Many organizations are made up of other organizations that have decided to act collectively — as with research and development consortia, industrial alliances, trade associations, and formal political coalitioins. These collective action organizations (CAOs) have differing strategies: Some are general in scope, while others specialize to a more narrow purpose. We ask what explains the prevalence of generalism and specialism among collective action organizations. We propose a social matching model of collective action, assuming that organizations are only boundedly rational. The model predicts a generalism bias in the ecology of founding and growth among collective action organizations. This outcome is predicted to be parth dependent, however emerging over time according to relatively minor differences in initial conditions. These predictions are supported in an analysis of founding and growth rates among U.S. & R&D consortia registered from the passage of the National Cooperative Research Act in 1984 until March 31, 1993.