For discussion - is Head of R&D and Managing Director, Harold de Boer of Transtrend right?
Views on conventional VaR and correlation measures
Measuring risk factors in different ways is a distinctive part of Transtrend’s risk approach. Unlike some CTAs, Transtrend has never targeted a constant volatility level, and eschews conventional Pearson correlation measures of linear correlation and standard VaR measures. Transtrend has found these measures to be unreliable estimators and instead uses estimators that are not affected by kurtosis or fat tails. “Since 1963 it has been clear that financial markets do not follow a normal distribution,” De Boer says, alluding to Benoit Mandelbrot’s seminal paper, “The Variation of Certain Speculative Prices”, published in October 1963.
Transtrend’s January 2016 paper “VaR measures for dynamic long/short investment portfolios” sets out Transtrend’s concerns about conventional VaR. It can be unstable and may throw up false alarms while also underestimating extreme moves at other times, Transtrend has seen. De Boer suspects that either this risk approach, or discretionary decisions, may have led some CTAs to reduce risk around events such as the Brexit and Italian referenda in 2016. Of particular concern is the risk of a synchronised exodus if many investors heed the same VaR measures, possibly at the behest of some regulators. “Since standard Value at Risk measures overreact to short term price moves, it could lead to trading at expensive moments,” warns De Boer. Transtrend’s risk philosophy is to be proactive rather than reactive and to manage risk “by anticipation, not by response” De Boer explains. The intention is to size positions appropriately so that the impact of a large price move on the portfolio should be within the programme’s risk tolerance. “Managing money starts by not losing too much. While that may be easy in normal market circumstances, it is the extreme events that really matter,” De Boer says.
Transtrend has developed its own risk measures, especially designed for the specific strategy, focused on gauging extreme risk by looking at joint tail distributions. Transtrend’s proprietary extreme risk measure is usually much higher than conventional Value at Risk. With a variety of risk measures, the programme dynamically controls position sizes, seeking to avoid overly high factor exposures or extreme risk concentrations. Transtrend’s approach to correlation is multi-dimensional. De Boer observes that there are many ways to measure correlation; patterns of correlation change over time and the coefficient is in practice both asymmetric and nonlinear. Hence more sophisticated techniques than plain vanilla Pearson correlation are needed.