately

27% of the subjects choose the fair outcome in transparent treatment but decide to strategically

use the uncertainty in the hidden-information treatment to implement self-interested

outcome. These number of the subjects are higher than DWK s experiment. However, the

difference in result is not statistically significant.

Experimental Design:

In this section, we present the experimental design and the result of robustness test for

DWK s experiment. While DWK s design is between treatment, we have run within treatment

design to observe changes in the subjects choices over treatments. We use the second

treatment of DWK design adjusting monetary payoffs to the local money. The subjects are

a total of 97 undergraduate students from the department of economics at the University of

Tehran. The students participate in the experiment voluntarily at the end of their lectures.

There are five sessions in total, and at the beginning of each session, the instruction is presented

to participants.1 The subjects are not paid for their choices and at the end of the

experiment leave the experiment venue.

There was not any interaction between the subjects, and they played the game independently.

The subjects who were acting as dictators were told that there is another group

of students who are potential receivers. A role of either the dictator (Player A) and the

receiver(Player B) was assigned randomly to the subjects, and they were instructed that dictators

and receivers will not know each other even after the experiment to satisfy anonymity

condition. The dictators decide the allocation of payoffs to her own and the receiver. We

made similar payoff matrices for the subjects by converting DWK s payoffs to Iran s currency

using approximated current exchange rate. Therefore, the proportion of payoffs are the same

as Dana et al. (2007) s experiment.

Treatments and Conditions:

Baseline Treatment

Our baseline treatment is the same as DWK s baseline treatment with the difference

that we did not use the computer interface and used pen and paper.The subjects were told

that they would participate in a dictator game with an anonymous receiver who is in the

other room and they will never know each other. At the beginning of the experiment,

the participants were instructed how to play the dictator game. Also, the subjects were

given the payoff matrix below and explained how the payoffs for the dictators and receivers

are determined. Then, the dictators are asked to make the decision either choice A or B.

The payoff matrices are the same as Dana et al. (2007) experiment and only payoffs were

converted to Iran s currency.

In the transparent treatment, the payoffs are public information for both dictators and

receivers. Therefore, the dictator has complete information on the consequence of his action

and the receiver knows that the dictator is the sole cause of her payoff received. As we see in

matrices, there is a conflict between the dictator and the receiver in matrix 1. In the matrix

one choice, A is the self-interested outcome for the dictator (Player X) as it ensures higher

payoff for her (The dictator gains 18 units, and the receiver receives three units). For the

choice B, both players receive the equal amount of 15 units of the currency. The dictators

sacrifice three units less than choice A, and the receiver(Player Y) receives 15 units. We see

that there is a conflict of interest in this game. The dictator has to sacrifice her payoff to

help the receiver.

Player X s Choice

A X:18000

Y:3000

B X:15000

Y:15000

Hidden-Information Treatment:

Hidden information treatment was run within treatment with pen and paper, otherwise,

is the same as second treatment in DWK s experiment. The instructions were given to the

subjects along with a sealed envelope which has the actual matrix of the game. The payoff

matrices below show the payoff matrix of the dictator game in hidden-information treatment.

The dictators face the payoff matrix that their payoff in presented but the receivers payoff

is unknown.

Table 1: *

Payoff Matrix in Hidden Information Treatment

Player X s Choice

A X:18000

Y:?

B X:15000

Y:?

If the subjects decide to see the actual payoff matrix, they need to open the envelope

and see the payoff matrix. The payoff matrix might be one of the matrices below. There is a

conflict of interest in the matrix 1, and this matrix is the same as in transparent treatment.

On the other hand, in matrix 2, the payoff for player X and Y are in line, and both of them

earn the higher payoff when choice A is chosen. The participants do not know which of

matrices is in the envelope before they open it, but they were told that the matrix had been

determined by a flip coin before the experiment starts. In both matrices, player X s payoff

is the same in both choices, but the player Y payoff varies.

Table 2: *

Matrix 1 Player X s Choice

A X:18000

Y:3000

B X:15000

Y:15000

The dictators had to decide if they want to see the full actual payoff table before they make their choice. If the dictators do not want to see the actual payoff table, they do not open

the envelope and choose without knowing the receiver s payoff ( the First table where the

payoff for player Y is unknown). This result suggests that the dictators strategically decide

to stay ignorant about the receiver s payoff. If the dictators decide to see the payoff matrix

with the receiver s payoff, then she opens the envelope, where she sees the appropriate matrix

(one of the matrices above), and choose either choice A or B. The order of the treatment are

swapped to avoid any order effect in the result. lt;/p>

Result:

Table 1 below shows the result of comparison between the transparent and the hidden

information treatment. Number and proportion of the subjects that decide to see the full

payoff matrix and the corresponding proportion in the transparent treatment are presented.

As it is shown in Table 1, 46 subjects out of 72 (64%) decided to choose outcome B which

has equal payoffs for dictator and receiver in the transparent treatment. Although this

percentage is 10 percent less than Dana et al. (2007) but the difference is not significantly

different(Fisher Exact Test, p = 0.5877).

In hidden information treatment, 40 subjects out of 72 (56%) decided to reveal the game

by opening the envelope. Although opening the envelope has no cost, only 56% decided to

see the actual payoff matrix. This number is the aggregate number of people who had chosen

either outcome A or B, in the transparent treatment. This result is the same as DWK.

