Living Up to Reputation: Cooperation in Public Good Experiments in - - PowerPoint PPT Presentation

Living Up to Reputation:

Cooperation in Public Good Experiments in Rural Kenyan Communities

Olga Rostapshova NBER/Social Impact Inc. October 19, 2013 SPI Annual conference 2013

Motivation

• In developing countries, voluntary contributions often

necessary to finance local public goods

• Observe high heterogeneity in:

▫ 1) incidence of successful cooperation ▫ 2) individual contributions

• Context:

▫ Western Kenyan villages contributing to local threshold public good ▫ 14 of 25 communities succeeded

Research Approach

• What factors and mechanisms are

responsible?

• Lab-in-the-field games + survey data
• Address questions hard to answer in lab
• Context specificity
• Measure cooperation and factors of influence

Context: rural Kenyan villages

• Small villages in rural Western Kenya

▫ Closed environment, little mobility ▫ Low income, low education ▫ Ethnically mostly homogeneous ▫ Strong social and family ties ▫ History of collective action

– Most belong to community groups (e.g. credit, church, burial) – Local norms, sanction institutions

Local public good: communal spring

Research questions

• What factors promote cooperative behavior in

local public goods financing?

(1) Social capital: norms and shared experiential history (2) Composition: proportion of residents with cooperative preferences and certain beliefs

• Examine roles of:

– Social accountability mechanisms

– Social capital/reputation – Inequality and fairness norms

– Preferences and beliefs – Asymmetric conditions

Research Approach

• Simulation via threshold public goods game (PPM)

▫ Symmetry/social accountability/fairness norms ▫ 4 experimental treatments: 1) Transparency: Anonymous vs. revealed

▫ Sensitivity to external social peer pressure

2) Restricted contributions: focal point removed 3) Asymmetrical endowments 4) Second chance/conditional contribution ▫ Anticipated & unanticipated

Game Structure: Threshold Public Good

• Subject pool

▫ 10 villages ▫ 20 randomly selected women (aged 18-50) / village

• Parameters

▫ 10 Players ▫ 10 Ksh Endowment / player / round ▫ 50 Ksh Threshold (50% of total E) ▫ 5 Ksh Focal, ”fair” contribution per player ▫ 100 Ksh Payout if threshold reached (equally divided: 10Ksh each)

• Rules:

▫ No refund or rebate ▫ Simultaneous play ▫ Single shot ▫ No feedback ▫ No communication ▫ No sanctions

• 4 Treatments: Within subject design

Model

• Anonymous game:
• Revealed game:
• First order conditions:

Where:

• ri=reputation
• Mi= non-monetary component of utility
• T = threshold
• Eit =endowment
• cit= contribution
• V/N = monetary value of the good
• pt ()= probability threshold will be reached

Pivotal areas of influence

Contribution distributions Belief distributions

Earnings of player i (πi) Total group contribution excluding i, C-i

Ci = E Ci = C*= E/2 Ci= 0

E .5E T-E T 1.5E ¡ 2E ¡

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Cumulaitve subjective probability Mean of group contribution excluding i, c-i

B1(C-i): optimistic beliefs B2(C-i): pessimistic beliefs B3(C-i): pivotal beliefs

E ¡ .5 ¡E ¡

Methodological contributions

1) Asymmetrical (no focal point) treatment

▫ Disallow symmetric focal point contributions to elicit preferences and sort players

2) Multi-stage conditional contributions

▫ Allow second chance to contribute if threshold is unmet: surprise and anticipated rounds

3) Develop experimental measure of social capital

▫ Leverage individual relationships within existing social networks, resulting reputation mechanisms ▫ Test external validity

Main Results

• Social accountability & reputation matter

▫ Transparency increases cooperation slightly, on average ▫ Low endowment players not responsive to transparency ▫ Many subjects give less under higher transparency to coordinate

• Inequality/fairness norms

▫ Asymmetric conditions produce more cooperation ▫ Heterogeneous endowments lead to higher contributions ▫ When “fair” contributions disallowed most subjects give more but some give much less

• Social capital

▫ Varies significantly by village

Successful coordination

• In majority of games and conditions, groups able

to cooperate: 82% met threshold

▫ But, much success attributable to the second-chance feature, conditional contributions instrumental for reaching threshold.

Heterogeneity in social capital

• Aggregate levels of cooperative social capital differ

drastically across villages

▫ Village 1 achieved cooperation every time, but village 10 only in 1/3rd of time

• Why?

▫ Leadership, cooperative history, local cultural factors ▫ Institutional and historical processes, interplay of civic, sociopolitical and economic factors, influence characteristics of social relationship set

Transparency: Slightly increases contributions

Distribution of anonymous and revealed contributions

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 1 2 3 4 5 6 7 8 9 10

Cumulative Distribution (%) Contribution (Ksh)

Anonymous Revealed 0% 5% 10% 15% 20% 25% 30% 35% 40% 1 2 3 4 5 6 7 8 9 10

Frequency (%) Contribution (Ksh)

Anonymous Revealed Revealed contributions are higher

Transparency: Many give more in secret

• Motivation: Coordination
• Pessimistic prediction of others’ contributions + elevated importance of
• wn contribution in secret round

Distribution of Difference in Revealed vs. Anonymous Contributions

5 10 15 20 25 30

• 7 -6 -5 -4 -3 -2
• 1

1 2 3 4 5 6 7

Frequency (%) Revealed - Anonymous Contribution

10 20 30 40 50 60 70 80 90 100

• 7
• 6
• 5
• 4 -3
• 2
• 1

1 2 3 4 5 6 7

Cumulative Frequency (%) Revealed - Anonymous Contribution

Don’t change contribution Give much more in anonymous Give much more in revealed 30% give more in anonymous Median unchanged

Beliefs: what will others do?

