Front Matter Page
Research Department
Contents
Summary
I. Introduction
II. The Framework
1. The maintained hypothesis
2. Testing for aggregation
3. Structural shifts and data breaks
4. Expected rates of depreciation
5. The measure of money
6. Multi-country aggregation
III. Empirical Results
1. Integration tests
2. Testing for cointegration
3. Dynamic error correction models
IV. Conclusion
Text Tables
1. Shift Dummies for Ml Demand Equations
2. Tests for Integration, 1972:4-1990:4
3. Cointegration Tests, 1970:4-1990:4
4. Cointegration Tests, 1978:4-1990:4
5(a). Estimated Cointegrating Vectors, 1972:4-1990:4, OLS
5(b). Estimated Cointegrating Vectors, 1972:4-1990:4, SUR
5(c). Estimated Cointegrating Vectors, 1972:4-1990:4, Restricted SUR
6(a). Money Demand Excluding Currency Substitution, 1972:4-1990:4, SUR
6(b). Estimates Excluding Currency Substitution, 1972:4-1990:4, Restricted SUR
7. Cross-Equation Residual Correlations
8. Error Correction Models – Summary Statistics, 1972:4-1990:4, SUR
9. Cross-Equation Residual Correlations
References
Summary
Recent studies have suggested that one can specify a stable aggregate demand for money for the countries participating in the European Monetary System (EMS). This paper evaluates two alternative interpretations of this result: first, the relationship between money demand and its determinants may be similar enough in the different countries that not much is lost by aggregating across national borders. Second, there may be currency substitution: if residents of EMS countries hold their money in a variety of European currencies and shift among them in response to exchange rate expectations and other difficult-to-measure factors, the demand for money in the EMS as a whole may be more predictable than in any one country.
The paper presents estimates of demand for narrow money in the Group of Seven (G-7) industrial countries; the smaller EMS countries are omitted because of data limitations, and the three non-European G-7 countries are included to allow for possible currency substitution outside as well as inside the EMS. Within a two-stage error-correction framework, Seemingly Unrelated Regressions (SUR) estimation is used to capture possible interaction between demand for money in the different countries and to permit tests of aggregation restrictions, namely, that the coefficients on income and interest rate variables for the money demand equations are the same for the four European G-7 countries. A common specification for the money demand equation is used for all seven countries, with dummy variables added to account for breaks in the data series (and for episodes of financial innovation documented in the literature). Exchange rates and foreign incomes are used to capture currency substitution.
In the first set of results, a cointegrating equation is estimated for the levels of the variables. It is found that for most countries, the currency substitution variables are needed to obtain a cointegrating relationship, while the aggregation restrictions do not appear to be consistent with the data. These results are borne out when dynamic error-correction equations are estimated using SUR: tests reject the hypothesis that currency substitution does not affect money demand and also the imposition of the aggregation restrictions on the four EMS countries. Moreover, both the static and dynamic equations yield significant and often negative correlations among the errors in money demand in different countries, and this also suggests cross-border shifts in money holdings.
These results support the view that currency substitution, and not merely similarities in money demand relationships across countries, may be responsible for the success of cross-border aggregation in money demand estimation. If borne out in further research, they would imply that national moneys may become more difficult to predict and control, strengthening the case for implementing monetary control on a supranational level, such as through a European central bank.