Monetary policy and unemployment: The US, Euro-area and Japan

Semmler, Willi (Ed.) Book Monetary policy and unemployment: The US, Euro-area and Japan Routledge International Studies in Money and Banking, No. 31 Provided in Cooperation with: Taylor & Francis Group Suggested Citation: Semmler, Willi (Ed.) (2005) : Monetary policy and unemployment: The US, Euro-area and Japan, Routledge International Studies in Money and Banking, No. 31, ISBN 978-0-203-32958-0, Routledge, London, https://doi.org/10.4324/9780203329580 This Version is available at: https://hdl.handle.net/10419/243254 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/ Monetary Policy and Unemployment This book pulls together papers presented at a conference in honor of the 1981 Nobel Prize Winner for Economic Science, the late James Tobin. Among the contributors are Masanao Aoki (UCLA), Olivier Blanchard (MIT), Edmund Phelps (Columbia University), Charles Goodhart (LSE), Marco Buti (European Commission), Hiroshi Yoshikawa (Tokyo University and BoJ), Athanasios Orphanides (Fed, US), and Jerome Henry (ECB). Written in the spirit of the long time Yale Professor, James Tobin, who has held the view that monetary policy is not neutral, this volume provides an analysis of the different economic performances exhibited by the USA, the Euro-area, and Japan in the last decade. Through addressing the potential role monetary policy has on economic growth and unemployment, this book also discusses the new policy rules that, perhaps, should have been and should be used in the future to improve the economic performance of the three regions. This book will be of great interest to both undergraduate and graduate economic students, academics, and practitioners. Willi Semmler is Professor of Economics at the New School University New York, and the Center for Empirical Macroeconomics at Bielefeld University. Routledge international studies in money and banking 1 Private Banking in Europe Lynn Bicker 2 Bank Deregulation and Monetary Order George Selgin 3 Money in Islam A study in Islamic political economy Masudul Alam Choudhury 4 The Future of European Financial Centres Kirsten Bindemann 5 Payment Systems in Global Perspective Maxwell J.Fry, Isaak Kilato, Sandra Roger, Krzysztof Senderowicz, David Sheppard, Francisco Solis, and John Trundle 6 What is Money? John Smithin 7 Finance A characteristics approach Edited by David Blake 8 Organisational Change and Retail Finance An ethnographic perspective Richard Harper, Dave Randall, and Mark Rouncefield 9 The History of the Bundesbank Lessons for the European Central Bank Jakob de Haan 10 The Euro A challenge and opportunity for financial markets Published on behalf of Société Universitaire Européenne de Recherches Financières (SUERF) Edited by Michael Artis, Axel Weber, and Elizabeth Hennessy 11 Central Banking in Eastern Europe Nigel Healey 12 Money, Credit and Prices Stability Paul Dalziel 13 Monetary Policy, Capital Flows and Exchange Rates Essays in memory of Maxwell Fry Edited by William Allen and David Dickinson 14 Adapting to Financial Globalisation Published on behalf of Société Universitaire Européenne de Recherches Financières (SUERF) Edited by Morten Balling, Eduard H.Hochreiter, and Elizabeth Hennessy 15 Monetary Macroeconomics A new approach Alvaro Cencini 16 Monetary Stability in Europe Stefan Collignon 17 Technology and Finance Challenges for financial markets, business strategies and policy makers Published on behalf of Société Universitaire Européenne de Recherches Financières (SUERF) Edited by Morten Balling, Frank Lierman, and Andrew Mullineux 18 Monetary Unions Theory history public choice Edited by Fonest H.Capie and Geoffrey E.Wood 19 HRM and Occupational Health and Safety Carol Boyd 20 Central Banking Systems Compared The ECB, the pre-Euro Bundesbank and the Federal Reserve System Emmanuel Apel 21 A History of Monetary Unions John Chown 22 Dollarization Lessons from Europe and the Americas Edited by Louis-Philippe Rochon and Mario Seccareccia 23 Islamic Economics and Finance: A Glossary, 2nd Edition Muhammad Akram Khan 24 Financial Market Risk Measurement and analysis Cornelis A.Los 25 Financial Geography A Banker’s view Risto Laulajainen 26 Money Doctors The experience of international financial advising 1850–2000 Edited by Marc Flandreau 27 Exchange Rate Dynamics A new open economy macroeconomics perspective Edited by Jean-Oliver Hairault and Thepthida Sopraseuth 28 Fixing Financial Crises in the 21st Century Edited by Andrew G.Haldane 29 Central Banking in Eastern Europe Edited by Nigel Healey and Barry Harrison 30 Exchange Rates, Capital Flows and Policy Edited by Peter Sinclair, Rebecca Driver, and Christoph Thoenissen 31 Monetary Policy and Unemployment The US, Euro-area, and Japan Edited by Willi Semmler Monetary Policy and Unemployment The US, Euro-area, and Japan Edited by Willi Semmler First published 2005 by Routledge Typeset in Baskerville MT by Newgen Imaging Systems (P) Ltd, Chennai, India British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Library of Congress Cataloging in Publication Data Monetary policy and unemployment: the US, Euro-area, and Japan/ edited by Willi Semmler. p. cm. Papers presented at a conference hosted by the Economics Dept. and the Center for Economic Policy Analysis of the New School University held from Nov. 22–23, 2002. Includes bibliographical references and index. 1. Monetary policy—United States—Congresses. 2. Monetary policy—European Union countries—Congresses. 3. Monetary policy—Japan—Congresses. 4. Unemployment—United States— Congresses. 5. Unemployment—European Union countries—Congresses. 6. Unemployment—Japan—Congresses. I. Semmler, Willi. II. New School Universiy. Economics Dept. III. New School University. Center for Economic Policy Analysis. HG540.M659 2005 331.137–dc22 2004013132 This project was generously supported by the Bernard Schwartz Center for Economic Policy Analysis at New School University Copyright © 2005 Willi Semmler for selection and editorial matters; individual chapters, the contributors First issued in paperback 2012 ISBN 13: 978-0-415-65025-0 (pbk) ISBN 13: 978-0-415-70087-0 (hbk) Routledge is an imprint of the Taylor & Francis Group, an informa business 2 Park Square, Milton Park, Abingdon, Oxon OX14 4RN 711 Third Avenue, New York, NY 10017, USA Published 2017 by Routledge The Open Access version of this book, available at www.tandfebooks.com, has been made available under a Creative Commons Attribution-Non Commercial-No Derivatives 4.0 license. Contents List of contributors ix 1 Introduction WILLI SEMMLER 1 PART I Overview: unemployment and monetary policy in the three currency areas 6 2 Monetary policy and unemployment OLIVIER BLANCHARD 8 3 Some notes on monetary policy and unemployment EDMUND PHELPS 14 4 The long stagnation of the Japanese economy during the 1990s and macroeconomic policies HIROSHI YOSHIKAWA 18 PART II Labor market institutions and unemployment 21 5 The role of shocks and institutions in the rise of European unemployment: the aggregate evidence OLIVIER BLANGHARD AND JUSTIN WOLFERS 23 6 Labor market institutions and unemployment in Europe: a comment on Blanchard and Wolfers DAVID R.HOWELL 56 7 Labor market dynamics in the Euro-area: a model-based sensitivity analysis ALISTAIR DIEPPE, JÉRÔME HENRY, AND PETER McADAM 63 PART III Structuralist causes of unemployment and monetary policy 101 8 The structuralist perspective on real exchange rate, share price level, and employment path: what room is left for money? HIAN TECK HOON, EDMUND PHELPS, AND GYLFI ZOEGA 103 the Japanese economy Among the main factors, as the authors point out, is the increased uncertainty In a stochastic version of a model with multiple equilibria, they show that the economy got stuck, due to the rise of uncertainty in a bad equilibrium. In the chapter by Peter Flaschel, Gang Gong, and Willi Semmler, it is shown, using the example of the German economy that both the dynamics of the economy as well as monetary policy rules are significantly impacted and constrained by the exchange rate system. The core of their model is an estimated open economy price and wage Phillips curve for Germany which allows evaluating the different monetary policy rules and their success to impact employment and inflation in the open economy context. In Part IV monetary policy rules and fiscal policy are discussed more specifically Charles Goodhart describes the institutional changes that, in his opinion, have affected both monetary and fiscal authorities in Europe. He argues that monetary policy has to be considered in the context of fiscal policy and macromonetary policies against asset price movements. There were many constraints to effective monetary policy in the Euro-area such as the lack of reputation of the new ECB, the decentralization of fiscal policies, and the absence of real labor mobility across regions. Recent academic studies have proposed direct inflation targeting as a possible optimum solution for the threat of high inflation. Many of these studies suggest replacing traditional rules, based on the control of money growth, by other rules such as the Taylor Rule. The Taylor Rule sets both output and inflation targeting goals for the monetary authorities, although in practice they mostly emphasize the latter. Chapters by Orphanides and Moreno concentrate on these new monetary policy rules and study the new policy rules in action. The last chapter studies the constraints on fiscal policy Marco Buti and Paul Van den Noord, by discussing the currently ongoing tax reform as a policy tool in the Euro-area countries, suggest that there may be a trade-off between efficiency and stability in the Euro-area economies. An increase in economic efficiency—through tax cuts and a reduction in public spending— may lead to a rise in the long run instability of the Euro-area economies. Finally we want to note that frequently the most important cause for a constrained monetary policy which was also initially stressed by the Bundesbank and more recently by the ECB, has been seen in the lack of reputation of the ECB and the threat of high inflation. Yet, as also the chapters in Part I of the volume confirm, there may be considerable risk of deflation rather than inflation in the three currency areas. If this is so, as Olivier Blanchard in his contribution in Part I argues, this appears as a new challenge to monetary as well as fiscal policies in the three currency zones. Monetary policy and unemployment 4 Part I Overview Unemployment and monetary policy in the three currency areas 2 Monetary policy and unemployment Olivier Blanchard I was asked for my thoughts on monetary policy and unemployment. I shall build on the themes developed at this conference, and do my best to be provocative. 1. Monetary policy can have large and long-lasting effects on real interest rates, and by implication, on activity. What I mean here is really large, and really long lasting, a decade or more. This conclusion is at odds with much of both the recent empirical work and the recent theoretical work on the topic: The large empirical literature based on structural vector autoregressions (VARs) suggests that the effect of an innovation in money on activity peaks after a year or so, and is largely gone within 2 or 3 years. The large theoretical literature based on an equation for inflation derived from Taylor-Calvo foundations gives roughly the same results. A change in money growth has its maximum effect on activity after a year or so, and the effect is again largely gone within 2 or 3 years. Neither literature is totally convincing. The type of money shocks whose effects are traced by VAR impulse responses are deviations from normal monetary behavior, and thus (even if identification is convincingly achieved and these are truly deviations, rather than noise) are likely to have different effects from the nondeviation part of policy The Taylor-Calvo inflation equations have many merits. They capture something essential, namely the staggering of price and wage decisions. They can be derived from microfoundations. They provide a simple and elegant characterization of the relation between inflation and activity But, as we all know, they do not fit the data. There is much more inertia in the behavior of inflation than these equations imply And, taking a step back, I see the evidence on the relation between monetary policy and real interest rates as speaking very strongly and very differently Think of the evolution of ex ante real interest rates (use your favorite measure of inflation expectations to do that; my point is robust to all plausible variations) over the last 30 years in OECD countries: For most of the 1970s, ex ante real rates were very low in most countries. This was due—as a matter of accounting, not in a causal sense—to a large increase in inflation, and a less than one-for-one increase in nominal interest rates. Who can doubt that the evolution of real rates was due to monetary policy? That, faced with an increase in inflation triggered by supply-side shocks, central banks were too slow and too reluctant to increase nominal interest rates, leading to low or even negative real interest rates for a good part of the decade. There may be other interpretations, arguing that the evolution of real interest rates was the result of shifts in investment or saving, and had nothing to do with monetary policy I have not seen a plausible account along those lines. For most of the 1980s, ex ante real rates were high in most countries. This was due, again as a matter of accounting, to a large increase in nominal interest rates, together with a decrease in the rate of inflation. Again, who can doubt that this evolution was primarily due to monetary policy? In every country one can trace the sharp increase in interest rates to an explicit change in monetary policy be it the change under Margaret Thatcher in the United Kingdom in the late 1970s, the Paul Volcker disinflation in the United States in the early 1980s, the competitive disinflation strategy in France a few years later. The case can also be made a contrario: The experience of Germany with a much more stable monetary policy and little change in real interest rates, either in the 1970s or the 1980s, reinforces the argument. Again, there may be plausible nonmonetary accounts for these high real rates (Here, for the sake of internal consistency I must mention one, that I explored in a paper with Larry Summers in the mid-1980s, in the face of the joint increase in interest rates and stock prices: An increase in anticipated profitability increasing present values and putting pressure on long real rates. I still believe that this was a relevant factor. But I also believe that much of the evolution of real interest rates in the United States during the decade had to do with monetary policy). If we accept those two facts, we must reach the conclusion that, while money is eventually neutral, and the Fisher hypothesis holds in the long run, it takes a long time to get there. (This was indeed Milton Friedman’s view.) But, if we accept the fact that monetary policy can affect the real interest rate for a decade and perhaps more, then, we must accept, as a matter of logic, that it can affect activity be it output or unemployment, for a roughly equal time. (Maybe one can think of models where the real rate returns to the natural real rate slowly but output returns to its natural level faster. The models we use imply that the two should return to their natural level at roughly the same speed.) In short, monetary policy is potentially much more powerful (although we may not want to use that power) than is often assumed in current debates. 2. Monetary policy affects both the actual and the natural rate of unemployment. The first part of the proposition is obviously not controversial. But, studying the evolution of European unemployment, I have become convinced that the second part is also true, that monetary policy can, and does, affect the natural rate of unemployment: Monetary policy and unemployment 9 Again for the sake of internal consistency let me start with a channel I explored, again with Larry Summers, in the late 1980s, namely hysteresis. There, we argued that anything that increased the actual rate of unemployment for sufficiently long—such as, for example, a sustained increase in real interest rates induced by monetary policy—was likely to lead to an increase in the natural rate. Our original explanation, that the goal of those employed was simply to keep their jobs, not create jobs for the unemployed, was too crude. It ignored the pressure that unemployment puts on wages, even when bargaining is only between employed workers and firms. But, even if full hysteresis (a unit root) is unlikely one can think of many channels, from the unemployed given up search, to the unemployed losing skills, to endogenous changes in labor market institutions, which imply that sustained high unemployment will lead to an increase in the natural rate itself. Sadly I must admit, I still do not have a good sense today of how important this channel really is. A much more conventional channel for the effects of real rates on the natural rate is through capital accumulation. Real interest rates affect the cost of capital; the cost of capital affects capital accumulation; the capital stock affects the demand for labor; the demand for labor affects unemployment. For all this to be of relevance for monetary policy monetary policy must be able to affect real interest rates for a long period of time. But this is the point I just argued earlier was also true. I believe that this mechanism plays an important role in accounting for the history of unemployment in Europe over the 30 years. Low real interest rates in the 1970s probably partly mitigated the increase in labor costs on profit, limiting the decline in capital accumulation, and thus limiting the increase in the natural rate of unemployment in the 1970s. High real interest rates in the 1980s (and then again, as a result of the German monetary policy response to German reunification, in the early 1990s) had the reverse effect of leading to a larger increase in the natural rate of unemployment during that period. And the decrease in real interest rates since the mid-1990s is probably contributing to the slow decline in unemployment in Europe. Are there other mechanisms at work? The real business cycle has focused on effects of the real interest rate on labor supply Ned Phelps has focused on the effects of the real interest rate on the markup of firms. My sense is that interest rate induced movements in the markup may be of relevance, but the capital accumulation channel strikes me as more obvious, and probably more important. A detour here on an exotic but perhaps important labor supply channel. I have been struck in the recent past by the (so far anecdotal) evidence on the effects of stock market movements on retirement decisions. In an economy in which most people have defined contribution plans, and in which there is no mandatory retirement age (both conditions are necessary and are satisfied in the United States), a decrease in the stock market appears to lead many older workers to continue working, so as to maintain Monetary policy and unemployment 10 their desired level of consumption during retirement. The recent stock market decline has not been due to high interest rates. But the logic would be the same if it had. It may well be that we have moved to an economy in which increases in the interest rate lead to a fall in asset prices, and, in response, an increase in the participation rate of older workers. A last point here, on the relation between unemployment and inflation. The implication of the earlier argument is that a sustained increase in real interest rates leads first to an increase in the actual unemployment rate (the usual aggregate demand effect) and later, as capital accumulation decreases, to an increase in the natural rate itself. If we think of the pressure on inflation as depending on the difference between the actual and the natural unemployment rates, then, as the natural rate increases, the pressure on inflation from a given unemployment rate will decrease over time. In other words, sustained tight money may have less and less of an effect on inflation over time (the same argument applies if hysteresis, i.e. some effect of the actual rate on the natural rate, is at work). 3. The ECB failing is in its words, not in its deeds. But words matter very much. ECB bashing is a popular sport, especially on this side of the ocean. I am not sure it is justified. The ECB, like many other Central Banks, has adopted inflation targeting (the other pillar, M3, is mostly for show). Inflation has remained for most of the period above the ECB target, so it is no surprise that the ECB has not embarked on the same kind of drastic interest rate cuts as the Fed over the past few years. There is, however, an irony to the use of inflation targeting To noneconomists—that is, to most economic agents, from consumers to firms—inflation targeting as the exclusive goal of Central Bank policy sounds heartless: How can the Central Bank put no weight on output stabilization? In fact, as we (economists) know, inflation targeting is actually an activist policy a commitment by the Central Bank to keep output close to its natural level, and so unemployment close to the natural rate: If inflation is kept close to the target, expected inflation will be close to inflation, and so, by the definition of the natural rate—that unemployment rate such that actual and expected inflation are the same—unemployment will be close to the natural rate. The problem of the ECB is not therefore with the policy it has followed. But it is with the way it has sold it to the public. Its public relations have been dismal. It has not explained what inflation targeting actually did, how it was as much of a commitment to help Euro economies get out of a recession, as to fight inflation. Worse than that, it has been ambiguous about the symmetry of the target, and thus about its commitment to decrease interest rates if inflation became low. (Compare the rhetoric of the ECB to the careful explanations given by the Central Bank in the United Kingdom. The policies are much more similar than the words.) The issue is that not only policy but also public relations matter very much: They shape expectations, which in turn determine spending, and output. Today in Europe, the private sector feels that it is very much on its own. It is not sure the ECB will help if the slump continues. It is not sure, given the constraints imposed by the Stability and Growth Pact, that fiscal policy can help. I suspect this explains, in part, the pessimism which permeates Europe at this point, and in turn contributes to the current slump. The contrast with US policy could not be stronger. Monetary policy and unemployment 11 4. Europe could easily fall in the liquidity trap. I worry very much about the liquidity trap. Ten years ago, we thought of this as an exotic case. Japan has shown it could happen. Japanese economic policy bashing is also a popular sport, and it strikes me also as largely unwarranted. Japanese policy was not that crazy for most of the 1990s. Interest rates were decreased, in retrospect a bit too slowly Expansionary fiscal policy was used, admittedly with ebbs and flows, but who would not be scared about running such large deficits for so long? The hope was that, with a turn-around in the economy asset prices would recover, and balance sheets of banks would improve. These hopes did not pan out, but how many of the current critics predicted this outcome in the early 1990s? (A major question is why this fiscal cum money expansion is insufficient to avoid getting into the trap. I do not know the answer.) I am also unconvinced by a number of recent papers arguing that, under existing policies, this is unlikely to happen elsewhere. I think the same set of events could well happen again. Economies which try to aim for very low inflation (0–2 percent), and put sharp constraints on fiscal policy are playing with fire. Let me sketch a scenario on which I put positive probability The current account of the United States is very large. It is absorbing about 30 percent of non-US world net saving, and this will not last forever. When foreign capital flows slow down (and they will) the current account will have to decrease, the dollar will have to depreciate. And the only currency it can really depreciate against is the Euro. My sense is that macroeconomic policy in the Euro zone is not ready to react to a major appreciation of the Euro. The room for monetary policy is small, the room under current Pact rules for fiscal policy equally limited. The risk of going to the two limits and still being in a recession with deflation strikes me as substantial. What is there to do? The usual and unsastisfactory response: Europe should not have gotten there in the first place. I believe that a 2 percent inflation target, and the associated 4 or 5 percent nominal interest rate are too low, leaving too little room to decrease interest rates if needed. The second answer is: Beware of analogies. Analogies are only pseudologic, and pseudologic can be dangerous. A really dangerous analogy is “Keep your powder dry.” Central banks should do precisely the opposite: Try by all means to avoid getting into the trap. When close to it, do more rather than less. The third is: Think harder about how to use fiscal policy. This takes me to my last point. 5. We need to rethink fiscal policy and redesign automatic stabilizers. Discussions of fiscal policy suffer from schizophrenia: We all seem happy to accept variations in the budget due to automatic stabilizers. The argument for allowing the automatic stabilizers to operate is indeed a convincing one. They allow for countercyclical fiscal policy but avoid the dangers of discretionary fiscal policy Because of their automatic nature, they are more likely to avoid the perverse effects—the negative fiscal multipliers—that appear to characterize some discretionary fiscal expansions. What automatic stabilizers a country has however, and how strong they are, is entirely based on past decisions that typically gave no weight to Monetary policy and unemployment 12 output stabilization. A country with a more progressive income tax structure has stronger stabilizers. Was this intended? Almost surely not. Clearly if we like automatic stabilizers, we should not be blindly accepting what history left us, but thinking hard instead about how to design the tax/transfer system so as to achieve the optimal degree of optimal automatic stabilization (an argument made recently by Martin Feldstein, that I strongly second). Our profession is nearly silent on this issue, and I believe we can do much better. Many of the things that monetary policy does could be done by fiscal policy This will be most useful if the economy is in a liquidity trap, but may be useful even away from it. This is not to say fiscal policy can do everything monetary policy can do. Suppose, for example, that you want to decrease the cost of capital. This can be done through expansionary monetary policy It can also be done through fiscal policy and interest rate subsidies. The problem, however, is that the cost to the budget is likely to be enormous. Suppose you want to decrease the cost of borrowing on mortgages by 1 percent. You can do this through expansionary monetary policy and a decrease in the appropriate long real rate. While the effect will be initially on flows, refinancing, if sufficiently attractive, will eventually lead to an effect on the whole stock. Or you can do it through a 1 percent tax subsidy for existing mortgages. Mortgages outstanding in the United States at this point are around 6 trillion dollars, so the subsidy will be equal to roughly 60 billion dollars. This is a very large number, and if this is to be a balanced budget, requires a large increase in taxes elsewhere. Much better to leave this to monetary policy (The size of the transfers between borrowers and lenders are exactly of the same magnitude under monetary policy But they are stealthy and do not explicitly involve the budget.) There are however fiscal policy instruments, which can have a strong effect on spending at a much lower cost to the budget. Conceptually they are those that lead firms or consumers to shift spending over time, and work through intertemporal substitution. The best known example here is that of the investment tax credit. Starting from an example from Sweden, John Taylor wrote a beautiful Brookings paper 20 years ago, showing how such a cyclical investment tax credit could be put in place and used to smooth fluctuations. We should explore it again, together with other cyclical tax credits, on consumer durables for example. In the new monetary policy environment, choosing automatic stabilizers optimally and having a fiscal policy that responds quickly and strongly to movements in activity is a high priority Acknowledgments Remarks made at the Conference “Monetary policy and the labor market. A conference in honor of James Tobin,” held at the New School, November 2002. I thank Francesco Giavazzi and Justin Wolfers for comments. Monetary policy and unemployment 13 Finally let me make a comment on fiscal policy Here the scope is very limited because debt outstanding is 140 percent relative to GDP. Still many of us think that the type of tax cut that Professor Blanchard proposed in his presentation—tax cuts that would push investment such as the investment tax credit—appears to be promising. In fact, probably within the next month, the Japanese government will announce a tax cut of this kind. Monetary policy and unemployment 20 Part II Labor market institutions and unemployment 5 The role of shocks and institutions in the rise of European unemployment The aggregate evidence Olivier Blanchard and Justin Wolfers Figure 5.1 shows the evolution of unemployment in Europe since 1960. The figure plots average unemployment rates over 5-year intervals, starting in 1960, both for the OECD- Europe as a whole (the line) and for 15 individual OECD-Europe countries.1 It shows the increase in the overall unemployment rate, from 1.7 percent in the early 1960s