Event Papers: Progress Report PhD-Year 1: Towards Quantifying Liquidity Premia on Corporate Bonds

Paul van Loon (100019500) Supervisors: Prof. Andrew Cairns Prof. Alexander McNeil Alex Veys August 2013

The Liquidity Premium on corporate bonds and their potential significance as an additional discount factor on long-dated insurance liabilities has been a high priority for Solvency regulators. The robust decomposition of the credit spread into credit- and liquidity components has been the principal objective of this report. Methodologies used in previous academic work, most notably Webber (2007), are reviewed and the proposed modeling approach addresses issues that made previous work unsuitable for industry wide applications. Using the iBoxx GBP investment grade bond index a measure of relative liquidity is calculated. The Relative Bid-Ask Spread (RBAS) is a relative liquidity measure that allows liquidity comparisons between bonds of different characteristics such as rating, duration and age. A series of daily regressions model the Credit Spread using readily available bond characteristics such as duration, age, coupon, rating and RBAS. It also takes into account five levels of seniority of debt, collateralization and classifications such as financial/non-financial. Previous literature, especially those focusing on calibrating structural models, relied on data sources that are not easily available and calibrated models using carefully chosen samples of bonds. The proposed modeling approached uses data that is readily available or easily computed for all bonds on a daily basis, producing daily estimates of liquidity premia on an individual bond level. Results regarding the size of the illiquidity premium are, on aggregate, similar to those reported previously for investment grade corporate bonds. We find that for the average bond, between 30%- 40% of the credit spread is due to liquidity effects, but there is variation over time. In the time period just prior to the financial crisis, the average liquidity premium was between 5%-10%.

Liquidity Premium estimates for individual bonds can substantially differ from reported average; bonds with above an average level of illiquidity (50%-80% quantile) command a premium of 35%-55% and bonds with very high levels of illiquidity command in the region of 55%-85%. Bond specific Liquidity Premium estimates are crucial when considering the possibility of a discount factor for insurance liabilities. Only the actual assets held to maturity, used to replicate liabilities’ cash flows, can potentially be used in to calculate a discount factor. Moreover, a frequent review of a portfolio’s (relative) liquidity and subsequent liquidity premia is an essential part of good risk management practice. The proposed modeling approach allows for both the Liquidity Premium to be estimated for individual bonds as well as constant (daily) review of these estimates.