910 Macroeconometric Modeling for Policy
and from that perspective structural breaks are seen to be not entirely negative,
since there can be progress in macroeconometric modeling through forecast failure.
It follows from our approach that we agree with those who conclude that guid-
ance from economic theory is important in the forecasting process (cf. Elliott
and Timmermann, 2008). But theory and modern econometrics do not provide
immunity from new structural breaks. Nevertheless, the case for macroeconomet-
ric system of equations models may, to a large extent, depend on their ability to
forecast relatively well compared to competing forecasting methods. Making mod-
els sufficiently adaptive to a structural break, once the evidence is there that it has
occurred, is a necessary step in that project.
It is possible to look to other forecasting disciplines for inspiration. Meteorol-
ogists developed their forecasting theory, models and routines between the two
world wars. Those forecasts were based on a deterministic model of the dynamics of
the atmosphere. Then, in the mid 1960s, the understanding of the dynamics were
completely altered with the development and acceptance of chaos theory, with the
logical consequence that accurate weather forecasting is impossible. Thirty years
after the arrival of the meteorologists’ “impossibility theory” for forecasting, the
weather forecasts are undeniably more accurate than ever. Hence, the weather is
predictable even ten days ahead, despite the chaos represented by the underlying
forecasting models. It is perhaps unlikely that we will witness something similar in
economics, but there are nevertheless parallels. Meteorologists have precise theory
and almost continuous updating of initial conditions. Policy-oriented modeling
may have to live with idealized or partial theories, and with variables that are mea-
sured at relatively long time intervals, and which are influenced by measurement
errors. Yet, as a discipline we have developed methods and strategies that are quite
good at making the most of “small data amounts.” Hence, while meteorologists
can rely on the theoretically specified model of weather dynamics, the interaction
between economic theory, econometric theory, good model selection procedures,
and diagnostic testing have together greatly improved our capability of modeling
the macroeconomy, thus providing models that can aid policy decisions.
Appendix: Data definitions and equation statistics
Variables
The model employs seasonally unadjusted data. Unless another source is given, all
data are taken from FPAS, the database of Norges Bank.
The model is developed and estimated with Oxmetrics 5 (http://www.oxmetrics.net)
and then re-estimated and simulated with Eviews 6 (http://www.eviews.com).
VTrade-weighted nominal value of the krone based on import shares of trading
countries.
GGovernment sector consumption expenditure, fixed 1991 prices. Millions (Mill):
Norwegian krone (NOK).
LNominal credit volume. Mill. NOK.
RMoney market interest rate (three month euro–krone interest rate).