Our paper o�ers a minimalist model of a run on a financial market. The prime ingredient is that each risk-neutral investor fears having to liquidate after a run, but before prices can recover back to fundamental values. During the run, only the risk-averse market-making sector is willing to absorb shares. To avoid having to possibly liquidate shares at the marginal post-run price—in which case the market-making sector will already hold a lot of share inventory and thus be more reluctant to absorb additional shares—all investors may prefer selling their shares into the market today at the average run price, thereby causing the run itself. Consequently, stock prices are low and risk is allocated ineciently. Liquidity runs and crises are not caused by liquidity shocks per se, but by the fear of future liquidity shocks.

The first chapter of my dissertation explores the roles “quants” or technically trained workers play at hedge funds. I quantify the impact of hiring “quants” on hedge fund strategy and risk taking by exploiting H-1B lottery results and a policy shock that significantly reduced the future supply of foreign high-skilled labor in the U.S. I find that the H-1B visa program allows hedge funds to pursue strategies that are more quantitative (e.g., hedging, systematic trading, etc.) as opposed to fundamentals-based (e.g., event driven). Although there is evidence of substitution between fund strategies, I also find that hedge funds overall become more diversified, both in terms of regional focus and fund investment style.

Traditional theories on corporate financial policy mostly focus on firm-specific determinants and assume that capital structure choices are made independently of the actions of peer firms. The second chapter of my dissertation introduces a theoretical model that embodies the dynamic of how peer behavior affects a firm’s optimal choice of debt level. Specifically, peer effect is captured by a measure of how much the liquidation value of assets are adversely impacted if peer firms choose to liquidate simultaneously, the effect of which is more prominent for industries with highly specialized assets – hence low asset redeployability. The theoretical model predicts that a firm’s optional debt level is lower when the liquidation prices of its assets are expected to be more severely impacted by the concurrent liquidation of peer firms. Moreover, in those situations where peer firms’ cash flows are more closely correlated, the optional debt level will be higher. Individual firms’ attempts to maximize their own utility lead to an industry-wide over leverage, further exacerbating the risk of general crises in an already highly correlated industry.

The third chapter of my dissertation explores the potential of using deep learning models to enhance equity trading strategies such as momentum and reversal trading. A deep feed forward neural network model (DFN) is fed with a training data set (1965-2000) that uses the rolling Z-scored cumulative returns of various horizons as predicative variables. With the model parameters generated by the training set, the model is applied to the validation set (2000-2016) to generate a signal for each stock predicting its likelihood of outperforming the cross-sectional medium. Based on this model-predicted signal, I have constructed an long-short investment portfolio out of the validation data set. The deep learning portfolios yield an annualized return of over 20 percent and generate significantly large alphas over commonly used factor models.

We consider a firm with two investment projects (divisions) each run by a manager who can provide (i) (unverifiable) information about the quality of either or both projects and (ii) (unverifiable) access to valuable resources that can enhance the cash flows of either or both projects. We then examine the extent to which capital allocation and managerial compensation schemes can be designed jointly to mitigate the information and incentive problems within the firm. We find that the firm underinvests in capital in both projects (relative to the first-best level), optimal managerial profit-sharing increases in the quality of both projects, and managers underinvest resources in both projects. Interestingly, the underinvestment problem in one project is less severe when the other project is higher quality. This effect will be stronger when the moral hazard problem is more acute and project returns are more highly correlated. We also predict that multi-division firms will invest more (less) than single-division firms when other divisions in the firm perform well (poorly).

In the first chapter, I apply machine learning techniques to numerically solve high-dimensional continuous time models in finance. Traditional methods rely on finite difference schemes for solutions to partial differential equations. By approximating the solution with a deep neural network, I am able to leverage the computational efficiency of neural networks and batch gradient descent to accurately compute solutions involving many state variables. I demonstrate the accuracy and efficiency of this method for Black-Scholes options pricing problems and dynamic programming problems in up to 50 spatial dimensions, far beyond the capability of grid methods. I also develop a solution method to mean field game type problems, where both a value function and a distribution function must solve a system of differential equations, utilizing mixture density networks.

In the second chapter (with Ivo Welch), we develop a model where buyers prefer local over lower-cost vendors even in the absence of direct preferences, taxes, subsidies, contracts, sanctions, information asymmetries, audits, etc. Instead, they prefer locals because they internalize the fact that local agents will in turn be more likely to buy from them in the future. Local sellers understand that buyers' preferences give them limited local market power, and therefore raise their prices and earn surplus in equilibrium. Our model can explain how voluntary reciprocity among subsets of identical agents can sustain itself, and how ex-ante identical goods from ex-ante identical sellers can acquire and maintain sustainably differentiated prices.

In the third chapter (with Antonio Bernardo and Ivo Welch), we develop a model where firms with lower leverage are not only less likely to experience financial distress but are also better positioned to acquire assets from other distressed firms. With endogenous asset sales and values, each firm's debt choice then depends on the choices of its industry peers. With indivisible assets, otherwise identical firms may adopt different debt policies---some choosing highly levered operations (to take advantage of ongoing debt benefits), others choosing more conservative policies to wait for acquisition opportunities. Our key empirical implication is that the acquisition channel can induce firms to reduce debt when assets become more redeployable. This article has been accepted for publication and is forthcoming in the Journal of Financial and Quantitative Analysis.

We develop a model of a two-division firm in which the “strong” division has,on average, higher quality investment projects than the “weak” division. We show that the firm optimally biases its project selection policy in favor of the weak division and this bias is stronger when there is a greater spread in average project quality. The cost of such a policy is that the firm sometimes funds an inferior project but the benefit is that it motivates the manager of the strong division to set (and meet) more aggressive cash flow targets.

We consider the problem of motivating privately informed managers to engage in entrepreneurial activity to improve the quality of the firm's investment opportunities. The firm's investment and compensation policy must balance the manager's incentives to provide entrepreneurial effort and to report her private information truthfully. The optimal policy is to underinvest (compared to first-best) and provide weak incentive pay in low-quality projects and overinvest (compared to first-best) and provide strong incentive pay in high-quality projects. We also show that, unlike the standard agency model, uncertainty and incentives can be positively related.