Exchange Rate Models
Where is the profession now?

The system of floating exchange rates which has emerged in the last decade has naturally focused attention on modelling the behaviour of the foreign exchange markets. Yet despite this renewed interest, many aspects of exchange rate behaviour remain unresolved. The Centre held a workshop on 30 June to assess the current state of exchange rate modelling and to suggest future directions in which research might usefully be extended.

David Currie (Queen Mary College, London, and CEPR) and Stephen Hall (National Institute of Economic and Social Research) presented the first paper of the workshop, on 'A Stock/Flow Model of the Determination of the UK Effective Exchange Rate'. Currie and Hall developed a model of the effective exchange rate based on a view of capital flows that combined both stock and flow theories. They also explored the nature of longer-run external balance of payments equilibrium, particularly the possible long- run trade-off between larger external deficits and higher domestic interest rates. This, they noted, was an important issue when considering the longer-run viability of policy measures aimed at domestic expansion.

Currie and Hall observed that the literature on modelling the sterling exchange rate presented a picture of 'general failure'. The best available model, from a forecasting point of view, appeared to be a simple random walk, in which the value of the exchange rate in the next period was equal to its value today plus a random disturbance which could not be forecast on the basis of information available today. This view is reflected in some of the large macro models of the United Kingdom, according to Currie and Hall. The authors themselves took a less pessimistic view, however, noting that the current NIESR model (Model 8) contains an estimated model of the exchange rate, as did its two predecessors, Models 6 and 7, which had been used in practical forecasting exercises for a number of years.

The literature on exchange rate determination has generally favoured a stock view, in which capital flows reflect the international supply and demand for financial assets and the adjustment of stocks of these assets in portfolios. Stock theories suggest that a change in domestic interest rates will produce an adjustment of international portfolios and a once-off capital inflow or outflow. An interest rate differential will not, on this view, produce a sustained capital inflow. Currie and Hall noted that, in a world where financial portfolios are growing, one would expect a change in relative yields to alter a country's share of world financial wealth, and thereby to create a sustained capital inflow for a given interest rate differential and exchange rate. This would tend to enhance the flow aspects and play down the stock aspects of exchange rate determination.

Currie and Hall attempted to discriminate between the stock and flow views by estimating a model in which the real effective exchange rate depended on the lagged exchange rate, the expected future exchange rate, interest rate differentials, and exports and imports. Their estimates, they concluded, favoured the flow view of exchange rate determination and suggested a moderate long-run trade-off between larger external deficits and higher domestic interest rates.

In the discussion that followed, Barry Eichengreen (Harvard University and CEPR) asked for clarification of the theoretical model which underlay the estimated equations. David Currie interpreted it as deriving from a net asset demand equation. Participants argued, however that if this was the case, then the estimated equation should include variables which measured the growth in domestic and world portfolios. Measures in even adequate proxies for such variables were difficult to construct, the authors replied.

Mike Wickens (University of Southampton and CEPR) presented the second paper of the workshop, written with Peter Smith (London Business School), entitled 'A Stylised Econometric Model of an Open Economy: UK 1973-1981'. Smith and Wickens's paper also addressed the pessimistic conclusions which had emerged from tests of empirical models of the determination of flexible exchange rates, namely that the underlying economic theories should be rejected and that none can outperform a random walk model. The failure to find satisfactory alternative economic theories, the authors noted, has lent credence to irrational theories, involving speculative bubbles. Despite the lack of supporting evidence, however, the appeal of the standard textbook theories remains.

Smith and Wickens argued that nearly all of the tests of exchange rate models made to date have been carried out on the reduced form rather than on the structural form of the exchange rate equations. These tests have generally failed to support the underlying economic theories, but Smith and Wickens argued that this may result from their failure to incorporate a reasonable dynamic specification.

In their paper, Smith and Wickens estimated a structural econometric model of a small open economy, based on an adaptation of the Buiter and Miller version of the Dornbusch 'overshooting' model. The principal characteristics of this model are sticky prices, portfolio balance and uncovered interest parity. Buiter and Miller's model allows there to be steady-state inflation by using an expectations augmented Phillips equation. Smith and Wickens modified this model to include the effects of certain exogenous variables such as oil prices production, and to permit endogenous core inflation.

In CEPR Discussion Paper No. 7, Smith and Wickens had estimated a stylized structural version of the monetary model for the United Kingdom in the period 1973-1981. In their workshop paper, using the same data set, they estimated the full structure of the rational expectations Dornbusch/Buiter-Miller model. They then used the estimated econometric model to provide an estimate of a simplified rational expectations version of the Buiter-Miller model which could easily be used to analyse the short-run behaviour of the exchange rate. The simulation results for the model confirmed that the exchange rate does overshoot following a change in the level of the money supply, and they indicated that this overshooting is of the order of 21%. Initially, the price level is unaffected, and thereafter it rises to its new equilibrium level fairly smoothly.

Smith and Wickens concluded that the standard textbook model of the determination of exchange rates for a small open economy is capable of giving a reasonably good empirical explanation, provided that the model is dynamically well specified and that its structural (and not its reduced form) is estimated.

William Branson (Princeton University and CEPR), in his presentation 'US Fiscal Policy and the Dollar', argued that while structural economic models explained medium-term exchange rate movements quite well, they were very bad at explaining short-run fluctuations. His model concentrated on the relationship between the US exchange rate, the long and short interest rates and the fiscal deficit. The risk premium in the interest-parity condition was related to the stock of government bonds outstanding, on the grounds that a larger stock of bonds increased the risk of devaluation, as foreign residents became less willing to hold US assets. Branson argued that numerical simulations with plausible parameter values reproduced the stylized facts of US exchange rate and interest rate behaviour.

The final session of the workshop was devoted to a general discussion of alternative approaches to exchange rate modelling. This discussion was based around a survey paper by Peter Isard (IMF), entitled 'Alternative Approaches to the Empirical Modelling of Exchange Rates: Where is the Profession Now?'; Gerry Holtham (OECD) reported on progress at the OECD in modelling exchange rates. Putting more effort into building models which assumed uncovered interest parity was generally seen as unpromising: more attention should be paid to providing a more sophisticated treatment of risk premia. Portfolio balance models tended to perform poorly for a variety of reasons, which included difficulties in measuring asset stocks accurately. A number of participants argued that by putting so much emphasis on rationality economic models were often incapable of capturing important real world phenomena. Speculative bubbles were a plausible explanation of many exchange rate fluctuations and more effort should be made to incorporate them into econometric modelling. There was also some discussion of the use of the current account balance to 'tie down' the long-run exchange rate equilibrium, although it was stressed that this often led to perverse short-run exchange rate movements.