Mixed records, complexity, and ethnic voting in African elections

Abstract

• Utilizes a survey experiment in a representative exit poll during Kenya’s election. • Explores effects of incumbent co-ethnicity and mixed performance on vote choice. • Effect of co-ethnicity is highest under mixed (versus uniform) performance. • Mixed performance increases evaluation complexity, inducing ethnic heuristic. • Suggests challenges of evaluating performance create barriers to accountability. Incumbent politicians often deliver mixed performance records: good results in some areas but poor ones in others. We explore the challenges these records generate for voters attempting to use elections to incentivize better governance, improved development outcomes, and greater accountability. We argue that evaluation of mixed records poses a more complex cognitive task for citizens than the evaluation of either uniformly good or bad records. Unlike uniform records, mixed records require weighting and aggregating information across multiple dimensions, raising the difficulty of arriving at a single, clear evaluation of performance. Evaluative complexity, in turn, induces voters to rely more heavily on informational shortcuts like ethnicity, reducing the effectiveness of elections as accountability mechanisms. We evaluate the link between ethnic voting and mixed performance records using a survey experiment implemented in a nationally representative exit poll during Kenya’s 2013 election. The survey experiment manipulated two factors, the ethnicity of a hypothetical candidate running for president and their performance record (with uniformly good, uniformly bad, and mixed performance conditions). We find that co-ethnicity between respondent and candidate had a large effect on voting when the candidate’s prior record was mixed, but no effect when it was uniform. Our results suggest that ethnic voting is situational rather than dispositional: even performance-oriented citizens vote ethnically when performance records are difficult to evaluate. They also suggest an under-appreciated source of accountability problems in developing countries: not fixed voter attachment to co-ethics, but rather reliance on informational shortcuts to deal with the challenges of evaluating performance in complex environments.