Macroeconometric Models and European Monetary Union

Hall, Stephen G. (Ed.); Heilemann, Ullrich (Ed.); Pauly, Peter (Ed.) Book Macroeconometric Models and European Monetary Union RWI Schriften, No. 73 Provided in Cooperation with: RWI – Leibniz-Institut für Wirtschaftsforschung, Essen Suggested Citation: Hall, Stephen G. (Ed.); Heilemann, Ullrich (Ed.); Pauly, Peter (Ed.) (2004) : Macroeconometric Models and European Monetary Union, RWI Schriften, No. 73, ISBN 978-3-428-51398-7, Duncker & Humblot, Berlin, https://doi.org/10.3790/978-3-428-51398-7 This Version is available at: https://hdl.handle.net/10419/150030 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-nd/4.0/ Stephen G. Hall, Ullrich Heilemann, and Peter Pauly (Eds.) Heft 73 RW I ESSEN asdfghjk Duncker & Humblot ´ Berlin RWI : Schriften Rheinisch-Westfälisches Institut für Wirtschaftsforschung Vorstand: Prof. Dr. Christoph M. Schmidt, Ph.D. (Präsident), Prof. Dr. Wim Kösters Verwaltungsrat: Heinrich Frommknecht (Vorsitzender); Eberhard Heinke, Dr. Dietmar Kuhnt, Dr. Henning Osthues-Albrecht (stellv. Vorsitzende); Prof. Dr.-Ing. Dieter Ameling, Manfred Breuer, Prof. Dr. Walter Eberhard, Prof. Dr. Harald B. Giesel, Marianne Halstrick-Schwenk, Dr. Thomas Köster, Hartmut Krebs, Rolf Hermann Nienaber, Heinz Putzhammer, Dr. Günter Sandermann, Dr. Gerd Willamowski Forschungsbeirat: Prof. David Card, Ph.D., Prof. Dr. Clemens Fuest, Prof. Dr. Walter Krämer, Prof. Dr. Michael Lechner, Prof. Dr. Till Requate, Prof. Nina Smith, Ph.D., Prof. Dr. Harald Uhlig, Prof. Dr. Josef Zweimüller RWI : Schriften Heft 73 Schriftleitung: Prof. Dr. Christoph M. Schmidt, Ph.D. Redaktionelle Bearbeitung: Joachim Schmidt Stephen G. Hall, Ullrich Heilemann, and Peter Pauly (Eds.) Macroeconometric Models and European Monetary Union RWI : Schriften Heft 73 Stephen G. Hall, Ullrich Heilemann, and Peter Pauly (Eds.) RW I ESSEN asdfghjk Duncker & Humblot ´ Berlin Bibliografische Information Der Deutschen Bibliothek Die Deutsche Bibliothek verzeichnet diese Publikation in der Deutschen Nationalbibliografie; detaillierte bibliografische Daten sind im Internet u¨ber <http://dnb.ddb.de>abrufbar. Alle Rechte, auch die des auszugsweisen Nachdrucks, der fotomechanischen Wiedergabe und der U ¨bersetzung, fu¨rsa¨mtliche Beitra¨ge vorbehalten #2004 Duncker & Humblot GmbH, Berlin Fotoprint: Berliner Buchdruckerei Union GmbH, Berlin Printed in Germany ISSN 0720-7212 ISBN 3-428-11398-5 Gedruckt auf alterungsbesta¨ndigem (sa¨urefreiem) Papier entsprechend ISO 9706 ∞ * Internet: http://www.duncker-humblot.de Preface This volume contains the contributions of a conference dealing with the consequences of the European Monetary Union for the macroeconometric modelling of the Euro area, which took place in Essen on November 17 and 18, 2000.The objective of the conference was to give an overview of the status and development of the econometric modelling in the Euro area based on a concentrated exchange of information and discussion among a comparatively small group of scientists. At the end of the conference the participants were convinced that the discussions including a great variety of theoretical, methodical and factual aspects from the producers’ as well as the consumers’ perspective will not fail to have a certain impact on the future development of macroeconometric modelling. Once more it became clear, however, that an ideal way to a solution of the problems is still not in sight.The future development will be characterized by a plurality of approaches and models. Thus trends continue which have had a more or less strong, durable or temporary influence on the model landscape since the emergence of the monetarist revolution, the „rational expectations“ or the “real business cycle”-models; their influence was clearly perceptible during the conference. We hope that the results and reported experiences will soon be included in the experts’ discussions.We are still at the beginning of the theoretical and empirical exploration of the macroeconomic development of the Euro area,it is not always clearly perceptible what is transitory and what is permanent, and this openness should facilitate the reception of the experiences and results which have been presented. The idea for this event was developed in the course of the Project LINK, our special thanks go to Professor Dr.Stephen Hall,London Business School,Professor Dr. Ullrich Heilemann, RWI, and Professor Dr. Peter Pauly, University of Toronto.One of the highlights of the conference was the participation of the nobel prize winner Professor Dr. Lawrence Klein – pioneer and Nestor of macroeconometric modelling – who,as his contribution shows, is following up the creation of the European Monetary Union with critical interest.The meeting was organized by Mrs. Hiltrud Nehls, graduated economist, she was sup- ported by Mrs.Claudia Lohkamp.The editorial work was done by the very experienced Mr. Joachim Schmidt, who was assisted by Mrs. Anette Hermanowski.The organization of individual meetings was taken over by the editors of this volume and by Professor Dr. Bert Hickman, Stanford University, Professor Dr. Jean-Louis Brillet, INSEE, Paris, and Professor Dr. Manfred Deistler, Technical University Vienna. The institute would like to thank all of them as well as the speakers and those who participated in the discussions. Last not least we have to mention the “Gesellschaft der Freunde und Förderer des RWI”; without their generous financial support this event would not have been possible. Essen, October 2003 Rheinisch-Westfälisches Institut für Wirtschaftsforschung Christoph M. Schmidt 6 Preface Policy in the new environment is the subject of two papers. Ray Barrell,Karen Dury and Ian Hurst examine the decision making within the ECB.They do this by evaluating simple monetary policy rules in an encompassing framework. They develop different types of rules that the ECB may consider to implement to assess their effectiveness in stabilising EMU member economies.Stochastic simulations on the National Institutes Global Econometric Model (NiGEM) are used to evaluate different types of rules and different parameterisations of the rules on the country level and for the aggregated Euro area.They find that more or less for both levels, the combined nominal GDP and inflation targeting rule, with a coefficient of 1.0 on inflation, is the best rule for maximising output and minimising inflation. Pure inflation targeting with a coefficient of 1.0 appears to be the least effective at stabilising the Euro area aggregates.It is also shown that the covariance structure of Euro area inflation can matter in terms of determining the best policy rule. As to the results and their implications for policy making at the ECB,the authors find that there is no immediate conflict within the decision making bodies of the ECB. In the last paper Christian Schumacher and Christian Dreger look for evidence from panel cointegration tests on broad money demand (MH3) in Europe.For a group of fourteen European countries the null of no cointegration cannot be rejected. Only in a core sample consisting of Germany, France, Austria, the Netherlands and Belgium cointegration is found. Other studies using Euro area wide aggregated money demand functions show that aggregated time series money demand equations perform statistically better than their counterparts for single countries. The results found here indicate that this could be partly due to an averaging effect that overcompensates the missing cointegration feature in some of the countries. Although the panel methods have the advantage of higher efficiency than usual time series cointegration tests, they are not as much developed. Some drawbacks of the panel cointegration tests are possibly neglected cross section dependence and non availability of methods to control for structural breaks. It can be questioned whether the advantage of higher efficiency due to the inclusion of the cross section dimension overcompensates these drawbacks. The discussion of this conference has made it clear that to date only a small part of the challenges the EMU poses for macroeconometric modelling has been resolved. Most progress has been made by simply building aggregate models covering the entire Euro area.The empirical quality of these models is difficult to compare to that of national country models – which by no means will vanish – but at least the new models seem to work. This indicates that there is the necessary amount of homogeneity of EMU member’s economic behaviour and that is encouraging for model builder – and for ECB policy. It was not at all so clear that this is the case if we recall the many debates on “optimal currency area” that came up when in the mid 1990s the debate on EMU 13Introduction started. No problem seems to be the proper modelling of the institutional setting,while the question of a proper catch of future reactions of the Euro area is much harder to answer. It will be interesting to see whether past experiences of a gradual absorption of the various steps of European integration will hold here, too. Until then, the many results presented at the conference have improved our understanding of EMU’s present and future problems and certainly also narrowed the debate. 14 Stephen G. Hall, Ullrich Heilemann, and Peter Pauly Lawrence R. Klein US and NAFTA – Some First Experiences and Modelling It is quite appropriate, at this meeting focusing on Macroeconometric Models and European Monetary Union, to look at similar issues from the perspective of the effects of NAFTA on the US and the other two members, Canada and Mexico. NAFTA is unusual in that it combines effects on both developed countries and a developing country. Although Mexico is a recent member of OECD, it has many of the characteristics of a developing country.The NAFTA organization is regional, as is the European Union, and it affects both trade in goods and services and in capital flows. It is not, however, involved with monetary union as is the European Union. First, let us look at NAFTA issues from the point of view of Canada. Here I draw upon an interesting analysis by a one-time colleague from econometric model building research, Donald J. Daly, of York University, Canada (Daly 1998, 2000). There was great fear that Canada would lose jobs to Mexico, as would the United States, after the implementation of NAFTA. While it is true that Mexican employment gained considerable advantage from NAFTA, since Mexican wages are far smaller than those in Canada and the US (possibly only one-tenth the hourly rates in the more advanced economies), there have been both losers and gainers from sector to sector, and the gainers more than compensated for the job losses. Not only were there gainers, but there was an overwhelming gain for Canada and the US, in the sense that overall unemployment rates dropped significantly in the post-NAFTA period. This is something to think about when reasoning on behalf of EU countries.Had they tried to reach Maastricht targets by high growth,as did Canada and the United States,they might be having,on an all round basis,the same kinds of labor market improvements, domestic budget surpluses, and low inflation, all at the same time. There is no unique route to the kinds of target values set by Maastricht criteria. Very low wage costs in Mexico were not sufficient to tip the balance against the weight of superior infrastructure (transport, communication, power, sani- tation, water, education and other infrastructural facilities). These and other features made for much stronger productivity gains in Canada and the US.The gains from trade, which have been very real, as a result of lowering barriers through NAFTA enabled the advanced countries to achieve lower average and marginal costs with greater output levels; this is achieved through realizing increasing returns to scale, which have played a considerable role in the process of technical changes in Canada and the US. The roots of economic efficiency go beyond NAFTA in an historical sense. Very good gains of this type were realized earlier by two predecessor agreements, namely the Canadian-US auto agreements (1965) and the Canadian-US Free Trade Agreement (FTA) which was crafted just before the introduction of NAFTA in 1994. These predecessor agreements permitted high degrees of specialization.Automobile manufacturing benefited by outsourcing parts and supplies for final assembly, use of modern inventory practices, and using all the infrastructured advantages,especially for shipping to and fro across the Canadian-US border. The concept of increasing returns to scale helped to identify gains for microchip production, software activities in a “business-to-business” mode. These kinds of gains can be detected in macroeconometric models that are open to exploitation of economies of scale, but not so obvious in computable general equilibrium models (CGE), where there is often preoccupation with constant returns to scale and fixed coefficients of production. The Free Trade Agreement opened possibilities on a broader scale, beyond autos,and trade in manufactured products expanded as barriers were reduced, thus enabling scale economies to be more pervasive. The analysis of economic relations between either Canada or the US with Mexico showed improvement after NAFTA was introduced, but these gains were temporarily halted by Mexico’s financial crisis of December 1994. It was not only financial, but also very political. Now, the political situation has changed a great deal, and it is to be hoped that Mexico can set out on a fresh and productive path from 2000 onwards, under a new kind of leadership. It should be noted that some large foreign direct investment activities (FDI) continued expansion in spite of the crisis. Economists broadly agree that free trade is a good thing, both from the viewpoint of welfare economics and for macroeconomic growth. At a regional level, as is the case with NAFTA, the conventional judgment is that a local agreement that does not divert trade but,instead creates trade, is economically desirable and I would surely argue that NAFTA creates trade.There are,however, winners and losers, but the prevailing economic environment in Canada, 16 Lawrence R. Klein US,and Mexico,after NAFTA began in 1994,was expansionary and helped all three countries. There were, nevertheless opponents and detractors who argued that Mexico’s crisis in December 1994, was because of NAFTA. Prevailing opinion and analysis, however, concluded that political turbulence in the Mexican presidential campaign of 1994 in which the favored candidate was assassinated,political unrest in Chiapas, and a cover-up (lack of transparency) on the part of Mexican officials was responsible for the ultimate depletion of international financial reserves. A stage was set for rapid recovery as a result of credit supplied by the IMF and the US government, continuing inflows of private capital (mainly through FDI),and very good performance in the border industries (Maquiladoras).All these activities contributed to a rapid recovery in which the loans of the United States were quickly repaid, with full interest. There was a sharp recession in 1995, but it should be pointed out that the previous financial crisis of 1982-83 (the World Debt Crisis) put Mexico into a two-year recession before expansion could be resumed. Alfredo Coutiño and I authored a study in early 1995 that argued in favor of a prediction for a recovery after a one-year recession because Mexican domestic reforms, with support from NAFTA trade, would lift the economy out of crisis much faster and better than in the period of a decade earlier (Klein, Coutiño 2000). Mexican trade, 1993-99, has expanded so much that Mexico’s ranking has grown from 17th in world export volume to 13th, surpassing many countries that are well known for trade performance. In this period, Mexico’s exports tripled, and imports more than doubled. Also FDI has expanded rapidly right after NAFTA, to reach values now in excess of $10 billion. Bond ratings have improved, and Mexico is building earnings from a stronger energy sector by taking advantage more wisely than before, of the favorable prices for crude oil. Future prospects The two major advanced countries of NAFTA, Canada and the US have been leaders in world economic expansion for nearly a decade. Output has been growing vigorously, with low inflation, and technically advanced activity in information, biological, and other innovative sectors. The US has a problem of trade and current account deficits, but much of this is due to the leadership role of the US in a troubled world economy. Mexico is in a different situation. After a period of political change it has an entirely different kind of government starting in 2001. President Fox has raised new hopes by taking an expansionist line. He aims for sustained GDP 17US and NAFTA – Some First Experiences and Modelling growth of 7%. He has yet to prove himself, but it is refreshing. Also he has found unusual cooperation from outgoing President Zedillo. Customarily, econometric models of Mexico have a “dummy variable” to designate economic discord at the end of each 6-year reign, but at the present time, it is expected that there will be a smooth transition from President Zedillo to President Fox. This is a good sign. Fox is expected to pursue an active international policy, building on the gains of the past few years that have stemmed from the NAFTA accord. In 2000, Mexico will have expanded GDP by more than 5.5%, well on the way to the 7% target. The US economy is deliberately slowing,in order to work off some excesses such as incipient inflation of wage rates and a highly distorted stock market. It should, however, be able to avoid recession in 2001 and Canada should also remain prosperous, but feel the effects of a slowdown in the US. Mexico has come to a position of low inflation (one at single-digit levels) and maintains an improving foreign exchange position. Unemployment remains steady. At the time of signing the NAFTA agreement,President Clinton declared that he would try to expand NAFTA, to include strong economies in South America. Chile has been a favored choice. During his second term, he did not succeed in bringing Chile into NAFTA, but the new US government does show some interest in enlarging NAFTA membership in this same direction.At this time, the verdict is favorable in support of the NAFTA concept, and we can look forward to seeing the gradual expansion of a strong institution. References Daly, D.J. (1998), Canadian Research on the Production Effects of Free Trade: A Summary and Implications for Mexico. North American Journal of Economics and Finance 9: 147–167. Daly, D.J. (2000), Will North American Manufacturing Jobs Move to Mexico with NAFTA? Paper presented at ASSA Meetings, Boston, January 9, 2000. Schulich School of Business, York University, North York, Ontario. Klein, L.R. and A. Coutiño (1996), The Mexican Financial Crisis of December 1994 and Lessons to Be Learned. Open Economies Review 7 (Supplement 1): 501–510. Klein, L.R. and A. Coutiño (2000), A Mexican Adjustment Scenario. Ensayos Sobre Aspectos Macroeconomicos de Mexico. Instituto de Investigation Economica y Social Lucas Alaman, A.C. 18 Lawrence R. Klein Michael Beeby, Stephen G. Hall, and S.G. Brian Henry1 Modelling the Euro-11 Economy: A Supply-Side Approach 1. Introduction In this paper a medium sized econometric model of the Euro-11 (or Eurozone/ Euroland) economy is constructed. The motivation for this is to obtain a model that can be used to conduct policy simulations and obtain forecasts for those eleven countries that will be the inaugural members of European Monetary Union. Underlying the construction of the model has been the attempt to incorporate into macro-modelling recent developments that have occurred elsewhere in macroeconomics. Three features of the model are worth noting. First, given the greater emphasis placed on the supply-side, we provide a consistent treatment by estimating a production function for the Eurozone and including its implied marginal products as cointegrating vectors (CVs) in the dynamic earnings, employment and investment equations. Second, wage bargaining theories of the type associated with Layard/ Nickell/Jackman (1991) are included in the labour market equations.Unobserved components models are then used to obtain estimates of the return from not working (unemployment benefits etc.).Third,to overcome the apparently counterfactual dynamic responses resulting from the assumption of rational expectations, agents are assumed to learn, i.e. to update their expectations each period (each of these techniques are discussed in more detail below). Overall, the addition of these features should result in a model that is not only more theoretically consistent, but also in a model that has improved dynamic properties relative to econometric models that assume rational expectations. At the heart of the model is an estimated Cobb-Douglas constant-returns-toscale production function. From this production function the marginal products for labour and capital are derived and equated to their marginal costs (the real wage and the real rental rate of capital respectively) as required by the standard neo-classical theory of the firm.These equations are then used as the long-run equilibrium relationships (CVs) in the dynamic earnings, employ1We gratefully acknowledge funding from ESRC ROPA Award No. R/022/25/0187. ment and investment equations. Given the size of econometric models, researchers typically proceed by estimating each equation individually and not as a system. One drawback from this approach is that usually only a single CV enters into each equation. This is an arbitrary restriction and not one that can be justified through either economic or econometric theory. Our approach is instead to take a more general modelling strategy. Although each equation is estimated individually we allow all CVs from the production side to enter into all the dynamic equations at the initial “general” stage. Testing down procedures then eliminate those CVs found to be insignificant. Thus more than one equilibrium relationship may enter into a single dynamic equation if their presence is not rejected by the data. This idea is based on the approach taken by Vector Error Correction models (VECMs). Although we do not estimate the equations as a system we do exploit the reduced rank characteristic of such systems. In brief, for a set of non-stationary variables let m cointegrating vectors exist, then the model may be written as, () ΔΘΔYLY X ttt t =++ −−11 αβ ε' whereαβ'Xtis the set of CVs with X=(Y, Z),and where the variables are partitioned into endogenous (Y) and weakly exogenous variables (Z).CVs do not enter the dynamic equations for the weakly exogenous variables. The above equation is then the conditional model that has an accompanying marginal model,which is not stated here. The adjustment matrix αis of dimension nxm but it is not typically diagonal, each dynamic equation may include more than one CV. The determinants of labour supply (and hence aggregate supply) within the model are set out in terms of a conventional structural econometric model enabling us to describe the main arguments in the standard union-firm model of Layard et al. In their model unions set the level of wages prevailing in the economy and in return unions agree to allow managers the right to set employment. To determine the level of wages demanded, unions need to know the value (or return) from not working. This is affected in part by the level of unemployment benefits. Data on the return from not working in the Euro-11 appears (as yet) unavailable and even if it was available, its interpretation might not be clear given the disparate levels of benefit offered by national governments. To overcome this an unobserved components model (i.e. Kalman Filter) is used to obtain some estimates of the return from not working. The advantage of this is twofold. First, estimates of the non-accelerating inflation rate of unemployment (NAIRU) can be obtained, an exercise that is conducted below. Second, the effects of policies that target changes in labour supply measures can be simulated and assessed. Our approach thus brings together structural (i.e. the union-firm model) and time-series models of the NAIRU. 