Interpreting the Effectiveness of Active Labor Market Policies in Combating Unemployment in OECD Countries Using a Causal Machine Learning Approach

Abstract

This study examines the effects of Active Labour Market Policies (ALMP) expenditures on unemployment in OECD countries. Using a large country-year panel dataset covering the period 2000–2022, policy effects were analyzed using causal machine learning methods. In this study, ALMP expenditures were treated as a continuous treatment variable expressed as a percentage of GDP, the unemployment rate was used as the outcome variable, and population size was included in the model as a confounding variable using logarithmic transformation. The average policy effect was estimated using the Double Machine Learning (DML) approach, while conditional mean treatment effects (CATE) were calculated using the Causal Forest method to reveal how the effects differed across countries and contexts. The findings show that ALMP expenditures significantly reduce the unemployment rate on average across the OECD. However, it was found that policy effects are quite heterogeneous across countries and contexts, with ALMP expenditures having much stronger unemployment-reducing effects in some countries, while marginal effects remained limited in others. The results highlight the inadequacy of a uniform policy approach in evaluating active labor market policies and emphasize the importance of context-sensitive, targeted policy designs.