The effect of social interactions on the self serving bias presentation

25
The effect of social interactions on the self-serving bias Ilya Bartenev Jethro Elsden Monika Pompeo Gian Luca Tedeschi University of Nottingham Experimental Methods in Economics 29 April 2016 1 / 25

Transcript of The effect of social interactions on the self serving bias presentation

The effect of social interactions on the self-serving bias

Ilya Bartenev Jethro Elsden Monika Pompeo Gian Luca Tedeschi

University of Nottingham

Experimental Methods in Economics

29 April 2016

1 / 25

Outline

1 What is the self-serving bias?DefinitionMotivationMeasurement

2 Experimental designDesign frameworkMethods

3 ResultsSummaryAnalysis

4 ConclusionsRelevanceImprovementsSummary

2 / 25

Outline

1 What is the self-serving bias?DefinitionMotivationMeasurement

2 Experimental designDesign frameworkMethods

3 ResultsSummaryAnalysis

4 ConclusionsRelevanceImprovementsSummary

3 / 25

What is the self-serving bias?

Murphy’s Law

If more than one person is responsible for a miscalculation, no one will beat fault.

What is the self-serving bias?

Blaming others or external factors for poor results

Claiming individual credit for success

It is but one of many biases contributing to forming a positive self-image.Others include:

Overestimating ones ability relative to others

Downward comparison under pressure or threat

4 / 25

Motivation

Implications for economics:

Distorted self-image can lead to difficulties in bargaining and preventmarket clearing (Babcock and Loewenstein, 1997)

Unusually high rejection rates in ultimatum games (Knez andCamerer, 1995)

Effect on (positive and negative) reciprocity – the reverse is our focustoday

5 / 25

Measurement

How is self-serving bias measured? The most common pattern appears tobe:

Subjects are asked to complete a task as a group

Subjects receive feedback on the overall result

Subjects are asked to estimate their own contribution relative to thereported result

Studies using this method:

“Among friends. An examination of friendship and self-servingbias”(Campbell et al, 2000)

“The self-serving bias in a relational context”(Campbell et al, 1998)

Both set a creative task to be completed in a pair people of a varyingdegree of closeness and then record perceived own contribution.

6 / 25

Measurement

This is not the only possible way; for example, some other studies deduceself-serving bias from reactions or descriptions instead of measuring itdirectly.

For example, in “Hurting hurts more than helping helps: the role of theself-serving bias ”(Offerman, 2002) subjects are put into a reciprocitygame situation and their emotional reactions to intentional and randomhurtful or kind moves are noted.

7 / 25

Outline

1 What is the self-serving bias?DefinitionMotivationMeasurement

2 Experimental designDesign frameworkMethods

3 ResultsSummaryAnalysis

4 ConclusionsRelevanceImprovementsSummary

8 / 25

Design framework

Issues to take into consideration in the design:- Subjects knew each other- Positive and negative feedback- Subjects perceive (to some extent) how well they are performing

Limitations:- Limited time- Limited number of subjects- Limited experience of the experimenters

Some of the questions we faced:- Should we present the subjects with the game payoffs?- Should we include several rounds?- How should we incentivize the subjects to report their true beliefs?

9 / 25

Experimental design

10 / 25

Experimental design

Hot response game (Offerman, 2002): The couples start with no

endowment.The individual payoff isrepresented by the sum ofthe consequences of bothplayer’s actions.The first mover starts bymaking his choice, whilethe second mover observesthe first mover’s choiceand decide a response be-tween the given options.

11 / 25

Experimental design

Quiz:

Easy question Hard question

12 / 25

Outline

1 What is the self-serving bias?DefinitionMotivationMeasurement

2 Experimental designDesign frameworkMethods

3 ResultsSummaryAnalysis

4 ConclusionsRelevanceImprovementsSummary

13 / 25

Summary

Score distribution

14 / 25

Summary

Summary of feedback results

TreatmentDifficulty 1 2DifficultMean 3.5 3SD (1.29) (0.81)Easy

Mean 5.75 5.75SD (2.36) (2.21)

15 / 25

Summary

Summary of attribution results

TreatmentDifficulty 1 2DifficultMean 6.25 5.5SD (1.26) (3.51)Easy

Mean 5.5 3.75SD (3.51) (1.26)

16 / 25

Analysis

Attribution distribution by feedback

17 / 25

Analysis

Two-sample Wilcoxon rank-sum tests

Test feedback difference by difficulty: Significant difference at 5%(p=0.0141);

Test attribution difference by treatment: Significant difference at 10%(p=0.0722) if feedback>5, no significant difference (p=0.9150) iffeedback <5;

Test attribution difference by type of relation: No significantdifference between kind and unkind behaviour for both feedback>5(p=0.3173) and feedback <5 (p=0.6547);

18 / 25

Analysis

Difference-in-differences estimation of treatment effect on attribution

Variable Coefficient (Std. Err.)Feedback -0.090 (0.456)Treatment -1.847 (3.177)Treatment#Feedback 0.675 (0.645)Intercept 5.020∗∗ (2.194)

* p < 0.10, ** p < 0.05, *** p < 0.01

19 / 25

Analysis

Difference-in-differences estimation of interaction effect on attribution

Variable Coefficient (Std. Err.)Feedback 0.500 (1.068)Kind -6.800 (7.630)Unkind 2.250 (9.904)Kind#Feedback 0.700 (1.368)Unkind#Feedback -1.000 (2.388)Intercept 5.000 (4.776)

* p < 0.10, ** p < 0.05, *** p < 0.01

20 / 25

Outline

1 What is the self-serving bias?DefinitionMotivationMeasurement

2 Experimental designDesign frameworkMethods

3 ResultsSummaryAnalysis

4 ConclusionsRelevanceImprovementsSummary

21 / 25

Conclusions

Hypothetical links between SSB and reciprocity:

Offerman (2002): SSB → Reciprocity

Our study: SSB ← Reciprocity

Both directions may be correct.

22 / 25

Conclusions

Possible improvements:

More subjects, and unbiased sample (e.g. strangers rather thanclassmates)

Better separation of subjects to prevent communication.

Give subjects a copy of general instructions.

Multiple games/rounds, but need to consider learning by doing.

Investigate different forms of self-serving bias, i.e. blaming externalcauses rather than internal causes, division of pie with partnerinflating own share.

Reciprocity by real task instead of computerized one.

23 / 25

Conclusions

Summary:

Prior interaction seems to influence reported contributions.

But small sample prevents us from establishing the nature of therelationship.

Similar picture emerges when we focus on kind/unkind interaction.

The effects are not significant.

Bigger sample is needed.

24 / 25

THANK YOU

25 / 25