This paper presents an empirical test of a subclass of poverty traps hypotheses. The test is based on the observation that the nonconvexities in the production function necessary to generate multiple equilibria need only be present in the region between the equilibria. Increasing returns should therefore be strongest when the economy is transitioning between steady states than when it is at or near one of those steady states. I implement this idea by estimating the degree of increasing returns during growth accelerations and growth transitions for a panel of developing and developed economies using UNIDO's Database of Industrial Statistics. I find no evidence of systematic differences in economies of scale between transition and non-transition episodes, shedding doubt on the idea that increasing returns in manufacturing generate poverty traps.