20 Michael Beeby, Stephen G. Hall, and S.G. Brian Henry The third innovative feature of this paper is the assumption that individuals update their expectations each period (i.e. they learn), rather than simply assuming that agents have rational expectations. Introducing learning into econometric models has potentially large rewards. Unlike in small analytical models, the introduction of RE has changed the short-run policy responses of the larger models in ways that are not totally plausible. For instance, if the government announces that interest rates will increase by one per cent,exchange rates in large-scale models will typically jump by approximately fifty per cent – a highly implausible result (Fisher et al. 1990). However RE does have big theoretical advantages since it assumes that agents use their information set optimally, and it is the dominant assumption in modern macroeconomics. Learning can potentially maintain the long-run advantages of RE (since learning has been shown to converge asymptotically to the RE solution under rather general conditions),while overcoming its poor short-run policy responses in large-scale models. Expectations can enter these models in many areas, but it is in the exchange rate sector where they have been shown to have the largest effect.Typically expectations have been introduced via a forward-looking open arbitrage exchange rate equation. The easiest way to introduce a shock is through an exogenous price, usually the world price of oil. In this paper learning also enters via the exchange rate equation, though the model will be subjected to more than just a single shock. The outline of the paper is as follows.Section 2 describes the data,the methods used to construct a single aggregate from the eleven individual series, and plots the behavior of the main series over the past fifteen to twenty years. The model is presented in Section 3 with the estimation results shown in Section 4. Section 5 contains estimates of the NAIRU for Euroland and Section 6 describes the results from some simulation exercises. Section 7 concludes. 2. The data Methodological issues Aggregating data across countries is of course problematic. Movements in exchange rates rule out simply adding up series across countries. Methods that use estimates of exchange rates that are instead based on purchasing power parity (PPP) are also unsatisfactory since PPP estimates are themselves not uncontroversial and may be sample dependent.This means that before any aggregation method can be chosen the following decisions need to be made.First is the aggregation across levels or growth rates? Second, are the weights on each country’s series assumed to be fixed (e.g. as with Paasche and Laspeyres indices) or time varying? There are four available choices. The approach here is to use data that is aggregated across growth rates and where the weights are time varying.Beyer et al.(2000) have subsequently shown that this approach is superior to other methods. Their reasons are discussed below. 21Modelling the Euro-11 Economy: A Supply-Side Approach The weights attached to each individual country’s series are assumed to vary over time. This approach was chosen since one major drawback with using fixed weights is that they fail to capture any large changes to relative prices that may occur. With data from the Euro-11 there have clearly been periods where large devaluations and exchange rate movements have occurred. However, applying varying weights to series in levels results in distortions to the data following exchange rate movements, distortions that can be shown to not occur if the aggregated data is in growth rates. For these reasons the chosen method of aggregation was to use the series in growth rates with the applied weights allowed to vary over time. For an aggregate real series Y, and its two country components Y1and Y2, the aggregation method used and the derivation of the weights is as follows. ΔΔΔΔ ΔyYYY Y YY Yy tt t t t tt t tt =− ≈≡ +=+ −−− − log log 1 1 12 1 111 ωΔytt221 ω− where Δis the first difference operator, yis the log of Y, and the weights ωi, for i = 1 to 2 are ω1 11 11 2 2 1 1221 t tt tt t t t tttt EY EY E Y Y YEYE =+=+/ where Eis the exchange rate used to convert the values into a common currency (a similar formula exists for ω2). Exchange rate movements then alter the magnitude of the weights attached to a country’s series and can be expected to vary over time. However the magnitudes of any individual series are computed from growth rates and are not affected by changes to exchange rates. Euro-11 data Figure 1 plots the Euro-11 data for some key Euroland series. The upper part plots the annual change in output (GDP), consumption (C) and business investment (I). The series display a strong degree of procyclicality with consumption in particular tracking changes to GDP. Throughout the 1980s the Euro-11 economy exhibited steady growth without any of the booms/recessions that were observed in the individual countries. Aggregating across the eleven economies has had the effect of evening out these deviations from trend, implying that country specific cycles may have become less synchronised. During the 1990s however the smoothness of the series becomes less prevalent with the beginning of this decade marked by a strong boom in GDP and consumption. A sharp downturn in the economy then begins, bottoming out in 1993, before a sustained period of steady growth is again maintained. One last fact to note is the high volatility of the investment series.Fluctuations in investment are much more severe than for any of the other series documented, a fact that has been well documented within individual countries. 22 Michael Beeby, Stephen G. Hall, and S.G. Brian Henry variable, or measure of competitiveness, is relative wage costs (WCR). Variations in quantities are captured by total final expenditure (TFE=GDP+M) in the imports equation, and by a world trade (WT) measure in the exports equation. () MfTFEWCR tt =,, (18) () X f WT WCR ttt =,, (19) We also have a technical relation for the consumer price (Pc) linking it to the producer price (P) in (5), ΔΔPP ct t =(20) In aggregate a resource constraint binds, GDP C I GC X M ttt t t t =++ +− (21) where Iis business investment and government consumption (GC) is assumed to be exogenous. Equation (14) describes the evolution of the capital stock with investment changing over time in the usual manner, () IK K tt t =−− − 11 δ.(22) Finally, we can note the policy rules used.The monetary rule adjusts the nominal interest rate such that deviations of inflation from its target rate are minimised (using the conventional proportional, integral, differential (PID) formula). Fiscal rules (again using a PID form) adjust tax rates such that the government’s budget constraint is met,where for the moment government spending is assumed exogenous. 4. Estimation results Results from estimating each of the equations are given in the appendix. The production function was calibrated such that the share of capital in output was 0.36, the value typically chosen in the real business cycle literature (e.g. Kydland, Prescott 1982). Constant returns to scale was also assumed. From these equations it is straightforward to obtain estimates of the marginal products of capital and labour. In the long-run consumption function wealth effects are reasonably strong. A one per cent increase in the ratio of real net wealth to real personal disposable income leads to a rise in consumption of 0.13 per cent.Long-term real interest rates also enter the CV for consumption though its effects are not particularly large. The hypothesis of homogeneity in consumption and income could not be rejected. Underlying the estimation process was the decision to obtain re29Modelling the Euro-11 Economy: A Supply-Side Approach sults that are parsimonious and that have good dynamic properties. The dynamic consumption equation typifies this. Dynamics enter solely from the change in real personal disposable income (lagged one quarter).Lagged terms in consumption, income, inflation and real interest rates were also included at the initial first stage but their t-statistics were either very low or entered with the incorrect sign. The estimated dynamic investment equation contains lagged dynamic terms in both GDP and investment. Its CV is the difference between the real interest rate (used to proxy capital’s marginal cost) and the marginal product of capital.If the interest rate is above capital’s marginal product,investment falls so that the capital stock can return to its long-run level. The dynamic employment equation contains two long-run relationships. One is the difference between actual output and the level of output that can currently be produced by supplyconstrained firms. This term enters with a coefficient of 0.13. The second CV is the difference between the real wage (labour’s marginal cost) and the marginal product of labour.If labour’s marginal product is above its marginal cost then the level of employment increases since firms seek to employ more workers. The coefficient on this term is 0.02. Dynamics enter via two lagged changes in employment. The earnings equation also contains two long-run terms. One being the difference between actual output and a measure of output by the supply constrained firms. The second term is from the wage bargaining model. Unions choose the wage dependent upon the level of unemployment (a proxy for the probability that workers may lose their job if the chosen wage is too high) and the return from not working (A). The series for A was obtained by a Kalman Filter estimation process with A assumed to be an unobserved variable. Also included in the Kalman Filter was the level of unemployment and the real wage. The residuals from this regression then enter the dynamic wage equation. Interpretation of the coefficients isn’t straightforward. For instance, the coefficient on the lagged unemployment term is –0.02, which seems small but simulation exercises suggest that this has a strong effect on the simulation properties of the model. Aggregate prices in the long run are a weighted average of unit wage costs and import prices, with UWC having the largest weight (0.82). The dynamic price equation then contains this CV (with a coefficient of 0.2) plus a lagged dynamic price term and the change in earnings lagged one quarter. In turn, import prices are a weighted average of the log of prices plus the log of world prices, with both having approximately equal weight. The final set of estimated equations concern the open economy effects from the trade equations. In the long run exports equation, exports move one-for-one with world trade, and the coefficient of relative wage costs is 0.38. This CV enters the dynamic exports equation with a coefficient of 0.19. The other term is a lagged dynamics term. The CV for imports has total final ex30 Michael Beeby, Stephen G. Hall, and S.G. Brian Henry penditure as the quantity variable, again imposed with a unit coefficient, and relative wage costs which are more important for imports since its coefficient is now 0.64. In the dynamic imports equation the error-correction mechanism enters with a coefficient of –0.18, implying a return to equilibrium reasonably quickly each quarter.Also included are three lagged dependent variables with coefficients of 0.13, 0.25 and 0.18 respectively. This completes a description of the econometric results. 5. Estimating the NAIRU in the Euro-11 In this section the model is used to obtain two estimates of the NAIRU. The first estimate is perhaps more appropriately termed the NAWRU, since it is earnings inflation that is being targeted and not price inflation.The estimate is obtained by asking what the unemployment rate should be to set the current period’s wage inflation rate equal to the previous period’s rate.The mechanics of this exercise simply involve inverting the earnings equation. A plot of the resulting estimate is shown in Figure 2. From these figures it appears that the short-run NAWRU closely tracks the actual unemployment rate. The biggest discernable difference between the two series is over the period 1991 – 1995 when the estimated NAWRU is consistently lower than the actual unemployment rate. This occurs at a time when the inflation rate was steadily falling. Figure 2 lower part plots a long-run version of the NAIRU. This series was obtained by asking what level of unemployment would be necessary for the price level to be equal to the long-run level of prices implied by the cointegrating vector in the price equation. Based on these figures and in contrast to the short-run NAWRU estimates, the NAIRU has proved remarkably invariant given the changes that have occurred in the Eurozone over this period. From 1993 the NAIRU begins to gradually rise as the actual unemployment rate increases into double digits. The net effect is that the NAIRU is now half of one per cent higher in the 1990s than it was in the 1980s. It may seem surprising that our estimates for the NAIRU are relatively invariant over time. However these results are not unusual. Laubach used a Kalman Filter to estimate NAIRU’s for the G7 and also found his estimates to be approximately constant for each of the countries2. 6. Simulation exercises In this section the response of the model to two types of shocks is presented. First, the model is subjected to a temporary shock to government consumption. Second, the return to not working is shocked temporarily. 31Modelling the Euro-11 Economy: A Supply-Side Approach 2It may be that our specification of the dynamic wage equation (8) is responsible for the constancy of the NAIRU. The return from not working (A) is modelled as a Kalman Filter process and it may be that any movements in the NAIRU are being captured by this variable. A temporary shock to government consumption For this shock government consumption is increased by 5 % over the period 1987Q1 to 1990Q1 (results are presented in Figures 3 and 4).Figure 3 plots the responses of GDP, consumption and investment. Since government consumption is a component of GDP from the resource constraint, GDP is higher for the duration of the increase in GC. Consumption, which is a function of real disposable income, rises gradually following GDP’s increase. Surprisingly there is little effect upon investment. By the end of the sample period invest32 The short-run NAIRU for the Euro11 1985-1 to 1999-4; in % Short-Run Long-Run 6 6 6 6 7 7 7 7 8 8 8 8 9 9 9 9 10 10 10 10 11 11 11 11 12 12 12 12 13 13 13 13 U* U* UP % UP % 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 Figure 2 Authors’ calculations. Explanations see text. Michael Beeby, Stephen G. Hall, and S.G. Brian Henry 33 Temporary shock to government consumpt i on 1988-1 to 1999-4; in % GDP, C, I INF, RLG, UP X, M, RX -2,5 -2,5 -20 -20 -7 -7 -2,0 -2,0 -1,0 -1,0 -1,5 -1,5 -0,5 -0,5 0 0 -15 -15 -6 -6 -10 -10 -5 -5 0,5 0,5 0 0 5 5 -3 -3 -4 -4 1,0 1,0 10 10 5 5 1 1 0 0 -2 -2 -1 -1 1,5 1,5 15 15 2 2 GDP INF RX M X RLG UP C I 88 89 90 91 92 93 94 95 96 97 98 99 Figure 3 Authors’ calculations. Explanations see text. Modelling the Euro-11 Economy: A Supply-Side Approach ment begins to rise as interest rates fall (see Figure 3 middle) following the decreases in output. Inflation too is unchanged following the shock but unemployment falls during the period of the increased expenditure before beginning a sustained period of recovery. Positively correlated with the decline in GDP is the depreciation of the nominal effective exchange rate (RX), plotted in Figure 3 lower part. Occurring with the depreciation is a decline in imports, though the change in exports is much smaller. Expectations enter into the model via an open arbitrage equation requiring the difference in the levels of real interest rates between two countries to be equal to the expected change in the real exchange rate. Under rational expectations the expected exchange rate next period would be equal to the actual exchange rate, subject to any shock. Under learning it is instead assumed that agents use a rule to predict the expected exchange rate. This expected exchange rate rule can be a function of any of the other variables, though for the purposes of modelling a parsimonious specification is usually chosen. Each period, as new information becomes available, agents update the rule as they learn which variables are relevant for forecasting the exchange rate and which variables are not. Marcet/Sargent (1989) have shown that so long as the learning rule contains some information that is correlated with the true process driving exchange rates then the exchange rate under learning will converge to the exchange rate equilibrium derived under rational expectations. 34 Learn i ng parameters f ollow i ng a government shock 1988-1 to 1998-4; in % -4 -4 -3 -3 -2 -2 -1 -1 0 0 1 1 2 2 a1 a2 88 8 99 0 9 1 9 2 9 3 9 4 9 5 9 6 9 7 9 8 Figure 4 Authors’ calculations. Explanations see text. Michael Beeby, Stephen G. Hall, and S.G. Brian Henry 35 T emporary s h oc k to t h e l a b our supp l y 1987-1 to 1990-1; in % GDP, C, I INF, RLG, UP X, M, RX -1 -10 -3 -3 -10 -1 -0,5 -8 -2 -2 -8 -0,5 0 -6 -1 -1 -6 0 0,5 -2 -2 -4 0 0 -4 0,5 1,0 2 2 2 2 0 1 1 0 1,0 1,5 4 3 3 4 1,5 C GDP INF INF I UP X RX M 88 89 90 91 92 93 94 95 96 97 98 99 Figure 5 Authors’ calculations. Explanations see text. Modelling the Euro-11 Economy: A Supply-Side Approach In our model the expected real effective exchange rate (between Euroland and the rest of the world) is assumed to be a linear function of lagged nominal effective exchange rates. Figure 4 plots how the learning parameters evolve over time,with a1being the constant and a2the coefficient on lagged exchange rates. One would expect these parameters to converge over time as the exchange rate converges asymptotically onto the rational expectations solution although here, ten years after the shock, there is still some way to go before this has occurred. A temporary shock to the return from not working As above, the shock to the return from not working (which can be interpreted as a decrease in unemployment benefits) is temporary and also lasts for the period 1987Q1 to 1990Q1. Given the bargaining framework, any decrease in unemployment benefits results in a fall in the wage paid by firms, and demanded by unions,since there is a lower return available elsewhere in the economy.Figure 5 indicates that GDP,consumption and investment are all higher than prior to the decrease in benefits.The amount of labour supplied increases in this economy as the wage falls resulting in higher GDP,consumption and investment. Inflation (a function of unit wage costs) rises along with the real long rate. Unemployment falls as the amount of labour supplied in the economy is raised. Nominal exchange rates appreciate for the five years following the shock with imports rising as their price has fallen (following a short lag). 36 Learning parameters following a shock to the labour supply 1980-1 to 1999-4; in % -3 -3 -4 -4 -1 -1 0 0 1 1 2 2 3 3 a1 a2 88 89 90 91 92 93 94 95 96 97 98 99 Figure 6 Authors’ calculations. Explanations see text. Michael Beeby, Stephen G. Hall, and S.G. Brian Henry Exports are again unresponsive to these exchange rate movements.The learning parameters a1and a2show a higher degree of convergence following this shock and appear well on their way to converging (Figure 6). 