Inequality, Redistribution, and Discrimination:

An Experiment

Brandon Williams

Applied Micro Brown Bag

April 21, 2025

Motivation: Some Facts

• 71% of the population live in country where wealth inequality has risen since 1990
• The source of inequality matters for preferences about redistribution (Young, 1958; Fong, 2001; Alesina and Angelitos, 2005; Andre, 2024)
  • Some accept all forms of inequality
  • Some accept inequality only if it is earned
  • Some do not accept any form of inequality
• A considerable amount of literature has used spectator designs to study the share of these types across country, political views, and gender (Cappelen et al., 2007; Almas et al., 2020; Almas et al., 2024)
• Preferences about inequality are important because they strongly correlate with preferences about policy

Types Around the World

Motivation: Some More Facts

• Discrimination and in-group bias are real phenomena (Becker, 1958; Tajfel et al., 1971; Chen and Li, 2009)
• Social desirability bias can lead to attenuated estimates of discriminatory intent in experiments (Kuklinski et al., 1997a, Kuklinski et al., 1997b)
• Evidence of discrimination tends to emerge more strongly when it can be hidden by other possible explanations, like list randomization techniques (e.g. Coffman et al., 2013)
• Or when observation is not apparent
  • Correspondence resume studies (Bertrand and Mullainathan, 2004)
  • Choice of flowers to allocate to downstream workers (Hjort, 2013)

Research Question

• Are redistributive preferences (partially) determined by who will benefit from redistribution?

Research Question

• Are redistributive preferences (partially) determined by who will benefit from redistribution?

Research Question

• Are redistributive preferences (partially) determined by who will benefit from redistribution?
  • Preferences about implemented inequality

Research Question

• Are redistributive preferences (partially) determined by who will benefit from redistribution?
  • Preferences about implemented inequality
  • Social in- and out-group, focusing on ethnicity

Research Question

• Are redistributive preferences (partially) determined by who will benefit from redistribution?
  • Preferences about implemented inequality
  • Social in- and out-group, focusing on ethnicity
  • Under-performing or unlucky in-group benefits from redistribution; advantaged in-group benefits from "libertarian" outcomes

Research Question

• Are redistributive preferences (partially) determined by who will benefit from redistribution?
  • Preferences about implemented inequality
  • Social in- and out-group, focusing on ethnicity
  • Under-performing or unlucky in-group benefits from redistribution; advantaged in-group benefits from "libertarian" outcomes

•  
  • What is the function of in-group relationship that informs one's willingness to redistribute?

Research Question

• Are redistributive preferences (partially) determined by who will benefit from redistribution?
  • Preferences about implemented inequality
  • Social in- and out-group, focusing on ethnicity
  • Under-performing or unlucky in-group benefits from redistribution; advantaged in-group benefits from "libertarian" outcomes

•  
  • What is the function of in-group relationship that informs one's willingness to redistribute?

Research Question

• Are redistributive preferences (partially) determined by who will benefit from redistribution?
  • Preferences about implemented inequality
  • Social in- and out-group, focusing on ethnicity
  • Under-performing or unlucky in-group benefits from redistribution; advantaged in-group benefits from "libertarian" outcomes

•  
  • What is the function of in-group relationship that informs one's willingness to redistribute?

% share disadvantaged co-ethnic group

Research Question

• Are redistributive preferences (partially) determined by who will benefit from redistribution?
  • Preferences about implemented inequality
  • Social in- and out-group, focusing on ethnicity
  • Under-performing or unlucky in-group benefits from redistribution; advantaged in-group benefits from "libertarian" outcomes

•  
  • What is the function of in-group relationship that informs one's willingness to redistribute?

% share disadvantaged co-ethnic group

Research Question

• Are redistributive preferences (partially) determined by who will benefit from redistribution?
  • Preferences about implemented inequality
  • Social in- and out-group, focusing on ethnicity
  • Under-performing or unlucky in-group benefits from redistribution; advantaged in-group benefits from "libertarian" outcomes

•  
  • What is the function of in-group relationship that informs one's willingness to redistribute?