• Subjects believe others will give far more in the

revealed than in anonymous rounds

▫ Expectation of deviation of avg. group contribution from focal point:

– Anonymous: .07 (3.09) – Revealed: 2.61 (3.45)

Belief Distribution

3% 3% 2% 3% 23% 5% 22% 16% 22% 0% 5% 10% 15% 20% 25% 30% 35% 40% 45%

Undercontribute Focal Overcontribute Revealed game belief (%) Anonymous game belief Overcontribute Focal Undercontribute

28% 42%$ 30%$

Sorting by beliefs/preferences

• Lack of anonymity has a heterogeneous impact on

contributions, allowing sorting of subject pool by preference and belief profiles

• Restricting contributions by removing the focal point

effectively separates players pooled at the focal point.

Distribution of player types

Player types revealed by anonymous vs. revealed contributions relative to focal point

Anonymous game Revealed game Less than Focal Focal More than Focal Total Less than Focal 23% 10% 8% 40% (Free-riders) (Reputation insensitive) (Reputation insensitive) Focal 11% 15% 11% 37% (Reputation sensitive) (Focalists) (Reputation insensitive cooperators) More than Focal 3% 7% 13% 23% (Reputation sensitive) (Reputation sensitive) (Full cooperators) Total 36% 31% 32% 100%

Heterogeneity in reputation sensitivity

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

Free-riders Full cooperators

Reputation insensitive Reputation indifferent Reputation sensitive

Distribution of cooperators and free-riders by reputation sensitivity

Sorting: Under and Over-contributors

Anonymous Contributions: Unrestricted vs. Restricted Revealed Contributions: Unrestricted vs. Restricted

0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 1 2 3 4 5 6 7 8 9 10 Unrestricted Restricted 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 1 2 3 4 5 6 7 8 9 10 Unrestricted Restricted

Contribution Contribution

Removing the focal point

Sorting: Focal Contributors

• Focal contributors: players giving 5Ksh when unrestricted
• 1/3rd of players
• In “restricted/no focal point” round contribute on average:
• Secret:

5.01 Ksh

• Revealed: 5.2 Ksh

0% 10% 20% 30% 40% 50% 5 10

Frequency (%) Contribution in restricted round, no focal point (ci)

Anonymous Revealed 0% 25% 50% 75% 100% 5 10

Cumulative frequency 9%) Contribution in restrictred round, no focal point (ci)

Anonymous Revealed

Sorting of focalists

Anonymous game Revealed game Less than Focal More than Focal Total Less than Focal 27% 26% 52% (Focal free-riders) (Focal reputation insensitive) More than Focal 18% 30% 48% (Focal reputation sensitive) (Focal cooperators) Total 45% 55% 100%

Revealed in separating restricted treatment

Fairness norms

• Under asymmetric endowments

▫ Probability of success rises from 30 to 89% ▫ “Poor” subjects will give larger percentage of their endowment ▫ “Rich” players give more when giving revealed

Amount ¡Given ¡ % ¡Given ¡ Endowment ¡ Anonymous ¡ Revealed ¡ Anonymous ¡ Revealed ¡ Diff ¡ Diff% ¡ Z-­‑stat ¡ P ¡ 5 ¡(Poor) ¡ 3.24 ¡ 3.24 ¡ 65% ¡ 65% ¡ 0% ¡ 0% ¡

• ­‑0.04 ¡

97% ¡ 15 ¡(Rich) ¡ ¡ 7.58 ¡ 8.05 ¡ 51% ¡ 54% ¡ 3% ¡ 6%* ¡

• ­‑1.92 ¡

6% ¡

Individual heterogeneity

• Player contributions vary with individual

characteristics

• Contributions increase in

▫ Ethnic minority ▫ Monogamous marriage ▫ No wage job ▫ House with iron roof (wealth proxy) ▫ Higher trust (weakly)

Reputation scores

• Reputation score as within-context reputation measure
• Higher reputation players tend to be:

– Older – More educated – Have a wage job – Belong to community groups – Less trusting

5 10 15 20 25 30 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 21 23 24 27 32 38

Number of subjects Reputation score

Second chances

• Second chance conditional contributions:

▫ Subjective beliefs about others’ play only affect revealed second chance conditional contributions ▫ Players giving less in 1st round give significantly more in revealed game ▫ More conditional contributions come from under-contributors, equalizing the distribution of contributions to the public account across the players

More second chances

• Surprise 2nd round allowed almost all groups to reach the

threshold: 57% of groups only able to cooperate with help of 2nd chance

• When two-stage play is anticipated, unconditional

contributions drop significantly relative to surprise two- stage play; conditional contributions remain the same => total contributions in anticipated 2-stage rounds are lower

Next Steps

• Match game results to actual public good contributions
• Replicate in lab with American population to test for inter-

cultural differences?