to 11.0 percent in the mid-1990s, together with the large dispersion in unemployment rates across countries, from 4.0 percent in Switzerland to more than 20 percent in Spain in the mid-1990s. Explanations for these evolutions fall into three classes: • Explanations that focus on the role of adverse economic shocks. Adverse shocks can indeed increase the unemployment rate, at least for some time. And there are many plausible candidates for such adverse shocks over the last 30 years. As unemployment started rising in the 1970s, the focus was on oil price increases and the total factor productivity (TFP) growth slowdown. Since then, the evolution of the real interest rate, and other shifts in labor demand have been added to the list. Explanations based solely on shocks run, however, into a major empirical problem. Shocks can potentially explain the general increase in unemployment over time. But, as we shall see, they do not differ enough across countries to explain the cross-country variation so evident in Figure 5.1. • Explanations that focus on the role of adverse labor market institutions. Labor market institutions affect the nature of unemployment, and some can indeed potentially generate a high unemployment rate. With the persistence of high unemployment now for more than two decades, explanations based on adverse institutions (“labor market rigidities”) have become steadily more popular. Explanations based solely on institutions also run, however, into a major empirical problem: many of these institutions were already present when unemployment was low (and similar across countries), and, while many became less employment-friendly in the 1970s, the movement since then has been mostly in the opposite direction. Thus, while labor market institutions can potentially explain cross-country differences today they do not appear able to explain the general evolution of unemployment over time. Figure 5.1 Unemployment Rate, E15. Note Line links average unemployment rate for the E15. Mnemonics are listed in Note 1. • Explanations that focus on the interaction of adverse shocks with adverse market institutions. Some institutions may affect the impact of shocks on unemployment. For example, better coordination in bargaining may lead to a faster adjustment of real wages to a slowdown in productivity growth. Some institutions may affect the persistence of unemployment in response to shocks. For example, if labor market institutions lead to a labor market with long unemployment duration, adverse shocks are more likely to lead some of the unemployed to become disenfranchised, reducing the pressure of unemployment on wages, thereby slowing, and possibly even halting the return to lower unemployment. It is easy to see what makes this third class of explanations attractive. It has the potential to explain not only the increase in unemployment over time (through adverse shocks), but also the heterogeneity of unemployment evolutions (through the interaction of the shocks with different labor market institutions). In an earlier work (Blanchard 1999), we took stock of the underlying alternative theories. We looked at whether and how different shocks and different institutions may affect the unemployment rate. We looked at the channels through which shocks and institutions might interact. This led us to argue in favor of the third class of explanations. In this chapter, we look at the aggregate empirical evidence more formally, at the role of shocks, institutions, and interactions, in accounting for the evolution of European unemployment. Monetary policy and unemployment 24 To do so, we look at the data through two panel data specifications. In the first, we assume unobservable but common shocks across countries. In the second, we construct and use country-specific time series for a number of shocks. In both specifications, we allow for an interaction between shocks and institutions: The effect of a given shock on unemployment is allowed to depend on the set of labor market institutions of the country. We see the results as surprisingly (at least given our priors) good: specifications that allow for shocks, institutions, and interactions can account both for much of the rise and much of the heterogeneity in the evolution of unemployment in Europe. The magnitudes of the effects of the shocks on unemployment are plausible. The magnitudes of the effects of institutions are equally so. And their interactions explain much of the difference across countries. These results notwithstanding, three caveats are in order. First, the results are preliminary In many cases, we do not have time series for institutions, and the series we have may not be very good. Second, the results are typically weaker when we allow for time-varying, rather than time-invariant, measures for institutions. This gives some reasons to worry. Last, the fact that the specifications fit the data does not prove that the underlying theories are right; just that they are not obviously inconsistent with the aggregate data. We believe we are the first to analyze the panel data evidence looking simultaneously at shocks, institutions, and interactions. But we build on a large number of previous studies. Bruno and Sachs (1985) were among the first to emphasize both shocks and institutions in the initial rise in unemployment. An empirical attempt to explain UK unemployment as a result of shocks, institutions, and interactions was presented by Layard et al. (1991) in their book on unemployment. Two recent influential studies are by Phelps (1994) and by Nickell (1997). We differ mostly from Phelps by allowing for institutions, and for interactions. We differ mostly from Nickell by allowing for observable shocks, and by having a panel data dimension going back to the 1960s. Our results are partly consistent with those of Phelps with respect to shocks, and largely consistent with those of Nickell with respect to institutions. Our chapter is organized as follows: The first section looks at shocks, both across countries and over time. The second section does the same for institutions. The third section discusses potential interactions between shocks and institutions. The fourth section reports the results of estimation under the assumption of unobservable but common shocks across countries. The fifth section reports the results of estimation using country-specific time series for shocks. The final section concludes.2 Shocks Three shocks appear to have played an important role in the increase in European unemployment. (This short declarative sentence conveys more certainty than is justified. Caveats follow.) The role of shocks and institutions in the rise of European unemployment 25 The decline in TFP growth Starting in the early 1970s, Europe suffered a large decrease in the underlying rate of TFP growth. This is shown in Figure 5.2.3 The two lines in Figure 5.2 (a) give the evolution of the average rate of TFP growth for the 15 countries of OECD-Europe (E15 in what follows) and for the five largest European countries, France, Germany, Italy, Spain, and the United Kingdom (E5). To give a sense of the heterogeneity across countries, Figure 5.2(b) gives the evolution of TFP growth in each of the E5 countries. (Showing all 15 countries would clutter the figure but yield similar conclusion.) TFP growth which had been close to 5 percent in the 1960s decreased to 3 percent in the first half of the 1970s, and to 2 percent in the second half of the 1970s. It has remained around 2 percent since then. The decline has affected countries in roughly similar fashion.4 The decrease in TFP growth was initially partially hidden by the large increase in the relative price of oil and other raw materials. Thus, much of the focus of the initial research (e.g. Bruno and Sachs (1985)) was on this increase in relative prices rather than on the slowdown in TFP growth. In retrospect, the slowdown in TFP growth from its unusually high level in the first 30 years after Second World War was surely the most important shock of the period.5 There is no question that a slowdown in TFP growth can lead to a higher equilibrium unemployment rate for some time (we prefer to use “equilibrium rate” rather than “natural rate,” but the meaning is the same). All that is needed is that it takes some time for workers and firms to adjust expectations to the new lower underlying rate, leading to wage growth in excess of productivity growth for some time. Gan the effects of such a slowdown on unemployment be permanent? Theory suggests that the answer, to a first approximation, is no. Once expectations have adjusted, the effect on unemployment should mostly go away There lies the first puzzle of European unemployment. The initial shock is clearly identified. But, after more than 20 years, it is hard to believe that its effects are not largely gone. So, what accounts for today’s high unemployment? There is much less agreement here, but two other shocks appear relevant. The real interest rate Figure 5.3(a) gives the evolution of the average real interest rate for both the E15 and the E5. Figure 5.3(b) gives the real interest rate for each of the E5 countries.6 Figure 5.3 shows that, both for the E15 and E5 countries, the real rate turned from positive in the 1960s to sharply negative in the second half of the 1970s, and then to large and positive in the 1980s and the 1990s. For some countries, the decline in the 1970s was nearly as dramatic as the ensuing increase. Figure 5.3(b) shows how the real rate in Spain went down from 2 percent in the 1960s to −5 percent in the mid-1970s, back to 5 percent in the 1980s and the 1990s. For others, such as Germany the real rate has remained much more stable. Monetary policy and unemployment 26 Figure 5.2 TFP Growth (a) E15 and E5; (b) E5. The role of shocks and institutions in the rise of European unemployment 27 (a) 0 E15 (simple aver age ) fl. ES (s i mp le average) 0.06 .c ~ e 0> 0.04 0. u. >- 0 ., 1! 0.02 ;;; " c c < 0 · 1960 1 970 19 80 1990 2000 5-year peri od (b ) 0.1 .c "" ~ 'Ill.' e '"' - 0> 0. 05 &!!l 0. ,. u. ""' -a:u lit. >- -. ,.. 0 <illR 1m ""' lim ""' ~ .,., Iii: - """ "" ;;; 0 "' r:JOJ .... " - c c < -0.05 19 60 1970 1980 1990 2000 5-year peri od Why might such changes in the real interest rate affect the equilibrium unemployment rate?7 Because they are likely to affect capital accumulation, and so, at a given wage (and thus a given ratio of employment to capital), to shift labor demand. Are the effects on unemployment likely to be permanent? Theory is largely agnostic here. Again, a plausible answer is that long run effects, if present, are likely to be small. It is clear from Figure 5.3 that the pattern of interest rates may help explain why unemployment kept increasing in the 1980s, even as the effects of lower TFP growth on unemployment were—presumably—declining. This suggests that, had real interest rates been stable, unemployment would have been higher in the 1970s, and lower in the 1980s. Put another way the low real interest rates of the 1970s delayed some of the increase in unemployment by a decade or so. The higher real interest rates since the early 1980s may help explain why unemployment has remained high in the 1980s and the 1990s. Figure 5.3 Real interest rate. (a) E15 and E5; (b) E5. Monetary policy and unemployment 28 Shifts in labor demand Figure 5.4 gives the evolution of the log of the labor share for both the E15 and the E5 (normalized to equal zero in 1960). For both groups of countries, the evolution of the share is quite striking. After increasing in the 1970s, the labor share started decreasing in 1980s and the decline has continued since then. For the E5, the labor share is now 10 percent lower than it was in 1960; for the E15, it is 8 percent lower. Why look at the evolution of the labor share? Suppose that technology were characterized by a Cobb-Douglas production function, both in the short and the Figure 5.4 Log labor share, E15 and E5. long run. The decrease in the share since the 1980s would then reflect either technological bias away from labor—a decrease in the coefficient on labor in the production function—or a decrease in the wage relative to the marginal product of labor. In either case, the implication would be an adverse shift in labor demand and thus a potential source of unemployment in the 1980s and 1990s.8 The elasticity of substitution may be equal to one in the long run, but is surely less than one in the short run. In that case, movements in the share will also reflect the dynamic response of factor proportions to factor prices. Indeed, much of the increase in the labor share in the 1970s surely reflects the effects of the increase in the real wage relative to TFP growth together with a low short run elasticity of substitution, and some of the decrease since then reflects the adjustment of proportions over time. In Blanchard (1997), we argued, however, that more has been at work than the adjustment of factor proportions to factor prices, and that the large decline in the share reflects a genuine adverse shift in labor demand. The role of shocks and institutions in the rise of European unemployment 29 converged from a low maximum replacement rate to the European average) it has decreased since then. In other words, the worst excesses have been largely eliminated. This may be more important than changes in the average replacement rate. • Putting together series on employment protection is difficult. We have taken a first step by constructing series based on recent work by the OECD (see OECD Figure 5.7 Replacement Rates, E5: (a) OECD; (b) Maximum. Monetary policy and unemployment 36 (a) 40 I -- "" £ !!! ., E ' oru ~ .... . .. 30 '"" "' ""' C<!J ow E ""' C<U 8 ""' ""' -- \'IJi - !ll "" <m a. .... ~ 20 ""' "" 2:' ., -- ., E E .,., " } "' 0 () w 0 "' "' "" "' "' ' 1960 1970 1980 1990 2000 5· year period (b) 100 .,. ~ ESP .,., E "" ""' - .... "' --- E om ~ - - ., - <.> "" .!l! 50 ""' a. §1!6 ~ """ -"" OE\I cw "" "" E ""' -- " "' "" E .,., " i;j "" ::; "' "" "" 0 1960 1970 1 980 1 990 2000 5· year peri od (1999)), as well as on earlier work by Lazear (1990). Details of construction are given in the appendix. There are a number of reasons why these series are at best rough approximations to the evolution of employment protection. In particular, the OECD data, which we use to construct the measures from 1985 on, are based on a much broader set of dimensions of employment protection than the Lazear series (notice period and severance pay for a blue collar worker with 10 years seniority), which we use to construct the series before 1985. This caveat notwithstanding, Figure 5.8 shows the evolution of the employment protection index for the E5 countries since 1960. (The figure for the E15 would be harder to read, but yield similar conclusions.) Note again the diversity of evol utions, and the lack of a simple answer. Figure 5.8 Constructed employment protection index, E5. Spain and Italy appear to have had high employment protection throughout. Employment protection in Spain was high even under Franco, before unemployment increased. In both countries, employment protection has decreased since the mid-1980s–in Spain, largely because of the development of fixed term contracts rather than the weakening of protection for workers on indefinite contracts. In France and Germany, employment protection was low to start with, then increased in the late 1960s and early 1970s, and has been roughly stable since then.14 To conclude: There is enough heterogeneity in labor market institutions within Europe to explain potentially, differences in unemployment rates today. As to the evolution of institutions over time, it is clear that neither the view that labor market institutions have been stable through time, nor the view that labor market rigidities are a recent development are right. Some countries have had these institutions for a long time, others have acquired them more recently. There clearly was The role of shocks and institutions in the rise of European unemployment 37 an increase in employment-unfriendly institutions in the late 1960s and early 1970s. Since then, there appears to have been a small but steady decline. Interactions Our review of facts makes clear why it is tempting to look for explanations of the rise of European unemployment based on the interaction of shocks and institutions: Adverse shocks can potentially explain the general increase in unemployment. Differences in institutions can potentially explain differences in outcomes across countries. This is indeed the direction that has been explored in much of the recent research on unemployment. This section gives a brief assessment of the current state of knowledge.15 One can think of labor market institutions as shaping the effects of shocks on unemployment in two ways. First, they can affect the impact of shocks on unemployment. Second, they can affect the persistence of unemployment in response to shocks. Most of the initial research explored the first direction, focusing on how the nature and the details of collective bargaining might determine the response of unemployment to various shocks.16 It pointed, for example, to the importance of indexation clauses in labor contracts. It also pointed to the potential importance of the level and the structure of collective bargaining: it might be easier, for example, to achieve a slowdown in wage growth in response to a slowdown in productivity growth if bargaining takes place at the national rather than the firm or sectoral level—where aggregate trends may be less well perceived and understood, and coordination of the slowdown may be more difficult to achieve. As unemployment remained high, the research shifted to how labor market institutions might also explain the persistence of unemployment in response to shocks.17 The general idea is as follows. Take an adverse shock which leads to higher unemployment. The normal adjustment mechanism is then for unemployment to put downward pressure on wages until unemployment has returned to normal. To the extent that some labor market institutions reduce the effect of unemployment on wages, they will increase the persistence of unemployment in response to shocks. Research has identified a number of such channels. Here is a nonexhaustive list: • A rise in unemployment typically comes with higher unemployment duration (rather than higher flows in and out of unemployment). If some of the unemployed remain unemployed for a long time, they may either stop searching or lose skills. Indeed, the two factors reinforce each other: if firms perceive the long-term unemployed as more risky they may be reluctant to hire them, decreasing the incentives of the long-term unemployed to search for a job. But if they are not actively searching or employable, these unemployed workers become irrelevant to wage formation. Firms do not consider them as competition. Employed workers do not see them as competition. The pressure of unemployment on wages decreases, and unemployment becomes more persistent. Layard and Nickell (1987) were the first to point to the potential macroeconomic relevance of such duration dependence. Why should institutions matter in this context? It is because of their effect on the average duration of unemployment. A well documented fact about European Monetary policy and unemployment 38 labor markets is that, probably because of institutions such as more generous benefits and employment protection, a given unemployment rate is associated with much longer duration than in the United States.18 And the longer the average duration of unemployment to begin with, the more likely the effects mentioned earlier are to play an important role. If an increase in the unemployment rate from 5 to 10 percent is associated with an increase in unemployment duration from 3 to 6 months, few of the unemployed will become long-term unemployed. If instead, the same increase in the unemployment rate implies an increase in duration from 1 to 2 years, then disenfranchising effects are much more likely to be important. • Higher unemployment falls unevenly on different groups in the labor market. In most countries, higher unemployment tends to fall disproportionally on the youngest workers and the less educated. Labor market institutions affect the compostion of the unemployed, thus affecting the effects of unemployment back on wages. For example, a high minimum wage can both increase the effect of adverse shocks on the unemployment rate of the less-educated workers, and—because the minimum wage is fixed—reduce the effect of unemployment on wages. Collective bargaining, to the extent that it reflects primarily the preferences and the labor market prospects of prime-age workers, may also lead to little response of wages to youth unemployment, and thus lead to more persistence in unemployment. • Higher unemployment may lead to a change in norms—an argument developed in particular by Wilson (1987) in the context of urban poverty in the United States, and by Lindbeck in the context of European unemployment (e.g. Lindbeck (1995)). As long as unemployment is low, workers may be largely ignorant of the rules governing unemployment insurance, or there may be a stigma attached to being unemployed. After a period of high unemployment, ignorance is likely to disappear; attitudes vis-à- vis unemployment are likely to change. Thus, countries with a more generous welfare system may end up with higher unemployment, even when the shocks are gone. Other channels have been explored as well: Sargent and Ljundqvist (1995) have explored the effect of unemployment insurance rules on the relation between “turbulence” shocks and equilibrium unemployment. Mortensen and Pissarides (1999) have explored the effect of unemployment insurance and employment protection on the relation between relative demand shifts and equilibrium unemployment. Our understanding of the specific channels and their empirical relevance remains rather primitive. This is still very much work in progress, and there is a need for substantially more theoretical and empirical work. Nevertheless, the general thrust is sufficiently clear for us to explore the potential role of interactions in explaining the evolution of unemployment. This is what we do in the rest of the chapter. Common unobservable shocks and interactions In looking more formally at the data, we proceed in two steps. In this section, we treat shocks as unobservable, but common across countries—in effect we treat them as time effects. In the next, we treat shocks as observable and country-specific. The role of shocks and institutions in the rise of European unemployment 39 Our first specification in this section relies on the set of time invariant measures of institutions used by Nickell (1997).19 The specification we use is the following: (1) where i is a country index, t a (5-year) period index, and j an institution index. The dependent variable, uit, is the unemployment rate in country i in period t. ci is the country effect for country i. dt is the time effect for period t. Xij is the value of institution j in country i (in this first specification, we do not allow for time variation in institutions, so there is no index t). The specification allows for the effects of the common time effects on unemployment to depend on the specific set of labor market institutions of a country This dependence is captured by the parameters bj. The specification of (1) is clearly more a description of the data than the outcome of a tightly specified theory of interactions. It does not distinguish in particular between the effects of institutions on the impact or on the persistence of shocks on unemployment. But it captures the basic hypothesis that, given the same shocks, countries with worse institutions will experience higher unemployment. We estimate this equation using data from 20 countries—the E15 countries listed and examined earlier, plus the United States, Canada, New Zealand, Australia, and Japan. (These countries are clearly important controls for any story about European unemployment.) There seems to be little point in looking at year-to-year movements in institutions or in shocks unless one wants to learn more about dynamic effects, and this would take us too far. So, as in earlier figures, we divide time into eight 5-year periods, from 1960–64 to 1995+ . Following Nickell, we use measures for eight “labor market institutions” (the reader is referred to Nickell (1997) for more details): • Three are measures of different dimensions of the unemployment insurance system: the replacement rate (RR), the number of years over which unemployment benefits are paid (Ben), and a measure of active labor market policies (AIMP). • One is a measure of employment protection (EP). • One is a measure of the tax wedge (Tax). • The last three measure aspects of collective bargaining: union contract coverage (Cov), union density (Den), and (union and employer) coordination of bargaining (Coor). The results of estimation of (1) (by nonlinear least squares) are presented in Table 5.1. All the measures of labor market institutions are defined so that an increase in the measure is expected to increase the effect of an adverse shock on unemployment: the expected sign of each bj is positive.20 Also, all measures of institutions are constructed as deviations from the cross-country mean; this way the time effects gives the evolution of unemployment for a country with mean values for all eight institutions. The results of Table 5.1 are surprisingly strong (relative to our priors). The estimated equation gives the following description of the data: • Estimated time effects account for an increase in the unemployment rate equal to 7.3 percent. That is, the equation implies that, if a country had had mean values for all Monetary policy and unemployment 40 eight institutions, its unemployment rate would have grown by 7.3 percent over the period. Table 5.1 Time effects interacted with fixed institutions (1) Coefficients (2) Range of independent variable (3) Implied range of effect of shock (mean=1) Time effects* 7.3% RR 0.017 (5.1) −46.3 32.6 0.21 1.55 Ben 0.206 (4.9) −2.0 1.6 0.60 1.33 ALMP 0.017 (3.0) −47.2 9.5 0.20 1.16 EP 0.045 (3.1) −9.5 9.5 0.58 1.42 Tax 0.018 (3.2) −17.8 22.2 0.68 1.40 Cov 0.098 (0.6) −1.7 0.3 0.83 1.03 Den 0.009 (2.1) −30.4 39.6 0.73 1.36 Coor 0.304 (5.1) −2.0 2.0 0.40 1.60 Country effects (CE) Yes 0.863 Note * Estimated time effect for 1995+ minus estimated time effect for 1960–64; Column (1): regression results, tstatistics in parentheses; Number of observations: 159. • Coefficients on all eight institutions have the predicted sign: Higher RRs, longer duration of Ben, higher EP, a higher Tax, higher Cov and Den, lead to a larger effect of shocks on unemployment. ALMP and Coor lead to a smaller effect (remember our sign convention in defining each institution). All coefficients, except for the union coverage variable, are statistically significant.21 To give a sense of magnitudes, column (2) gives the range for each institutional measure (recall that these are deviations from the cross-country mean). Column (3) then shows the effect of a given shock for the lowest and highest value of the corresponding institution. The way to read the column is as follows. Take three countries, each with mean values for all institutions except one—say, employment protection (line 5). Take an adverse shock which would raise unemployment by 1 percentage point in the country with the mean value of employment protection. Then, the same shock will have an effect of only 0.58 percentage point in the country with the lowest employment protection, but an effect of 1.42 percentage point in the country with the highest employment protection. The conclusion one should draw from column (3) is, given the existing The role of shocks and institutions in the rise of European unemployment 41 variation in labor market institutions, the range of the effects of institutions on the impact of a given shock on unemployment is roughly similar across institutions. • Not only are the coefficients on institutions plausible, but the model does a good job of explaining the differential evolution of unemployment rates across countries. Figure 5.9 plots the change in the actual and the fitted unemployment rates from 1965–69 to 1995+. The fit is quite good. Interactions between common shocks and different institutions can account Figure 5.9 Actual and predicted change in u, 1995+ over 1965–69. for much of the actual difference in the evolution of unemployment rates across countries. (Recall that a pure time effect model with no interactions would predict no variation in predicted unemployment rates across countries: all the points would lie on a horizontal line.) • Another way of thinking about these results is as follows. Consider a model with unobservable shocks and unobservable institutions—equivalently a model with time, country and interacted time and country effects: uit=ci+dt (1+bi)+eit (1′) • Equation (1) can then be thought of as imposing the restriction that bi be a linear function of country i’s institutions: bi=ΣXijbj. This raises the question of how much better we would do if we did not impose this restriction and estimated (1′) instead. One way to answer the question is to look at two . The from estimation of (1′) is 0.903, compared to 0.863 in Table 5.1. The from a second state regression Monetary policy and unemployment 42 of the estimated on labor market institutions Xijs is 0.57. We read these results as saying that (1) the statistical description of the evolution of unemployment as the result between shocks and institutions has the potential to give a good description of the data (as reflected in the first stage ), and that (2) labor market institutions do a good job of explaining country interaction effects (as reflected in the second stage ). In short, (1) gives a good description of the heterogeneity of unemployment evolutions, as the result of interactions between shocks and institutions. These results are indeed consistent with the two cross-sections estimated by Nickell, and show that his results are robust both to the use of a longer time period and the introduction of country effects.22 One must worry however that these results are in part the result of research Darwinism. The measures