We examine the cyclicality of employer-sponsored work-based training using Registered Apprenticeship in the United States. Unlike schooling, which often expands in downturns as individuals respond to lower opportunity costs, apprenticeship depends on firms' willingness to create and finance training positions. Its cyclical behavior is therefore theoretically ambiguous and may differ across segments. We combine administrative data on apprenticeship starts from RAPIDS with county-industry real GDP from 2002 to 2024 and estimate how local economic fluctuations relate to apprenticeship entry. Exploiting variation within county-industry-year cells, we also examine heterogeneity by starting wages, unionization, and gender. We find that apprenticeship starts are procyclical overall: stronger local economic growth is associated with more new apprentices. This average effect masks substantial heterogeneity. Procyclicality is concentrated in lower- and middle-wage segments and weakens as starting wages rise. Unionization significantly amplifies the cyclical response of apprenticeship starts. Finally, men's apprenticeship starts are procyclical, whereas women's are acyclical once comparisons are made within local industries and over time. These results show that the cyclicality of work-based training is shaped by labor-market institutions and that business-cycle shocks affect both the volume and the distribution of training opportunities.