This is an extended version of the article that appeared in the May 2016 issue of Investment Advisor.
The Oracle of Omaha once said, “Diversification is protection against ignorance; it makes little sense for those who know what they’re doing.” With respect to Warren Buffett, we’d disagree, at least with the second part of that statement. Even investors who know what they are doing need protection against ignorance – ignorance of, or uncertainty about, the future that is. Proper diversification is a must if your goal is to build a portfolio that you can actually stick with over multiple market cycles, including tumultuous times.
Of course, it’s one thing to diversify the idiosyncratic risk in a stock or bond portfolio – that isn’t particularly difficult. It’s another problem entirely to diversify a multi-asset class, multi-manager portfolio. Surely a multi-asset class, multi-manager portfolio is, by definition, diversified, right?
Investors routinely overdiversify at the manager or fund level, filling portfolios with so many SMAs, mutual funds and ETFs that they end up with expensive index portfolios that still aren’t diversified in a way that is beneficial to the investor. Simply adding more funds isn’t effective if the risk exposures those new funds bring to the portfolio are redundant or highly correlated to those already contained in the portfolio.
True diversification is difficult because volatility reduction is only achieved when correlations are really low. For example, if you take two assets with a correlation of 0.7, equally weighted and with the same volatility, you only diversify away about 8% of the volatility. If that correlation drops to 0.5, 87% of the undiversified volatility still remains. Even with a correlation of 0, about 70% of the undiversified volatility remains.
With that knowledge in hand, consider the correlation table below.
With the notable exceptions of investment-grade debt and managed futures, the correlations are rather high in the context of the math we just laid out. Now, that can’t be applied directly to any pair of assets in this matrix because these assets all have different volatilities, but with the full admission that this is an oversimplified exercise, a back-of-the-envelope analysis would suggest that many “diversifying assets” don’t reduce volatility all that much. The takeaway is that those few assets that are truly uncorrelated are hugely valuable to a portfolio.