used by Nickell have all been constructed ex-post facto, by researchers who were not unaware of unemployment developments. When constructing a measure of employment protection for Spain, it is hard to forget that unemployment in Spain is very high…. Also, given the complexity in measuring institutions, measures which do well in explaining unemployment have survived better than those that did not. Thus, in the rest of this section, we look at robustness. Dropping institutions, countries, or country fixed effects To give a sense of robustness with respect to the set of institutions, column (1) in Table 5.2 reports the results of eight separate regressions, each regression allowing interactions with only one of the eight measures for institutions. When introduced on their own three measures are highly significant: Ben, EP, and Cov (which is insignificant in the multivariate specification). In contrast, the RR, which is highly significant in the multivariate specification, is insignificant when introduced alone. Another strategy is to see what happens when we drop one institution at a time. The results (not reported) indicate that the coefficients reported in Table 5.1 are robust to such a variation. Table 5.2 Time effects interacted with fixed institutions: alternative specifications (1) Institutions entered individually (2) No country effects Time effects 7.1% RR 0.004 (1.0) 0.017 (4.1) Ben 0.268 (6.6) 0.213 (4.1) ALMP 0.007 (1.4) 0.017 (2.4) EP 0.043 (4.0) 0.049 (2.8) Tax 0.012 (2.2) 0.017 (2.4) Cov 0.532 (4.9) 0.049 (0.2) Den −0.002 (−0.5) 0.009(1.8) Coor 0.048 (1.1) 0.301 (4.3) The role of shocks and institutions in the rise of European unemployment 43 CE Yes No 0.797 Note Column (1): each coefficient is estimated using a different regression, allowing interactions between the time effects and the specific institution variable. Column (2): Levels of institutional measures entered, but coefficients not reported. Number of observations: 159. Second, we look at robustness with respect to the set of countries. In general, dropping one country at a time makes little difference to the results (not reported here). The only exception is the importance of Spain in determining the coefficient on EP. When dropping Spain, the coefficient on EP goes from 0.045 in Table 5.1 to 0.015. Third, we look at robustness with respect to the treatment of country effects. Column (2) in Table 5.2 reports the results of estimation of (1), replacing country effects by the set of (time invariant) measures of labor market institutions for each country. That is, it imposes the constraint that all differences in unemployment rates be explained by differences in institutions; such a constraint is surely too strong, but it is worth seeing how it affects the results. Only the coefficients on interactions are reported in column (2). They are roughly the same as in Table 5.1. The coefficients on the levels of the labor market institutions (not reported) are typically insignificant. The fit is significantly worse than in Table 5.1. Looking at alternative measures of institutions Table 5.3 looks at the implications of using alternative measures for some of the institutions. This is the work-in-progress part of our chapter. Our goal is eventually Table 5.3 Time effects interacted with institutions: alternative measures (1) Alternative RRs (2) Timevarying RRs (3) Alternative EP (4) Timevarying EP Time effects 7.3% 6.2% 7.3% 7.1% (N)RR 0.017 (5.2) 0.017 (4.7) (N) Ben 0.238 (5.6) 0.205 (4.4) (Alt) RR1 0.009 (2.6) 0.007 (2.0) (Alt) RR25 0.009(1.4) 0.019 (2.7) ( N ) 0.014(1.6) 0.005 (0.5) 0.019 (3.2) 0.017 Monetary policy and unemployment 44 ALMP (2.6) (N)EP 0.024(1.4) 0.032 (1.7) (Alt) EP 0.294 (4.3) 0.167 (2.2) (N) Tax 0.016 (2.4) 0.015 (2.1) 0.019 (3.5) 0.021 (3.7) (N) Cov 0.413 (2.1) 0.395 (1.9) 0.085 (0.5) 0.287 (1.8) (N) Dens 0.004 (0.8) 0.000 (0.0) 0.010 (2.5) 0.008 (1.7) (N) Coor 0.272 (4.9) 0.325 (4.5) 0.392 (6.5) 0.361 (5.3) CE Yes Yes Yes Yes 0.824 0.831 0.872 0.857 Note (N) means Nickell measure. Column (1): estimation using time-invariant values of RR1 and RR25, equal to their average values for 1985–89. Column (2): estimation using the time series for RR1 and RR25. Column (3): estimation using the value of EP for the late 1980s. Column (4): estimation using the time series for EP. Number of observations: 159. to construct time series for all eight institutions. So far, we have done so only for RRs and for EP. Columns (1) and (2) report our results using alternative measures for RRs. Columns (3) and (4) report our results using alternative measures for EP. Using the OECD database on RRs for each country since 1961, we construct an alternative set of measures for the generosity of unemployment insurance. The first measure, RR1, is the RR during the first year of an unemployment spell, averaged over all categories. The second, RR25, is the average RR during years 2 to 5 of an unemployment spell, averaged over all categories. Column (1) shows the results of estimation using time-invariant values for RR1 and RR25. For comparisons with the results using Nickell’s measures which apply to the late 1980s and early 1990s, we use the mean value of the two RRs for the period 1985–89. Measures for the other six institutions are the same as in Table 5.1. The fit is a bit worse than in Table 5.1. The two RRs are both individually significant, and jointly highly significant. Coefficients on the other labor market institutions are often less significant than in Table 5.1. In particular, the coefficient on EP is smaller, and less significant. Column (2) shows the results of estimation using time-varying measures for RR1 and RR25. Relative to column (1), the fit, measured by is marginally improved (but is still worse than in Table 5.1). The part of the increase in unemployment due to time effects decreases from 7.3 to 6.2 percent. Coefficients on labor market institutions are largely the same as in column (1). Columns (3) and (4) use the index of employment protection discussed under “Institutions”. In contrast to the Nickell index, which is a ranking of countries and thus The role of shocks and institutions in the rise of European unemployment 45 Table 5.6 looks at alternative measures of institutions. Its structure is the same as that of Table 5.3. Columns (1) and (2) look at the effects of using the two alternative measures of replacement rates using OECD data. Column (1) uses a time-invariant value equal to the average for 1985–89; column (2) uses the time series. Columns (3) and (4) do the same for employment protection. The table suggests two conclusions, both worrisome: Replacing the Nickell measures by alternative, but time-invariant measures, substantially decreases the . Going from the time-invariant to the time-varying measures further decreases the fit. The coefficients on institutions remain consistently positive, but are typically smaller than in Table 5.5, and less significant. These results lead to the same discussion as in “Interactions”: Luck, or data mining, when the standard set of measures is used? Poor time series for institutions, interacting here with the fact that we are looking at their product with time-varying and also imperfectly measured shocks? Or reverse causality (although the fact that the deterioration of fit happens when replacing one timeinvariant measure by another is not supportive of this hypothesis). To conclude, one can indeed give a good account of the evolution of unemployment across countries and times by relying on observable shocks and interactions with labor market institutions. The fact that the results are weaker when using time-varying institutions is worrying. But, again, the results strike us as surprisingly good overall. Conclusions We see our results as preliminary. We see our dynamic specification of the effects of shocks as much too crude. We still need to construct and introduce time series for some labor market institutions. We worry about the endogeneity of labor market institutions. Nevertheless, we believe that the results so far suggest that an account of the evolution of unemployment based on the interaction of shocks and institutions can do a good job of fitting the evolution of European unemployment, both over time and across countries. If our account is correct, one can be mildly optimistic about the future of European unemployment. The effects of some of the adverse shocks should go away The real interest rate is likely to be lower in the future than in the recent past. The dynamic effects of what we have identified as adverse labor demand shifts should eventually prove favorable to employment. Institutions are also slowly becoming employment-friendly Our results suggest that the more favorable macroeconomic environment and the improvement in institutions should lead to a substantial decline in unemployment. Acknowledgments Harry Johnson Lecture. We thank Steve Nickell, Ed Lazear, John Addison, and Paula Adam at the OECD for providing us with some of the data. We also thank Daron Accmoglu, Alberto Alesina, Tito Boeri, Bill Brainard, David Blanchflower, Peter Diamond, Ben Friedman, Jenny Hunt, Larry Katz, Steve Nickell. Andrew Oswald, Steve Pischke, Chris Pissarides, Chris Sims, Betsey Stevenson, and Robert Solow for useful suggestions and comments. We would like to thank Blackwell Publishing Ltd for permission to publish this chapter, which first appeared in The Economic Journal, 110 Monetary policy and unemployment 52 (March 2000). An appendix containing the data, the programs, and describing the construction of the data, is available at http://web.mit.edu/blanchar/www/articles.html. Notes 1 The eight time periods are 1960–64 to 1990–94, and 1995+ (typically 1995–96.) The 15 countries included in OECD-Europe are Austria (AUT), Belgium (BEL), Denmark (DNK), Finland (FIN), France (FRA), Germany (DEU), Ireland (IRE), Italy (ITA), the Netherlands (NLD), Norway (NOR), Portugal (PRT), Spain (ESP), Sweden (SWE), Switzerland (GHE), and the United Kingdom (GBR). Left out are Greece, Iceland, and Luxembourg, for which we could not construct time series for all the explanatory variables used later in the chapter. The unemployment rates are the rates according to national definitions, rather than standardized rates—which typically do not exist back to 1960. (For the period when both unemployment rates exist, using one or the other makes little difference.) Also, while the figures only show what has happened in Europe, the regressions we run later look at all available OECD countries; they include, in addition to Europe, the United States (USA), Canada (CAN), Australia (AUS), New Zealand (NZL), and Japan (JPN). 2 We shall use the existence of the earlier work as an excuse for keeping our discussion of theoretical issues, and of relevant references, to a minimum. 3 We first construct the rate of TFP growth for each year and each country. We do so by computing the Solow residual for the business sector, and then dividing it by the labor share in the sector. Under the assumption of Harrod neutral technological progress—the assumption that allows for steady state growth—this is the right measure of technological progress, and gives the rate at which real wages can grow along the balanced growth path. We then take averages for each 5-year period, for each country. E5 and E15 are constructed as simple (unweighted) averages of TFP growth over countries. 4 Note that, in contrast to the other observations which are based on five yearly observations, the observation for 1995 is typically based on only one year (1995) or two years (1995 and 1996). Thus, one year can make a lot of difference. This is the case for Italy in this figure. 5 An early article on that theme is Grubb et al. (1982). 6 We first compute the real interest rate for each year and each country, as the nominal long rate on government bonds minus a 5-year average of lagged inflation. We then take averages for each 5-year period. 7 The focus here is on the effects on the equilibrium unemployment rate. Changes in the real interest rate also affect the deviation of actual unemployment from the equilibrium rate. We focus on that effect later. 8 Let . Let the ratio of the wage to the marginal produce of labor . µ is equal to 1 under perfect competition in both goods and labor markets, but may differ from 1 otherwise. Then the share of labor α=aµ. A decrease in α reflects a decrease in a or a decrease in µ. Also labor demand can be written as . A decrease in log α leads to an equal decrease in log given output and the wage. This is why we look at the log share. 9 This distinction between Anglo-Saxon and Continental countries is discussed in Blanchard (1997). The differences in evolutions reflects divergence rather than convergence of the share in levels: For the last period (1995+), the labor share in the business sector was 62 percent for France and Spain, vs 70 percent for the United Kingdom and 67 percent for the United States. (The caveat about the dangers of comparing share levels across countries applies.) 10 We first construct the change in inflation (using the business sector GDP deflator for each year and each country. We then take the average for each 5-year period. The role of shocks and institutions in the rise of European unemployment 53 11 A longer discussion is given in our earlier work. A nice theoretical discussion is given by Mortensen and Pissarides (1998). A wider ranging presentation of both theory and facts is given by Nickell and Layard (1998). 12 The steady state unemployment rate is equal to unemployment duration times the flow into unemployment as a ratio to the labor force. Unemployment benefits increase duration, and leave the flow roughly unchanged, increasing the unemployment rate. 13 In addition to the references in these two articles: For a recent comparison of various measures of unemployment insurance, see Salomaki and Munzi (1999). For a recent comparison of measures of employment protection, see OECD (1999), chapter 2. 14 Informal evidence suggests that employment protection was high in France even in the 1960s. Again, this is not reflected in the Lazear measure, and by implication, not reflected in our measure either. This may be an issue for other countries as well. 15 Again, see our earlier work for references, and discussion. 16 This was indeed one of the main themes of Bruno and Sachs (1985). 17 This was the motivation behind the admittedly crude “hysteresis model” of unemployment in Blanchard and Summers (1986). Research since then has shown that while full hysteresis (permanent effects of shocks) is unlikely institutions can lead to high persistence. 18 See for example, the comparison of the labor markets in Portugal and the United States in Blanchard and Portugal (1999). 19 Nickell gives values for these institutions for both 1983–88, and 1989–94. We use the average of the two. 20 Thus, we multiply the original Nickell measures of active labor market policies and of coordination by −1. We take the expected effect of employment protection to be that more employment protection leads to a larger effect of adverse shocks on unemployment, and the expected effect of coordination that more coordination reduces the effects of adverse shocks on unemployment. 21 The t-statistics are computed under the assumption of iid residuals. The residuals show however both spatial and serial correlation, and adjusted t-statistics would probably be lower. 22 There are however some differences between estimated coefficients. In particular: employment protection is signifi cant here, not in Nickell. Union contract coverage is not significant here, but is significant in Nickell. 23 Most theories predict that the interaction of institutions and shocks may be different for different shocks. But allowing for different interactions between each shock and each institutions struck us as asking too much from our limited data set (131 data points for the regressions in this section). 24 The difference between macro and labor panel data regressions is that, in macro, each data point is intimately known by the researcher… 25 If our approach to measuring the equilibrium unemployment rate is right however, then most existing estimates of a, which rely on a much rougher measure of equilibrium unemployment, are not right. We did not take up the task of estimating a in this chapter. 26 In doing so, we are implicitly assuming that the sacrifice ratio is not related to institutions. This is probably incorrect. References Blanchard, O. (1997). “The medium run,” Brookings Papers on Economic Activity, No. 2, 89–158. Blanchard, O. (1999). “European unemployment: the role of shocks and institutions,” Baffi Lecture, Rome, (forthcoming). Monetary policy and unemployment 54 Blanchard, O. and Portugal, P. (1998). “What hides behind an unemployment rate. Gomparing Portuguese and U.S. unemployment,” NBER Working Paper 6636. Blanchard, O. and Summers, L. (1986). “Hysteresis and the European unemployment problem,” in S.Fischer (ed.), NBER Macroeconomics Annual, No. 1, 15–78, Cambridge, MA: MIT Press. Bruno, M. and Sachs, J. (1985). The Economics of Worldwide Stagflation, Oxford: Basil Blackwell. Grubb, O., Jackman, R. and Layard, R. (1982). “Causes of the current stagflation,” Review of Economic Studies 49, No. 5, 707–30. Layard, R. and Nickell, S. (1987). “The labour market,” in R.Dornbusch and R.Layard (eds), The Performance of the British Economy, Oxford: Clarendon Press. Layard, R., Nickell, S., and Jackman, R. (1991). Unemployment; Macroeconomic Performance and the Labour Market, Oxford: Oxford University Press. Lazear, E. (1990). “Job security provisions and employment,” Quarterly Journal of Economics 105–3, 699–725. Lindbeck, A. (1995). “Hazardous welfare-state dynamics,” American Economic Review 85, No. 2, 9–15. Mortensen, D. and Pissarides, C. (1998). “Job reallocation, employment fluctuations, and unemployment differences,” mimeo (forthcoming, Handbook of Macroeconomics). Mortensen, D. and Pissarides, C. (1999). “Unemployment responses to ‘skill biased’ shocks: the role of labour market policy” Economic Journal 109, 1–24. Nickell, S. (1997). “Unemployment and labour market rigidities: Europe versus North America,” Journal of Economic Perspectives 11, No. 3, 55–74. Nickell, S. and Bell, B. (1994). “Would cutting payroll taxes on the unskilled have a significant effect on unemployment?,” presented at CEPR Conference on Unemployment Policy, Vigo, Spain, September. Nickell, S. and Layard, R. (1998). “Labour market institutions and economic performance,” CEP Discussion Paper 407 (forthcoming, Handbook of Labor Economics). OECD (1999). OECD Employment Outlook, OECD, Paris. Phelps, E. (1994). Stmctural Slumps. The Modern Equilibrium Theory of Unemployment, Interest and Assets, Cambridge, MA: Harvard University Press. Salomaki, A. and Munzi, T. (1999). “Net replacement rates of the unemployed. Gomparisons of various approaches,” Economic Paper 133, European Commission, Directorate General for Economic and Financial Affairs. Sargent, T. and Ljundqvist, L. (1995). “The European unemployment dilemma,” WP 95–17, Federal Reserve Bank of Chicago. Wilson, WJ. (1987). The Truly Disadvantaged, Chicago, IL: University of Chicago Press. The role of shocks and institutions in the rise of European unemployment 55 6 Labor market institutions and unemployment in Europe A comment on Blanchard and Wolfers David R.Howell The statistical examination of the links between labor market institutions and unemployment across developed countries is a relatively recent development, pioneered by Stephen Nickell and his colleagues (Layard et al. 1991; Nickell and Bell 1995; Nickell and Layard 1997). Chapter 5 by Blanchard and Wolfers in this volume, previously published as an NBER working paper in 1999 and in the Economic Journal in 2000, has been a hugely influential contribution in what might be termed the second generation of research on this question. It focuses on the interaction of macroeconomic shocks and institutions, expands the time period, and adds time-varying measures of institutions. But most importantly it is cautious—persistently checking the initial results with various robustness tests and questioning the meaningfulness of the regression results given the quality and potential endogeneity of the key institutional measures. I will begin with a brief overview of what may be termed the “first generation” studies, followed by short comments on the Blanchard-Wolfers contribution, and then will turn to more recent work that challenges the mainstream view that cross-country regression research confirms a central role for labor market institutions in the explanation for Europe’s high unemployment in the 1980s and 1990s. The first generation A hallmark of the classical thinking attacked by Keynes in the 1930s was that wage rigidity must explain persistent high unemployment. Sensible economic policy would then consist of promoting labor market flexibility which in turn would entail—according to the general view—attacking those labor market institutions designed to shelter workers from the harshest effects of the competitive marketplace. But it is notable that it was not until the late 1980s that leading economists began to carefully explore the statistical links between “employment-unfriendly” labor market institutions and the cross-country pattern of unemployment for developed (OECD) countries. The explanation for this apparent anomaly is easily found—until the 1970s, countries rich in strong, highly “unfriendly” institutions, such as labor unions, unemployment benefit systems, and employment protection laws, were consistently outperforming the more Figure 6.1 Standarized unemployment rates for OECD countries, 1960–2002 (second quarter). Sources: Baker et al. (1960–99) “5-year unemployment rates,” Appendix 2 (see Chapter 1 of this volume). Unemployment 2000: OECD Employment Outlook, July 2002. Unemployment 2002: OECD online (http://www.oecd.org/). Medians/standard deviations: and author’s calculations. laissez-faire countries. As Figure 6.1 shows, the US unemployment rate was well above the median for these 19 OECD countries as recently as 1980–84. But as employment performance across much of Europe worsened and the crosscountry pattern of unemployment became more aligned with classical theory which had experienced a considerable resurgence in the 1970s, leading economists turned their attention to the links between institutions, rigidities, and unemployment (Blanchard and Summers 1986; Lindbeck and Snower 1988). A statistical demonstration of the classical rigidity story required consistent, meaningful measures of the key labor market institutions, the development of which was greatly advanced by Stephen Nickell and several colleagues (Layard et al. 1991; Nickell and Bell 1995; Nickell and Layard 1997). In what is perhaps his most prominent paper in this area, Nickell (1997) examined the link between institutions and unemployment with a sample of 20 OECD countries for two 6-year periods, 1983–88 and 1989–94, and found strong support for the conventional wisdom—union coverage, unemployment benefits, and employment protection all Labor market institutions and unemployment in Europe 57 substantially increased the unemployment rate. But notably two other institutions had strong “good” effects: according to Nickell’s results, bargaining coordination and active labor market policies both tended to reduce unemployment. In sharp contrast to another paper in the same symposium, which argued that labor market rigidities alone accounted for high European unemployment (Siebert 1997), Nickell’s (1997) conclusion was cautious: “the broad-brush analysis that says that European unemployment is high because European labor markets are too ‘rigid’ is too vague and probably misleading.” Blanchard and Wolfers With the late 1990s, a number of empirical studies appeared that improved upon Nickell’s original institutional measures, added others, changed the time period covered, and experimented with the specification and econometric method (see Baker et al. 2002, for a detailed review of this literature). The high water mark of this second generation was, in this author’s view, the Blanchard-Wolfers chapter that appears in this volume. Building on Blanchard’s earlier work, this chapter shifts the focus from simple institution effects to the interaction of institutions with macroeconomic shocks, represented by the slowdown in total factor productivity (TFP) growth, trends in long-term real interest rates, and shifts in labor demand. Blanchard and Wolfers argue that labor market institutions may produce higher unemployment by limiting the ability of labor markets to respond to adverse shocks. This helps explain why the same institutions were not employment-unfriendly in previous decades. Their study is also distinguished by a much longer time period (eight 5-year periods from 1960–96; the last 2 years are treated as a full period), and while it relies heavily on Nickell’s institutional measures, it also employs alternative, OECD-generated measures of benefit replacement rates and employment protection laws that vary over time. The substantial skepticism with which Blanchard and Wolfers treat their results offers a striking contrast to much of this cross-country regression literature (again, see Baker et al. 2002). Along with numerous robustness tests, the authors challenge the inadvertent bias that may creep into the generation of the variables themselves. One must worry however that these results are in part the result of research Darwinism. The measures used by Nickell have all been constructed ex-post facto, by researchers who were not unaware of unemployment developments. When constructing a measure of employment protection for Spain, it is hard to forget that unemployment in Spain is very high…. Also, given the complexity in measuring institutions, measures which do well in explaining unemployment have survived better than those that did not. (Blanchard and Wolfers 2000:C22) Using Nickell’s (1997) time-invariant measures of institutions (the average for 1983–88 and 1989–94) and accounting for time and country effects, Blanchard and Wolfers get results for the entire 1960–96 period that are similar to Nickell’s for the late 1980s and early 1990s. But, the authors point out that the results are quite sensitive to the Monetary policy and unemployment 58 specification. Indeed, it appears that the use of alternative, arguably much superior OECD-generated measures of unemployment benefit replacement rates and the severity of employment protection laws worsens the results. Referring to Table 5.6, they write that “The table suggests two conclusions, both worrisome: replacing the Nickell measures by alternative, but time invariant measures, substantially decreases the R2. Going from the time invariant to the time varying measures further decreases the fit.” Nevertheless, on balance they conclude that” “the results strike us as surprisingly good overall”— institutions traditionally viewed to be employment-unfriendly are found to be statistically associated with higher unemployment across countries. Recent research Blanchard and Wolfers’ emphasis on the importance of the interaction of macroshocks and institutions follows from their view that “while labor market institutions can potentially explain cross-country differences today they do not appear able to explain the general evolution of unemployment over time” (2000:C2). In a recent study that relies on annual data for the 1960–92 period, Nickell and colleagues (2002) directly take up the challenge. As they put it, “our aim is to see how far it is possible to defend the proposition that the dramatic long term shifts in unemployment seen in the OECD countries over the period from the 1960s to the 1990s can be explained simply by changes in labor market institutions in the same period” (Nickell et al. 2002:1). This represents a striking shift from Nickell’s relatively cautious 1997 assessment. The new paper concludes that the data support their proposition: “broad movements in unemployment across the OECD can be explained by shifts in labor market institutions.” But there is reason for considerable skepticism about such a conclusion. First, it might be noted that the Nickell et al. results appear far more fragile than the authors suggest. As Baker et al. point out, Nickell et al. put out 2001 and 2002 versions of the paper (the latter published as Nickell et al. 20031), and the main difference seems to be that the more recent one extends the data from 1992–95. This change appears to have quite large effects.2 The fact that the inclusion of three additional years (from 31 to 34 years in the time series) leads to substantial changes in the regression results—suggesting entirely different conclusions about the effects of key institutional measures—would appear to raise serious questions about the robustness of their findings. Of course, this fragility is exactly what Blanchard and Wolfers found and expressed so much concern about. Along the same lines, alternative tests using the same data suggest far less impressive results. In his “Comment” on the Nickell et al. study Fitoussi runs separate country tests on their data and concludes that “What is striking is the weak, to say the least, explanatory power of the institutional variables, especially those considered as being the more important, namely the benefit replacement rate and employment protection…” (Fitoussi 2003:434). Fitoussi’s conclusion is unequivocal: “Until now, there has been no convincing evidence that labor market institutions are responsible for the high level of unemployment in Continental Europe or for the disappointing macroeconomic performances for Europe during the 1990s” (p. 434). The same conclusion is reached in a new study by Baker et al. (2002). One of the distinctive features of the “second generation” literature has been the improvement in the Labor market institutions and unemployment in Europe 59 quality of the institutional measures (recall Blanchard and Wolfers’ warning about “research Darwinism”). As a first test of the robustness of Nickell’s influential study (1997), Baker et al. simply use the Nickell et al. data in a specification that closely resembles that of Nickell (1997). Baker et al (2002) find that, Using the Nickell et al (2001) data in the Nickell (1997) regression produces results that differ markedly from those obtained in the original study. In Nickell (1997), seven of the eight institutional variables had the correct sign and were statistically significant at standard levels. The only exception was the employment protection variable, which was close to zero and not statistically significant. Using the Nickell et al (2001) data, however, three of the six institutional variables have the wrong sign (employment protection, union density and the tax wedge) and none are statistically significant. (Baker et al. 2002:114) Just as Blanchard and Wolfers found, the use of more recently constructed (and presumably better) measures of the “employment-unfriendly” institutions produces notably worse results. Indeed, Nickel’s 1997 results entirely disappear with the better Nickell et al data. The Baker et al study runs a number of tests of the full 1960–99 period, using measures drawn from the OECD (personal communication), Blanchard and Wolfers (2000), Belot and Van Ours (2000), and Nickell et al (2002), and consistently find weak and even wrong signed results for the key measures—unemployment benefits, employment protection, and union coverage. Indeed, they find that, comparing the results before and after the mid-1980s, “if anything the results for the more recent period offer even weaker support for the deregulationist position than does the 1960–84 period.” Assessment The first generation empirical work on labor market institutions and unemployment, exemplified by Nickell (1997), has spawned a large literature which has utilized better (often time-varying) measures of institutions and increasingly sophisticated estimation methods. Given the direct implications for public policy getting the effects right is particularly important. Blanchard and Wolfers’ chapter not only advanced this research, but was distinguished by a concern over the quality of the data and fragility of their results, a concern that appears to be strongly confirmed by recent research. Their healthy skepticism is not particularly characteristic of this research literature, which may be the result of a bias towards confirmation, stemming from the dominance of classical thinking (rigidities must explain persistent high unemployment). The alternative perspective is one that stresses the ability of capitalism to thrive under a variety of institutional arrangements (Freeman 2000; Hall and Soskice 2001). It is notable that the European unemployment crisis appears concentrated in just the 1985–99 period, and for many countries (Denmark, the Netherlands, Sweden, Germany and so on) only for a portion of that decade and a half. Figure 6.1 shows a recent, strong downward convergence in unemployment rates, one that has taken place without radical reform of Monetary policy and unemployment 60 the European welfare state—by any measure, northern European countries have maintained a dramatically stronger welfare state than the United States (see Howell 2004). Indeed, by 2002, the United States again had an unemployment rate above the median of these 19 major OECD countries. The conventional wisdom, that employment-unfriendly labor market institutions are the source of the relatively recent crisis in unemployment in much of Europe, needs to be put in a larger perspective, one nicely put by Fitoussi in his comment on the Nickell et al. chapter: If we had followed conventional wisdom in each decade we would have recommended that every country in the world adopt the French institutional model in the 1960s, the Japanese one in the 1970s, the German one in the 1980s, and the U.S. one in the 1990s. (Fitoussi 2003:437) As the American “roaring nineties” come to an end, it is time for a more balanced view of the role of labor market institutions and, following Blanchard and Wolfers’ example, a less confirmatory approach to the cross-country research on unemployment. Notes 1 Actually, it appears that the new results (through 1995) are published in table 13 of Nickell et al (2003) with a heading that mistakenly reads 1961–92 instead of 1961–95. 