RDP 33 Inflation: Prices and Earnings in Australia V Implications for the Analysis of Inflation and Anti-Inflation Policy

The estimated results for the rates of change of earnings and prices reported in the preceding two sections need to be interpreted with caution. These are only preliminary results, and as highlighted by the paper, several deficiencies still exist in both the theoretical specification of the basic model, the time dimensions of the empirical representation of the model and some of the data. For these reasons, this section will be brief and comparisons with other studies will be minimal.

While both unconstrained and constrained versions of each equation are presented, the preferred equations are the constrained ones with the former only being presented for the purpose of comparison. Formal testing of the various constraints generally shows that their imposition as sets causes a statistically significant deterioration in the fit of the equations relative to the unconstrained estimates. Nevertheless, the constrained estimates have attractive statistical properties and appear to lend considerable support to the hypothesis that the data are consistent with the theoretical specification of the model.

Before interpreting the estimated results, it should be re-emphasised that the estimated coefficients are impact coefficients and take no account of general equilibrium interactions between the various independent variables. Further, as the equations are applied in explaining the historical movements of prices and earnings, their predictive ability for the future needs to be interpreted with caution.

The constrained equations for the rate of change of average weekly earnings indicate that increases in indirect and payroll tax rates have a well-determined deflationary impact on earnings of a similar magnitude to that of the tax rate change. Changes in the income tax rate have a much smaller but still deflationary impact on earnings. It is the introduction of tax variables that is the main departure from earlier Australian inflation analyses and the results appear to justify their inclusion. Of the variables appearing in other studies, past changes in award wages appear to contribute more, and world prices less, to the explanation of movements in earnings in this study than in previous ones. This may be partly due to the different data periods used, partly to the different specifications used (perhaps suggesting that the true specification has yet to be found) and in the case of world prices, partly to the adjustment for the effective exchange rate (an adjustment previously ignored). The reasonably good performance of the price expectations series in the second formulation of the earnings equation suggests that a certain degree of substitutability exists between aggregate price expectations and award wages. This may lend some weight to the representation in equation A.45 of the rate of change of award wages as a function of expectations of price and productivity changes. However, the exact extent of the relationship cannot be ascertained from the estimated coefficients.

With respect to the characteristics of wage inflation in Australia, no comment can be made about the natural rate of unemployment.^[27] But in this model there are two tests for a long run trade-off. For there to be no long run trade-off between inflation and excess demand (and by implication, unemployment) the sum of the coefficients on past rates of change of award wages and the expected rate of change of world prices should add to unity in the first specification and those on changes in aggregate price expectations and expectations of changes in world prices should add to unity in the second specification. A second test,^[28] applying to both specifications, is that the sum of the coefficients on the payroll and income tax variables should add to minus unity. Both of these tests are constrained to occur in the constrained equations. However, the formal test of these constraints indicates that it is very unlikely that they would occur naturally from the given data. Casual observation of the unconstrained equations suggests that the first test could well apply naturally in each case while the second test is very unlikely to occur. Given the insignificance of the coefficients on the tax variables in each equation it is reasonable that the first test is the more reliable of the two. Imposing the first test independently of the second, provides F statistics^[29] of .261 and 1.475 for the first and second specifications respectively. Thus the estimated equations both appear to be consistent with the hypothesis that Australia does not have a long run trade-off between inflation and unemployment.

From the estimated price equations it appears that the main stimulus for increases in each measure of prices comes from expectations of increases in the earnings rate with a moderating influence coming from increases in productivity. This provides some support for the often quoted long-run equilibrium condition that the rate of change of prices will tend to equal the rate of change of earnings less the rate of change of productivity. World prices and the exchange rate appear to have a well determined positive influence on domestic prices in the short-run although the lack of significance on the variable entering the equations in the unconstrained form must cast some doubt on the strength of this relationship. Excess demand does not appear to have an important or well-determined relationship with prices in the short run. This may be due to either an under-specification of the theoretical model for prices or the use of an inefficient proxy for excess demand in the estimated equations. The need to correct for first order autocorrelation supports the former hypothesis.

In further work, it is intended to correct the deficiencies of the current study already highlighted, and to use the wage and price equations in a complete model of the economy to simulate various anti-inflationary policy packages and examine the costs, in terms of unemployment, of reducing inflation over differing time horizons. In addition, the various equations will be estimated over several sub-periods in an attempt to determine whether the characteristics, and thus the implications for anti-inflationary policy, of inflation in Australia have altered significantly over the fourteen year period used in this study.

Footnotes

For this, a further equation linking the constructed excess demand variable to the unemployment rate would be required. [27]

Strictly speaking these tests are not independent but are coefficient restrictions which should hold simultaneously. [28]

For an explanation of the F test see footnote 21 on page 18. [29]