Abstract
In this chapter a structural portfolio model of exchange rate determination is estimated for the Netherlands. The specification of the equations and the estimation procedure used seek to circumvent some of the weak points of many other studies on exchange rate determination mentioned in Chapter 2. This structural model of exchange rate determination forms the monetary sector of a quarterly model of the Dutch economy. The other relations of this model and the simulation results of the entire model are presented in the next chapters.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
References
For a rationale of this assumption see Friedman (1977), p. 664, and the literature cited in his footnote 4.
A similar approach has been used in Backus et al. (1980) and Van Loo (1983).
A survey of estimation methods for portfolio models subject to adding-up restrictions can be found in Owen (1986), pp. 59–76.
A more extensive defence of this Bayesian approach can be found in Smith and Brainard (1976).
See e.g. Johnston (1972), pp. 238–241 for a brief description of SURE.
The LSQ-procedure in the TSP-package is used for simultaneously estimating the equations.
The essence of the instrumental variables approach can be found in textbooks such as Johnston (1972) and Theil (1971). An extensive study of this method is presented in Bowden and Turkington (1984).
See e.g. Jager(1981).
Other methods for reducing the number of instrumental variables are listed in Brundy and Jorgenson (1971) and Bowden and Turkington (1984).
The use of a different set of instruments per endogenous variable would have led to inconsistent estimates. See Mitchell and Fisher (1970).
Other methods for reducing the number of instrumental variables are listed in Brundy and Jorgenson (1971) and Bowden and Turkington (1984).
Driehuis (1972), pp. 9 and 10 lists some reasons in favour of the use of seasonally adjusted data.
The same argument against the use of seasonally adjusted times series can be found in Meese and Rogoff (1983a), p. 9.
See Den Dunnen (1985), p.9. Quotations on the seasonal influences of the Dutch monetary policy can be found in Den Dunnen (1985), pp. 58, 61, 76, 80 and in De Wilde (1975).
See Wallis (1974).
See the studies mentioned in Thomas and Wallis (1971).
Anema and Jepma (1978) and Rhomberg (1976) discuss the various characteristics of effective exchange rate indices.
Brodsky (1982) extensively discusses the differences between the geometric and arithmetic effective exchange rates. The advantages of the geometric effective exchange rate presented in the text are based on his analysis.
For the geometric average this property follows at once from the fact that [Ri2/RiQ]/[Ri1/Ri0] = Ri2/Ri1. The derivation of the relative change in the arithmetic indices can be found in Brodsky (1982), p. 552.
See Van Nieuwkerk (1981), p. 259.
No data on capital flows are considered because no reliable data on the currency decomposition of these flows are available.
According to Frankel and Froot (1985), the expected relative inflation is among the determinants of survey data of the expected exchange rate. Thus, although according to the literature mentioned in section 2.6.3, the purchasing power parity does not hold, agents use this parity for deriving their expectations. Since in this section we are modelling the parivate agents’ expectation formation, the relative prices are used as an explanatory variable.
Note that although Pauly and Peterson use the substitution method, they consider the rolling regression technique as a probably superior alternative. Their most important argument for not applying this procedure is that it introduces “the need for an extra loop in the maxi-LINK solution to update the expectational variable” (Pauly and Peterson, 1986, p. 155).
The principal components analysis of interest rates is presented in Dongelmans and Fase (1975). The studies of the Dutch financial sector are De Nederlandsche Bank (1984), Van Loo (1983), Van den Berg, Don and Sandee (1983).
Roley showed that, if the assets demand equations are derived from the frequently used mean-variance model, the symmetry restrictions imply that investors exhibit constant risk aversion with respect to the mean of the argument of the utility function (see Roley, 1983, for details).
More details on the gross credit ceilings can be found in Van Loo (1983), p.181.
The same reasoning can be found in Sterken (1986), p. 25.
An extensive description of the instruments of Dutch monetary policy can be found in Den Dunnen (1985). Van Loo (1983) presents the most detailed econometric study of the influence of monetary policy measures on the banks’ behaviour.
It should be noted that the change in the central bank’s monetary reserves is a rough measure of the bank’s intervention on the foreign exchange market. Therefore, the central bank’s holdings of foreign currency reserves have been corrected for some institutional aspects. See the list of symbols for details. This proxy has been used, because the Dutch central bank does not publish a time series of its interventions on the foreign exchange market. See Den Dunnen (1985), pp. 92 and 93 for details.
A description of this instrument can be found in De Nederlandsche Bank, Kwartaalberichten, 1973 (2), pp. 43–54.
See De Nederlandsche Bank, Kwartaalberichten, 1970 (4) and 1971 (1) for a brief description of these facts.
See De Nederlandsche Bank, Kwartaalberichten 1975 (2), pp. 19 and 31.
See De Nederlandsche Bank, Kwartaalberichten 1979 (1), p. 32.
This method assumes that a polynomial of fairly low degree can represent the lag coefficients. The degree of the polynomial and the number of lags have to be specified a priori. Given these values, the method defines new variables, whose coefficients correspond with the parameters of the polynomial. The number of these parameters is less than the number of lags. Thereafter these newly created variables are substituted in the original equation, which is estimated by OLS. The estimated parameters and t-values of the lagged variables’ coefficients are derived from the estimated value and covariance of the polynomial’s parameters. Two additional constraints can be imposed on the polynomial. The first one sets a hypothetical coefficient of the variable lead by one period at zero. The second constraint sets the hypothetical coefficient of the the variable lagged by one extra period than the number of lags at zero. These constraints are known as the near and far constraints respectively. Details with regard to the Almon scheme can be found in Almon (1965) and Johnston (1972), pp. 294–298. A description of the Almon procedure’s implementation in the TSP-package is presented in Hall and Hall (1980), pp. 25, 26. For estimating the coefficients in the equation of r ., the polynomial is assumed to be quadratic and far constraints are introduced.
Compare the discussion of the multicollinearity problem in Section 2.5.4. In fact this is an example of perfect correlation between a set of variables and another variable.
See Van Loo (1983), p 38.
Author information
Authors and Affiliations
Rights and permissions
Copyright information
© 1991 Springer-Verlag Berlin Heidelberg
About this chapter
Cite this chapter
de Jong, E. (1991). The Financial Sector of the Model. In: Exchange Rate Determination and Optimal Economic Policy Under Various Exchange Rate Regimes. Lecture Notes in Economics and Mathematical Systems, vol 359. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-51668-9_4
Download citation
DOI: https://doi.org/10.1007/978-3-642-51668-9_4
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-54021-2
Online ISBN: 978-3-642-51668-9
eBook Packages: Springer Book Archive