CANADA adopted a fluctuating exchange rate system in September 1950.1 Since then, the foreign exchange value of the Canadian dollar has been determined by the interplay of market forces, except for some official intervention in the exchange market reported as intended to prevent excessive short-term fluctuations. Since 1952, this official intervention has been exercised on only a moderate scale, and official sales or purchases of exchange by the Canadian Exchange Fund Account have not been the dominating influence in determining exchange rate fluctuations from one quarter to another. The combined current and long-term capital account has shown substantial quarterly net balances. These balances have varied between plus $200 million and minus $200 million, and have exceeded $100 million, or about 10 per cent of Canada’s quarterly export receipts, in nearly half of the quarters. In view of these relatively large swings in the current and long-term capital account and of the moderate extent of official intervention in the exchange market, the stability of the Canadian dollar during the last eight years is remarkable. Since 1952, the exchange rate has remained within the range of US$1.00 to US$1.06, and the change from one quarter to another has never exceeded 2 per cent. It will be argued below that this high degree of stability was chiefly the result of the responsiveness of equilibrating short-term capital movements to changes in the exchange rate.
The theory of the determination of the foreign exchange rate, which has been developed and refined during the last three or four decades, is largely a theory of long-run equilibrium. The discussion has centered on the problem of the stability of the foreign exchange market in the absence of capital movements.2 Lack of national income and balance of payments data for periods shorter than one year has prevented any empirical study of the short-run balance of payments adjustment process under flexible exchange rates even in the few historical cases where a fluctuating exchange system was in operation for a longer time period and under reasonably normal conditions.3 Canada’s foreign exchange experience during the 1950’s and the availability of quarterly data on Canada’s national accounts and balance of payments afford an opportunity to add to our knowledge of the process by which the level of a fluctuating exchange rate is determined in the short run, and of the role played by a fluctuating exchange system in the economic policy of a dependent economy.
Mr. Rhomberg, who is a graduate of the University of Vienna and of Yale University and has been a member of the faculty of the University of Connecticut, is an economist in the Finance Division.
This paper was presented at the Winter Meeting 1959 of the Econometric Society in Washington, D.C. It presents part of the research for the author’s doctoral dissertation, Fluctuating Exchange Rates in Canada: Short-Term Capital Movements and Domestic Stability (Yale University, 1959). The author wishes to express his gratitude to the M.I.T. Computation Center and the International Business Machines Corporation for the use of the Electronic Data Processing Machine 704 at the Massachusetts Institute of Technology, where the largest part of the computational work was done, and to the National Science Foundation and the Cowles Commission for Research in Economics for the use of the IBM 650 at the Yale Computation Center, where some of the preliminary computations were carried out. He also wishes to thank Mr. H. C. Lampe, Department of Agricultural Economics, University of Rhode Island, who cooperated with him in the coding of the limited-information maximum-likelihood program.
At least one exception should be explicitly mentioned; see Milton Friedman, “The Case for Flexible Exchange Rates,” Essays in Positive Economics (Chicago, 1953), which discusses the short-run balance of payments adjustment process with special emphasis on short-term capital movements.
In his study, “An Experiment with a Flexible Exchange Rate System: The Case of Peru, 1950–54,” Staff Papers, Vol. V (1956–57), pp. 449–76, S. C. Tsiang had to rely on annual data and on a period of only four years.
For a similar treatment, see S. C. Tsiang, “The Theory of Forward Exchange and Effects of Government Intervention on the Forward Exchange Market,” Staff Papers, Vol. VII (1959–60), pp. 75–106.
While the Canadian authorities operated in the forward market on a fairly large scale through 1951, their intervention since 1952 has been negligible.
See Studies in Econometric Method (Cowles Commission for Research in Economics, Monograph No. 14, edited by Wm. C. Hood and Tjalling C. Koop-mans, New York, 1953).
Computations for the 28-quarter period 1951–57 yielded on the whole less good results, presumably because of certain special circumstances which characterized the first year of operation of the fluctuating exchange system, including the continuance of exchange control until December 1951. Experiments were also made with several ways of handling the problem of seasonal variation. Neither the use of seasonally adjusted values for imports, exports, consumption, income, etc., nor the use of special seasonal dummy variables gave satisfactory results for the model as a whole. The model presented here is seasonally adjusted in two respects: First, the change in farm inventories, which shows a very marked seasonal pattern, has been subtracted from the domestic investment and disposable income variables appearing in the import demand equation and in the consumption function. Second, exports and the variable which stands for foreign income in the export demand function have been seasonally adjusted. The appropriate additions have been made in the list of predetermined variables. If the number of observations had been larger, a more satisfactory and more consistent solution of the seasonal problem might have been found. The small number of observations also accounts for the absence of lagged variables. Lagged income variables were originally included in the import demand, export demand, and consumption functions. But since their inclusion did not improve the model, they are omitted in the version presented here. Combinations of such variables as price ratios have been treated as one variable in the model reported here. Computations were made with linear approximations for the products or ratios of combined variables without the results being materially affected.