2 In the 2001 version, the employment protection legislation variable was highly significant in all three of the published unemployment regressions (table 13) and quite large in its economic impact. In contrast, the coefficient of this variable in the regressions in the more recent version is not close to being significant. The additional 3 years also seems to have a substantial affect on the impact of other variables. In the 2002 version, the effect of higher taxes is more than 30 percent lower, the effect of coordination is nearly 40 percent lower, and the effect of benefit duration is cut by more than 50 percent. The additional 3 years of data also now make the coefficient of the interest rate variable significant. It had been very close to zero and not close to significant in the earlier regressions. In the EPOP regressions in the earlier version, only the replacement rate and benefit duration variables were found to have significant negative effects and the employment tax variable was not close to being significant. References Baker, D., Glyn, A., Howell, D.R., and Schmitt, J. (2002). “Labor market institutions and unemployment: a critical assessment of the cross-country evidence,” Center for Economic Analysis Working Paper, New School University, forthcoming in D.R.Howell (ed.), Unemployment and Labor Market Flexibility: International Perspectives on the Limits of Deregulation, NewYork: Oxford University Press. Belot, M. and Van Ours, J. (2000). “Does the recent success of some OECD countries in lowering their unemployment rate lie in the clever design of their economic reforms?” IZA Discussion Paper N0. 147. Labor market institutions and unemployment in Europe 61 Source: OECD, ECB. Notes 1 Based on ADF regression: where k=1. 2 Figures in ()s are 90 percent confidence intervals. p-values are in []s. below one (0.819) while in the two other countries, only unit values appear—this may be the result of not controlling for structural breaks. McAdam and Mestre (2002) document a number of further “stylized facts” at the Euro-area level, which are relevant to our discussion, for example, in terms of persistence of the various series of interests but also of the relation between unemployment and nominal/real variables. As can be seen (Table 7.2), the price level, unit labor costs, the mark-up, nominal wages, and the unemployment rate are all countercyclical; others are pro-cyclical.8 Unemployment therefore is the only real magnitude, which has cyclical features similar to those of nominal series. On the other hand, series that appear to be leading the output cycle are only nominal ones, such as the price level, unit labor costs, and nominal wages whereas inflation, real wages, employment, and unemployment are lagging indicators. Finally, all series are highly persistent as judged by their AR (1) values, but unemployment as well as employment appear more sluggish than productivity and real wages, as if the latter two were adjusting to cyclical developments to a larger extent than the former two. Much attention has been devoted to explain the relatively poor record for Euro-area employment and job creation. Typical themes in this debate being differences in the adoption of new technologies (new economy) (Temple 2002); shifts in rents reflecting changes in union bargaining and technology biases (Blanchard 1997; Bentolila and Saint- Paul 1998); differences in product-market regulations (Jean and Nicoletti 2001); the interaction between shocks and national institutions (Blanchard and Wolfers 2002), etc. Finally, Morgan and Mourougane (2001) offer Table 7.2 Euro-area historical stylized facts St. dev. St. dev. ratio AR (1) t−4 t−3 t−2 t−1 t t+1 t+2 t+3 t+4 Real GDP 1.027 1.000 0.834 0.137 0.365 0.623 0.839 1.000 0.839 0.623 0.365 0.137 Price level 1.196 1.164 0.955−0.669−0.713 −0.680 −0.591 −0.453−0.279 −0.119 0.011 0.109 Unit labor costs 0.014 0.014 0.900 −0.589−0.676 −0.725 −0.696 −0.605−0.365 −0.121 0.117 0.288 Mark-up 0.862 0.839 0.794 −0.285−0.389 −0.505 −0.554 −0.548−0.319 −0.078 0.184 0.38 Nominal wages 1.074 1.045 0.866 −0.635−0.644 −0.586 −0.462 −0.265−0.138 −0.009 0.12 0.205 Real (Producer) wages 0.664 0.646 0.678 −0.132−0.100 −0.039 0.046 0.195 0.165 0.159 0.198 0.223 Real consumer wages 0.834 0.811 0.737 0.167 0.190 0.207 0.245 0.321 0.235 0.171 0.151 0.123 Monetary policy and unemployment 68 Labor Productivity 0.719 0.700 0.705 0.228 0.379 0.563 0.69 0.799 0.512 0.226−0.051 −0.26 Employment 0.621 0.6040.928 −0.018 0.173 0.375 0.577 0.72 0.787 0.765 0.659 0.503 Unemployment 0.344 0.335 0.913 −0.094−0.234 −0.398 −0.585 −0.734−0.799 −0.76 −0.645 −0.471 Source: Abbreviated from McAdam and Mestre (2002). Notes Sample for all series is 1970Q1 to 1999Q4. We first log each series, then use the difference between the logged series and its HP filter to derive the associated moments and cross-correlations: y−HP (y). a comprehensive overview of the effect of structural and institutional factors on the labor markets of the major EU countries. While the standard interpretation—see our earlier cited references—is that European labor markets are subject to institutional features that slows their responsiveness down compared to the United States, it should be pointed out that econometrically pinning down the wage-price-employment nexus is a difficult task. For instance, to take one important illustration, while standard (as opposed to “New”) Phillips curves fit the data robustly their recent forecasting performance appears to have weakened (e.g. Anderson and Wascher 2000). This might reflect such things as measurement errors in output-gap estimates (Orphanides et al. 2000); new economy effects or, the failure of backwardlooking expectations to capture improved policy credibility The evidence of whether European labor markets are chronically slower to adjust than, say US ones is therefore mixed and not clear-cut. For instance, Chagny et al. (2002) estimate standard wages equations and find generally that the absolute value of the elasticity of wages with respect to unemployment is higher in the United States (i.e. a more flexible labor market); there are signs that France has an even higher elasticity. Notwithstanding the uncertainty on such parameter values, it seems useful to review the estimates we will employ as benchmark parameters in the subsequent steps of the exercise. In what follows, we provide details in particular on the long run unemployment rate as well as the behavioral equations for employment and wages. An estimate of the Euro-area Nairu The Euro-area Nairu (the non-accelerating inflationary rate of unemployment) referred to here is computed as suggested by Fabiani and Mestre (2002), namely following the Gordon (1997) approach. They consider a number of additional estimation methods and models and, in general, find that a Nairu can be identified that is relatively robust to changes in the underlying models. The Nairu that they identify is imposed on the AWM as an exogenous but time-varying component. The resulting time series for the Nairu is on Figure 7.2, starting from low values around 3 percent in the early 1970s and increasing almost continuously between 1973 and the late 1980s. Thereafter, it stabilizes at a value under 9 percent, with a small additional increase around 1993, which the recent decrease has not fully offset. In the following exercise, no particular assumption is made regarding further exogenous changes that could drive the Nairu—the long run value is fixed at 9.1 percent. Exercises to vary the Nairu are possible in the model, but the default case is to keep the Nairu fixed, equal to its baseline value in history to a constant value out-of-sample.9 Labor market dynamics in the Euro-area 69 An estimated employment equation The specification employed for the employment equation of the AWM is solved from the inversion of the (Gobb-Douglas) production function—from this static Figure 7.2 Trend unemployment rate. expression, there is a dynamic equivalent.10 The resulting estimated ECM equation is the following: where : Total employment (including self-employed); DLNNSS: a parameter set equal to trend labor force growth; : real (product) wage growth minus trend productivity growth KSR: Total capital stock; : Trend total factor productivity—labor augmenting, HP filtered Solow residual; : Real GDP growth minus trend productivity growth; β: Capital-share parameter in the Cobb- Douglas production function (=0.41). An interesting feature of this equation is that it incorporates some significant impact of real wage on the dynamics of employment, with an elasticity of about 0.25. One of our experiments will incorporate a scenario where this response is twice as strong, with a view to mimicking a more flexible labor market.11 Monetary policy and unemployment 70 An estimated equation for wage determination where LPROD: Labor productivity; PCD: Gonsumption deflator; ULC: Unit labor costs; : Trend unit labor costs; : Trend unemployment rate; URX: Unemployment rate; : Average compensation per head; WIN: Compensation to employees; : GDP real; : GDP deflator at factor cost; : Inflation expectations—set equal to steady-state inflation, also enters the Taylor rule. Wages are modeled as a Phillips curve in levels, with wage growth depending on productivity current and lagged inflation—in terms of consumer prices—and the deviation of unemployment from its structural level (the Nairu). Since dynamic homogeneity holds, the long run Phillips curve is vertical. As in Chagny et al. (2002) who provide estimates of wage equations for a number of large OECD countries, we focus on the wage equation within a system also comprising a price equation, that is, we do not take the price–price Phillips curve approach (see e.g. Turner and Seghezza 1999, as an illustration of the latter, also on OECD countries). The price formation is taken as given, identifying therefore the wage equation as embedding most of the labor market features, whereas the price equation is assumed to relate to goods market features. Short run dynamics also include a calibrated term in expected inflation. This expectations term may be viewed as a crude proxy for forward-looking behavior (inflation expectations being set exogenously) such as the one modeled in Clarida et al. (1998) or Gerlach and Svensson (2000). Having both this term and the similar one entering the price equation equal to 0.2 yields similar full model simulation results as the configuration in which either the wage or price-setting curve includes an expectation term with a 0.8 coefficient, which comes close to available estimates from the reduced form approach. On the basis of the equation mentioned in the preceding paragraph, two key parameters can be identified, first the one pinning down the response of (real) wages to employment, second the extent to which wage formation incorporates exogenous inflation expectations—in line with the steady-state value, in turn equal to the Central Labor market dynamics in the Euro-area 71 Bank objective—or model-consistent forward-looking ones. In terms of structural change affecting behavior, shifts in these two parameters seem worth examining. The experimental framework In this section, we clarify how the sensitivity analysis is to be conducted, that is, using a model of the economy with both estimated and calibrated equations, both fitting the data and reflecting long run “well-behaved” structure, allowing for changes in the earlier mentioned parameters—real wage effects on employment, real rigidity and inflation expectations in wage formation. In addition, there is a need for both deterministic and stochastic simulations to capture the “average” response that may differ from the deterministic one (see e.g. Clark et al. 1997). The AWM structure—an overview The AWM model structure is standard, built on an aggregate demand/aggregate supply framework, with a well-defined long run classical supply-side equilibrium and a vertical long-run Phillips curve, with an exogenous Nairu.12 The steadystate real interest rates pins down the capital to output ratio, via the marginal productivity optimality condition of the firm. Since labor force is given in the long run, steady-state output is then equal to the production function outcome with a zero unemployment gap, that is, with no deviation from the Nairu. In that equilibrium, all real variables grow at the same rate as potential output, real wages grow in line with long run labor productivity and relative prices are constant, in particular the real exchange rate.13 The model also takes into account stock-flow consistency so that, for example, households’ wealth (comprising total capital stock, public debt, and net foreign assets) is also determined in GDP points at steady state. With the steady-state public debt, ratio pinned down by a fiscal rule and the steady-state capital stock determined by the marginal productivity condition, this relation pins down the steady-state ratio of net foreign assets to GDP In the short run, since nominal variables do not immediately adjust to their steadystate values, the model has some Keynesian features, whereby output is given by the sum of demand components. The resulting departure from equilibrium in both the goods and labor markets exerts influence on short run price and wage developments, via an output gap and an unemployment gap term, respectively The convergence to the long run is ensured by the responses of policies to deviations from equilibrium. A number of alternative options can be used in the AWM (as well as in other models). In a standard configuration, fiscal policy is modeled as a change in direct tax rates responding to deviation from an objective that can be formulated in terms of, for example, a fiscal deficit to GDP ratio. Monetary policy in turn can be assumed to follow a standard rule—such as the Taylor rule—whereby the Central Bank responds to deviation from some inflation objective and to the output gap (which can be interpreted as an indicator of future inflationary pressures). Irrespective of their precise specification, both types of policy reactions are necessary for the model to reach its steady-state balanced growth path. Monetary policy and unemployment 72 An additional important feature of the model is the possibility to pin down the nominal equilibrium in a number of alternative ways. Since the earlier mentioned equilibrium is defined in real terms or relative prices only a nominal anchor has to be specified, which can be done, for example, via a Taylor rule, determining the steady-state inflation rate, and therefore, the price level path, depending on the initial conditions for the price level. In total, the AWM comprises 89 equations, of which 15 are behavioral—which still makes it a fairly tractable tool. Box 7.1 provides an overview of the main Box 7.1 A summary of the AWM equations Supply side I≡∆К+δK−1 (1) (2) (3) (4) (5) (6) (7) (8) Demand side (9) ∆c=γ∆y−γ1·r−γ2·(c−γ3·yd−(1−γ3)·a)−1 (10) (11) (12) (13) Labor market dynamics in the Euro-area 73 (14) (15) Note I: Investment; K: Gapital; ypot: Potential output; Ogap. Output gap; l: Labor; U: Unemployment rate; w: Wages; p: Prices; GDP; c. Consumption; yd: disposable income; d: Deficit in GDP percentage points; A: Wealth accumulation; X: Exports; m: Imports. δ depreciation rate, β long run capital income share, equal to the elasticity of capital in the production function, r real interest rate—defmed as the difference between the monetary policy interest rate and inflation. Trend labor augmenting total factor productivity t direct tax rate—defined by the fiscal policy rule, aiming at stabilizing the fiscal deficit ratio to GDP, e nominal exchange rate—defined by a (forward-looking) UIP condition. Variables in italics are in logarithms, a bar denotes exogenous variables such as labor supply, public spending, world demand, world prices. Symbols: =is used for econometric equations and≡for accounting identities. λ, ω, π, γ, ξ, µ are coefficients. All econometric equations in the AWM are specified as ECMs, while in this Box, for simplification purposes, only those equations with specific short run features are presented with their dynamics (e.g. for prices). equations in the model, in a simplified version comprising seven econometric relations (missing equations are mostly related to additional deflators). The main elements in the supply-side of the AWM are, first, equation (2), which links the long run capital to output ratio, to the real interest rate augmented by the depreciation rate (as proxy for the user cost). Investment obtains via the accounting identity (1). Inverting the Cobb-Douglas production function, equation (3), labor demand can be derived as in equation (5)—with some short run negative impact of the real wages. The marginal productivity condition on labor appears in the long run of the wage equation (7), with, in addition, some Phillips curve impact of the unemployment rate—as defined by equation (6). The price behavior is based on a standard mark-up on Unit Labor Cost expression (see equation (7)), the mark-up increasing with the output gap defined by equation (4). It should be noted that inflation terms in both wages and prices can be rewritten in terms of a weighted average of past and expected future inflation, as in Gerlach and Svensson (2000) (see Fagan et al. 2001). These expectational terms can alternatively be set exogenous (e.g equal to the inflation objective in the Taylor rule) or fully model-consistent. As to the demand-side, GDP has five components, of which public spending is exogenous in real terms (see equation (9)). The consumption equation (10) can be interpreted as a backward-looking version of a standard theoretical consumption function with some habit formation term. Consumption is proportional to income supplemented with financial wealth, with moreover an intertemporal substitution effect, hence the real interest rate term. There is moreover in the short run a proxy for liquidity constraints, Monetary policy and unemployment 74 captured by changes in income. Disposable income, defined by equation (11), is equal to wage income net of taxes (transfers are also considered in the full model framework). Equation (13) describes the accumulation of house-holds’ wealth, which increases with the trade balance, public deficit and investment net of capital depreciation. Public deficit is given by equation (12), which is expressed in GDP points, to reflect the possible endogeneity of the direct tax rate via a fiscal rule based on a deficit ratio target. Finally equations (13) and (14) show the specification of the exports and imports, both being assumed to respond with unit elasticity to the corresponding demand, conditional on competitiveness developments. A particular aspect of the model is that it comprises a number of indicators for prices and costs, namely the following full system of deflators: • GDP (factor cost) deflator • GDP (Market Prices deflator, i.e. including indirect taxes) • Average whole-economy earnings (compensation per head) • Consumer Expenditure Deflator PCD—average import +output prices • HICP—bridge equation from the latter price, specific effect of oil • Import and Export Deflators— and —intra vs extra Euro-area variables • The investment deflator— —weighted average of import and output prices. The system mentioned in the preceding paragraph has also to be supplemented with closure rules and a UIP (forward-looking) condition to be closed. For illustrative purposes, the model can therefore be further approximated by a simple 4-equation deviation from steady state one comprising the UIP exchange rate condition (e), a Taylor rule for the nominal interest rate (r), an aggregate demand curve (y) and a Phillips curve for inflation (π): The system can be solved, for example, for the change in the exchange rate (e)—in this simple case, the only forward-looking variable, as in the standard version of the AWM— the dynamics of which then are found to satisfy: The system clearly displays saddle-path stability with a unitary unstable root and the stable root given by λ. Thus, as standard, there is an initial overshoot of the nominal Labor market dynamics in the Euro-area 75 exchange rate for a given monetary policy change, and a reversion to base whose dynamics are driven by the value of the stable root and by an amount which is a function of the whole sequence of shocks (in this model, that shock derives purely from the inflation equation: a cost-push shock in normal parlance). This simple exercise demonstrates important aspects of our system. After the initial saddle-path jump, the reversion to baseline will be a function of parameters such as those measuring the Phillips curve slope, γ, the imported inflation pass through, β, and also the respective weights of inflation and output gap in the postulated Taylor rule—assumed earlier to take their standard values. In practice, however, the dynamics of the AWM are much richer. Hence there is the need for more complete exercises. Nevertheless this simple model highlights the key role of parameters such as the slopes of both the IS and the Phillips curves as well as the pass through of the exchange rate to prices in defining core dynamics of the system. Three labor market configurations As already indicated, both deterministic and stochastic simulations will be conducted with three alternative sets of parameters and specifications for the labor market equations, namely the earlier mentioned employment and (real) wage equations. First, we look at the Base case, namely the standard version of the AWM is to be used as an already largely documented benchmark—simulation results have been reported and assessed by, for example, Dieppe and Henry (2002) who present simulation results for a number of shocks under alternative monetary policy rules or McAdam and Morgan (2001) who give a systematic comparative assessment of the model responses to a normalized interest shock. Second, we investigate the impact of a so-called Flexible case, in which both the real wage term in the employment equation and the Phillips curve term in wages are multiplied by a similar factor (two) to reflect the idea that, in line with conventional wisdom, such elasticities should be much higher than those estimated to date for the Euro-area, in the event where the labor market would be more flexible, with both labor demand more sensitive to real wages and real wages more responsive to unemployment (Poret et al. 1989 took a similar approach when looking at the impact of deregulation in goods and labor market in the OECD countries, by running counterfactual simulations giving to European economies elasticities similar to those estimated for the United States). Third, we define a nonlinear configuration, which will be referred to as Hysteresis thereafter. As documented in a number of empirical papers, there are signs of nonlinearities in the process underlying the Nairu, such as asymmetric impact of shocks of different size or threshold effects (see e.g Bianchi and Zoega 1998; Akram 1998; Røed 1997; León-Ledesma and McAdam 2002). The particular type of configuration we want to investigate builds on the smooth transition models, see, for example, Teräsvirta (1994); this allows us to functionally endogenize the elasticity of wages with respect to unemployment, without however affecting ex ante the long run behavior of the model.14 In our application, we choose the following logistic specification: Monetary policy and unemployment 76 With such a specification, the long run Nairu is unchanged although in the short- to medium run, the first derivative of inflation with respect to unemployment is no longer a constant but a function of the unemployment level itself, so that for a sizeable number of periods, unemployment could possibly widely differ from its steady-state value. With the earlier mentioned formulation, the elasticity γ is exactly equal to its point estimate (– 0.0147) in the case where actual and long run unemployment rates (URX and , respectively) are equal. There is, however, some increasing divergence from this value as the unemployment deviation from its steady-state value increases. For illustration, the elasticity is at −0.0086 for URX=11 percent and at −0.0382 for URX=7 percent. An additional non linearity then appears when comparing the elasticity from its central point value to that taken on these two extremes, with a stronger impact of lower unemployment on the elasticity than what is observed for higher unemployment.15 Monetary policy environments In common with many other macromodels, the AWM includes a Taylor rule to operationalize monetary policy. Taylor rules have a number of advantages. First, they are empirically well-founded and second, in the theoretical literature on policy rules, they are often considered well micro-founded, being a good approximation to the true welfare function and an approximation to optimized rule. Against that, there is the problem of the often-cited unreliability of real-time output gap estimates,16 and the need to know the steady-state rate of interest, which can be highly model-dependent and highly uncertain. (See the discussions in Taylor 1999; Gali 2001.) The generalized Taylor rule is given by: Where ρ is a smoothing parameter, r* is the steady-state real rate of interest, πt is the contemporaneous inflation rate with target is the contemporaneous real output gap, α, β are parameters defining the feedback from, respectively inflation and output-gap targets to the nominal interest rate, i, and E defines the expectations operator. The integer parameters θ, k define the policy-maker’s planning horizon. Typical configuration include: θ=k=0 (i.e. an outcome-based rule), or a forecast-based rule for θ>0, k≥0. In our exercises, we consider the following parameter variations to be of interest: a, β, ρ, θ.17 Invariably however, we assume α (1, αMax] in line with the so-called Taylor Principle, which states that if inflation rises then nominal interest rates will rise further to stabilize the economy (i.e. real interest rate increases).18 More specifically the conducted set of experiments is based on the following range of parameters and specifications for the Taylor rule (all in all eight cases): First we use the “Standard” configuration, namely a standard Taylor rule, with the usual parameters on both its arguments. This has to be considered as a bench-mark in Labor market dynamics in the Euro-area 77 Beyond these descriptive elements, further statistics can be computed from the stacked observations, for both the“shocked” and the full samples. In Tables 7.3 and 7.4, we present summary statistics to represent each of the distribution’s mean, median, standard error, selected percentiles, selected cut-off point critical values as well as more formal metrics—Skewness, Кurtosis, and Normality. Skewness refers to lack of symmetry: a distribution has positive skewness when it has a long thin tail to the right. Kurtosis refers to the extent to which the peak of a distribution departs from that of a Normal, being either pointed (leptokurtic) or flatter (platykurtic). Normal has a (excess) Kurtosis of zero. Formal definitions of these concepts are given below: For background to these tests, see, for example, Stuart et al. (1999). 