The target zone proposal was first put forward by John Williamson in 1985 and subsequently extended by him with Marcus Miller. It is a `blueprint' for the coordination of monetary, fiscal and exchange rate policies among the G7 countries. In its extended form, the target zone proposal envisages rules for the conduct of monetary and fiscal policy to stabilize real exchange rates and nominal demand growth. These rules take the following form. First, countries determine a set of real, `fundamental equilibrium exchange rates', chosen so as to ensure medium- to longer-run current account equilibrium. They then set targets for the growth of nominal demand, taking into account internal and external policy goals. Interest rate differentials are varied in order to limit the deviation of currencies from their target levels, aiming to keep exchange rates within a given band around the target, while the average level of interest rates across countries is fixed in order to stabilize aggregate growth of nominal demand around the sum of national targets for nominal demand growth. In this context, national fiscal policy is varied with a view to achieving national targets for nominal demand growth.
On 13 May, CEPR held a workshop to discuss exchange rate target zones, organized by Marcus Miller, Co-Director of the Centre's International Macroeconomics programme. The meeting was part of a research programme on `Macroeconomic Interactions and Policy Design in an Interdependent World', with financial support provided by the Ford Foundation and the Alfred P Sloan Foundation.
George Alogoskoufis (Birkbeck College, London, and CEPR) began the meeting with his paper `On Optimal World Stabilization and the Target Zones Proposal'. In the context of a theoretical model, Alogoskoufis compared the performance of optimal policies for world stabilization with a regime of target zones. The model assumed a similar economic structure across different nations and allowed countries to produce both traded and non-traded goods. Nominal wages were imperfectly indexed to the price level, so that, in the absence of an appropriate stabilization policy, unanticipated disturbances may impose welfare losses on the economy. In addition to a standard measure of welfare loss, a Harberger triangle for the labour market, Alogoskoufis also used a `Bailey money market triangle', which measures the welfare loss induced by deviations of interest rates from their long-run equilibrium due to unanticipated shocks.
When all economies are free to use both monetary and fiscal policy, the analysis suggested that target zones are the optimal arrangement relatively fixed exchange rates and independent fiscal policies ensure the first-best optimum. If countries are unable to use fiscal policy, however, then a `second-best' outcome resulted, in which overall welfare is less than the `first-best' outcome (although the loss is minimized). Thus, in a second-best world, Alogoskoufis argued, the additional exchange rate constraint imposed by the target zone proposal might hinder rather than promote world stabilization. He also showed, however, that if the only economy constrained in its use of fiscal policy is the largest one the `Stackelberg leader' then the use of target zones might in fact reproduce the optimal world monetary arrangement quite closely.
David Begg (Birkbeck College, London, and CEPR) pointed out that Alogoskoufis's model was couched entirely in terms of percentage deviations of variables from their long-term equilibrium levels. Thus, in taking the long-run position for granted, Begg argued, the paper was essentially about exchange rate volatility rather than misalignments, and it was misalignment i.e. persistent deviations from long-run equilibrium that the target zone proposal was intended to correct. Begg also wondered whether a dir ect measure of the welfare loss due to inflation should have been used in place of the Bailey money market welfare loss measure: since many components of the money supply are now interest-bearing, interest rates can no longer be treated as the opportunity cost of money holding. Alogoskoufis respond ed that his paper was really concerned with cases where the target zone proposal was in operation, so that it was unnecessary to model sustained deviations from equilibrium. Furthermore, he claimed, all his conclusions were robust to adding inflation into the objective function, and one could always narrowly define money to exclude interest-bearing deposits. Simon Wren-Lewis (National Institute of Economic and Social Research and CEPR) argued, however, that a fully specified model ought to be able to accommodate persistent deviations from equilibrium, or `fads', rather than ironing them out from the beginning. Moreover, Wren-Lewis added, the model was deficient in not setting up any policy conflicts between the target zone proposal and domestic monetary issues.
John Driffill (University of Southampton and CEPR) suggested that the analysis was not sufficiently forward-looking and gave too little attention to the role of capital flows. In essence, Driffill argued, the model was Mundellian in not allowing flows to have stock effects. David Begg pointed out that although the model was essentially Keynesian, it did not fully allow for market imperfections such as adjustment problems and core inflation. Thorvaldur Gylfason (Institute for International Economic Studies, Stockholm, and CEPR) pointed out that the informational requirements in Alogoskoufis's model were very high. Although it may be reasonable to assume that each country observes its own supply shocks, a competent `World Central Banker' (the Stackelberg leader) would have to observe all aggregate supply shocks as well as world monetary shocks. Thus it might be fruitful to analyse a `third-best' outcome, where informational requirements were less stringent.
