Correlation Numbers Only Tell So Much..

We have written before (probably numerous times) about correlation. Sparing you from the mathematics behind the calculations and what the ratios mean, it’s safe to say that the output between two variables ranges between +1 (perfectly correlated) to -1 (perfectly uncorrelated).?0 means there is no correlation at all.

With almost thirty years behind me as a fund selector, asset allocator, fund manager, investment director and chief investment officer, I’ve interviewed thousands of fund managers all over the world; I’ve seen thousands of presentations. All presentations are slightly different (even those that are passive and benchmarked to the same index) and markets are constantly changing which means due diligence is constant too – regardless of whether the fundamental research being undertaken is on active or passive managers.

Of course, these presentations will have performance statistics, stock examples, team and manger biographies, investment philosophy slides and it’s incredibly rewarding to see in the last couple of years a number of slides relating to ESG being integrated into pitch decks too.

Many presentations will include observations relating to how well the fund helps diversify a portfolio, and this is predominantly done by showing a correlation table. The fund manager or sales team will argue that if you can include the fund they are promoting to your portfolio the diversification benefits help spread risk (maybe even reduce it) and correspondingly improve the risk/reward profile thus assisting the “efficient frontier” making it a win-win situation.

But, as previously mentioned, markets are constantly changing. Therefore, correlations are constantly changing. Taking two sectors – UK All Companies and UK Gilts - the last twenty years sees these two sectors having a -0.15 correlation (data taken from Financial Express to the end of October 2021) which basically means that the movement of one sector has very little influence on the movement of the other.?Looking at 10 years and the number is -0.14. Five years has a -0.13 correlation which would suggest some degree of consistency.

The yield curve – well used in the fund management profession basically tracks the yield payable for a government bond time along the horizontal axis and yield on the vertical.?By plotting all the different durations, you get to see how much money “costs” according to time.?Taking this principle, you can do the same for correlations.?What has the correlation been over 1 month, 2 months, 3 months and so on.?Taking the two sectors as above, we have created a chart and the output is below:

It’s best to read the chart from right to left. The data looks back ten years, updated monthly, so there are 120 observations to compare.?As can be seen, the correlation between these two sectors is basically negative from 2 years out which ultimately means that if you want to own these two sectors in your portfolio for diversification reasons, you have to own both for at least this time frame for the correlation benefit to take shape.

The chart below is a really interesting on in my view – it compares the UK All Companies Sector with the UK Equity Income Sector over the same time frame and it basically states there is no correlation benefits whatsoever by owning funds in both sectors (at a sector level).

The difference in performance terms over the time period in question favours an exposure to the UK All Companies sector as the chart below shows, although in reality it looks like the difference has been exacerbated since the Covid-19 outbreak as the two charts are very similar to that point.

There are – of course – lies, damned lies and statistics and correlations only work against two variables and only tell part of the story – take for instance Liontrust Special Situations – the correlation data suggests the fund is very highly correlated…

…and yet the performance statistics tell a very different story…

The Liontrust example above shows that fund managers can add performance without having to be massively different to the peers from a correlation perspective.?We could have easily shown funds with high correlations to the peers combined with poor performance in the same sector, and funds with low correlations with high performance and funds with low correlations and low performance track records.

In the pitch decks of fund management companies, many will show the correlation number of their fund relative to the index, benchmark, peers or other assets – how that number is interpreted is very subjective as the tables / charts above have shown, but it can also be a useful indicator in the tool kit of portfolio construction.