Free and fair to whom? Expert and citizen assessments of election quality in twenty-two African countries

Abstract

Purpose. Two instruments claim to measure the quality of the same election. Expert panels score it against a checklist of procedural standards; mass surveys ask the people who voted in it. Both are used, often interchangeably, as measures of electoral integrity. This article asks how far the two agree across African elections, and what follows for research that treats either as a proxy for the other.

Methods. Citizen assessments come from the Afrobarometer Round 10 country reports, which print the full response distribution for the question on the freeness and fairness of the last national election. Thirty-five countries were transcribed and each was accepted only after its two positive categories reconciled, within one percentage point, with the figure Afrobarometer published for that country in Dispatch 1128. Expert assessments come from Perceptions of Electoral Integrity 11.0. The citizen item was rescaled to 0 to 100 so that it shares the metric of the expert index, and countries were retained only where the election the survey asked about is the election the expert panel scored. Twenty-two countries survive both filters.

Findings. Citizens rate their elections 14.6 index points more favourably than experts do on average (SD 20.7; t = 3.31, p = 0.003), with 17 of 22 gaps positive. The direction is the opposite of the global pattern, where citizens are the harsher judges. More consequentially, the two measures barely track each other. Regressing the citizen measure on the expert measure gives a slope of 0.17 (SE 0.16) and an R-squared of 0.04; the slope cannot be distinguished from zero but is decisively different from one (p = 0.0001). Press freedom does not account for the divergence, contrary to the mechanism established on a 25-nation sample. Perceived freedom of speech enters with the sign opposite to the one preference falsification predicts.

Implications. The divergence is not well explained by citizens being misinformed or intimidated. It is better read as the two instruments measuring different things: experts score the administration of a contest, citizens appear to render a verdict on the political order that contest belongs to. Neither is thereby wrong, but they are not substitutes, and the common practice of using a survey item as a proxy for election quality, or as a validity check on expert scores, does not survive the evidence reported here.