Long Run Equilibrium Estimation and Inference : A Non-Parametric Application

The principal aim of this paper is to present some initial estimates and inferences based on Phillips' (Phillips 1991) theoretical results and suggested empirical procedures. The innovative aspects are: (1) it provides an early application of this methodology to macroeconomic time-series data, thereby giving initial evidence on both the potential gains and difficulties relative to traditional methods; (2) it extends the Engle and Granger (1987) type scalar cointegration methods to vector cases, in a way which is not conditional on the precise modeling of short-run dynamics as is required, for example, in the maximum likelihood procedure developed by Johansen (1988, 1991); and (3) it illustrates the outcome of some simple hypothesis tests on the long-run parameter values. This is possible because Wald test statistics involving coefficients on the I(1) regressors are not misleading. It therefore overcomes a major limitation of the Engle-Granger procedure. More specifically, our application is to a vector of Australian data on household disposable income, and aggregate and disaggregated consumer expenditure. The long-run consumer expenditure equations estimated could help to underpin a full continuous-time or discrete-time macroeconomic model, once satisfactory long-term relationships have been developed in other key areas