23 Monetary policy and unemployment 84 Figure 7.7 Base case and standard Taylor rule, quantiles. Labor market dynamics in the Euro-area 85 Figure 7.8 Unemployment distribution under different regimes, Base case. Monetary policy and unemployment 86 Figure 7.9 Inflation distribution under different regimes, Base case. Labor market dynamics in the Euro-area 87 Figure 7.10 Unempolyment distribution under different regimes, Hysteresis case. Monetary policy and unemployment 88 Figure 7.11 Inflation distributions under different regimes, Hysteresis case. Labor market dynamics in the Euro-area 89 Table 7.3 Statistics for the unemployment stacked observations, “shocked” sample Urx TR std. dev. TRFL inf 4 inf 2 inf 0.5 inf 0.25 TR smoot Expect Mean base 9.05 9.07 9.09 9.02 8.96 9.04 9.01 8.94 flex 9 9.05 9.09 9.04 9 9.04 9 9.02 hyst 9.13 9.14 9.39 9.35 9.14 9.04 9.18 9.15 Ser base 1.376 1.17 1.486 1.46 1.367 1.024 1.44 1.586 flex 1.37 1.19 1.578 1.434 1.2 0.876 1.43 1.208 hyst 1.145 1.142 1.282 1.361 1.142 0.788 1.09 1.15 Kurt base 0.546 0.15 0.266 0.313 0.277 0.281 0.313 0.558 flex 0.384 0.324 0.346 0.469 0.62 0.462 0.42 −0.128 hyst 0.906 0.795 –0.003 0.819 0.795 0.139 0.34 0.976 Skew base −0.155 −0.05−0.037 −0.144 −0.232 −0.223 −0.156 −0.338 flex −0.037 −0.095 0.038 −0.079 −0.22 −0.256 −0.06 −0.077 hyst 0.368 0.485 0.205 0.383 0.485 −0.023 0.424 0.461 Norm base 311 42 86 59 97 92 62.4 195 flex 174 140 142 127 475 192 207 12.94 hyst 346.7 413 89.39 357 413 11.97 317 413 T0.05 base 6.65 7.09 6.56 6.49 6.61 7.25 6.53 6.08 flex 6.68 7.03 6.39 6.63 6.96 7.5 6.62 6.96 hyst 7.4 7.37 7.29 7.27 7.37 7.66 7.52 7.42 T0.10 base 7.25 7.54 7.2 7.12 7.15 7.68 7.16 6.91 flex 7.21 7.73 7.04 7.23 7.45 7.9 7.17 7.38 hyst 7.77 7.76 7.83 7.75 7.76 8.02 7.84 7.77 Med base 9.02 9.06 9.07 9.05 8.96 9.01 9.05 9.03 flex 8.98 9.05 9.03 9.04 9.04 9.05 9.03 9.04 hyst 9.05 9.05 9.31 9.22 9.05 9.05 9.08 9.06 T0.90 base 10.55 10.6 11 10.87 10.42 10.08 10.85 10.89 flex 10.49 10.43 10.78 10.87 10.51 10.13 10.83 10.52 hyst 10.6 10.63 11.09 11.09 10.63 10.03 10.63 10.63 T0.95 Monetary policy and unemployment 90 base 10.88 11.02 11.59 11.43 10.8 10.28 11.34 11.44 flex 10.86 10.99 11.24 11.46 10.93 10.43 11.34 10.96 hyst 11.1 11.16 11.69 11.74 11.16 10.33 11.09 11.17 CV 1 ser base 25.8 23.3 27.7 25.9 24 19.9 25.6 25.8 flex 23.9 15.3 28.1 25.7 21.3 15.3 24.5 13.8 hyst 22.7 22.5 33.1 30.3 22.5 13.2 23.7 23 Table 7.4 Statistics for the inflation on stacked observation, “shocked” sample Summary statistics inflation Inf TR std. dev. TRFL inf 4 inf 2 inf 0.5 inf 0.25 Tr smoot Expect Mean base 2.02/2.05 2.001/2.01 2.02/2.04 2.02/2.05 2.03/2.08 2.01/2.04 2.02/2.06 2.06/219 flex 2.03/2.09 1.999/1.990 2.03/2.05 2.03/2.06 2.02/2.08 2.01/2.03 2.03/210 2.04/2.06 hyst 2.05/2.19 2.05/2.17 2.03/2.10 2.05/2.16 2.05/2.17 2.04/2.12 2.04/2.12 2.05/2.18 Ser base 0.42/0.709 0.361/0.630.451/0.722 0.453/0.7310.418/0.732 0.36/0.6440.443/0.733 1.043/1.341 flex 0.722/1.062 0.596/0.9100.814/1.105 0.840/1.0940.629/0.978 0.480/0.7830.777/1.105 1.434/1.286 hyst 0.480/0.834 0.505/0.7990.500/0.811 0.649/1.0830.505/0.799 0.322/0.5580.379/0.624 0.486/0.832 Kurt base 6.36/0.808 5.47/0.12 4.22/0.375 5.77/0.75 7.1/0.550 7.06/0.49 6.09/0.711 4.22/2.57 flex 3.36/0.938 3.825/0.418 2.57/0.723 6.56/1.067 7.23/1.61 5.23/0.742 4.13/1.32 −0.633/−0.054 hyst 68.2/24.013 76.3/8.81225.44/8.919 85.11/33.27 76.31/8.81 9.89/1.493 32.33/3607 60.883/23.09 Skew base 0.636/0.297 0.264/0.10 0.325/0.173 0.653/0.32 0.82/0.252 0.412/0.1240.681/0.292 0.84/1.01 flex 0.629/0.444 0.273/0.2380.268/0.427 0.725/0.5020.979/0.593 0.740/0.448 0.626/0.52 0.081/0.153 hyst 5.775/3.554 5.662/1.904 3.25/2.110 6.875/4.441 5.66/1.904 1.934/0.865 3.13/1.897 5.203/3.302 Norm base 41986/621 50200/60 34493/239 14132/228 16580/155 23099/102 15009/208 6632.1/1239 flex 26200/1058 29499/332 16205/887 24239/610 41945/1771 15940/456 25244/1422 656/31 hyst 12220/16910 64685/375131278/6143 289180/3922 64685/3751 16611/1520 21748/3607 70666/10767 T0.05 base 1.35/0.91 1.40/0.99 1.28/0.879 1.28/0.89 1.38/0.89 1.44/0.94 1.30/0.89 0.49/0.27 flex 0.88/0.43 0.99/0.56 0.69/0.32 0.72/0.35 1.04/0.59 1.23/0.82 0.81/0.40 −0.22/0.00 hyst 1.53/1.26 1.48/1.15 1.38/1.09 1.38/1.09 1.48/1.15 1.60/1.33 1.54/1.25 1.51/1.22 T0.10 base 1.60/1.15 1.66/1.21 1.54/1.15 1.56/1.16 1.65/1.15 1.68/1.19 1.58/1.16 0.91/0.67 flex 1.25/0.77 1.35/0.86 1.1/0.69 1.10/0.741.38/0.91 1.50/1.091.19/0.76 0.14/0.44 hyst 1.70/1.44 1.68/1.33 1.60/1.29 1.61/1.30 1.68/1.33 1.76/1.48 1.71/1.43 1.68/1.41 Labor market dynamics in the Euro-area 91 Med base 2.00/2.00 2.00/2.01 2.01/2.02 2/2.025 2.00/2.02 2/1.99 2.01/2.03 2.02/2.04 flex 2.01/1.99 2.01/1.98 2.02/1.95 2.02/2.012.00/2.02 1.99/1.992.01/2.02 2.02/2.02 hyst 2.00/2.06 2.00/2.08 1.99/1.99 2.00/2.01 2.00/2.02 2.00/2.05 2.00/2.05 2.00/2.07 T0.90 base 2.44/2.80 2.35/2.84 2.49/2.98 2.47/2.11 2.42/2.84 2.28/2.65 2.46/2.94 3.21/3.79 flex 2.81/3.23 2.79/3.18 2.95/3.23 2.94/3.472.65/3.31 2.48/3.042.89/3.50 4.10/3.86 hyst 2.37/2.92 2.43/3.03 2.42/2.93 2.40/2.94 2.43/3.03 2.32/2.80 2.36/2.84 2.39/2.95 T0.95 base 2.75/3.00 2.63/3.09 2.79/3.27 2.77/2.95 2.77/3.03 2.49/2.80 2.77/3.30 3.76/4.63 flex 3.26/3.55 3.50/3.58 3.38/3.59 3.41/4.033.08/3.77 2.83/3.433.34/3.99 4.44/4.40 hyst 2.72/3.35 2.80/3.48 2.76/3.52 2.77/3.63 2.80/3.48 2.61/3.18 2.70/3.15 2.76/3.38 GV 1 ser base 9.9/26.6 8.1/23.8 11.3/27.2 10.7/27.3 9.7/28.9 8.0/24.810.4/27.7 30.3/34.8 flex 19.1/32.7 12.4/30.2 24.7/32.0 24.1/31.514.7/30.9 11.0/26.7 21.5/33.2 60.0/53.2 hyst 8.7/25.6 9.7/29.0 9.50/22.4 9.4/23.1 9.74/29.0 7.57/23.0 8.5/24.1 9.25/26.3 In order to summarize the wealth of information supplied, we proceed by looking in turn at each of the statistics that are reported, focusing on the one hand, on the differences across monetary policy responses (a given column in the Tables) and across labor market configurations, on the other (a given row in the Tables). Only results for the “shocked” sample are reported, “control” sample ones being available upon request. In most cases, the whole sample results are similar in terms of ranking the various configurations, and anyhow less relevant from an analytical viewpoint, since they include 75 percent of observations without shocks, therefore tracing more the return to baseline feature than the actual properties of the simulated economy Taking the earlier mentioned approach, and focusing on the “shocked” sample, the main results are as follows: In terms of the mean values of both inflation and unemployment, some gaps appear between the average and the steady-state values, which are not negligible, although relatively small. The results are particularly notable for the Hysteresis case, irrespective of monetary policy with both variables being higher than their long run value, up to 0.3 percentage points. Both the Base and the Flexible case yield lower unemployment than its steady-state value, with however the opposite result for inflation. Finally The Flexible case is doing better than the Base one for unemployment, only if standard weights are used in the Taylor rule. As to the resulting standard deviations, the Hysteresis case is characterized by a more concentrated unemployment distribution, that is, a low variance, and this holds across policy regimes. This could be related to a higher degree of stickiness in the unemployment rate, around higher values on average. In addition, it seems that, across labor market configurations, using forward-looking inflation expectations either in the rule or in the wage equation leads to lower variance for both unemployment and inflation. Another stylized fact is that increasing weights on inflation in the (contemporaneous) Taylor rule creates usually more dispersion, in particular, on unemployment. Looking at the normality features, among a bunch of configurations that are all quite non-Gaussian, Hysteresis appears clearly as departing more markedly from normality Monetary policy and unemployment 92 than the other cases. Irrespective of the policy regime, Hysteresis creates very fat tails, positive skewness for unemployment (whereas it is negative for other cases) and stronger skewness for inflation. This can be related to the earlier mentioned discrepancy between the average and steady-state values.24 As regard the quantiles, the whole distribution under Hysteresis is shifted to the right, that is, with higher values at a given percentage. Another finding is that mostly upper bounds for both inflation and unemployment are affected by the weight put on inflation in the Taylor rule, whereas changes in the labor market configurations impact more on the lower bounds of the distributions. The quantiles also provide information on the empirical confidence band around the mean value of a given variable. It appears that, consistent with the standard error results, Hysteresis corresponds to a narrower band for unemployment, whereas the Flexible case yields the wider band for inflation. Overall, the derived 90 percent confidence bands— taking the difference between the 5 percent and the 95 percent critical values—are about 2 percent wide for inflation and 4 percent for unemployment in almost all cases. The unemployment band increases however with the weight put on inflation in the Taylor rule; this is less the case for inflation. A paradoxical result is that, neither for inflation nor unemployment, does the Hysteresis case generate the highest share of the distribution—the p-value—beyond a given threshold.25 The “unemployment trap” configuration seems therefore not to be observed, in spite of the very strong nonlinearity assumed in the employed calibration. On the other hand, density figures show that in some cases there is some accumulation of observations around values at 12 percent or so. This is the case, for example, for the Taylor rule and the interest rate smoothing—see urxhystr or urxhysis respectively—but these observations do not seem to have much weight overall. In many cases, however, the Flexible case does result in a lower share of high unemployment values, but this finding is almost reversed for inflation. Across all configurations, relatively high inflation is observed for between one-fourth and one-third of the observations, whereas high unemployment obtains for 15–25 percent of the then cases. Of course, by construction, those figures drop—to values around 10 percent—when considering the whole sample. An additional general remark is that, for almost all indicators, differences across labor market configurations are largely dampened by using a forward-looking Taylor rule. This holds especially for unemployment performance, be it in terms of first and second moments or quantiles; this is the case also, albeit to a lesser extent, for inflation outcomes. Another finding holding across configurations is that aggressiveness (i.e. high weight on inflation)—and without forward-lookingness—does not necessarily result in lower inflation, it may even cause higher unemployment and increased variability in the economy Similarly interest rate smoothing leads to lower unemployment, but at the expense of variability with moreover stronger and more volatile inflation. A final general remark is that the mean unemployment and its variability are often at the lowest level with either a Flexible labor market configuration or with model-consistent inflation expectations entering wage formation. Labor market dynamics in the Euro-area 93 Part III Structuralist causes of unemployment and monetary policy 8 The structuralist perspective on real exchange rate, share price level, and employment path What room is left for money? Hian Teck Hoon, Edmund Phelps, and Gylfi Zoega The current sluggish performance of the US economy follows one of the most remarkable booms in recorded history. The late 1990s was a period of striking expansion of both output and employment, with the unemployment rate hitting 3.9 percent in 2000; productivity growth was much improved, in part because of higher utilization, though not exceptional.1 The absence of rising inflation during this period came as a surprise to many since the level of the natural rate of unemployment was commonly estimated to be in the range of 5–6 percent by the mid-1990s. The noninflationary boom, however, reminds one of another episode where nonmonetary forces were strongly at work, namely the nondeflationary slump in Europe and elsewhere in the 1980s and 1990s, which appeared to signal a move to a higher natural rate of unemployment. The modeling of such structural slumps and booms is the task that we have tackled in a series of works in recent years, the book Structural Slumps being a milestone.2 The theory set out in Structural Slumps is based on intertemporal nonmonetary models of the modern (thus incentive-compatible) kind and provides microeconomic foundations for a moving natural rate of unemployment. Involuntary unemployment occurs, since incentive wages and consequent job rationing are allowed, and this unemployment is structural, not a result of deficient aggregate demand. The determining structure includes tax rates and regulations, the focus of supply-side (SS) theory and includes fluctuations in technical progress, the focus of “real business cycle” (RBC) theory but includes much more. The implications exhibit some sharp differences from those of Keynesian theory and some striking parallels, as we shall see. What basically characterizes the structuralist perspective and differentiates it from both RBC and SS models is its view of business life: the imperfect information, the business assets firms need, and the expectations they have to form. A firm incurs costs to acquire and retain employees (workers who know their job, have learned the ropes) and customers (buyers who know how to reach it), not just equipment and plant. The rate of investment in each asset is a function of the value, or shadow price, placed by firms on that asset and the cost of investing in it. A raft of nonmonetary fundamentals—world real interest rates, expectations for technical advances and thus productivity growth, entitlements, the stocks of the business assets, the wealth of the workers, tax rates, the political climate, investor trust, etc.—drive the values placed on the business assets, the cost of investing in them, and thus the rates of investment in them.3 Our “baseline” models of business asset investment and the employment path are restricted to the case of (intertemporal) equilibrium: more accurately to a “punctuated” equilibrium in which, infrequently or more frequently as the case may be, a wholly unforeseen shift in one or more of the fundamentals occasionally occurs—a parametric shift or even a loss of some kind of capital—causing jumps in the real values of each of the business assets onto their new correct expectations time-paths. Yet some of the key forces among these fundamentals, such as the visions of (each of) the economy’s entrepreneurs about future profit opportunities and the judgments of financiers and professional investors, are plainly speculative, unobservable, unmeasurable; moreover, the consequences of both these unobservable forces and the observable ones, such as the world real interest rate and the long-term national productivity trend, for the values (the shadow prices) entrepreneurs put on the customer, the employee, and much else are likewise unobservable. This poses difficulties (and opportunities!) for testing and using the theory as it did for testing and using Keynes’ theory4 Two recent studies of ours meet that problem by hypothesizing that the net overall influence of these unobservable forces on the assets’ shadow prices (and thus on the employment path) is reflected, alongside the net influence of some other observable forces, by the level of the stock market. Since either the level of employment or its growth is an increasing function of every one of the (three) shadow prices, it is plausible that an increase in the current index of real share prices, interpreted as the current shadow price of the representative basket of the firms’ several business assets, is also expansionary for employment on the average. The studies found a statistical relationship between the first difference of employment and the real share price index taken as a ratio to some indicator of the cost of investing in employees and customers.5 An alternative measure of the unobservable share prices is total market capitalization as a ratio to GDP. Figure 8.1 shows the relationship (cross-section) between stock market capitalization to GDP ratio and the employment rate. A clear positive relationship is visible. It is to be noted that Switzerland is way off the line. The very high value of stock market capitalization in Switzerland would lead us to predict an even higher employment rate. However, as allowed by the theory the supply function may be convex and only asymptotically approaches full employment, in which case the position of Switzerland in the figure may not come as a big surprise after all. This chapter, in pursuing that strategy faces up to some questions. First, there is another asset price that was neglected in the two previous studies, namely the real exchange rate. In the customer-market model, an appreciation (strengthening) of the real exchange rate in a country causes firms to moderate their price markups, Monetary policy and unemployment 104 Figure 8.1 Employment and the stock market. thus pulling up the product wage and employment—an upward move along what is called the wage curve: the real appreciation may hence lower the natural rate of unemployment! This is the provocative hypothesis at the heart of this chapter. So we ask whether employment rises in response to a strengthened real exchange rate just as it rises in response to a strengthening of real share prices. The question is especially interesting since, as is well known, monetary theories of economic activity say that the stock market and the foreign exchange market pull in opposing directions: in those Keynesian and monetarist models, a strengthening of real share prices increases economic activity by boosting “effective demand” but a strengthening (appreciation) of the real exchange rate decreases activity by cutting effective demand.6 Real exchange rates are typically reported in the form of indices that enable a comparison over time but not across countries. For this reason we use data from the World Bank on hypothetical exchange rates that would give purchasing-power parity (PPP) between a country and the United States. The ratio of this hypothetical exchange rate and the actual exchange rate—observed in foreign exchange markets—can be used to test our hypothesis on the relationship between real exchange rates and the natural rate of employment and unemployment. Figure 8.2 has the relationship between the real exchange rate—defined as described in the previous paragraph—and the employment rate (one minus the rate of unemployment) in a cross-section of the same OECD economies. Note the upward sloping relationship: a real exchange rate appreciation appears to go hand in hand with higher employment rates—when domestic output becomes relatively more expensive, the rate of employment goes up, instead of falling as Keynesian theory might lead us to The structuralist perspective 105 Figure 8.2 Employment and real exchange rates. believe. Though not perfect, the relationship is surprisingly strong (correlation is 0.68). This simple graph is indicative that the relationship between real exchange rates and employment may be more involved than the textbook version of the open economy (New) Keynesian model would lead one to believe. We first lay out the theory and the answer to the key question, whether a real exchange rate appreciation tends to raise or lower the rate of employment. Figure 8.2 appears to imply that a stronger real exchange rate acts to raise the employment rate. However, it is not clear whether it is the cause or the effect. After all, our model also says that a weakening of profit prospects and the consequent drop in investment and ultimate employment causes a weaker real exchange rate as well as lower real share prices. From a forecasting standpoint, this distinction makes little difference: either way, whether as causes or effects, the strength of the real exchange rate and that of the real share-price level are theory-grounded predictors of where present forces are taking the economy one or two years ahead—absent a shift in the winds. A weak real exchange rate, like a weak stock market, spells weak activity ahead. We then consider the implications of our model for the conduct of monetary policy. Clearly, a Central Bank faces a daunting task during structural booms and slumps because the underlying natural rate of unemployment is changing over time. Our model yields a solution for the domestic real interest rate that is compatible with the endogenously determined natural rate of unemployment. This is the natural rate of interest, discussed by Knut Wicksell. By keeping the current short-term real interest rate in tandem with the natural rate of interest, a Central Bank is able to control the equilibrium inflation rate and keep the economy along a path of time-varying natural rate Monetary policy and unemployment 106 of unemployment. The natural rate of interest could, therefore, serve as the guiding light of a Central Bank’s interest rate policy during structural booms and slumps. The chapter’s third section examines the data and finds that a real exchange rate appreciation may raise the employment rate, hence providing support for our theoretical prediction and casting doubt on the simplest versions of the Keynesian model. The result is a rather hopeful step in the confirmation of our structuralist model. We find that a weaker real exchange rate, in sheltering firms from overseas competitors, invites higher markups—effectively a contraction in the supply of output and jobs—which causes employment to contract, not expand as in the monetary views. The final section takes a look at recent US experience, asking whether the current slump is of a structuralist nature. Again, the results are promising. The economic boom experienced in the United States in the late 1990s is almost entirely explained by our model, while the petering out of that boom and the recent rise in unemployment is to a large extent compatible with our model. Theory Here we set out a model of the small open economy in which all firms, foreign and domestic, operate in a market subject to informational frictions. We first examine a case where, initially all the relevant customers of national firms—firms that produce only with national labor—are nationals. Although the small open economy is too small, by definition, to affect perceptibly the world real rate of interest, national firms will certainly feel changes in demand by its national customers, and so will the exchange rate and the real interest rate in terms of the goods supplied by national firms and their price. With regard to the i-th firm, we let xi, a continuous variable, denote (the size of) its customer stock; let csi denote the amount of consumer output it supplies per customer; and let pi denote its price, say in units of the domestic good. We will let p denote the price at the other domestic firms and pe denote the price that the firm and its customers expect is being charged by other domestic firms (all measured in units of the domestic good). We introduce a variable e, where e tells us how many units of the foreign good must be given up in exchange for one unit of the domestic good. Gonsequently an increase in e is a real exchange rate appreciation. In product-market equilibrium, by definition, every firm and its customers have correct expectations about the other firms, that is, p=pe. With their expectations thus identical in product-market equilibrium, the identically situated domestic firms will then behave alike, so that pi=p=pe. A firm, in maximizing the value of its shares, has to strike a balance between the benefits of a high price, which are increased revenue and reduced cost, thus increased profit, in the present, and the benefits of a low price, which are an increased profit base in the future as customers elsewhere gradually learn of the firm’s price advantage. The key dynamic is therefore the law of motion of the firm’s customer stock, The structuralist perspective 107 (1) The joint assumption that g1<0; g11≤0 means that the marginal returns to price concessions are nonincreasing, in the sense that successive price reductions of an equal amount by firm i yield a nonincreasing sequence of increments to the exponential growth rate of customers. The inequality g2<0 implies a gain of customers at the expense of foreign suppliers when the real exchange rate depreciates, though successive weakening of the real exchange rate yields a non-increasing sequence of increments to the exponential growth rate of customers since g22≤0. What the sign of g12 is relates to the question of what the effect of foreign competition on domestic firms’ market power is. Suppose that e<1 so there has been a real exchange rate depreciation, hence foreign goods are selling at a premium. Then each identically situated domestic firm is increasing its market share at the expense of foreign suppliers. In such an environment, a reduction in pi, given p, can be expected to generate a smaller increase in the rate of inflow of customers compared to a situation where e>1 (and each identically situated domestic firm is losing customers to foreign suppliers). Since stiffer foreign competition (higher e) confers a higher marginal return to a price concession, firm i is induced to go further in reducing its markup, holding other things constant. In our theory therefore, the assumption that g2<0; g22<0 taken alone or jointly with g12<0 implies that an appreciation of the real exchange rate will lead to lower domestic markups and hence increased output supplied due to the increased competition that domestic producers face from foreign suppliers. It turns out that our key theoretical results that follow will depend on the assumption of g2<0, g22<0, and g12<0. Under this assumption, a real exchange appreciation will raise the marginal benefit of cutting domestic prices—in terms of retaining more customers— and such price cuts will appear in the labor market as upward shifts of the labor demand curve, raising employment and reducing unemployment (i.e. the natural rate of unemployment). The assumption implies that when domestic goods are relatively expensive, the marginal benefit from cutting prices—in terms of customers recruited—is greater, hence prices are lower given nominal wages, the real demand wage is higher, and so is the rate of employment in equilibrium. Intuitively high domestic prices may have made consumers aware or suspicious of further price increases. When customers pay closer attention to price decisions, this increases the gain domestic firms reap from price cuts—in the form of an expanded market share—and the loss inflicted on the domestic market share from price increases. Readers may wonder whether a policy of “pricing to market” might nullify our results. If foreign producers sell their output in our market at a fixed domestic price that does not respond to changes in the (nominal) exchange rate, so the degree of exchange rate passthrough is zero, the real exchange rate will be unchanged. In contrast, when foreign producers fix the foreign price of their product, or at any rate do not change it equiproportionately in response to a nominal exchange rate change, and allow the domestic (import) price to fluctuate, the real exchange rate is bound to fluctuate. The high correlation in the data between nominal and real exchange rates suggests that the latter scenario is by no means unrealistic, so our model has applicability despite our abstracting from pricing-to-market behavior. The degree of exchange rate pass-through is, in Monetary policy and unemployment 108 particular, high when (nominal) exchange rate changes are perceived to have a large permanent component.7 The representative firm has to choose the price at which to sell to its current customers. Raising its price causes a decrease, and lowering the price an increase, in the quantity demanded by its current customers according to a per-customer demand relationship, D(pi/p, cs), where cs in this context is set equal to the average expenditure per customer, cd, at the other firms. For simplicity assume that D(pi/p,cs) is homogeneous of degree one in total sales, cs, and so rewrite csi=η(pi/p) cs; η′(pi/p)<0; η(1)=1. Each firm chooses the path of its real price or, equivalently the path of its supply per customer to its consumers, to maximize the present discounted value of its cash flows. The maximum at the i-th firm is the value of the firm, Vi, which depends upon xi: where ς is unit cost. The maximization is subject to the differential equation giving the motion of the stock of customers of the i-th firm as a function of its relative, or real, price, and the real exchange rate given by (1) and an initial . The current-value Hamiltonian is expressed as [(pi/p)−ς]η(pi/p)cs xi+qig(pi/p, pi/p*)xi where qi is the shadow price, or worth, of an additional customer and