Validity of Okun’s Law in a spatially dependent and cyclical asymmetric context

Abstract

The current article analyzes the validity of Okun’s Law and sizable distortions that can occur in the estimation when spatial dependence and cyclical asymmetric impacts are not considered, which is a concern commonly ignored by the existing literature. Primarily spatial panel regressions (SDM, SAR, and SEM) and nonparametric regressions along with specification tests are adopted in terms of the methodology (such as panel unit root tests, panel cointegration, Moran’s I and Geary’s C tests of global spatial dependence, spatial LM, and Hausman tests). Additionally, spatial heterogeneity and cross-regional variation in Okun’s Law are investigated by adopting geographically weighted regression, LISA (local indicators for spatial association), and local Geary’s C analysis. A panel of 26 Turkish NUTS-2 regions from 2004 to 2018 was analyzed. The results clearly revealed that failing to incorporate spatial proximity and asymmetric cycle impacts leads to the biased estimation of Okun’s coefficient, such that during the downswing years of the national economy, Okun’s Law holds robustly: unemployment increases quickly in response to a decline in output. In contrast, during upswing years, the size of Okun’s coefficient is relatively much lower. Moreover, spatial dependence and heterogeneity are sizably evident. Okun’s coefficient is demonstrated to vary significantly across regions that have different industrial and labor market characteristics. As a policy implication, it has been understood that the reduction of unemployment is more difficult than initially understood, as economic growth itself does not provide a solution during upswing periods. The necessary special and region-specific policies are discussed throughout the text.