7. Conclusion This paper has estimated a medium-sized model of the Euro-11 area. Three features of the model are worth noting. First, a specific supply-side built up from a CobbDouglas production model was constructed and estimated. Second, a general form for each equation was specified so that more than one long-run relationship could enter into each dynamic regression. Third, the model was simulated under the assumption that agents learn. The model was then used for two purposes. First, to estimate the level of the NAIRU in the Euro-11 area.Two measures were constructed:one short run and one long run measure. The short-run measure was found to closely track actual unemployment over the sample period.In contrast the long-run measure of the NAIRU was relatively stable at 9 per cent, though there was some evidence to suggest that it may be beginning to fall. Second, the model was used to study how the Euro-11 economy responds to different types of shocks. Temporary shocks to government expenditure and unemployment benefits were each found to act as stimulants to GDP and other key series for the first five years at least before the effect of higher real interest rates began to feed in. In future work, we intend to extend the model’s database by approximately fifty years so that questions concerning the stability of the model can be addressed and experiments conducted where the duration of the shock (policy change) is much longer. Appendix: Estimation Results Production () y k n dummies tt t =+− +036 1 036.. (A1) where lower case letters denote logs. From (i) the marginal products of capital (MPK) and labour (MPL) are next derived which in turn are set equal to the real rental rate of capital and real wage respectively. Consumption ()( ) ECMC c y nw y RLG INF dummies ttt tt t t =−+ − − + − +0 08 014 0 002.. . +εt ΔΔc rpdi ECMC dummies tttt =+ − + + − 0 004 0 24 012 1 .. . ε(A3) (5.15) (2.84) () () () Rsc NH 20 86 0 003 4 2 84 2 151 1 0 48== = = =.; . ; .; .; ..σχ χ χ 37 (A2) Modelling the Euro-11 Economy: A Supply-Side Approach Investment ECMK RLG INF MPK ttt t =−−+026.(A4) ()ΔΔΔΔi y y i ECMK du ttt t t =+ − + − + −− − − 0 006 0 37 0 22 0 03 12 1 1 .. . . mmies t +ε (A5) (2.47)(1.45) (1.70) (2.33) () () () Rsc NH 20 42 0 014 4 359 2 0 26 1 2 03== ===.; . ; .; .; ..σχ χ χ Employment ECMW w p mpl ttt t =−+ −506.(A6) () ECMY y k n dummies tt t t =− −− −036 1 036.. (A7) i.e. the difference between actual output and supply constrained output. ΔΔΔnnnecmy ttt t =+ + + − −− − 0 0005 019 0 45 0133 12 1 .. . . (A8) (1.62) (1.99) (4.85) (2.10) 0 002 1 .ecmw dummies tt −++ε (1.26) () () () Rsc NH 20 48 0 003 4 753 2 1 86 1 051== = = =.; . ; .; .; ..σχ χ χ Earnings ()Δer A ecmw er p tt ttt =− + − − − − −−− 0 74 1 266 0 38 056 111 .. . . (A9) (11.5) (6.52) (9.29) 002 1 .UP dummies tt −++ε (12.27) () () Rsc H 2094 0 011 4 7 49 1 0 80== = =.; . ; .; ..σχ χ Prices () () ECMP p n er y pm tt ttt t =− − +− −−006082 1082.. .* (A10) ΔΔpp ecm ttt t −=− + −−11 0 003 0 2.. ε(A11) (2.06) () () () Rsc NH 20 79 0 003 4 2 49 2 0 20 1 068== = = =.; . ; .; .; ..σχ χ χ Δpm p wp dummies tttt =− + + + +067 049 051.. . ε(A12) (85.2)(9.13) (9.13) () () () Rsc NH 2092 0 04 4 490 2 098 1 0 04== = = =.; .; .; .; ..σχ χ χ 38 Michael Beeby, Stephen G. Hall, and S.G. Brian Henry The bulk of this paper is concerned with the model’s simulation properties, which we turn to in the next section. Here we deal with the earlier steps in the calibration process, using the database described in the previous section. The key parameters to be established for the steady state (Table 1) are g : the real rate of growth δ: the rate of depreciation to set the equilibrium path χ: the equity premium β: the share of capital in the economy and a series of other parameters to set the steady-state ratios for inventories, government transfers, debt, real consumption, real investment and other income to GDP.Weights need to be determined for the components of the price indexes and the trade equations need to be calibrated as a whole in a manner consistent with stable exchange rate determination. Values have to be assigned to the NAIRU and to the two key parameters in the consumption function – the subjective discount factor and the probability of death. Lastly the adjustment speeds of labour demand, wages and the GDP deflator need to be set. The results of these choices are set out in the Appendix.Where series were stable we used estimated or average values from the data period available for the euro area. Where they were unstable we used values prevailing at the end of the period, particularly for the projection of the model into the future. Thus, for example, gwas calibrated as 0.5 % per quarter to omit the slower growth associated with convergence to the Maastricht criteria but without a return to the higher rates occurring before the 1990s or any strong new economy effect. Such prospects are best left to simulations. In the same way the NAIRU was estimated as an HP filter through the data period with its most recent value projected forward. Again it may very well be appropriate to argue that structural reforms will enable this value to fall in the future and hence this can be entered into the simulations as we explore in the next section. The calibration of the equity premium is largely based on the US experience (Siegel 1992).Calibrating the consumption equation proved a little difficult as the parameter values that fit the data well imply a sensible marginal propensity to consume but rather implausibly high values for the subjective discount factor and probability of death.Rather than solve this by the traditional method of adding somewhat arbitrary lags to the consumption equation we decided to remain data consistent as an initial step. While we can fix the government sector according to practice prevailing in the data period as described, the monetary authority’s reaction function is proba45Using EDGE – A Dynamic General Equilibrium Model of the Euro Area bly the most important choice from our point of view.Again what is necessary is a simple starting point from which simulations of different functions can be run as we will demonstrate in the next section.We therefore posit a Taylor rule based on the foreign real rate of interest and equal weights on the inflation and unemployment gaps.No smoothing is assumed, although the specification of the equation allows this.Taking all these calibrations together and filling the model across the data period generates a reasonable fit, as indicated by Table 2. The second step is to calibrate the dynamic model. This involves the computation of leads and lags for the principal equations (Table 3). Kortelainen (2002) shows the results from some stochastic simulations to determine how well the calibrated model seems to track the properties of the data. 3. Simulations to set the characteristics of the model EDGE is coded and solved in TROLL, using the Laffarque-Boucekkine-Juillard algorithm to solve the model forward. In setting up the simulations we first need to run the steady state model far enough forward to generate suitable terminal conditions for the dynamic model.We have used 800 periods (200 years). Then in solving the dynamic model over the same time horizon we insert a correcting factor for all the nominal and price variables in the terminal period to equate the differences between the dynamic and steady state models. Thus while the real variables converge to their steady state values there is no such requirement for the nominal variables. We did not experience problems of convergence and EDGE appears to be dynamically stable over the long run. This therefore should give a suitable base from which to compute policy simulations. In order to set the properties of the model we ran a series of standard shocks to the policy variables: taxes, government spending and the inflation target; to drivers of the model: rate of growth of the labour force, world demand, equity premium etc. However, some shocks can only be temporary, such as those to the exchange rate and interest rates, if the steady state is to be regained. To illustrate this we show examples of the following shocks in this section: 1. A shock from government policy in the form of a permanent increase in public consumption equivalent to 1 % of GDP. 2. A domestic shock in the form of a permanent increase in the equity premium by 1 %. 3. A foreign shock in the form of a permanent increase in world demand by 1 %. 46 Mika Kortelainen and David G. Mayes It is important to have some yardstick against which to judge the resultant paths.Hunt (2000) was particularly helpful in providing a comparison with the responses of the IMF’s MULTIMOD. (There is always a danger of circularity here in that if models are calibrated2against each other they may embody modellers’ prejudices rather than observed behaviour.) 3.1 A shock from government policy The key feature of the model that this simulation, increasing public consumption permanently by the equivalent of1%ofGDP,illustrates is that a shift of resources towards the government reduces overall GDP,as productivity in the public sector tends to be below that in the private sector (the picture is shown in Figure 1). This of course is mainly because of the higher levels and growth rates of productivity that are possible in manufacturing industry compared with more service based activities.In the short run there is an increase in activity as it takes time for the increase in taxation required to finance the increased public expenditure to reduce private sector spending. The counterweight to this is that unemployment falls as a result of the sectoral shift towards more labour intensive activities. Because spending runs ahead of tax revenue in the short run public debt increases, inflation also rises and along with that nominal interest rates rise as the central bank tries to maintain price stability. The interest rate increase is sufficiently large for it to be a real as well as a nominal increase.This in turn leads the real and nominal exchange rates to increase as well. Because of the size of the initial shock to government debt it takes a long time for nominal magnitudes to return to equilibrium and a noticeable proportion of the adjustment process is still to come after the 15 years shown in the Figure. 3.2 A permanent increase in the equity premium An increase in the equity premium is in effect a downward shock to wealth as a result of an increase in risk in the corporate sector. The immediate effect is a cutback in consumption as the private sector tries to adjust (Figure 2). This slows the economy, inflation falls, the real exchange rate falls and unemployment rises. The jump in these variables is largely the result of the impact of expectations of the future problems being discounted back into current asset prices. Monetary policy can ease under these circumstances and real wages fall. What is interesting in this example is that monetary policy cannot solve the adjustment problem. Because the shock is to wealth, a stock variable, the adjustment is not nearly complete even within the 15 year period shown. If nominal interest rates were to be cut even further in the hope of having a rath47Using EDGE – A Dynamic General Equilibrium Model of the Euro Area 2The simulations shown stem from an earlier version of the model but the changes do not affect the qualitative results and in most cases have little visible impact on the graphs shown. 48 Mika Kortelainen and David G. Mayes A shock f rom government pol i cy i nthe f orm o f a permanent i ncrease in public consumption equivalent to 1% of GDP Difference from baseline years years imports exports investment real wage real product wage consumption CPI deflator GDP deflator 2 0.2 0.3 2.0 0.20 0 1.5 0 1 0.1 0.2 1.5 1.0 0.1 0.5 0.15 -0.01 1.0 0.5 0.10 -0.02 0 -5 -1 -0.1 -0.1 -0.5 0 -0.04 -1.0 -15 0 0 0 0 0.05 -0.03 -0.5 -10 -2 -0.2 -0.2 -1.0 -0.05 -0.05 -1.5 -20 real nominal nominal real government net lending/GDP government debt/GDP current account/GDP net foreign assets/GDP GDP Main indicators; in % Wages; in % Interest rates; in %-points Debt and lending; in %-points Inflation; in %-points Unemployment; in %-points Exchange rates; in % Account and assets; in %-points 33669912 1215 15 Figure 1 Authors’ calculations. Explanations see text. 49Using EDGE – A Dynamic General Equilibrium Model of the Euro Area A domestic shock in the form of a permanent increase in equity premium by 1%-point Difference from baseline years years imports exports investment real wageproduct real wage consumption CPI deflator GDP deflator 5 0 2 2.0 0.5 0.30 15 160 -0.5 -1.0 1 1.5 1.0 0 0.5 0 0.25 0.20 10 5 -0.5 0.15 0 120 -10 -2.0 -2 -0.5 -1.5 0.05 -10 40 -5 0 -1.5 -1 0 -1.0 0.10 -5 80 -15 -2.5 -3 -1.0 -2.0 0 -15 0 real nominal nominal real government net lending/GDP government debt/GDP current account/GDP net foreign assets/GDP GDP Main indicators; in % Wages; in % Interest rates; in %-points Debt and lending; in %-points Inflation; in %-points Unemployment; in %-points Exchange rates; in % Account and assets; in %-points 33669912 1215 15 Figure 2 Authors’ calculations. Explanations see text. 50 A foreign shock in the form of a permanent increase in world demand by 1% Difference from baseline years years imports exports investment real wage real product wage consumption CPI deflator GDP deflator 0.5 0.05 0.04 0.04 0.02 0.004 0 0 0.4 0.3 0.04 0.03 0.03 0.02 0.02 0.01 -0.1 -0.2 -0.3 -0.1 0 0.01 -0.02 -0.01 0.01 -0.02 -0.01 0 0.002 -0.006 -0.004 -0.002 0 -0.5 0.1 0.2 0.02 0 0 -0.4 -0.2 -0.1 0 -0.04 -0.02 -0.03 -0.008 -0.6 -0.3 real nominal nominal real government net lending/GDP government debt/GDP current account/GDP net foreign assets/GDP GDP Main indicators; in % Wages; in % Interest rates; in %-points Debt and lending; in %-points Inflation; in %-points Unemployment; in %-points Exchange rates; in % Account and assets; in %-points 33669912 1215 15 Figure 3 Authors’ calculations. Explanations see text. Mika Kortelainen and David G. Mayes er smaller cut in inflation then the real adjustment process for wealth would merely be dragged out rather longer, making the loss on the unemployment side of the Taylor rule greater.The result would therefore look rather different if the central bank were purely targeting inflation. 3.3 A permanent increase in world demand What is interesting from the simulation of a foreign demand shock is that it has very little impact on the economy (Figure 3), with the exception of the trade variables themselves. Because the shock is seen to be permanent it has an immediate effect on behaviour through expectations, even though the realisation, period by period will come through much more steadily. The immediate effect of this increase in demand comes through partly on trade volumes and partly through an appreciation in the real and nominal exchange rate. As a result imports increase more than exports as in effect the terms of trade move in favour of the euro area. 4. Monetary policy simulations Thus far the shocks we have imposed show the central bank responding following a simple Taylor rule. Clearly the areas of greatest interest to us are to explore what happens when the central bank is itself the initiator of shocks and how the way the bank operates can affect the operation of the economy. One of the obvious changes, which we do not explore here, is that rule itself could be changed. We could alter the weights in the rule or indeed replace the Taylor rule by an inflation forecast targeting rule (Amano et al. 1999). An inflation targeting rule is more difficult to implement as we have to be able to solve the model for the inflation forecast before then implementing the rule. Since the rule itself is part of the forecast this is a tedious process. However, in this paper we focus on three aspects of the operation of the Taylor rule itself. The first is simply to assess what happens when monetary policy settings change in the form of 4. A temporary shock in the form of a two year increase in interest rates by 1 %3. In this case the shock is not anticipated but it is not a “surprise” in the sense that central bank is deviating from its anticipated rule for a short-run advantage.It is merely responding to information that it has,in the expected manner. There is thus an asymmetry in the first period when the change is implemented. 51Using EDGE – A Dynamic General Equilibrium Model of the Euro Area 3This simulation is labelled 4 as it follows on from the three in the previous section. However this “straightforward”form of shock has only limited interest for the policy maker. We, therefore go on to show what happens if the central bank tries to implement a change in policy in the form of a change in the inflation target.We are concerned in this instance to show the importance of “credibility”. If a central bank is “credible” in this sense then the private sector will expect the policy change to succeed and inflation expectations will shift by the full extent of the change in target. Our simulation in this case is thus: 5. An exploration of the importance of credibility in the form of a monetary policy shock of a 1 % increase in the inflation target a when the central bank is credible b when the central bank is not believed and the central bank and private sectors act simultaneously c when the central bank is not believed and the central bank acts first. The importance of simulation 5c is that in this case the private sector has the opportunity of observing the central bank’s action. We can perhaps relate this simulation to the discussion of transparency. If the private sector can be better informed about what the central bank is doing then the costs of policy will be lower. Even so the case explored here is rather extreme. It seems unlikely on the one hand that a large policy change would be fully credible immediately. At least some of the private sector would doubt that it would be sustained. Hence to some extent credibility would be earned by experience (see Vilmunen 1998 and Mattila 1998 for a discussion of these “peso” problems). On the other hand it also seems unlikely that the central bank would not eventually gain a substantial measure of credibility if it persevered with its policy. In some respects this is akin to the process of learning (Tetlow et al. 1999). If events do not turn out as anticipated then one would expect that both the central bank and the private sector we realise that their view of the world may be incorrect and slowly adjust their behaviour towards the new circumstances. Our next step therefore is to take the case of an external shock and show the difference in impact when central bank recognises the shock from when it does not. The particular shock is as follows: 6. An examination of the impact of the central bank’s failure to recognise a structural shift in the economy in the form of a 1 % fall in the NAIRU. Two simulations have to be run in this instance,the first showing what happens when the central bank does recognise the structural shift and the second when it does not.This issue is one of the most important in monetary policy as structural shifts are always difficult to detect (unless due to regulatory change, in which case the debate is over the size of the response) and confusion by the 52 Mika Kortelainen and David G. Mayes central bank of a shift in a relationship with a shock to the relationship4can have major consequences,particularly since it may mislead the private sector. We take one such example of learning but rather than just exploring how the private sector might learn whether the monetary authority has actually changed its behaviour we take the case of a change in the supply-side of the economy and consider the effect on the behaviour of the model of different learning processes: 7. An exploration of learning in the form of a 1 % reduction in the real rate of interest a when both central bank and private sector perceive it immediately b when the central bank fails to realise it c when the central bank and the private sector learn of the change linearly over 5 years 4.1 A temporary interest rate shock The three shocks we have shown thus far (in the previous Section) can be permanent,although in the case of the government spending shock it is because of a matching increase in financing. It is rational for the private sector to act as if these shocks were permanent in the light of no other evidence as future shocks could be of either sign, unless of course public expenditure is reaching the bounds of plausibility as a share of GDP.A change in interest rates on the other hand is inherently a short run temporary phenomenon unless there has been a change to expected growth rates,productivity or the inflation target.As is clear from Figure 4,a 2 year (nominal) interest rate rise has no long-term impact on the economy. In almost all cases variables have returned to the steady state after 3 years, i.e. within one year of the ending of the shock. However, as there is a one-off fall in inflation,this will result in a permanent appreciation in the nominal exchange rate (temporary appreciation in the real exchange rate). Net foreign assets will also make a one-off permanent adjustment. The rise in unemployment (fall in GDP) is substantial – over2%ofGDP–intheshort run but rapidly disappears. As before the impacts of the shock are spread among quantities and prices, with the real wage falling temporarily as well as unemployment rising. Thus stickiness in the system is clearly limited. Even in the case of a temporary interest rate shock it is necessary to find the cause somewhere in order to conduct a logically coherent simulation. Otherwise it will merely appear as a monetary policy surprise that will generate expectations of changes in the target of monetary policy.In this case we assume it 53Using EDGE – A Dynamic General Equilibrium Model of the Euro Area 4i.e.confusing a shift in a curve with a shift along it. 54 Mika Kortelainen and David G. Mayes A temporary shock in the form of a two years increase in interest rates by 1% Difference from baseline years years imports exports investment real wage real product wage consumption CPI deflator GDP deflator 1.0 0 2.0 2.0 0.1 0.8 1 0.5 0.5 0 -0.1 -0.2 1.5 1.5 1.0 1.0 0.5 0 -0.1 0 0 -2.0 -1.5 -0.5 -0.5 -0.5 -0.4 0.2 -2 -1.5 -1.0 -0.5 -0.4 -0.3 0 0.5 0 -0.3 -0.2 0.6 0.4 -1 -1.0 -0.5 -2.5 -0.6 -1.0 -1.0 -0.5 0 -3 -2.0 real nominal nominal real government net lending/GDP government debt/GDP current account/GDP net foreign assets/GDP GDP Main indicators; in % Wages; in % Interest rates; in %-points Debt and lending; in %-points Inflation; in %-points Unemployment; in %-points Exchange rates; in % Account and assets; in %-points 33669912 1215 15 Figure 4 Authors’ calculations. Explanations see text. 61Using EDGE – A Dynamic General Equilibrium Model of the Euro Area A permanent fall in NAIRU by 1 % -point Myopics vs. sharp-eyed central bank Difference from baseline Main indicators; in % Wages; in % Interest rates; in %-points Debt and lending; in %-points Inflation; in %-points Unemployment; in %-points Exchange rates; in % Account and assets; in %-points 33669912 1215 15 years years imports exports investment real wage real product wage consumption CPI deflator GDP deflator 0.04 0.005 0.35 0.4 0.004 0.012 0 0 0.02 0 0 -0.005 0.30 0.25 0.2 0.010 0.008 -0.02 0.002 0 0.006 -0.1 -0.2 -0.08 -0.06 -0.025 -0.020 0.05 0.10 -0.2 -0.004 0.002 -0.06 -0.4 -0.04 -0.02 -0.015 -0.010 0.15 0.20 0 -0.002 0.004 -0.04 -0.3 -0.10 -0.030 0 -0.4 -0.006 0 -0.08 -0.5 real nominal nominal real government net lending/GDP government debt/GDP current account/GDP net foreign assets/GDP GDP nominal wage Figure 6b Authors’ calculations. Explanations see text. 62 Mika Kortelainen and David G. Mayes An exploration of learning in the form of a 1%-point reduction in the real rate of interest when the central bank fails to realise it Difference from baseline Main indicators; in % Wages; in % Interest rates; in %-points Debt and lending; in %-points Inflation; in %-points Unemployment; in %-points Exchange rates; in % Account and assets; in %-points 33669912 1215 15 years years imports exports investment real wage real product wage consumption CPI deflator GDP deflator 0.10 0.02 1.2 0.8 0.008 0.04 0.10 1.0 0.05 0 0 1.0 0.8 0.4 0.03 0.05 0 0.004 0 0.02 0.8 0.6 0.4 0.2 -0.20 -0.15 -0.06 -0.04 0 0.2 -0.4 -0.008 0 -0.20 -0.15 -0.2 -0.10 -0.05 -0.02 0.4 0.6 0 -0.004 0.01 -0.10 -0.05 0 -0.25 -0.08 -0.2 -0.8 -0.012 -0,01 -0.25 -0.4 real nominal nominal real government net lending/GDP government debt/GDP current account/GDP net foreign assets/GDP GDP nominal wage Figure 7a Authors’ calculations. Explanations see text. 63Using EDGE – A Dynamic General Equilibrium Model of the Euro Area An exploration of learning in the form of a 1 % -point reduction in the real rate of interest when the central bank and the private sector learn of the change linearly over 5 years Difference from baseline years years imports exports investment real wage nominal wage real product wage consumption CPI deflator GDP deflator 0.3 0.8 1.5 0.9 0.30 0.05 0.8 1.5 0.2 0.1 1.2 0.6 0.3 0.9 0 0.25 0.20 0.15 0 0.6 0.4 0.10 -0.05 1.2 0.9 0.3 0.6 -0.2 0 0.2 0.3 0 -0.6 0 -0.15 -0.2 -0.1 0 0.4 0.6 0.6 -0.3 0.05 -0.10 0 0.2 0 -0.3 -0.2 -0.3 -0.9 -0.05 -0.20 -0.4 -0.3 real nominal nominal real government net lending/GDP government debt/GDP current account/GDP net foreign assets/GDP GDP Main indicators; in % Wages; in % Interest rates; in %-points Debt and lending; in %-points Inflation; in %-points Unemployment; in %-points Exchange rates; in % Account and assets; in %-points 33669912 1215 15 Figure 7b Authors’ calculations. Explanations see text. the real economy falls back to the baseline path.The short-run oscillations are quite complex; financial, product and labour markets adjust at different rates, leading to an uneven path for inflation.Unemployment,government debt and net lending, net foreign assets and the current account all show reversals in their time paths. If we assume that the same exogenous process of linear learning also applies to the private sector then the results are both more substantial and different in character (Figure 7b).The initial effect is now positive as the private sector expands activity, assuming that the policy reaction implies a change in the inflation target (upward).The impact on inflation is much more substantial.Unemployment falls by 0.24 percentage points instead of rising by 0.35 points. Real wages converge to the steady-state from above and not below. Thus there are actually real gains to the economy from the slower learning,although inflation performance is worse. This is, however, a result of the particular simulation and other forms of slower learning would generate different results. The key feature is that if the central bank is slower at learning than the private sector then there are real costs, in part because the private sector confuses the slow learning with a policy change. If the central bank were very transparent then its thinking could be clear to the private sector and this error would not be made. An exploration of transparency in that sort of detail is however beyond the current analysis. 5. Implications The EDGE model that we have developed illustrates five important lessons for euro area monetary policy. –The first is the importance of credibility. If the private sector does not believe that the central bank will succeed in its actions then this forecast tends to be self-fulfilling as expectations of inflation do not change.In so far as the bank can achieve its policy aims this will be at much greater cost in terms of output and unemployment. The other side of this relationship is also worth recalling,as it implies that a credible central bank that responds as expected will have to do relatively little to achieve its policy objectives, as inflation expectations are not jolted by shocks. –Secondly,it shows the importance of transparency.If the private sector cannot readily detect when the central bank has reacted to new information it will tend to assume that it is the goals of policy that have changed. This also will add to the costs of monetary policy. –Thirdly the model suggests that the faster the private sector can learn the smaller the loss. –Fourthly, if the central bank fails to recognise a structural shift it can negate the benefits of that shift by applying an unchanged policy rule.The problem 64 Mika Kortelainen and David G. Mayes is worsened if the private sector notices the shift but the central bank does not. –Taking these together, if the central bank thinks that a structural shift may be taking place and wants to adjust policy in the light of that probability, it needs to make its actions explicit rather than trying to hedge its assessment in secrecy. This last circumstance is probably the most important for the euro area as a new system.Neither the Eurosystem nor any of the other participants in the economy have really good evidence on how it works. They form judgements based on past behaviour and their knowledge of the requirements of the future and update them in the light of experience. Mistakes in that process are inevitable but the way the process of learning is undertaken affects its cost. The simulations illustrate the importance of three key features of the model itself. In the first place it illustrates the importance of forward-looking behaviour with respect to wealth and the valuation of assets. Shocks affecting those values have substantial effects in the short run but adjustment processes can be very long lived, exceeding the 15 years illustrated in the Figures. Secondly behaviour differs markedly if a shock is perceived to be transitory rather than permanent and the impact is much more limited. This also has a clear implication for monetary policy. Monetary policy actions that are not expected to endure will be relatively ineffective. Thirdly it illustrates the key importance of building the reaction of both fiscal and monetary policy into the model. Private sector actions depend crucially on what they think the monetary and fiscal authorities will do in the future. The effectiveness of current policy depends on private sector expectations of future policy. Non-convergent rules will not be credible. The policy rules illustrated in the paper are just that,illustrations. They do not imply that authorities have to follow rigid rules but they illustrate the interaction between the behaviour patterns that the authorities have and the behaviour patterns the private sector thinks they have.In the circumstances we illustrate it is not normally beneficial for the authorities to disguise their intentions because the private sector is well aware of the incentives and the longer term consequences of nonsustainable actions. Particularly in the case of monetary policy, there are substantial payoffs to designing some sort of “precommitment technology” that allows the private sector to believe that the monetary authority will actually carry out the actions necessary to maintain its objective of price stability in the future. This is a young model,which will develop as we gain experience in using it.The calibration process is a continuing one as new evidence about parameter values and plausible properties appears. With the euro area only being in place for only two years the learning curve is likely to be steep for some time to come. 65Using EDGE – A Dynamic General Equilibrium Model of the Euro Area Appendix 1. List of equations Labour demand: LEL L Y TK tt tt t t =⋅ +⋅+⋅ ⎛ ⎝ ⎜⎞ ⎠ ⎟ +− 0 483 0 492 0 025 11 041 1 ... . .69 (A1) Capital stock: ΔΔ Δlog . log . logKEK EK tt ttt =− ⋅ + ⋅ + ++ 0 324 0986 21 (A2) 0 3378 0 0005 041 1 .log. . ⋅+⋅ ⋅ ⎛ ⎝ ⎜⎞ ⎠ ⎟−++ − ΔKPF PI Y K r t t t t t tχ () 001 1 . ++ ⎛ ⎝ ⎜⎞ ⎠ ⎟ r PI PC t t t χ Nominal wages: WN E WN WN Y N PF tt tt tt t =⋅ +⋅ +⋅ ⋅ ⋅⋅− +− 0 49 05 0 01 059 1 3 11 ..../() ()[]⋅−UU tt GDP deflator: () PP EP WN L tttt t t indirect t =⋅+⋅ +⋅ −⋅ −+ 0 495 0 485 0 02 1 11 .. . τ059.⋅ ⎛ ⎝ ⎜⎞ ⎠ ⎟ Yt (A4) Consumption: () CrEC A PC YDN PC t t ttt t t t t =⋅ ++⋅+⋅+⋅+ ⎛ ⎝ + − 06 1 10 015 1 1 1 .. χς ⎜⎞ ⎠ ⎟(A5) Windfall gain: ςπ t t t tt t t t A A YDN C PC A =−− −⋅ − −− − 11 1 1(A6) Wealth: ()AREA GDN NFA t t tttt =++ +++ +++ 1 1 400 111 /χ(A7) ()PF Y WN L PI K GOY GDN NFA tt tt t tttt ⋅− ⋅− ⋅ ⋅ − ⋅ + + − 001 033 1 .. Inventories: () KI TL K Y TL K tttttt =⋅⋅ −⋅−⋅ +088 05 059 041 059 041 .. .. .. (A8) ()0 494 11 059 1 041 ... ⋅−⋅ +++ EY TL K tttt 66 Mika Kortelainen and David G. Mayes (A3) Exports: log . log . log . logXYDD PX Pe tt t t tt =⋅ +⋅ −⋅ ⋅ ⎛ ⎝ ⎜ ∗∗ 048 072 041 ⎞ ⎠ ⎟+063.(A9) Imports: log . log . log .MDD PM P tt t t =⋅ −⋅⎛ ⎝ ⎜⎞ ⎠ ⎟−12 09 39 (A10) Export prices: () log . log . log .PX P P e tttt =⋅ +⋅ ⋅− ∗ 032 068 005 (A11) Import prices: () ( ) log . log . log . logPM PX P e PC e tttt tt =⋅ +⋅ ⋅+⋅ ⋅− ∗∗ 0 48 0 38 014 065. (A12) Consumer price deflator: log . log . log .PC P PM tt t =⋅ +⋅ +090 010 0 01 (A13) Investment deflator: log . log . log .PI PF PM tt t =⋅ +⋅ +0 85 015 010 (A14) Identities Private nominal disposable income: YDN YFN TAX INN TRF GOY NFN PI K ttttttt t t =−++−+−⋅⋅ − 001 1 .(A15) Real GDP: YCCGIXM KI tt ttt t t =+ ++− +Δ (A16) Capital accumulation equation: IK K tt t =− ⋅ − 099 1 .(A17) Indirect taxes: TIN YEN tt indirect t =⋅τ(A18) Direct taxes: TAX YEN tt direct t =⋅τ(A19) Public disposable income: GYN TAX TIN GOY TRF INN tttttt =++−− (A20) 67Using EDGE – A Dynamic General Equilibrium Model of the Euro Area Interest outlays of government: INN R GDN tt t =⋅ − / 400 1(A21) Net foreign assets: ()NFA NFA e e CA ttttt =⋅ + −−11 /(A22) Net factor income from abroad: NFN R NFA tt t =⋅ ∗ − / 400 1(A23) Current account: CA X PX M PM NFN tt tt t t =⋅ −⋅ + (A24) Public debt: GDN GDN GLN ttt =− −1(A25) Public net lending: GLN GCN GIN GYN tttt =− − + (A26) Domestic demand: DD C CG I KI tt tt t =+ ++Δ (A27) Nominal GDP as factor cost: YFN Y PF tt t =⋅ (A28) Nominal GDP: YEN Y P ttt =⋅ (A29) GDP deflator at factor price: () PF P tt t indirect =⋅−1τ(A30) Expected inflation rate, quarterly: πttt PC PC=− + log log 1(A31) Expected real interest rate: rR tt t =−/ 400 π(A32) Effective exchange rate (UIRP): () log log /eeRR tttt =+− + ∗ 1400 (A33) Unemployment rate: () UNLN tttt =−/(A34) 68 Mika Kortelainen and David G. Mayes Public nominal consumption: GCN CG P ttt =⋅ (A35) Public nominal investment: GIN IG PI ttt =⋅ (A36) Public other income: GOY YEN tt =⋅020.(A37) Public real consumption: CG Y tt =⋅γ(A38) Policy parameters Transfers: TRF YEN U tt t /..=⋅+025 020 (A39) Direct tax rate: () ()()Δτ ψ ψ π t direct tt tt t GDN YEN GLN YEN g=⋅ −−⋅ +⋅ + − 005 01 1 ./ ./ (A40) Inflation rate target: πt=0 0027. (A41) Taylor rule: () () RR rPCPC PC PC tt tt tt =−⋅ ⋅ ⋅+ ⋅ +⋅ − − 1 400 100 50 1 4 ΩΩ *log / log / ()[] () −−⋅ −⋅ ⋅ − ⎡ ⎣ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ 44 50 059 πt tt UU. (A42) 2. The steady state model Output: YTK L=041 059.. (B1) Capital stock: () () KPFPI Yr=⋅⋅++/./ .041 001χ(B2) Wages: WN PF Y L=⋅⋅059./ (B3) 69Using EDGE – A Dynamic General Equilibrium Model of the Euro Area Consumption: () () Crg rAYDN PC=++− ⋅ + ++ ⎛ ⎝ ⎜⎞ ⎠ ⎟⋅⋅+ − 1061 10 015 1 χ χ ../ (B4) Private wealth: () () A r g PF Y PI K GOY GDN NFA=+− ⋅ ⋅ ⋅− ⋅⋅− ⋅ + + − χ1041 001 033.. . (B5) Change in inventories: ΔKI g Y=⋅⋅088.(B6) Exports: () log . log . logXY CCGIKI=⋅ +⋅ +++ − ∗ 048 072 Δ(B7) () () 041 063.log / .⋅⋅+ ∗ PX P e Imports: ()() log . log . log / .MCCGIKI PMP=⋅ + ++ − ⋅ −12 064 39Δ(B8) Export prices: () log . log . log .PX P P e=⋅ +⋅ ⋅− ∗ 032 068 005 (B9) Import prices: () ( ) log . log . log . log .PM PX P e PC e=⋅ +⋅ ⋅+⋅ ⋅− ∗∗ 0 48 0 38 014 065 (B10) Consumer price deflator: log . log . log .PC P PM=⋅ +⋅ +090 010 0 01 (B11) Investment deflator: log . log . log .PI PF PM=⋅ +⋅ +0 85 015 010 (B12) Identities Employment: () LN U=⋅−1(B13) Technical progress: log log .TTg=+⋅ −1059 (B14) Public interest outlays: INN R GDN=⋅/ 400 (B15) 70 Mika Kortelainen and David G. Mayes References Amano, R., D. Coletti and T. Macklem (1999), Monetary Rules when Economic Behaviour Changes. In B. Hunt and A. Orr (eds.) (1999), 157–200. Blanchard, O. (1985), Debt Deficits and Finite Horizons. Journal of Political Economy 93 (2): 223–247. Calvo, G. (1983), Staggered Prices in a Utility-Maximizing Framework. Journal of Monetary Economics 12 (3): 383–398. Fagan, G., J. Henry and R. Mestre (2001), An Area-Wide Model (AWM) for the Euro Area. ECB Working Paper 42. ECB, Frakfurt. Hunt, B. (2000), Comments on “Actual and Perceived Monetary Policy Rules in a Dynamic General Equilibrium Model of the Euro Area”, Bank of Canada Workshop on “Advances in Econometric Model Building”, Ottawa, August 2000, printed in Kortelainen (2001). Hunt,B.and A. Orr (eds.) (1999),Monetary Policy Under Uncertainty.Reserve Bank of New Zealand, Wellington. Kortelainen, M. (2001), Actual and Perceived Monetary Policy Rules in a Dynamic General Equilibrium Model of the Euro Area. Bank of Finland Discussion Papers 3/2001. Bank of Finland, Helsinki. Kortelainen, M.(2002), EDGE:a Model of the Euro Area with Applications to Monetary Policy. Bank of Finland, Helsinki, Mimeo, forthcoming. Mattila, V.-M. (1998), Simulating the Effects of Imperfect Credibility: How Does the Peso Problem Affect the Real Economy. Bank of Finland Discussion Paper 24/98. Bank of Finland, Helsinki. Orphanides, A. (2000), Activist Stabilization Policy and Inflation: The Taylor Rule in the 1970s. Finance and Economics Discussion Series 2000-13. Board of Governors of the Federal Reserve System, Washington, DC. Ripatti, A. and J. Vilmunen (2001), Declining Labour Share – Evidence of a Change in Underlying Production Technology? Bank of Finland Discussion Papers 10/2001. Bank of Finland, Helsinki. Sefton,J.and J. in’t Veld (1999),Consumption and Wealth:An International Comparison. Manchester School of Economic Studies 67 (4): 525–544. Siegel,J.J.(1992),The Real Rate of Interest from 1800–1900:A Study of the US and the UK. Journal of Monetary Economics 29 (2): 227–252. Taylor,J.B. (1993), Discretion versus Policy Rules in Practice. Carnegie-Rochester Conference Series on Public Policy 39: 195–214. Tetlow, R., P. von zur Muehlen and F. Finan (1999), Learning and the Complexity of Monetary Policy Rules. In B. Hunt and A. Orr (eds.) (1999), 113–153. Vilmunen, J. (1998), Macroeconomic Effects of Looming Policy Shifts: Non-Falsified Expectations and Peso Problems. Bank of Finland Discussion Paper 13/98. Bank of Finland, Helsinki. 77Using EDGE – A Dynamic General Equilibrium Model of the Euro Area Wieland, V. (1998), Monetary Policy under Uncertainty about the Natural Unemployment Rate. Finance and Economics Discussion Series 1998-22. Board of Governors of the Federal Reserve System, Washington, DC. Willman,A.,M. Kortelainen,H.-L. Männistö and M. Tujula (1998),The BOF5 Macroeconomic Model of Finland. Structure and Equations. Bank of Finland Discussion Paper 10/98. Bank of Finland, Helsinki. 78 Mika Kortelainen and David G. Mayes David Rae and David Turner1 A Small Global Forecasting Model 1. Introduction This paper describes the OECD’s new small global forecasting model. The main focus of the model is the production of globally-consistent short-term forecasts of the major aggregates for the three main OECD economic regions: the United States, the euro area, and Japan. The rest of the world is modelled as a fourth composite region, albeit in a crude way. The key variables – which include output, inflation, the trade balance, and import prices – are driven by monetary and fiscal policy,exchange rates,and world demand.The projections from the model are used as a starting point to help animate the early stages of the OECD’s forecasting round. A particular focus of the model is the impact of global linkages and the transmission of influences between regions. Consequently, the three regional models are linked directly via trade, interest rates, and exchange rates. There are two additional linkages. First, output and inflation in the rest of the world depend on developments in the three main regions, and feed back on them through the trade equations. Second, commodity prices are endogenous and depend on world output and inflation. Both linkages provide important additional channels through which shocks are propagated across regions. The model is essentially a demand-side model. Output is based on an IS-style relationship,although this has been split into domestic demand and net export components rather than being modelled as a single reduced form equation. Potential output is assumed to be exogenous, and the model can therefore be written in terms of an output gap. In other words, the model explains why growth may differ from the potential growth rate but does not attempt to ex1The paper reflects helpful comments from numerous colleagues, including Laurence Boone, Thomas Dalsgaard, Jorgen Elmeskov, Michael Feiner, Pete Richardson, and Ignazio Visco. Special thanks to Laurence Le Fouler and Isabelle Wanner for their excellent research assistance;and also to Rosemary Chahed and Jan-Cathryn Davies for document preparation. – The paper was also published as OECD Working Paper ECO/WKP(2001)12. plain changes in potential growth. This approach seems to be a reasonable simplification given the model’s primary roles of short term forecasting and analysis of global linkages. However, it does mean that it has a limited ability to analyse the impact of supply side factors that may be expected to change potential output. Subject to the above constraints,the primary design criterion is that it be small in order to provide simple direct insights into specific forecast judgements on the basis of clear model properties.In addition,being small implies that few inputs or exogenous assumptions are required and makes it easier to decompose the influences behind the forecasts of each variable. In particular, the main equations have been solved out in terms of their explanatory variables so that the particular contributions to inflation or growth can be identified at any point in time. A further design criterion has been to ensure that extra relationships can be added without major re-estimation or re-coding.For example,a standard forecasting application would have exogenous exchange rates and short-term and long-term interest rates, and (mostly) backward-looking inflation expectations. However, monetary policy reaction functions can be added (as demonstrated later in the paper), along with alternative assumptions regarding the formation of expectations. This allows a little more economic richness to be temporarily added to the model when it is used for policy analyses, especially for those situations in which financial markets and expectations play important roles in the transmission of shocks within and between regions. Another feature is that it incorporates several concepts that provide a consistent point of contact between the model and the larger projection exercise. Demand and the composition of output are modelled relative to a specific (and exogenous) view about potential output,2inflation is modelled in a framework in which the output gap is important and real exchange rates and relative demand pressures play important roles. While very different in size and structure, the small model can be thought of as a simplified version of INTERLINK’s demand side. One difference, however, is that this model is based on quarterly data whereas INTERLINK and the forecasting round use semi-annual data. In this respect, the model is able to take better account of short-term developments in key variables. As with any model, there is a trade-off between the goodness-of-fit of individual equations and the model having properties that conform to priors about macroeconomics. The equations reported in this paper reflect a compromise between these choices.The estimation philosophy is that differences in equati80 David Rae and David Turner 2Potential output is estimated in a consistent way across countries, based on a production function approach; see Giorno et al. 1995 for details. on specifications across regions appear only where there is a clear economic rationale.For example,the empirical importance of stock market wealth to the United States economy has led to the inclusion of such a variable in the United States domestic demand equation.3Otherwise, coefficient values are relatively freely estimated. However, a few coefficients have been restricted, particularly where they were poorly determined, in order to deliver properties closer to our priors or to be more consistent with results for other regions. Various homogeneity and global closure restrictions are also imposed to ensure that the model settles down to a sensible steady-state path. Goodness-of-fit is an important criterion for a forecasting model but particular weight has been given to accuracy in recent years, partly because the short-run dynamics of highly-reduced form equations may not remain stable over long spans of history. 