18
Roland Dohrn
Even though correlations between short term indicators and forecast errors
were found in a considerable number of cases, the results give little reason for
optimism that a better use of information may help to improve forecast ac-
curacy. Quite often the correlations show a low level of significance. There are
only a few candidates whose ability to improve forecasts should be scrutinized
more thoroughly. One of them is the OECD leading indicator and another
share prices. More important is that co-variations are detected mostly either
for GDP only and not for any of its demand side components, or they appear in
some components, but not in GDP. As short term forecasts are made bottom
up, the study gives no hint, where to integrate additional iformation.
However, the limitations of this analysis should not be forgotten. Firstly, the
selection of indicators as well as the choice of lags is arbitrary. Maybe other in-
dicators or longer lags show better results. Secondly, only pair wise corre-
lations are calculated. It is not tested, whether forecasters draw the right con-
clusions from combinations of indicators.