The failure of the monetary exchange rate model for the Canadian‐U.S. dollar

Abstract

In this paper the validity of the monetary exchange rate model in the long run for the Canadian-U.S. dollar exchange rate is examined. The primary test employed is the Johansen (1991) and Johansen and Juselius (1990) cointegration technique. The effects of dummy variables and lag specification on the statistical inference are considered, and Monte Carlo simulations based on the estimated parameters are employed. Despite the use of the longest data set yet for the Canadian case, no evidence is found in favour of the monetary exchange rate model using the Johansen procedures. This result is confirmed by several other cointegration procedures. JEL Classification: F31; F41 La faillite du modèle monétaire du taux de change pour les dollars canadien et américain. Ce mémoire examine le modèle monétaire du taux de change pour savoir s'il est validé par l'expérience canado-américaine en longue période. Le test principal utilisé est la technique de cointégration de Johansen (1991) et Johansen et Juselius (1990). Les effets des variables fictives et de la spécification de la structure de délais sur l'inférence statistique sont considérés, et on utilise des simulations de Monte Carlo fondées sur les paramètres calibrés. Malgré le fait qu'on fait appel à la plus longue série chronologique utilisée dans ce type d'analyse à ce jour, on ne trouve aucun support du modèle monétaire du taux de change en utilisant les procédures de Johansen. Ce résultat est confirmé par l'utilisation de plusieurs autres procédures de cointégration.

Testing the monetary model of exchange rate determination: a closer look at panels

“…13 Even for the six countries where the degress-of-freedom corrected trace statistic is significant at the 10 percent level, this is not sufficient evidence for the monetary model, since (as we emphasized above) the JOH-ML coefficient estimates reported in Table 2 are not in accord with the monetary model for these six countries. Note that our results for Canada are similar to those reported in Cushman (2000). Using the Johansen (1991) procedure, Cushman (2000) finds evidence of a cointegrating relationship among the nominal US dollar-Canadian dollar exchange rate and a set of monetary fundamentals over the post-Bretton Woods float, but the estimated cointegrating coefficients are not at all in line with the monetary model.…”

“…Even studies that find evidence of cointegration between nominal exchange rates and monetary fundamentals during the modern float lend little support to the monetary model, as the estimated cointegrating vectors themselves often fail to conform to the monetary model. For example, Cushman (2000) finds evidence of cointegration between the US dollar-Canadian dollar exchange rate and a set of monetary fundamentals over the modern float, but the estimated cointegrating coefficients differ widely from those predicted by the monetary model. Cushman (2000) therefore concludes that there is no support for the monetary model in USCanadian data.…”

“…For example, Cushman (2000) finds evidence of cointegration between the US dollar-Canadian dollar exchange rate and a set of monetary fundamentals over the modern float, but the estimated cointegrating coefficients differ widely from those predicted by the monetary model. Cushman (2000) therefore concludes that there is no support for the monetary model in USCanadian data. 1 It is not surprising that the monetary model performs poorly on a country-bycountry basis during the modern float, as long-run purchasing power parity (PPP)-a building block of the monetary model-also fares poorly on a countryby-country basis during this period.…”

Long-run PPP in a system context: No favorable evidence after all for the U.S., Germany, and Japan

“…13 The degree-of-freedom adjustment is a correction for small-sample size distortion in the direction of an excessive tendency to reject if the null is true. It has been suggested by Reimers (1992) and supported by simulations in Cheung and Lai (1993a), Gregory (1994), Cushman et al (1996), andCushman (2000). The simulations in the last two papers suggest that the correction is also useful for various chi square tests in the Johansen cointegration procedures, and so I apply it here to all subsequent tests.…”

“…TSP 4.5 is used. Alternatively, one could bootstrap the tests as in Cushman et al (1996) and Cushman (2000). 11 The source is the IMF's International Financial Statistics online, November, 2006.…”