12
Roland Dohrn
ment index (CSI) which is published by the European Commission. IFOC
and IFOE are available during the last week of each month presenting the
results for current month. For our test it is assumed that the forecasters
know the data of the month before their forecast is published. The CSI is is-
sued with a longer lag. It is assumed forecasters know the result of the last
but one month. For forecasts published in July, e.g., the May CSI is included
in our test.
- Three variants of leading indicators from official sources are considered:
Total (NOMT) as well as foreign new orders in manufacturing (NOMF) are
published by the German statistical office approximately 6 weeks after the
end of the month the data were collected. New orders in construction
(NOC) are issued about 2 weeks later. When forecasters issue their July
forecast, they know the May data on orderbooks as a rule. As there are huge
short term fluctuations in the data sometimes, they are smoothed by calcu-
lating two month averages of seasonal adjusted indicators. Using month
over month changes makes sure that the data are stationary.
- Various interest rates as well as the real effective exchange rate are consid-
ered as monetary variables. These data are available very shortly after the
end of a month. Therefore averages of the rates for the month before the
forecast is published are included in the analysis. Thus, the yield curve (YC
= long term rate LR minus short term rate SR), and the real short term rate
(RSR) as an indicator of monetary policy stance. As monetary policy often
shows its impact with long lags, also averages of the short term (SR-1) , the
long term (LR-1), and the real short term interest rate (RSR-1) as well as of
the yield curve (YC-1) over the entire year before the forecast are intro-
duced into our calculations. Kirchgassner/Savioz (2001) found a correlation
between forecast errors and interest rates when they tested for such long
lags4. As the interest rates are already stationary, no transformation is re-
quired. Furthermore the real effective exchange rate is included as a mone-
tary indicator, in two variants: the year over year change in the last month
(REER) as well as of the average of the last three months (REER3).
- As financial market indicators share price index CDAX is included in our
study. Again, two transformations are tested: Firstly, the year over year
change in the month before the forecast (CDAX) is published, secondly the
change of the three previous months (CDAX3).
- Finally, the OECD composite leading indicator (OECD) is tested. It com-
bines various data already considered here, namely the ifo business climate
(IFOC) and total new orders in manufacturing (NOMT) with financial data
(YC) and additional information taken from the ifo business survey
(OECD 2002: 31).
4 We refrained from including monetary aggregates such as M3 into the calculations as there is a
break in the time series due to the start of EMU.