With the advent of European Monetary Union (EMU), nominal exchange rates will cease to exist in the EU and the policy relevance of real exchange rates will increase. Against this background, a conference was held at the Oesterreichische Nationalbank in Vienna on 3/4 April 1998 for the purpose of bringing together academics and central bank policy-makers to discuss the policy implications of recent theoretical and empirical research on structural explanations of real exchange rates. The conference was organized by Eduard Hochreiter (Oesterreichische Nationalbank) and Axel Weber (Universität Bonn and CEPR).

Much work on purchasing power parity (PPP) has been unable to produce strong rejections of unit roots in US dollar real exchange rates for industrial countries in the post-1973 period. In 'The Great Appreciation, the Great Depreciation, and the Purchasing Power Parity Hypothesis', David Papell (University of Houston) tested the hypothesis that these non-rejections can be explained by the large appreciation and depreciation of the dollar in the 1980s. Treating the large rise and fall in the dollar as exogenous events which lie outside the data-generating process, Papell developed unit-root tests that account for these events while imposing the subsequent return to long-run PPP. Using panel methods, the unit root is rejected for those countries which adhere to the typical pattern of the dollar’s rise and fall. Moreover, the half life of deviation from PPP drops from almost three years to half a year.

Fabio Canova (Universitat Pompeu Fabra, Barcelona, and CEPR) criticized the exogeneity assumption, arguing instead for an economic explanation of the reasons behind such a persistent deviation from PPP. Mark Salmon (City University Business School, London, and CEPR) suggested that, instead of assuming three breaks in the slope of the real exchange rate, allowance should be made for different distributions in time in seeking to explain why the market switches from one regime to the other. Jacques Mélitz (CREST-INSEE, Paris, and CEPR) proposed that other relative prices – such as oil or commodity prices – be used to explain the real exchange rate because economic theory does not necessarily suggest mean-reversion of the real exchange rate, i.e. a constant equilibrium real exchange rate.

Asking 'What Do we Really Know About Real Exchange Rates?', Ronald MacDonald (University of Strathclyde) provided an overview of the recent literature. Widely researched questions include the extent of mean-reversion, and the importance of real, relative to nominal, shocks in the determination of real exchange rates. Because the recent period of floating rates has been too short to produce strong results, more powerful tests have been sought by using either panel methods or long-run data extending back to the 1870s. Though evidence for mean-reversion is found, the half-life adjustment back to equilibrium is too slow to be consistent with traditional PPP. Transport costs, generating a band within which it proves unprofitable to arbitrage away deviations from PPP, provide an explanation for the slow adjustment. Although sticky-price models suggest that the nominal rate drives real exchange rates, models focusing on the relative price of traded and non-traded goods imply the reverse, namely that volatility in real rates explains the behaviour of the nominal rate. Real exchange rates contain business cycle-related components, although the failure to capture important supply-side effects may be more a reflection of the methods used than of their lack of importance. A number of studies have shown that the systematic component of real exchange rates is related to productivity differentials, fiscal balances, terms of trade effects and net foreign-asset positions.

Anne Sibert (Birkbeck College, London, and CEPR) doubted the usefulness of some of the questions posed in real exchange rate research. Looking for mean-reversion with univariate methods implies the assumption that real shocks are temporary. Causality between nominal and real exchange rate movements is meaningless if both variables are endogenous. The use of long-run data implies the assumption that the stochastic nature of the process has not changed – an assumption that is obviously violated with regard to transport costs, trade restrictions, exchange rate regimes, capital mobility or the predominance of monetary over real shocks.

Klaus Liebscher (president of the Oesterreichische Nationalbank) explained the Austrian perspective on EMU and on the accountability of the European Central Bank (ECB). The Maastricht Treaty provided a variety of measures for transparency and accountability of ECB policy. Reporting to parliament would be useful if it related to the ex-post explanation of central bank policy. Nevertheless, publication of the minutes and the voting behaviour of the ECB council members could lead to national pressures if European monetary policy were to conflict with national views. Ultimately, the ECB would be accountable to European citizens with respect to maintenance of the purchasing power of the currency. After the announcement of the conversion rates in May 1998, an intensified coordination process would be necessary during the interim period up to the start of monetary union in January 1999. Concerning the decision on a targeting procedure, Liebscher advocated a broad approach based on more than one target variable, since uncertainty would be high at the beginning of EMU.

