The Multiplicity of the Labor-market Structure and Gender Inequality: Decomposition analysis of inequality with a finite mixture model (Japanese)

This paper employs a statistical model recently developed by the author that combines a decomposition analysis of inequality with a finite mixture model that assumes multiple latent classes in the outcome equation. The method permits a decomposition analysis of gender inequality in earnings for each latent class with a distinct income determination function, thereby clarifying the multidimensionality of the mechanism that generates gender inequality in earnings. The analytical results show that there are three latent classes with different income determination functions. (1) In latent class 1 of the labor market, where the average wage is highest, there are relatively many more college graduates and employees at large companies, and the majority are regular employees. For this class, variables related to long-term employment, such as age and years of service for the current employer, are the main mediating factors of the gender gap in earnings. (2) In latent class 2 of the labor market, where the average income is lower, there are relatively many more high school graduates and employees at small and medium-sized enterprises, and about one-third are non-regular employees. For this class, variables related to working hours, such as long working hours and the proportion of full-time employees, are the main mediating factors of the gender gap in earnings. (3) In latent class 3, where the average income is lowest, there are relatively many more junior high school graduates and employees at micro enterprises of less than 30 employees, and more than half are non-regular employees. For this small class, the proportions of managers and part-time workers are the main mediating factors of the gender gap in earnings. Furthermore, (4) employment status and educational attainment mostly create gender inequality indirectly through between-class disparities in earnings, rather than within classes. In other words, the gender gap in earnings also arises because the high proportions of regular employees and college graduates, who are more prevalent among men in Japan, are associated with latent class 1, which has the highest average income. This paper also uses these new findings to examine the segmented labor-market theory and discuss the consistency/inconsistency of the analytical results with the theory.