% share disadvantaged co-ethnic group

An Ideal Experiment...

• Have workers complete a task with an equivalent ex-ante expected outcome by group
• Identify "blind" redistribution preferences of spectators
• Have clear, salient and preferably binary in- and out-groups
• Natural variation in the distribution of groups that is well-known by decision-makers
• Choices in redistribution unbiased by social desirability (e.g. a plausible "out")
• Leverage the changing distribution to test the slope of redistribution probability function
• Match the blind elicitation to the corresponding point in the redistribution function

Overall Design

• Lab-in-the-field design using a worker-spectator redistribution paradigm
  • Control treatment: generic workers
  • Information treatment: reveal location of workers
• Leverage Fiji as "technology":
  • Stark and salient in- and out-group differences between ethnic groups
  • Wide-ranging geographical variation in group distribution
• Measure within-group preferences about redistribution
• Measure shift in group preferences about redistribution
  • Redistribution towards in-group with unfavorable allocation
  • No/limited redistribution with favorable allocation

Experimental Paradigm

• From Almas et al. (2024), two types of participants:
  • Workers - complete task for possible payment
  • Spectators - can decide on possible redistribution of payment

• The task requires little skill but real effort (e.g. ball-catching task, counting zeros)

Experimental Paradigm

• One worker gets initial allocation according to treatment
  • Luck - a fair coin determines who is paid the whole amount
  • Merit - the more productive worker is paid the whole amount

$

$

$

$

$

$

Experimental Paradigm

• Spectators know the source of the allocation
• Workers are not informed of the initial distribution, but are aware a spectator will have final say

$

$

$

$

$

$

Experimental Paradigm

• Before the workers are paid, the spectator can choose to redistribute: none

$

$

$

$

$

$

Experimental Paradigm

• Before the workers are paid, the spectator can choose to redistribute: none, some,

$

$

$

$

$

$

Experimental Paradigm

• Before the workers are paid, the spectator can choose to redistribute: none, some, or all

$

$

$

$

$

$

Experimental Paradigm

• Spectators can be classified by their choices over redistribution depending on treatment

$

$

$

$

$

$

Experimental Paradigm

• Spectators can be classified by their choices over redistribution depending on treatment

$

$

$

$

$

$

Experimental Paradigm

• Spectators can be classified by their choices over redistribution depending on treatment

$

$

$

$

$

$

Experimental Paradigm

• Spectators can be classified by their choices over redistribution depending on treatment

$

$

$

$

$

$

Meritocrat*

Experimental Paradigm

• Spectators can be classified by their choices over redistribution depending on treatment

$

$

$

$

$

$

Experimental Paradigm

• Spectators can be classified by their choices over redistribution depending on treatment

$

$

$

$

$

$

Egalitarian

Experimental Paradigm

• Spectators can be classified by their choices over redistribution depending on treatment

$

$

$

$

$

$

Experimental Paradigm

• Spectators can be classified by their choices over redistribution depending on treatment

$

$

$

$

$

$

Experimental Paradigm

• Spectators can be classified by their choices over redistribution depending on treatment

$

$

$

$

$

$

Libertarian

Experimental Paradigm

• Spectators can be classified by their choices over redistribution depending on treatment

$

$

$

$

$

$

Experimental Paradigm

• Spectators can be classified by their choices over redistribution depending on treatment

$

$

$

$

$

$

Egalitarian*

Experimental Paradigm

• Spectators can be classified by their choices over redistribution depending on treatment
  • Egalitarians - share dividing equally in the merit treatment
  • Libertarians - share allocating everything to the lucky worker in the luck treatment
  • Meritocrats - difference between the share of participants allocating more to the productive worker in the merit treatment and the share allocating more to the lucky worker in the luck treatment
  • Other - share redistributing in other ways

