Preferences for affirmative action policies

Last registered on March 14, 2024

Pre-Trial

Trial Information

General Information

Title
Preferences for affirmative action policies
RCT ID
AEARCTR-0007131
Initial registration date
May 23, 2021

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published
May 24, 2021, 8:54 AM EDT

First published corresponds to when the trial was first made public on the Registry after being reviewed.

Last updated
March 14, 2024, 5:57 AM EDT

Last updated is the most recent time when changes to the trial's registration were published.

Locations

Region

Primary Investigator

Affiliation
University of Duesseldorf

Other Primary Investigator(s)

PI Affiliation
University of Duesseldorf
PI Affiliation
University of Duesseldorf
PI Affiliation
University of Duesseldorf

Additional Trial Information

Status
On going
Start date
2021-05-25
End date
2024-12-31
Secondary IDs
Prior work
This trial does not extend or rely on any prior RCTs.
Abstract
Support for affirmative action policies is typically higher among those who would be favored by them. There are several possible explanations for this gap. One explanation are the different experiences these groups have made, for example in the workplace, while another is that opinions on affirmative action are influenced by in-group favoritism and/or self-serving motives.
We test these possible explanations in a stylized environment that allows us to vary the stakes individuals have in the decision in favor of or against an affirmative action policy. For this purpose, we conduct an online experiment using a US sample provided by Prolific that is representative with respect to age, gender, and ethnicity. The experiment includes a tournament where half of the tournament participants are randomly assigned to a disadvantaged group. Subjects decide whether or not they prefer a quota rule compensating for that disadvantage to be used in determining the tournament winners.Three different treatments allow us to disentangle impartial preferences for affirmative action from other motives like in-group favoritism and self-interest.

External Link(s)

Registration Citation

Citation
Herzog, Sabrina et al. 2024. "Preferences for affirmative action policies." AEA RCT Registry. March 14. https://doi.org/10.1257/rct.7131-1.1
Experimental Details

Interventions

Intervention(s)
experimental treatments: PARTIAL, SPEC-TYPE, SPEC (details are provided below, see experimental design)

Addendum March, 14, 2024:
Two additional treatments are added: PARTIAL-GENDER and SPEC-GENDER (see details below)
Intervention Start Date
2021-05-25
Intervention End Date
2024-07-31

Primary Outcomes

Primary Outcomes (end points)
Preference for implementing a quota favoring disadvantaged tournament participants
Primary Outcomes (explanation)

Secondary Outcomes

Secondary Outcomes (end points)
Preference for real-world affirmative action policies
Secondary Outcomes (explanation)
Three survey questions:
Do you generally favor or oppose affirmative action programs for women? Favor - Oppose - No opinion
Do you generally favor or oppose affirmative action programs for racial minorities? Favor - Oppose - No opinion
Do you generally favor or oppose affirmative action programs for people with disabilities? Favor - Oppose - no opinion

Experimental Design

Experimental Design
We employ a between-subject design with three treatments.

Affirmative action is represented by a quota rule that, depending on subjects' decisions, may or may not be applied to govern the outcome of a tournament. In the tournament, subjects complete a real-effort task that consists of encoding words by substituting the letters of the alphabet with specific numbers given in a table. Subjects participate in the tournament in groups of six. Without the quota rule, the two subjects who encode most words correctly in five minutes are the winners and receive a monetary prize. The remaining four tournament participants are losers and don't receive any monetary payment. Three of the six tournament participants are assigned to the "Green Group" that is disadvantaged since every correctly encoded word is only counted as 0.9 words, while the other three tournament participants are assigned to the "Blue Group" whose performance is not downgraded. Before the start of the tournament subjects decide whether they want to implement a quota rule that ensures that at least one, namely the highest performing, member of the disadvantaged "Green Group" has to be selected as a winner.

