close× Call Us
Title Experimental Tests of Duverger’s Law in the Context of Salient Ethnic Divisions
Post date 02/23/2019
C1 Background and Explanation of Rationale Duverger’s Law/Hypothesis, one of the most well established theoretical expectations in Political Science, holds that restrictive electoral systems (like Single Member Plurality electoral rules) tend to induce coordination around a few large parties capable of capturing majorities, whereas proportional systems encourage larger number of parties to win office, enabling or requiring coalition governments. The key behavioral mechanism underlying Duverger is strategic voting, or the propensity of voters to defect from candidates or parties they believe cannot win. This project explores: 1) whether Kenyan voters engage this type of strategic voting; and 2) whether salient co-ethnicity between voter and candidate interferes with this behavior. It uses a 3x2 factorial survey experiment delivered through a door-to-door survey in Nairobi, Kenya to evaluate these hypotheses. (For more, see PAP).
C2 What are the hypotheses to be tested?

The core hypotheses are:

H1: Co-ethnic bias: respondents will expect more effectiveness from co-ethnics.

H2: Information updating: “popular with voters” treatment will increase beliefs about p[winning] versus no information; “unpopular with voters” treatment will decrease beliefs about p[winning] versus no information.

H3: Disrupted information updating: respondent beliefs about viability and p[winning] will respond less to “unpopular with voters” informational treatment when it pertains to a co-ethnic.

H4: Strategic voting: “unpopular with voters” treatment will decrease p[voting for candidate] versus no information. The “popular with voters” treatment should increase p[voting for candidate] versus no information.

H5: Disrupted strategic voting: co-ethnicity will attenuate H4.

H6: Ethnic voting: respondents will prefer co-ethnics.

(There are also several expected heterogeneous treatment effects, see PAP).

C3 How will these hypotheses be tested? * The hypotheses will be tested in a 3x2 factorial design survey experimented embedded in a door-to-door survey. The experiment involves a hypothetical candidate for governor and the primary outcome variable is beliefs about the candidate’s chances of winning and probability of voting for the candidate. The first factor (viability) will provide information that a candidate is likely to win, not likely to win, or no information. The second factor (co-ethnicity) will randomize co-ethnicity between respondent and hypothetical candidate.
C4 Country Kenya
C5 Scale (# of Units) 2400
C6 Was a power analysis conducted prior to data collection? An informal analysis was conducted. Main effects are well powered, the interaction less so. Effects below .08 pp may be difficult to pick up.
C7 Has this research received Insitutional Review Board (IRB) or ethics committee approval? Yes
C8 IRB Number 171699
C9 Date of IRB Approval 9/5/2018
C10 Will the intervention be implemented by the researcher or a third party? The experimental treatment will be delivered by survey interviewers from the survey company, Ipsos.
C11 Did any of the research team receive remuneration from the implementing agency for taking part in this research? No
C12 If relevant, is there an advance agreement with the implementation group that all results can be published? Yes
C13 JEL Classification(s) not provided by authors