Why do monetary models of the exchange rate break down?
New evidence from a new approach.

There have been many attempts to find a structural model capable of explaining the behaviour of floating exchange rates. None of the models proposed has proved entirely satisfactory, and most have performed poorly on the data of the past decade.

Virtually all of these structural models have been based on the asset market theory of exchange rate determination. According to this theory, foreign exchange should be viewed as a financial asset with its price determined by the demand and supply for the stock of foreign exchange. A common further assumption is that asset markets are efficient and fully reflect all available information. In particular, both the current and the expected future price of foreign exchange would immediately respond to new information, but the actual future price would be affected by events currently unpredicted. This would cause the spot exchange rate and the expected future rate to be closely linked but the change in the spot rate to be highly volatile. These features of the asset market theory appear to be broadly in accordance with the facts and account for its attractiveness as an explanation of the behaviour of exchange rates.

To explain the exchange rate fully, however it is necessary to embed the foreign exchange market within a complete economic model. Unfortunately, the resulting model equations explaining the exchange rate usually fail tests of misspecification or forecast poorly. Thus whatever the theoretical merits of the asset market approach, when incorporated into a complete structural model it does not provide a satisfactory empirical description. Understanding why these equations perform badly is complicated, because they embody restrictions derived from other equations in the model, which are required to complete the model but which are not part of the asset market theory itself. Consequently it is difficult to know which part of the model has gone wrong.

In a recent CEPR Discussion Paper, Michael Wickens and Peter Smith apply the rational expectations monetary model of Mussa and Frenkel to the sterling-US dollar and Deutschemark-US dollar exchange rates. They find that the model does break down and attempt to assess the relative contribution of various factors to this model failure.

Wickens and Smith propose a novel methodology for measuring the relative importance of these factors. The technique has wide applicability elsewhere. They construct time series models of the various sources of error or "misspecification" in the model equations. The exchange rate equation can then be solved to include these additional misspecification terms. By estimating the extra contribution of these terms to the equation explaining the exchange rate, Wickens and Smith can derive a measure of how important is each source of misspecification in the failure of the exchange rate equation.

A common explanation of the failure of monetary exchange rate models is the breakdown of the assumption that price levels converted at market exchange rates are the same in all countries (purchasing power parity). Using their methodology Smith and Wickens find that the inadequacy of the purchasing power parity assumption is indeed important. Misspecification of the money market equations is equally important, however, even though these equations are not fundamental to the asset market explanation of exchange rates.

Smith and Wickens also assess the random walk model of the exchange rate. This model posits that the exchange rate today is its value yesterday, plus a shock or disturbance which cannot be forecast using information available yesterday. They find that although the random walk model does appear to fit the data well, past information can nevertheless be used to explain the spot exchange rate, so the random walk hypothesis can be rejected.

An Empirical Investigation into the Causes of the Failure of the Monetary Model of the Exchange Rate
P N Smith and M R Wickens

Discussion Paper no. 7, March 1984 (IM)