2. Overview of the model Each of the three OECD regions (the United States,the euro area, and Japan) consists of four main blocks:4 –Output is determined through an IS-style relationship, although domestic demand and net exports are modelled separately, partly for econometric reasons,and partly to emphasise the model’s role of capturing international linkages. Potential output growth and fiscal policy are exogenous. –Inflation. The main inflation variable is core CPI inflation, which is modelled using a Phillips curve. Inflation therefore depends on the output gap and various components of imported inflation. Headline CPI inflation depends on core inflation plus a wedge that is determined by commodity and oil price inflation. –Import Prices. Manufacturing and service import prices are modelled, and depend on foreign and domestic consumer prices, the exchange rate, and commodity prices. –Financial Variables. For forecasting purposes, short-term and long-term interest rates and nominal exchange rates are exogenous. For simulations, short-term interest rates can be determined by forward-looking monetary policy rules in which short-term interest rates depend on the output gap and the expected future core inflation rate (relative to an exogenous target rate). Bond rates will then depend on expected future short-term rates. Exchange rates can be endogenised using a (risk-adjusted) uncovered interest parity condition. 81A Small Global Forecasting Model 3Such effects are less easy to identify in other regions. 4Recent examples of this style of model include Duguay 1994; Bharucha, Kent 1998; Ball 1998; Hargreaves 1999; Beechey et al. 2000. A separate block covers the rest of the world: –Commodity Prices are modelled explicitly in order to capture an important mechanism through which global demand shocks can have inflationary consequences and be propagated between regions. Oil prices are exogenous. –Output and Inflation. The output gap and inflation in the rest of the world are linked to output and inflation in the three main regions, and therefore provide other feedback channels for the main regions. Key features and estimation results for each block are discussed in more detail below. 2.1 Inflation The main inflation variable is core CPI inflation, defined as the CPI excluding food and energy.This was chosen because it is possibly the best single measure of the general inflationary pressures that monetary policy is concerned with. The exclusion of energy prices is also useful when monetary policy reaction functions are added the model, to ensure for example that policy does not react to (i.e., “lets through”) the direct effects of an oil shock while reacting to second-round effects, such as the shock feeding through to inflation expectations. Core inflation is determined by a Phillips curve, where the explanatory variables are a pressure of demand term, in the form of the output gap, and supply shocks in the form of various components of import prices: () () ()( ) ππ α ωπ π ωπ νπ γπ=+ + − + + + emmcomo ygap lags lags lags lags 0ΔΔ Δ () il ,(1) where πis core CPI inflation, πeis expected inflation, and ygap is the output gap.5The remaining terms capture import prices, where the inflation rate of import prices of manufactures and services (πm) is separated out from commodity price inflation (πcom) and oil price inflation (πoil). All import prices are measured in local currency terms, and the lags are designed to capture slow passthrough. Non-commodity, non-oil import prices are weighted by the degree of openness of the economy, ω, which is measured by the share of these imports in total value added. This measure of openness has risen over time in all the regions to reach its current level of 11 per cent for the United States, 12 per cent in the euro area, and 6 per cent for Japan. Commodity price inflation is weighted by the share of manufacturing in GDP (ν) while oil price inflation is weighted by an index of oil intensity in production (γ) which has halved since the early 1970s in all regions.All these weights are exogenous in model simulations. 82 David Rae and David Turner 5The lags function is shorthand for a general distributed lag, which may include current-dated values. Detailed data definitions and sources are given in an Annex. 83 Core CPI inflation Dependent Variable: Δπ =Δ(100 Δlog core-CPI) United States Euro Area Japan Lagged dependent variables Lag 1 –0.551*** –0.473*** –0.636*** Lag 2 –0.366*** –0.168** –0.251*** Lag 4 –0.145** Gaps Output Gap 0.045*** 0.055*** Output Gap-1 0.075a** ΔDomestic demand-1 0.107** Import prices1 () ωπ π −− − 11 m0.501* 0.518*** 0.302b Δπm0.242 0.515*** Δπm−20.515*** Commodity Prices2 Average, lags 1–8 0.411*** 0.255** Average, lags 1–4 0.117*** Oil Prices3 Full sample 0.149*** pre-1980 0.141*** 0.040** post-1980 0.066 0.029 Sacrifice Ratio 2.9 1.9 1.6 (6.3)c Dummies 93q1 97q2 Estimation period 63:2 – 00:4 74:2 – 00:1 71:2 – 00:1 Standard error 0.23 % 0.16 % 0.46 % Authors’c calculations.Data definitions are in an Annex.- One,two,and three stars denote significance at 10,5,and 1 percent levels.– 1Manufactures and services.– 2Calculated asνπΔcom −1where ν is the weight of manufacturing in OECD value added.– 3Calculated asγπΔoil −1,where γis a measure of intensity of oil use in production. The lag structure is as follows: US: Average of lags 1–3; Euro Area: lag 1; Japan: lags 0 and 1. – aIf gap is negative and inflation is less than1%p.a.then coefficient is one-quarter of the reported value. – bThree quarter lag. – cThe sacrifice ratio in brackets is for the “flat” portion of the Phillips curve (i.e., low and falling inflation). Recent residuals (positive value means under-prediction) 1997 Q1 0.12 0.00 –0.14 Q2 0.10 –0.06 0.00 Q3 –0.12 0.01 0.00 Q4 0.10 0.07 –0.22 1998 Q1 0.08 0.12 –0.37 Q2 0.19 0.23 –0.11 Q3 0.04 0.12 –0.17 Q4 –0.04 0.13 0.77 1999 Q1 –0.18 –0.18 0.30 Q2 0.05 0.14 –0.01 Q3 –0.13 0.02 –0.16 Q4 –0.11 0.00 0.29 2000 Q1 –0.19 –0.09 0.35 Authors’ calculations. Table 1 A Small Global Forecasting Model For estimation purposes and in the standard version of the model, the coefficients on the lagged inflation terms are assumed to sum to unity, proxying the combined effects of nominal inertia and (backward-looking) inflation expectations. In that case the Phillips curve can be written in terms of the change in the inflation rate (Δπ). Alternative specifications could include a weighted average of forward-looking and backward-looking expectations. Each of the three regions also has a simple equation linking core CPI to headline CPI: () () ππ π π head com oil lags lags=+ + .(2) Here the headline CPI inflation rate (πhead) is built up from the core rate by adding the direct (or accounting) impact of oil and commodity prices. With 84 Impacts on headline inflation (single equation properties) Deviations from baseline (annual inflation rate, percentage points) 0 0 0 0 12345 Temporary rise in domestic demand Rise in import prices (1 percent increase for two years) (10 percent increase in all import prices, incl. oil and commodities) 0,2 0,5 0,5 0,2 0,4 1,0 1,0 0,4 0,6 1,5 1,5 0,6 0,8 0,8 1,2 2,0 2,0 1,2 1,0 1,0 1,4 2,5 2,5 1,4 Years Euro Area Euro Area US US Japan Japan Figure 1 Authors’ calculations. Explanations see text. David Rae and David Turner this formulation it is possible to distinguish the direct (or accounting) effects of oil and commodity price shocks from the indirect or second-round effects, in which they may get built into the general inflation process. The estimation results are shown in Tables 1 and 2, and the single equation dynamic properties are presented in Figure 1.Detailed data definitions are given in an Annex. The output gap is strongly significant for each region, although the gap appears in a non-linear form in the Japanese equation. Japan’s equation has a goal-line effect,in which it is difficult to drive inflation lower when it is already very low.6Specifically, when inflation is below 1 per cent per annum, a negative output gap will only reduce inflation by 1 quarter of the amount that it would otherwise. This feature is important when trying to explain Japanese inflation over the past few years. A similar effect was tested for in the United States and euro equations but was not found to be empirically important, possibly because those two regions have not had Japan’s experience of a prolonged period of low inflation. Although measured with some uncertainty, the sacrifice ratios are broadly consistent with those found in other Phillips curve work, including Richardson et al. (2000) and Turner/Seghezza (1999). Manufacturing and services import prices are statistically significant for each region, and their impact on inflation is quite large. The (πm–π) term ensures that manufacturing and service import prices are eventually fully passed on to 85 Headline CPI inflation Dependent Variable: πh= 100 Δlog CPI United States Euro Area Japan Core Inflation 0.964*** 0.976*** 0.964*** Commodity Price Inflation Lag 0 0.010** Lag 1 0.019*** Lag 2 0.010*** Lag 3 0.013*** Oil Price Inflation Lag 0 0.016*** 0.006** 0.006*** Lag 1 0.005** Lag 2 0.011*** Estimation period 63:2 - 00:1 80:1 - 00:1 70:4 - 00:1 Standard error 0.27 % 0.20 % 0.30 % Authors’ calculations. Data definitions are in an Annex. – One, two, and three stars denote significance at 10, 5, and 1 percent levels. – Coefficients restricted to sum to one. Table 2 A Small Global Forecasting Model 6The analogy comes from American football, where the closer to the goal-line you are, the harder it is to gain extra yardage. The cut-off of 1 per cent per annum is fairly arbitrary but was chosen after experimenting with several values. consumer prices,7but the adjustment is significantly slower in Japan than in the other two regions.That is partly because the import price coefficient is estimated to be lower, and partly because Japan’s economy is more closed in the 86 Impacts on inflation of a 50 % rise in oil prices (single equation properties) Deviations from baseline (annual inflation rate, percentage points) -0,2 -0,2 -0,2 -0,2 -0,2 -0,2 United States Euro Area Japan 0 0 0 0 0 0 0,2 0,2 0,2 0,2 0,2 0,2 0,4 0,4 0,4 0,4 0,4 0,4 0,6 0,6 0,6 0,6 0,6 0,6 0,8 0,8 0,8 0,8 0,8 0,8 1,0 1,0 1,0 1,0 1,0 1,0 Years Headline Headline Headline Core Core Core 123 Figure 2 Authors’ calculations. Explanations see text. David Rae and David Turner 7The model-wide implications of this term depend on the assumed exchange rate regime.Under fixed exchange rates, this term forces all countries to have the same steady state inflation rate (in order that real exchange rates are untrended). However, under floating rates each country can have its own inflation rate with the nominal exchange rate moving according to the inflation differentials between countries. –Finally,two specification checks were performed.First,export growth (relative to potential) was added to each equation.This “trade multiplier” effect was added because it is possible that domestic demand will respond to an increase in the volume of trade even if net exports remain unchanged.However, this additional multiplier was insignificant in each region suggesting that the net export formulation is a useful simplification. Second,richer dynamic adjustment from potential output to actual output was tested by adding lags of potential growth. However, they were insignificant in each region. Net exports Net exports as a proportion of potential output are explained by the real effective exchange rate, the local domestic demand gap, and the trade-weighted 93 Impacts on net exports (s i ngle equat i on propert i es) Deviations from baseline (percentage points of potential GDP) 0 0 0 0 12345 Depreciation of real exchange rate Increase in foreign demand (permanent 10 percent depreciation) (1 pp increase in foreign domestic demand gap) 0,2 0,2 0,1 0,1 0,4 0,4 0,6 0,6 0,8 0,8 0,2 0,2 1,0 1,0 0,3 0,3 Y ea r s Euro Area Euro Area US US Japan Japan Figure 4 Authors’ calculations. Explanations see text. A Small Global Forecasting Model foreign domestic demand gap. Although differences in openness or import propensities may lead to the coefficients being different on local and foreign domestic demand gaps, in practice the restriction that the coefficients are equal but opposite in sign was accepted for each region.In that case, the equation can be rewritten in terms of the relative gap: 94 Net exports Dependent variable: Δ(net exports/potential) .100 United States Euro Area Japan Δlog Real Exchange Rate Lag 0 –1.157** –1.856*** –1.237*** Lag 1 –1.157** –1.237*** Average, lags 2–8 –3.652*** –5.445*** Average, lags 1–8 –5.202*** Relative Domestic Demand Gap1 ΔRelative gap 0.130*** 0.258*** 0.135*** ΔRelative gap-1 0.076*** Dummies 78q1 Estimation period 74:4 – 00:1 74:3 – 00:1 74:3 – 00:1 Standard error 0.19 0.18 0.30 Authors’ calculations. – Data definitions are in an Annex. – One, two, and three stars denote significance at 10, 5, and 1 percent levels. – 1Foreign minus domestic gap. Recent residuals (positive value means under-prediction) 1997 Q1 –0.2 –0.2 0.1 Q2 0.0 0.2 0.0 Q3 –0.1 0.1 –0.2 Q4 0.1 –0.1 0.1 1998 Q1 –0.1 0.0 –0.4 Q2 –0.2 0.1 0.1 Q3 0.1 0.0 –0.3 Q4 0.3 –0.5 –0.3 1999 Q1 –0.4 –0.1 0.1 Q2 –0.3 0.0 0.1 Q3 0.0 0.4 0.2 Q4 0.2 0.0 –0.5 2000 Q1 –0.2 0.0 1.1 Authors’ calculations. Table 4 David Rae and David Turner xmgap = lags ( xmgap) + lags ( rer) + lags ( relgap) (7) where rer is the log real effective exchange rate (based on relative CPIs) and relgap is the relative domestic demand gap (foreign minus domestic). The estimation results are in Table 4 and Figure 4.The real exchange rate has a strong and significant impact on net exports in each region.The lag structure is quite long in each case, up to two years, implying that a sustained real exchange rate change is more important than an equal-sized short-term blip.The long-term elasticities are fairly similar across regions: a ten per cent depreciation of the real exchange rate will raise net exports as a share of GDP by 0.6 per cent in the United States, 0.7 per cent in the euro area, and 0.8 per cent in Japan. The size and timing of these responses is consistent with a short-run J-curve effect and with other evidence, including from interlink. 2.3 Import prices Manufacturing and service import prices are assumed to be a weighted average of price-taking and price-making behaviour. For price takers, the import price is simply equal to foreign prices divided by the nominal exchange rate. For price makers, the import price is determined by the local price of competing goods, which is proxied by the domestic CPI. With this formulation, long run import prices can be written as a function of the real exchange rate, commodity prices, and domestic consumer prices. An error-correction equation is used to determine short run import prices. A time trend is also included to capture the long term decline in import prices relative to consumer prices. Unfortunately there is no suitable time-series for euro area import prices that excludes intra-euro-area trade. This is particularly a problem when trying to estimate a real exchange rate elasticity because the trade-weighted real exchange rate excludes intra-euro currencies. Consequently, the coefficients of the euro equation have been imposed at values similar to the United States-Japan average but making adjustments in order to improve recent forecasting performance.14 Results are shown in Table 5. The relative weight on domestic prices versus foreign prices implies that Japan is significantly more of a price-taker than is the United States.15 Speed of adjustment to long-run equilibrium is reasonably fast, and around half of the exchange rate impact comes through in the first quarter. Commodity prices have a significant impact on non-commodity 95A Small Global Forecasting Model 14 Despite being imposed,the residuals from the long-run part of the equation are stationary implying that the long run represents a valid cointegrating relationship. 15 The long-run equation can be rewritten so that relative import prices (PM/CPI) are a function of the real exchange rate. In that case, the real exchange rate elasticity is –0.45 for the United States and –0.74 for Japan. 96 Import prices Dependent variable: p = log (PM/CPI), PM is manufactures and services import prices Implicit Long Run US log PM = const + 0.547 log CPI + 0.453 log pf/e - 0.008 trend1 Euro log PM = const + 0.4 log CPI + 0.6 log pf/e - 0.0057 trend Japan log PM = const + 0.260 log CPI + 0.740 log pf/e - 0.00083 trend2 Short Run United States Euro Area3Japan Equilibrium Correction p-1 –0.163*** –0.2 –0.254*** rer-1 (= log p.e/pt) –0.074*** –0.12 –0.187*** Lagged dependents Lag 1 0.139** 0.1 Lag 4 ΔReal exchange rate Lag 0 –0.267*** –0.2 –0.483*** Lag 4 –0.158*** Commodity Prices Δlog (pcom/cpi) 0.07 0.131*** Δlog (pcom/cpi)-2 0.052*** Trend/100 –0.130*** –0.114*** –0.207*** Dummies 87q1 Estimation period 77:1 – 00:1 (imposed) 80:1 – 00:1 Standard error 0.67 % 0.82 %41.50 % Authors’ calculations. Data definitions are in an Annex. – One, two, and three stars denote significance at 10, 5, and 1 percent levels. – 1Trend applies after 1980 only. – 2Trend applies before 1994 only.– 3Coefficients imposed.See text for a discussion.– 4Standard error and residuals (below) for euro based on implicit residuals from imposed equation, 1980 – 2000. Recent residuals (positive value means under-prediction), in % 1997 Q1 0.0 –1.2 –1.9 Q2 –0.3 0.0 –0.2 Q3 –0.2 –0.6 1.8 Q4 0.2 –0.2 0.5 1998 Q1 0.2 0.3 3.6 Q2 0.2 –0.6 –0.1 Q3 0.6 0.5 0.1 Q4 –0.2 –0.1 –1.5 1999 Q1 0.4 –0.8 –1.0 Q2 0.2 –1.2 –0.7 Q3 0.3 –0.9 0.1 Q4 –0.2 0.4 –0.2 2000 Q1 0.9 –1.5 –2.1 Authors’ calculations. Table 5 David Rae and David Turner import prices,implying effects that work through the production chain,but the effect is only temporary. Oil prices were not significant. 2.4 Commodity prices As mentioned earlier, commodity prices can be an important channel through which global demand shocks are magnified and propagated across regions. Commodity prices tend to be much more volatile than prices for final output, but have an asymmetry in their behaviour. Price rises tend to be large and quick, while price declines tend to be milder but to last longer. In addition, there is a strong commodity price cycle,and this cycle is highly correlated with the world demand cycle. Commodity prices are determined by a complex interaction of supply and demand factors,but for the purposes of this model the key features can be simplified and modelled as follows. It is assumed that the inflation rate of non-oil commodity prices measured in US Dollars (πcom) depends on world inflation and the world output gap. In the absence of shocks and with the world economy growing at potential, commodity price inflation will settle down to the world inflation rate (after adjusting for a constant “drift” term that captures the trend decline in real commodity prices). Modelling πcom as a function of the world output gap implies that if the world economy is growing at its potential rate then there will be no excess demand and no pressure on manufacturing capacity,and consequently no pressure on real commodity prices (whether they be commodities that are used as inputs to the production process,or commodities for final consumption). Several forms of asymmetry were tested in estimation to capture the apparent asymmetry in the commodity price cycle, including distinguishing between rises and falls in inflation, between positive and negative output gaps, and between positive and negative changes in the gap. There was little strong evidence to help choose between the alternatives but the following equation was chosen as the simplest econometrically sound equation that captures the key features: () ππ ππ com oecd com oecd wldgap−= + − + − 0 0082 0 467 418 10 1 .. . /Δ0+(8) (3.9) (7.1) (4.2) +− +++ −− 0577 0 409 0 358 1635 12 ....ΔΔΔΔ −3 ππππ com com c oecd 3 207 3925 12 ..ΔΔππ −− + oecd oecd (7.2) (5.8) (4.8) (1.9) (3.9) (4.8) Estimation period: 1970-2 – 1999-4. t-values in brackets. R2= 0.79. Std. Error = 2.4%; DW = 1.9; AR(4) p-value = 0.59; Jarque-Bera Normality p-value = 0.99 The single equation properties are shown in Figure 5.Changes in world output are estimated to have a large and statistically significant impact on commodity 97A Small Global Forecasting Model price inflation but only if world output growth exceeds potential (as denoted by the Δwldgap/100 +term). The OECD output gap was tested as an alternative measure of global excess demand but performed less well, the difference being most important during the recent Asian crisis. The OECD inflation rate excluding high inflation countries (πoecd) is used to proxy world inflation, and the large coefficients imply a substantial degree of overshooting of commodity prices relative to consumer prices. Finally, the constant term implies that real commodity prices will fall by approximately 3 ½ per cent per annum, ceteris paribus. 