2008 changed everything; the global economic crisis proved that diversification alone could no longer be the panacea for risk management in equity-driven retail portfolios.
Many hedge-funds were able to side-step the crisis through the creative use of derivative products and low-correlation (to core equity), alternative investment vehicles. The ordinary retail investor, however, was left with an all equity portfolio, that no matter how well globally diversified, saw a dramatic loss in portfolio value.
Whether or not 2008 represents a “failure of diversification” is still up for debate; what is certain is that diversification alone in an all-equity portfolio is no longer adequate to manage risk. Until retail investors have access to sophisticated investment vehicles and alternative asset classes, other solutions must be found for their largely-equity portfolios.
Portfolio optimization techniques have always existed that focus on managing risk: target volatility, target value-at-risk, target expected-shortfall, and even target draw-down to name a few. Recently, target volatility portfolios have become popular retail solutions, with multiple ETF strategies being rolled out and Standard & Poors releasing a “risk-controled” variation of nearly every index they publish.
These risk-management methodologies all have one thing in common: they are predictive in nature, using simulated, hypothetical future return streams to construct their allocations. A target volatility portfolio, for example, may simulate many hypothetical return paths and change the allocation of the portfolio until the variation between these return streams falls within a certain level. Since volatility is a symmetric measure, this methodology necessarily limits upside potential to limit downside loss (note, however, that even the lowest volatility target portfolios won’t prevent a “smooth ride down”).
Optimization strategies that specifically target down-side protection can suffer from this symmetry issue as well. The most advanced simulation techniques will often still create a fairly symmetric distribution of future returns, so unless special care is taken in optimization, the “easiest” way to limit down-side loss is to sacrifice upside gain. Even when special care is given, because of event-driven correlations, tail-dependence, and fat-tails, it is still extremely difficult to construct a portfolio that limits downside risk and while allowing significant upside potential when optimizing through simulation.
The effectiveness of simulation-based risk management is also typically restricted to the similarity of the historical period the model is built on and the time period the model is predicting over. The reality is, however, that there are exogenous events that can neither be predicted nor controlled which will always fall outside of the historical data set. The further the model attempts to predict out, the greater this risk.
Finally, for simulation-based models to be accurate, they must capture the realities of capital markets: time-varying volatility, volatility clustering, time-varying correlations, tail-dependence of security returns, and fat-tails — to name a few. The effectiveness of these models is contingent on their ability to consistently capture these effects.
These are some of the inherent limitations and realities of simulation based risk management. Despite these drawbacks, this sort of risk management has its place; at Newfound Research LLC, we contend that it is best used to manage portfolio targets. We define portfolio targets as goals and requirements specified by the investor. For example, a pension plan may be willing to sacrifice upside potential to ensure that there is less than a 1% chance that their portfolio loses 10% this year. In this case, simulation based risk management is an extremely appropriate solution.
Newfound Research offers a different kind of risk management solution. Our models are reactive in nature, which means that they don’t use simulation of hypothetical return streams to manage risk but rather model the realized return stream to determine if the time-series is appreciating or depreciating. In this way, we can create a skewed upside-downside capture ratio that isn’t possible with simulation based techniques, since our risk models only kick in when the realized return stream indicates they should. Furthermore, we don’t need to worry about unforeseen events or accurately modeling the intricacies of capital markets since the model is not trying to accurately model market returns.
Not much can be said with certainty regarding reactive models (in the “statistics” definition of the word) and expected portfolio behavior because the reactive methodology does not aim to control behavior by restricting it, but rather by reacting to it.
When markets are appreciating, we expect that the Newfound Research risk models will allow for upside participation without limitation, and when markets are depreciating, the Newfound Research risk models will get you out of harm’s way. In this way, our risk models seek to enhance the underlying returns of an investment.
Newfound Research believes that both of these methodologies for risk management have a role in portfolio construction. Investors who do not want to sacrifice returns for loss protection can benefit from exploring reactive based risk methods, such as those offered by Newfound Research.