Empirical essays on credit risk based asset pricing

Xu, Yaofei (2021) Empirical essays on credit risk based asset pricing. PhD thesis, University of Glasgow.

Due to Embargo and/or Third Party Copyright restrictions, this thesis is not available in this service.


The credit risk is of significant importance in the current financial market. For instance, unlike most financial institutions whose long-run performance is affected by the tail-risk induced extreme negative return, hedge funds focusing on trading credit risk are able to obtain tremendous return during critical periods, such as 2020 Covid-19 outbreak, 2012 European debt crisis and 2008 global financial crisis. These hedge funds mainly choose Credit Default Swap (CDS) or Deep Out-Of-The-Money (DOOM) put option to trade the credit risk, utilizing the leverage. Given this interesting phenomenon, this doctoral thesis aims at exploring some related topics in credit-risk-based asset pricing.
The first chapter of this thesis provides a comprehensive review of credit-risk-based asset pricing. The first part briefly introduces definitions of CDS and put option, their connection and related concepts. The second part reviews credit risk related structure models, reduced-form models, CDS-implied stock volatility, default boundary on stock/asset level as well as risk factors driving the interaction between stock, bond, CDS and option markets. The third part identifies current research gap and illustrates my motivations for each topic one by one.
The remaining four chapters investigate four topics about credit-risk-based asset pricing: volatility information difference between CDS and option markets and option pricing, CDS-inferred stock volatility and trading on its deviation with option implied volatility, default boundary at stock price level and co-movement between CDS market and option market.
Chapter 2 examines the information content difference between the credit and option markets. I extract volatilities from corporate CDSs and options. The volatility difference between two markets is positively related to future option returns. I rank firms based on the normalized volatility spread and analyze the returns for straddle portfolios. A zero-cost trading strategy that long (short) in the portfolio with the largest (smallest) spread generates a significant average monthly return, even after controlling for stock characteristics, traditional risk factors, and moderate transaction costs.
Chapter 3 explores trading strategies based on the cross-market information between CDS market and option market. I propose a new measure of CDS implied volatility (CIV) inferred from a reduced-form model called as Unit Recovery Claim Theory (URC) (Carr and Wu, 2011), mapping the CDS spread into a CDS-inferred American put option through URC theory and back out the CIV through numerical method with 200-step binomial tree model. My CIV is more like an estimator of option implied volatility (OIV) of long-term Deep OTM put option, which has a higher correlation with corresponding OIV and higher standard deviation compared with other Merton-style CIVs. Then, I use the standardized spread between CIV and OIV as the deviation to trade CDS contracts and OTM put options after considering the biased estimator of CIV on OIV and mean-reverting feature of the spread between CIV and OIV. The resulted long-short quintile CDS portfolios (OTM put option portfolios) can achieve a significant annualized return with high Sharp ratio, after controlling for jump and volatility risk factors.
Chapter 4 investigates the possible default boundary at the stock price level. I extract the default boundary jointly from CDS and option market at the aggregate level. Then, I apply the Unscented Kalman Filter to extract the time series of hidden market-level default boundary and default-level implied volatility from the implied volatility surface and CDS spread. The implied volatility surface is modelled through a deterministic linear function. The CDS spread is modeled through a reduced-form Black-Cox model, or URC-theory-based model alternatively. Finally, I check the bankrupt firms between 2002 and 2017 with my default boundary at the stock level. The default boundary calibrated by the reduced-form Black-Cox model can provide a closer estimator on historical one, while that inferred from URC theory is much more upward biased than the historical one. I argue that my default boundary at stock price level can be a promising indicator predicting firm bankruptcy.
Chapter 5 studies the co-movement between CDS market and option market at the index level. I firstly prove that parameters in N-S model are consistent with those in no-arbitrage models when modeling CDS curve, which lends further support to the validity of decomposing CDS term structure via the N-S model. Given that, I apply Unscented Kalman Filter to extract three components (level, slope and curvature) of CDS term structure through the N-S model and whole volatility surface through the deterministic linear function respectively. I examine the co-movement between CDS market and option market, finding that after controlling for stock return, the strong relation between CDS market and option market in terms of the three components mostly disappears or significantly decreases. Additionally, I also find that the co-movements between CDS market and option market is stronger during the financial crisis.
Chapter 6 draws the conclusions and identifies gaps for further research on credit-risk based asset pricing.

Item Type: Thesis (PhD)
Qualification Level: Doctoral
Keywords: option, CDS, stock, default probability, portfolio.
Subjects: H Social Sciences > HG Finance
Colleges/Schools: College of Social Sciences > Adam Smith Business School
Supervisor's Name: Shi, Dr. Yukun and Stasinakis, Dr. Charalampos
Date of Award: 2021
Embargo Date: 8 February 2024
Depositing User: Yaofei Xu
Unique ID: glathesis:2021-82009
Copyright: Copyright of this thesis is held by the author.
Date Deposited: 15 Feb 2021 13:35
Last Modified: 15 Feb 2021 13:35
Thesis DOI: 10.5525/gla.thesis.82009
URI: http://theses.gla.ac.uk/id/eprint/82009

Actions (login required)

View Item View Item