The second paper, presented by David Currie (Queen Mary College, London, and CEPR) and Simon Wren-Lewis, was entitled `A Comparison of Alternative Regimes for International Macropolicy Coordination'. The paper reported the results of simulation exercises for the period 1975-84 for the United States, Japan and Germany, using GEM, the National Institute's Global Econometric Model. One aim of their analysis was to provide guidelines for a number of quantitative issues concerning the extended target zone proposal: for example, how fast should interest rates adjust to exchange rate dis equilibrium? should we rely on fiscal or monetary policy to control world demand?
The study derived feedback rules and substituted them into GEM in place of existing equations explaining government expenditure and interest rates in the G3. These feedback rules stipulated that changes in each government's expenditure should depend on the level of and change in the gap between actual and target growth of nominal GNP in that country. Their optimal parameters were chosen by simulating the model. The results confirmed the importance of real exchange rates as an intermediate target. Capacity utilization should play a major role in setting nominal GNP targets, with fiscal policy used fairly actively to move nominal GNP growth towards its target. The results also suggested that the influence of global GNP dis equilibrium on changes in the target level of world interest rates should be fairly small, mainly because in GEM the size and timing of interest rate effects vary widely across countries. Thus, an attempt to reduce global aggregate demand by raising the general level of interest rates would affect some countries more than others.

The implementation of `optimal' feedback rules would, Currie and Wren-Lewis suggested, have led to a significantly better output and inflation performance in all three countries over the 1975-84 period, if the GEM model is accepted. In particular, the G3 countries might have avoided the 1980-1 recession by preventing the large increase in US interest rates which occurred around 1980. These results, and others reported in the paper, were model-specific, however, and should be tested using a different model of the G3. In an earlier paper, Currie and Wren-Lewis had conducted a sensitivity analysis of changes in the results produced by varying the objective functions, and had found that they were reasonably robust. Also, the authors had optimized the parameters of the feedback rules ex post, using data up to 1984, and these were then used to evaluate policy over the 1984-6 period; the results were qualitatively unaffected.

Currie and Wren-Lewis also considered an alternative to the target zones blueprint, sometimes referred to as the `IMF view', which reverses the Miller Williamson assignment by using fiscal policy primarily to achieve external current account equilibrium, while monetary policy stabilizes nominal demand growth. The authors noted that such a policy should not really be termed the `IMF view', since the IMF is typically also concerned with the balance between savings and investment in an economy. In any event, the target zone blueprint appeared to yield much greater welfare gains than this alternative scheme. The superior performance of the target zone proposal depended in part, however, on the presence of real exchange rate disequilibrium in the objective function; this is controversial, even though recent research suggests that the hysteresis effects of exchange rate disequilibrium can be substantial.
Warwick Hood (HM Treasury) suggested that a more complete analysis might allow for differences in objective functions among the G3. Wren-Lewis agreed with the spirit of this suggestion, but pointed out that actual German macro policy for much of this period was a long way from the optimal solution, so that the optimal German weights would be rather incredible. Peter Kenen (Princeton University and CEPR) asked whether this analysis constituted a fair test of the target zone proposal versus the `IMF view'. He pointed out that the first blueprint is geared towards real exchange rate targets, while the second is geared towards current account targets. Kenen also wondered whether it might be worth examining real tax cuts as the instrument of fiscal policy, instead of or in addition to public expenditure. On the first of these issues, Currie argued that the exchange rate and current account targets examined in their analysis were in fact consistent. On the second issue, he noted that government spending and tax policies have essentially the same effects in GEM. Gerry Holtham (Credit Suisse First Boston) questioned how the government budget constraint was incorporated into GEM, how the hysteresis effects of different policies were allowed for, and how the model was closed in terms of incorporating the rest of the world, besides the G3. George Alogoskoufis pointed out that, if fiscal policy were not in fact available, then the target zone proposal and the `IMF' blueprint would be equivalent.
