Demand for money and financial assets in New Zealand

Abstract

When a long run equilibrium relationship is found between the demand shares and the explanatory variables in the portfolio model, there exist a dynamic error correction representation of Engle and Granger (1987). (The study by Engle and Granger (1987) and others has introduced the statistical notion of cointegration of time series which can be taken as corresponding to the theoretical notion of long-run equilibrium relationship).

However, of the relatively small sample (when quarterly data is used there are 54 observations) and the autoregressive nature of the data, biases in estimating using the Engle and Granger (1987) procedure are likely to be high, Adam, 1998: Banatjee et al. 1993).

Assuming all variables in the long run share equations are I (1) and are cointegrated (Engle and Granger, 1987), the Engle Granger's 2-step procedure, "general to specific" methodology, is applied. The first stage regression of the long run share equations yields consistent estimates of matrix of long run parameters. In the second stage, the residuals obtained from the share equations are substituted in to the dynamic system error feedback equations (EFE) of Andersen and Blundel (1983) which is used to obtain parsimonious dynamic equations while holding the long run parameters fixed (Hendry, 1986). Thus, the adjustment of actual shares St. to desired long run shares is assumed to operate via a generalized error feedback mechanism, EFE (error feedback equations) of Andersen and Blundel (1983): The long run demand share equations are incorporated in the general single equation dynamic specifications using the residuals obtained from the cointegration equations.

The dynamic AIDS model allows to test the theoretical restrictions of homogeneity, symmetry, and negativity which must hold if the behaviour of the representative agent is to conform with the basic axioms of rational choice (e.g. convexity, transitivity).