Market Risk Analysis Volume II: Practical Financial Econometrics

II.3

Classical Models of Volatility and Correlation

II.3.1 INTRODUCTION

This chapter introduces the time series models of volatility and correlation that became popular in the industry more than a decade before this book was published. The point of the chapter is to make readers aware of the pitfalls they may encounter when using simple statistical techniques for estimating and forecasting portfolio risk.

We begin with the models of volatility and correlation, made popular by JP Morgan in the 1990s and still employed to construct the RiskMetrics^TM data. The 1990s were a time when the profession of financial risk management was still in its infancy. Up to this point very few banks, fund managers, corporates or consultants used any sort of time series data to quantify and track the risks they faced. A breakthrough was made in the mid 1990s when JP Morgan released its RiskMetrics data. These are moving average estimates of volatility and correlation for major risk factors such as equity indices, exchange rates and interest rates, updated daily, and they used to be freely available to download. The first two versions of RiskMetrics applied incorrect time series analysis and had to be amended, but by the end of the decade a correct if rather simple time series methodology for constructing covariance matrices was made generally available to the industry.

Volatilities and correlations of financial asset returns and changes in interest rates may be summarized in their covariance matrix. ...