The third paper, entitled `Target Zones, Currency Options and the Dollar', was presented by Marcus Miller (University of Warwick and CEPR) and Paul Weller (University of Warwick). The first section of the paper analysed a (deterministic) Dornbusch exchange rate overshooting model with the imposition of target zones. The commitment to keep the nominal exchange rate within a target zone implies that monetary policy must be adjusted when the fundamentals stray too far from equilibrium. When this happens, the money supply and exchange rate must essentially swap roles, the latter being kept fixed at the edge of the band and the former being adjusted so as to bring the interest rate differential back to zero.
In the second part of the paper, Miller and Weller analysed the target zone proposal in the context of a stochastic model of overshooting in addition to the systematic effect of economic activity, random shocks are allowed to affect the price level. Using a different model, Krugman has shown that if fundamentals follow a pure random walk in discrete time, there is an explicit analytical solution which exhibits `bias in the band' i.e. the exchange rate is in general closer to the middle of the band than it would be under a free float. Miller and Weller are able to extend this result to a continuous-time, stochastic version of the Dornbusch model, but with two qualifications. First, when the fundamentals show a systematic tendency to return to equilibrium, then the explicit analytical solution is no longer applicable and numerical methods must be used. Second, it appears that, in order to retain certain characteristics of Krugman's solution, the exact arbitrage (uncovered interest parity) condition must be replaced by an approximation.
Finally, Miller and Weller explored a link with the finance literature to show that the effect of introducing target zones is formally equivalent to that of a large-scale issue of compound currency options. In particular, since the top of the band represents a commitment by the authorities to sell the domestic currency in unlimited amounts against other major currencies, it essentially constitutes a set of US `call' options on the domestic currency, with an exercise or strike price equal to the level of the top of the band. Conversely, the bottom of the band should be equivalent to a set of US `put' options on the domestic currency written by the authorities, with a strike price at this level. Since these options would be worthless so long as the exchange rate remained within the band, the authorities could, in principle, greatly enhance the credibility of their exchange rate policy by a free large-scale issue (i.e. `helicopter drops') of such options.
David Currie particularly welcomed the final section of the paper, which linked macroeconomic analysis with financial economics, with great potential for future research. Currie also pointed out that in the kind of models examined in this paper and by Krugman, random walk exchange rate fundamentals essentially cointegrate with a random walk price level, so that the real exchange rate is mean-reverting. Analysis would be a lot more difficult if, as seems the case in reality, the real exchange rate were nonstationary. On the use of currency options, Gerry Holtham noted that if the market view differed from the government view, then the authorities could continually make profits. He also suggested that many important problems were absent from Miller and Weller's analysis. In particular, fundamentals may follow non-stationary processes. Thus the model might be improved by introducing strongly auto regressive behaviour in the system. Referring to the paper by Currie and Wren-Lewis, Holtham pointed out that the simulations were carried out on a model which may not have high status with policy-makers. On a general level, he argued that academics often conduct empirical work using models which do not incorporate intertemporal budget constraints. Policy-makers may be more concerned with public debt and other medium-term issues than they are with fine-tuning. Thus the Mundell-Fleming-type models used in each of the three workshop papers may be of little interest.

Peter Kenen began his general remarks on the workshop by asking what was the motivation for analyses similar to Krugman's, such as that of Miller and Weller. In essence, Kenen argued, the `intra-band dynamics' of such analyses are uninteresting, because the credibility of the arrangements is never itself tested. He did, however, find merit in the `equivalence theorem' linking the target zone proposal and the issue of compound currency options, but he wondered what implicit assumptions were made with respect to expectations. In particular, there may be a case for further incorporating rational exchange rate bubbles or `fads' into such models, and perhaps making expectations non-homogeneous. In general, Kenen foresaw two fruitful lines of inquiry in this area the first involving further work on policy rules, the second involving further work on how participants in foreign exchange markets actually behave.
Nigel Jenkinson (Bank of England) began his remarks by pointing out that all the workshop papers had examined `hard' rather than `soft' target zones, and suggested that it might be worthwhile examining what differences would arise if the `soft' options were pursued. While accepting that the Currie and Wren-Lewis analysis was encouraging, Jenkinson called for further work to allow international differences in objective functions and to build more stock-flow consistency into the simulation model. More generally, he argued that there was a problem in working out an interesting counterfactual to a situation where policy-makers firmly believe that they are assimilating all available, relevant information and pursuing flexible policies.