Using Exit Poll Data from Kenya to Explore Ethnicity and Government Performance in Vote Choice. What Explains the African Vote?. Clark C. Gibson James D. Long. Department of Political Science UC San Diego.
Using Exit Poll Data from Kenya to Explore Ethnicity and Government Performance in Vote Choice
Clark C. Gibson
James D. Long
Department of Political Science
UC San Diego
a.) descriptive and cross tabs
b.) multivariate tests
c.) survey experiment
And why didn't a U.S.-funded group release its exit-poll data?
Slate Magazine, 1/2/08
Kenyan president lost election, according to U.S. exit poll
“IRI will not release any polling results unless and until we are confident in the integrity of the data.”
International Republican Institute 1/15/08
U.S. Ambassador Ranneberger: “…it is my understanding that this ‘exit poll’ was part of a training exercise and was never intended for publication.”
Kenya winner lost, U.S. poll indicates
Belated tally suggests election was stolen
Chicago Tribune 7/9/08
Testify before the Kriegler Commission 7/15/08
N = 5,495
Margin of error = 1.32
Approaches to explaining voting behavior in Africa
Identity/Expressive voting (Horowitz)
Elections become ethnic head counts
May even vote against their policy preferences
Policy voting (Hechter, Bates, Bratton, Mattes)
Co-ethnics care about same policies
Giving your co-ethnic a break on policy evaluation
Can be observationally equivalent to identity
Kibaki’s (Kikuyu) and Raila’s (Luo) groups cannot win election alone. Must at least have coalitions.
Some ethnic groups split their presidential votes.
Leaders from same ethnic groups join multiple coalitions.
Hypothesis 1.(Identity voting) If a voter has a co-ethnic candidate, she will vote for that candidate (if possible).
Hypothesis 2.(Policy voting)
Hypothesis 3.(Government performance) If a voter believes that the government has performed well, she will be more likely to vote for the incumbent.
Kikuyu22% (President Kibaki’s group)
Luo13% (Odinga’s group)
Kamba11% (Musyoka’s group)
Other African 15%, Non-African 1%
5,495 surveys, nationally representative:
8/8 provinces; 69/71 districts; 179/210 constituencies.
Good for provincial estimates (“25% in 5 provinces” rule).
Multi-stage cluster sampling proportionate to size, using final ECK published registration.
Random selection of polling stations within constituencies, random selection of respondents (every 5th person).
Process and timing of voting
Performance of local, parliamentary, and central government
Attitudes about policies, issues and ethnicity
Vote choice for local, MP, president
sig * p<0.05; ** p<0.01; ***p<0.001
Constants and controls suppressed.
What Explains the African Vote?
and/or source of the exit poll"