Gunnar Bårdsen and Ragnar Nymoen 8836
4
22005 2010 2015(a) Consumer price inflation2020 2025 2030 20350
–27
6
5
4
32005 2010 2015(b) Unemployment rate (c) Wage growth2020 2025 2030 20352
110
8
6
4
2(^0) 2005 2010 2015 2020 2025 2030 2035
12
8
4
2005 2010 2015
(d) GDP growth
2020 2025 2030 2035
0
–4
12
8
4
0
–4
2005 2010 2015
(e) Import price inflation (f) Currency depreciation rate
–8 2020 2025 2030 2035
15
10
5
0
–5
–102005 2010 2015 2020 2025 2030 2035
Figure 17.3 Stochastic dynamic simulation of NAM for the period 2007(4)–2035(4)
The money market interest rate is kept constant at the 2007(3) level for the length of the simulation period.
The distance between the dotted lines indicates 95% prediction intervals.
In order to match the theoretical steady state above as closely as possible, the
solution in Figure 17.3 is based on a constant domestic money market interest rate
(Rt=R 2007 ( 3 )). The exogenousI( 1 )variables have also been given constant growth
rates for the length of the simulation period, e.g., foreign inflation is fixed at 2%.
The impression is clearly that NAM has an asymptotically stable steady state, even
under the assumption of a constant interest rate (the money market is in equilib-
rium by an endogenous money supply in the counterfactual “regime” defined by
the constancy of the interest rate). The annual rate of inflation is seen to stabilize at
a level just above 2.5%, and the natural rate of unemployment (in the Frisch sense)
appears to be 3%. Clearly, in this scenario an inflation rate of 2.5% is attainable
with very moderate instrument use. The seasonality of unemployment, modeled
by dummies, is clearly visible and represents no problem in terms of stability.
17.4 Macroeconometric models as tools for policy analysis
In this last section we discuss several aspects of model usage that are of relevance
for policy analysis. We begin with a practical problem, namely that congruent
modeling, or a high degree of “data coherence” to cite the influential Pagan (2003)
report on monetary policy modeling, gives rise to dynamic specifications that are
too complicated to be of any help when the task is to explain the basic policy
channels and lags between instruments and target. Nevertheless, section 17.4.1
shows that the need for simplicity in communication is not an argument for com-
promising empirical validity at the modeling stage, since it is always possible and
convenient to work from the empirically valid and complicated model to a sim-
ple and stylized model that contains the essential dynamics of the full model. In