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: • r i =reputation • Mi= non-monetary component of utility • T = threshold • E it =endowment • c it = contribution • V/N = monetary value of the good • p t ()= probability threshold will be reached

Pivotal areas of influence Contribution distributions Belief distributions 1 2E ¡ 0.9 Cumulaitve subjective probability 0.8 1.5E ¡ 0.7 B1(C-i): optimistic beliefs Earnings of player i ( π i) 0.6 B2(C-i): pessimistic beliefs E 0.5 B3(C-i): pivotal beliefs 0.4 0.3 .5E Ci = E 0.2 Ci = C*= E/2 0.1 Ci= 0 0 T-E T .5 ¡E ¡ E ¡ Mean of group contribution excluding i, c-i Total group contribution excluding i, C -i

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 1/3 rd 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 40% Anonymous 100% 35% Revealed 90% Cumulative Distribution (%) 30% 80% Frequency (%) 70% 25% Revealed contributions 60% 20% are higher 50% 15% 40% Anonymous 30% 10% Revealed 20% 5% 10% 0% 0% 0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10 Contribution (Ksh) Contribution (Ksh)

Transparency: Many give more in secret Distribution of Difference in Revealed vs. Anonymous Contributions 100 30 Frequency (%) Cumulative Frequency (%) Don’t change contribution 90 25 80 70 20 60 Median unchanged 50 15 Give much more Give much more 40 in anonymous in revealed 10 30% give more in anonymous 30 20 5 10 0 0 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 Revealed - Anonymous Contribution Revealed - Anonymous Contribution • Motivation: Coordination • Pessimistic prediction of others’ contributions + elevated importance of own contribution in secret round

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 45% 42%$ Overcontribute Focal 40% Undercontribute 35% 16% Revealed game belief (%) 30%$ 28% 30% 25% 20% 22% 22% 15% 23% 10% 5% 5% 3% 3% 3% 2% 0% Undercontribute Focal Overcontribute Anonymous game belief

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 insensitive (Reputation sensitive) (Focalists) cooperators) More than Focal 3% 7% 13% 23% (Reputation sensitive) (Reputation sensitive) (Full cooperators) Total 36% 31% 32% 100%

Heterogeneity in reputation sensitivity Distribution of cooperators and free-riders by reputation sensitivity 100% 90% 80% 70% 60% Reputation insensitive 50% Reputation indifferent 40% Reputation sensitive 30% 20% 10% 0% Free-riders Full cooperators

Sorting: Under and Over-contributors Removing the focal point 45% 45% 40% 40% 35% 35% Unrestricted 30% Unrestricted 30% Restricted 25% Restricted 25% 20% 20% 15% 15% 10% 10% 5% 5% 0% 0% 0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10 Contribution Contribution Anonymous Contributions: Unrestricted vs. Restricted Revealed Contributions: Unrestricted vs. Restricted

Sorting: Focal Contributors • Focal contributors: players giving 5Ksh when unrestricted • 1/3 rd of players • In “restricted/no focal point” round contribute on average: • Secret: 5.01 Ksh • Revealed: 5.2 Ksh 50% 100% Anonymous Revealed Cumulative frequency 9%) 40% 75% Frequency (%) 30% 50% 20% 25% 10% Anonymous Revealed 0% 0% 0 5 10 0 5 10 Contribution in restrictred round, no focal point (c i ) Contribution in restricted round, no focal point (c i )

Sorting of focalists Revealed in separating restricted treatment Anonymous game Less than Focal More than Focal Total Revealed game 27% 26% 52% Less than Focal (Focal free-riders) (Focal reputation insensitive) 18% 30% 48% More than Focal (Focal reputation sensitive) (Focal cooperators) 45% 55% 100% Total

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% ¡