Secrets of Success: Divide and conquer: the optimal portfolio

Diversification, all sensible people know, is the key to an effective investment strategy. But it is a principle that is harder to put into practice than we mostly realise.

Diversification, all sensible people know, is the key to an effective investment strategy. But it is a principle that is harder to put into practice than we mostly realise. The main reason, I think, is that it is one thing to sign on to the principle, and quite another to know how much to have in different types of asset and security.

What, for example, is the right proportion of your assets to have invested in the stock market? And how much, to take the most topical example, should your financial welfare be dependent on property?

Thanks to sterling efforts of the Nobel prizewinner Professor Harry Markowitz and a generation of later academics, what we do know is that there are some clever mathematical ways of working out the optimal weights in a portfolio.

The technique of mean variance optimisation makes it possible to work out the best combination of risky assets to hold, given knowledge of each asset's risk and return characteristics (risk in this case being defined as the volatility, or standard deviation, of returns) and the correlations between the different assets. The calculations will confirm that the less well correlated the components of your portfolio are, the better diversified you are.

That is fine as far as it goes, but - apart from requiring some fancy maths - the academic approach suffers from the disadvantage that it is wholly reliant on historic data. In the wrong hands, this can produce some bizarre results, as for example a few years ago when one of the City's grandest investment banks argued that investors should have 80 per cent in equities and 0 per cent in gilts - at the very point when the equity market was about to crater and gilts continued to produce their best returns for decades! Garbage in, garbage out, in other words.

Another danger is that even an optimised portfolio can lead you into difficulties, because short-term cash flows can force you into changing your asset mix even though it may be the optimal one for the longer term. You only have to look at how three years of a bear market turned many life companies into forced sellers of equities to see how this principle - which is a variant of the better-known syndrome known as gambler's ruin - is no idle danger.

Common sense can probably do at least as good a job. The truth, as far as I can see, is that investment is not a precise science, but a classic example of a business where it is better to be roughly right than precisely wrong. The common sense approach must be to take historic data as a starting point but then add some simple overriding adjustments on the simple contrarian principle that you want to reduce somewhat your exposure to assets that are overvalued and increase your exposure to undervalued ones.

The psychological drawback of this approach, as with all diversification, is that it means you are guaranteed not to do as well as you might have done if you had happened to have 100 per cent exposure to the asset class that in practice turns out to do best. You will in other words miss the very top of every bull market but do better when things are going badly. You cannot have it both ways, unfortunately.

So how then to assess the proportions of different assets to hold? One simple starting point is to look at how different combinations of assets would have worked in the past. I did this exercise some while ago with the data on UK equities, gilts and cash from 1899-2000. (Since then some of the data for the first half of the 20th century has been revised, and we have had a new bear market, but the general principles remain broadly the same). The table below shows the results of this exercise for various mixes of equities and gilts and equities and cash.

You can see at once how mixing either set of assets succeeds in bringing down both the volatility and the returns of each portfolio. Note how introducing gilts into the portfolio does virtually nothing to reduce the number of down years you would have experienced as a 20th-century investor, whereas introducing cash does have that effect. It also reduces the maximum loss in any one year. The band of potential returns is somewhat wider with gilts in the portfolio than it is with cash.

It is also of course possible to mix all three assets together and produce a new set of results. In recent years we now have the option of including other assets, such as property (for which we have good data going back about 35 years), index-linked gilts, corporate bonds and overseas equities. Putting all this information together in a usable format is not easy (and beyond the space of a newspaper column), but it is easy enough to study their various returns over time and see which ones tend to move together and which ones do not.

On top of this, finally, you have to introduce your judgements about the relative valuations of each class of asset. The simplest way to do this with equities and property is to compare them to their long-term averages and also look at how their yields (dividends or rental value) compare to alternatives.

With gilts and other fixed-interest securities, it is usually enough to take the current yield for say a 10-year-gilt and assume that this will be its total return over the next ten years (the correlation is very strong). Then look at the current and expected rate of inflation: if gilts are offering a real yield (after inflation) of 3 per cent per annum on these two measures, then they will normally be good value. Otherwise be wary.

jd@intelligent-investor.co.uk