(with Haitao Li, Xiaoxia Ye, and Fan Yu), presented at the 4th Contemporary Issues in Financial Markets and Banking conference (Trent, U.K.), Jan. 2026. Most recent version
Government bond yields display pronounced non-Markov dynamics: moving averages of long-lagged yields substantially enhance the predictability of excess bond returns. To capture these features, we develop a systematic framework for constructing non-Markov Gaussian Dynamic Term Structure Models within the Heath-Jarrow-Morton setting. Compared with existing approaches, our framework is both more flexible and more parsimonious, enabling the estimation of economically meaningful non-Markov effects that improve forecasts of excess returns in-sample and out-of-sample. The models outperform traditional Markov specifications by more accurately capturing bond risk premiums, particularly through improved modelling of the conditional mean of risk factors.
(with Xiaoxia Ye, Fan Yu, and Ran Zhao), scheduled for presentation at the 6th Finance and Accounting 2026 Annual Research Symposium (Westminster, U.K.), Jun. 2026. Most recent version
We examine the role of dealer network centrality in the U.S. corporate bond market by developing a comprehensive set of bond-level measures that explicitly incorporate dealer inventory management. We assess whether bonds traded by more central dealers earn higher mean excess returns and show that the corresponding portfolio returns are largely driven by systematic exposure to the bond market factor. Our measures highlight the importance of idiosyncratic liquidity frictions in OTC bond trading and help rationalize how established illiquidity metrics are transmitted into bond-level credit spreads. In addition, we estimate a variance decomposition model with an explicit interdealer extension and document that firms whose bonds are traded by more central dealers are more responsive to market-wide information flows, underscoring the role of OTC trading networks in corporate bond price formation.
(with Enrico Onali and Xiaoxia Ye), submitted to the World Finance Conference 2026 under peer review. Most recent version
We develop a term-structure model of analysts' earnings forecasts that embeds a latent analyst-firm uncertainty state and yields a structural measure of long-run belief dispersion. We show that conventional cross-sectional dispersion in short-horizon EPS forecasts, while often viewed as scale-invariant, becomes positively associated with stock price levels after controlling for firm fundamentals. In contrast, our model-implied dispersion in long-run valuation beliefs is negatively related to price scale. We reconcile this discrepancy by identifying earnings uncertainty as a latent friction channel that amplifies long-run belief dispersion and interacts with price levels in shaping valuation disagreement. We further demonstrate that analysts' earnings uncertainty significantly predicts investors' ex ante uncertainty, as reflected in option-implied volatility prior to earnings announcements. Overall, our findings show that analyst-based belief dispersion captures firm-level information frictions that are capitalised into equity valuation and the pricing of downside risk, providing a structural complement to conventional measures of forecast disagreement.
(with Xi Chen, Aliki G. Karagrigoriou, and Thao N. Nguyen).