98 Impacts on commod i ty pr i ce i n f lat i on (s i ngle equat i on propert i es) Deviations from baseline (annual inflation rate, percentage points) -4 -4 0 0 12345 Rise in World demand Rise in OECD inflation (Permanent 1 percent increase) (1 percent p.a. increase) -2 -2 1 1 2 2 0 0 4 4 2 2 8 8 6 6 3 3 5 5 4 4 6 6 10 10 7 7 Y ea r s Figure 5 Authors’ calculations. Explanations see text. David Rae and David Turner 2.5 Other variables Monetary policy rules The large recent literature on modelling monetary policy in small models of this type has two branches: optimal rules; and the relative performance of simple “rules of thumb” or interest rate reaction functions. Examples include Drew/Hunt (1998), Ball (1998), Svensson (1998), Fair (2000), Rudebusch (1999), and Smets (1998).Looking at optimal policy is certainly feasible in the context of this model, but beyond the scope of this paper. Instead, some preliminary experiments have been performed using simple interest rate reaction functions. A general précis of that branch of the literature suggest that policy rules work better if:(a) interest rates respond to expected future inflation,rather than current inflation; (b) the current output gap is included; and (c) the weights are higher than the simple Taylor rule. Experiments have been made with the following rule: ()r r ygap t e =+ + − + **αβΠΠ 6(9) where r=i–Πis the real short term interest rate, r* is the equilibrium real interest rate, Πis the annual rate of core inflation, and Πeand Π*are the expected and target annual inflation rates respectively. In this way, interest rates are increased if the output gap is currently above zero or if the expected core inflation rate in eighteen months time is above its target level. A rule that looks ahead eighteen months was chosen partly because the 18-months to 2-year period is typically regarded as the period over which monetary policy has its greatest influence. In addition, it ensures that policy does not react to short-term blips in inflation (i.e., those in the next 1 or 2 quarters) unless they lead to longer lasting inflationary pressures. The weights chosen for the version of the model discussed here are α= 0.75 and β= 1.0,although they are experimental and may be revised after further research. For comparison, the “standard” Taylor rule depends on the output gap and current inflation, with weights of 0.5 on each variable.16 This rule is not intended to mimic actual central bank behaviour, but to approximate an optimal policy rule in the context of this model. Long-term interest rates Long-term interest rates feed into domestic demand and can be modelled using an approximation to the expectations theory of the term structure. The bond rate (iB) is assumed to be a weighted average of all future short rates (iS) where the weights decline geometrically in the future: 99A Small Global Forecasting Model 16 Attempts to estimate the coefficients of the policy rule for the United States include Judd/Rudebusch (1998) and Clarida et al. (1998). Both papers include an interest rate smoothing term in order to fit historical policy.Ball (1997) and Levin (1996) show that higher weights than implicit in the Taylor rule are more successful at stabilising output and inflation in a small model of this sort (Ball) and in the Fed’s FRB model (Levin). () ()iEit t Bi iti S =− + = ∞ + ∑ 1 0 λλ ϕ|(10) where λdetermines the speed with which the weights decline and ϕis an exogenous term or liquidity premium to capture the fact that the yield curve slopes up on average. With this formulation, bond rates are purely forward looking but put more weight on the near future than would be the case under the pure expectations theory.In contrast,the pure theory gives equal weight to next quarter’s 90-day bill rate as it does to the 90-day rate in each of the next 39 quarters but zero weight to anything after 40 quarters. Aside from (10) being more a plausible guide to investment decisions in the domestic demand equation, it greatly simplifies the model solution.It can be rewritten by taking a Koyck lead: () () () iEii t B t B t S =− + + + 11 λλϕ (11) so that today’s bond rate is a weighted average of next period’s expected bond rate and the current short-term rate. The parameter λis set to 0.9 to give a mean lead of 2 ½ years between shortand long-rates, which is approximately consistent with the observed relative volatility of bonds and bills.17 The term premium (ϕ) has been set to 1 per cent. Nominal exchange rates Exchange rates against the United States Dollar (USD) (et) can be endogenised using uncovered interest rate parity (UIP). () () log log /eE e ii tt f ≈+−− +1400η(12) where iand ifare the domestic and trade-weighted foreign short term interest rates respectively, and is an exogenous risk premium. Since UIP assumes perfect capital markets, equilibrium requires that real interest rates in all regions be equal in the long run,adjusted for a risk premium. Under UIP the exchange rate is a jumping variable.A possible alternative that would reduce the degree of jumping is to model the exchange rate as a weighted average of the current rate and the UIP rate. Rest of the world output gap The output gap in the rest-of-the-world (i.e., the world minus the three major regions) is determined by a simple form of trade multiplier equation, but with an error-correction term to ensure that the rest-of-the-world gap returns to zero. The lag structure has been determined empirically: 100 David Rae and David Turner 17 Bond rates appear to move “too much”to be consistent with the pure expectations theory,but the formulation used in the model will mimic the observed “excess”volatility because it puts more weight on the near future. ΔΔΔrowgap rowgap rowgap rowgap=+ + −−− 0 055 0 300 0198 11 .. . 2(13) (2.9) (3.2) (2.4) ()0114 0 0 047 1 ..Δ .132Δ Δusygap usygap eurygap japygap++ + −−2 (2.5) (2.5) (1.3) Estimation period: 1974-3–1999-3; t-values in brackets; R2= 0.49; Std. Error = 0.27; DW = 2.0 The coefficient estimates imply that output in the United States has the largest impact on demand in the rest-of-the-world,followed by the euro area and then Japan. The World Output Gap is then an accounting identity: worldgap = 0.26 usgap + 0.21 eurgap + 0.19 japgap + 0.34 rowgap (14) 3. Simulation properties This section describes some simulations in order to demonstrate the major properties of the model. The simulations have been chosen specifically to emphasise the nature and size of the international linkages in the model. A range of “standard multiplier” shock results are reported in the Annex. 3.1 The inflationary consequences of a global boom The first simulation illustrates the role of commodity prices in propagating a global demand shock. World domestic demand is assumed to be two per cent above baseline for two years. It is assumed that monetary policy does not react so that the demand and commodity price channels can be separated from the monetary policy channel as influences on global inflation.18 The impacts are summarised in Figure 6. The thin line shows the impacts with commodity prices held at their baseline level,while the thick line assumes that commodity prices move according to equation (8). Even with commodity prices held fixed at their baseline level, the demand shock has relatively large impacts on inflation. For example, United States headline inflation is 1.4 per cent higher after two years,and euro area inflation rises even higher but at a slightly slower pace. However, with endogenous commodity prices the increased demand pressure leads to a 20 per cent rise in commodity prices which pushes United States and euro inflation more than 2 per cent above baseline. Overall, the commodity price channel adds around half as much again to the inflationary consequences of a demand shock (slightly less than half in the euro area). The key to this result is that the demand 101A Small Global Forecasting Model 18 More precisely, output in each region jumps by two per cent in the first quarter, stays at that level until quarter 8, then returns to baseline. Nominal interest rates and exchange rates are unchanged. 102 World demand shock – impacts on headline inflation Annual rate, deviations from baseline in percent 0 0 0 0 0 0 0 0 12345 United States Euro Area Japan OECD 0,5 0,5 0,5 0,5 0,5 0,5 0,5 0,5 1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,0 1,5 1,5 1,5 1,5 1,5 1,5 1,5 1,5 2,0 2,0 2,0 2,0 2,0 2,0 2,0 2,0 2,5 2,5 2,5 2,5 2,5 2,5 2,5 2,5 Years With commodity prices adjusting With commodity prices adjusting With commodity prices adjusting With commodity prices adjusting Commodity prices unchanged Commodity prices unchanged Commodity prices unchanged Commodity prices unchanged Figure 6 Authors’ calculations. Explanations see text. David Rae and David Turner 109 Impact of a rise in government spending Permanent increase of 1 percent of potential output; deviations from baseline in percent Quarters after shock 123481216 Rise in spending in the United States United States Output gap 0.8 0.9 0.7 0.4 0.2 0.0 –0.1 Inflation10.0 0.1 0.2 0.3 0.4 0.3 0.3 Net exports2–0.1 –0.2 –0.2 –0.1 –0.1 –0.1 –0.1 Euro Area Output gap 0.0 0.1 0.1 0.1 0.0 0.0 0.0 Inflation10.0 0.0 0.1 0.1 0.1 0.0 0.0 Net exports20.0 0.1 0.1 0.1 0.0 0.0 0.0 Japan Output gap 0.0 0.1 0.1 0.1 0.0 0.0 0.0 Inflation10.0 0.0 0.1 0.1 0.1 0.0 0.0 Net exports20.0 0.1 0.1 0.0 0.0 0.0 0.0 Rise in spending in the Euro Area United States Output gap 0.0 0.1 0.1 0.1 0.0 0.0 0.0 Inflation10.0 0.0 0.1 0.1 0.1 0.1 0.0 Net exports20.0 0.1 0.1 0.1 0.0 0.0 0.0 Euro Area Output gap 1.2 0.7 0.9 0.8 0.1 –0.1 –0.1 Inflation10.1 0.2 0.3 0.5 0.6 0.5 0.4 Net exports2–0.4 –0.2 –0.3 –0.2 –0.1 –0.1 –0.1 Japan Output gap 0.1 0.0 0.1 0.1 0.0 0.0 0.0 Inflation10.0 0.0 0.1 0.1 0.1 0.0 0.0 Net exports20.1 0.0 0.0 0.0 0.0 0.0 0.0 Rise in spending in Japan United States Output gap 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Inflation10.0 0.0 0.1 0.1 0.1 0.0 0.0 Net exports20.0 0.0 0.0 0.0 0.0 0.0 0.0 Euro Area Output gap 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Inflation10.0 0.0 0.0 0.1 0.1 0.0 0.0 Net exports20.0 0.0 0.0 0.0 0.0 0.0 0.0 Japan Output gap 0.6 1.3 0.4 0.5 0.3 0.1 0.0 Inflation10.0 0.1 0.3 0.3 0.3 0.2 0.2 Net exports2–0.1 –0.2 –0.1 –0.1 –0.1 –0.1 –0.1 Authors’ calculations. – 1Annual rate of headline inflation. – 2As a percent of potential output. Table 7 A Small Global Forecasting Model 110 Impact of an exchange rate depreciation Permanent ten percent nominal depreciation; deviations from baseline in percent Quarters after shock 123481216 Depreciation of the United States Dollar United States Output gap 0.1 0.2 0.3 0.4 0.5 0.5 0.4 Inflation10.1 0.2 0.4 0.5 0.8 1.0 1.1 Net exports20.1 0.2 0.3 0.3 0.5 0.5 0.4 Euro Area Output gap 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Inflation1–0.1 –0.3 –0.4 –0.6 0.0 0.0 –0.1 Net exports20.0 0.0 0.0 0.0 0.0 0.0 0.0 Japan Output gap 0.0 –0.1 –0.1 –0.1 –0.2 –0.2 –0.1 Inflation1–0.2 –0.4 –0.6 –0.7 –0.2 –0.1 –0.1 Net exports20.0 –0.1 –0.1 –0.1 –0.1 –0.1 –0.1 Depreciation of the Euro United States Output gap 0.0 0.0 0.0 0.0 –0.1 –0.1 0.0 Inflation10.0 0.0 0.0 0.0 0.1 0.0 –0.1 Net exports20.0 0.0 0.0 0.0 0.0 0.0 0.0 Euro Area Output gap 0.2 0.3 0.3 0.4 0.6 0.5 0.4 Inflation10.1 0.4 0.7 1.2 1.2 1.4 1.5 Net exports20.2 0.2 0.3 0.3 0.5 0.5 0.4 Japan Output gap 0.0 0.0 –0.1 –0.1 –0.1 –0.1 –0.1 Inflation10.0 0.0 0.0 0.0 0.1 0.0 –0.1 Net exports20.0 0.0 0.0 –0.1 –0.1 –0.1 –0.1 Depreciation of the Yen United States Output gap 0.0 0.0 0.0 0.0 –0.1 0.0 0.0 Inflation10.0 0.0 0.0 0.0 0.0 –0.1 –0.1 Net exports20.0 0.0 0.0 0.0 –0.1 0.0 0.0 Euro Area Output gap 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Inflation10.0 0.0 0.0 0.0 0.0 –0.1 –0.1 Net exports20.0 0.0 0.0 0.0 0.0 0.0 0.0 Japan Output gap 0.1 0.2 0.3 0.5 0.8 0.9 0.7 Inflation10.3 0.5 0.9 1.0 0.6 0.6 0.7 Net exports20.1 0.2 0.3 0.4 0.6 0.7 0.6 Authors’ calculations. – 1Annual rate of headline inflation. – 2As a percent of potential output. Table 8 David Rae and David Turner Appendix – Data Definitions All data is quarterly,seasonally adjusted, and is based on the OECD’s Analytic Database (ADB). The euro area volume aggregates are calculated by aggregating the growth rates of individual countries, weighted by the previous period’s share of nominal GDP at current exchange rates. Euro area financial aggregates are weighted averages of the levels of country variables, using the same GDP weights as above. A detailed description of ADB data can be found on the OECD’s web site at www.oecd.org/eco(data/eoinv.pdf. A description of variables by category is given below (ADB mnemonics are in capitals). 111 Inflation, prices, exchange rates Core inflation () π=100. logΔcoreCPI Core CPI CPI excluding food and energy. Source: Main Economic Indicators (MEI) Headline CPI CPI all items, seasonally adjusted. Source: MEI. Euro area CPI is the harmonised CPI index from 1995, non-harmonised before that. Import prices πimp = rate of change of implicit deflator for imports of manufactures and services, where the deflator has been detrended from 1980 using a time trend. Weight on import prices ω= (MM+MSR)/(GDP+MM+MSR) where MM and MSR are imports of manufactures and services respectively. Commodity prices WPHD: primary commodities excl. energy, world price, HWWA-index US-$ Weight on commodity prices Weight of manufacturing output in GDP. Oil prices WPOIL: OECD crude oil import price, cif, US$ per barrel Weight on oil prices ν= oil intensity in output = production plus imports of oil relative to GDP. Source: OECD Energy Yearbook. OECD Inflation GDP weighted average of OECD consumer price inflation, excluding high inflation countries. Real exchange rate rer = log real exchange rate (CPIDR) based on relative CPIs. Weights based on manufacturing exports adjusting for third-country competitors. Euro area weights exclude intra-euro trade. Output and Spending Domestic demand gap ddgap = 100.(TDDV/GDPVTR-1) where TDDV is domestic demand and GDPVTR is potential output (both in volume terms). Net export gap xmgap = 100.(XGSV-MGSV)/GDPVTR, where XGSV and MGSV are exports and imports of goods and services, volume, respectively. Output gap ygap = 100.(GDPV/GDPVTR-1). Government spending gspend = 100.(CGV+IGV)/GDPVTR Government revenue grev = NLGQA + gspend, where NLGQA is the cyclically adjusted general government balance. World gap and rest of world gap wldgap = GDP weighted world output gap. Based on Secretariat's estimates of output gaps for OECD members. The gap for the rest of the world is based on taking a Hodrick Prescott filter through an aggregate of rest-of-the-world GDP. Relative gap relgap =ddgap – foreign ddgap, where the foreign ddgap is a trade weighted average of foreign domestic demand gaps. Interest Rates Short-term interest rate IRS Long-term interest rate IRL Real long-term interest rate (domestic demand eqn) r = IRL – πe where πeis inflation expectations, proxied by very smooth Hodrick Prescott filter through actual inflation (λ= 20,000). Real short-term interest rate (policy rule) IRS - annual core inflation rate. Other US Sharemarket wealth US household sharemarket wealth relative to disposable income. Wealth source: US Federal Reserve Board Flow of Funds Table L100. Japanese land prices Japanese urban land price index. Source: Japan real estate institute. A Small Global Forecasting Model References Ball,L. (1998), Policy Rules for Open Economies.Reserve Bank of Australia Research Discussion Paper 1998-06. Sydney. Beechey, M., N. Bharucha, A. Cagliarini, D. Gruen and C. Thompson (2000), A Small Model of the Australian Macroeconomy. Reserve Bank of Australia Research Discussion Paper 2000-05. Sydney. Bharucha, N. and C. Kent (1998), Inflation Targeting in a Small Open Economy. Reserve Bank of Australia Research Discussion Paper 1998-07. Sydney. Coenen,G.and V. Wieland (2000),A Small Estimated Euro Area Model with Rational Expectations and Nominal Rigidities. ECB Working Paper 30. ECB, Frankfurt. De Brouwer,G.and L. Ellis (1998),Forward-Looking Behaviour and Credibility:Some Evidence and Implications for Policy. Reserve Bank of Australia Research Discussion Paper 1998-03. Sydney. Duguay, P. (1994), Empirical Evidence on the Strength of the Monetary Transmission Mechanism in Canada: An Aggregate Approach. Journal of Monetary Economics 33: 39–61. Fair, R.C. (2000), Estimated, Calibrated, and Optimal Interest Rate Rules. http://fairmodel.econ.yale.edu/rayfair/pdf/1999D.PDF. Giorno, C., P. Richardson, D. Roseveare and P. Van Den Noord (1995), Potential Output, Output Gaps and Structural Budget Balances. OECD Economic Studies 1995 (1): 167–209. Hall,S.G.and J.D. Whitley (1999),Linkages between Countries in International Models. Unpublished. Hargreaves, D.(1999), SDS-FPS:A Small Demand-side Version of the Forecasting and Policy System Core Model. Reserve Bank of New Zealand Discussion Paper G99/10. Reserve Bank of New Zealand, Wellington. Hooker,M.(1999),Are Oil Shocks Inflationary? Asymmetric and Nonlinear Specifications versus Changes in Regime. Finance and Economics Discussion Paper 1999-65. Board of Governors of the Federal Reserve System, Washington, DC. Meredith, G. (1997), Effect of Equity Prices on Aggregate Demand. Unpublished IMF Office Memorandum. Murata, K., D. Turner, D. Rae and L. Le Fouler (2000), Modelling Manufacturing Export Volumes Equations – A System Estimation Approach. OECD Working Paper 235. OECD, Paris. Richardson,P.,L. Boone,C. Giorno,M. Meacci,D. Rae and D. Turner (2000),The Concept,Policy Use and Measurement of Structural Unemployment:Estimating a Time Varying NAIRU across 21 OECD Countries. OECD Working Paper 250. OECD, Paris. Rudebusch, G. (1999), Is the Fed Too Timid? Monetary Policy in an Uncertain World. Federal Reserve Bank of San Francisco, mimeo. Smets, F. (1998), Output Gap Uncertainty – Does it Matter for the Taylor Rule? BIS Working Paper 60. BIS, Basle. 112 David Rae and David Turner Stiglitz, J. (1997), Reflections on the Natural Rate Hypothesis. Journal of Economic Perspectives 11(1): 3–10. Turner, D. and E. Seghezza (1999), Testing for a Common OECD Phillips Curve. OECD Working Paper 219. OECD, Paris. 113A Small Global Forecasting Model Gabriel Fagan, Jérôme Henry, and Ricardo Mestre Structural Modelling of the Euro Area1 The econometric work on the euro area has already a long history at the European Central Bank. Indeed, much inspiration has been drawn from the analyses made ahead of monetary union by the European Monetary Institute (see e.g. EMI 1997),which indicated that the ECB would require a comprehensive econometric infrastructure. In order to meet this need in a timely way, given the standard and well-known lead times involved in the process of model-building, related work had therefore to start quite early (even before knowing precisly the country composition of the euro area). The first efforts, already at the EMI times, mostly concentrated on data collection for the EU countries and on tests on aggregation of country data. Estimation and simulation tasks conducted afterwards also implied successive rounds of amendments to the data and the model before publication of the first results obtained could be considered – see Fagan et al. (2001). In line with suggestions made at the EMI times,the modelling of the euro area conducted at the ECB aims at producing a „suite of models“, since, in principle,a variety of models would be better suited to address issues of different nature. In this respect, the type of models that were considered useful ranged from time-series ones – univariate, multivariate, trend-cycle, dynamic factor models – to structural models of various size, various country coverage, and various degree of disaggregation. This objective has to be considered in a quite specific context, where issues to be faced go beyond those met when modelling single countries in isolation. In particular, aggregation across countries with possibly differing behaviour and conceptual data definitions is not a trivial issue. Another issue arises from the fact that structural breaks around the onset of EMU are more likely to occur than in other periods,so that the stability of relations estimated on past behaviour may be a less reliable feature than is normally the case. 1The opinions expressed in this paper are solely those of the authors and do not necessarily represent those of the ECB. Against this background, this paper presents an overview of the work on the models which have been developed over the recent years by ECB staff, with a particular focus on the structural modelling of the euro area. In the first section, the history of modelling activities at the EMI and at the ECB is recalled, with a particular emphasis on the approach followed and the various tools developed/being developed. The second section concentrates on a illustration of this approach, based on the Area Wide Model (Fagan et al. 2001), presenting its structure, the key parameters,and how the model works in terms of both its long run steady state and the dynamic adjustement to it following a shock. Finally, the third section concludes and mentions further work to be envisaged on such models. 1. The story underlying the history of euro-area modelling at the ECB It was mentioned early enough that, as a new central bank, the ECB should have at its disposal a refined econometric infrastructure, comprising in particular structural macro models for the euro area. For instance, the published EMI Report (1997) on the strategy for the single monetary policy put some clear emphasis on the need for such work to be conducted. Such tools were deemed necessary for tasks such as forecasting and the analysis of the „transmission mechanism“ for the single currency area.Therefore,those tools would provide help to decision-making at the area level. 