Textbooks in finance are written about the benefits of diversification and how to achieve your portfolio objectives. If you can find two assets that you estimate have the same expected return, in theory it makes sense to split your portfolio 50/50 among them to reduce the risk to achieve the expected value. Implicit behind this is that the returns achieved by these assets are not correlated. For instance, if your two assets are CNR and CP, if Canada goes bust, your diversification is not going to help. But if your two assets are CP and some boring and stable power generation utility out in India, chances are that the returns from the two assets are likely to be much less correlated. Computer algorithms can sort out all of these historical correlations and give you a pretty good idea of the mathematical risk, just from historical trading data.
Then we get into the business of asset allocation. Traditionally, equities and government bonds are inversely correlated to each other, and it has been a layer of portfolio protection when equities rise, you sell a little bit and buy (relative to before, lower priced) treasuries and vice versa.
However, it all goes haywire when traditional correlations do not manifest themselves.
One example is the usage of gold as a “world is going to hell” hedge and also a hedge against inflationary monetary policy decisions. In panicked market conditions, gold is just as susceptible as other asset classes for being liquidated.
Another example is the market for unsecured debt (e.g. TSX debentures or any other corporate bond that trades publicly in a reasonably liquid manner) – although many of these companies are sure-guarantees to pay out at maturity, the value of their debt trades down in market panic conditions.
Finally, another example is the usage of Bitcoin. Since there is limited historical data, there is a considerably higher element of human intuition that goes behind what the true risk profile of this asset is.
When traditional correlations break, it forces portfolio managers to either stay the course (assuming it will regress to some sort of ‘mean’), or to adjust the asset allocation to reflect the new reality with the correlations between various assets. In general, my gut feel is that markets are moving ‘faster’ than they were before, which will make institutional managers that much more challenged to adjust their models to reflect market reality.
I have a paper somewhere that demonstrates mathematically that the benefits of diversification fall off dramatically above about 16 holdings. This makes the typical mutual fund grossly overdiversified.
Also, there are benefits to hedging but “protecting” the day to day mark-to-market value of private portfolio is much less important than it is for managed money (capital return risk, reputational risk, etc.). If fact, the main “loss” when traditional correlations break is opportunity cost (i.e. the ability to be flush when others are not). The costs of hedging, IMO, in a private portfolio ought to weighed against this latter basis. For a real life example, see Prem Watsa’s multi-billion hedging losses after the 2008-9 financial crisis. Ironically, if he had just held those positions another couple of years he would have looked brilliant again.
Prem is an interesting case study on this. It is always easy to retrospectively look back and say what-if, so I’m not entirely slamming this, but if he had just:
1) Not hedged against the S&P 500 post-economic crisis
2) Not invested a couple billions in CPI-linked derivatives (i.e. betting on negative CPI)
3) Invested in Apple instead of Blackberry
There’s be tens of billions more in book value on Fairfax.
All three of the above he had a reasonable thesis for. He reversed #1 with Trump’s election, he let #2 lapse (I believe it has nearly expired) and #3… that’s still on-going but suffice to say that BB hasn’t gone anywhere although it wasn’t a loss (the investment was primarily in low coupon convertible debt and now roughly a 20% equity stake).