In the first chapter, I estimate dynamic factors from the term structure of credit spreads and the term structure of equity option implied volatilities, and I provide a comprehensive characterization of the dynamic relationships among those credit spread factors and equity volatility factors. I find strong evidence that the volatility factors, especially the volatility level factor, Granger cause credit spread levels, confirming the theoretical predictions of Merton (1974) in a significantly richer and more nuanced environment than previously achieved. Simultaneously, I also find evidence of reverse Granger causality from credit spreads to equity volatility, operating through the slope factors, consistent with the market microstructure literature such as Fleming and Remolona (1999a, 1999b), which finds that price discovery often happens first in bond markets. Hence my results extend and unify both the corporate bond pricing and market microstructure literatures, deepening our understanding of stock and bond market interaction and suggesting profitable trading strategies. In the second chapter, which is a joint work with Frank X. Diebold, we study the dynamics of the U.S. Treasury yield term structure by applying the Nelson-Siegel models introduced in Diebold and Li (2006) and its arbitrage-free version developed by Christensen, Diebold, and Rudebusch (2007). We analyze in-sample by testing the risk-neutral restrictions, which establish the absence from arbitrage, on the Nelson-Siegel factors estimated under the physical measure. We thus compare the term structure modeling between the risk-neutral measure and the physical measure. Specifically, we show that the risk-neutral restrictions on the factor dynamics are well-satisfied, and those factors have the following properties: 1) the level factor is a unit-root process and does not affect the other two factors; 2) the slope and curvature factors are mean-reverting processes that revert at the same rate; 3) the curvature factor forecasts the slope factor, but not conversely. Moreover, we find that utilizing these restrictions can improve out-of-sample forecast performance.
