Norms of Distributive Justice in Rural Malawi Annika Mueller Harvard University amueller@fas.harvard.edu 2012 World Bank Conference on Equity
Two-Part Study Research Questions Part 1 Which norms of distributive justice do subjects in central Malawi subscribe to? How are those preferences distributed within the study population? Do demographic and socio-economic characteristics serve as good predictors for fairness preferences?
Motivation (Part 1) Culture as Determinant of Economic Growth (Weber, 1904): Anthropologists and experimental economists have documented other-regarding preferences for many cultures all over the world.
Motivation (Part 1) cont d In Malawi, as in other developing countries, these types of preferences are of special importance. Norms of distributive justice translate into redistributive practices: Set incentives or disincentives for production Co-determine pace of economic development
Literature (Part 1) Pamela Jakiela (2011): Lab-in-the-field experiment on whether effort is rewarded in rural Kenya and the USA Cappelen et al. (2007): Lab experiment on pluralism of fairness ideals in Scandinavia
Key Features One-shot two-player dictator games Effort generated pie/social surplus to which multiple norms of distributive justice and associated sharing rules may apply Analysis of relative weight on specific sharing norms and selfishness Random sample of population in a developing country
Preview of Main Findings (Part 1) Equality based norms have largest influence on sharing behavior Behavior according to contribution based norms also widely observable Moreover, individuals take equality of opportunity to exert effort and luck into account when distributing social surplus
Two-Part Study Research Questions Part 2: Are the norms subjects subscribe to fully internalized by the specific population or domain-specific to a particular informational environment?
Motivation (Part 2) If norms are domain-specific to a particular. informational environment changes in community structures are arguably leading to more abrupt changes in sharing behavior
Background Literature Review (Part 2) Ultimatum game studies with incomplete information treatments For example Mitzkewitz and Nagel (1993), Gueth et al. (1996) Pie size can vary With incomplete information about pie size, responder cannot be sure about how to evaluate a small offer in terms of fairness of the proposer
Literature Review (Part 2) cont d Proposers pretend to be what is perceived as fair by many for the small pie size if the responder does not know that the pie is de facto large In line with outcome-based models of social preferences Fehr and Schmidt (1999) Bolton and Ockenfels (2000)
Literature Review (Part 2) cont d Additional explanations must be considered: Dictator game experiments, where subjects act more selfish when able to conceal or delegate their choices or if they can stay ignorant about the allocation to the recipient For example Dana et al. (2007), Andreoni and Bernheim (2009), Grossmann (2010), Hamman et al. (2010), Matthey and Regner (2010)
Literature Review (Part 2) cont d Concern with what others believe Some people like to be perceived as fair Andreoni and Bernheim (2009), Grossman (2010) Some people feel guilty if their behavior falls short of others expectations Battigalli and Dufwenberg (2007) Ockenfels and Werner (2012) find hiding behind a small cake -effect in dictator games conducted with newspaper readers
Key Features (Part 2) One-shot two-player dictator games Receivers knew or did not know size of the pie Analysis of relative importance of specific sharing norms, selfishness, and the information effect on observed behavior
Preview of Main Findings (Part 2) Dictators act more selfishly under incomplete information, i.e. in situations where their image will not be damaged by doing so Effect is small relative to influence of norms of distributive justice on sharing behavior
Principles of Distributive Justice Interpreted for: Two-player one-shot dictator games with production stage, where Both players contribute to a common pie A real effort task generates income
Norms of Distributive Justice Strict Egalitarianism Equal distribution of the common pie Libertarianism Individuals rewarded in proportion to their contribution Liberal Egalitarianism Individuals rewarded in proportion to effort exerted Inequity Aversion Equalizing final payoffs
Experimental Design Benchmark Treatment and Equality of Opportunity Treatments without Effort Cap Endowment: 30 MK 2 bags of beans Income Shock Treatments Endowment: 30 MK / 0 MK 2 bags of beans
Experimental Design cont d Equality of Opportunity Treatments with Effort Cap Endowment: 50 MK 1 bag of beans
Experimental Design cont d Incomplete Information Treatment Rate of Return of Dictators not Revealed to Receivers Known to Both Players Prior to any Decision Making
Data Data collection took part in July and August 2010 in Ntchisi District, Central Region, Malawi Data consists of baseline, game outcomes 30+ rounds of piloting 1280 subjects enrolled in the actual study One round of data collection in each location over a period of 20 days Random sample
Incentives Token gift worth approximately 30 MK Between 0 MK up to 350 MK (2.18 US-Dollars) depending on the outcome of the experiment In 2005, 46.7 percent of the population lived at or below 44.29 MK per day, with 16.1 percent classified as ultrapoor" according to the World Bank (2007) Data collected in the dry season: characterized by especially low, constant opportunity costs for participants Goldberg (2010) found that over 70 percent of her sample of 529 subjects chose to do hard physical labor at a rate of only 30 MK per day during the dry season
Sorting Decisions Of those who were given 2 bags 27.25 percent chose to sort 0 bags 44.63 percent chose to sort 1 bag 28.13 percent chose to sort 2 bags Of those who were given 1 bag 34.06 percent chose to sort 0 bags 65.94 percent chose to sort 1 bag