Experimental Paradigm

Experimental Paradigm

Experimental Paradigm

• Under "blind" elicitation, spectator should choose their optimal preference for inequality, but may be informed by assumptions about who would benefit
• Hypothesis/claims to be tested:
  • Introducing information about workers changes the redistributive choice towards a preferred group
  • Suddenly egalitarians look like meritocrats (or vice versa) when their choice benefits a certain type of worker
  • "Real" policy preferences exist with knowledge and inference about who will benefit from redistribution schemes
  • It is vital to know if inequality preferences are robust or subject to the context of who benefits

Context: Fiji

• Island country in Melanesia north of New Zealand
• Consists of 330 islands, with about 100 inhabited, most of the population (87%) lives on one of two islands

• 1874: Fiji as a British colony
• A catastrophic measles outbreak (mortality rate of 540 out of 1000 workers) and rebellion led to limited labor force
• Beginning in 1878, the UK began bringing indentured "laborers" from India to work, eventually over 60,000
• Today, 35% of the population is Indo-Fijian, with 60% indigenous iTaukei

Context: Fiji

Context: Fiji

• Meanwhile "Fiji is for the Fijians" was adopted as official land ownership policy
  • Traditional chiefs granted control over governing
  • Most land could not be sold, only leased
• Since independence in 1970, "Two Fijis" is the norm
  • iTaukei own land rights and receive rent payments
    • Redistributive scheme within tribes
    • Often iTaukei do not work the land they own, or only subsistence farm on small percentages of it
  • Indo-Fijians largely comprise the cash crop production and urban business development
    • Lower poverty rate (20% vs 36%)
    • Generate considerable economic activity through land use

Context: Fiji

• Cultural differences are extremely prevalent (and salient):
  • iTaukei have a stronger social identity, in-group identification, and are more collectivistic than Indo-Fijians
  • iTaukei more likely to be Christians, while Indo-Fijians more likely to be Hindus or Muslims


• Geographic differences are also clearly defined:
  • iTaukei: native land under traditional clan structure
  • Indo-Fijian: urban centers and towns


• Political strength largely vested in iTaukei (esp. following 2000 coup), but economic strength with Indo-Fijians

Context: Fiji

Where do the iTaukei live?

Experimental Procedure

• Standard spectator game
• Each with two sub-treatments for determining types:
  • Luck
  • Merit
• Two experimental treatments:
  • Blind treatment - no information about the workers provided to the spectator
  • Information - location information about each worker, such that ethnic group of worker is readily inferred
• N = 400, balanced on iTaukei and Indo-Fijian split
• Collect their justification for making redistributive choice

Empirical Specification

• There are two key measures from the spectator game:                    
  1. The share of each type (as defined before)
  2. Implemented inequality:

Empirical Specification

• There are two key measures from the spectator game:                    
  1. The share of each type (as defined before)
  2. Implemented inequality:

• In the information treatment this becomes:                                      

Hypothesis 1: the collectivist tradition of the iTaukei and the redistribution scheme for rents will lead to more egalitarian views

• Hypothesis 1a: there is more inequality implementation by Indo-Fijian spectators than iTaukei spectators in blind treatments

• Hypothesis 1b: there are fewer Indo-Fijian egalitarians than iTaukei egalitarians

Hypotheses

Hypotheses

Additional Things I'm Thinking About

• Adding a treatment with specific details about the workers to test direct pathway to in-group preference
• Developing a meaningful / useful task for the workers, and determining logistically how to pay the workers
• Repeating the task within-subject to determine if spectators attenuate the bias shift in order to stay consistent
  • For example, going from blind in first rounds to geographic in later rounds
  • Or, seeing the immediate beneficiaries of payment switch from in-group to out-group

Conclusion

• Understand how in- and out-group differences affect inequality preferences
  • E.g. to what degree does who benefits from redistribution change those preferences?
• Leverage the stark and salient binary ethnic groups in Fiji
  • Natural variation in geographical location allows for inferring co-ethnic beneficiaries with minimal social desirability bias
• Experimental procedure that uses an established paradigm and introduces an in- and out-group dimension with testable implications

Thank you!

Preferences about inequality and policy

Norway and the US

Power Simulation