Participants at a CEPR London workshop on 16 March focused on changes of policy regime, with a particular emphasis on the learning procedures adopted by private agents in updating their expectations immediately following a change. The workshop was organized by Marcus Miller and George Alogoskoufis, Co-Director and Associate Director of the Centre's International Macroeconomics programme. It formed part of CEPR's research programme on `Financial and Monetary Integration in Europe', supported by a grant from the Commission of the European Communities under its SPES programme.

In his paper, `Should Rules Be Simple?', Paul Levine (University of Leicester and CEPR) used an overlapping contracts version of the Barro-Gordon model to study how monetary authorities can precommit their aggregate demand policy to avoid the inefficiencies of discretionary policy. For a deterministic version, he derived the optimal rule with precommitment, the discretionary time-consistent rule and a simple rule constrained to depend only on the current values of the system. When discretionary policy is replaced with a new rule after a regime change, the authorities stand to gain credibility if the private sector learns their chosen rule quickly. Levine concluded that simple policy rules are more suitable than `optimal' rules when the private sector is badly informed about the government's preferences.

Martin Klein (Universität Bonn) noted that the authority takes no account of learning in its optimization process: he suggested that a game-theoretic approach would enrich the analysis and that the appearance of multiple equilibria would strengthen the case for simple rules.

In a paper on `Bounded Rationality and Learning', Mark Salmon (European University Institute, Firenze, and CEPR) focused on agents' processing of information, learning and revision of expectations. Most traditional models of learning processes assume that private agents know the true model of the economy, which if untrue may lead to divergence. More realistic learning schemes entail moving away from the rational expectations hypothesis, which has received little empirical support despite its theoretical attractiveness.

Salmon focused on `procedural' rationality, which requires only that actions are the outcome of deliberation: agents' learning methods require no `structural information', only knowledge of the exogenous and output variables. He used the `neural network', more commonly employed in biological studies, to compute expected monetary growth in a macro policy game. This method provides non- linear approximations to unknown functions, which may change over time, to any degree of accuracy required. He found this method robust, since it converged to the rational expectations equilibrium and permits the parallel processing of information.
In the discussion that followed, Andrew Hughes Hallett (University of Strathclyde and CEPR) agreed that an improved understanding of learning processes was required, but he warned that transplanting techniques from other sciences may be inappropriate: in particular, he argued that the neural network does not adequately capture variability.

Jacques Mélitz (Institut National de la Statistique et des Etudes Economiques, Paris, and CEPR) analysed an important recent regime switch in a paper on `German Reunification and Exchange Rate Policy in the EMS'. He focused on reunification's short- and long-run effects on the Deutschmark under various assumptions concerning EMU. In the long run, the relative price of German goods must fall to generate a trade surplus and repay the debt accumulated in the wake of reunification, so German economic, monetary and social union will entail a depreciation.

Mélitz noted that the short-run pressures on demand following reunification will tend to push the Deutschmark upwards; and he emphasized the need for EMS member countries to coordinate their short- and long-run behaviour. If regular realignments are to follow in order to accommodate the pressures from reunification, any tendency to resist them in the short run will simply amplify the appreciation of the Deutschmark vis-à-vis the dollar and put the other EMS currencies out of line with the dollar as well. On the other hand, if there are to be no such future realignments, the monetary authorities should coordinate their policies to convince the public of this: otherwise the tendency to follow the Deutschmark in the short run will only aggravate the upward pressure on the Deutschmark and the other EMS currencies.

Martin Klein agreed with Mélitz's short-run analysis, but he was surprised that the long-run values of the real exchange rate in the `old' EMS, characterized by frequent realignments, and the present EMS were the same. Marcus Miller (University of Warwick and CEPR) observed that the long-run value of the Deutschmark may be lower within a monetary union than outside while still remaining above its current level.

In a paper on `Monetary Accommodation, Exchange Rate Regimes and Inflation Persistence', George Alogoskoufis noted that forward- looking agents' expectations of higher prices will be embodied in contracts and induce higher inflation persistence through the wage-price spiral. He examined the view that inflation displays less persistence and the variance of inflationary shocks may be smaller under fixed exchange rates than in a system of managed floating.

First, the time-series data on the price level from 1880 for the UK and the US and for 21 OECD countries during the post-war period suggest that inflation was less persistent under fixed rates, although the regime appears to have had no significant effect on the variance of inflationary shocks. Second, for an open economy model of aggregate demand and supply with overlapping contracts, wage and price setters claim higher contracts if world monetary policy accommodates price shocks. Further, if the exchange rate is not fixed, the policy-maker can pursue an independent monetary policy devoted to output stabilization rather than price stability, which in turn will affect the public's expectations and hence contracts.

Alogoskoufis argued that his results did not necessarily imply a need to return to the Gold Standard. That had many well-known shortcomings, and the UK's low degree of monetary accommodation may have resulted from the Bank of England's policy of stabilizing reserves and not directly from the link between gold and money. Achieving low inflation persistence requires a strong commitment to non-accommodative monetary policy, however, which might be obtained instead by creating an independent central bank or through the G3 countries' agreement to a fixed exchange rate system and to constraints on the conduct of monetary policy.

In his paper, `Stochastic Process Switching Probabilities in Exchange Rate Target Zones: An Empirical Evaluation of the EMS', Axel Weber (Universität Gesamthochschule Siegen and CEPR) noted that the empirical testing of recent target zone models focuses on testing the non-linearity of the exchange rate trajectory assuming that the fundamentals are unobservable. Following this approach, Weber estimated the credibility of a target zone by evaluating the probability of stochastic process switching arising from intervention to stabilize the exchange rate or from realignments. His results indicated that while the `old' (pre-1983) EMS displayed a high probability, this has subsequently fallen, which suggests that the system's credibility has increased. In the discussion that followed, various participants questioned whether the linearizations used in his statistical fitting were consistent with the essentially non-linear nature of the models tested.