Returns on the UK stock market have been predictable over the last 25 years, according to new research by Professors Hashem Pesaran and Allan Timmermann, published in the latest issue of the Economic Journal. Their work suggests that without the benefit of hindsight, an investor could have correctly predicted the direction of the UK stock market roughly 60% of the time on a monthly basis. When exploited in a market timing investment strategy in real time, investors could have more than doubled their returns per unit of risk, even after accounting for transaction costs.
Such findings are particularly interesting in view of UK investment funds'' poor track record in market timing strategies. Recent studies have found that UK funds systematically attempt to market-time rather than follow passive investment strategies. But these same studies also find that returns to market timing are predominantly negative. Pesaran and Timmermann''s research shows that, in fact, market timing based on simple forecasting methods could have lead to successful market timing.
The researchers are careful to select forecasting models as if in ''real time'', without assuming that investors actually knew anything about which forecasting model was ''true''. Stock returns are generally difficult to predict because, even if markets are inefficient, such inefficiencies are likely to be very small given the amount of capital dedicated towards exploiting them. Using publicly available information, such as interest rates, money growth, oil prices and growth in industrial production, the researchers find that small degrees of predictability still remain, and these could have been exploited over the period 1970-93 by investors who searched carefully for a good forecasting model.
Historical predictions would have lead to large variations in portfolio weights when used in a market timing strategy that mixes cash and the stock index. Some periods see the portfolio weights change from zero cash and 100% stocks to 100% cash and zero stocks. Such a ''switching rule'' was in fact advocated early on by John Maynard Keynes, who combined investment in real assets with switching between short-dated and long-dated securities based on movements in interest rates.
Pesaran and Timmermann discuss two possible interpretations of these findings. The efficient market interpretation regards variations in expected stock returns as changing risk premia since the ''price of risk'' may well vary over time. It identifies two periods (1973-74 and 1981) where negative stock returns were systematically (and correctly) predicted.
It is hard to believe that investors would have been willing to hold stocks at times when the price of risk is negative. This leads to a second, more plausible explanation. Predictability of stock returns could reflect an inefficient stock market dominated by investors who do not use publicly available information efficiently, possibly because they systematically overreact to news.
''A Recursive Modelling Approach to Predicting UK Stock Returns'' by Hashem Pesaran and Allan Timmermann is published in the January 2000 issue of the Economic Journal. Pesaran is at the University of Cambridge; Timmermann is at the University of California, San Diego.
