City Research Online

Abstract

This thesis consists of empirical and theoretical studies on the liquidity of bond markets.

In the first study, we present an extended model for the estimation of the effective bid-ask spread that improves the existing models and offers a new direction of generalisation. The quoted bid-ask spread represents the prices available at a given time for transactions only up to some relatively small trade size. Trades can be executed inside or outside the quoted bid-ask spread. Thus, we extend Roll's model to include multiple spreads of different sizes and their associated probabilities. The extended model is estimated via a Bayesian approach, and the fit of the model to a time series of a year of corporate bond transaction data is assessed by a Bayesian model selection method. Results show that our extended model fits the data better.

Our second study examines the relationships between different liquidity proxies and the non-default corporate yield spread as well as the effective bid-ask spread. We first separate the non-default component of bond spreads from the default one by using the information contained in credit default swaps. We then apply our state-space extension of the Roll model to disentangle the unobservable non-default yield spread from the effective bid-ask spread. The empirical results show that the non-default yield spread has a nonlinear relationship with time to maturity and a positive correlation with the bid-ask spread as well as with the default risk, and therefore may reflect the future expected liquidity. We find that the effective bid-ask spread is related to bond characteristics associated with illiquidity (e.g. timeto-maturity and issue amount) and trading activity measures (e.g. daily turnover, and daily average trade size), indicating that transactions costs are more likely to be associated with the current level of liquidity rather than the future expected liquidity. We also find that the non-default component accounted for a bigger proportion of the yield spread before the financial crisis 2007 - 2009, whereas during the crisis credit risk played a more influential role in determining the yield spread. Common factors such as the underlying volatility and CDS spread explain more of the variation in the non-default yield spread and the bid-ask spread than idiosyncratic factors such as timeto-maturity, issue size, and trading activity proxies do.

The third study presents an equilibrium model in which the heterogeneity of liquidity among bonds is determined endogenously. In particular we show that bonds differ in their liquidity despite having identical cash flow, riskiness and issue amount. Under certain conditions, we show that investors have strong preference for concentrating
trading on a small number of bonds. We conjecture that the identity of the ones which are traded may result from a `Sunspot' equilibrium where it is optimal for traders to randomly label a subset of the bonds as the `liquid' ones and concentrating trading on them. We also show that changing the model assumptions leads to different equilibrium configurations where trading is spread over the bonds. In addition, by utilising the concepts of stochastic dominance, utility indifference pricing, and some specific assumptions on asset value and order arrival rate, the equilibrium prices and bid-ask spreads can be quantified.

Publication Type:	Thesis (Doctoral)
Subjects:	H Social Sciences > HG Finance
Departments:	Bayes Business School Doctoral Theses Bayes Business School > Bayes Business School Doctoral Theses

Export

Downloads