Treatment Effects on Sorting Decisions Low vs. high rate of return: Wilcoxon-Mann-Whitney test: p~0.48 Info vs. No-Info: Wilcoxon-Mann-Whitney test: p~0.61 Info vs. No-Info for *high rate of return*: Wilcoxon-Mann-Whitney test: p~0.32 Effort Cap vs. No Effort Cap Test of Proportions: p-value~0.017
Empirical Strategy Determining influence of experimental variations and effort choices on decision making Assigning observations and individuals to norms Regression analysis Mixed Logit Conditional Logit (McFadden s Choice Model)
Shares of Common Pot Kept by Dictators (Strategy Method)
Shares of Common Pot Kept by Dictators for Dictators Exerting More Effort
Shares of Common Pot Kept by Dictators for Dictators Exerting Less Effort
Ranksum Tests Experimental Variations Different Levels of Effort, Same Rates of Return Dictator: p=0.0000 Receiver: p=0.0000 Different Rates of Return, Same Effort Dictator: p=0.0010 Receiver: p=0.6920 Different Contributions Dictator: p=0.0000 Receiver: p=0.0000
Findings Subjects respond as intuitively expected to all variations for which at least some norms predict changes of the optimal share Norms seem to be relevant in describing behavior of subject population
Regressions of Dictator Shares on Model Variables (1) (2) (3) (4) Receiver's Effort (Measured -0.112*** -0.112*** in Number of Bags Sorted) (0.005) (0.005) Difference in Rates of 0.008** 0.008*** Return (0.003) (0.003) Difference in Contributions 0.013*** (Measured in Bottle Caps) (0.001) Effort Cap -0.034** 0.029* -0.010-0.033*** (Receiver) (0.015) (0.016) (0.015) (0.015) Effort Cap 0.020 0.030** 0.028** 0.021 (Dictator) (0.013) (0.014) (0.014) (0.013) Effort Cap -0.017 0.051*** 0.007-0.017 (Both Players) (0.015) (0.016) (0.015) (0.015)
Income Shock -0.055*** -0.052*** -0.050*** -0.055*** (Receiver) (0.013) (0.014) (0.013) (0.013) Income Shock 0.019 0.020 0.021 0.019 (Dictator) (0.013) (0.014) (0.013) (0.013) Income Shock -0.015-0.012-0.016-0.016 (Both Players) (0.013) (0.014) (0.013) (0.013) Constant 0.694*** 0.540*** 0.641*** 0.695** (0.013) (0.012) (0.013) (0.013) Num Obs 2751 2751 2751 2751 R-Squared 0.24 0.10 0.20 0.24 Demographic and Socio-Economic Controls YES YES YES YES
Findings Evidence for norm-adherent behavior in OLS framework Confirmation that absent to theory effects hold despite inclusion of controls
Hand Matching Strong vs. weak matches Unconditional vs. exclusive matching Observation vs. individual level
Tables of Norms at Observation Level for Strong Match Unconditional Matches of Preferences to Norms
Findings Decision makers mainly guided by strict egalitarianism libertarianism liberal egalitarianism Strict egalitarianism followed even under exclusive matching Even under strong matching high level of self-consistency, but scope for estimating mixed logit
OLS Regression of Matched Norms on Baseline Variables Unconditional Weak Matches of Preferences to Norms
Findings Socio-economic and demographic predictors Limited explanatory power but overall intuitive Outcomes vary mainly with experimental variations and effort choices of subjects
Mixed logit Allows for random taste variation Allows for correlation in unobserved factors over time Model is defined on the basis of the functional form for its choice probabilities
Conditional Logit (McFadden s Choice Model)
Preliminary Results Equality based norms have largest influence on sharing behavior Behavior according to contribution based and effort based norms widely observable Moreover, individuals take equality of opportunity to exert effort and luck into account when distributing social surplus Socio-economic and demographic variables have little explanatory power regarding norm adherence
Ranksum Tests Informational Environment High Rate of Return Dictators: p~0.67 + Receivers Sorted: p~0.035 Low Rate of Return Dictators: p~0.47 + Receivers Sorted: p~0.55
Finding Subjects respond as intuitively expected to informational variation Conditional on receivers sorting a positive amount No theoretical prior Matches observations in the field
Regressions of Dictator Shares on Model Variables
Results Dictators act more selfishly if they can hide part of the social surplus However, this effect seems small compared to the effects the norms of distributive justice have on decision making Hence, changes in community structures are likely to lead to abrupt but by no means large changes to sharing behavior if we believe the results not to be exceptionally domain-specific
Thank you!
Dictator Shares (Strategy Method)
Dictator Shares (Strategy Method)
Dictator Shares (Strategy Method)
Dictator Shares (Strategy Method)
Dictator Shares (Strategy Method)
Dictator Shares (Strategy Method)
Dictator Shares (Strategy Method)
Findings: 35 percent of the participants assign little importance to fairness considerations (no interior maximum in their choice problem) 30 percent choose intermediate solution 35 percent of the participants act mostly in line with their view of fairness: 43.5 percent strict egalitarians 38.1 percent liberal egalitarians 18.4 percent libertarians
Tables of Norms at Observation Level for Strong Match Unconditional Matches of Preferences to Norms Exclusive Matches of Preferences to Norms
Tables of Norms at Observation Level for Weak Match Unconditional Matches of Preferences to Norms
Tables of Norms at Observation Level for Weak Match Unconditional Matches of Preferences to Norms Exclusive Matches of Preferences to Norms
Tables of Norms at Person Level for Strong Match Unconditional Matches of Preferences to Norms
Tables of Norms at Person Level for Strong Match Unconditional Matches of Preferences to Norms Exclusive Matches of Preferences to Norms
Tables of Norms at Person Level for Weak Match Unconditional Matches of Preferences to Norms Exclusive Matches of Preferences to Norms
Tables of Norms at Person Level for Weak Match Unconditional Matches of Preferences to Norms