@techreport{Gluck2004Forecast,
abstract = {We start from the assertion that a useful monetary policy design should be founded on
more realistic assumptions about what policymakers can know at the time when policy
decisions have to be made. Since the Taylor rule \textendash{} if used as an operational device -
implies a forward looking behaviour, we analyze the reliability of the input information.
We investigate the forecasting performance of OECD projections for GDP growth rates
and inflation. We diagnose a much better forecasting record for inflation rates compared
to GDP growth rates, which for most countries are almost uninformative at the time a
Taylor rule should sensibly be applied. Using this data set, we find significant
differences between Taylor rules estimated over revised data compared to real-time
data. There is evidence that monetary policy seems to react more actively in real time
than rules estimated over revised data suggest.
Given the evidence of systematic errors in OECD forecasts, in a next step we attempt to
correct for these forecast biases and check to which extent this can lower the errors in
interest rate policy setting. An ex-ante simulation for the years 1991 to 2001 supports
the proposal that correcting for forecast errors and biases based on an error model can
lower the resulting policy error in interest rate setting for most countries under
consideration. In addition we investigate to what extent structural changes in the policy
reaction behaviour can be handled with moving instead of expanding samples.
Our results point out that the information set available needs a careful examination
when applied to instrument rules like those of the Taylor type. Limited forecast quality
and significant data revisions recommend a more sophisticated handling of the dated
information, for which we present an operational procedure that has the potential of
reducing the risk of severe policy errors.},
author = {Heinz Gl\"{u}ck and Stefan P. Schleicher},
copyright = {http://www.econstor.eu/dspace/Nutzungsbedingungen},
keywords = {C82; C53; E52; 330; Monetary policy rules; economic forecasting; OECD; real-time data; Konjunkturprognose; Prognoseverfahren; Statistischer Fehler; Taylor-Regel; Geldpolitik; Sch\"{a}tzung; G-7-Staaten},
language = {eng},
number = {2004,30},
title = {Forecast quality and simple instrument rules: a real-time data approach},
type = {Discussion paper Series 1 / Volkswirtschaftliches Forschungszentrum der Deutschen Bundesbank},
url = {http://hdl.handle.net/10419/19497},
year = {2004}
}
