Target Zones
A New Generation

Recent research into international monetary economics has investigated the agreements whereby national monetary authorities' attempt to keep their exchange rates within currency bands or `target zones' in particular the Exchange Rate Mechanism of the European Monetary System. This research has used the analytical techniques of `stochastic process switching' taken from modern finance theory. The `first generation' of these models based on the target zone model developed by Paul Krugman presented quite simple economic ideas and focused mainly on the technical novelties. Empirical support for the simple Krugman model proved poor, however, and this stimulated the development of `second generation' models to improve the fit of the data by incorporating factors such as devaluation risk, monetary authorities' intervention within currency bands, and learning processes. Many of the key participants in this field discussed both theoretical and empirical work with other researchers at a workshop on `Exchange Rate Target Zones', held in London on 27/28 October and organized by Lars Svensson and Charles Wyplosz, Research Fellows in CEPR's International Macroeconomics programme. Financial support was provided by the Commission of the European Communities under its SPES programme.
Giuseppe Bertola (Princeton University and CEPR) opened the workshop with a survey paper on `Continuous-Time Models of Exchange Rates and Interventions'. Researchers seeking to match the theory with the observed facts have shifted their attention to study formal exchange rate arrangements such as the ERM and the unilateral pegging regimes of the Nordic countries. Their analyses have incorporated the authorities' rules for intervention within the band, realignments whose probabilities may be assumed either infinitesimal or finite, and the possible influence of variables other than the identified fundamentals on the exchange rate. Introducing devaluation risks seems to have improved the model's fit to the data significantly, while allowing for price stickiness has improved its realism; but relaxing the assumption of risk neutrality appears to have added no major new insights. Bertola concluded that the second generation target zone models approximate reality much more closely than the first, and may be used for normative purposes; but sound economic explanations for the existence of target zones have yet to appear.
Andrew Rose (University of California at Berkeley) then presented `An Empirical Evaluation of Exchange Rate Target-Zones', written with Robert Flood and Donald Mathieson. This was one of the first attempts to test target zone models empirically. They applied various methodologies to data from the six long-term ERM member currencies to test numerous implications of the simple target zone model, but they found very little empirical support. Using the monetary model and assuming uncovered interest parity (i.e. that the risk premium is in general small enough to be neglected), they analysed daily data graphically to detect the presumed non-linearities between the exchange rate and the fundamentals driving it. They found few such non-linearities, however, and even these did not follow the predicted pattern. Countries with greater credibility appeared to exhibit more linear relationships, and the non-linearities also diminished over time; and the observations appear clustered in the middle of the band, rather than at the edges as the theory suggests. Rose concluded that these major empirical shortcomings require a more sophisticated description of the fundamentals driving the exchange rate.
Lars Svensson (Institute for International Economic Studies, Stockholm, and CEPR) then presented `Expected and Predicted Realignments: The FF/DM Exchange Rate During the EMS' (available as CEPR Discussion Paper No. 552), written with Andrew Rose, which focused on expectations of future realignments. They estimated the financial markets' expected rate of devaluation in the central parity for the FF/DM exchange rate, adjusting observed interest rate differentials by the expected rate of exchange rate depreciation within the band. The estimated expected change in the parity appeared to be lower in the later years of the EMS, although significant variability in the devaluation expectations remained. Svensson also noted that the estimated devaluation risk predicted movements in the central parity fairly accurately, but it was less successful in forecasting jumps in the actual exchange rate.
Hans Lindberg (Sveriges Riksbank) presented a paper, `Target Zone Models and Intervention Policy: The Swedish Case', written with Paul Söderlind. Their second generation model incorporates mean-reverting intramarginal interventions (from a paper by Delgado and Dumas) and devaluation expectations (from a paper by Bertola and Svensson). Lindberg reported that with these modifications the model fitted the Swedish data quite well and that the empirical problems reported by Flood, Rose and Mathieson and others seemed to vanish. The estimated parameters had reasonable values; and the degree of mean reversion in the interventions is quite strong. The model's parameter values imply that the exchange rate function is close to linear and the exchange rate behaviour is close to that of a managed float, which explains why non-linearities have been difficult to find empirically.
Presenting their paper, `Learning About a Shift in Exchange Rate Regime', John Driffill (Queen Mary and Westfield College, London, and CEPR) and Marcus Miller (University of Warwick and CEPR) considered how the public learns about the probability of devaluation when the government wishes to realign `often' and `rarely', by updating initial estimates for the two cases over time. If no realignment is observed, the public gradually reduces its expectations of its occurrence in the future; if a realignment does take place, there will be a `jump' in the estimated devaluation risk. This type of learning process may account for the occurrence of prolonged output losses following a regime shift: the gradual adjustment of private agents' expectations when an inflationary country pegs against a hard currency may produce inertia in inflation and hence shift demand in favour of foreign goods, thus protracting the slump at home. In conclusion, Driffill and Miller suggested extending their work by introducing strategic interactions between the government's signals and the public's learning process to explore the government's optimal strategy.
Axel Weber (Universität Gesamthochschule Siegen and CEPR) presented his study of `Time-Varying Devaluation Risk, Interest Rate Differentials and Exchange Rates in Target Zones: Empirical Evidence from the EMS'. Here the devaluation risk can vary over time, and market participants update their estimates of the realignment probability on the basis of interest rate differentials and exchange rate data. He found that the expected devaluations within the EMS have decreased over time, but they have not yet disappeared completely. Applying the same procedure to speculative attacks, he found that the market did seem to anticipate the realignments in the early period of the ERM.
Bernard Dumas (University of Pennsylvania) and Lars Svensson (Institute for International Economic Studies, Stockholm, and CEPR) presented a paper, `How Long Do Unilateral Target Zones Last?', which focused on the design of an exchange rate band for an individual country. In their model, the present regime's expected time to collapse depends directly on the quantity of reserves and inversely on the real and monetary divergence between the country and the rest of the world. Dumas and Svensson found that the width of the band played a role in determining the exchange rate's variability, but its effect on its durability was surprisingly limited; so a government narrowing a unilateral band need not worry much about its sustainability.
Alan Sutherland (University of York and CEPR) presented his paper on `Target Zone Models with Price Inertia: Some Testable Implications'. He first noted that all tests of first generation target zone models are joint tests of the theory and the model in which the target zone is embedded. Since the monetary model never fitted the data well, any target zone model that assumes flexible prices may therefore be expected to fall short of explaining the facts. Extending an otherwise standard target zone model to incorporate sticky prices, he found that this yielded standard results for nominal variables but that different implications emerged for the real variables. Sutherland concluded that price stickiness does not account for the disappointing empirical performance of first generation models and that the introduction of realignments remains the more promising approach.
Marcus Miller and Bernard Dumas concluded the workshop with suggestions for further research. They noted how the first generation models' poor empirical performance had motivated and stimulated the formulation of second generation models that already provide more complete explanations of the observed facts. Miller identified the introduction of heterogeneous beliefs among agents as a critical and challenging issue for future research. Dumas argued for the explicit introduction of optimization: in particular the commonly reported existence of implicit narrower target zones inside the official bands supports the case for investigating the optimal width of the band. Many participants agreed in conclusion that the case for operating target zones rather than a fixed exchange rate regime or a managed float still remains to be established.