RDP 8701: The Australian Demand Function for Money: Another Look at Stability 1. Introduction

Milton Friedman's assertion that the demand for money is “a stable function of a few arguments” has arguably been one of the more influential statements made in macroeconomics. By the late 1970s, there were not too many industrial countries that did not adopt some sort of target or projection for growth in the money stock.

Yet at the very time this was taking place, the stability of “the” demand function for money went from being an accepted (and apparently empirically-documented) article of faith in many circles, to being seriously questioned. This was particularly so in the United States in the wake of the “missing money” episode beginning in 1974, but was not exclusive to that country. (See for example, Judd and Scaddlng (1982), Boughton (1980), and Atkinson et. al. (1984).)

In Australia, evidence on the stability or otherwise of the demand for money has been mixed. For M3,^[1] Juttner and Tuckwell (1973) concluded in favour of stability, but Sharpe and volker (1977), using essentially the same model but different techniques for testing stability, found that the equation was only stable when institutional changes were appropriately treated. The assessment of Davis and Lewis (1978), on the other hand, was that the need to incorporate extra variables to improve goodness of fit, the existence of large prediction errors from around 1972, and variation in the estimated co-efficient on the lagged dependent variable as new observations were added, constituted evidence that the instability that characterised demand functions for money overseas also operated in Australia.

In the case of M1^[2] there has been a difference of opinion as well, with Adams and Porter (1976) arguing that the demand for money is unstable, while pagan and Volker (1980, 1981), respecifying the Adams and Porter model, concluded that there was no convincing evidence of instability.

It is now some years since those studies were completed. Given the far-reaching change In the Australian economy generally and the financial system in particular in the last five to ten years, it is natural to ask what results still hold.

Recently, work by Drane, Marzouk and Valentine (1984), and Freeland (1984) has concluded, the latter somewhat tentatively, in favour of stability. This paper attempts to provide systematic evidence on this question. It does so by updating some existing studies of the demand for money and then subjecting them to accepted tests for stability, with particular attention paid to the early 1980's and to an important technical issue for the test procedures. In so doing, it seeks to go further than earlier work of Veale et. al. (1985), where stability was not formally tested.

The models considered are those of Sharpe and Volker (1977), Pagan and Volker (1981), Porter (1979) and Freeland (1984). The predominant monetary aggregate of interest in the paper is M3, although M1 is the subject of the Pagan and Volker work. This is defensible since M3 was arguably the central indicator of monetary policy for many years, at least in public debate. For example, from early 1976 until early 1985, the Australian Government maintained a conditional projection for M3 on an annual basis.

There are couple of important omissions in this approach. Firstly, the debate about stability of the demand for money might usefully be expanded to consider broad money,^[3] which is now accorded considerable attention in the assessment of financial conditions and monetary policy. To date this has been done only to a limited extent (see Drane, Marzouk and Valentine (1985), and Veale et. al. (1985)). This question is not addressed formally here; forthcoming work involving one of the authors will take up the stability of broad money further.

Secondly, it is noteworthy that the studies of concern here, in following a conventional pattern of single-equation estimation, diverge from another modelling tradition in the Australian literature, and one which has been particularly important in previous work at the Reserve Bank. The role of money as a buffer stock, and the explicit modelling of the effects of monetary disequilibrium on the course of other macroeconomic variables was discussed theoretically by Jonson (1976), and developed empirically in several generations of the simultaneously-estimated system first described by Jonson,

Moses and Wymer (1976). This latter approach potentially allows for a richer expression of the interaction between the observed money stock, interest rates and income than can be captured in a single equation.

The scope of this study is therefore necessarily limited. But it is of interest to systematically address the question of stability using single equations, if only because claims about stability or instability have been made based on such equations.

The procedure adopted is to reproduce, as closely as possible, the estimation results of the original study in the original sample period.^[4] Then the equation in question is tested for stability, both within the original sample period and over a longer sample which is extended in most cases up to the end of 1985.

Footnotes

Defined as currency and all deposits at banks held by the non-bank public. [1]

Defined as currency and current accounts at banks held by the non-bank public. [2]

Defined as deposits of the public with bank and non-bank institutions and public holdings of currency. [3]

It has been pointed out to us that an alternative approach might be to use the equation with the “best fit”. However, our objective was to test equations which had already been put forward as models of the demand for money. It seemed no less aribtrary to try to reproduce the original result and then test that than to search the data for some “best” alternative, which could not really be attributed to the original authors. [4]