Do Politicians Discriminate Against (polarized) Constituents? A Field Experiment on State Legislators

Last registered on March 30, 2021

Pre-Trial

Trial Information

General Information

Title
Do Politicians Discriminate Against (polarized) Constituents? A Field Experiment on State Legislators
RCT ID
AEARCTR-0006599
Initial registration date
October 22, 2020

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
October 26, 2020, 8:18 AM EDT

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

Last updated
March 30, 2021, 5:25 AM EDT

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

Locations

Region

Primary Investigator

Affiliation
Siegen University

Other Primary Investigator(s)

PI Affiliation
PhD

Additional Trial Information

Status
In development
Start date
2020-10-26
End date
2021-04-30
Secondary IDs
Abstract
Experiments are often replicated in discrimination research. These replication studies are a necessary means of checking and confirming effects. With this motivation we want to replicate a study by Butler and Broockman (2011) on discrimination against blacks in the light of the approaching US election campaign.
Butler and Broockmann (2011) investigated the effect of the first and last name of a person who makes a request to “his" or "her" MP. The representatives of the US citizens are known as so-called 'State Legislators'. The 'State Legislators' are current members of the Lower Chamber (State Assembly, House of Representatives etc.) or the Upper Chamber (Senate) of a US state. Following the experimental design of Butler and Broockman (2011), inquiries were made to these 'State Legislators'. The inquiries were sent by two different people with fictitious first and last names, which, according to the US census, are very likely to be “black” or “white”. The request was for information on how to register to vote. With the help of these name signals, politicians can be put in the position of responding to inquiries from different US citizens. If the request is answered, this is an indication of responsiveness. The time it takes to receive a response can also be examined.
Study design
We want to use the same design during the US election campaign. To do this, we use an experimental survey to investigate whether the race of the inquirer has an influence on whether and how quickly the addressees respond to inquiries from these two inquirers. By varying the inquirer, we achieve an experimental character, just as Butler and Broockman (2011) and many other studies from political science have presented: While the name 'Matthew Mueller' can be connected with a white, the name 'Deshawn Jackson 'more likely to be associated with a black. We want to use these names and ask about the prevalence of police violence in each constituency.
External Link(s)

Registration Citation

Citation
Köhler, Ekkehard and Marius May. 2021. "Do Politicians Discriminate Against (polarized) Constituents? A Field Experiment on State Legislators ." AEA RCT Registry. March 30. https://doi.org/10.1257/rct.6599-1.5
Experimental Details

Interventions

Intervention(s)
We ask question
Intervention Start Date
2020-10-26
Intervention End Date
2020-11-02

Primary Outcomes

Primary Outcomes (end points)
(i) We expect that the Democrats’ responsiveness towards blacks is higher and that the Republicans’ responsiveness towards whites is higher.
(ii) We expect that the Democrats’ responsiveness is increased towards inquirers supporting the hypothesis that blacks are subject to police violence more often. Contrarily, we expect that the Republicans’ responsiveness is increased towards inquirers supporting the hypothesis that whites are about as often as blacks subject to police violence.
(iii) We expect that the candidates’ responsiveness is higher in case the candidate’s and the inquirer’s race are the same.
Primary Outcomes (explanation)
(i) There is evidence that blacks tend to vote for Democrats and whites rather vote for Republicans, which we expect to be reflected in a higher responsiveness.
(ii) Compared to Democrats, Republicans rather tend to support the hypothesis that there is no significant difference between police violence for each race. Thus, we expect this ideological view to account for strategic voting behavior.
(iii) There is evidence that politicians tend to answer requests more often in case the inquirer’s race matches her or his race.

