Shared information and the limits of firm performance

Abstract

This paper develops a microfounded model in which integrating decisions through a shared signal improves local inference but weakens the diversification of decision errors inside the firm. Shared information lowers posterior variance at the decision level while raising the covariance of errors across decisions, generating conditional and potentially non-monotonic effects on aggregate firm performance. Digital integration provides the motivating application, but the mechanism is more general: changing the informational structure of the firm changes the extent to which errors diversify across choices. The model delivers an interior optimum and a structural interpretation for heterogeneous productivity effects and persistent partial adoption.