Palgrave Handbook of Econometrics: Applied Econometrics

892 Macroeconometric Modeling for Policy

.0036

.0033

.0030

.0027

0045

0042

0039

0036

0.008

0.006

0.004

0.002

0.002

0.003 0.004 0.005 0.006 0.007 0.008

0.003 0.004

H = 0

H = 0

H = 0

σinf × σgap; ICM

σinf × σgap; PCM

ΔL × Η; ICM

ΔL × Η; PCMr

↓

↓

↓

0.005 0.010 0.015 0.020

σinf × σgap; PCMr

40

30

20

10

(^0121086420)
H = 3
ΔL × Η; PCM
150
50
100
(^0121086420)
H = 6
1,500
1,000
500
(^0121086420)
H* = 11
Figure 17.8 Economic performance and optimal policy suggested by three specifications of
the supply side in the face of a supply shock
Left column: trade-offs between standard deviations of inflation gap,σinf, and output gap,σgap(horizontal
axis) associated with different (policy) horizon-specific rules in response to the demand shock. The trade-
offs are plotted for rules associated with policy horizons (H) in the range of 0–12 quarters. The trade-offs
associated with different horizons follow each other, where the one forH= 0 is indicated.Right column:
values of the relative loss function, denotedL(in %), at the different policy horizons (horizontal axis)
(Akram and Nymoen, 2008).
parameter values across models can lead to widely different policy implications.
Interestingly, it appears that imposing a set of parameter restrictions may have
stronger influence on policy implications than choosing a different functional form
of the model. Monetary policy based on a model that turns out to be an invalid
characterization of the economy and its transmission mechanism may lead to sub-
stantial losses in terms of economic performance, even when policy is guided by
gradualism, e.g., in the form of a long policy horizon.
17.4.4 Theory evaluation: the New Keynesian Phillips curve
Macroeconomics is an evolving science. New hypotheses and theories are put for-
ward, sometimes with far-reaching consequences, for policy, teaching and, of
course, also for model-building. A macroeconomic model project that lasts for
some time is therefore bound to face the need to adapt to new theoretical develop-
ments. However, new ideas in macroeconomics are usually partial, and can claim
superiority with respect to existing ones only by replacing oldceteris paribusclauses
by new ones. In a macroeconometric modeling context it therefore makes sense
to test the new theories before they are implemented. If the model is in opera-
tional use, say for policy recommendations, this step may be as much of a virtue