1.1 A specifically challenging context When trying to cope with such a recommendation,EMI and later ECB modellers had to face a quite specific context, in terms of both data and economic issues. First, as regard data, measures of economic variables have not always been – and are still not fully – based on harmonised concepts across countries now constituting the euro area. Second, from a more purely economic standpoint, the heterogeneity of behaviour across countries may also complicate the understanding of the mechanisms driving the euro area as a whole. In addition,econometric work on euro area data is likely to a particular extent to be affected by „structural“ change. A first, albeit mechanical, example of such structural change relates to the number of new definitions that are employed for a number of key variables, the definition of which had to be changed because of the monetary union, such as money stocks (the ECB euro area new aggregates) and similarly the new Balance of Payments statistics. In addition, other series have been harmonised, such as the key consumer price indices (the Eurostat HICP), and also those affected by the overall changeover to a new system of National Accounts (the Eurostat ESA95). More fundamentally, from an economic viewpoint, monetary union is expected to also affect underlying behaviour,since a new central bank is created,implementing 116 Gabriel Fagan, Jérôme Henry, and Ricardo Mestre a new and single monetary policy for the whole area. This could have implications on the functioning of financial markets,on responses of agents to interest rate changes, on formation of expectations, etc.2 1.2 A suite of models, with priority on structural models As a result of the above mentioned considerations, it was felt that a flexible framework was needed, in terms of data, country coverage, and also functions of models.Ideally,a „suite“ of models would be constructed and become available, in the vein of e.g. the Bank of England approach, whereby specific tools could be built and used for specific purposes. Along these lines of thought, simple unior multi-variate time series models could first be used for forecasting. In addition, some more economic structure could be embedded in such models by resorting to Vector ECMs. Further, with a view to refining the policy analysis, small size macro models would be employed.Finally,at the end of this spectrum of econometric tools, medium size structural macro models would be designed and run to carry out policy analysis simulations and also back forecasts with an economic “story” consistent with recognised standard economic relations. However, some prioritisation had to be defined, which led to priority being given to medium size structural models (such as the AWM, see Fagan et al. 2001).A number of reasons explain that.First,such modelling work clearly implies long lead times – in terms of data collection, estimation, test simulations, respecificaiton, etc. Second, there was a clear favourable spillover of database building on to other activities, by making the model data base available for other empirical work. Third, such models were thought to have a potential to understand and describe the transmission mechanism. Finally,the structure of such models is usually rich enough to document a forecast or a simulation with a sufficient degree of realistic detail whereas smaller models,albeit illustrative are not as useful in this respect. In order to produce tools that could be employed in time, a number of early moves had to be made. First, as regard data, compilation of the database started as early as in 1996. Of course, at the time the list of EMU members was not yet known, hence the need to prepare data construction and aggregation routines that would work for any final list of participating countries, with no restriction ex ante. Second, alternative aggregation methodes were programmed, tested and compared (levels vs. indices / time-varying vs. fixed weights / simple sum with fixed vs. current exchange rates, etc.). The first product addressing those issues, namely a money demand study (Fagan, Henry 1998), could therefore be considered as our starting point for structural euro 117Structural Modelling of the Euro Area 2The approach followed is further documented in Henry 1999. area modelling. Third, other central bank modellers, from both North America and Europe,were consulted to gather ideas on the appropriate specification for a euro area model, given the constraints faced. 1.3 Two structural models, the AWM and the MCM, supplemented with other tools A number of econometric tools have been developed by ECB staff, among which structural models for the euro area as a whole or for the largest countries of the area, supplemented with a number of smaller size econometric tools. In most cases, research is still on-going on the devlopment of those various tools, and only some of the work has been published to date. A first example of this modelling work is the AWM (cf. Fagan et al. 2001), which treats the whole of the euro area as a single economy. This quarterly model comprises 89 equations, of which 15 are behavioural – with mostly estimated and sometimes calibrated parameters. The sample period over which the estimation has been carried out spans over 1970 to 1998. The corresponding database – an original construction given the lack of official backdata for the euro area – is available on the ECB website. Another example of such work is the ECB contribution to the construction of a quarterly multi country model (MCM) in collaboration with the National Central Banks of the ESCB. As the name indicates, this model will comprise models for each of the countries in the euro area, which then could be reconstructed by aggregation of those. In addition a trade link block is being developed at the ECB to connect the various countries when simulating the various country models jointly.Each of the country block is similar in size and structure to the the AWM.3The period used for estimation at the ECB covers the sample 1980 to 1997. In addition to these two examples of medium size structural macroeconometric models,a number of other ECB researchers have been working on a variety of projects. In some cases work is still in progress so that no publication has been undertaken. This is so, e.g. for time series models for both monthly and quarterly inflation forecasts, with a particular emphasis on the HICP. There are however other publications available which present work already well advanced on alternative tools. Starting from the less to the most structural models, a dynamic factor model for trend inflation and inflation forecasts is presented in Angelini et al. (2001a, 2001b). Different VARs based on standard money demand system (comprising money, inflation, interest rates, GDP) are documented in Coenen/Vega (1999), Brand/Cassola (2000) and such models can be used for forecasting purposes, as in Trecoci/Vega (2000) and Nicoletti 118 Gabriel Fagan, Jérôme Henry, and Ricardo Mestre 3The Central Bank of Ireland (see Mc Guire,Ryan 2000) and the National Bank of Belgium (see Jeanfils 2000) have already published material on their own contribution to the MCM. to restore equilibrium,a number of mechanisms operate, involving disequilibrium terms and policy responses. Assume for illustration purposes that the model is supplemented with a forward-looking UIP condition, a Taylor rule and a fiscal rule. The main mechanisms are then as follows (taking the example of a positive aggregate demand shock): First,the shock mechanically increases output and employment,leading therefore to an increase in inflation via the Phillips curve. This triggers a rise in real short-term interest rates,since both arguments in the Taylor rule are deviating from their equilibrium or long-run values.This puts downward pressure on domestic demand, by weakening investment and therefore aggregate demand. Second, some external channel will operate too, although limited it might be, for a relatively closed economy. In line with the expected change in interest rates, the UIP condition would lead to an initial jump in nominal exchange rate. There would be ceteris paribus an appreciation of the real exchange rate, therefore exerting downward pressures on both prices (via diminished imported inflation) and demand (via lower net trade and also lower net foreign assets). Third, this initial nominal and real appreciation is reinforced by a further real appreciation and a crowding out via net trade caused by the additional inflation and growth resulting from the shock per se. First, the additional inflation induces a real appreciation of the exchange rate, which would tend to weaken net trade and, in part, offset the initial increase in output. Second, increased demand would boost imports, leading to a further weakening of trade contribution to growth. Fourth, the “automatic stabilisers” of fiscal policy imply in the case at hand that transfers to households should fall on foot of lower unemployment, helping to further dampen the growth of disposable income.In addition,in the case where the shock emanated from a fiscal expansion, the fiscal solvency rule gradually “kicks-in” and the rise in direct taxes also dampens demand. As an illustration we report one simulation to document the impact of the various adjustement mechanisms that are at play. In that exercise, a shock to public consumption amounting to 1% GDP is assumed. The shock is permanent – with monetary policy response and endogenous forward-looking exchange rate. As can be seen in Figure 1, there is a positive response of GDP to the fiscal stimulus,with a multiplier however remaining small (close to one only the first year) and short lived (negative impact after 5 years).The various crowding out mechanisms tend to counteract the initial positive shock to demand, namely 125Structural Modelling of the Euro Area the competitiveness, fiscal stabilisers, and monetary policy response, as above mentioned. In particular, monetary policy has an effect on net trade through the initial appreciation in the exchange rate – reflecting future increases in the interest rate. Also investment is dampened by the impact of the changes in interest rates.Beyond the one year horizon, fiscal policy also adjusts, via taxes to ensure that the debt target is met,so that consumption is also affected.In addition, the price level albeit lower than baseline initially – thanks to the appreciation of the currency – very quicly increases beyond the baseline level, remaining higher than the latter for a substantial number of years. Inflation remains higher than baseline until about 8 years for both the GDP and consumption deflators. Eventually, let alone some cyclical fluctuations, inflation and the GDP level are consistent with their baseline values, both unemployment and output gaps being closed and the additional public demand compensated by other demand components being lower. 126 Impact on GDP and prices of a permanent increase in real public consumption deviation from baseline in % GDP Prices 1 2 6 8 10 124 14161820 years 1,2 1,2 1,2 1,2 0,9 0,9 0,9 0,9 -0,3 -0,3 -0,3 -0,3 0 0 0 0 0,3 0,3 0,3 0,3 0,6 0,6 0,6 0,6 -0,6 -0,6 -0,6 GDP-Deflator Consumption-Deflator Figure 1 Authors’ calculations. Explanations see text. Gabriel Fagan, Jérôme Henry, and Ricardo Mestre 2.5 Evaluation and use of the model Prior to finalising the published version, the model had been regularly evaluated, mostly via diagnostic simulations. Successive versions were tested, comparing the corresponding simulation results with a number of single country or multi country models. This led to successive rounds of respecification and additional and / or changed calibration for some of the parameters – such as e.g. the role of expectations in price and wage behaviour. In its present version,the model can be used for a number of purposes,making the most of the flexibility of the framework.Diagnostic simulations are carried out, in and out of sample experiments, involving permanent or temporary shocks, both with respect to historical and steady state baseline, using either forward or backward looking formulations. Similarly, the simulation environment can be defined alternatively with endogenous or exogenous real interest rates, endogenous or exogenous nominal exchange rate, endogenous or exogenous fiscal policy variables. Shocks that could be analysed with the model affect public expenditure, world demand, world prices, exchange rate, interest rate,taxes,etc.Stochastic simulations can also be implemented with the model, as documented in Fagan et al. (2001). 3. Conclusions A number of conclusions can be drawn from the experience already accumulated in terms of euro area structural econometric modelling: –First of all, the investment, although quite heavy initially, appears worthwhile eventually,in view of the properties of the resulting model and also its potential use for a variety of purposes. –Second,the new “object”which appeared in 1999,namely the „euro area“ – strictly speaking the-countries-now-comprising-the-euro-area – is now better known.A first set of stylised facts can e.g.be derived from the estimation of standard relations on euro area data, which also contributes to focus the discussion on euro area level facts. –Third, there are of course elements of uncertainty affecting those facts, that are clearly due to key factors,such as data quality and availability,structural change around EMU,the absence of long expertise,and also the issue of aggregation across a priori relatively heterogenous countries. The combination of country level analysis with the euro area one may there be a way forward. –Fourth, a benchmark is now available, as against which to assess judgements. The process of modelling the euro area has only started; there is a clear need e.g. to systematically compare estimated standard behavioural 127Structural Modelling of the Euro Area equations based on synthetic euro-area data to those obtained on US or other countries’ data. –Fifth,more generally,such modelling work has to follow a continuous development and research process – already making plans for the next generation of our models, not having yet fully finalised the 1st one. In any event, such structural models are only one among other tools, econometric or not,that can and have to be employed to assess the euro area economic situation. References Angelini, E., J. Henry and R. Mestre (2001a), A Multi-Country Trend Indicator for Euro Area Average Inflation:Computation and Properties.ECB Working Paper 60. European Central Bank, Frankfurt a.M. Angelini,E., J. Henry and R. Mestre (2001b), Diffusion Index Based Inflation Forecast for the Euro Area.ECB Working Paper 61.European Central Bank,Frankfurt a.M. Arnold,I.(1996),Money Demand Stability in the EMU:Lessons from the US.Working Paper. Nijenrode University. Brand, C. and N. Cassola (2000), A Money Demand System for Euro Area M3. ECB Working Paper 39. European Central Bank, Frankfurt a.M. Coenen,G.and J.-L. Vega (1999),The Demand for M3 in the Euro Area.ECB Working Paper 6. European Central Bank, Frankfurt a.M. Coenen,G.and V. Wieland (2000),A Small Estimated Euro Area Model with Rational Expectations and Nominal Rigidities. ECB Working Paper 30. European Central Bank, Frankfurt a.M. EMI (ed.) (1997),The Single Monetary Policy in Stage Three: Elements of the Monetary Policy Strategy of the ESCB. European Monetary Institute, Frankfurt a.M. Fabiani, S. and R. Mestre (2000), Alternative Measures of the NAIRU in the Euro Area: Estimates and Assessment. ECB Working Paper 17. European Central Bank, Frankfurt a.M. Fagan, G. and J. Henry (1998), Long Run Money Demand in the EU: Evidence for Area-Wide Aggregates. Empirical Economics 23 (3): 483–506. Fagan G., J. Henry and R. Mestre (2001), An Area Wide Model (AWM) for the Euro Area. ECB Working Paper 42. European Central Bank, Frankfurt a.M. Gerlach, S. and L. Svensson (2000), Money and Inflation in the Euro Area: A Case for Monetary Indicators? Mimeo, BIS, IIES, Basel, Stockholm. Gordon, R. (1997), The Time-Varying NAIRU and its Implications for Economic Policy. Journal of Economic Perspectives 11 (1): 11–32. Henry, J.(1999), Euro Area-Wide and Country Modelling at the Start of EMU.Economic and Financial Modelling 1999 (Autumn): 103–148. 128 Gabriel Fagan, Jérôme Henry, and Ricardo Mestre Jeanfils, P. (2000), A Model with Explicit Expectations for Belgium. Working Paper 4. Banque Nationale de Belgique, Brussels. Mc Guire, M. and M. Ryan (2000), Macroeconomic modelling developments in the Central Bank. Spring Bulletin. Central Bank of Ireland, Dublin. Mitchell, P.R., J.E. Sault and K. Wallis (2000), Fiscal Policy Rules in Macroeconomic Models: Principles and Practices. Economic Modelling 13 (2): 169–184. Nicoletti Altimari,S. (2001),Does Money Lead Inflation in the Euro Area? ECB Working Paper 63. European Central Bank, Frankfurt a.M. Taylor,J.B. (1993),Discretion Versus Policy Rules in Practice.Carnegie Rochester Conference Series on Public Policy 39: 195–214. Trecroci, C.and J.-L. Vega (2000),The Information Content of M3 for Future Inflation. ECB Working Paper 33. European Central Bank, Frankfurt a.M. 129Structural Modelling of the Euro Area Jean Louis Brillet and Maria Dos Santos The Consequences of EMU Entry for Spain and Portugal – Simulations Using the MacSim System 1. The model We shall begin by a short description of the single country models,focusing on the originality of the two new countries. Then we shall describe, also shortly, the interactive system. A more detailed presentation of both single country behaviours and international interactions can be found in Augier et al.(1999). 1.1 Introduction The Macsim package is based on a set of simplified models, associated with some of the main countries in the European Union. It brings together single-country mechanisms, including some financial elements, and international trade, represented by bilateral flows. It considers essentially the consequences of shocks, associated to fiscal and financial policies. For a single shock, the results will depend on the rules for determining the interest rate and the exchange rate, as well as the subset of countries belonging to the European Monetary Union. The standard version of the model considers 6 countries:France, Germany,Italy, Netherlands, Sweden, United Kingdom. In the present paper, simulations will use Spain and Portugal instead of Netherlands and Sweden. However, in our discussion on single country models, we shall consider tables for the whole set of 8 countries. 1.2 The single-country basic model The single country model uses the structure of the MicroDMS model (Brillet 1997a) adding a few error correction mechanisms (Brillet 1997b). We have associated behaviours to the following concepts: 132 Figure 1 The s i ngle country model target investment investment actual competitiveness exports imports target capital actual capital actual unemployment target capacity actual capacity rate of use agents accounts demand: consumption, productive investment, change in inventories, total work force target employment employment employment value added Capital productivity Foreign prices World demand target rate of use Labour productivity civil servants working age population indirect taxes firms social contributions subventions income tax tax on profits workers social contributions social benefits government demand wages prices Jean Louis Brillet and Maria Dos Santos –Production factors: investment and employment, unemployment. –Prices: wages, value added prices, export and import prices. –Firms: Changes in inventories. –Households: consumption. –External trade: exports and imports. –Interest rate: real exogenous value in the base version. –Exchange rate: exogenous in the base version. In all we shall estimate 11 equations per country.When country models are assembled, it is clear that some estimations concerning external trade will lead to over identifications, imports by one country being composed of exports by other countries, which introduces constraints on quantities and values exchanged.We shall present however the full set of equations,which were useful for testing individual models. We shall now present the main behaviors, and the estimations for the eight countries selected. 1.2.1 Productive investment ()[]ik ci k c qq q ut ct tttt ttt t // //. −−− −− =⋅ +⋅ − + +⋅ 112 11 12 0853prob c t+4. (1) We suppose here that firms set an investment target, depending on: –the profits rate, representing both the expected profits on new investments, and the potential to finance them. –a desire to adapt productive capacity at the next period to the expected demand.The last element implies the expected growth for the next period,but also the present adaptation of capacities to production. –and that this process is affected by a strong inertia. One can observe that the presence of the rate of use will ensure in the long run a full adaptation of capacity to production,at a level depending on the profitability of capital. Results appear in table 1 (Appendix). If the lagged term and the “real” effect are almost always significant,the contribution of the profits rate shows often a poor quality. We have kept it nonetheless, and even fixed it for Italy. 1.2.2 Employment We suppose that firms have a target labor productivity,associated with a structural trend. Knowing production this defines a target employment, to which 133Consequences of EMU Entry for Spain and Portugal – Simulations actual employment adapts dynamically.Observing the graphs and using statistical breakpoint tests (such as the Chow or Perron tests) allows to identify two structural trends for labor productivity, with a negative break around 1973. () () ()()ΔΔLog le c Log q c Log q le c t c t c tt tt =⋅ +⋅ −⋅−⋅ + −− 1 2 3 4 73 5 11 /. (2) Where t73 is zero until 1972, then grows by 1 for each successive period. Results appear in table 2. They are always very significant, with rather similar results from one country to another (lower for Italy). The first coefficient (immediate response) is almost always higher, and our new countries present the highest values. The break is also identified for all countries. 1.2.3 Unemployment The variations of employment do not translate fully into unemployment,as an improved employment situation will attract to the labour market previously inactive persons. The work force (employed + unemployed) will increase. As usual, we shall use an error correction framework. ΔΔΔcho c le c pop c cho c le c pop tt t tt =⋅ +⋅ +⋅ −⋅ −⋅ −− 1 2 65 3 4 5 11 ()65 6 1tc −−.(3) Results are presented in table 3.Results are generally good, with variable sensitivity to labor across countries. A little surprisingly, our two new countries present quite different short-term dynamics. While created jobs are mostly filled by the unemployed in Spain, they essentially increase the work force in Portugal. 1.2.4 The value added price () ( ) ()ΔΔLog pva c Log c c ut c Log c pva ttt tt =⋅ +⋅ +⋅ + −− 123 11 sup sup /()c4.(4) We suppose that firms use the price level to optimize between quantities sold (at a given capacity) and margins on each unit. This introduces a positive link between rate of use and margins. Going from one optimum to another due to changes in external conditions, firms will move both targets in the same direction. As usual, we shall apply an error-correction mechanism. Results are presented in table 4. 1.2.5 The trade prices When defining their prices, exporters can take into account their own costs or the price of their competitors (in the same currency). The first behaviour will leave margins unchanged, but affect competitiveness. With the second behaviour, the reverse will happen. In our model, all exporters will apply both be134 Jean Louis Brillet and Maria Dos Santos ()()[] m m b a pex ch ch pim b ut utx ij i ij j i j i j i,, ./=− −−−1(20) which means that (as for the single country models) exporters to one country will increase a “natural” share with competitiveness and available capacity, this time relative to their competitors. One will observe that this technique guarantees the identity of the sum of individual exports with its global value, without any correction. Of course, the coefficients can be different from one market to another,but not within one market. The system can be summarized following figure 2. In addition to the above, we have introduced some accounting equations: –Exports from country ito j: 141 Th etra d e bl oc k Price compet i Production Price Price compet of j market i Export price of j market i Capacity compet i Rate of use i Cap compet of j on marketi Rate of use j Imports i Total Exports of j Demand Exports of j to i Rate of exports to i Import price i Production price j Imports of country i Exports of country j Figure 2 Consequences of EMU Entry for Spain and Portugal – Simulations xmusdusd ijjiji,, /,=⋅00 (21) where usd0irepresents the base year value of the currency of country i,in US Dollars. –Total exports computed as a sum xx iij =∑,.