Abstract: This paper investigates how firms invest under demand uncertainty focusing on the role of information. I develop a dynamic oligopoly model that allows uncertainty about the demand process: firms do not know the true parameters in the demand process, but form and revise expectations about demand based on information available at each decision-making moment. I estimate the model using firm-level data from the container shipping industry. The analysis shows that learning amplifies investment cycles and raises the correlation between investment and demand, which helps us explain the boom-bust investment patterns. I examine how learning interacts with firms' strategic incentives through counterfactual analysis. The results indicate that strategic incentives increase both the level and the volatility of investment and that learning intensifies these forces. I show that the regulator's modeling choice for firms' expectations has important policy implications, namely in merger evaluation.
The Competitive Effects of Information Sharing (with John Asker, Chaim Fershtman, and Ariel Pakes)
NBER Working Paper No. 22836.
Abstract: We investigate the impact of information sharing between rivals in a dynamic auction with asymmetric information. Firms bid in sequential auctions to obtain inputs. Their inventory of inputs, determined by the results of past auctions, are privately known state variables that determine bidding incentives. The model is analyzed numerically under different information sharing rules. The analysis uses the restricted experience based equilibrium concept of Fershtman and Pakes (2012) which we refine to mitigate multiplicity issues. We find that increased information about competitors' states increases participation and inventories, as the firms are more able to avoid the intense competition in low inventory states. While average bids are lower, social welfare is unchanged and output is increased. Implications for the posture of antitrust regulation toward information sharing agreements are discussed.