The change in reserves is taken to be a predetermined variable since its correlation with exchange rate changes on a quarterly basis was not significant.
The exchange rate variable used in the model is the quarterly average of daily noon rates.
Since there is only one jointly dependent variable in this equation, the least-squares estimate and the limited-information estimate are identical.
In view of the crudity of the equation, it should be noted that it “predicts” the net long-term capital balance during the six quarters following the period on which the model is based with an average error of only 8 per cent, and without any help from the constant term, which is negative while the long-term capital balance is positive throughout.
This relation ignores any magnification of domestic demand through an export-induced increase in domestic investment.
According to the coefficients presented in Appendix I, equations 2, 6, and 7, an increase in investment, ΔI, will lead to an increase in demand for domestic output of 0.4ΔI, while ΔY will be 0.8ΔI, and ΔYd, the change in disposable personal income, will be 0.64ΔI. Thus
Similar conclusions were reached by S. C. Tsiang in “Fluctuating Exchange Rates in Countries with Relatively Stable Economies: Some European Experiences After World War I,” Staff Papers, Vol. VII (1959–60), pp. 244–73.
where γ’n is the exchange rate expected to rule after n periods. In the text, we shall continue to make the simplifying assumption that such investments are planned for one period, although they may be renewed at the end of each period.
The relation is
In the case here being considered, the exchange rate is close to unity and the expected changes in the exchange rate are ordinarily small. Therefore, the right-hand expression yields the approximation in the text.
The yield wc to a Canadian investor in U.S. Treasury bills over and above the yield ic which he could earn on Canadian Treasury bills is
which, by the reasoning in the preceding footnote, gives the approximation in the text.
This formulation includes the special cases where E, the elasticity of expectations, equals zero, St = f(− Δγt − ht), and where E equals approximately unity, St = f(− ht).
If the short-term credit market were not imperfect, a tendency toward such a shortage of funds for arbitrage purposes, say in Canada, would raise the Canadian short-term interest rate until the return on arbitrage operations vanished. Persistence of such a return is, therefore, evidence of credit rationing on the part of the banks. In fact, Canadian banks are reported to be reluctant to extend credit for pure exchange transactions, even to the extent of requiring a cash margin or deposit to cover their risk of nonfulfillment of the contract (Sidney A. Shepherd, Foreign Exchange in Canada: An Outline, Toronto, 1953, p. 59). Negative values of e, indicating a short-term investment opportunity for owners of Canadian funds, can thus approach values at which the use of funds for interest arbitrage begins to compete with higher-yielding assets. For instance, in December 1952 and January 1953, Canadian funds converted into U.S. dollars and invested in U.S. Treasury bills would have yielded 3.7 per cent per annum, while the Canadian Treasury bill rate was 1.35 per cent; and in December 1956, this yield was 5.6 per cent, with the Canadian bill rate at 3.67 per cent at that time. On the other hand, on the U.S. side of the market, no such limitation on available arbitrage funds seems to have existed after 1951. The largest yields over and above the U.S. Treasury bill rate obtainable for arbitrageurs with U.S. funds were % per cent. This explains why from 1952 to 1957 the value of e, while sometimes larger and sometimes smaller, was almost always negative. Another factor which helps to explain this phenomenon is the Canadian trade deficit with the United States. In the course of normal operations of Canadian exporters and importers to cover exchange risks, the demand for U.S. dollars on the part of the importers will tend to exceed the exporters’ supply of forward U.S. dollars, and the forward price of the Canadian dollar will tend to be weak, i.e., the forward U.S. dollar will be at a premium. Similarly, forward sterling was almost always at a discount, because of the Canadian trade surplus with the United Kingdom, and also because of the exchange control restrictions on arbitrage operations in the sterling area.
To the extent that covered interest arbitrage does not fully close the gap measured by e, the size of this gap may be interpreted as a rough indicator of the market’s expectations concerning the future of the exchange rate. Large negative values of e imply that the Canadian dollar is expected to depreciate. Unfortunately, e is probably not a good indicator of expected appreciation, since U.S. interest arbitrage does not allow large positive values of e to develop.