A few tenths of a percentage point
Economic forecasting has increasingly been falling into both public and academic disrepute. Unfortunately, research institutions that conduct economic forecasts are not entirely blameless.
This has nothing to do with the poorly constructed pun made in the 1930s by the famous comedian Karl Valentin, who supposedly said that forecasting is a particularly difficult matter when you’re trying to predict the future. Maybe this is the reason why the German Federal Labour Court chose its words carefully in its decision on protection against dismissal in the beginning of the 1980s, referring to “future predictions” in order to avoid any possible misinterpretations in judicial practice. After all, it seems plausible enough that we could – much like in the movie Back to the Future – go back in time and make those “future predictions” with casual ease.
There is also no question that the researchers who conduct these forecasts do so with a high level of professionalism. These institutes employ elaborate and scientifically grounded methods and provide careful explanations with regard to the assumptions that form the basis of their prognoses.
In the discussions surrounding the validity of economic forecasting, what is at issue is a matter of a few tenths of a percentage point. Within a period of a few weeks, forecasting institutes changed their predictions regarding this year’s real GDP change rate from 0.75 per cent to 0.5 per cent. The rate was then adjusted again to 0.3, and then to 0.0 per cent, only to be corrected once more to -0.2 per cent. In some cases, these downward corrections were made by the same institutes. These developments were accompanied by headlines in the business section of prominent national daily newspapers, which certainly did not encounter opposition from many of the forecasting institutes. However, the aforementioned forecast values – and this is the important point – were all published within plausible projection intervals. There is no point in discussing whether the GDP will grow by 0.5 or 0.3 per cent, because there is no significant difference between these two values. To put it in other words: the probability of the GDP growing by 0.3 is only slightly higher than the probability of it growing by 0.5 or 0.1 per cent. And were it not for the serious effects on the reputation of empirical economic research, many experts would struggle to stifle a laugh at the uproar over the “latest” forecast values.
By now, the public must have the impression that the half-life period of forecasts values is decreasing at a rapid pace. This is, however, only true for a small number of cases, as the forecasting values hardly differ from one another. And yet, the public uses these different values to draw conclusions about the quality of empirical economic research, while media representatives focus their attention on the performance of economic forecasts, asking why German taxpayers have to pay for these practices – not to mention more substantive discussions concerning the very raison d’être of those institutes.
In order to remedy this unfortunate development, two different measures could be considered. Firstly, forecasters could focus more closely, or even exclusively, on projection intervals, and disregard the fact that the media puts too much emphasis on the mean value of the intervals. Secondly, forecasting institutes could limit their predictions to spring and/or autumn forecasts, even if this might offer them fewer opportunities to raise their public profiles. For the periods in between, it might be enough to publish sentiment indicators, which reflect the expectations of companies or analysts – and shame on anyone who thinks ill of it.