In total in the hidden-information treatment, 32(44%) of the dictators preferred not to reveal and choose based solely on their payoff. From these 32 subjects, 18 had chosen

outcome A in baseline game, and 14 had chosen outcome B in the transparent treatment.

These 14 subjects(19.4%) are people who changed their behaviour from the transparent to

the hidden information treatment. This 19.4% percent are the subjects who decide to stay

ignorant and choose the self-interested outcome strategically.

Among subjects who had chosen outcome B in the transparent treatment (46), 32(70%)

decided to reveal. Among these subjects, 18 of them faced table 1(same as the transparent

treatment), and 13(72.2%) chose the fair outcome and 5(27%) of them preferred to choose

outcome A. This 27 percent are the subjects who use what DWK call Moral Wiggle Room'lt;/p>to implement self-interested outcome. 3

Table 2 presents the choices in the hidden-information treatment separated on the revealing

and actual payoff matrices that the subjects face. 60% of the subjects who face matrix 1

decided to reveal the payoff matrix. This percentage is not statistically different from DWK s

experiment (Fisher Exact test, p=0.5584). From these subjects who revealed and faced the payoff matrix 1, 42% chose outcome A. This percentage is higher than DWK experiment

(25%), but the difference is not statistically significant (Fisher Exact test, p=0.65). This

result suggests that in this experiment, more subjects revealed the actual payoff and chose

the selfish outcome.

The subjects who faced matrix 2, 16 subjects (25%) decided to reveal the game. 13

subjects (81%) of them chose outcome A. The difference with the result of DWK is not

significant (Fisher Exact test, p=0.54). Also, 11 subjects (68%) chose outcome A, although

they have not revealed the game. This percentage is also the same a DWK.

Table 1 shows the use of Moral Wiggle Room" by the subjects when the payoff matrix is

uncertain. Around 27% of the subjects who chose fair outcome in transparent treatment used

the excuse, either not to reveal the game or reveal the game and choose the self-interested

outcome, to implement self-interested choice.

Discussion: lt;/b>

In this experimental study, we investigate the robustness of Dana et al. (2007) s experiment

with two changes in the implementation of the experiment. Different from DWK s

experiment, we use the within-subject design to investigate the change in the subjects decision.

We were concerned that the between-treatment design does not capture the effect of

uncertainty in the payoff matrix. Within-treatment design can show the choices in transparent

and hidden information treatment and show the change in the choices across treatments for all subjects.

In particular, we examine if the change in the result of strategic ignorance is robust in

the within-subject treatment. Also, there is no financial payment for this experiment which

is the same as the hypothetical choices in DWK experiment but different from the dictator

choices.

We design an experiment to let the subjects to make their choices for the transparent

treatment and the hidden-information treatment sequentially. We could observe the subjects

choice and changes in their decision across two different treatment. We observed that, in

confirmation of DWK s result, a majority of the participants, whenever they get the chance,

decide to stay ignorant about their counterparts payoffs and choose the self-interested outcome.

From the subjects who had chosen outcome B (fair outcome) in the transparent treatment

(46 subjects), 14(30%) decided not to reveal the game. More importantly, 15 subjects

(32%) chose outcome A (self-interested outcome) in the hidden information treatment and

19 subjects(41%) implemented fair outcome. Percentage of fair choices in this experiment is

slightly more than Dana et al. (2007) s experiment which shows that more people stay with

their fair outcome in the baseline game.

In comparison with Dana et al. (2007) s experiment, we find a higher percentage of

the participants who choose self-interested outcome after revealing the actual payoff matrix

of their counterparts. This result suggests that within-subjects design sharpen the effect of

uncertainty in matched players payoff. We observed the change in the choices of the subjects

in around 27% of the participants.

Thus, we find that DWK s results are robust to the change in experimental design. Our

replication of the results of DWK put forward more question on the motivations of altruistic

and fair behaviour that are observed in the lab. Further experimental studies needed to

investigate the effect of situational pressure, self-image and other external factors on prosocial

behaviour. From theoretical points of view, a comprehensive theoretical model has to &explain such changes in the choices of subjects in the lab.

References

Bardsley, N. (2008). Dictator game giving: altruism or artefact? Experimental Economics,

11(2):122 133.

Camerer, C. (2003). Behavioral game theory: Experiments in strategic interaction. Princeton

University Press.

Dana, J., Roberto, W., and Jason, K. (2007). Exploiting moral wiggle room: experiments

demonstrating an illusory preference for fairness. Economic Theory, Springer, 33(1):67 80.

Hoffman, E., McCabe, K., Shachat, K., and Smith, V. (1994). Preferences, property rights,

and anonymity in bargaining games. Games and Economic behavior, 7(3):346 380.

Hoffman, E., McCabe, K., and Smith, V. L. (1996). Social distance and other-regarding

behavior in dictator games. The American Economic Review, 86(3):653 660.

Jalava, N., Joensen, J. S., and Pellas, E. (2015). Grades and rank: Impacts of non-financial

incentives on test performance. Journal of Economic Behavior Organization, 115:161

196.

List, J. A. (2007). On the interpretation of giving in dictator games. Journal of Political

economy, 115(3):482 493.