THE ECONOMETRICS OF MACROECONOMIC MODELLING

The New Keynesian Phillips curve

We estimate a hybrid NPCM as described in Section 8.5.3 (cf. Chapter 7 for further details). Using the instruments of Gall et al. (2001)[77]—five lags of infla­tion, Apt, and two lags in the wage share, wst, and output gap (gap)—we are able to replicate the results for the hybrid model in Chapter 7, which in turn are representative for the empirical results reported in Gall et al. (2001). We have chosen to estimate a small simultaneous model where the inflation lead Apt+1 and the wage share wst are specified as functions of the instruments and full information maximum likelihood estimation[78] then yields the following inflation equation:

Apt = — 0.0008 + 0.72Apt+1 + 0.31Apt_ 1 + 0.002 + dummies (0.006) (0.07) (0.07) (0.008)

a = 0.00232 (8.22)

1972(4)-2000(3)

Single equation diagnostics

Far(i-S)(5, 96) = 4.55(0.001**] FArch(i-4)(4, 97) = 0.87[0.48]

X2normality(2) = 5.16[0.08] Fhetx? (18, 86) = 1.56[0.09]

Systems diagnostics

FAr(i-5)(45, 262) = 9.45[0.000**]

X2’" normality (6) = 8.64[0.19]

F2HETx2(108, 471) = 1.38[0.01*]

In (8.22) we have augmented the NPCM equation with the significant dum­mies from the other models. Increasing the information set by adding more instruments does not change the estimates for the NPCM equation. The dum­mies reduce the estimated a for the NPCM by 10%, but this is still 10-20% higher than the other three model classes. The highly significant FAR(1-5)-test in (8.22) is not only due to first-order autocorrelation (which is consistent with the New Keynesian Phillips curve theory[79]), but reflects also higher order auto­correlation. Figure 8.14 underscores that the coefficients of the forward and the backward terms of the NPCM are recursively stable, as is also the wage share coefficient at a zero value.