Revisiting the trade openness–unemployment nexus: an application of the novel JKS panel causality test with static and dynamic panel models

Abstract

PurposeThis paper documents a robust empirical regularity: higher trade openness is associated with a lower unemployment rate. This paper also examines whether or not the effects of trade liberalisation depend on countries' income levels. Further, the dynamic causation between trade openness and unemployment is also examined.Design/methodology/approachIn order to obtain insight into the openness–unemployment nexus, following empirical methods were utilised - static panel models, dynamic panel models and a novel panel Granger causality approach proposed by Juodis et al. (2021).FindingsResults suggest that openness negatively affects unemployment; the extent to which trade liberalisation affects unemployment depends on the income level of each country. The Juodis, Karavias, and Sarafidis (JKS) test confirmed that the past values of trade openness, inflation, foreign direct investment and gross domestic product per capita contain information that helps to predict unemployment in a more robust manner. To simply put, opening upto trade may eventually become a requirement for creating more job opportunities, but this alone may not be enough. The extent to which nations benefit from trade liberalisation is largely dependent on the overall economic conditions and their capability to move up the income scale.Originality/valueA major difference between this study and those performed previously is that this study does not only examine the impact of trade openness on unemployment, but also investigates whether the unemployment effect of liberalisation is affected by countries' income levels – an issue that has received little attention in the past. Additionally, the unique panel non-causality approach put forth by Juodis et al. (2021) is used in the first instance to look into the causal link between trade openness and unemployment. This method has advantages in that the method enables capturing Granger-causality in homogeneous or heterogeneous panels amongst multiple variables.