| Summary: | In Chapter 1, I investigate abnormal behavior in individual stocks using two decades of high-frequency U.S. stock market data. I identify hundreds of thousands of short episodes where stocks exhibit ”explosive” behavior, deviating from the unit-root null hypothesis. These phenomena span multiple days, differ from typical return movements, and affect a wide range of stocks, including liquid and large-cap stocks. Explosive episodes account for a considerable portion of stocks’ idiosyncratic variance. These are transitional episodes with partial reversal, providing predictable returns, setting them apart from large overnight and high-frequency jumps. I analyze stocks and their susceptibility to explosive behavior in connection with aggregate market fluctuations. While downward explosions tend to cluster among stocks and are more pro-cyclical, upward explosions appear as an idiosyncratic phenomenon. Explosive episodes involve significant buying and selling pressure along with trading volume. To explain explosive price movements, the paper introduces a model involving inelastic buyers, insiders, and competitive sellers. It emphasizes the role of explosions in the price discovery process and addresses the observed reversal. The frequency, severity, and reversal of explosiveness are explained by the expected size of inelastic demands, the knowledge possessed by a representative insider, and the frequency of seeing both in the market. Using short interest dissemination dates, empirical tests validate the model’s predictions, indicating a higher likelihood of explosive behavior in stocks with substantial reported short interest. Chapter 2, joint work with Roman Sigalov, studies the stability of factor structure by analyzing its variation on different market events. We start by documenting variation in distributions, means, volatilities, and correlations in a set of characteristics managed longshort portfolios on the weeks with large market moves, leading earnings announcements, and FOMC announcements with unexpected shocks to interest rates. This variation manifests in differences in factors extracted using characteristics based on statistical methods that we document using Instrumented PCA. The factor structure shows variation in the factor loadings and in the distribution of factors itself. We propose two ways of capturing eventspecific variation in the factor structure. The first method, Treatment-IPCA, estimates orthogonal factors specific to the events we consider. We find significant premia associated with some treatment factors. The second method, Boosted- IPCA allows us to test the differential importance of firm characteristics in describing the cross-section of stock returns on market events relative to base periods. Chapter 3 explores market making under imperfect competition. Using a dataset on individual-level intraday market making in an option market, I demonstrate a significant level of concentration in liquidity provision across options. I propose a dynamic duopoly market making model wherein inventory distribution shapes agents’ strategic behavior and observed liquidity provision on best quotes. I characterize the solution up to the optimal actions, enabling straightforward numerical solutions under both non-cooperative and cooperative equilibria. Qualitatively, the equilibria differ under various sets of parameters, allowing for a wide range of possible inventory and liquidity dynamics, some of which are non-trivial. Tight capital constraints and a high rate of order arrival lead to violations of a monotonic principle. In particular, this results in observing “resting” market maker behavior when an agent does not provide liquidity. Conversely, relaxed constraints lead to a more standard equilibrium where market makers reduce inventory imbalances. Analyzing a grim-trigger non-Markov equilibrium, I find that collusive behavior among market makers increases liquidity prices but reduces their variability.
|
|---|