On the Detection of State Dependence Using Aggregate Outflow Data: Comments on Previous Studies

Abstract

In this paper we examine a rather popular eyeball check which has been used to establish the presence of state dependence in aggregate unemployment duration data. We correct the literature by showing that this check may well lead to wrong conclusions. Empirical analysis of unemployment dynamics is based on a variety of methods ranging from the description of tabulated data to the application of highly sophisticated econometric methods. In between are lines of analytical reasoning supported by so called 'stylised facts', which include so-called 'eyeball tests' or 'eyeball checks'. An eyeball check provides evidence for the presence or absence of a particular phenomenon by way or eyeballing certain characteristics of graphs or tables. Eyeball checks are supposed to be easy-to-perform, to be robust with respect to (parametric) assumptions on the underlying theoretical framework, and, preferably, should have limited data requirements. Note that eyeball checks are not tests in the usual statistical sense. Vocabulary specific to statistical testing should therefore be used with caution in this context.