Messenger: Forecasting or fortune-telling?

Each day we read someone’s predictions that the US economy will or won’t recover, meaning that corporate profits will or won’t increase, implying that US shares will or won’t rebound. Fund managers are especially fond of sharing their predictions with the world.

Inexplicably, many people take this seriously and some planners even take it into account in portfolio construction.

Yet history suggests that the act of economic forecasting is so unreliable that, in my view, it ranks with reading tea leaves. Further, it can be riddled with conflicts of interest.

Finally, even if it was accurate, it should be peripheral in portfolio construction.

Table 1 contrasts consensus earnings forecasts with actual earnings growth in the US, over 17 years. The actual result averaged 6.8 per cent, and ranged between 25.8 per cent and -19.4 per cent. Predictions, however, averaged 17.9 per cent and were in a range of 30.4 per cent to 9.1 per cent.

On average, the forecasters were 11 per cent too high, and never predicted a negative year. Indeed, they never forecast a year as low as the actual 16-year average. According to these ‘experts’, every year will be a good year.

In other words, they are trying to talk the market up.

There has been much discussion of conflicts of interest on Wall Street, but no one seems to have noted this issue. If these people aren’t biasing their predictions to paint a rosy scenario, then they are simply lousy at their job. In either case, these forecasts should be completely ignored, along with most other attempts to precisely predict the future.

Forecasts about levels of markets are also demonstrably unreliable. Table 2 reports the predictions of a range of financial institutions for the Australian dollar. It reports five different six- and 12-month predictions (the periods being chosen simply because I have access to the data). In each case there were 10 to 19 institutions involved.

Predictions for the first period, the six months to June 1995, ranged from a high of 82.0 to a low of 76.9 and averaged 80.1.

The actual result was 71.9.

No institution was exactly correct. However, that is surely understandable — this is a difficult task. However, the actual result was well outside the entire range of expectations. In aggregate, the whole group entirely missed the mark.

The same occurred in the second period. And the third. In the fourth, the very lowest estimate was correct and all the others too high. In the fifth the average call was reasonably right.

Table 3 shows predictions for 10-year bond yields. Again, three times out of five, the full range of predictions missed the mark.

Table 4 reports predictions of cash rates over the two most recent periods. In both, the result was outside the range.

Table 5 reports forecasts for the All Ordinaries. Over the earlier period the lowest estimate was close to exactly right. In the latest period, despite the predictions covering a range of 300 points, the lowest was 243 points too high.

On the basis of these outcomes why would anyone ever publicly make another prediction? And if they did why would any planner give it sufficient credence to consider it relevant in determining their recommendations?

Let us return to the question of predicting the direction of a stockmarket, based on expectations of economic recovery. A survey of the last 101 years(Triumph of the Optimistsby Dimson, Marsh & Staunton) found very little correlation between gross domestic product (GDP) and annual equity returns. So much for predicting price moves based on GDP.

This finding is counter-intuitive, but is not incomprehensible. Take the example of recent years in the US. In 2000 the stockmarket had reached a ludicrous degree of over-valuation on any sensible measure.

As one example, the S&P 500 traded on a price/earnings ratio (P/E) in the mid-30s compared to its long-term average of around 15. Earnings growth was irrelevant; they could not possibly grow rapidly enough to justify these prices. Consequently economic growth forecasts were irrelevant.

Twice last year I wrote inMoney Managementabout the widespread over-exposure of equities in portfolios — not because I can precisely foretell the future any better than anyone else — but simply because prices were so far divorced from fair value that it was self-evidently dangerous.

This raises interesting challenges for today. If we ignore the forecasters, how can we form a view on US stocks? On many measures the degree of excess has reduced.

However, while there is certainly far more to valuations than P/Es, it is arresting that, at October 31, the S&P P/E was 33. Despite the huge slump in share prices, P/Es have hardly declined and remain above twice their average level.

It is hard to feel unreserved optimism in the face of this. Yet most managers continue to forecast a rebound — as many have done through the duration of the decline (with some notable exceptions).

Perhaps there is an inherent bias to optimism in this. If you were promoting an equity fund with “risk control” parameters designed to prevent it from diverging far from a market index you’d want to believe the index would rise, wouldn’t you?

It’s just another reason to take forecasts with a grain of salt.

I also have some advice for those who are trying to succeed as forecasters. Forecast often, but don’t keep records.