One of six subjects' choices is randomly drawn to be implemented. Depending on random treatment assignment, subjects are either i) first assigned to either the disadvantaged or advantaged group, then decide on the quota rule for their own group and participate in the tournament themselves (PARTIAL treatment), ii) first assigned to either the disadvantaged or advantaged group, then decide on the quota rule for a group of tournament participants but only observe the tournament outcome without participating in the tournament themselves (SPEC-G treatment), or iii) are not assigned to any group, decide on the quota rule for a group of tournament participants and only observe the tournament outcome without participating in the tournament themselves (SPEC treatment). A fourth group of subjects acts as tournament participants, for whom subjects in the SPEC-TYPE and SPEC treatment decide on the implementation of the quota rule, but is not of major interest for our research question. Between-subject comparisons between the disadvantaged and standard group as well as between treatments allow us to disentangle impartial preferences for affirmative action policies from other motives, in particular in-group favoritism and self-interest (relating to own chances of winning).

In all three treatments, subjects answer a questionnaire that elicits the following measures: beliefs about chances of winning (with and without the quota rule), in-group favoritism, risk aversion, altruism, socio-demographics, political orientation, prior experienced discrimination, preferences for efficiency, overconfidence, and perceived fairness of the quota rule. These measures serve to analytically show what drives preferences for implementing affirmative action on a within-subject level.

Addendum March,14, 2024:

While we used a minimal group paradigm (Green and Blue group) before, we add two new treatments (PARTIAL-GENDER and SPEC-GENDER) that use groups that supposedly subjects identify with even more strongly. We conduct the same experiment as before but instead of random assignment to either the disadvantaged or advantaged group, women are always assigned to the disadvantaged and men to the advantaged group and subjects know this.
Experimental Design Details
Not available
Randomization Method
computerized random assignment (part of the experiment software) to treatments
Randomization Unit
individual
Was the treatment clustered?
No

Experiment Characteristics

Sample size: planned number of clusters
see "Sample size"
Sample size: planned number of observations
780 decision makers in three treatments (+ 468 observations from 78 additional subjects who participate in six tournaments each) Addendum March, 14, 2024: 624 decision makers in two treatments (+ 312 observations from 52 additional subjects who participate in six tournaments each)
Sample size (or number of clusters) by treatment arms
PARTIAL treatment: 312 subjects (156 in the disadvantaged, 156 in the advantaged group)
SPEC treatment: 156 subjects (no sub-groups)
SPEC-G treatment: 312 subjects (156 in the disadvantaged, 156 in the advantaged group)

(additional tournament participants for SPEC and SPEC-G: 78 subjects, each participating in 6 tournaments)

Addendum March, 14, 2024:
PARTIAL-GENDER treatment: 312 subjects (156 in the disadvantaged/female, 156 in the advantaged/male group)
SPEC-GENDER treatment: 312 subjects (156 in the disadvantaged/female, 1506 in the advantaged/male group)

(additional tournament participants for SPEC-GENDER: 52 subjects, each participating in 6 tournaments)
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
With 780 decision makers, that is, 156 decision makers per treatment and group (disadvantaged/advantaged), we can detect an effect of at least -10 percentage points at conventional levels of power (one-sided chi^2 test, significance level 5%, power 80%). This is based on the assumption that in treatment PARTIAL 90% of subjects in the disadvantaged group ("Green Group") favor a quota rule over no quota rule. Should baseline agreement be smaller, we may need to increase the number of observations. At lower baseline agreement rates of 85% or 80%, respectively, but the same number of decision makers, we could still detect the same effect size at a power of 71% and 65%, respectively.
IRB

Institutional Review Boards (IRBs)

IRB Name
German Association for Experimental Economic Research e.V. Institutional Review Board
IRB Approval Date
2021-04-28
IRB Approval Number
qtbZuqg8
IRB Name
German Association for Experimental Economic Research e.V. Institutional Review Board
IRB Approval Date
2024-03-13
IRB Approval Number
6hnVZuQf
Analysis Plan

Analysis Plan Documents

Analysis plan

MD5: a5700c68e31dbc0f9c8cb1603b20760c

SHA1: d667a81f904d97d32254bb9515bec19422057e02

Uploaded At: May 21, 2021

Addendum to Data analysis plan.pdf

MD5:

SHA1:

Uploaded At: March 14, 2024