Ethnicity and Partisan Bias in African Survey Data: A Cross-National Overview

• In Event: Poster Session: Dynamics of Authoritarianism in China & New Research in African Politics
  In Poster Session: Poster Session: New Research in African Politics

Abstract

A critical limitation of survey data is response bias caused by social group identities. When asked to evaluate the performance of politicians and institutions, respondents may be guided by group identities, rate political allies positively, and discredit political opponents irrespective of their actual performance. Unobserved bias may produce spurious individual-level correlations and distort country-level statistics in cross-national comparisons. At the same time, systematic group bias can be insightful and reveal political alliances and the intensity of group politicization. The Afrobarometer represents a particularly impactful survey project. Researchers, policymakers, and citizens alike draw on the data to study political preferences and track progress across Africa. While several works have demonstrated that ethnic and partisan identities influence the performance perceptions of ordinary Africans in certain societies, we have little knowledge of the prevalence and the magnitude of identity bias in cross-national perspective. This poster seeks to fill this gap by providing an overview of ethnic and partisan bias across Africa, drawing on several waves of Afrobarometer data. The analysis considers more than 40 relevant ethnic groups and partisans of all major parties. Beyond deviations from the sample average, I focus on gaps between opposing camps. While my approach clearly highlights ethnic polarization in certain countries (e.g. Ghana, Kenya, Sierra Leone, South Africa), consistent biases in favour of coethnics are surprisingly rare overall. In several instances, the performance ratings of groups in power contradict expectations of coethnic loyalty or turn around over time under the same administration. Partisan biases are more consistent across surveys, but the intensity varies substantively.

Author

• Simon Primus, University of Erlangen-Nuremberg (FAU)