Nonparametric Estimation of Matching Efficiency and Mismatch: An Application to Japanese Labor Markets

Abstract

I identify significant biases in the traditional Cobb-Douglas function under misspecification of nonadditive, time-varying matching efficiency, and evaluate the finite-sample performance of the nonparametric estimation method of Lange and Papageorgiou (2020). Additionally, I extend the mismatch index by Sahin et al (2014) to a nonparametric framework and develop a computational methodology. Applying the method to Japanese labor market data from government surveys and Hello Work, I analyze changes in matching efficiency, elasticities, and mismatch. Both datasets show declining matching efficiency, aligned with decreasing job-finding and vacancy-filling rates. The unemployment elasticity is around 0.5-0.9 in the Hello Work data and lower in the nationally representative aggregate, while the vacancy elasticity is higher in the nationally representative aggregate. Furthermore, I demonstrate that occupational mismatch is more severe than geographic mismatch, with the Cobb-Douglas mismatch index significantly underestimating the true extent of mismatch.