RDP 9314: The Demand for Money in Australia: New Tests on an old Topic 2. The State of Play

There is an extensive theoretical and empirical literature on money demand and excellent summaries of recent developments are provided by Goodhart (1989) and Cuthbertson (1991). At its most rudimentary, demand for money is motivated by the need to fund transactions, the need to hold money to fund unanticipated payments, and the decision to hold wealth in the form of money, which may be motivated by Keynes' speculative motive or by general portfolio theory. According to the transactions and precautionary motives for money demand, a scale transactions variable such as income, expenditure or wealth is a key variable in any money demand equation. The Baumol-Tobin inventory-theoretic approach to transactions demand also implies that money demand is inversely related to the opportunity cost of funds, which may be proxied by an interest rate or vector of interest rates. The speculative motive implies that the demand for funds is also inversely related to the interest rate while portfolio models imply that the own-rate of return enters positively and rates on substitute assets enter negatively.

Assuming that agents do not suffer from money illusion, the demand for money is a demand for real balances and this is conventionally expressed as a function of real income and a nominal interest rate or vector of interest rates. While this is the approach followed here, we also test the relationship between money, income and interest rates all defined in nominal terms. The results do not differ significantly from those obtained using real money and real activity. When the interest rate is defined as the own-rate of return on the monetary asset, the inflation rate and the two-year bond yield are also included as cost variables. In the presentation of the results, the inflation rate is not included since we found that it did not contain explanatory power in addition to nominal interest rates.[1]

This representation is purely static. The view that agents adjust their money holdings to some desired level is well accepted and attained more fulsome expression in the Goldfeld (1973) equation. The subsequent failure of this and other specifications in the 1980s is well documented. If money, income and interest rates are cointegrated, however, then the short-run dynamic relationship can be neatly modelled in the corresponding error-correction representation, which is a more general representation of money stock adjustment than in the Goldfeld equation (see Ericsson, Campos and Tran (1991)).

A summary of the Australian literature is provided in Table 1. The studies tend to define money in real terms, with the exceptions being Orden and Fisher (1993) and Juselius (1991). The studies also tend to focus on M3, which probably reflects its historical prominence, but tests have been conducted on the other well-known aggregates. The researchers draw on a variety of estimation techniques, though the Engle-Granger and Johansen procedures are the most popular. Overall, there is evidence supporting a long-run money-income relationship in the cases of currency and broad money. The evidence in the case of M1 and M3 is more ambiguous, with the result depending to some extent on the definition of independent variables, econometric technique and time period. Both Orden and Fisher and de Haan and Zelhorst find that M3 and GDP were cointegrated before deregulation but not after. Lim and Martin (1991), on the other hand, conclude that M3 and GDP are cointegrated even after deregulation.

This paper reviews the evidence for all these aggregates. With financial liberalisation and innovation, the range of bank deposits and their substitutes has expanded rapidly, and distinctions between money and non-monetary financial assets have become even less clear than before. A study of a range of aggregates is therefore appropriate.

The paper also conducts tests for various measures of the money base. There is no shortage of proposals for using the monetary base as an indicator or target (Harper (1988), Sieper and Wells (1989), Porter (1989) and McTaggart and Rogers (1990)). A feature of the existing literature in Australia is the lack of much empirical evidence on the indicator properties of the money base, and this paper seeks to redress this.


If the Fisher effect holds, expected inflation is included in the nominal interest rate. There is, however, the Friedman-type argument that the inflation rate should enter the equation separately as the opportunity cost of the monetary asset relative to real assets or other excluded financial assets, though the empirical importance of this has been questioned in the US (Emery (1991)). Baba, Hendry and Starr (1992) argue that if the Fisher effect is somehow prevented from working, for example by nominal interest rate ceilings, or if inflation and interest rates are only imperfectly correlated, then the inflation rate should enter the estimating equation separately. To check this, we included inflation in the estimation using the Johansen procedure but with little change to the results. Consequently, we report only the most conventional money demand equation test results. [1]