On the Use of Computer-Simulation for Qualitative Tests with Theoretical Samples

Abstract

Random samples are the «gold-standard» of quantitative empirical research: they allow to make inferences from a sample of cases to the corresponding population. When based on small samples (N ≤ 30), such inferences are often unreliable due to the so-called sampling errors. Moreover, critical cases required for a correct evaluation of a theory may be excluded from a small random sample. Consequently, this article favours theoretical samples, which are non-random but focus on cases that are especially important for the empirical test of a theory.
In order to construct theoretical samples the author proposes to use computer-models, which simulate for different scenarios the outcomes of the theory to be tested. The quantitative output of the simulations are reinterpreted in qualitative terms, e. g. referring to the (non-)​existence of an equilibrium or the growth or shrinking of a group of actors. Thus, by computer-simulation it is possible to gain theoretical samples of parameters and qualitative model-outcomes that can be compared with observational data.
As an illustrative case the author presents a computer-model of the spiral of silence in public discourses. The computer-simulations can identify the conditions, under which originally very visible opinions finally disappear. One of them is e. g. the shame attributed to the public support of an opinion. This way it becomes possible to construct theoretical samples of the conditions for the triggering of a spiral of silence, which can be compared with real data and consequently serve for testing the related theory.