How much diversification is enough?
Mark Twain once said that “History never repeats itself, but it does often rhyme”.
I prefer to live by the maxim “ignore history at your peril” because (to use another well-worn and beloved idiom), “those who cannot remember the past are condemned to repeat it”.
At the moment, we are living in uncertain times. Sailing in uncharted waters, perhaps. So, what can investors do to reduce risk and future-proof their portfolios?
The best way to navigate future turbulence is to have a robust, well-diversified portfolio. But how much diversification is enough? I unequivocally support diversification. So much so that I believe diversification is still the only free lunch available to investors. However, there are some pitfalls when considering how much diversification is optimal and sometimes diversification might not even be appropriate.
Spurious diversification is not helpful. For example, if cross-sectional volatility is low and most individual assets are going up or down together, then purported diversification is a cost with negligible added benefit.
A case study where this occurred was in 2007-2009 with Collateralised Debt Obligations (CDOs). Ratings agencies assumed that having a large number of individual sub-investment grade debt instruments in a single portfolio would make that portfolio so diversified that the portfolio as a whole could be rated as investment grade. This would allow investors who were only able to buy products above a certain rating to invest in CDOs.
The methodology employed by Moody’s et al. proved to be a terrible under-estimation of the true risk of CDO portfolios because some investments were so correlated that in the GFC that they lost value together and provided little or no diversification benefits.
The moral of this story is that it is important to look carefully at covariance matrices and assess whether adding additional assets is likely to reduce enough risk to justify the cost and will work in a market dislocation. (Hoping that the dislocation is not so severe that all correlations move towards one and the only thing that rises is my blood pressure)
Perhaps this anecdote illustrates the point: I have three daughters who would all love to hear the words “I love you”. They are not likely to hear the three words “attention K-Mart shoppers”. But they should be terrified of the three words “Serial Auto Correlation” because in that environment diversification can do little to save investors from large drawdowns.
Over-diversification is an expensive waste of time and will certainly reduce alpha-generating opportunities. Just like investment strategies, diversification works until it doesn’t. But where is the point at which further diversification is no longer appropriate?
Some very active investors would have us believe that diversification should be re-named “diworsification” as it is not a good idea at all. For example, Buffet called diversification “protection against ignorance” and said that it “makes very little sense for anyone who knows what they’re doing”, but followed that by saying “it is a perfectly sound approach for somebody who does not feel they know how to analyse businesses”.
Because the Your Future, Your Super performance assessment makes it exceedingly dangerous to take too much active risk, and for sound prudential reasons, I do not recommend a highly concentrated portfolio for most investors. The risk of excessive divergence against indices and peers is just too high.
I have a more moderate approach and prefer to use a measure such as portfolio redundancy (or active-risk percentage) to strike the right balance between diversity and portfolio risk.
When we restructured the Australian equities portfolio at Australian Catholic Super, we began with a portfolio that had so many managers that it was really just a super expensive index portfolio. This was seen anecdotally when some managers were long some stocks whilst others were short the same stocks making their collective active decisions redundant. If I recall correctly, the portfolio had a redundancy factor of around 70 per cent.
By carefully reducing the number of managers we were able to construct a portfolio where the redundancy was reduced to around 27 per cent and we thought we had a good mix of active risk without too much tracking error. By and large this was successful, though the mid cap bias of our portfolio still exhibited some tracking error to the ASX300 over some time periods.
Very long-term investors should be less concerned about volatility – perhaps they should even embrace it. All three of my daughters have an investment horizon of over 50 years and they can afford to take a lot more risk, whether that’s in terms of absolute volatility or tracking error. Accordingly, they all have portfolios with at least 100 per cent equities.
The best rationale I can give for their asset allocation is as follows:
- Even if shares might make a return between -30 per cent and +40 per cent in any given year, the very long-term expectation is that returns will average 8-12 per cent per annum over the long term; and;
- Diversifying with lower risk, lower returning assets is not necessary for those with a very long investment horizon.
This of course brings me back to my crusade (another religious reference) in favour of lifecycle investing whereby younger investors should take more risk than older investors. I will just leave it at that for now.
I believe that it is a term of art rather than of science as to how much diversification is optimal for any given portfolio and a complex question for investment practitioners and their stakeholders to discuss. At the end of the day, the only guidance I can add (continuing with the Goldilocks theme from previous articles) is not too much diversification, not too little diversification, but just right!