Discuss the Risk Management at LTCM. How does “Value at Risk” limit loss exposure? What information does “Stress Testing” provide? What is the effect of correlation among trades on portfolio risk? How can you test LTCM’s correlation assumption with the Fund Performance Data?
Risk Management at LTCM
LTCM principals had strong preference for strategies that have little or no default risk. LTCM generally avoided outright long positions in high yield debts like emerging countries debt or high risk corporates. However, LTCM holds short equities and long debt of corporate where it finds risk manageable.
LTCM traded in markets where the underlying is credit enhanced by a strong party like mortgage pools backed by the full faith and credit of the United States government. LTCM also participate UK, French, Germany and Japanese markets as they markets are mature and relatively stronger than emerging markets.
Evaluating own risk
LTCM consistently assess own risk measured in terms of the probability distribution of potential profits and losses. It uses “value at risk” which was commonly employed by financial institutions. The value is ascertained as it runs scenarios ranging from completely correlated to completely uncorrelated portfolios based on its stress tests.
Fundamental risk assumption
The fundamental assumption of LTCM risk assessment was the belief that as pricing discrepancies became more pronounced, downside risk would be lowered as it would attract attention and subsequent profit seeking activities would revert valuation to its fair value.
I believe that this fundamental understanding could lead LTCM to lose money as it may take long time for market to revert to its fair value. In the meantime, LTCM may continue to add positions which could be interpreted as “loss averaging”.
“Value at Risk” limits loss exposure
The value at risk is a numerical method of providing a risk measure of the risk of loss on a portfolio of financial assets. The loss is defined as within a probability and time horizon that is convertible into a dollar value.
The trader can opt to view this value as the loss a trade can cause on the capital of an investment. It is used as a risk management tool for trading strategies like setting “stop loss” trades or alternatively, for aggressive investors, it can be used to increase portfolio positions in uncorrelated trades where returns expectation are higher to partially offset potential loss.
It also sets a limit on exposure for any one trade or total number of trades, size of trades and type of trades in a portfolio. There are criticisms against the usefulness of VAR on limiting loss exposure. The economist and thinker, Nassim Taleb, writer of “Black Swan”, noted that VAR is a probabilistic model that are developed based on current mathematics but might not be mature enough to cater for real life scenarios. It created false sense of security and resulted in excessive risk taking.
I agree with Taleb as the VAR model failed to account for the interplay of fear and greed in a market.
It is a method of evaluating the performance of any investment or all investment under stressed scenario. The description of a “stressed” scenario is often an art rather than science. It is reliant on the skilled management to determine what assets are under stressed.
The recent banking crisis sees central banks performing stress tests on many bank balance sheets. The final information provided is the additional capital required by bank to survive the stress tested scenario. The scenarios included ranged from default by individual borrowers, to sovereign debt default (like Greece). This has led to failed banks raising cash as a buffer in case scenario occurs.
Performing stress tests provides a feel behavior of portfolio under different stressed circumstances. There are many mathematical method of stress testing including Monte Carlo simulations. However, it could be prone to “rubbish in rubbish” as it is dependent on the skills of administrators who are human.
LTCM assesses risk to profit of a portfolio depending on the correlation between the profits of the individual position. For example, in the case of M perfectly correlated positions, each with the same standalone standard deviation of annual profits of $10million, at the margin, each position contributes $10 million to the risk of portfolio. In the case of M correlated position, the standard deviation of the annual risk or profit is $10M while M uncorrelated positions, the standard deviation of the annual risk or profit is square root of M multiply by $10 million.
LTCM is of the opinion that by adding additional uncorrelated positions, the quantum addition of risk contribution is reduced. LTCM also believe that adding correlated position will not improve risk reward as there is no quantum reduction in risk on expected reward.
LTCM had 2 approaches to risk correlations. In the monthly view, LTCM is of the position that the positions are always positively correlated while on the annual basis, LTCM is of the position that there is only a small correlation.
Test Risk Correlation assumption with Fund Performance Data
If we evaluate the monthly performance of the fund in dollar terms, then obtain the standard deviation of the gross monthly performance. It is possible to test the monthly risk correlations by comparing the gross monthly performance of the fund over the period. It should not deviate beyond 20% of the NAV per annum, if the risk model is correct.
The same method can be obtained for the annual basis, where the fund presumes that the correlation is small, but not completely uncorrelated.
This method made use of assumptions that monthly returns on profit are attributable solely to the risk portfolio. It also assumes that the fund capital is direct result of investment in the fund based on only risk portfolio returns, and the net fund withdrawals and investments by investors is very small.
Net monthly performance does not apply net of fees paid is not useful as it is deduced after rewards deductions, that is dependent of the returns of the fund which contained both a fixed component and a “high water mark” component.
Despite modern mathematical tools, risk assessment, evaluation of risk using the value at risk method, and advanced data inference and computer trading, the hedge fund was still caught off guard by the end of 1997.