高技術勞動市場的性別差距：以臺灣西醫師為例

Abstract

本文探討臺灣醫師在生產力與薪資上的性別差異，使用涵蓋 2000–2020 年之間全民健康保險（健保）系統的紀錄。健保資料具備少見的特性：完整涵蓋所有受僱的西醫師，於所有醫療院所中採用統一的產出衡量指標（健保點數），並同時紀錄醫師的工作場所（醫院）以及該場所內的詳細工作任務。這樣的獨特背景，使我們得以首次對性別差距進行「工作場所分派」與「任務層級生產力差異」的聯合分解分析。

我們發現，原始性別薪資差距為 26%，生產力差距達 60%。若使用邊際分解，結果顯示個人異質性在病例數量上的差異是主要因素；然而，我們的聯合分解框架揭示，超過一半的總體生產力差距來自醫院固定效果對病例數量的影響：女性醫師系統性地被分派至病患量較低的醫院，限制了她們的整體產出。病例表現的性別則差異極小，而可觀測特徵對差距的解釋力有限。

本研究有兩項主要貢獻。首先，我們展示了如何利用豐富的行政微觀資料，對跨工作場所與工作場所內的生產力差距進行聯合分解，此方法在其他勞動市場中幾乎難以實現。其次，我們針對 Goldin（2014）所強調的高技術、績效導向職業之一提供了實證證據，顯示造成觀察到性別差距的主要原因是制度性職位分派，而非每件病例的生產力差異。

This paper investigates the sources of gender disparities in productivity and earnings among physicians in Taiwan, leveraging 2000–2020 administrative records from the National Health Insurance (NHI) system. The NHI database offers a rare combination of features: it covers the universe of employed Western physicians, applies a standardized output metric (NHI points) across all institutions, and records both the workplace (hospital) and the detailed tasks performed within each workplace. This unique setting allows us, for the first time, to jointly decompose gender gaps into across-hospital sorting and within-hospital task-level productivity differences.

We document a raw gender wage gap of 26% and a productivity gap of 60%. While marginal decompositions suggest that individual heterogeneity in case volume is the dominant factor, our joint framework reveals that more than half of the total productivity gap arises from hospital fixed effects on case volume: female physicians are systematically placed in lower-volume hospitals, directly limiting their overall output. Differences in case performance are minimal, and observables account for little of the gap.

Our contributions are twofold. First, we demonstrate how rich administrative microdata can be used to perform a joint decomposition of productivity gaps across and within workplaces, a task rarely feasible in other labor market settings. Second, we provide evidence on gender inequality in one of the high-skilled, performance-based professions emphasized by Goldin (2014), showing that institutional placement, instead of differences in per-case productivity, drives most of the observed gender gap.