Project Federal German election 2021 – Taste-based discrimination and strategic behavior

INTERVENTIONS
None

2021-08-24

2021-10-31

PRIMARY OUTCOMES
Primary Outcomes (end points)
(i) We expect that the overall candidates’ responsiveness towards inquirers without immigrant background is higher and that the candidates’ responsiveness towards inquirers with immigrant background.
(ii) We expect that candidates with immigrant background answer more frequently to inquirers with immigrant background than candidates without immigrant background.
(iii) We expect that candidates from parties in favor of dual citizenship are more responsive to inquirers signaling a preference for dual citizenship. We also expect that candidates from parties against dual citizenship are more responsive to inquirers signaling a preference against dual citizenship.
(iv) We expect that candidates from parties in favor of dual citizenship are more responsive to inquirers with immigrant background. We also expect that candidates from parties against dual citizenship are more responsive to inquirers without immigrant background.

Primary Outcomes (explanation)
(i) There is evidence that German politicians rather respond to inquiries made by people without immigrant background compared to inquiries made by people with immigrant background.
(ii) We expect this outcome due to affiliation with people of the same race.
(iii) These expectations are based on partisanship effects which have been in corresponding literature. It can also be explained by strategic behavior of the candidate. An answer in opposition of the inquirer’s opinion might deteriorate the chances to gain the inquirer’s vote, which is anticipated by the candidate.
(iv) This outcome can be driven by two reasons: Firstly, inquirers with immigrant background are often affected directly by dual citizenship. Thus, parties in favor of this policy should be more responsive to these inquirers and vice versa due to partisanship. Secondly, parties in favor of dual citizenship are rather considered liberal and vice versa. As people with immigrant people background tend to vote for liberal parties and vice versa, there is another partisanship effect.

Secondary Outcomes (end points)
(i) We expect that if a candidate is confronted with an opinion against the candidate’s party, the candidate answers more often contrary to the party’s preference compared to the case in which the candidate is confronted with an opinion in line with the candidate’s preference.
(ii) We expect that the non-incumbents’ responsiveness is higher if the inquirer’s and the candidate’s attitude towards dual citizenship match but disagree with the preference of the incumbent’s party than if the inquirer’s, the candidate’s and the incumbent’s attitude match.
(iii) We expect a lower responsiveness of non-incumbents if the attitude on dual citizenship of the candidate’s party matches the preference of the incumbent’s party.
(iv) We expect that non-incumbents state an opinion contrary to their party’s but in line with the inquirer’s opinion on dual citizenship more often if the candidate’s and the incumbent’s party share their preference on dual citizenship than if the candidate’s and incumbent’s party have different positions in this matter.

Secondary Outcomes (explanation)
All effects are based on strategic behavior:
(i) The candidate can confirm the inquirer’s preference to increase the chances to be voted for.
(ii) Non-incumbents have an incentive to signal different preferences than the incumbent to increase voting chances, given that the nonincumbent’s and the inquirer’s opinion match.
(iii) Non-incumbents with the same ideological interests as the incumbent have fewer incentives to answer inquiries. Satisficing voters vote again for the incumbent if expectations are met. Thus, signaling the same preference as the incumbent has does not increase voting chances.
(iv) Contrary to the prior case, non-incumbents sharing the incumbent’s preference may confirm the inquirer’s preference to increase their voting chances even though the non-incumbent’s party prefers the opposite.

This experiment aims to find effects of taste-based discrimination within German politicians in the responsiveness towards voters with immigrant background and without immigrant background as well towards voters of different political attitudes.
Therefore, we designed an experiment with two binary dimensions on side of the inquirers who ask candidates of districts for the German Bundestag about their opinion on dual citizenship. We take the candidates of the 6 parties currently represented in the German Bundestag into account. Candidates of districts constitute half of the parliament. Within each district, the candidate with the most votes is elected. The first of the two dimensions of the inquirers is a potential immigrant background. The inquirer signals, from a German point of view, an immigrant background or not. The second dimension is the opinion on dual citizenship, in an affirmative or rejecting sense. Each candidate 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 candidate responds to the request and if, how much time was required and what opinion on dual citizenship is conveyed.

Inquiries are sent via email by two fictional people. First names and surnames are chosen such that they imply that the corresponding inquirer has immigrant background (e. g. Yusuf Wolf) or not (e. g. Linus Müller) with a high probability according to German census. This is the signal we employ to convey a potential immigrant background of the inquirer. The ideological view on dual citizenship is transmitted by the text in the inquiry.
The text of the inquiry also signals that the inquirers are high school students. Conclusively, inquirers, irrespective of treatments, are first-time voters. This specific group of the electorate is of special interest for politicians.
We consider all candidates of districts running for the German federal election on September 26, 2021. In order to create our sample, we refer to abgeordnetenwatch. In addition, we gather data about each candidate’s party. Moreover, we collect the photo of each legislator from the official website. All data concerning candidates on individual level is compiled from Wikipedia. The photos are analyzed on the race, emotions and gender utilizing artificial intelligence. A potential immigrant background is also estimated. We collected the attitude towards dual citizenship of all 6 parties in concern from their preliminary election programs. As an exception, due to lack of such a program, we collected the attitude of CDU/CSU from their government program for the current election period. The attitudes are as follows:

Party View on dual citizenship
Die Linke Affirmative
Bündnis 90 / Die Grünen Affirmative
SPD Affirmative
FDP Not unconditionally affirmative
Union Dual citizenship has to be an exception
AfD Declining

During our experiment, we generate two samples of data. Firstly, we create an anonymous data set in which each candidate 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 candidates 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
Randomization is done in office by a computer.

Randomization Unit
District level

Yes

Sample size: planned number of clusters
299 clusters with respect to randomization unit (district level). 4 clusters with respect to treatment clusters (two binary dimensions) within randomization unit.

Sample size: planned number of observations 1794, which is the number of candidates in districts of the 6 parties currently represented in the German Bundestag. The respective election takes place on September 26, 2021.

Approximately 449 for each of the 4 possible treatments:
(i) Immigrant background + in favor of dual citizenship
(ii) Immigrant background + against dual citizenship
(iii) No immigrant background + in favor of dual citizenship
(iv) No immigrant background + against dual citizenship

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.