Labour market flexibility and unemployment: evidence from dynamic panel models

Deregulation of the labour market has previously been supported by both the OECD and the World Bank. However, by 2013 they had abandoned the idea that deregulation would necessarily reduce unemployment. Most empirical studies of the effect of labour market institutions have focused on samples of OECD economies and results have been conflicting, depending not only on the countries in the sample and the period covered but also on the specific measures of employment protection and the estimation methods used. A critical issue in using panel data models is how to deal with potential endogeneity. The Arellano-Bond and the Arellano-Bover/Blundell-Bond GMM estimators have been widely used to address this problem, but these estimators are sensitive to both model specification and estimation strategies. Given the difficulty of deciding precisely how to set up a model as well as which estimation strategy to use, we estimate a range of dynamic models of unemployment in a panel of 27 OECD countries with data from 1990 to 2019 using the OECD’s indicator of employment protection (EPL) and the Fraser Institute’s labour market regulations index (FRASER), which is part of its index of economic freedom. The EPL index measures labour market rigidity while the FRASER index is a measure of flexibility on the labour market so that they are inversely related. We first run random effects and fixed effects models, and then models using the Arellano-Bond (AB) and Arellano-Bover/Blundell-Bond (ABB) approaches in both one-step and two-step variants. Only the AB one-step models pass all diagnostic requirements. We find evidence of strong persistence in unemployment as well as a robust negative relationship between economic growth and unemployment. The average point estimate of the effect of a one standard deviation increase in labour market rigidity as measured by EPL, excluding the countries for which there is no measured change in EPL, is an increase of 0.42 percentage points in unemployment with a 95% confidence interval of (0.27, 0.57). The average point estimate of the effect of a one standard deviation increase in labour market flexibility as measured by FRASER is a reduction of 0.20 percentage points in unemployment with a 95% confidence interval of (-0.36, -0.05).