The Simple Truth about Good Forecasting

There is always a lot to be said for keeping things simple, and it is gratifying to see how robustly accurate some of the simplest methods of asset class forecasting continue to be. The secret of building a robust long term record, experience suggests, lies in giving yourself a sensible time horizon (at least five years), concentrating on a few fundamental valuation metrics and downplaying or better still ignoring wherever possible macro-economic forecasts.

Jeremy Grantham of GMO is one of the leading exponents of this kind of approach and he was able recently to point to some impressive evidence of the success of his methods. His firm’s asset allocation decisions are built around the idea that profit margins and price-earnings ratios are mean-reverting over a period of years. He has been doing this since 1994 and the results, so he noted in his latest quarterly letter, have turned out to be pleasingly accurate.

Looking back at the ten-year forecasts for eleven asset classes his firm made back in December 2001, he notes that the actual out-turns for the ten-year period ending in December 2011 were, with minor exceptions, both ranked in more or less the right order and not far out either in terms of annualised real returns – not a perfect track record by any means, but one for which most economic forecasters would give their eye teeth.

Particularly noteworthy was his firm’s prediction that emerging market equities (up 11.4% per annum against a forecast of 9.4%) would be the star performers of the decade while the S&P 500 – which scratched out an annualised return of just 0.4% per annum against his predicted minus 1.0% – would languish firmly at the bottom of the list. There were not, unsurprisingly, many firms on Wall Street publicly predicting a “lost decade” for US equities at the time GMO was making its original forecast.

Mr Grantham points out that there is nothing particularly profound about his predictive methods. The discovery that when stocks and bonds are expensive, they tend to perform badly, and when they are cheap, do much better is hardly breakthrough thinking. Back in 2001 the S&P 500 was trading on 30 times earnings and emerging markets on 13 times earnings.

Yet as so often happens, not many professional forecasters drew the obvious, right conclusions. No doubt this was partly for professional reasons. The investment business, as we all know, is dominated by short term pressures and its default position tends to be biased towards orthodoxy and bullishness. (A more subtle problem, in a period dominated by a belief in rational expectations, has been the tendency for investors to seek a rationalisation for any current valuation anomaly, however improbable it looks in retrospect).

GMO’s estimates, says Mr Grantham, “are not about nuances or PhDs. They are about ignoring the crowd, working out simple ratios, and being patient. But if you are a professional, they would also be about colossal business risk….The problem is that though they may be simple to produce, they are hard for professionals to implement”. GMO’s all-too-accurate market forecasts cost it a lot of business at the time.

Another veteran of the industry who favours a keep-it-simple approach to valuations and asset allocation is Jack Bogle, the founder of Vanguard. In his early books about the fund business, he uses a relatively straightforward variant of the cyclically-adjusted p/e ratio to come up with broad ten-year forecasts for future stock returns, while noting that the current 10-year bond yield is in practice a perfectly workable and normally reliable estimate of the next decade’s most likely return from bonds. (He was of course writing before the Federal Reserve and other central banks got to work to try and manipulate the yield curve for their own purposes as blatantly as they are doing today).

His successors at Vanguard now use a range of sophisticated techniques, such as Monte Carlo simulations, to model ranges for future bond and equity market returns, but, as I read their material, the underlying philosophy remains essentially the same as that of the founder. “Our long held view” write the firm’s strategists Joseph Davis and Roger Aliaga-Diaz in the firm’s 2012 annual investment outlook, “is that market valuations generally correlate with future stock returns” while “consensus economic growth expectations and initial dividend yields do not”.

Applying this approach today, they argue that there is no reason why future stock market returns should disappoint today even if economic growth disappoints, as it tends to do in the aftermath of global financial crises. Their central case is that over the next ten years equities have a 50% chance of matching or beating the 6% annualised real return they have achieved since 1926.

Bonds will produce lower nominal returns than their historical average, but will retain their diversification value even as bond yields start to rise, as they are bound to do over the course of the next decade. Implicit therefore in their view is that the correlation between bond and equity market performance will remain low, continuing the experience of the new century, but reversing the experience of the 1990s.

GMO’s blunter advice is to be underweight long term bonds and broadly neutral on global equities (though heavily biased in the United States towards quality stocks). If you believe in the value of these simple methods, the dull but reassuring message is that on a ten-year view equity markets are probably priced about right for once, which is not a universally shared opinion.