Gender in Experimental Elections

How to cite

Poutvaara, Panu; Graefe, Andreas. 2020. "Gender in Experimental Elections." AEA RCT Registry. AEARCTR-0006653. https://doi.org/10.1257/rct.6653-2.0

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

Panu Poutvaara* , Andreas Graefe

* Primary investigator

Registry ID
AEARCTR-0006653
DOI
https://doi.org/10.1257/rct.6653-2.0
First registered
Last updated
Status
Completed
Trial dates
2020-10-23 – 2020-11-30
Countries
United States of America ()
JEL codes
D72, J16
Registry record
http://www.socialscienceregistry.org/trials/6653

Abstract

We study the role of the candidates' gender as a cue for voting in the United States in a low-informational setting. In this setting, voters can observe candidates' gender from headshots but have no additional information about the candidates. Our conjecture is that
(a) Democrats of either gender are more likely to vote for women than Republicans of the same gender
(b) those supporting Biden of either gender are more likely to vote for a female candidate than those supporting Trump of the same gender
(c) men tend to support male candidates more often than women
(d) women tend to support female candidates more often than men.
Additionally, we study how confident respondents are in their choices in experimental elections, as well as to what extent respondent confidence in their choice depends on the appearance gap between the two candidates. In this respect, our conjectures are:
(e) Republicans are, on average, more certain in their choice than Democrats,
(f) Trump supporters are, on average, more certain in their choice than Biden supporters,
(g) Republicans are, on average, more likely to choose better-looking candidates than Democrats,
(h) Trump supporters are, on average, more likely to choose better-looking candidate than Biden supporters,
…at least when it comes to choices between two candidates of the same gender.

Intervention

Each respondent is shown pairs of randomly selected photos. They are then asked which person they would vote for as a member of the House of Representatives if they would have to decide based on photos only, or if they would rather abstain.

Experimental design

Each respondent is shown 99 pairs of randomly selected photos, so that each pair depicts with equal probability (0.25) (a) two males, (b) two females, (c) one female and one male, and (d) one male and one female. They are then asked which person they would vote for as a member of the House of Representatives if they would have to decide based on photos only, or if they would rather abstain. The order in which the different combinations are shown is randomized. To ensure that American respondents do not know the candidates in photos, we use photos of MEPs, collected from the European Parliament’s web site. Subsequently, the respondent is asked a number of background questions, including their party identification and which candidate they would like to see elected as the president. We use Amazon MTurk to recruit the respondents.

Design details

Randomization method
Computer program. The survey is programmed using https://www.soscisurvey.de/
Randomization unit
Randomization takes place at two levels for each respondent. First, we randomize for each pair of photos, whether it depicts two females, two males, one female on the left and one male on the right or one male on the left and one female on the right. Then, we randomly select the photos of the chosen gender to be shown.
Clusters
360 respondents
Observations
360 respondents are invited to make choices in 99 experimental elections. If everyone replies to all choices, we have 35,640 observations.
Arms
360 respondents are invited to make choices in 99 experimental elections. If everyone replies, this gives us 35,640 observations. The expected number of evaluations that include one female and one male candidate is 17,820, and the expected number of observations of choices between two female or two male candidates is also 17,820.
Primary outcomes
Our primary interest is the share of female candidates selected in experimental elections that include one female and one male candidate, excluding experimental pairs in which the respondent would abstain.
Secondary outcomes
The level of confidence that respondents have in their choice, as well as how candidates' performance in experimental elections relates to the candidates’ average looks.

Papers from this trial

  1. Poutvaara, Panu and Graefe, Andreas. (2024). Do Americans Favor Female or Male Politicians? Evidence from Experimental Elections. CESifo Working Paper No. 11414. https://www.cesifo.org/DocDL/cesifo1_wp11414.pdf

Cited by

No citation data available for this registration yet.

Keywords