p* is the price charged by the foreign supplier expressed in our domestic currency The first-order condition for optimal pi is (2) Another two necessary first-order conditions (which are also sufficient under our assumptions) from solving the optimal control problem are: (3) (4) One can readily show that “marginal q” is equal to “average q” so we have qi=Vi/xi. Equating pi to p, and setting qi=q, delivers the condition on consumer-good supply per firm for product-market equilibrium: 1+[η(1)/η′(1)]−S=−(q/cs)[1/η′(1)][g1(1, e)+eg2(1, e)]; η(1)=1 (5) The structuralist perspective 109 The model of Part I that was developed under the assumption of full flexibility of prices gave us a theory of the natural rate of interest. We can therefore derive an expression giving the natural rate of interest, and show that it is a function of the real share price normalized by productivity the real exchange rate, and the stock of customers. By taking note that cs/Λ=Ω(q/Λ, e, x) we obtain through various substitutions the following expression for the Wicksellian natural rate of interest in our model of Part I: (17) where µ, the markup is a function of q/Λ, e, x. It is readily checked from (17) that a rise in q, holding other things constant, is associated with a higher natural rate of interest. Intuitively a higher q raises the wealth to per capita consumption ratio and also increases the rate of growth of per capita consumption, consequently increasing the household’s required real rate of interest. A real exchange rate appreciation, however, has an ambiguous effect on the natural rate of interest as it lowers the wealth to per capita consumption ratio but increases the rate of growth of a representative household’s consumption. If the former effect dominates, which is a sufficient condition for saddlepath stability then a real exchange rate appreciation in our model lowers the natural rate of interest. In steady state, of course, we have r=r*. What are the implications of the Central Bank’s misestimating the natural rate of interest? We can solve the model to show that the equilibrium rate of inflation is a function of the current and expected future gaps between the natural rate of interest and the intercept term in the Taylor rule .10 Replacing zt in equation (13) with (15), and noting that the required path of the short-term nominal interest rate is given by we can solve forward to obtain (18) If the Central Bank fails to adjust the intercept term upwards when the natural rate of interest increases or the intercept term is decreased without there being a decline in the natural rate of interest, this would tend to raise inflation and the output gap. In the converse situation, there would tend to be deflation and a decline in the output gap. To prevent the inflation rate from either rising or falling, it would be necessary to adjust the intercept term in tandem with the natural rate of interest. In a recent study two European Central Bank economists discuss the consequences of taking into account movements in the natural rate of interest in simple monetary policy rules based on data of the Euro-area since the early 1970s.11 They found that taking into account the time-varying properties Monetary policy and unemployment 116 of the natural rate of interest led to increased stabilization of the output gap and inflation rate. Apart from misestimating the natural rate of interest, a Central Bank could also misestimate the natural rate of unemployment, which leads to another sort of error. Suppose we observe an episode where an expectation of bright future prospects leads to a booming stock market together with a real exchange rate appreciation—similar to that which the US economy experienced in the second half of the 1990s. According to our theory both the rise of q/Λ and e has the effect of lowering un. In the extreme case that helps make our point most starkly suppose that the actual decline of u observed was entirely the result of the decline of un but the Central Bank attributes it entirely to a fall relative to un, that is, a fall of u−un. Then, although a correct application of the Taylor rule would suggest that the short-term nominal interest rate be left unchanged on account of employment or output stabilization, a Central Bank that does not see that the booming stock market and the stronger real exchange rate has lowered the natural rate of unemployment would incorrectly raise the short-term nominal interest rate to a level that is not justifiable. (The short-term nominal interest rate should solely be raised in tandem with a rise of the natural rate of interest in accordance with an upward adjustment of the intercept term in the Taylor rule in (15).) The result would be a decline in the inflation rate and the output gap. Evidence on employment and the real exchange rate We have seen that the model outlined in the earlier section yields a positive relationship between employment, on the one hand, and share prices and the real exchange rate, on the other hand. In the customer-market model, an increase in share prices and an appreciation of real exchange rates induce domestic firms to cut their markup, which implies an increase of the demand wage in terms of domestic product. The upward shift in the demand-wage schedule pulls the economy rightwards and upwards along its “wage curve,” causing employment as well as the product wage to increase. We now wish to test our proposition empirically by first using OECD data and then focusing exclusively on the recent US experience. OECD unemployment The countries included in our statistical study are Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany Italy Japan, the Netherlands, Norway Spain, Sweden, the United Kingdom, and the United States. We have data on share prices, productivity (for normalizing), and real (effective) exchange rates for these countries covering the period 1977–2000. The real (effective) exchange rate is compiled by the International Monetary Fund (IMF). The estimated equation is of the error-correction variety (19) The structuralist perspective 117 where i is the country index and t denotes the years (i=1,2,…, 16; t=1976, 77,…,2000). The equation postulates a long run relationship between unemployment, on the one hand, and—in light of our theory—real exchange rates e, real share prices s, world real interest rates r* (measured in decimals), and real oil prices poil, on the other hand.12 The terms in the square bracket represent an upward sloping supply curve in the employment/real exchange rate and the employment/real share price planes. Inflation and exchange rate shocks push unemployment off its long run equilibrium path but—assuming that β1 is positive, unemployment gradually converges back to its long run equilibrium following such shocks.13 The speed of adjustment towards equilibrium is measured by this coefficient β1, which one hopes will take a value somewhere between zero and one. Singling out the potential influence of effective demand in creating disequilibrium and having in mind, in particular, the Central Bank’s imperfect demand management, as just discussed, the equation says that unemployment will be above (below) its equilibrium path if and only if price inflation is falling (rising). The function f is a nonlinear function of the unemployment rate; (u0.5−1)/0.5, following Bean (1994).14The idea is to capture the (strict) convexity of the wage-setting relationship—each consecutive fall in unemployment requires ever-larger shifts of labor demand. Note that αi is a country-specific fixed effect that captures any omitted countryspecific effects. While each country has its own fixed effect, groups of countries are made to share a sensitivity coefficient β1, as well as the sensitivity to inflation shocks (β2) and changes in the real exchange rate (β3). Table 8.1 has the definition of the variables. The equation was estimated with a panel of 326 observations. The reported estimates were derived using weighted least squares. Table 8.2 has the coefficient estimates γ1,γ2, γ3 and γ4. Table 8.1 Definition of variables u Unemployment rate (source: OECD). e Real effective (trade-weighted) exchange rate, measuring the relative price of domestic and foreign consumer goods (source: IMF). s Real share prices normalized by real GDP per employed worker (source: IMF). r* World real rate of interest (weighted average of G7 yield on government bonds) (source: IMF). poil Real price of oil (source: Citibase) π Inflation (GDP deflator) (source: IMF). The results in the table are consistent with predictions of our structuralist model. First, an appreciation of the real exchange rate causes the steady-state unemployment rate to fall (t=4.12). Second, an elevation of real stock prices also lowers unemployment (t=8.27). An increase in the world real rate of interest raises unemployment (t=9.09). Finally an increase in the real price of oil causes unemployment to rise (t=4.93). Importantly the results from estimating equation (19) confirm a negative association between real exchange rates and the unemployment rate as suggested by structuralist theory. Monetary policy and unemployment 118 The reader may wonder why real share prices and the world real rate of interest are included side by side since the former might be thought to encapsulate the latter. One reason for this inclusion is that if share prices are highly volatile owing to misguided speculation, that volatility may obscure the effect of changes in the world interest rate on unemployment unless it is entered explicitly on the right-hand side of the estimating equation.15 Table 8.3 has the group-specific coefficients. The effect of surprise inflation causes unemployment to fall for all country groups. The employment effect of the inflation shock is smallest in Japan, then in the United States, and roughly the same in the other four areas. The short-term unemployment effect of a real exchange rate appreciation is less robust. The sign of the estimated coefficient is also only correct in Japan. The speed of adjustment Table 8.2 Estimation results Coefficient Estimate t-statistic γ1 −0.02 4.12 γ2 1.28 8.27 γ3 2.76 9.09 γ4 1.28 4.93 Table 8.3 Further estimation results: group effects Sensitivity coefficient Inflation shock Real exchange shock Areas Estimate tstatistic Estimate tstatistic Estimate tstatistic Australia 0.51 4.71 −7.81 2.88 −0.02 2.40 Europe* 0.15 7.93 −5.01 3.94 −0.01 1.95 Japan 0.08 3.85 −1.79 1.51 0.00 1.31 Scandinavia** 0.06 2.14 −8.02 3.05 −0.03 2.33 Ganada 1.15 6.39 −6.00 2.41 –0.03 2.36 United States 0.85 5.00 −2.41 0.81 −0.00 0.29 Notes * Including Denmark. ** Excluding Denmark. to steady state is greatest in Australia, the United States, and Canada and much smaller in Europe, Scandinavia, and Japan. This confirms our prior expectations. We have so far omitted one important variable in our model. This is the market share of domestic producers, xi. Clearly domestic output and employment are an increasing function of the market share. The reason for this omission is simply lack of data. However, we did experiment with calculating the market share by assuming that it takes the value 1 in year 1978—all domestic customers are customers of domestic firms—and then updating it using the following difference equation: (20) The structuralist perspective 119 where denotes the average real exchange rate over the period 1978–2000—which is our proxy for the PPP real exchange rate and the number 0.2 is only a rough guesstimate of the responsiveness of the market share to real exchange rates. The equation suggests that an elevated (i.e. appreciated) real exchange rate makes customers drift away to foreign firms while a lower value of e makes new customers join domestic firms. Including the market share xi in equation (20) gave a negative coefficient with a t-ratio of 2.21, as predicted by our structuralist theory. This suggests that a transiently larger market share goes together with transiently lower unemployment (hence higher employment rate and higher output). This is what we expected. Apart from this, the results were qualitatively unaffected. We now turn to the most recent employment experience of the United States in light of our theory. The 1990s boom in the United States Figure 8.5 has a plot of the rate of employment in the United States (one minus the rate of unemployment) against the SP500 index (in logs), when the latter has been normalized by labor productivity (all sectors) and the consumer price index (CPI). The long swings in the two series are clearly related—and in a sustained rather than purely transient way. A positive relationship between the two series has also been verified for a range of countries (see Phelps and Zoega (2001)). The persistent unemployment found in a number of Continental economies is simultaneously reflected in the failure of stock prices (normalized) to recover. The fall in the employment rate in the United States in the early 1970s corresponded to a fall in the normalized share price with a common trough in year 1975 and then again in year 1982. There followed a joint recovery peaking in year 2000, followed by a decline in both series. There are also instances when the two series go separate ways: employment expanded in the late 1960s, the late 1970s, and the late 1980s without a corresponding elevation of stock prices. It follows that these may possibly have brought in rising inflation, since a rise in employment above its noninflationary level—or natural rate—creates rising inflation in our model. The recession in 1990–92 also seems to fit this mold although with the reverse sign—employment fell without a corresponding fall in stock prices. In contrast, the rise in share prices in the late 1990s appears not fully reflected in the employment rate. At its peak in year 2000, the employment rate had not yet reached the peak of the late 1960s, although the stock market was much higher. Similarly, the employment rate in 2003 is lower than what could be expected from the stock market, which is still high by historical standards. This may suggest that employment could expand without risking inflation. (More on this later.) Monetary policy and unemployment 120 Figure 8.5 Share prices (normalized) and employment in the United States. Note The employment series is one minus the rate of unemployment (in decimals). The share price series is the SP500 normalized by the CPI index (1995=1) and a measure of labor productivity (1995=1). The value of the SP in 1995 was 470, which is therefore also the value of the normalized series. Figure 8.6 shows the inflation rate and its first difference over the same period. The periods of rising price inflation are the late 1960s, the late 1970s, and—to a lesser extent—the late 1980s. These periods correspond to those when employment expanded without any accompanying elevation of share prices. In contrast, the inflation shock in the mid-1970s is clearly caused by the oil price hikes in 1973–74. Wage inflation also picked up in the late 1970s and the late 1980s. Interestingly, wage inflation rose in the late 1990s to a greater extent than price inflation—the real wage rose during this period. If share prices truly affect the level of the natural rate of unemployment, they should be of use in explaining and predicting inflation. As a prelude, Figure 8.7 has a plot of the relationship between the first difference of the inflation rate (GPI and wages)—that is, unexpected inflation—and the employment rate. Not surprisingly, there emerges no clear The structuralist perspective 121 Figure 8.6 Price and wage inflation in the United States. Note Inflation (in decimals) is measured as the rate of change of the CPI from last quarter of the previous year to the last quarter of the current year. Figure 8.7 The (non) relationship between inflation and employment in the United States. relationship between the two variables. The data clearly reject the joint hypothesis of an expectations-augmented Phillips curve and a constant natural rate of unemployment. The incorporation of share prices should help clarify the relationship between inflation and employment if changes in share prices go hand in hand with changes in the nonobservable natural rate of unemployment. In effect, changes in share prices affect the position of the inflation–unemployment trade-off. We test this by estimating an expectations-augmented Phillips curve of the following form: Monetary policy and unemployment 122 (21) where π denotes inflation (either price inflation (GPI) or wage inflation), 1−u is the employment rate, s is the SP500 share price index normalized by labor productivity g is the rate of (labor) productivity growth and e is the (trade-weighted) effective real exchange rate. The first difference of the employment rate is also included because a rapid expansion—or a rapid convergence to steady state—may be more prone to generate rising inflation. The results appear in Table 8.4. Columns (1) and (2) show results when inflation is measured with price (CPI) inflation while in columns (3) and (4) wage inflation is used instead (wages and salaries in private industry). In the first column, only stock prices are used to predict inflation, in addition to the employment rate, the first difference of the employment rate, and lagged inflation. All three variables have a statistically significant coefficient with the expected sign. We then add (labor) productivity growth and the (effective) real exchange rate to the equation since these should affect the level of the natural rate of unemployment; both an acceleration of productivity growth as well as a real exchange rate appreciation should raise employment in our model. The productivity growth rate has a statistically significant and a positive coefficient while the real exchange rate has an insignificant coefficient. The positive and significant coefficient of the productivity rate implies that higher expected productivity growth lowers the natural rate of unemployment and hence also inflation in equation (21).16 Regrettably when adding the effective real exchange rate one is forced to discard the first 10 observations due to missing data. In columns (3) and (4) wage inflation (rate of change of wages and salaries in private business) is the dependent variable instead of changes in the CPI. The wage data start in year 1975, which shortens the sample period by 15 years. Table 8.4 Estimation of Phillips curves Variables Price inflation Wage inflation (1) (2) (3) (4) Estimate tratio Estimate tratio Estimate tratio Estimate tratio Constant term −0.77 4.07 −0.69 3.08 −0.51 6.23 −0.37 4.07 Employment rate 0.94 3.57 1.29 5.21 0.67 6.17 0.68 6.89 Change of employment 0.91 3.88 0.81 2.88 0.61 6.09 0.45 4.12 Share prices (logs) 0.02 2.74 0.01 1.56 0.02 6.14 0.01 3.50 Productivity growth 0.86 1.87 0.28 1.18 Real exchange rates 0.00 0.14 0.03 1.53 The structuralist perspective 123 Observations 42 42 27 27 R-squared 0.34 0.61 0.67 0.73 R-squared (adj.) 0.29 0.54 0.63 0.67 Notes Em p loyment is measured in decimals. Share prices are normalized by prices (CPI, annual averages) and the level of labor productivity where its value in year 1975 is equal to the unnormalized one. Productivity growth is also measured in decimals and measures the rate of change of average (annual) productivity. Real exchange rates are measured by an index that takes the value 100 in year 1995. However, one gets even better results in this case. The R-squared of the equation is higher and the statistical significance of share prices and the real exchange rate is now much higher, although productivity growth has a somewhat lower t-ratio. The coefficient for the real exchange rate is now clearly positive; a real appreciation reduces inflation. Now take the estimation results from Table 8.4 (columns (1) and (3)) and calculate the difference between the actual and the natural rate of employment—using only share prices—and then use this to predict the first difference of the inflation rate. The results are shown in Figure 8.8. The left-hand-side panel shows actual and predicted change of price inflation while the right-hand-side panel shows the change of wage inflation. In contrast to Figure 8.7, we now have a clear relationship between inflation and its causal variables. The most notable prediction failures are the price inflation shocks in the mid- 1970s and the late 1970s that correspond to the two oil crises. Now invert equation (21) so that it explains the employment rate, and move inflation to the right-hand side: (22) The results follow in Table 8.5. Not surprisingly in light of equation (21), rising inflation causes employment to go up and higher share prices and higher employment go together. Monetary policy and unemployment 124 Figure 8.8 Actual and predicted change of price and wage inflation. Note Predicted change of inflation uses the estimated results in columns (1) and (3) of Table 8.4. Table 8.5 Estimation of employment equation Price inflation (on right-hand side) Wage inflation (on right-hand side) Variables Estimate t-ratio Estimate t-ratio Constant term 0.41 4.37 0.57 3.90 Lagged employment rate 0.50 4.06 0.31 1.75 Inflation shock 0.25 2.49 0.87 6.07 Share prices (logs) 0.01 2.23 0.01 2.80 Observations 42 27 R-squared 0.75 0.87 R-squared (adj.) 0.73 0.86 Note See explanations below Table 8.4. The structuralist perspective 125 Table 9.1 Contribution of demand components to GDP growth (%) GDP growth Consump tion H ousing invest ment Fixed invest ment I nventory invest ment Public consum ption Public invest ment Exports Imports 1990 5.1 2.6 0.3 2.0 −0.2 0.1 0.3 0.7 −0.8 1991 3.8 1.5 −0.5 1.2 0.3 0.2 0.3 0.6 0.3 1992 1.0 1.2 −0.3 −1.1 −0.5 0.2 1.0 0.5 0.1 1993 0.3 0.7 0.1 −1.9 −0.1 0.2 1.2 0.2 0.0 1994 0.6 1.1 0.4 −0.9 −0.3 0.2 0.2 0.5 −0.8 1995 1.5 1.2 −0.3 0.8 0.2 0.3 0.1 0.6 −1.4 1996 5.1 1.7 0.7 1.8 0.4 0.2 0.8 0.8 −1.3 1997 1.6 0.3 0.9 1.5 0.1 0.1 −0.9 1.4 −0.1 1998 −2.5 −0.3 −0.6 −1.4 −0.6 0.1 −0.2 −0.3 0.9 1999 0.2 0.7 0.1 −1.0 0.1 0.1 0.6 0.3 −0.6 Figure 9.1 GDP growth 1950:1– 2001:1. After the asset price bubbles bursted, the Japanese economy officially entered the recession in 1991. At first, it appeared as a normal cyclical downturn, but it was actually only the beginning of the decade long stagnation. The average growth rate of Japan during 1992–98 is exactly 1.0 percent (the first column of Table 9.1). During the same period, the US economy enjoyed the 3 percent growth. The 1 percent growth is even lower than that of the EU which suffers from such high unemployment. As shown in Figure 9.1, during the high growth era of the 1950s and 1960s, the Japanese economy grew by almost 10 percent every year. After the first oil shock of 1973–74, the growth rate fell,1 but still it was 4 percent on average through the end of the 1980s. It was higher than those of most OECD economies. Then the growth rate declined Monetary policy and unemployment 132 further from 4 to 1 percent during the 1990s. The important question is, of course, why the Japanese economy suffered from such a long stagnation. On a close examination of Table 9.1, one finds that the generally depressed 1990s is actually divided into three subperiods: (1) the 1992–93 recession, (2) the 1994–96 recovery and (3) another recession2 during 1997–98. A sensible way to get an overview of the Japanese economy during the 1990s is to look at the demand-decomposition of the growth rate of real GDP. Table 9.1 presents contribution of demand components such as consumption, investment, and exports to growth of GDP. The contribution is here defined as the growth rate of each demand component, say investment, times its share in real GDP. By construction, the figures sum up to the growth rate of GDP. The table shows that fixed investment is by far the most important factor to account for the 1992–93 recession, the 1994–96 recovery and also the 1997–98 recession. In fact, investment is the most important explanatory variable for the Japanese business cycles throughout the postwar period (Yoshikawa 1993). This stylized fact applies to the 1990s. When the growth rate fell from 3.8 to 0.3 percent during 1991–93, for example, the contribution of investment fell from 1.2 to −1.9 percent, accounting for nearly 90 percent of a fall in the growth rate. Similarly when growth accelerated from 0.3 to 5.1 percent during 1993–96, the contribution of investment rose from –1.9 to 1.8 percent, again accounting for 80 percent of the recovery. Thus, to explain the long stagnation of the Japanese economy during the 1990s, we must explain depressed fixed investment. For the 1991–94 recession, we must refer to normal stock adjustment after the long boom during the bubble period (Yoshitomi 1998). And for the 1997–98 recession, the credit crunch played the major role. However, we need to explain why investment stagnated for such a long period on average. After all, investment basically responds to demand; when demand grows, investment also grows whereas if demand stagnates, so does investment. We must, therefore, explain the long stagnation of demand. Other than fixed investment, depressed consumption is notable.3 For 1998, we even observe an unprecedented decline in consumption. Contrary to the common belief, however, a fall in asset prices had relatively small effects on consumption. One might expect that the negative wealth effects depressed consumption after the bubble burst in the early 1990s. Altogether, households enjoyed almost ¥1,200 trillion worth of capital gains on their assets (¥200 trillion on stock, and ¥1,000 trillion on land, respectively) during the bubble period of 1986–90. Subsequently they suffered from the ¥400 trillion worth of capital losses during 1990–92. The analysis of consumption by type of household reveals that capital losses on stock did exert the negative wealth effects on consumption of the retirees and a portion of the self-employed who were the major stock owners. These types of households share only 12 percent, however. The major capital gains and subsequent losses accrued on land. As one would expect, most land which households own is indivisibly related to housing. Therefore, to the extent that housing service and other consumables are weak substitutes, and that land and housing are indivisible, it is not so surprising nor irrrational as it might first appear, that sizable capital gains and losses on land left most households to keep their houses and their consumption basically intact. Capital gains and losses on stock and land affected household consumption only marginally. Bayoumi (1999), using vector autoregressions (VARs) also finds that the The long stagnation and monetary policy in Japan 133 effects of land prices on output largely disappears once bank lending is added as an explanatory variable, and concludes that the “pure” wealth effects are quite limited. Among the factors not taken up as yet to explain depressed consumption we take up job insecurity It is well known that the unemployment rate in Japan had been very low by international standard. During the 1980s, when the unemployment rate reached 10 percent in many EU countries, that in Japan remained 2 percent. The unemployment rate was traditionally low in Japan for several reasons. Thanks to bonus payments and the synchronized economy-wide wage settlements called the Shunto (Spring Offense), wages in Japan are believed to be more flexible than in other countries.4 Furthermore, the necessary adjustment of labor is done through changes in working hours of workers rather than changes in the numbers of workers. On the supply-side, cyclical fluctuations in the labor force participation rate are large; In recessions, when the so-called “marginal” workers (typically female) lose jobs, they often get out of labor force rather than remain in the labor force and keep searching for jobs. These factors kept the unemployment rate from rising.5 Even during the 1992–94 recession, the unemployment rate, though rising, did not reach 3 percent (Table 9.2). However, the long stagnation during the 1990s has thoroughly changed the structure of the Japanese labor market. Most important, with the slogan of “restructuring,” firms are now ready to discharge workers. Table 9.2 shows that the number of involuntary job losers has been more than tripled between 1992 and 1999. In 1999, the unemployment rate in Japan finally became higher than the US counterpart. Until 5 years ago, nobody had expected that it would ever happen. In the autumn of 1997, big financial institutions such as the Hokkaido Takushoku Bank and the Yamaichi Security went into bankruptcy. These events made an unmistakable announcement that the celebrated employment for life in Japan was Table 9.2 Unemployment Year 1992 1993 1994 1995 1996 1997 1998 1999 Unemployment rate (%) 2.2 2.5 2.9 3.2 3.4 3.4 3.9 4.8 The unemployed (10,000) 142 166 192 210 225 230 277 339 Involuntary job separation 32 41 50 55 59 54 74 106 Voluntary job separation 61 69 78 83 87 95 106 107 New school leavers 6 7 9 11 13 12 26 30 Others 36 39 45 50 55 59 63 82 Source: Statistics Bureau, Management and Co-ordination Agency, Monthly Report on the Labour Force Survey. Note For 1998 and 1999, the figures are for March, not the annual average. Monetary policy and unemployment 134 over. Understandably, job insecurity depressed consumption. In 1998, consumption actually fell. This unprecedented event can be well explained by a sharp rise in the unemployment, particularly involuntary unemployment. Nakagawa (1999) demonstrates that uncertainty surrounding the public pension system has also depressed consumption. Besides the major factors such as fixed investment and consumption, there were other relatively minor but still important factors to explain the stagnation of the Japanese economy For example, a sharp increase in imports during 1994–96 hindered the feeble economy from recovery Imports are much more cyclical in Japan than in other countries such as the United States. However, an increase in imports during 1994–96 (the average growth rate 12 percent) was anomalous even by the Japanese standard; Namely the propensity to import sharply rose during the period. Some economists such as McKinnon and Ohno (1997), in fact, attribute the stagnation of the Japanese economy to the high yen.6 However, the appreciation of the yen from 240 per dollar (1985) to 120 (1988) was