A diversification risk rating gives an idea of the true diversification of a portfolio. This depends on the level of cross-correlation that exists within the portfolio but it is often a misunderstood indicator of effective portfolio diversification.
To determine if the portfolio is well-diversified, we need to look beyond the number of different investments in the portfolio and understand the portfolio’s correlation to different asset classes. However, correlation is misused in the financial industry. This is because it is very easy to gather data and calculate a correlation matrix with a tool like Excel. But just because something can be calculated doesn’t mean it is correct or useful.
Often times many assets in a seemingly diversified portfolio begin to fall in unison. High Yield, Real Estate, Commodities, and even possibly US Treasuries start to behave like equities. Investors are often negatively convex to market losses. What does this mean? Being negatively convex to the markets means that when the markets fall sharply investor losses are amplified as opposed to being minimized by diversification.
Here is the problem, not only are the measured correlation relationships unstable and change unpredictably, but invariably, when one is looking at assets such as stocks and bonds, the solution to the diversification problem is to always add lots of bonds to the portfolio because, more often than not, stocks and bonds are anti-correlated. Stocks down, bonds up, and vice versa.
However, not only is this not always the case, like in the Taper Tantrum but now that interest rates are so low worldwide, there is little room for them to fall further. This means that bonds are less effective as a diversifying agent in the post-quantitative easing realm we are in now that central banks worldwide have run out of ammunition to drive interest rates lower.
At StratiFi, we specifically designed a proprietary Diversification Risk Rating to quantify this very risk. For example, an extreme correlation with SPY will very likely reflect a poor-performing and crisis-sensitive portfolio, whereas the same correlation amplitude with respect to Real Estate only is less likely to reflect the same behavior. For this rating, a threshold-based approach is used where the threshold values and the weights (normalization factors) are learned using a machine learning approach such that a set of benchmark portfolios (including the 60/40) would have a rating close to 5.
We use Volatility Index (VIX), Bonds (BND), S&P 500 (SPY), Gold (GLD), and Real Estate (VNQ) as indicators of different asset classes. We compute rolling six-month cross-correlations between the portfolio and different asset classes to get a set of cross-correlation distributions.
Each of those distributions has tails (2% and 98% percentiles) that give the number of extreme events, which in turn gives us the percentage of extreme events and extreme correlations.
Those percentages are compared against a set of threshold portfolios in order to tell us whether they are indicative of a “healthy” or “unhealthy” portfolio. Namely, a “healthy” portfolio will have percentage values that are similar to what a reasonable portfolio would do, for each asset class.
A portfolio will exhibit:
Mike Tyson, the boxing champ, once said everyone has a plan until they get hit in the mouth. This is why analyzing diversification is so important. StratiFi advisors use this to know if a client or prospect portfolio is diversified or pseudo-diversified. That way, when they get hit in the mouth by Mr. Market, they’ll be among that small group of investors who actually have a plan. Most everyone else will panic, and react by doubting themselves, their advisors, and their allocations.
StratiFi’s PRISM Ratings™ risk scoring technology provides RIAs, asset managers, and broker-dealers more insight into the risks in their clients’portfolios and their own business, so they can pick the risks they want to take. Contact the StratiFi team to find out more and get started today!