Nelson C Mark (Ohio State University) presented 'Price Level Convergence Among United States Cities: Lessons for the European Central Bank', written with Steve G Checchetti (Ohio State University) and Robert Sonora (Ohio State University). The authors studied the dynamics of price indices for 19 major US cities over the period from 1918 to 1995. Using panel econometric methods, they found that relative price levels among cities were mean-reverting, but at a surprisingly slow rate. The estimated half-life of convergence was approximately nine years. The inclusion of non-traded goods prices in the price index did not seem to provide an explanation for the persistence in real exchange rates, as deviations from PPP for traded goods were equally persistent. Instead, arbitrage impediments created by transport costs seemed to be present since large deviations were shorter-lived than small ones. As the United States is more integrated than the EU, the estimate provided an upper bound on convergence rates for the participants in EMU.

Riccardo Rovelli (Università degli Studi di Bologna) asked why the nine-year half-life of disturbances to the real exchange rate within the United States was higher than that of the three to five years which is generally found for real exchange rates between different countries. He supposed it might be owing to the failure to account for real factors, such as non-tradables prices. Different policy implications could be drawn for the ECB. If persistent nominal rigidity was present, the ECB could use monetary policy for output stabilization. In contrast, if persistence was due to real rigidities, monetary policy could do nothing about it. Other participants questioned the accuracy of the price data and the use of the CPI in the paper.

Matthew Canzoneri (Georgetown University and CEPR) investigated 'Trends in European Productivity: Implications for Real Exchange Rates, Real Interest Rates and Inflation Differentials', written with Robert Cumby, Behzad Diba and Gwen Eudey (all Georgetown University). They argued that the rapid increases in relative prices of home goods in southern European countries and in Belgium were owing largely to productivity gains in their traded goods sectors, and fundamentally should not be a cause for concern. If these trends in prices and productivity continued, however, these countries should experience inflation and real interest rate differentials with Germany of the order of 2–2.5%, which was about ten times the size of differentials observed across regions in the United States. As the real interestrate differential was a consequence of different productivity trends, it would be independent of the chosen monetary arrangement. If the policy objective of the ECB was to hold German inflation on the pre-union level, the European inflation target would need to be set about 1% higher than the current German target. For countries about to join the union in a second round, incorporating an inflation and an exchange rate criterion for convergence could create a competitiveness problem.

Mark Salmon remarked that some of the implications hinged on the functional form of the production function used. Jacques Mélitz asserted that real interest-rate differentials did not imply inefficiencies in the allocation of credit but only different shoe-leather costs in different regions of a monetary union. Timo Tyrväinen (Bank of Finland) suggested excluding agriculture and the public sector from the analysis because prices in these sectors were not determined competitively and there was no clear concept of productivity.

Menzie David Chinn (University of California, Santa Cruz, and NBER) examined the evidence for a productivity-based explanation of real exchange rate behaviour among East Asian currencies in 'The Usual Suspects? Productivity and Demand Shocks and Asia-Pacific Real Interest Rates'. Since the Asia-Pacific economies had been the fastest growing in the last decades, they were the most likely to offer evidence for the Balassa-Samuelson effect, which predicts that richer countries with higher labour productivity in the tradables sector would have higher non-tradables prices and higher price levels than poorer countries. Yet panel regressions incorporating productivity variables, as well as other demand-side factors, were hardly encouraging. Relative per capita incomes, a proxy for preferences towards services, or government spending did not appear to be determinants of real exchange rates in the region.

Jacques Mélitz suggested that a systems approach to estimation would be more efficient. Chinn’s results showed that the model did not work well for countries in which the variables displayed different degrees of integration. This could indicate that some explanatory variable was missing. Enrique Alberola-Ila (Banco de España) suggested that the current account deficit of the United States could be relevant for the Asia-Pacific countries. Axel Weber mentioned that equity prices could exert an influence on the real exchange rate via asset markets.

