Maybe you can reach more accurate data from much more different sources today than yesterday. Measuring techniques can be more accurate. Analysis methods can be more precise. So, I don't think forecasts are going worse.

"Forecasts are always wrong" it is said. In Air Transportation usualy +/- 30% wrong on the long term. Through the emergence of dynamic strategic planning and similar concepts forecasts will be adjusted over time as events unfold and operating conditions change. An interesting aspect would be the integration of "black swans" into the forecasting models since they tend to appear a lot lately (in last 15 years i could think of at least 4) and, obvisouly, with a huge impact.

Long term analyses of the accuray of macroeconomic forecasts for Germany have shown that they have not become more accurate. However, one must be careful in making such comparisons. A rather large portion of the forecast error can be associated to large errors in a relatively small number of years. Thus results are strongly influeced by the period you ar looking at. Furthermore one must not forget that there are "natural" linitations tor forecast accuracy, which are set by the accuray of data. For example, the mean absolute error of GDP forecasts made in the middle of a year are of about the same size as the average absolute revison of the GDP rate from the first release to the finally published figure.

"Great advances have been made in the development and validation of useful forecasting procedures over the past century. (...) However, it is difficult to find evidence that forecasting has improved in practice. Ascher (1978) concluded that forecast accuracy had not improved over time in his review of forecasting for population, economics, energy, transportation, and technology. In his review of the research on agriculture forecasting, Allen (1994) was unable to find evidence that forecasting practice in economics had improved over time; he then compared the accuracy of forecasts from 12 studies (22 series) before 1985 and 11 studies after 1985, finding only trivial differences in accuracy. Based on her review of 25 years of population forecasting research, Booth (2006) finds no evidence that the accuracy of population forecasts has improved over time. When McCarthy, Davis, Golicic and Mentzer (2006) replicated two surveys on sales forecast accuracy conducted 20 years earlier, they concluded that accuracy had fallen."

This paragraph, which is taken from the Golden Rule paper (pp. 28-29, see attached), summarizes our findings on that question to date. We are still looking for additional evidence, so please provide citations to other relevant sources.

Article Golden Rule of Forecasting: Be Conservative

Not included in Booth's review is my paper "European demographic forecasts have not become more accurate during the past 25 years", Population and Development Review 34(1)2008, 137-153. It supports her conclusion with the results from detailed analyses for 14 countries.

Note, however, that many forecast users do not distinguish between expected value and outcome. Before I cast a fair die, my best guess (expected value, median) is 3.5. But when I observe a result of 5 later on, does that mean that my forecast was wrong? No!

Hello all. Recently I found this place in the net: International Journal of Forecasting, with many documents about the question plus a critical and autocritical vision. They afirm that the problem is in the methods and the way they are applied without enough care, mainly in social and very wide contexts of research with too many variables, and without prior-meta-analysis about main variables. Thanks, emilio