No. 3 / 2026 — Autumn

Abstract. Modern economics modeling practice involves highly unrealistic assumptions. Since testing such models is not always an easy enterprise, researchers face the problem of determining whether a result is dependent (or not) on the unrealistic details of the model. A solution for this problem is conducting robustness analysis. In its classical form, robustness analysis is a non-empirical method of confirmation – it raises our trust in a given result by implying it with from several different models. In this paper I argue that robustness analysis could be thought as a method of post-empirical failure. This form of robustness analysis involves assigning guilt for the empirical failure to a certain part of the model. Starting from this notion of robustness, I analyze a case of empirical failure from public choice theory or the economic approach of politics. Using the fundamental methodological principles of neoclassical economics, the first model of voting behavior implied that almost no one would vote. This was clearly an empirical failure. Public choice scholars faced the problem of either restraining the domain of their discipline or giving up to some of their neoclassical methodological features. The second solution was chosen and several different models of voting behavior were built. I will treat these models as a case for performing robustness analysis and I will determine which assumption from the original model is guilty for the empirical failure.