(22) –The average export price of country i: pex b pex iijij =∑,, .(23) 1.4.2 The Rest of the World The Rest of the World will not be associated with a model. Actually, we our goal will only be to give to its trade elements similar properties to the countries we consider. But this does not mean that we can keep its prices exogenous, or its capacities infinite: in case our six countries lead the same policy simultaneously (lowering the social contributions for example) we cannot assume that inflation and demand in the Rest of the World will remain unaffected. We have chosen the following system, introducing the main mechanisms through a limited set of equations. Production: We consider only production generated by exports. We start from the supply-demand equilibrium: Q = D + X – M. (24) Ex ante, Xgenerates Q = X. Ex post, an increase in production will generate revenue and demand: D = a Q. (25) Of which a share b will be imported: M = b FD. (26) Finally we get: Q = 1/(1–a (1–b)X. (27) Final demand will be defined as: FD = c a Q + (1–c)FD0, (28) where FD0 represents the part of final demand independent from trade. 142 Jean Louis Brillet and Maria Dos Santos In the absence of investment and capital, the rate of use follows an autoregressive behavior: Log(UT)= c [d Log(UT_1) + e Log(Q/(Q–1(1+txq)))) + (1–c) Log(UT0).(29) As above, the share of production concerned with trade is treated endogenously. It separates two effects : –the previous disequilibrium on capacities at a normal growth of production (it disappears gradually through additional investment); –the present gap on production between the actual level and the one associated with a normal growth. The production price differs from an exogenous track through the evolutions of the rate of use, and the import price (actually their deviations from a base track): () ( ) ( ) ( ) Log pp Log pp f Log ut ut g Log pim pim ttt tt =+⋅ +⋅00 0//. (30) The export price will use exactly the same equation as country models. Finally, the import equation uses the same framework as the single counties, except that it is not estimated: () ( ) ( ) ( ) Log m c Log df ouv c Log ut c Log pim pp ttt t tt =⋅ ⋅ +⋅ − +⋅ −1213/ … (31) 2. The model properties 2.1 Simulation context We shall now examine the properties of the full model.For this we need a base simulation. To make the diagnosis easier,we shall produce it on the future, using simplified assumptions. We shall suppose that the economy grows at structural rates, the same for each country: –growth rate of population: 0.3 % per year, –growth rate of labor productivity: 2.4 % per year, –growth rate of prices: 2.4 % per year. For this analytic simulation, we shall use a very long period, for several reasons: –We want to control the presence of a long-term equilibrium,and of a steady state path.We do not necessarily believe conclusions drawn from the text of the equations. –We want to free the results from any short-term fluctuations. As the model is not linear, an irregular base solution could disturb the sensitivities to shocks, and make the diagnosis less clear.If everything goes well,we should 143Consequences of EMU Entry for Spain and Portugal – Simulations get regular curves, easier to interpret. On the opposite, any irregular trajectory will be attributed to the model. Our simulations will be conducted on a 200 years period, which appear to guarantee long term convergence.In the same way,shocks will start in 2101.Of course,we have checked that the general conclusions drawn from these shocks can also be applied to the period we shall actually use: 2000 and the following periods. A definition of the “structural” shares of partner countries in imports and exports of a given country. We have used the definition equations for the foreign markets of one country, presented by the NiGEM model, and the global imports from the year 1995.The shares of partners in the exports of our six countries (and of the Rest of the World by difference) give in turn: –imports by the Rest of the World from each country, then total imports of the Rest of the World, –imports of countries from each single country, then individual imports from the Rest of the World (by difference), then total exports of the Rest of the World. In table 9, each line gives the share of each market in the exports of the associated country. Of course, only the lines and columns associated to the new countries and the rest of the world will be affected. 2.2 The shocks We shall now consider how the composition of the Euro zone affects out two countries through the consequences of policy shocks. We shall concentrate on 144 Export market shares by exporting country Market Germ UK Fran Ital Neth Swed RoW1 Port Spai RoW2 Exporter Germany . 0.080 0.120 0.075 0.075 0.023 0.627 0.009 0.032 0.684 UK 0.121 . 0.094 0.047 0.066 0.025 0.647 0.009 0.035 0.683 France 0.168 0.097 . 0.092 0.045 0.011 0.587 0.015 0.070 0.558 Italy 0.190 0.065 0.131 . 0.029 0.009 0.576 0.014 0.047 0.553 Netherlands 0.281 0.093 0.105 0.052 . 0.017 0.452... Sweden 0.133 0.102 0.051 0.034 0.053 . 0.627 . . Rest of World 1 0.227 0.158 0.167 0.230 0.152 0.066.... Portugal 0.187 0.117 0.147 0.034....0.146 0.369 Spain 0.142 0.082 0.201 0.092...0.078 . 0.405 Rest of World 2 0.370 0.204 0.191 0.139...0.083 0,.015 . Authors’ calculations. Table 9 Jean Louis Brillet and Maria Dos Santos asymmetric shocks,affecting either Spain or Portugal.Indeed,we have already observed (see Augier et al.1999) that symmetric shocks had very similar properties notwithstanding the composition of the zone. For simplicity, we shall suppose that the rest of the countries follow the actual status: Germany, France and Italy belong to EMU, while UK does not. We shall only consider four cases, with Spain and Portugal in or out of the system, independently from each other. Logic will lead us to start as a basic case with both countries outside EMU.This will allow us to observe what happens when they join, and not when they quit, which is actually almost unfeasible. 2.2.1 A demand shock in Spain We shall increase Government demand in Spain by one full GDP percentage point. As stand earlier, shocks will begin in 2101, when the steady state is almost reached,but we have checked that conclusions apply to the present situation. Spain and Portugal do not belong to EMU As this is our basic case, we shall have first to give some basic explanations on the macroeconomic effects of such a decision. But we shall be quite short, as the issues are well known,and our model does not present very particular features,compared to the previous MacSim system or to any model of this family. Figure 3 shows that under the assumption of a fixed nominal exchange rate and a fixed real exchange rate, the short-term multiplier for Spain is well above 1,but decreases with time to about 0.4, due to losses in competitiveness, as shown in figure 4. We have seen that for Spain, the sensitivity of imports to prices was very strong; actually not much lower the one. In the short-medium term, we observe the usual cycles, coming from the inertia on capacity adjustment (figure 5). The trade balance deteriorates significantly, even in current terms (figure 6). The PPP assumption eliminates the loss in competitiveness, and stabilizes the multiplier at a much higher level (1.5), with the same medium term cycles. The Taylor rule almost eliminates the cycles, by going against the accelerator as to the influence of the output gap: higher tensions increase investment directly but also increase the interest rate. In practical terms, if a pursuer keeps the same inertia on its previous speed, but can close the present gap much faster, he is more likely to overshoot. The evolution of the Portuguese economy is perhaps more interesting,as some elements were not necessarily obvious. Of course, it profits a lot from the Spanish demand, as it represents about 15 % of its exports (see table 8).It will be noted that Spanish imports from Portugal account only for8%ofthetotal, due to the difference in country size. 145Consequences of EMU Entry for Spain and Portugal – Simulations We can see (figure 7) that in the short medium term,the Portuguese trade balance actually deteriorates.This comes from several sources.First,to supply the Spanish demand, it has to import equipment goods (about 2 units per additional output unit), and the jobs created will create additional consumption. Second, the limits on capacity will reduce the exports to the other countries, and the satisfaction of local demand, which leads to imports. As capacities build up, these effects will disappear, but Portuguese inflation will increase more than the rest (except for Spain, of course) (figure 8). This country will lose competitiveness compared to EMU+UK, and even on average if we include Spain. And we have seen that Portuguese imports are quite sensitive to prices.This loss in competitiveness reduces the need for additional capacity, leading investment down and GDP further down. This accounts for the fact that in the long run Portugal does not profit from the Spanish policy, while Spain itself does (but at some cost) (figure 9). But we have still to explain the higher inflation itself, in the absence of local growth. It actually comes from the lingering effects of initial tensions on capacities and labour, and from the inflation imported from Spain,as the change in the relative prices has a limited substitution effect on import shares. The changes in rules bring logical results. With PPP the GDP increase reappears,to a normal level (0.2 is about 15 % of the Spanish 1.5 multiplier), which shows that the loss in competitiveness was indeed to blame. The Taylor rule with a fixed exchange rate smoothes the profile without bringing any long term growth, and with PPP it the additional inflation increases the real rate and limits the growth. Spain does not belong to EMU, but Portugal does (figures 10 to 12) From now on, we shall present the differences with the previous case. First, let us remark that the Spanish situation is not significantly affected. This will always be the case: the properties for one country depend essentially from the conditions of that country. This is truer of course for Portugal, considering its size. Considering the basic case, we can observe that Portuguese growth is somewhat improved. The reason lies in the (disputable) fact that the interest rate is decided at the EMU level, and that the higher Portuguese inflation reduces the local real rate, bringing more growth (and more inflation). With the other rules, we can observe interesting evolutions: –First, PPP reduces inflation, compared to a fixed rate. In the previous case, Portuguese inflation increased more than the average of its partners, due to local tensions and the Spanish influence.Now we consider EMU as a whole: in EMU (including Portugal), inflation increases a little more than at its partners. Weighted by the specific Portuguese trade (in which the devalua146 Jean Louis Brillet and Maria Dos Santos ted Spanish peseta plays a large role), the Portuguese exchange rate actually improves, bringing deflation.Of course, this brings also a loss in competitiveness, and eliminates any GDP growth. –Second, the Taylor rule is less efficient in its dampening effects. The rate of use, and the over indexation on inflation, are computed at the EMU level. Spain belongs to EMU, but Portugal does not (figures 13 and 14) Comparing results with the basic model shows the usual effects. With fixed nominal exchange rate and real interest rate determined at EMU level, growth is improved by the decrease of the real Spanish rate. With PPP, the loss in competitiveness is only very partially limited,as the Euro adapts to the evolution of EMU inflation. The increase in inflation is only a little higher. This means that the Taylor rule will have similar effects in both cases, close to the previous fixed exchange rate case. Defined at the EMU level, its dampening effects are a little lower, but do not affect growth on average. For Portugal, it is interesting to note that local inflation being higher than in the EMU zone, PPP actually devaluates the currency relative to the Euro and brings large competitiveness gains compared to Spain, which has devaluated much lower than its own inflation. The long run GDP increase stabilizes at a much higher level. Spain and Portugal belong to EMU This case is less interesting, as it gives the usual results of any asymmetric shock inside EMU. PPP gets closer to the fixed rate assumption, the Taylor rule stabilizes less, and does not differentiate between exchange rate rules.We can observe however that Portugal profits from the decrease of its real interest rate,and the lower inflationary effect compared to Spain.The improvement of growth is quite significant. 2.2.2 A supply shock in Spain We shall now decrease the rate of firms’ social contributions by one percentage point. This reduces ex ante the wage cost by around 0.75 %. Spain and Portugal do not belong to EMU (figures 15 and 16) The base assumption produces the usual results. Firms decrease their prices, gaining competitiveness, and invest as profitability increases. Both effects improve GDP,but take some time in reaching their full impact,as the investment decisions are rather inert, and capacity is needed for external trade to profit from competitiveness. In the first period, the long-term interest rate follows only partially the reduction in inflation,especially in the “real”case.Its real value increases,leading to a globally negative effect on GDP in the first period.As usual,PPP reduces the 147Consequences of EMU Entry for Spain and Portugal – Simulations efficiency with this elimination of competitiveness gains. But the effect is almost compensated by the higher disinflation, and the decrease in the price of imports, which improves profitability through the price of capital. And the Taylor rule mixes an over indexation on reduced prices with a decrease of the output gap, with uncertain results. In our particular case, the most interesting observations come from Portugal. But this time results are less clear, even in the basic case. As a main partner of Spain, Portugal suffers higher losses in competitiveness, but not too important as the reduced price of products imported from Spain lowers local inflation. And at the same time Spanish imports increase, a large share coming from Portugal. On the whole, Portuguese GDP does increase in the short-medium term, but by a small amount. In later periods, the variations coming from strong cyclical effects are quite higher than the small average improvement. PPP leads to a significant improvement of GDP. If Spain loses its competitiveness gains, the shock brings still growth through capital profitability and the increase in the real holdings of households. Portugal profits from it, the more so as competitiveness is maintained. And the Taylor rule, as usual, stabilizes the cycles, and reduces the efficiency in the PPP case. Spain does not belong to EMU, but Portugal does Now Portugal has to follow the interest rate and exchange rate of the Euro zone. As its deflationary effect was higher than its partners, the real interest rate increases, and PPP maintains some competitiveness gains. The basic case leads to a decrease of activity in the short run, and the interest rate plays again a stabilizing role, through quite different channels than the Taylor rule.This is somewhat compensated somewhat by PPP.But all these effects are not very high, as we have not changed the option for Spain. And the size of Portugal limits the backward effects of the new assumption on the Spanish economy. In the long run, the shock becomes demand-oriented, and the situation improves as the EMU weighted inflation is lower than the Portuguese one, in particular in the PPP case. But this also brings down the Portuguese real interest rate. Spain belongs to EMU, but Portugal does not (figure 17) Now things will be quite different, as the origin of GDP growth in Spain, the improvement of price competitiveness, is maintained for all options, including PPP. But at the same time, we observe an increase in the real interest rate, as inflation decreases more in Spain than in its partner countries.But as these options showed very comparable efficiencies in the non-EMU case, this does not 148 Jean Louis Brillet and Maria Dos Santos have a significant effect, at least in the short and medium term. We observe only a small improvement of the efficiency for Spain. Spain and Portugal belong to EMU As the consequences from each entry were not individually very important, combining them gives results comparable to their sum. 2.2.3 A demand shock in Portugal Spain and Portugal do not belong to EMU (figure 18) Considering the basic case, the results are consistent with the Spanish equivalent. But we can observe that the multiplier is quite lower. Being a much smaller country (about 7 times in GDP terms) the share of imports in demand is much higher, and tensions on capacities appear much faster. The cycles are stronger, as the importance of external trade increases the role of the rate of use of capacities. It is also interesting to note that the gains from the Spanish demand shock were actually higher. Ex ante, the share of Spain in Portuguese exports is around 15 %, and Portugal was not submitted, as in the present case, to inflationary and capacity effects. Finally, the consequences for Spain are symmetrical to the previous case. We observe a short term loss on the real trade balance, as the need to supply the Portuguese demand leads to investment and consumption,and the Portuguese inflation is transmitted though imports. But the effect is somewhat lower, as Portugal is less important for Spain than the reverse. Spain does not belong to EMU, but Portugal does (figure 19) We observe the usual consequences of EMU participation for a country producing an asymmetric shock. But now the size of the country means that the effects of the shock on the interest rate and the exchange rate are very small. The results are almost identical to the basic case,for all the assumptions.This is what we already observed for Spain, where the consequences of a local shock converged to the real exchange rate case. The reason for the limited convergence was the weight of Spain in the determination of the exchange rate. With a weight 7 times smaller, the convergence is now almost full. This was not the case when the shock originated in Spain, as the assumptions used affected directly the consequences for Portugal. As a comparatively important partner of Spain, the shock created disequilibria between Portugal and its partners as to inflation and tensions,and the consequences of these disequilibria on the exchange and interest rates changed a lot with participation to the EMU. Moreover, the exchange rate affected directly the source of the Portuguese growth, exports to Spain. 149Consequences of EMU Entry for Spain and Portugal – Simulations In the present case,the same disequilibria are produced,but they do not significantly affect the nominal EMU rates, leading all cases to the basic one. Spain belongs to EMU, but Portugal does not Belonging to the EMU allows Spain to reduce the increase of interest rates, improving growth. The exchange rate keeps close to the fixed assumption, introducing permanent losses in competitiveness. But this also limits the inflationary effect, with positive consequences. Spain and Portugal belong to EMU As seen earlier, the entry of Portugal in EMU leads all assumptions to the basic case.Again, we find here the consequences of Portugal’s small size. Even with a five-country EMU, the change in rules has virtually no effect on GDP improvement, in the case of a local shock. This is at least partly due to the fact that local inflation has a small effect on the Euro rate; moreover, as its main partner has strong inflationary properties, the fixed exchange rate assumption had properties closer to PPP. 2.2.4 A supply shock in Portugal For simplicity, we shall not address this issue individually. The main observation is that the small size of the country precludes any sizable backward effect, including the changes in the interest rate and the exchange rate. Once this is considered, we get the same conclusions as for Spain, but with a higher efficiency, due to the weight of external trade in the Portuguese economy. However, this is countered by the lower Portuguese disinflation, coming from the smaller dynamic coefficients of the price-wage loop. 2.3 General conclusion This study has confirmed some elements observed in the previous papers, and some new ones. General consequences of the rules in a single country framework Facing an asymmetric demand shock, PPP stabilizes competitiveness, but leads to higher inflationary effects. The change in GDP increases in the original country, while the diagnosis is less clear for its partners, with contrary effects of a larger external market and the disappearance of competitiveness gains. The Taylor rule stabilizes the shocks, by delaying investment, but does not affect the average efficiency. Combined with PPP, the higher inflation and activity increases the real interest rate and exert a permanent downward effect.For supply shocks, PPP limits gains but improvement still comes from the profitability of capital. The Taylor rule has a limited role, as disinflation and tensions have opposite effects. 150 Jean Louis Brillet and Maria Dos Santos