RDP 9105: Inflation in Australia: Causes, Inertia and Policy 5. The Structural Model: Estimation

The variance decompositions pertain to the determination of inflation in the short-run. In this section we examine the determinants of the long-run, equilibrium, inflation rate.

The estimating equation for the structural model of Section 2 is derived by substituting equation (5) into equation (13):

Equation (14) is non-linear in the parameters β, γ, λ₀, λ₁, λ₂, λ₃ and λ₄. There are five such equations in our model, corresponding to each of the five countries that we examine. We think it is reasonable to presume that contemporaneous inflation disturbances could be correlated across countries and so we estimate the model by non-linear Seemingly Unrelated Regressions. The model is estimated using annual data over the period 1964–89.

The estimation results are reported in Table 3. They are very encouraging. The fit of all the equations is good; there is no evidence of any serial correlation in the residuals and only for Australia is there any evidence of heteroskedasticity.^[11] In general, the estimated standard errors are small relative to the coefficient estimates. Very few of the estimated parameters have the wrong sign and, of the ones that do, none is significantly different from zero at the one percent level.

The estimates of γ show that in Japan only about 30 percent of the disequilibrium inflation is reflected in the change in inflation the following year, compared with about 50 per cent in Australia, 60 percent in the United Kingdom, 80 per cent in the United States and 90 per cent in New Zealand. The estimates of β, which indicate the proportion of the change in the equilibrium inflation rate that is observed contemporaneously, also vary markedly across countries. Here the relative rankings of Japan and New Zealand are reversed, with the greatest change occurring in Japan (about 80 percent) and the least change in New Zealand (about 50 percent). Australia has the median estimate with 63 percent.

Generally speaking, the overall degree of inflation inertia (given by the sum of β and γ) does not vary greatly across countries. However, the divergence of the estimated β′s and γ′s indicates that the nature of the inflation dynamics does vary across countries. Australian inflation displays slightly more inertia than the international average, but this difference is not terribly large.

The estimated values of the equilibrium inflation parameters (the λ′s) show a consistent pattern across the five countries. In particular, in every country, nominal wage growth appears to be far more important than the growth of money in determining the steady state inflation rate (λ₃ > λ₁). In the Australian case, for example, a five percent increase in wage growth leads to an increase in the equilibrium inflation rate of about four per cent, while the same increase in money growth leads to an increase in equilibrium inflation of slightly more than one percent.

A determined quantity theorist would no doubt interpret these results differently, arguing that nominal wages are not exogenous and the apparent causation from wages to prices masks the true structural relationship, which runs from money to wages and prices. While we do not deny the likelihood that tighter monetary policy will exert some downward pressure on wages growth, either through expectations of lower price inflation or through a weaker labour market, we nevertheless believe that such an argument would be beside the point.

Our interpretation of these results is that whatever the source of any disinflationary impulse, the response of wages is of paramount importance if inflation is to be reduced on a sustained basis. In a world where labour markets resemble the textbook model of perfect competition, a tightening of monetary policy will almost certainly be accompanied by a rapid fall in nominal wage growth, and concomitantly, price inflation. In the real world where the nature and behaviour of labour market institutions matters, such a response is not guaranteed. A tightening of monetary policy that is not accompanied by a slowing of wage growth will lead to a fall in inflation, but only via a contraction of profit margins. This strategy might be effective in the short-run but is unlikely to prove an acceptable method of lowering the inflation rate over a long time horizon.

The estimates of λ₄ show that increased productivity growth reduces inflation, as we would expect, but the extent of this reduction varies markedly across countries. The size of the coefficients shows, moreover, that the quantitative effect on inflation of increased productivity growth is likely to be modest. For example, suppose that “microeconomic reform” permanently increases the rate of total factor productivity growth in Australia by two percentage points per year, which would seem to us to be the upper bound for what might be achieved. The resultant fall in the steady state inflation rate will only be about one percent.

The estimates of λ₂ show that foreign inflation (measured in domestic currency), in every country except New Zealand, has negligible effects on the equilibrium inflation rate. This means that, New Zealand aside, foreign inflation has no explanatory power beyond that of the other variables.

Footnote

Specifically, at the five per cent level of significance, we can reject the null hypothesis that the variance of inflation errors in Australia is unrelated to predicted inflation. However, since we are conducting 25 tests, we would expect at this level of significance to spuriously reject one null hypothesis. [11]