How credible are Okun coefficients? The gap version of Okun’s law for G7 economies

Abstract

The paper draws attention to the fact that findings that follow from estimation of Okun’s law are extremely sensitive to methodological choices. The argument rests in a case study oriented upon G7 countries for a period 1991/Q1–2021/Q4 and accounts for a possible asymmetry in the output–unemployment relationship. First, business and unemployment fluctuations are estimated by six purely statistical approaches that arise by casting the Hodrick–Prescott filter, the Hamilton filter and the unobserved component model into a univariate or bivariate framework. Second, the gap version of Okun’s law is modelled by means of an auto-regressive distributed lag model or its nonlinear threshold counterpart according as asymmetry is allowed or not. The results indicate huge heterogeneity in Okun coefficients for every country caused by differences even in the basal methodological aspects accounted for in the case study. The diversity of results demonstrates that initial modelling choices may provide economic policy-makers with conflicting insights and advice. This issue follows merely from the absence of general standards that might decide which particular result is more credible.