OF FORECASTS AND FORECASTERS

One of the major interests at the annual conference of the National Association of Business Economists is the release of the results of the survey of the economic forecasts of the 4,000 members of the organization. This year's results, presented in mid-September in Boston, show that the members are somewhat more gloomy than the Administration about the prospects for 1987.

The median forecast for real growth for 1987 was 2.9 percent, as compared with the Office of Manage ment and Budget forecast of 4.2 percent and a Congressional forecast of 3.9 percent. The median inflation forecast of the economists was 3.7 percent, which would represent a substantial increase over the 2 percent likely for 1986.It should be emphasized that the forecast figures do not represent a group forecast. Rather they are the medians of the individual forecasts, and sometimes conceal a wide range of opinion among individual forecasters. This, in fact, was the case for the inflation outlook for 1987, where forecasts ranged from a 0 percent increase to a whopping 8 percent, a figure that would certainly once again set the policy alarm bells ringing.

Nevertheless, there is no doubt that there is an element of consensus forecasting in the survey result. Each economist, like everyone else, looks for peer support in his opinions and is influenced by the forecasts of others. Each realizes that a wrong forecast widely shared will attract less attention than an error made only by himself.

Another factor that tends to draw the forecasts together is the fact that they are almost all very heavily influenced by the latest data available. For example, the slow growth of the second quarter of 1986 has resulted in a downward revision of the median forecast for real business investment in 1987,

from a 5 percent forecast in the early summer to only 2.5 percent in the current survey. While such influences are probably inevitable, forecasts virtually always incorporate a bias in favor of a continuation of the trends indicated by the most recent data.

This can result in embarrassment when figures are revised. As an economist with the Organiza tion of Economic Cooperation and Development (OECD), I was closely involved with forecasting myself. In one instance a "weak growth" forecast was developed that relied heavily on the previous quarter's rise in a reported increase in the personal savings rate. After the release of the forecasts, however, the figure was revised to show no change, thus undermining the rationale for the forecast.

Another strong influence in forecasting is the appeal of past averages. If the average of real gross national product growth for the past 10 years was 3.0 percent, then forecasts for next year are not likely to diverge strongly

from 3.0 percent. This is partly an inevitable result of the statistical techniques used, since regression and other forecasting tools represent types of weighted averages of past relationships. But it is also caused by the same risk avoidance of the economists: a forecast that deviates substantially from past trends and turns out wrong will attract much more attention than a ''conservative" forecast that turns out wrong.

The results of these influences are readily observable to users of forecast. At the OECD we received many forecasts from other sources, both public and private. I found that it was quite difficult to forecast the economy - but much easier to forecast the forecasts of the economy. A fairly reliable advance estimate of the forecasts for a particular variable, say, real GNP growth, could be obtained by taking an average of the most recent quarterly GNP growth figure and the trend level for GNP growth over the previous five to 10 years. If the economy last quarter grew much more rapidly than the average of recent years, then it was a good bet that new forecasts would split the difference, projecting near-term growth lower than last quarter, but still higher than the past averages.

Admittedly, reliable forecasts of forecasts are not quite as useful as reliable forecasts of the economy. On the other hand, if you knew what the forecasts would be, you could save yourself the cost of purchasing them!

The innate conservatism of forecasters is also observable in comparing past survey forecasts of the business economists with the actual outcomes. The forecasts for real GNP growth and for inflation are much less volatile than the actual outcomes for the corresponding years. (The standard deviation of the annual forecasts for inflation of the period 1976-1985, for example, was 1.3 percent, whereas the standard deviation for the corresponding outcomes was 2.2 percent.) A study presented at the conference also showed that the relatively poor results of many forecasts for industrial production in the first half of the 1980s was related to the greater volatility of industrial production figures during that period.

What this means is that economists' forecasts are likely to be reasonably good when there is little volatility in the economy. If real GNP and its components are growing at an even pace, inflation is stable, and there is little deviation from quarter to quarter, the forecasts probably will be accurate. On the other hand, if there is much volatility, with rapid reversals of direction and major deviations from trend, then many of the forecasts are likely to go badly astray.

Unfortunately, this is almost the same as saying that the forecasts are fairly accurate when you don't need them, but not much use when you do.

For the full story: Log In, Register for Free or Subscribe