Dynamic project selection

We study a normative model of an internal capital market, used by a company to choose between its two divisions’ pet projects. Each project’s value is initially unknown to all but can be dynamically learned by the corresponding division. Learning can be suspended or resumed at any time and is costly. We characterize an internal capital market that maximizes the company’s expected cash flow. This market has indicative bidding by the two divisions, followed by a spell of learning and then firm bidding, which occurs at an endogenous deadline or as soon as either division requests it.