We investigate the security design problem in an initial public offering (IPO) in which some investors are better informed than others about the prospects of the firm, resulting in a winner's curse problem. To raise capital, the owners of the firm must underprice the securities they issue in order to compensate the less informed invest ors for their willingness to participate in the issue. In this context, we first show that firms can sometimes lower the cost of going public by using unit IPOs, in which equity and warrants are combined into a non-divisible package. Because warrants are less sensitive to low cash flow realizations, unit IPOs tend to be valuable to firms that face large downside risks or whose uncertainty revolves around the eventual performance of their assets in place. Second, we show that firms may be able to completely eliminate the winner's curse problem by making the warrants callable. Such a first-best scenario is possible when a firm's growth potential is sizeable even in bad states of the world, as the callability feature of the warrant allows the firm to dynamically create payoffs that are insensitive to the investors' private information.

We introduce and study two-sided matching with incomplete information and interdependent valuations on one side of the market. An example of such setting is a matching market between colleges and students in which colleges receive partially informative signals about students. Stability in such markets depends on the amount of information about matchings available to colleges. We show that when colleges observe the entire matching, a stable matching mechanism does not generally exist. When colleges observe only their own matches but not those of other colleges, a stable mechanism exists if students have identical preferences over colleges, but may also fail to exist if students have different preferences.

We consider multidimensional cheap talk by a biased expert, e.g., comparison of different spending proposals by an industry lobbyist, evaluation of different stocks by a sell-side analyst, or analysis of different topics by a biased newspaper. When the expert's motives are sufficiently transparent, we find that cheap talk is credible and influential even when the expert strongly favors one issue over another, e.g., negative advertising against a rival by a political or marketing campaign. The transparency condition holds for state-independent preferences and for standard Euclidean preferences as the expert's bias toward a higher (or lower) action on each issue becomes arbitrarily large.

Asymmetric information regarding project prospects causes dilution, leading to adverse selection and inefficiencies in the market for new investments. However, if the market obtains information about the firm over time, issuing callable convertible securities with restrictive call provisions is optimal. Even when the market's information is noisy, such securities can be designed to make the payoff to new claimholders independent of the private information of the manager. The restrictive call provision serves as a commitment device, enabling the manager to call only when the stock price rises in the future. This solves the dilution and adverse selection problem costlessly. The same first-best efficient outcome can also be implemented by issuing floating price and mandatory convertibles.

In symmetric common value auctions where bidders differ ex--post in information quality, a seller may benefit from imposing a ceiling on allowable bids. By reducing the winner's curse facing poorly informed bidders, a ceiling encourages them to bid aggressively. This may reduce information rents earned by better informed bidders, yielding the seller higher expected revenues compared to selling the object in a standard ascending auction or at a fixed posted price. Such a ceiling may be explicit (a firm commitment not to accept bids above the ceiling) or implicit (a credible threat not to honor the outcome if anyone bids higher than the ceiling). Either situation can be interpreted as one where the object is offered for sale at a fixed price or the best offer.

We characterize efficient equilibria of common interest voting games with privately informed voters and study the implications for Condorcet jury theorems. We show that larger juries can do no worse than smaller ones and derive a simple necessary and sufficient condition for asymptotic efficiency of different voting rules. This condition implies that the unanimity as well as near unanimity rules are asymptotically inefficient regardless of equilibrium selection. However, if the signal distribution fails a non-degeneracy condition the unanimity rule dominates any other rule. Finally, if signals are conditionally independent, full information equivalence can be exactly achieved for any rule that allows the divisibility of individual votes, and for any finite number of voters.

As participants reveal their orders to a market, prices are discovered and trading volume (quantity) is "found" (i.e. books get deeper). We refer to this as "bookbuilding," a non-trivial process that depends on a market's rules of order handling and order information disclosure. The paper shows that, for large participants, uncertainty about the profile of other traders in the market distorts trading volume and reduces the gains from trading. These inefficiencies are mitigated when participants can place multiple (scaled) orders and when submitted orders are displayed in an open book. In effect, the open book permits participants to engage in non-binding communication, thus facilitating efficient bookbuilding. These findings are robust to the inclusion of order submission costs and the ability to costlessly withdraw previously placed orders.