I model positive sorting of entrepreneurs across the high and low value-added segments of the venture capital market. Aiming to attract high-quality entrepreneurs, inefficiently many venture capitalists (VCs) commit to provide high value-added by forming small portfolios. This draws the marginal entrepreneur away from the low value-added segment, reducing match quality in the high value-added segment too. There is underinvestment. Multiple equilibria may emerge, and they differ in aggregate investment. The model rationalizes evidence on VC returns and value-added along fundraising “waves” and when the cost of entrepreneurship falls, and generates untested predictions on the size and value-added of venture capital. 

In a “lemons” market, a shock to gains from trade precedes the buyers' offer. Lower gains exacerbate adverse selection. By trading with intermediaries beforehand, sellers commit not to keep high-quality assets in such states, improving surplus despite impeding efficient use of information. To add value, intermediaries need not possess superior skills or information. Intermediation emerges endogenously if exclusive trading agreements are enforceable. Alternatively, if sellers choose intermediaries to overcome search frictions, welfare decreases with search costs. The theory rationalizes the underwrite-to-distribute model in leveraged loans, and links dealers' market-making activity with the severity of adverse selection in the corporate bonds market. 

Financial institutions (FIs) hedge their portfolio risks using over-the-counter contracts, becoming connected. We derive conditions under which, in the “bad” state of the economy, the number of defaults is arbitrarily high and (almost) all defaults are due to contagion. In these cases, increasing the value of collateral necessarily causes a negligible change in each FI's capacity to absorb counterparty losses, yet it alleviates systemic risk. However, when assets used as collateral are volatile: (i) margins can increase defaults precisely in those states in which defaults concentrate, exacerbating crises and (ii) privately optimal margins can be inefficiently high, motivating maximum margin requirements. 

In OTC markets, privately informed asset owners face investors who differ in expertise: the ability to evaluate assets. Investors incentive to use expertise is motivated by rent-seeking, and depends on their beliefs about average assets quality—the state of the economy. I characterize the optimal policy to disclose the state of the economy. The policy always improves over, and can generate large gains against, the fully observable state benchmark. It affects equilibrium expertise, which is inefficient due to the cream-skimming externality and non-internalized effects on assets origination incentives. The policy does not necessarily aim at minimizing the probability of a market dry-up. 

We model an economy where managers create value through their ability to learn at an intermediate stage about the intrinsic profitability of a risky investment. Managers are heterogeneous in their ability to extract information from experiments, and care about their reputation. Their incentive to take on risk is distorted by career concerns, and can result in under or over risk-taking. This is determined by whether discarding a risky project following the experiment is more typical of better managers. We also show that, in an effort to appear informed, managers at the same time reduce risk-taking when investments are only marginally positive NPV, and engage in excessive costly experimentation when investments are more profitable.