The use of value at risk by institutional investors - analysis of risk management - Statistical Data Included

New England Economic Review, Nov-Dec, 2000 by Katerina Simons

Senior Economist, Federal Reserve Bank of Boston. The author wishes to thank Richard Kopcke for helpful comments. David schramm provided valuable research assistance.

**********

In recent years, risk management has been of growing interest to institutional investors, including pension funds, insurance companies, endowments, and foundations as well as the asset management firms that manage funds on their behalf. Traditionally, institutional investors, and particularly pension funds, have emphasized measuring and rewarding investment performance by their portfolio managers. In the past decade, however, many U.S. pension funds have significantly increased the complexity of their portfolios by broadening the menu of acceptable investments. These investments can include foreign securities, commodities, futures, swaps, options, and collateralized mortgage obligations. At the same time, well-publicized losses among pension funds, hedge funds, and municipalities have underlined the importance of risk management and measuring performance on a risk-adjusted basis.

One approach to risk management, known as Value at Risk (or VaR), has gained increasing acceptance in the last five years. However, institutional investors' quest for a VaR-based risk-management system has been hampered by several factors. One is a lack of generally accepted standards that would apply to them. Most work in the area of VaR-based risk measurement and standard-setting has been done at commercial and investment banks in conjunction with managing market risk. VaR originated on derivatives trading desks and then spread to other trading operations. The implementations of VaR developed at these institutions naturally reflected the needs and characteristics of their trading operations, such as very short time horizons, generally liquid securities, and market-neutral positions. In contrast, investment managers generally stay invested in the market, can have illiquid securities in their portfolios, and hold positions for a long time.

Moreover, many risk-management systems developed for trading operations are expensive to implement and beyond the budget and manpower of smaller pension funds. Nevertheless, recent developments in web-based technologies, which application service providers use to make risk measurement available to clients over the Internet, hold promise of bringing affordable risk management to the cross-section of smaller institutional investors. This makes it important to explore the practical issues institutional investors have to consider while implementing a VaR-based risk management system.

VaR is a measure of risk based on a probability of loss and a specific time horizon in which this loss can be expected to occur. Bank regulators use VaR to set capital requirements for bank trading accounts because VaR models can be used to estimate the loss of capital due to market risk. Pension plans are generally concerned not with the loss of capital, but with under-performing their benchmarks. Pension plans distinguish between a long-term or strategic asset allocation, also known as the "policy portfolio," and a short-term or tactical asset allocation. The policy portfolio is typically aimed to match the plan's liabilities. The actual portfolio, which represents the tactical asset allocation, can differ from the policy portfolio because fund managers implement market views with the goal of outperforming the policy portfolio. Thus, the policy port-folio represents the benchmark against which the actual portfolio performance is measured. Because performance is measured against the benchmark, the risk shoul d be measured the same way. At the same time, for defined-benefit plans, VaR can represent the risk that assets fall below a certain target, in particular the risk that assets would be insufficient to fund the benefits due employees.

VaR has advantages as a risk measure for institutional investors. Specifically, it is based on the current portfolio composition rather than the historical return on the portfolio, and it can be aggregated across many asset classes. The more traditional risk measures used in investment management have one of these characteristics, but not both. For example, tracking error is a measure of the deviation of the portfolio's historical return from the return on the benchmark index. It may not be useful if the current composition of the portfolio differs from the one that produced these historical returns. On the other hand, two traditional asset-specific measures, beta for stocks and duration for bonds, are based on the current portfolio composition. Beta measures the portfolio's systematic risk, that is, the degree to which its return is correlated with the return on the market as a whole. Duration measures the sensitivity of a bond portfolio to changes in interest rates. The higher the duration, the more sensiti ve it is to changes in interest rates. These measures, while useful, cannot be combined to provide an overall measure of risk.