Diversification is the best! Except when it actually matters...

There are some rare "truths" in finance: concepts that work no matter what. Diversification is one of them. But how much does it help when the tide goes out?

A little note on diversification

Let me take a minute to review why diversification is such an imperative aspect of one's portfolio in the first place, and to crystallize the fact that I'm not writing against diversification at all.

Diversification within an asset class takes out the business/industry-specific risk, which the market wasn't going to pay you for in the first place.

For example, let's say you didn't diversify in equities and only bought automobile stocks. A major risk your portfolio faces is the government rolling out large public transport plans next month, leaving automobiles in less demand. However, the market is not going to pay you any extra ex-ante return for having taken this risk even if it doles out in your favour ex-post. This is because you had the chance to diversify your portfolio into other industries as well.

Diversifying one's portfolio across asset classes (equities, bonds, commodities...) always lowers risk as long as the asset classes are sufficiently uncorrelated. This means that if one of the asset classes goes down, the other one would likely go up and reduce your net loss.

So, the main metric of the diversifying power of a new asset class is simply its correlation with the other asset classes. Moreover, the correlation across the asset classes in one's portfolio is also a metric of the extent to which one's portfolio is diversified. This is what I'll base the rest of my article on.

Historical correlations across asset classes

To research this matter, I classify all investable assets into 11 non-exclusive but largely exhaustive categories, and assign them with proxies:

Equities
- US large-cap equities (S&P 500)
- Emerging market equities (MSCI EM Index)
- US growth equities (Russell 1000 Growth Index)
- US value equities (Russell 1000 Value Index)
Bonds
- US corporate bonds (MSCI US Corporate Bonds Index)
- US sovereign bonds (US Govt Generic 10-Yr Bonds)
Commodities
- Gold (iShares S&P/TSX Global Gold)
- All commodities (Bloomberg Commodities Index)
Real Estate Investment Trusts (iShares US Real Estate ETF)
The US Dollar (DXY ETF)
Bitcoin-USD rate (XBT-USD cross rate)

Historical returns correlation between major asset classes for the decade 2010-2019

Overall, a few trends are clear:

Equities as an asset class is highly (75%+) correlated within its sub-categories
Gold is historically a great diversifier, with low correlations with every other class
Real Estate is moderately correlated with both equities and bonds, because it has characteristics of both
For a long-only portfolio, USD is also a great diversifier, with negative correlations

Why these correlations mean nothing

We've so far established that we diversify our portfolio to reduce its overall risk. And of course, why won't we? We do not want our hard-earned earnings to go down the drain when a tsunami hits.

A looming assumption in this whole logic is that the correlations (diversifying power) of our asset classes won't hoodwink us when that tide goes out. So, let's test that out.

We take 20 years from the start of this century, and separate out 2 periods: 2007-09 (global financial crisis) and 2020-21 (COVID crisis) out to see how correlations rank up in these periods versus other normal times.

Here's the 2-point summary for those with a job:

The median correlation between any two asset classes goes up from 26% to 51% in crisis periods
In the COVID crisis, even the 1st quartile of correlations was 37%, meaning you had virtually no way of sufficiently diversifying your portfolio within these asset classes, unless you absolutely knew the returns of each asset class pre-hand. By the way, this number was 7-11% for other periods.

Box-and-whisker plot of correlations through different periods

Explanations and silver linings

What I have highlighted here is definitely not some new discovery. However, in my experience, it is quite understated and undermined when talking about diversification with real-world constraints.

It shouldn't be surprising to note that when asset prices go down, they go down together. A very high value of assets of the world is owned by a very few number of financial institutions. So, when one of the assets takes a plunge for the worse (for example, real estate and related derivatives in 2007-08), it introduces a cascading effect on the other assets as margin calls and fund redemptions clock in for loss-making assets.

There are only a very few combinations of assets that retain their uncorrelated property in downturns. Conceptually, for instance, hedge funds using managed futures, short-biased or long/short equity mandates are bound to profit in periods where traditional asset classes take a loss.

In practice, using data from the last 20 years, the following are what I call the most liberating pairs of assets, assets that prove to be good diversifiers even when the tide is not in our favour:

Equities and sovereign bonds

This is a testament to the flight-to-safety phenomenon, in which money moves from high-risk equities to virtually-no-risk DM sovereign bonds

Gold and sovereign bonds

Our elders were right all-along to be investing heavily in gold, I guess. No more words needed here.

Feel free to play around with the Excel below. It contains the formatted data as downloaded from Bloomberg.

Asset class correlations.xlsx

Page updated

Google Sites

Report abuse