My dissertation attempts to understand the determinants of two asset prices, bank loan and corporate stocks. The thesis contains three chapters.

Chapter 1 studies a unique determinant of bank loan pricing, i.e. the protection of banks’ proprietary information, in a natural experiment setting. We find that a better protection of banks’ proprietary information enhances a reciprocal, long-lasting lending relationship. After banks’ trade secrets are protected by the inevitable disclosure doctrine (IDD), banks offer loans with lower interest rates and longer maturities. This effect is more pronounced for relationship loans. Meanwhile, a relationship bank is more likely to be chosen as the lead loan underwriter after IDD adoptions. Furthermore, banks whose trade secrets are better pro...[ Read more ]

My dissertation attempts to understand the determinants of two asset prices, bank loan and corporate stocks. The thesis contains three chapters.

Chapter 1 studies a unique determinant of bank loan pricing, i.e. the protection of banks’ proprietary information, in a natural experiment setting. We find that a better protection of banks’ proprietary information enhances a reciprocal, long-lasting lending relationship. After banks’ trade secrets are protected by the inevitable disclosure doctrine (IDD), banks offer loans with lower interest rates and longer maturities. This effect is more pronounced for relationship loans. Meanwhile, a relationship bank is more likely to be chosen as the lead loan underwriter after IDD adoptions. Furthermore, banks whose trade secrets are better protected can sell loans at higher loan bidding prices and lower bid-ask spreads in the secondary loan market. Our paper highlights the importance of bank proprietary information protection.

Chapter 2 investigates the stock price movement associated with a behaviour bias in the merger and acquisition market in the US. Contrast effect is a behavior bias where a decision-maker will change his subsequent perceptions based on the previous context. We document that investors’ stock market reaction to M&A deals announcement is adversely affected by the deal premium of the closest prior day M&A deals under contrast effect. This adverse relationship is confined to the interaction with closest prior day deals and the mispricing reverses in post-announcement period, consistent with a perceptual error, but not learning effect or information transmission from the prior day deals. Associated with high prior premium, the probability of offer price upward revision increases, and deal success rate decreases. Subsample tests show that the effect is concentrated among target firms with less sophisticated investors, target firms with smaller size, same industry prior deal and high-value prior deals.

Chapter 3 dissects bidirectional Long short-term memory network(LSTM), a state-of-the-art technique for time series prediction. I implement and analyze the effectiveness of bidirectional LSTM in predicting out-of-sample stock directional movements in Chinese stock market from 2009-2017. Utilizing only five dimensional features: daily stock price, daily turnover, daily volatility, deviation from the daily average price, percentage change from opening to closing price, my model achieved significantly good performance in predicting next day stock movement. A long strategy uses the prediction from the bidirectional LSTM outperforms traditional Random Forest and Logistic regression with an average annual return of 28% and a Sharpe Ratio of 0.8 after transaction cost.

My study explores the rational and behaviour hypothesis in explaining loan pricing and stock pricing cross-sectionally. My research indicates that behavioural bias plays a role in predicting cross-sectional asset pricing. Weak market efficiency is challenged.