Workforce Dynamics and the Role of Firms in U.S. Childcare, 2000-2025

• In Event: Empirical Evidence on Child Care Supply and the Child Care Workforce

Abstract

Turnover is a pervasive feature of the U.S. childcare workforce, with consequences that reach well beyond the sector itself. Frequent worker departures disrupt the stable caregiver-child attachments that support healthy early development, while imposing substantial costs on firms through unplanned vacancies, recruitment, and training. Yet despite broad recognition of the problem, the drivers of childcare turnover—and how they vary across workers, firms, and places—remain poorly understood, limiting the design of effective retention policy.

This paper asks three questions: How stable is the childcare workforce at quarterly and annual horizons? Where do workers come from and where do they go when they leave, and how has the pay premium relative to outside options evolved? And which individual, geographic, and firm-level characteristics most strongly predict turnover, and how do these predictors vary across regions and over time? Building on Cunha and Lee (2023) and Fares, Markowitz, and Bassok (2026), we leverage matched employer-employee records from the Longitudinal Employer-Household Dynamics (LEHD) database covering 29 states from 2000 to 2025—the most comprehensive administrative dataset assembled for this workforce to date—linked within the restricted Census environment to American Community Survey records that provide rich individual-level demographic and household characteristics unavailable in LEHD alone.

Our analysis proceeds in three steps. First, we establish stylized facts on workforce instability, including the share of workers for whom childcare is the primary job, quarterly survival rates at 1, 2, 3, 4, 8, and 12 quarters, and the average share of stable workers each quarter. Second, we document worker flows into and out of the sector alongside the evolving pay premium relative to destination industries, providing a longitudinal portrait of how workers' outside options have shifted. Third, using the LEHD-ACS linked sample, we estimate logistic regressions that decompose the correlates of turnover into individual predictors (e.g., race, age, marital status, household income, education, experience), geographic predictors (e.g., state median income, market concentration, urban/rural status), and firm predictors (e.g., average pay, size, chain affiliation). Results have not yet been disclosed from Census but will be available for presentation at the conference. By quantifying how the correlates of turnover vary across regions and across a quarter-century, this study provides the first granular empirical foundation for policymakers seeking to target retention interventions where they can most effectively stabilize the childcare workforce.

Author

• Katharine Sadowski, Stanford University
  Presenting Author