RDP 8803: Do Financial Aggregates Lead Activity?: A Preliminary Analysis 3. Overseas Experience

The question of the relationship between money and income was brought into prominence by Friedman and Schwartz (1963), in their voluminous study of the monetary history of the United States. Friedman and Schwartz argued that the empirical evidence in the U.S., especially the turning points in money and output over the period from 1867 to 1960, suggested a strong, stable relationship between money and nominal income, with the causality running from money to income.

The observation of Friedman and Schwartz was tested using reduced-form econometric models in later studies. These models usually involved regressions of current values of money or income on lags of both variables. The models were designed to allow tests of the predictive value of lags of money in explaining current values of income, or vice versa. If the researcher could establish the significance of lags of money in explaining income (even allowing for the information provided by own lags) then there was evidence that money “caused” income.

Sims (1972) tested the causal ordering of money and income for the U.S., using measures of money base and Ml, and nominal and real GNP, for the period 1947–69. Current values of GNP were regressed on future and lagged values of money and vice versa. Sims' test results showed a leading relationship running from money to GNP (both nominal and real), but not from income to money.

However, these results have been disputed on a number of fronts. Firstly, they have not been supported by evidence from other countries. Similar tests were applied to United Kingdom data over the period 1958–1971 by Williams, Goodhart and Gowland (1976). The U.K. study found evidence for one-way causality running from income to money, and some evidence of causality from money to prices, the opposite of Sims' findings for the U.S. On the basis of this evidence, the authors concluded that a more complicated causal relationship existed, in which both variables were determined simultaneously.

Cuddington (1981) proposed two reasons for these apparently contradictory results. He argued that the data were affected by the asymmetry existing between the large (relatively closed) U.S. economy and the small, open U.K. economy under Bretton-Woods. The difference could also be caused by the U.K. authorities' interest rate management policy. Cuddington found support for both propositions, particularly for the latter.

Secondly, the results were found to be sensitive to the inclusion of other variables. In a later study, Sims (1980b) added a short-term nominal interest rate to money and income, and found no evidence of causality from money to output.

Thirdly, the tests have been shown to be sensitive to the pre-filtering procedures applied to the data (see Feige and Pearce, 1979, and Stock and Watson, 1987). Cooley and LeRoy (1985) have shown that they are not strict tests of causality or exogeneity. They can be useful in testing one variable's value in forecasting another, but this type of “causality” is not equivalent to exogeneity. (This point is discussed in more detail in Section 4 below.)

On the basis of these studies, it would seem that the question of the causal relationship between money and income is still open and that the lead/lag relationship is not yet defined. After surveying the U.S. literature, Blanchard (1987) concludes that there is a strong relationship between money and output and that monetary policy affects output, at least for U.S. data, but that the evidence from the U.K. suggests otherwise. The technical critiques of this type of reduced-form analysis encourage care in the construction and interpretation of tests, especially with regard to the pre-filtering of the data and in drawing inferences about causality.

The relationship between credit aggregates and income has attracted less attention in the literature. Most of the discussion has assumed that financial aggregates lead (and cause) income or output and has concentrated on assessing the relative merits of money and credit as policy variables.

Benjamin Friedman is a prominent proponent of the use of credit as an indicator (and possibly as a target) of monetary policy. Friedman examines the comparative stability of money and credit aggregates with respect to income for U.S. data, using both simple regression and VAR techniques. Using results from this analysis, Friedman (1981) argues that credit is at least as stable in relation to activity as the major money aggregates, and that the inter-relationship between money and credit is important for activity. He concludes that credit aggregates should be used as an indicator in addition to money for the purposes of monetary policy.

Friedman (1982) conducted similar tests for data from Canada, Germany, Japan, and the United Kingdom, and again concluded that, in each country, credit aggregates exhibit stability comparable with that of money aggregates.

Offenbacher and Porter (1983), however, express doubts about the robustness of Friedman's results. They argue that slight changes in Friedman's use of VAR techniques or in the construction of the data used in the analysis, cause substantial changes in the results. From their own analysis, Offenbacher and Porter conclude that the evidence favours the use of money rather than credit aggregates as guides for policy.

Other U.S. studies which discuss the usefulness of credit aggregates differ in their methods and conclusions. Islam (1982) compares monetary and credit aggregates as intermediate targets by looking at income velocities and some simple regressions. Using evidence from the United States, Germany and Japan, he concludes that there is some support for the inclusion of a broad credit aggregate among financial indicators, rather than an exclusive focus on monetary targets.

On the other hand both Davis (1979) and Hafer (1984) find little evidence in favour of using a broad credit measure as an intermediate target. By using simple regression analysis, both authors conclude that credit aggregates add very little additional information about the economy once the monetary aggregates have been taken into account. Davis qualifies this conclusion, however, by noting that where innovation distorts the monetary aggregates, broad credit aggregates may become more useful as financial indicators. Fackler and Silver (1982) also conclude that although history provides no support for targeting a credit aggregate, these aggregates may contribute useful additional information until the innovations which distort the monetary aggregates subside.

These studies have focussed on the question of whether credit is a useful target for monetary policy: most have assumed that credit and money aggregates lead, or at least move contemporaneously with, activity. It is not clear whether credit would be a better target or instrument than money, but most studies support the consideration of credit as an indicator, especially during periods of deregulation and innovation.