Two essays on Volatility Transmission and Default Prediction for Individual Firms

Revenue Exposure : A volatility transmission mechanism between the Shanghai Composite Index and S&P500 firms : Volatility transmission studies are mostly concerned with the volatility spillover effects between equity markets and between asset classes. Few have considered the transmission of volatility between markets and firms. We examine the significance of revenue exposure as a proxy for trade links at the firm-level in explaining volatility transmission. Specifically, we consider how the volatility of Europe, Asia Ex-Japan, China and Japan equity markets affect the volatility of S&P500 firms, using revenue exposure as a volatility transmission mechanism. Robust evidence of such market volatility transmission effects can be found both at the individual firm and at the level of the portfolio of firms exposed to the same foreign market, when using adapted GARCH models at the firm-level. Economic value is achieved with a portfolio of stocks selected on the basis of their volatility transmission coefficient. Out-of-sample performance of this portfolio exceeded that achieved by other active portfolios optimized on GARCH(1,1), or passive and buy-and-hold (MSCI US) portfolios.The implication of this is that portfolio managers can diversify portfolio risks by investing in S&P500 firms that have revenue exposure to foreign markets, without having actually to invest in stocks of those markets which may further subject the portfolio to currency risks.

Liquidity inputs can improve default prediction: Models for predicting defaults fall into two main categories, structural and reduced form. This paper focuses on structural models to provide short to mid-term credit views, as opposed to through-cycle agency ratings. While it has been noted that systemic risk exacerbate defaults and liquidity has been observed to be poor prior to crises, yet existing models reviewed have not explicitly modeled for liquidity and the macro economic conditions that exasperate these crises. Four models are considered in which each succeeds the former with more variables. Base model A consists of modified Altman (1968) variables; latter models successively add firm-level variables, macro-level liquidity and non-liquidity macro variables. Discriminant functions defined for each model, using a dimension-reducing implementation of Multiple Discriminant Analysis (MDA) calibrated on an initial data set between 1980 – 2004, discriminate defaulting and non-defaulting firms out-of-sample. Optimal default predictability is achieved when macro-level liquidity and nonliquidity macro variables complemented firm-level variables. Excluding non-liquidity macro variables could improve default predictability further but compromised on the overall model accuracy. Industry segmentation before calibration separates segments that could achieve higher default predictability from segments where predictability is low. Models for Materials, Industrials, Consumer Cyclical and Energy had higher default predictability.