Identity of affirmed groups and the consequences of affirmative action policies

How to cite

Trieu, Chi. 2019. "Identity of affirmed groups and the consequences of affirmative action policies." AEA RCT Registry. AEARCTR-0005122. https://doi.org/10.1257/rct.5122-1.1

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Chi Trieu*

* Primary investigator

Registry ID
AEARCTR-0005122
DOI
https://doi.org/10.1257/rct.5122-1.1
First registered
Last updated
Status
Completed
Trial dates
2019-12-04 – 2020-03-31
Countries
Germany (Nordrhein-westfalen)
JEL codes
C91, C92, D02, D63
Registry record
http://www.socialscienceregistry.org/trials/5122

Abstract

In an increasingly diverse population, an ongoing debate is whether or not affirmative action policies should favor more than one minority group. When such a policy is implemented, the affirmed group is not identifiable with certainty. We experimentally study how this feature affects the outcomes of affirmative action policies in a tournament context. In the baseline treatment, no affirmative action is implemented. In the gender quota treatment (GQ), affirmative action in the form of a gender quota makes the affirmed group (female) certainly identified. In the other treatment, a quota for both gender and a randomly assigned characteristic is applied in order to reduce the identity salience of females as an affirmed group. We test the difference in efficiency, willingness to compete, and partner selection in teamwork.

Experimental design

We applied affirmative action on the performance of an arithmetic task introduced by Niederle and Vesterlund (QJE, 2007), in which subjects have to add as many sets of five 2-digit numbers as possible in a given time. Measuring performance by the arithmetic task serves two purposes. First, no special knowledge is required, and minimal learning effect has been documented in previous studies. Second and more importantly, gender difference in willingness to compete is sensitive to the nature of the task. In particular, females are less competitive in mathematical tasks than males (Sutter et al. (EER, 2019)). Thus, by using an arithmetic task to measure performance, we introduce a reasonable need for gender quota favoring females in our setup.

We implement a between-subject treatment design with three treatments. The baseline treatment (CTR) is without affirmative action. In the gender quota treatment (GQ), a gender quota is applied to favor female candidates, which makes the favored group identifiable with certainty. In the mix treatment (MIX), we use a set of criteria for affirmative action, including gender and a randomly assigned characteristic to reduce the salience of gender.

The experiment consists of five stages:

Stage 1: Piece rate. Subjects work on the task for five minutes and receive a piece-rate payment of 0.5 Euro per correct answer.

Stage 2: Tournament. Subjects work on the same task as in stage 1, but under a tournament incentive. We randomly assign subjects to the Blue group and the Green group with equal probability. Each subject competes in a group of six, with three females and three males, three subjects from the Blue group and three from the Green group. Subjects stay in the same group in the following stages. Two winners are selected in each group. The winners earn 1.5 Euro per correct answer, while losers earn nothing. A random tie-breaking rule is applied in case of a tie. At the beginning of stage 2, subjects are informed about the group structure and their type. Subjects are not informed about the type of each group member. The result of the tournament is only informed at the end of the experiment. The treatments differ in how winners are selected:
- Baseline treatment (CTR): winners are the two best performers.
- Gender quotas (GQ): at least one of the winners must be female.
- Mix treatment (MIX): with 50% chance, a gender quota is applied - at least one of the winners must be female. With 50% chance, at least one of the winners must be from the Green group.

Stage 3: Self-selection into the tournament. Subjects choose whether they want to work under piece-rate incentives or tournament incentives before working on the task again. The rules of the tournament are the same as in stage 2. If a subject chooses the tournament in stage 3, her performance will be compared to the performance of her five fellow group members in stage 2. This feature ensures that beliefs about winning the tournament are similar in stages 2 and 3 (Niederle et al. (MS, 2013)).

Stage 4: Submit stage 1 score to either piece-rate or tournament. Subjects are asked to choose which scheme (piece-rate or tournament) they want to apply for their scores in Stage 1. If the tournament is chosen, their stage 1 score is compared with stage 1 score of the other five group members. The winners are then selected with the same rules and payoffs as in stage 2. If the piece-rate is chosen, subjects receive 0.5 Euro per correct answer in stage 1.

Stage 5: Partner selection in teamwork. Subjects move to a new working setup with a new gender-neutral task (Grid task, Abeler et al. (AER, 2011)). Before working, subjects are asked to team up with two of five subjects who used to be their group members in previous stages. Each group of three must include one leader and two employees. Leaders are two winners in stage 2, employers are the losers in stage 2. Subjects do not know about decisions on partner selection of others. The choice of partners of one subject does not affect the partner choice of others such that partner selection is an individual decision. Subjects then work on the Grid task for five minutes under a team incentive. The overall payoff of the team is equally shared among team members. Each correctly solved grid generates 0.6 Euro for the team and 0.2 Euro for each team member. In addition, the leader receives a bonus of 5 Euro, while each employee receives a bonus of 2 Euro.

A final questionnaire will include control variables such as overconfidence, risk preference, social preference and socio-economic background.

Design details

Randomization method
Subjects are invited via ORSEE. When arriving in the lab, subjects are randomly assigned to computers.
Randomization unit
Each individual subject will be one observation.
Clusters
No clusters
Observations
We aim at collecting at least 60 observations per treatment.
Arms
We aim at collecting at least 180 observations.
Primary outcomes
1. Proportions of affirmed and unaffirmed subjects entering the competition in Stage 3 and Stage 4. 2. The frequency that affirmed and unaffirmed subjects are selected as partners for teamwork in Stage 5.
Secondary outcomes
Individual performance under piece-rate incentives (stage 1), tournament incentives (stage 2 and stage 3), and in teamwork incentives (stage 5).

Materials

Cited by

No citation data available for this registration yet.

Keywords