actually caused by high productivity growth in the Japanese export sector, and, therefore, followed the purchasing power parity (PPP) with respect to tradables (Yoshikawa 1990). Therefore, it is not plausible to regard the appreciation of the yen as the major cause for the long stagnation of the Japanese economy In fact, as shown in Table 9.1, exports had been the most stable component of GDP throughout the 1990s except for 1998 when the Asian financial crisis rather than the appreciation of the yen hindered exports. Having briefly seen the Japanese economy during the 1990s, we now turn to macroeconomic policies. We begin with fiscal policy. Fiscal policy Fiscal policy in the 1990s was in sharp contrast to that in the 1980s. In the fiscal year 1980, after two oil shocks during the 1970s, the budget deficits had become a serious problem; debt finance shared one-third of the budget, and the out-standing debt reached ¥70 trillion. Throughout the 1980s, the single objective of the Ministry of Finance (MOF) was to balance the budget. With the key phrase of the “minus ceiling,” the MOF effectively constrained expenditures. Thanks to an increase in tax revenues during the bubble-boom, the MOF’s goal had been basically achieved by 1990. As of 1990, the deficits/GDP ratio of Japan was lowest among major OECD countries (Figure 9.2); if the social security account is taken into account, the budget was actually in surplus. However, as the recession deepened beginning 1992, the expansionary fiscal policy was called for, and with it deficits mushroomed. The deficit/GDP ratio had reached 10.9 percent by 1999, which is comparable to that of Italy at the beginning of the 1990s. The monotonous worsening of Japan’s budgetary position during the 1990s is indeed in sharp contrast to the trend observed for other OECD countries. With such high costs, how do we assess the fiscal policy during the 1990s? Pessimists say that it was simply a failure because it did not produce any sustained The long stagnation and monetary policy in Japan 135 Figure 9.2 Deficit/GDP(%). Source: OECD (1999). Economics Outlook, No. 65 June. Note Japan* includes the social security account. Public deficit includes both the central government’s and the local government’s. growth. The contribution of public expenditures to growth can been seen in Table 9.1. The table shows that fiscal expenditures sustained growth during 1992–93, and also in 1996. Without fiscal expansion, the Japanese economy would have almost surely suffered from the negative growth in 1993. Posen (1998) goes so far as to argue that the 1994–96 recovery was also generated by fiscal expansion. Posen may exaggerate the role of fiscal expansion in that fixed investment was clearly the most important factor to explain the 1995–96 recovery and that a recovery of investment may not be directly linked to fiscal expansion. However, one can reasonably argue that without fiscal expansion during 1992–93, a prolonged recession with negative growth would have made the 1995–96 recovery impossible, and that in this sense fiscal expansion contributed to the 1995–96 recovery. The 5 percent growth in 1996 made the MOF confident enough to pursue the holy goal of budgetary balance. Many economists observe that a rise in the consumption (value added) tax from 3 to 5 percent and other social insurance contribution amounting to ¥9 trillion, depressed consumption and thereby triggered the 1997–98 recession. The fiscal tightening was actually not confined to the revenues. Table 9.1 shows that public investment was drastically cut in 1997; its contribution changed from 0.8 percent for 1996 to −0.9 percent for 1997, and, therefore, a cut in public expenditures by itself Monetary policy and unemployment 136 lowered growth of real GDP by 1.7 percent. Fiscal tightening through both revenues and expenditures, contributed to the 1997–98 recession. As of the year 2002, the pessimism prevails around fiscal policy. For one thing, the current deficit situation is so bad as to be unsustainable. The primary balance is in deficit. The outstanding public debt is ¥700 trillion (140 percent relative to GDP); this roughly amounts to ¥20 million per family of four, three times as high as its average annual income of ¥6 million. The pessimism not only stems from the size of debts and deficits, but also from mounting doubt about efficiency and justice behind deficits. In every economy the public finance involves transfers of income among households and firms. Transfers of income are, in fact, one of the major purposes of the public finance and, therefore, in itself there is nothing wrong about it. However, there is a broad consensus that aside from the size of deficits, there is a serious problem about the current situation of the public finance in Japan. To understand the problems facing Japan, one can visualize the Japanese economy as a two-sector economy: one consisting of highly efficient manufacturing sector, and the other consisting of inefficient small firms and the self-employed, particularly in the nonmanufacturing sector including agriculture. The existing political system allows significant income transfers from the former to the latter through both public expenditures and taxes, and other social security contributions.7 There is a broad consensus that the ¥100 trillion spent by the public sector during the 1990s contributed very little to raising profitability in the economy Kutter and Posen (2001) take efforts to show that the fiscal multiplier in Japan is 1.7. Most Japanese economists would agree that this figure is reasonable, but think that the size of the multiplier is not really the issue. If not full marks, there is no denying that fiscal policy was on the whole expansionary during the 1990s (Figure 9.2). The fiscal pessimism does not stem from the fear that the multiplier is small, but from the fact that fiscal expansion did not bring about sustainable growth. Even now, an increase in public expenditures surely raises GDP, but may not necessarily produce sustainable growth. Some say that if not, spend more. But until when? The debt/GDP ratio is currently 140 percent, and many economists question whether the deficits are sustainable. Monetary policy Monetary policy is widely believed to be responsible for the asset price bubbles during the late 1980s and the subsequent long stagnation during the 1990s.8 According to this view, during the 1980s, low interest rates produced the asset price bubbles, and the high land prices, in turn, allowed the liquidity constrained firms to make excessive investment by way of an increase in the collateral values. For the same reason, but now to the opposite direction, the collapse of the asset market entailed the stagnation of investment during the 1990s. Though this “standard” view contains a bit of truth, it does not actually stand up to careful analyses. There are actually a number of studies which demonstrate a significant relationship between real variables such as investment and real GDP on the one hand, and asset prices, land prices in paticular, on the other. Since asset prices and GDP went up and down broadly in tandem, these findings are not surprising. The problem is interpretation. The long stagnation and monetary policy in Japan 137 Most of the analyses interpret their findings as indicating that changes in asset prices affected investment of financially constrained firms by way of changes in their collateral values; Ogawa and Suzuki (1998), for example, find land prices significant in their investment functions. Bayoumi (1999) also finds in his VARs that land price changes were an important factor behind the rise in the output gap over the bubble period and the subsequent decline.9 However, this is not exactly what happened in Japan during the late 1980s and the 1990s. During the bubble period, it was believed (falsely in retrospect) that land-intensive sectors, such as holiday resorts and office spaces in Tokyo, would command high profits in the near future. These (false) expectations made land prices explode, and at the same time induced firms to make land-intensive investment. Firms purchased land with money borrowed from banks, and banks, based on their expectations of higher land prices in the future, often allowed more than 100 percent (!) collateral values for land which firms just purchased. Therefore, theoretically firms could borrow money from banks without any collaterals in advance to purchase land. This is very different from the standard story explained earlier, according to which an increase in the price of land which firms had owned in advance made it possible for the liquidity constrained firms to borrow more money to make desired investment. In fact, the ultimate cause of both a rise in land prices and an extraordinary surge in land-intensive investment was false expectations on future profitability of holiday resorts and office spaces in Tokyo. In short, misallocation of capital was the ultimate cause. After the bubbles burst, the asset prices collapsed, and at the same time investment also fell. However, it is once again not self-evident that this fact suggests that a fall in the asset prices cut investment by way of a fall in the firms’ collateral values. For example, investment of large firms and small firms fell during the 1992–94 recession roughly in the same magnitudes. Large firms do not finance their investment by borrowings from banks but rather by issuing bonds, and new equities in capital market. They are not financially constrained, and, therefore, the collateral story does not hold true for large firms at the outset. And yet, investment of large firms fell in the same magnitude as that of small firms. Meltzer (2001), and Hayashi and Prescott (2002) also express skeptical views against the significance of financial constraints. Thus, a careful study is necessary to determine how a fall in the asset prices affected investment during the 1990s. Now, let us briefly review the record of monetary policy during the period. The call rate (the overnight money market rate), the most important instrument of monetary policy in Japan, was kept high at the level of 7.5 percent during 1990–91 (see Table 9.3). The stock price which reached ¥38,915 at the end of 1989 had fallen below ¥15,000 by 1992. The land price also started falling the same year. The economy was in deep recession in 1992. The BOJ had already lowered the discount rate from 6.0 to 5.5 percent in July 1991. Through five successive cuts within a year, it had fallen to 3.25 percent by July 1992. Monetary policy and unemployment 138 Table 9.3 Interest rates (1) Call rate (%) (2) The 10- y eargovernmentbonds (%) (3) The long lending rate (%) (4) The term premium (2)−(1)(%) (5) The private risk p remimum (3)−(2) (%) 1990 7.4 6.8 8.1 −0.6 1.3 1991 7.5 5.8 6.9 −1.7 1.1 1992 4.7 4.8 5.5 0.1 0.7 1993 3.1 3.5 3.5 0.4 0.0 1994 2.2 4.6 4.9 2.4 0.3 1995 1.2 2.9 2.6 1.7 −0.3 1996 0.5 2.8 2.5 2.3 −0.3 1997 0.5 2.0 2.3 1.5 0.3 1998 0.3 1.0 2.2 0.7 1.2 1999 0.03 1.8 2.3 1.8 0.5 Despite the further cuts in the interest rates during 1993–94, the economy hardly revived. The annual growth rate of money supply (M2+CD) which was 12 percent in 1990, had fallen to zero by 1992. Since a sharp decline in bank lending was basically responsible for this fall in money growth, the question was why this sharp decline in bank lending occurred. Bayoumi (1999) interprets his finding that bank lending is more important than land price itself in explaining output gap as supporting the financial disintermediation hypothesis. He argues that “undercapitalized banks responded to falling asset prises and other balance sheet pressures by restraining lending to maintain capital adequacy standards.” Some economists in Japan suggested the same, and argued that the credit crunch was responsible for the weak investment. However, as shown in Table 9.3, during 1991–93, the interest rates kept declining. If the credit crunch had occurred, the interest rate would have risen. Therefore, the major cause of a sharp decline in bank lending during 1991–93 was a downward shift of demand curve (a fall in demand for bank lending) rather than an upward shift of supply curve (the credit crunch or a cut in supply of bank lending). As shown in Figure 9.1, the diffusion index of “Lending Attitude of Financial Institutions” of the BOJ Tankan (Short-term Economic Survey of Corporations) indeed shows that responding to successive cuts in the call rate, banks’ lending attitudes improved during 1992–95. Gibson (1995) also concludes that, although a firm’s investment is sensitive to the financial health of its main bank, the effect of the problems in the banking sector on aggregate investment during 1991–92 was small. The private risk premium defined as the difference between the long lending rate and the 10- year government bond rate, also declined during the period (Table 9.3). In summary the effects of a fall in land prices and consequent bad loans on bank lending was not significant during the 1992–94 recession. By looking at bank level data, Woo (1999) draws the same conclusion. The long stagnation and monetary policy in Japan 139 The economy recovered during 1994–96. With easy monetary policy the call rate had kept declining. However, the long-term interest rate had stopped declining to raise the term premium (Column (4) of Table 9.3). This is actually something one should expect when the economy is on the road to recovery. In fact, the growth rate of real GDP reached the respectable 5.1 percent in 1996. Facing this recovery the MOF decided to tighten budget for 1997. Meanwhile, a fall in the stock price created a serious problem for the Japanese banks to meet the BIS capital adequacy standards. The new legislation in April 1996 allowed the authority to step in if a bank were likely to fail to meet the BIS requirement. This new policy regime was to start in April 1998. In March 1997, the MOF made clear the new capital adequacy requirements. Unfortunately this basically correct policy action was not well-timed. Desperate to raise the capital–asset ratio within a short period of time, banks squeezed their assets by cutting lendings. In the autumn, the bankruptcy of big financial institutions such as the Yamaichi Security, and the Hokkaido Takushoku Bank triggered the real credit crunch. Figure 9.3 shows that the Tankan DI of lending attitude of banks abruptly worsened during this period despite the BOJ’s efforts to ease money.10 What was the impact of this credit crunch? Motonishi and Yoshikawa (1999) assess the macroeconomic magnitude of the credit crunch by estimating investment functions separate for large and small firms in both the manufacturing and non-manufacturing sectors. The explanatory variables are from the BOJ’s Tankan, which has the diffusion indices for business conditions and for credit constraints facing firms shown in Figure 9.3. As one might expect, they find that credit constraints are not significant for investment of large firms, but are significant for small firms, particularly in the nonmanufacturing sector. They conclude that the credit crunch, by way of depressing Figure 9.3 Lending attitude of financial institutions. Source: Bank of Japan, “Tankan Short-term Economic Survey of Enterprises in Japan.” Monetary policy and unemployment 140 investment of financially constrained firms, lowered the growth rate of real GDP by 1.3 percent during 1997–98. Their analysis takes into account only fixed investment, but actually two-thirds of bank lendings is for running costs and inventory investment rather than fixed investment. We can, therefore, reasonably argue that at the minimum, the credit crunch accounts for one half of −2.5 percent growth of real GDP in 1998. The low interest rates were of no help in the credit crunch. The renewed recession forced the BOJ to lower interest rates further. The call rate became 0.3 percent in 1998, and finally 0.03 percent in 1999. With transaction costs, 0.03 percent effectively means zero-interest rate, the absolute minimum for nominal interest rate. The BOJ, thus, lost the instrument for traditional monetary policy. Economists then started discussing how monetary policy could possibly affect the economy facing zero-interest rate. Krugman (1998) argues that with zero nominal interest rate, the Japanese economy is caught in the liquidity trap, and that the BOJ must create expected inflation to lower the real interest rate to get the economy out of this trap. How to generate inflationary expectations? Increase money supply! The answer seems so obvious. Is this policy effective and/or feasible? In Krugman’s model and the similar proposals for lowering the real interest rate by way of generating inflationary expectations, demand is assumed to be interest elastic. However, in the Japanese economy during the 1990s, a major problem facing monetary policy is low interest elasticity of demand, in particular, investment. Indeed, the Japanese economy was not trapped in the zero-interest rate from the beginning. Table 9.2 shows that in 1992, the call rate was still 4.7 percent, and that it had been lowered to zero by 1999. Low interest rates, in fact, together with preferential taxes, pushed housing investment, but did not revive fixed investment. Based on the interest elasticity for the US economy Krugman suggests that to fill the 5 percent GDP gap, the 3–3.75 percent inflationary expectations would be enough. However, with low interest elasticity which appears to hold for the Japanese economy during the 1990s, the necessary expected inflation will easily become as high as 30 percent! Beyond that, in Krugman’s model, the “future” is not in liquidity trap, and the simple quantity theory of money is assumed to hold in the future: price is proportional to money supply in the future. Thus, in theory it is easy for the Central Bank to generate the expected inflation despite the absence of the current actual inflation. The only thing the Central Bank must do is to persuade the public now to believe that money supply will increase enough to generate inflation in the future. However, in reality the most important factor to determine the expected inflation is the current actual inflation. Whatever the policy actions of the Central Bank, who would believe in inflation so easily in the economy actually facing deflation? As long as we believe in the Phillips curve wisdom, namely the story that only high pressure in the real economy produces inflation, then we are likely to be led to the catch 22 in our effort to cure recession by generating inflationary expectations! Blanchard (2000:190–3) states that “the Phillips curve wisdom remains largely true in modern treatments of the determination of prices, wages, and output: If output is above its natural level, then we are likely to see inflation increase.” Despite such a remark, he is optimistic that the BOJ can generate inflationary expectations to lower the real interest rate; “All that is needed is to convince markets that money growth will be cumulatively higher over the next 10 years by 20 percent.”11 He notes that monetary policy affects The long stagnation and monetary policy in Japan 141 of this function determines the sign of the denominator of the implicit function derivative in equation (9). Figure 13.2 Numerator slope (δ). Notes This figure graphs the function f(θ) in (9) depending on δ and holding the remaining parameters fixed. The slope of this function determines the sign of the denominator of the implicit function derivative in equation (9). To understand the intuition behind this result, recall that the loss function in (5) implies that the Fed will react more strongly to expected inflation when changes in the interest rate are more effective in reducing expected inflation. Our structural macro model contains two monetary transmission mechanisms from the interest rate to inflation. Interest rate changes affect the output gap through the real rate in the IS equation. In turn, output gap movements influence inflation through the Phillips curve relation in the AS equation. The second channel consists of expectational effects. As the monetary authority reacts more aggressively to inflation, the private sector adjusts its inflation expectations which directly affect the inflation rate. We now show how the effectiveness of monetary policy varies with the endogenous persistence of inflation and the output gap. At small values of δ, contractionary monetary policy is quite effective in reducing expected inflation, as future inflation still depends heavily on current inflation. Hence, as agents become more forward-looking when δ is still not very large, current inflation will experience a larger reduction following an interest rate increase (since the term δEtπt+1 in the supply equation will be larger in absolute value). This reinforces the decline in Monetary policy and unemployment 244 expected inflation, making contractionary monetary policy more effective. However, at high values of δ, expected inflation does not greatly depend on current inflation. Accordingly increases in δ will make expected inflation even less dependent on future inflation (and will also make the term δEtπt+1 smaller in absolute value), so that monetary policy will be less effective in reducing expected inflation. This implies the existence of a cutoff value δ* such that for δ>δ*, it is no longer optimal to react more aggressively to expected inflation as δ grows. Figure 13.2 performs a calibration exercise around different values of the Phillips curve parameter λ. It shows that when monetary policy is less effective in reducing inflation volatility (for smaller values of λ), the cutoff value is smaller, whereas for larger values of λ, the cutoff value is higher. Finally Figure 13.3 presents the impulse responses of inflation to a monetary policy shock. It shows that for low values of δ in the AS equation, as agents become more forward-looking, monetary policy is more effective in reducing inflation for about 15 periods. However, when δ is very large, the opposite is true during the 50 periods following the shock. These two pieces of evidence corroborate the ideas in the previous paragraph.6 Figure 13.4 graphs f(θ) as a function of µ, the forward-looking parameter in the IS equation. It is decreasing up to values of µ close to 0.4, when it starts to increase. As a result, for values of µ close to 0.5, as estimated in our sample, increases in the forwardlooking behavior of agents in the IS equation should be followed by a smaller reaction of the Fed to expected inflation. The reason is that for large values of µ, a smaller output gap persistence will make monetary policy less effective, since the way it influences inflation is by contracting the output gap. However, when µ is below 0.4, increases in the forward-looking parameter in the IS equation will reinforce the effect of contractionary monetary policy on the current output gap (through a larger value of the µEtyt+1 term in the IS equation). The mechanism is analogous to that of δ and is illustrated in Figure 13.5. When µ is small, as it grows, monetary policy becomes more effective in reducing inflation, whereas the opposite is true when µ is large. Our results resemble those obtained by Lansing and Trehan (2003) in a related optimal monetary policy exercise. There are, however, some differences. In their case, as µ grows, it is optimal to react less aggressively to inflation as long as µ is greater than 0.1, whereas in our chapter, the cutoff value µ* is 0.4. While Lansing and Trehan (2003) include output gap stabilization in their Central Bank loss function, they also consider a different structural model. Whereas in this chapter the expectations influence all the variables contemporaneously, in their case it is the lagged expectations which affect the current period variables. This implies that in our exercise the expectations and the current values affect each other simultaneously so that as µ grows, monetary policy is more effective up to higher values of µ (since the current output gap will react more strongly to interest rate changes). Our results also differ in the case of optimal monetary policy for different values of δ. In Lansing and Trehan (2003), it is not until δ is around 0.95 that increases in δ should restrain the monetary policy reaction, whereas in our case the cutoff value is around 0.60. This occurs because monetary policy starts being ineffective earlier in our case. The Fed’s monetary policy rule 245 Figure 13.3 Inflation response to a monetary policy shock for different values of δ. Notes This figure presents the response functions of inflation to a monetary policy shock. It presents responses under alternative values of δ. The remaining model’s parameters are held at their second period estimates. Monetary policy and unemployment 246 Figure 13.4 Numerator slope (µ). Notes This figure graphs the function f(θ) in (9) depending on µ and holding the remaining parameters fixed. The slope of these functions determine the sign of the numerator of the implicit function derivative in equation (9). The Fed’s monetary policy rule 247 Figure 13.5 Inflation response to a monetary policy shock for different values of µ. Notes This figure presents the response functions of inflation to a monetary policy shock. It presents responses under alternative values of µ. The remaining model’s parameters are held at their second period estimates. To understand the different optimal Fed’s behavior under increases of µ and δ in our exercise (µ* is larger with respect to Lansing and Trehan (2003) whereas δ* is smaller), notice that a given increase in a parameter results in a larger percentage increase at small parameter values. This effect dominates at small values of µ, where initial increases of µ have a sizable effect on the output gap, with the out-put gap being still quite persistent. However, in the case of δ, when this parameter is quite large, further increases of δ would have two effects. On the one hand, the percentage increase in δ is small and on the other hand, it exacerbates the already small degree of persistence in inflation, making expected inflation depend even less on current inflation. These two effects are amplified in our exercise, where expectations affect the macro variables contemporaneously. The results in this section illustrate the importance of the expectational effects in monetary policy management. As the private sector becomes more forward-looking, expectations of the future variables behave differently so that the effects of monetary policy actions also differ. One important implication of our study is that the monetary authority should not modify its reaction to inflation monotonically as changes in the private sector behavior occur. Monetary policy and unemployment 248 Conclusions This chapter shows that there was an economically but not statistically significant change in the preferences of the Federal Reserve after 1980 towards inflation stabilization. We also show, in the context of a strict inflation targeting regime, the optimal changes in the Fed’s reaction to expected inflation when the forward-looking behavior of the private sector changes. The importance of the private sector’s degree of forward-looking behavior has been high-lighted in this chapter. There have been several attempts in the literature to derive aggregate supply equations featuring both forward- and backward-looking components. Fuhrer and Moore (1995), for instance, develop a real wage contracting model with endogenous persistence. In this case, the persistence is induced by the existence of wagesetters who adjust their current real wages with respect to past real wages. One caveat of these works is that the endogenous persistence of inflation is not ultimately grounded in optimizing behavior. A better understanding of the sources of inflation persistence would be desirable as the Fed’s optimal policy changes with it. Appendix Computing the partial derivatives Equations (6) and (7) show that the optimal β and γ coefficients in our reaction function depend on partial derivatives of the target variables with respect to interest rate changes. In order to compute these partial derivatives, we recognize both an endogenous and an exogenous part in our model’s interest rate: it=ît+ĩi (10) where, in mean deviation, ît=ρit−1+(1−ρ)(βEtπt+1+γyt) and Therefore, ĩt constitutes the exogenous part. In this setting, we can proxy the partial derivative terms involving changes in ĩt changes in ĩt by applying vector differentiation rules to our model solution. The implied model’s solution is: , where in demeaned form. Since the next period expectations can be expressed as we can obtain: (11) Therefore , so that ∂Etπt+1/∂ĩt is simply the (1, 3) element of the product matrix ΩΓ, that is: (12) The Fed’s monetary policy rule 249 In order to obtain the partial derivative ∂yt/∂ĩt in equation (7), we follow the procedure described earlier and use the IS equation to obtain: (13) Optimal Fed behavior In this second part of the Appendix, we derive the optimal changes in the Fed’s reaction to expected inflation when the behavior of the private sector in the AS and IS equations varies. We reproduce equation (6) for ease of exposition: (14) In the previous section of the Appendix, we computed an approximation to the term. ∂Etπt+1/∂it. As can be seen in equation (12), this term depends on the reduced form elements of the model’s Rational Expectations solution which, in turn, also depend on β. Therefore, we can apply the Implicit Function theorem to equation (14) so as to determine how the Fed would change its reaction to expected inflation when the private sector becomes more forward-looking in the supply and demand equations. Let us first introduce some additional notation: (15) Then, we can rewrite (14) as: (16) Using the Implicit Function theorem: (17) In order to obtain the sign of the partial derivatives, we can compute numerically vectors of the Ωij and Γij terms in (12) as a function of δ, µ, and β holding the remaining parameters constant. In this way we can construct f(θ) in (12) as a function of each parameter, so that we can graphically identify the sign of the terms ∂f/∂θi, ∂f/∂β, and, in turn, the sign of the derivative for δ, . Figures 13.1, 13.2, and 13.4 graph the functions involved in the implicit function derivatives in (17). Monetary policy and unemployment 250 Notes 1 Clarida et al. (1999), Woodford (2003) or Lansing and Trehan (2003) are some examples. 