RIF quantile-band decompositions: An application to gender earnings gaps among the self-employed and paid employees

This paper provides the first comparative analysis of gender earnings gaps inself-employment and paid employment in South Korea. Using data from the 2023 Korean Working Conditions Survey, we introduce a simple extension of the recentered influence function (RIF) framework to decompose gender gaps in monthly earnings among low-, middle-, and high earners. By averaging RIFs within quantile bands, we address the instability of single-quantile estimates caused by distributionalbunching and obtain more stable inference. We find that gender gaps are larger among paid employees than among the self-employed—contrary to findings for other countries—particularly among low earners. In paid employment, these larger gaps among low earnersare driven by composition effects related to work hours and industrial segmentation. In self-employment, gender gaps are flatter across the distribution, with composition effects increasing at higher earnings levels, partly reflecting gender differences in business scale. JEL Classification: I31, J22, J16, J28