Abstract: Many popular techniques for determining a securities firm's value at risk are based upon the calculation of the historical volatility of returns to the assets that comprise the portfolio, and of the correlations between them. One such approach is the J.P. Morgan RiskMetrics methodology using Markowitz portfolio theory. An implicit assumption underlying this methodology is that the volatilities and correlations are constant throughout the sample period, and in particular that they are not systematically related to one another. However, it has been suggested in a number of studies that the correlation between markets increases when the individual volatilities are high. This paper demonstrates that this type of relationship between correlation and volatility can lead to a downward bias in the estimated value at risk, and proposes a number of pragmatic approaches that risk managers might adopt for dealing with this issue.