THE ECONOMETRICS OF MACROECONOMIC MODELLING

Summary of findings—Euro-area data

The model comparisons in this section do not allow us to draw decisive conclusions. Some caveats no doubt apply: the presumptions of a clearly defined monetary policy for the economy under study, which are underlying the P*-model as it is laid out in Gerlach and Svensson (2003), is not favoured by adopting an observation period which starts nearly 30 years before the intro­duction of the Euro.[85] Likewise, the ICM—with its focus on the labour market influx on inflation—is probably a better model description of the national economies than for the Euro area.

That said—from the model evaluation and the forecast comparisons—some comparative advantages seem to emerge in favour of the (reduced form) AWM inflation equation: it is the only model that encompasses a GUM and it forecast encompasses the competitors when tested on 20 quarters of one-step ahead forecasts. The P*-model—based on the extended (AWM) information set—forecast encompasses the other models based on 36 quarters of one-step forecasts. In that context the NPCM appears to be a particularly poor model.

The results of the forecast competition are in accordance with the model evaluation in the preceding sections. The ICM is likely to suffer in fore­casting due to recursive instability in the long-run coefficients (table 2 in Jansen 2004) as well as in the short-run coefficients (Figure 8.7). Generally, we find that the models that are derived from the wider information sets (AWM and P* enhanced) do better in forecasting than those based on a narrower information set, mainly prescribed by theory, like the P*-model proper and the NPCM.

8.7 Empirical evidence for Norway

In this and the following sections, we compare an inflation equation which is a reduced form of the dynamic ICM for Norway with variants of the P*-model and the hybrid NPCM, as defined in Section 8.5. We also include in this comparison the inverted money demand function MdInv of Section 8.4.2 and we show that models based on the P*-formulation are more successful than the alternatives in capturing effects on inflation from monetary aggregates. The models are estimated on a common sample covering 1969(1)-2001(1), and they are presented in turn below, whereas data sources and variable definitions are found in Eitrheim (2003).