Gunnar Bårdsen and Ragnar Nymoen 853markedly over the last decades. The Lucas (1976) critique and the Clements and
Hendry (1999) analysis of the sources of forecast failures with macroeconometric
models are milestones in that process. Interestingly, the methodological ramifica-
tions of those two critiques are different: the Lucas critique has led to the current
dominance of representative agent-based macroeconomic models. Hendry (2001a),
on the other hand, concludes that macroeconometric systems of equations, despite
their vulnerability to regime shifts, but because of their potential adaptability to
breaks, remain the best long-run hope for progress in macroeconomic forecasting.
Since monetary policy can be a function of the forecasts, as with inflation forecast
targeting (cf. Svensson, 1997), the choice of forecasting model(s) is important.
The class of macroeconometric models we present in this chapter requires coher-
ent use of economic theory, data, and mathematical and statistical techniques.
This approach, of course, has a long history in econometrics, going back to Tinber-
gen’s first macroeconometric models, and has enjoyed renewed interest in the last
decades. Recent advances in econometrics and in computing mean that we now
have much better tools than, say, 20 years ago for developing and maintaining
macroeconometric models in this tradition (see Garrattet al., 2006, for one recent
approach).
Regardless of underlying theory, a common aim of macroeconometric model-
building is identification of invariant relationships, if they exist at all (see
Haavelmo, 1944, Ch. II). A well-specified macroeconometric model is a good
starting point for such a quest, since it provides an ideal test-bed for further over-
identifying restrictions of microeconomic behavior. Such a strategy is, in particular,
relevant to the challenges from behavioral economics, with implications for time
inconsistency (hyperbolic discounting), changing expectations (learning), asset
bubbles (herd behavior), etc.
Macroeconomic models of the representative agent, intertemporal optimizing,
type are said to have structural interpretations, with “deep structural parameters”
that are immune to the Lucas critique. However, when the model’s purpose is to
describe observed macroeconomic behavior, its structural properties are concep-
tually different. Heuristically, we take a model to have structural properties if it
is invariant and interpretable (see Hendry, 1995b). Structural properties are nev-
ertheless relative to the history, nature and significance of regime shifts. There is
always the possibility that the next shocks to the system may incur real damage
to a model with hitherto high structural content. The approach implies that a
model’s structural properties must be evaluated along several dimensions, and the
following seem particularly relevant:
- theoretical interpretation
- ability to explain the data
- ability to explain earlier findings, i.e., encompass the properties of existing
models - robustness to new evidence in the form of updated/extended data series and
new economic analysis suggesting, e.g., new explanatory variables.