2 Söderlind (1999), and Cecchetti and Ehrman (2000) obtain estimates of deep parameters. However, their work is not focused on the shift of the preferences of the US Fed in the early 1980s. Their methodology is also quite different to ours. 3 Both estimations yield a stationary Rational Expectations solution. The first period estimates imply multiple equilibria. In this instance, we choose the equilibrium associated with the Recursive Method in Cho and Moreno (2003) which selects the bubble-free equilibrium. 4 Svensson (2003) criticizes forecast-based instrument rules of this kind on the grounds of time inconsistency. Our goal in this chapter is to provide a deeper interpretation of the coefficients in an interest rate rule which seems to capture the short-term interest rate dynamics quite closely. Accordingly, the loss function which we consider does not include any term of period other than the current one. This precludes the appearance of the term Etit+1 as an argument in the reaction function. 5 In a related exercise within a more stylized framework, Cecchetti and Ehrman (2000) show that the Central Banks of several countries which adopted inflation targeting put more weight on inflation deviations after the adoption date. 6 We stress that our results are limited to the case of strict inflation targeting. Acknowledgments I thank Seonghoon Cho, Michael Ehrmann, Kevin Lansing and Paolo Surico for helpful comments and suggestions. Financial support from the Fundación Ramón Areces is gratefully acknowledged. References Bai, Jushan, Robin L.Lumsdaine, and James H.Stock (1998). “Testing for and dating breaks in stationary and nonstationary multivariate time series,” Review of Economic Studies 65, 395– 432. Boivin, Jean, and Marc Giannoni (2003). “Has monetary policy become more effective?,” NBER Working Paper Mo. 9459. Calvo, Guillermo (1983). “Staggered prices in a utility maximizing framework,” Journal of Monetary Economics 12, 383–98. Cecchetti, Stephen, and Michael Ehrman (2000). “Does inflation targeting increase output volatility? An international comparison of policymakers’ preferences and outcomes,” Central Bank of Chile Working Paper Number 69. Cho, Seonghoon, and Antonio moreno (2003). A Structural Estimation and Interpretation of the Mew Keynesian Macro Model, mimeo, Columbia University. Clarida, Richard H., Jordi Gali, and Mark Gertler (1999). “The science of monetary policy: a new Keynesian perspective,” Journal of Economic Litemture 37, 1661–707. Clarida, Richard H., Jordi Gali, and Mark Gertler (2000). “Monetary policy rules and macroeconomic stability: evidence and some theory,” Quarterly Journal of Economics 115, 147–80. Fuhrer, Jeffrey G. (2000). “Habit formation in consumption and its implications for monetarypolicy models,” American Economic Review 90, 367–89. Fuhrer, Jeffrey G. and George Moore (1995). “Inflation persistence,” Quarterly Journal of Economics 440, 127–59. The Fed’s monetary policy rule 251 Lansing, Kevin J. and Bharat Trehan (2003). Forward-looking behavior and the optimal discretionary monetary policy Economic Letters 81, 103–10. Moreno, Antonio (2003). “Reaching inflation stability” forthcoming in the Journal of Money, Credit and Banking. Rotemberg, Julio J. and Michael Woodford (1998). “An optimization-based econometric framework for the evaluation of monetary policy: expanded version,” NBER Working Paper No. T0233. Soderlind, Paul (1999). “Solution and estimation of RE macromodels with optimal monetary policy,” European Economic Review 43, 813–23. Svensson, Lars (2003). “What Is wrong with Taylor rules? Using judgment in monetary policy through targeting rules,” Journal of Economic Literature 41, 426–77. Woodford, Michael (2003). Interest and Prices: Foundations of a Theory of Monetary Policy, Chapter 3, Princeton, NJ: Princeton University Press. Monetary policy and unemployment 252 14 What is the impact of tax and welfare reforms on fiscal stabilizers? A simple model and an application to EMU Marco Buti and Paul Van den Noord Introduction Taxation inevitably impinges on most aspects of economic activity, and thus careful consideration must be given to its design—in addition to its level and hence the level of related expenditure. So long as taxation affects incentives, it may alter economic behavior of consumers, producers or workers in ways that reduce the amount or utilization of physical, human and knowledge capital, and thus growth. Therefore, to the extent the tax system matters for economic efficiency its costs are likely to rise with the level of taxation. The widespread perception that in many European countries, the tax burden is too high and the tax system unduly distortive has led to calls for tax reforms. Empirical research suggests that a cut in the tax share in GDP by 1 percentage point raises output per working-age person in the long run by 0.6–0.7 percent (OECD 2000). While policy-makers’ efforts to streamline the welfare state and enact tax reforms that aim to bring down the tax burden may thus pay off in terms of better efficiency this may come at a cost in terms of weaker fiscal automatic stabilization. This trade-off between stabilization and efficiency would be particularly unpalatable in Economic and Monetary Union (EMU) countries, since they already have lost national monetary policy and the exchange rate as adjustment mechanisms to country-specific shocks. Indeed, EMU members would ideally aim for both stronger fiscal stabilization and higher economic efficiency and a trade-off between the two would be quite unwelcome. Fortunately this difficult trade-off may not always be relevant. In other papers (Buti et al. 2003a,b) we have shown that there may be a level of the tax burden beyond which reducing it may not only yield better efficiency but, depending on the nature of economic shocks, also render fiscal automatic stabilizers more effective. If supply shocks tend to prevail, a reduction in the tax burden might carry a “double dividend” of efficiency gains and better fiscal stabilization properties. This conclusion draws on evidence that lower taxation improves the terms of the short run inflation-unemployment trade-off (i.e. makes the Phillips curve flatter), by reducing the wedge between the marginal cost of labor and the marginal take-home pay. This is encouraging for countries with high tax burdens that are considering a reduction in the size of the public sector. The present chapter takes this analysis further, by introducing a distinction between the “optimal” tax burden at which, under supply shocks, the automatic stabilizers are most powerful and beyond which favorable stabilization properties decline, and a Figure 14.4 The effects of a negative supply shock under alternative tax rates. Taking into consideration the possibility of the supply curve becoming steeper as well, automatic stabilization may become, however, inflation destabilizing. From the second panel in Figure 14.3, one can notice that this will still lead to a closer output to its optimal level but to a higher inflation. Hence, in this case an increase in the tax rate risks becoming inflation destabilizing beyond a certain point if the slope of the supply curve is more sensitive to the tax burden than the slope of the demand curve. We turn now to the analysis of a supply shock. As shown in the left panel of Figure 14.4, an adverse supply shock induces a shift of the supply curve to the left. The new equilibrium point is now at A with a low tax burden, and at B with a high tax rate. One can easily notice that the new equilibrium level of output is further away from the initial level with a low tax rate than with a high one. The reverse emerges for inflation. Hence, in this case an increase in the tax rate from a low value to a high one is output stabilizing but inflation destabilizing. The increase of the tax rate may become, however, output destabilizing if the supply curve also becomes steeper due to high taxation, as shown in the second panel of Figure 14.4. The new equilibrium point is now at C with a high tax burden. It is clear from the graph that the new equilibrium level of output is further away from the initial level with a high tax rate than with a low one. Inflation is always further away from its optimal level with a higher tax rate. Hence, in this case an increase in the tax rate from a low value to a high one is both output destabilizing and inflation destabilizing. “Critical” levels of taxation The previous analysis shows that the changes of taxation to become output-destabilizing rise with the supply curve becoming steeper.11 On the other hand, the output-destabilizing effect diminishes as the demand curve become steeper. Since the slope of both curves depends on the tax rate, the threshold level for the tax rate beyond which further increase of taxation is destabilizing for output in the event of a supply shock depends on the Monetary policy and unemployment 260 relative sensitivity of demand and supply to taxation. This, in turn, depends on the openness of the economy: the more open the economy the lower will be the fiscal demand multiplier and therefore the steeper will be the supply curve relative to the demand curve for a given tax burden. Therefore, open economies are more likely to face adverse fiscal stabilization properties in the face of a supply shock than relatively closed economies for a given level of taxation (and progressivity).12 It is also easy to show that always ∂y/∂t<0 for a positive demand shock (εd>0) and ∂π/∂t<0 for an adverse supply shock (εs<0). As was shown in the graphs in the previous section, this implies that a higher t (or ξ) unambiguously increases the stabilization of output in the event of demand shocks, and destabilizes inflation in the event of a supply shock. However, in the case of a response of output in the case of supply shocks, or inflation in the case of demand shocks, the initial level of t matters. In line with the intuition, we show a higher t to entail stronger output stabilization in the event of demand shocks while it is inflation destabilizing in the event of demand shocks. The crucial result concerns output stabilization in the event of a supply shock and inflation stabilization in the case of a demand shock. In the traditional model in which taxes do not affect supply higher taxes tend to stabilize both variables. In our model, instead, there exists a threshold level of taxation beyond which a further increase in taxes has perverse stabilization effects. We consider two concepts of the threshold tax level: the “optimal” t, call it t*, which maximizes output and inflation stabilization in the event of supply and demand shocks, respectively; and the “critical” t, call it t**, which corresponds to the level of taxation resulting in zero fiscal stabilization (i.e. the same level of stabilization arising when t=0). t* is obtained by taking the derivative of the coefficient of εd in π or the coefficient of εs in y to t and equating the result to zero: (11) Hence, for t>t*, a rise in t reduces the degree of output stabilization in the event of supply shocks, and inflation stabilization in the event of demand shocks. t** is obtained by equating the coefficient of εd in π or the coefficient of εs in y to the same coefficient under t=0: (12) So t**=2t*. Some intuitively appealing conclusions can be drawn from this result: 1 It appears that there exists a trade-off between the redistributive thrust of the tax and benefit system (ξ) and the tax burden (t): the less the redistributive taxes and benefits are, the higher will be the critical tax rate, and hence the wider is the range of tax rates whereby automatic stabilizers are effective. 2 The same applies to the degree of wage resistance (γ): the higher it is, the lower will be the optimal (and critical) tax rate, because the more the level and redistributive thrust What is the impact of tax and welfare reforms on fiscal stabilizers? 261 of taxation and spending matter for wage formation, the bigger will be its impact through the supply channel. 3 The threshold level of the tax rate above which automatic stabilizers become destabilizing, depends on the responsiveness of demand to the fiscal impulses stemming from the automatic stabilizers . The weaker this responsiveness, (e.g. because of Ricardian behavior) the lower the tax rate that can be “afforded” without risking declining or perverse stabilization properties. 4 The threshold varies inversely with the weight of output stabilization in the Central Bank’s reaction function (β). A dovish Central Bank will choke off the output effect of automatic stabilizers and thus weaken their effectiveness. Interestingly this implies that the incentives to reform the tax and welfare system are lower under a hawkish central banker,13 although incentives to reform the tax system on efficiency grounds would obviously be decisive. 5 A greater openness of the economy reduces the threshold level of taxation. The reason is that the demand effects of automatic stabilizers leak out via foreign trade, implying that the negative supply effects predominate more quickly that is, even at a lower level of taxation. This is analytically similar to the third point of the list, but may be usefully highlighted separately This is so because while trade leakage is related to the openness of the economy policy transmission may be weak even in a closed economy Open economies in the EMU are thus facing stronger incentives to reform their tax systems than the relatively closed ones. How large are t* and t**?—some numerical simulations The typical tax burden in EMU countries is in the range of 40–50 percent of GDP. Is this exceeding the optimal level and would a reduction in the fiscal size thus work out favorably for stabilization? Is it empirically possible or even likely that the tax burden exceeds the critical tax burden? While a full-fledged analysis is well beyond the scope of this chapter, we can nonetheless provide some tentative indication of the possible values of t* and t**. It goes without saying that our computations are purely illustrative and that one should refrain from drawing policy conclusions from the simple comparison of the estimated t* and t** with the actual tax burden in Euro-area economies. Nevertheless, these estimates are helpful in exemplifying our reasoning. In Table 14.1, we report the chosen baseline values of the coefficients. With regard to the demand equation, we assumed that and which is broadly in line with the short run elasticities reported in ready-reckoners of the OECD’s INTERLINK model (Dalsgaard et al. 2001). The budget elasticity—encompassing both, spending and revenue—is set at based on Van den Noord (2000). We assume a hawkish banker, that is, and β=0, with the country’s weight in the monetary policy reaction function set at . Concerning the supply equation we assumed that ω=3, which corresponds to the mid-range of estimates of the price elasticity of aggregate Monetary policy and unemployment 262 supply reported in Clarida et al. (1998).14 To gauge the degree of wage resistance, we proceeded somewhat differently Rather than making a prior assumption for γ, we fixed the incidence of labor taxation on profits at one half, that is, γ. . This implies that γ=0.4. This is consistent with the evidence of Alesina and Perotti (1997), which estimate a coefficient of 0.4 for countries in continental Europe in the relation between labor taxes and unit labor costs in manufacturing in a sample of annual data from 14 OECD countries. Table 14.1 Baseline parameters ξ=1.25 ω=3 λ=0.25 α=1.5 γ=0.4 β=0 Figure 14.5 Baseline simulation. Note The horizontal axes indicate the tax burden (t) and the vertical axes, the impact of a shock (normalized at unity) on the output gap or inflation. What is the impact of tax and welfare reforms on fiscal stabilizers? 263 Table 14.2 Sensitivity analysis t* t** Base line 0.4 0.8 ξ=1 0.5 1 β=1 0.35 0.7 =0.75 0.3 0.6 =0.75 0.2 0.4 γ=0.5 0.2 0.4 β=1, λ=0.5 0.3 0.6 Based on these assumptions, we find that t*=0.4, and t**=0.8, which suggests that for countries in the upper end of the range, the tax burden would be suboptimal, but well below the critical level (see Figure 14.5). This implies that a country with an initial tax burden of 50 percent that would cut it by 10 percentage points realizes a slight improvement in the output stabilization properties after an adverse supply shock. The same holds true for the impact on prices after a positive demand shock. However, these results may be expected to be rather sensitive to the numerical assumptions and hence, if this proves true, the structural features of the economies in EMU. This is confirmed by sensitivity analysis. As shown in Table 14.2 and in the corresponding figures in the Annexure, a reduction in the budget elasticity from to 1 raises the value of t* to and t** to 1. In other words, a tax burden equal to one half of GDP may still be optimal from a stabilization point of view if the tax and benefit system is proportional. By contrast, a greater openness of the economy , a less effective fiscal policy and greater wage resistance all push t* into a range of 0.2−0.3 and t** into a range of 0.4–0.6. Under those conditions, slashing the size of government would pay substantially in terms of the gains in fiscal stabilization properties that would be realized. From Table 14.2 can be inferred that a similar scope for reductions in the size of government results if the central banker turned dovish to an extent where it gives a positive weight to output and inflation in its policy reaction function (β is set equal to 1). This effect is even more pronounced for larger countries that have a bigger weight in the reaction function (e.g. λ=). Interestingly, this result runs somewhat counter to the general perception that a hawkish central banker would be more successful in raising incentives for structural reform than a dovish one. Our results are broadly in line with recent empirical investigations, which have found evidence of a nonlinear relationship between the size of the government and macroeconomic stability. Martinez-Mongay and Sekkat (2003) test whether the structure of the tax system affects the impact of tax changes on output volatility. In a sample of 25 OECD countries over the period 1960–99, they find that the composition of tax and expenditure, in particular the tax mix, matters for output and price volatility: distorting taxes, namely taxes on labor and capital, tend to have negative effects on macroeconomic stability. Monetary policy and unemployment 264 Cuaresma et al. (2003) find that the smoothing effect of fiscal stabilizers may revert at high levels. In a panel of 14 EU countries over the period 1970–99, the stabilizing effect changes sign at a level of government expenditure of about 38 percent of GDP. According to their results, for a country displaying a public expenditure ratio around the median value of the distribution (40.6 percent of GDP), an increase in spending by 1 percentage point of GDP will raise the standard deviation of output growth by 0.02 points. The destabilizing effect is higher (0.04 percent) for a country with an expenditure ratio of 44.1 percent. However, this study is not entirely comparable to ours as it focuses solely on government spending and does not distinguish between automatic stabilizers and discretionary policy reactions. Conclusions Conventional AD-AS models imply that high and progressive tax systems are efficiencydecreasing but enhance output stabilization in the event of shocks. Progressive tax systems lead to a lower budget deficit (contraction of fiscal policy) in good times, while the deficit would increase in recessions (fiscal expansion). Moreover, large and progressive tax systems usually go hand in hand with more generous systems of social protection. Although social benefit programs mainly have an equity role, as well as potential efficiency effects when they correct market failures, most of them also act as automatic stabilizers. Unemployment benefits make up the clearest example, but more generally the relative robustness of expenditure programs to cyclical fluctuations serves to smooth economic activity and this smoothing effect is likely to increase with the size of government. However, since distorting taxes and benefits have a pervasive impact on potential growth, a trade-off between stabilization and efficiency seems to arise within the standard AD-AS framework. If there is a positive relationship between the size of automatic stabilizers and distortive taxation, any tax reform aiming at lowering distortions and enhancing efficiency will come at the expense of macroeconomic stability. This issue is at the heart of macroeconomic policy design in EMU. If, as suggested by the standard model, there were a trade-off between stability and flexibility EMU members—having given up national monetary independence—would not dispose of enough policy instruments to deal with idiosyncratic shocks. However, this chapter suggests that, in the event of supply shocks, such a tradeoff might not exist. Within our model, under the assumption of at least partial wage resistance, cutting tax rates reduces market distortions and enhances the output stabilization in the event of supply shocks, and inflation stabilization in the event of demand shocks. So, if our conclusions are right, unless there is a clear predominance of demand over supply shocks, one should not worry about the possible adverse effects on stabilization of the tax reforms that across the EU are lowering marginal and average tax rates across the whole income scale (European Commission 2000a,b, 2001). It is understood that the analysis in this chapter is only a first step into the analysis of the relations between efficiency and flexibility on the one hand, and cyclical stabilization, on the other hand. Obvious improvements concern the theoretical model (which is overly simple and static in nature) and the description of the behavior of policy-makers. What is the impact of tax and welfare reforms on fiscal stabilizers? 265 Moreover, the numerical simulations are only indicative and should be supplemented by more thorough econometric investigation. An issue that arises naturally is the apparent contradiction between our conclusion that adverse stabilization effects may arise at lower levels of taxation in smaller economies, and the finding that small, open economies tend to have larger governments (see the seminal contribution by Rodrick (1998), and, recently Martinez-Mongay (2002)). Two explanations can be offered. First, whatever their initial level, higher taxes are output stabilizing in the event of demand shocks. Hence, if output stabilization is the main goal of fiscal authorities, and demand shocks (are expected to) prevail, larger governments would ensue. However, EMU may bring a change in the composition of shocks by increasing the relative frequency of supply compared to demand shocks.15 If so, large automatic stabilizers may no longer be optimal. Second, to the extent the tax burden remains below the critical tax burden, a rise in it is stabilizing, although increasingly less so. This, coupled with a higher exposure to shocks, may imply larger governments in small open economies. Econometric analyses based on past data may capture this effect. However, in recent years, the actual tax burden may have reached or even exceeded the critical one. Fresh empirical evidence tends to lend support to our results. Our analysis indicates that tax reforms aiming at lowering marginal effective tax rates and the tax burden, under supply shocks may enhance the stabilization properties of automatic stabilizers, especially in small Euro-area economies. Hence, they face a lesser dilemma between structural reform and stabilization policy This may contribute to explain their greater reform efforts and better performance compared with the big “laggards.” However, if EMU brings about greater trade integration, the incentives to step up reform efforts would increase also in the large Euro-area countries. Acknowledgments This chapter was prepared as a paper for the New School conference on “Monetary Policy and the Labor Market,” New York, November 2002 and has been presented at workshops organized by the Banca d’Italia and the European University Institute. We would like to thank Torben Andersen, Giuseppe Garone, José Herce, Peter MacAdam, Sandro Momigliano, Pedro Neves, Peter Part, and Karl Pichelmann, as well as participants to the just mentioned events for useful comments. The usual caveats apply. The opinions expressed in this chapter are the authors’ only and should not be attributed to the institutions they are affiliated with. Monetary policy and unemployment 266 Annexure What is the impact of tax and welfare reforms on fiscal stabilizers? 267 <•l 0.100 0.000 0.500 0/.00 0. 3<)0 0.200 0.100 0.000 -0. 100 ~ O . I OS -0 .1 10 -0. 115 - 0.120 - 0 .1 2$ - 0.1 30 -0 .1SS -0.1¢0 -0 .1<5 (b) 0.700 MOO 0.500 0 . 400 0.3'>0 0200 Demand Shod~ : OU I I)UI ...... _ - ~ 0 .0 0.1 0.2 0.3 o,.: 0.5 0 .6 0.7 0.8 0.9 1.0 Sl4)ply &hoek: OU!pu't .0 0.1 02 0.3 0.4 0.5 0.6 0 .7 0.8 0.9 1 .0 ~~ ' ·· .. ··. 0.220 0210 0200 0 . 190 0.180 0.170 0.1$0 0. 1 50 o. eoo 0.500 OAOO 0.300 0200 0 ,1 00 0000 0220 02 10 0.200 0 .190 0 .180 0 . 170 Oem&n<l Shoclc i"'llatiOn / .. ··· ........__~ 0.0 0.1 0.2 0 .3 0..4 0.5 0 .6 0.7 0.8 0.9 1.0 SoPP"; &hoeie In!!;) :iO n ~ .- - 0.0 0.1 o.2 0.3 0,4 0.5 Q.$0.7 0.$ 0.9 1.0 0 .100 0 .1 60 0.000 0.150 ~---~~~---~ 0.0 0.1 02 0.3 0.4 0.5 0.5 0.7 0.8 0.0 1.0 0.00 .1 0 .2 0..30.4 0.$0 .G0.1 0 . 80 . 91 .0 -0 .1 00 SloiWIY Shodc output St.J;~ply $1'1«k: ltl!l.":ion 0 . 700 -O .I OS 0.0 0.1 0 .2 0.3 OA 0.5 0.60.7 0.8 0.1) 1.0 o.GOO o . ~ 0.400 0.300 0.200 -0.110 ..().115 ..(),1 20 ..0.125 -0. 130 -0 .1 35 -0 ,140 -0 . 145 . .. ... .. .. .. .. . ·· ··· ···· .. ···· .. .. 0. 100 0.000 1----------~ 0.00 .1 020.30 .• O.SO .G0. 70.80.91.0 Monetary policy and unemployment 268 1<1 0700 0.600 0.500 0.400 0.300 0. 200 ·· .. ~ 0.100 ·~~~~~,-~~~~~ 0.00 .1 020.30 .4 0.50.G07 OIO..t1.0 Sc.WJ ~ ClfA:M ..0.100 ~--=::::;;..::.;;::;.;;;::,. ...,.....,..~ .. o.1 os 0.00 .1 020.30A0.$0S0.7010.11.0 ..0. 110 ..0.115 ..0.120 ..0.125 ..0.130 -C).I3S -0.1-40 -C),ICS (d) 0.700 0.600 0.500 0.400 0.300 . ···· .... . .............. ... ... ··········· ... ~ ···~ 0220 0.210 0200 0.190 0. 180 0.1 70 0. 1 60 _ .. ·· 0'1 so~. - ~~~- . - . ~, o - 2~0--.30~.c-.-o.s..,..o . ~ • .-0~7.-o ~ .e.-o.o ~, ~~~ 0.100 0600 0.500 0400 0. 300 0200 -~- 0.100 oooo~~~,- ~~~~~ 02.00 0230 0.220 0.2 10 0200 O. l;G 0.180 0.170 0.00.1 0.2 0.3 0.4 0.$0.80.70.10.$1.0 ..... ·· .. . .................. .. ·· 0.100 0. 160 0. 000 . . ... +-~~~~ ~~...-...-.-.~ 0.0 0.1 0.2 0.3 0.4 O.S OC 0 7 0.8 O.t 1.0 0.00.1 0.2 0.30' 0 SOGO 7 0.8091.0 &.pp~y..-- ----- -G.100 o. eoo ..O . IIO 0. 00 .1 0.20 . 30 . 4050G07010.t 1.0 0.$00 -0.120 -o130 .. 0.140 ..0.150 .. o. 1eo ..0.170 .. . ...... ....... .. . ... ·· ········· ··· . .. . ..... 0.300 0200 0. 100 0.000 +--~ ~~~.-..~,-~ o.oo .1 0.20.30.4 o .s o.a 0.1 o.t o. o1 .o Sensitivity analysis: (a) ε=1; (b) β=1; (c) =0.75; (d) =0.75; (e) γ=0.5; (f) β=1, λ=0.5. What is the impact of tax and welfare reforms on fiscal stabilizers? 269 (e) 0.7 00 0.600 0. 500 OAOO 0.300 0. 200 0.1 00 Ocmond $1'10Cic output Su.P9i'J shock: cut>t;e -o .• oo r..- ........ ~'""":''""":':-:"':-:""- _0 _110 . . 0 0 .1 0.2 0.30.<. o. s 0.60 .7 0.8 0.9 1.0 -0.120 ..0.130 -0.140 -0 .150 ..0.160 -0.170 -0.180 - 0. 190 (Q 0. 700 0. 600 0.500 O AOO 0. 300 0.200 0.1 00 ·· · ········· · ···· ·· · ··· ·· ··· ···· .. ... ~ 0.290 0.270 0. 2!0 0.230 0. 210 0. 190 0.170 0.600 0. 700 0. 600 0. 500 &<00 0. 300 0.200 ·· ···· ······· ·· ····· ········· .... Supply t hO<: Ic in ll al:on .· .... ~ ~· 0.100 o. ooo +-,......,--,-""T'".,....,......,...,.....,...,.... o.m 0..210 · 0..200 · 0. 1 90 · 0. 1 80 0.170 0. 1 60 0.00 .1 020 . 30.4 O..S0.60.70B0 .9 1.0 . .. ··· 0. 000 · .-.-, 0. 150 ·· 1- ~~~~~~~~~ ~ M& 1 ~UUU~U~~ ~ · M&1~UUUMU~~~· Sup::ty AAoeJc ou~ n s ~, Shock; : ll"'f lat~e~n -0 .100 0.600 .0 0.1 0.2 0.30.4 0.5 O.G 0.7 0.8 0.9 1.0 ..0.110 o.soo -0.120 0.400 -o.t30 ... ... . .. . ......... . -0.140 -0.1 50 " · 0.300 0.200 0.100 -0. 160 0. 000 -!- ....... --.-~~.,.........,....,..._,_, -0.170 0.0 0.1 0.2 0.3 0.4 0. 50.6 0.7 0.80 .0 1.0 </> ., rational expectations theory 238 real business cycle theory 107, 134 real wage resistance 253, 256, 262 replacement rates 36–7, 59–60 Reserve Bank of New Zealand 207 Rodrick, D. 265 Rogoff, K. 133, 145 Rotemberg J.J. 239 Rudebusch, G.D. 194 Sachs, J. 27, 29 Semmler, W. 182, 197; co-author of Chapter 10 and Editor share prices 108–9, 121–3, 125, 126, 127, 129 Shimizu, T. 2–4, 20; co-author of Chapter 9 shocks, economic 25–53, 57–60, 64–8, 80–7, 93–8, 128, 134, 189–95, 212, 252–65 Snower, D.J. 58 Solow, R.M. 157 Stability and Growth Pact 13, 210, 257 stabilization policy 64–7, 149, 166, 188, 194–7, 218, 232–3, 252, 257–65; see also automatic stabilizers, Stability and Growth Pact stagnation 20, 133, 135, 136, 137, 138, 140, 158 Stein, H. 232 stochastic macro-equilibrium 134 structuralist view 130 supply-side theory 107, 208, 253, 258–61 Svensson, L.E. 73, 76, 194 taxation, critical level of 259–61, 265 taxation policy 209–12, 252–3, 264 tax wedge 35–6 Taylor, J.B. 14–15, 218–19 Taylor-Calvo equations 9 Taylor rule 4, 65–7, 74–81, 86, 96–8, 119–21, 167, 169, 181, 188–97, 219–24, 229–32, 257 ‘time inconsistency’ problem 208–10 Tobin, J. 3, 67, 134, 158, 167, 178, 197, 218–19, 232–3 Tobin’s Q 16 total factor productivity (TFP) 27–33, 47–9, 59, 157–8 Treasury, the 211 uncertainty modelling of 145–53, 158 unemployment trap 65, 97–8 unit-root process 64–5 Van den Noord, P. 4; co-author of Chapter 14 vector autoregressions (VARs) 9, 137, 141, 197 Volcker, P. 10, 238, 243 Index 276 wage-setting, modelling of 253–6 wage stickiness 96 Wald test 239 Wharton Econometrics 229 Wicksell, K. 110, 120 Wolfers, J. 3, 57, 59–62, 69; co-author of Chapter 5 Woodford, M. 239, 241 World Bank 109 Yamaichi Security 137, 143 Yoshikawa, H. 2–4, 143, 158; author of Chapter 4 and co-author of Chapter 9 Zoega, G. 4, 18–19, 124; co-author of Chapter 8 Index 277