Do Perceptions Match Provision? Conceptualizing and Modeling the Perception Gap

• In Event: Information and Perceptions

Abstract

Do citizens accurately evaluate the quality of government service provision? Advocates of retrospective democratic accountability argue that citizens are able to reward or sanction politicians according to their ability to provide services like water, electricity, health clinics, schools, and roads. Yet, the Political Science literature is replete with evidence that citizens frequently misevaluate the quality of policy implementation and service provision. These theories build on the insight that evaluating policy, even retrospectively, is cognitively difficult, and citizens in advanced and developing democracies alike systematically misevaluate performance. However, authors propose different sources of evaluation bias: Some focus on individual political sophistication, arguing that the politically uninformed and disengaged are less able to evaluate actual levels of service provision (Gilens 2001, Gomez and Wilson 2001). Contending that personal circumstances and ill fortune shape performance evaluations, another group of scholars examine how individuals’ sociotropic and egocentric economic evaluations are biased by irrelevant or misperceived events (Achen and Bartels 2017, Sances 2017). Finally, others maintain that ethnic and partisan identity are important moderators of how citizens receive and process information about service provision (Adida et al. 2016, Carlson 2016).

While these existing theories provide insight into how citizens (mis-)evaluate government performance, they are primarily focused on examining the variation in citizens’ assessment of government performance, rather than modeling the actual divergence between policy implementation and citizens’ perception of that service – a concept we term the “perception gap.” Moreover, these theories are generally distinct from one another, without engaging in a dialogue about when one source of bias might outweigh another. As a result, we are left with many well-defined theories that could explain this perception gap, but little guidance on how to best combine these theories to most effectively model and predict the perception gap between service provision and individual satisfaction with the provision of those services.

To address this oversight, our project puts these literatures in direct dialogue with one another, modeling the perception gap itself by comparing perceptions of government performance to objective service indicators in sub-Saharan Africa. We empirically test three competing determinants of voter behavior – political sophistication and interest, sociotropic economic voting, and partisan or ethnic bias – to examine where different retrospective voting theories hold the most explanatory power. Combining multiple rounds of Afrobarometer data with road network data, we directly model the direction (over-evaluation, under-evaluation, or congruent evaluation) and size of the perception gap. We then use a finite mixture model to systematically assess the relative explanatory power of variables related to each of these three theories. With a finite mixture model, we are able to estimate the probability that a single data point is consistent with each competing theory (Imai and Tingley 2012). By relaxing the assumption that one theory must explain all observations, this method provides us with a strategy for estimating whether political sophistication, personal economic circumstances, and partisan or ethnic identity shape the perception gap as well as which data points are best explained by each theory. This novel method allows us to investigate the contexts under which competing models of voter behavior apply, which holds important insights not only for theories of democratic accountability, but also for those seeking to understand on what basis citizens support particular politicians, independent of the policies they implement and services they actually provide.

Authors

• Paige Bollen, Massachusetts Institute of Technology

• Blair Read