Forecasting stock market returns is a tricky business but one can use the JSE’s Price-to-earnings (P/E) ratio to give pretty good estimates. Below is a chart showing the relationship between observed JSE P/E ratios and subsequent 5-year returns since 1993:
This is a pretty tight correlation as far as these things go. Clearly the higher the P/E, the more overvalued the stock market and the less the likely 5-year returns. Buying the JSE index when P/E’s are 12x earnings is likely to lead to you doubling your investment over 5 years. Currently, the P/E is 17 x which implies a 5-year return of 39% or 6.8% compound per annum.
Have a look below how well the model fits the actual outcomes:
What about something more useful and topical like 1-year ahead returns? Unfortunately, the P/E on any day has almost zero bearing on future one-year returns as shown below. Yes, there is a slight sloping regression that implies the higher the P/E the less likely your returns after a year (and vice versa) but the widely scattered data and low correlation means this is not really useful for the forecaster.
But all is not lost! We can use the "relatively accurate" 5-year forecast model to impute 1-year ahead returns using a really clever technique. It goes like this - on any given day "T", we look back 4 years to see what the 5-year forecasting model had predicted for the next 5 years "A" and compare this to how much the JSE has already grown in the last 4 years "B". This will then tell us what returns "C" we need to achieve in the year ahead to match the "accurate" forecast made by the 5-year regression model:
This trick yields remarkable accuracy (regression is 0.33) for what is essentially a very short forecast horizon. The accuracy of the model is shown on the regression chart below:
The next chart makes it easier to visualize the utility of this technique:
This monthly chart plots the J203 ALSH index (blue), the actual 1-year future returns achieved from each month (black), and the forecasted 1-year returns at the time (green) which is known 1 year before the black line (the actual achieved 1-year growth).
Lowly correlation aside, it is remarkable that the model accurately forecast the negative returns of the last 4 great bear markets - which is more important than month-to-month tracking error (which can be quite wide). On the last 5 occasions, whenever the 1-year forecast dips below zero, future 1-year returns are either very muted or even negative. The 1-year forecast going negative is so important, that just about every Analytics client has our systems set up to SMS alert them when this occurs.
Right now, the 1-year forecast is 23.8% but be careful before breaking out the champagne as the history shows how wide the forecasts and actuals can be on this short horizon. Remember this has absolutely nothing to do with today’s P/E but everything to do with the P/E exactly 4 years ago and the fact that we have not "used up" all the forecasted gains budget in the last 4 years. The most important thing, however, is that this model is not forecasting negative returns which means it is not forecasting a bear market in stocks. It makes sense as the black line shows us, we are probably on the cusp of emerging from a bear market.
Dwaine van Vuuren
RecessionAlert, Sharenet Analytics
Dwaine van Vuuren is a full-time trader, global investor and stock-market researcher. His passion for numbers and keen research & analytic ability has helped grow RecessionALERT.com (US based) and Sharenet Analytics (SA based) into subscription services used by thousands of hedge funds, brokerage firms, financial advisers and private investors around the world. An enthusiastic educator, he will have you trading and investing with confidence & discipline.