k-best feature selection and ranking via stochastic approximation

Abstract

The relevance and impact of political scientists’ professional activities outside of universities has become the focus of public attention, partly due to growing expectations that research should help address society’s grand challenges. One type of such activity is policy advising. However, little attention has been devoted to understanding the extent and type of policy advising activities political scientists engage in. This paper addresses this gap by adopting a classifcation that distinguishes four ideal types of policy advisors representing difering degrees of engagement. We test this classifcation by calculating a multi-level latent class model to estimate key factors explaining the prevalence of each type based on an original dataset obtained from a survey of political scientists across 39 European countries. Our results challenge the wisdom that political scientists are sitting in an “ivory tower”: the vast majority (80%) of political scientists in Europe are active policy advisers, with most of them providing not only expert guidance but also normative assessments.