THE ECONOMETRICS OF MACROECONOMIC MODELLING
Monetary effects in the inflation equation?
We find no effect of inflation in the money demand equations for Norway. Hence it does not make sense to interpret the money demand functions as inverted inflation equations. We have, however, experimented with a model where we consider money in real terms (mt — pt), the real interest rate on money and the yield spread as potential explanatory variables for inflation. These are the variables that enter the cointegrating relationship of the money demand equation in Eitrheim (1998), cf. Table 8.3.
This gives us a model which has several aspects in common with the inverted money demand relationship for the Euro area in Section 8.3.2. In addition to
6 Test statistics marked * and ** indicate significance at the 5% and 1% level.
Table 8.5
The MdInv model of inflation, including variables (in levels) from
the money demand relationship
Aypt = 1.1021Дзр(-1 + 0.2211Apt-2 + 0.0436Дре(
(0.0350) (0.0696) (0.0109)
+ 0.0587Д^_2 + 0.0272Дш^_ з —0.0208(mt_ і — pt-1)
(0.0115) (0.0170) (0.0103)
+ 0.0155 yt-1 — 0.0099(RTt-1 — Д4р^1) — 0.0262(RBt-1 — RMt-1) (0.0097) (0.0202) (0.0422)
— 0.0120 P dumt — 0.0586
(0.0010) (0.0577)
а = 0.45%
Diagnostics tests
Far(i-5) (5,113) = 1.6482[0.1530]
Farch(i-4)(4, 110) = 1.2934[0.2771]
X2normaiity(2) = 7.3731[0.0251]*
FHETx2(20, 97) = 2.6762[0.0007]**
FHETxiXj(65, 52) = 1.5171[0.0606]
FREsET (1,117) =7.6875[0.0065]**
Note: The sample is 1969(1)-2001(1), quarterly data.
the monetary variables we have included short-run effects of changes in energy prices Дpet, changes in output, and a composite dummy variable Pdumt which inter alia captures the effect of income policies in the late 1970s and 1980s. Even though the model fits the data reasonably well with an estimated standard error of а = 0.45%, the model does nonetheless fail in several of the mis-specification tests, reported in Table 8.5. The model captures the persistence in inflation through the included lags in price growth, and the effects from Дpet and Pdumt are reasonable, but the effects from the included monetary variables are more difficult to interpret. First, the results in Table 8.5 indicate that the effect on inflation of the real interest rate is insignificant. Assuming that the real interest rate is stationary, it is the other monetary variables from the long-run cointegrating relationship (i. e. the money demand equation in Eitrheim (1998)) that represent an excess money effect on inflation. A priori one would expect a positive rather than a negative effect from these variables on inflation.
Second, as will be clear when we present the forecasting properties of this relationship along with other inflation models for the Norwegian economy, that this model—which we have dubbed MdInv—suffers from severe parameter nonconstancies in the early 1980s and around 1994. This is a likely explanation why the model—when estimated on data up to 1990(4)—badly mispredicts inflation over the period 1991(1)-2000(4) (see Section 8.7.6). Thus, we conclude as we did for the Euro-area data: if we try to construct an inverted money demand relationship for Norway, forming an inflation equation based on the information set used in the money demand models in the preceding section, we find evidence of severe parameter non-constancy and resulting forecast failure.