Secondary Outcomes

Secondary Outcomes (end points)
(i) We expect a minor increase in the incumbent’s responsiveness if the crime statistic matches the inquirer’s opinion and the ideological view of the incumbent. For the opponent, there is a respective decrease in responsiveness.
(ii) We expect a medium increase in the opponent’s responsiveness if the crime statistic matches the inquirer’s opinion and not the ideological view of the incumbent. For the incumbent, there is a respective decrease in responsiveness.
(iii) We expect a major increase in the opponent’s responsiveness if the crime statistic does not match the inquirer’s opinion and neither the ideological view of the incumbent. For the incumbent, there is a respective decrease in responsiveness.
(iv) We expect a major increase in the incumbent’s responsiveness if the crime statistic does not match the inquirer’s opinion but the ideological view of the incumbent. For the opponent, there is a respective decrease in responsiveness. This is the opposite case to (iii).
Secondary Outcomes (explanation)
All effects are based on strategic behavior, inducing polarization if favorable and trying to avoid polarization if disadvantageous. The first effect confirms positive polarization with respect to the incumbent’s opinion, the second effect confirms negative polarization with respect to the incumbent’s opinion and both the third and fourth effect elicit a change of polarization. The magnitude of effects correspond to the strategic importance of sending a response implied by its potential effect on the inquirer’s voting behavior.

Experimental Design

Experimental Design
This experiment aims to find effects of taste-based discrimination within US-American politicians in the responsiveness towards blacks and whites as well towards voters of different political attitudes.
Therefore, we designed an experiment with two binary dimensions on side of the inquirers who request information concerning police violence in their election area from their respective state legislator. State legislators belong to the Lower Chamber and the Upper Chamber of each state. The first of the two dimensions of the inquirers is the inquirer’s race, which can be either black or white. The second dimension is the opinion concerning police violence in the respective state of residence, with one attitude being that blacks are subject to police violence significantly more often than whites and the other one being that whites are approximately as often subject to police violence as blacks. Each legislator is contacted via email by one of the four profiles which consist of all convex combinations of the two dimensions. We will then observe whether a state legislator responds to the request and if, how much time was required. Essentially, the statistics and other information potentially provided by the legislator are less important than the answer itself. For secondary outcomes (see above), we will check ourselves what political attitude is supported by empiric data.
Experimental Design Details
Inquiries are sent via email by two fictional people. First names and surnames are chosen such that they imply that the corresponding inquirer is black (e. g. Deshawn Jackson) or white (e. g. Matthew Mueller) with a high probability according to US census. This is the signal we employ to convey the race of the inquirer. The ideological view is transmitted by the text in the inquiry.
We only consider state legislators standing for election on November 3, 2020. In order to create our sample, we refer to Ballotpedia. In addition, we gather data about each legislator’s party, years in office and the tenure period. Moreover, we collect the photo of each legislator from the official website. All data concerning legislators on individual level is compiled from Wikipedia. The photos are analyzed on the race, emotions and gender utilizing artificial intelligence.
During our experiment, we generate two samples of data. Firstly, we create an anonymous data set in which each legislator receives a randomly generated ID. We collect information such as the reaction on our email. Secondly, we create a set of control data in which the aforementioned individual information on legislators is stored. The latter data set is password-protected and the first data set is anonymized, ensuring that legislators cannot be identified. Thus, issues regarding data protection are mitigated optimally.
Randomization Method
Simple Randomization is done in office by a computer.
Randomization Unit
individual level based on State and technology
Was the treatment clustered?
No

Experiment Characteristics

Sample size: planned number of clusters
5045, which is the number of state legislator positions to be elected again on November 3, 2020.
Sample size: planned number of observations
5045, which is the number of state legislator positions to be elected again on November 3, 2020.
Sample size (or number of clusters) by treatment arms
Approximately 1261 for each of the 4 possible treatments:
(1) Black + Police violence tends to harm blacks more
(2) Black + Police violence tends two harm both races equally
(3) White + Police violence tends to harm blacks more
(4) White + Police violence tends two harm both races equally
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
We establish one null-hypothesis for each of the expected primary and secondary outcomes and check for significance on 90 %, 95 % and 99 % levels. Determining the sample size a priori requires to know reasonable assumed effect sizes. However, we do not know of any study that analyses responsiveness on inquires on police violence to state legislators. Other priors are not available. This is why we maximize the number of subjects in our experiment.
IRB

Institutional Review Boards (IRBs)

IRB Name
Rat für Ethik und Forschung
IRB Approval Date
2020-10-21
IRB Approval Number
ER_26/2020

Post-Trial

Post Trial Information

Study Withdrawal

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information

Intervention

Is the intervention completed?
No
Data Collection Complete
Data Publication

Data Publication

Is public data available?
No

Program Files

Program Files
Reports, Papers & Other Materials

Relevant Paper(s)

Reports & Other Materials