Eliciting Preferences of Party Leaders for Candidate Characteristics

How to cite

Leucht, Lukas; Tukiainen, Janne; Kari, Tuomas. 2025. "Eliciting Preferences of Party Leaders for Candidate Characteristics." AEA RCT Registry. AEARCTR-0015917. https://doi.org/10.1257/rct.15917-1.2

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

Lukas Leucht* , Janne Tukiainen , Tuomas Kari

* Primary investigator

Registry ID
AEARCTR-0015917
DOI
https://doi.org/10.1257/rct.15917-1.2
First registered
Last updated
Status
In development
Trial dates
2025-06-11 – 2028-07-31
Countries
Private
JEL codes
D72, C93, P00, M50
Registry record
http://www.socialscienceregistry.org/trials/15917

Abstract

This study investigates the preferences of local party elites in Finland regarding the selection of municipal election candidates. Using a series of conjoint survey experiments, we elicit local party leaders’ evaluations of hypothetical candidate profiles varying across key attributes, including education, labor market experience, political experience, policy positions, age, and gender. This design allows us to isolate the importance that political leaders assign to each attribute in candidate selection, without the concern of confounding by potentially correlated candidate characteristics. We vary the framing of the choice tasks to shed light on how party leaders evaluate the importance of the attributes differently for electoral performance and the capacity to govern, and how these preferences depend on the composition of the rest of the candidate list. The results will provide novel evidence on the implicit trade-offs party actors make when balancing representational goals with other candidate characteristics. By explicitly modeling these preferences, the study offers new insights into intra-party decision-making and the micro-foundations of political representation in a proportional, multi-party democracy.

Intervention

Respondents complete 30 conjoint tasks in total, divided evenly across three sections. The first section includes 10 choices over candidate lists. The order of the two sections focusing on individual candidate evaluations is randomized between participants.

Experimental design

Our experiment is designed to estimate average marginal component effects (AMCEs) to assess the impact of each candidate attribute on respondents’ choices, averaged over the distribution of the other attributes. Additionally, conditional AMCES will be examined based on respondent characteristics, such as the demographic composition of their municipality. The study will also distinguish between preferences for candidates depending on the type of choice respondents have to make in the three types of conjoint experiments.

Design details

Randomization method
Randomization is done by the survey company via computer.
Randomization unit
We randomize on the level of each individual hypothetical candidate profile or list profile that respondents (i.e. local party leaders) are presented with to compare and choose from.
Clusters
We expect to receive around 6300 evaluations of candidate or list profiles by local party leaders. We will send the survey to 317 local party leaders, who will each be faced with 30 decisions among pairs of candidates or candidate lists (19020 = 317 x 30 x 2). We expect a response rate of less than a third of all respondents, based on existing studies of comparable populations (6340 = 19020 / 3). August Update: We received contact information for two additional local party leaders. We will add them to the study and send them an invitation to evaluate candidate and list profiles. At this moment, when we received the additional contact information and decided to include them, we do not yet have access to any of the data from the conjoint experiment. The survey company will share the data with us in September. We do not expect any significant changes to the resulting number of evaluations and our power calculation.
Observations
6300 evaluations.
Arms
Since this study employs a conjoint design, the concept of treatment arms with exact sample sizes for each treatment arm is not straightforwardly applicable. For each of the three types of conjoint experiments, we randomize each binary candidate attribute with equal probability. In expectation we should therefore observe 3150 evaluations for each possible candidate attribute, with 1050 evaluations for each attribute per type of conjoint experiment.
Primary outcomes
List Choice in the List Conjoint Candidate Choice in the Electoral Success Conjoint Candidate Choice in the Competence Conjoint Candidate Choice in the Pooled Sample of all Conjoints Differences in Candidate Choice between the Conjoints.

Cited by

No citation data available for this registration yet.

Keywords