An Interval Regression Analysis for Tenures of Japanese Elder Care Workers Using Matched Employer-Employee Data

This paper analyzes job tenures of Japanese elder care workers in the home care service sector, using an econometric framework that can fully utilize information of available data. This sector reveals a large betweenfirm difference in workers' separation rates, despite a regulation policy that induces a limited wage dispersion. I rationalize this puzzling observation by a screening model in which firms try to avoid adverse selection caused by information asymmetry regarding workers' motivation. My model induces a separating equilibrium in which several firms cover training costs for general human capital accumulation of workers. To examine a testable implication of my screening model, I construct an interval regression model using cross-section data with matched employer-employee information. A standard Bayesian estimation provides empirical results that support my economics model.