Reliance on a “representative firm” approach in studying industrial behavior during the Great Depression obscures economically interesting patterns. A newly discovered data source lets us form and study an establishment-level panel dataset on the motor vehicles industry, one of the largest in 1929. Substantial intraindustry heterogeneity led to large composition effects in employment, output, and productivity: the large number of plants that shut down were unlike the continuing ones. Oddly, output does not seem to have shifted among continuing producers to the relatively low-cost ones. Reconciling these should illuminate links between industrial organization and macroeconomics.