Mark P Taylor (University College, Oxford, and CEPR) examined 'Non-linear Mean-Reversion in Real Exchange Rates: Towards a Solution to the Purchasing Power Parity Puzzles', written with David A Peel (Cardiff Business School). The theoretical literature on the effects of transaction costs of international arbitrage suggests that real exchange rate adjustment towards long-run equilibrium may be highly non-linear. A smooth transition autoregressive (STAR) model with a quadratic transition function implied rapid adjustment to a shock if the exchange rate was far from its equilibrium value, and slow adjustment if it was near equilibrium. Impulse response functions had to be simulated, as the reaction to a shock was path-dependent and depended on the current position of the exchange rate relative to its equilibrium value. The half-lives of real exchange rate shocks varied with the size of the shock and the initial conditions, implying higher persistence of smaller shocks. For shocks of 15% or more, half of the shock was reversed in a year and a half or less.

Axel Weber suspected that it was not transaction costs, but policy actions – such as realignments – that accounted for most of the non-linearities found in European real exchange rates. Moreover, the discriminatory power between autoregressive, threshold autoregressive and STAR models may be low when only a few observations lie outside the bands. With current account deficits also asymmetric, non-linear adjustment seemed possible. Nelson C Mark doubted that non-linearities could solve the PPP puzzle as the dollar episode did not seem to fit into the model. Though the deviation from PPP was large, no mean-reversion was apparent for several years. Menzie David Chinn suggested specifying where the non-linearity arises in the model.

Holger C Wolf (New York University and NBER) assessed the aggregate importance of arbitrage in a paper entitled 'Is Real Exchange Rate Mean Reversion Caused by Arbitrage?', written with José Campa (New York University and NBER). To determine whether mean-reversion is caused by arbitrage it is necessary to know the size and the impact of arbitrage activities on prices. It is therefore necessary to investigate price and quantity data. The authors were unable to find evidence for a link between arbitrage and mean-reversion. Most episodes of evident mean-reversion were connected with three events: the collapse of the Bretton-Woods system; the end of the dollar bubble in the wake of the Plaza agreement; and the ERM crisis of 1992. One explanation for these findings is that arbitrage is not of sufficient quantitative importance to trigger mean-reversion. The other interpretation is that arbitrage is considered important by financial markets and governments with the result that exchange rates revert owing to markets' expectations of an appropriate exchange rate, based on expected future trade flows. This interpretation is supported by the finding that mean-reversion is faster among countries located close to each other, and among countries with sizable bilateral trade.

Philippe Martin (Graduate Institute of International Studies, Geneva, and CEPR) suggested that the model should discriminate between movements to equilibrium through trade in goods, trade in assets, and central bank interventions. Though deviations from PPP are assumed to trigger trade flows – leading to a positive correlation between the two variables, as assumed in the paper – trade also tends to reverse the deviation from PPP, and the sign of the correlation between trade and the real exchange rate is therefore not clear a priori. Thus it would be more natural to test for bidirectional Granger causality. Since the possibility that arbitrage could lead to mean-reversion is already accounted for, the actual trade volume becomes an endogenous variable in this case and should not be used in the estimation. Instead, a truly exogenous variable measuring the ease of conducting trade, e.g. openness or distance, would be preferable. Rainer König (Deutsche Bundesbank) considered the time horizon for the empirical tests to be too short for trade to show effects. In most cases, moreover, mean-reversions were brought about through a changing risk assessment in the asset markets.

Mathijs A van Dijk (Maastricht University) presented 'The Re-emergence of PPP in the 1990s', written with Kees G Koedijk and Peter C Schotman (both Maastricht University). They investigated PPP in a panel with 17 countries for the period from 1972 to 1996. The novel feature of their panel methodology was that the results were invariant with respect to the choice of a numeraire currency. Individual country effects in the relation between prices and exchange rates were allowed, which permitted identification of the currency pairs for which PPP held or did not hold. Evidence in favour of PPP was strongest for exchange rates relative to the Deutsche mark, and weakest for the Japanese yen. For the latter currency, a trend-like variable – such as productivity growth – was missing.

László Halpern (Hungarian Academy of Sciences, Budapest, and CEPR) questioned the restrictions imposed on the covariance matrix. Mark Taylor remarked that the model had no static equilibrium solution and therefore the tests maintained a hypothesis different from that in most of the literature. Moreover, the model did not specify the dynamics of the adjustment to PPP. Panel tests were more efficient because they imposed restrictions on the coefficients estimated. Taylor criticized the fact that the most restrictive assumptions in the paper were not tested.