Dissertation

Dissertation

Fiscal multipliers, unemployment and debt: Essays on the macroeconomics of fiscal policy coordination in Europe

Author:

Philipp Heimberger

Student-ID:

0708430

E-mail:

[email protected]

Doctoral thesis submitted to the Department of Economics of the Vienna University of Economics and Business in the program Doktorat der Sozial- und Wirtschaftswissenschaften Vienna, March 2018.

1

Declaration I certify that the thesis submitted for the doctoral degree at the Vienna University of Economics and Business is composed by my own work. The core of this cumulative dissertation consists of three papers, all of which have been published in peer-reviewed journals. 1. Heimberger, Philipp (2017): Did fiscal consolidation cause the double-dip recession in the euro area? Review of Keynesian Economics, 5(3), 439-458. 2. Heimberger, Philipp; Kapeller, Jakob (2017): The performativity of potential output: Pro-cyclicality and path dependency in coordinating European fiscal policies, Review of International Political Economy, 24(5), 904-928. 3. Heimberger, Philipp; Kapeller, Jakob; Schütz, Bernhard (2017): The NAIRU determinants: what’s structural about unemployment in Europe? Journal of Policy Modeling, 39(5), 883-908. From this list, it can be seen that the second paper was joint-work with Jakob Kapeller; I contributed approximately 70% of the work. The third essay was jointly co-authored with Jakob Kapeller and Bernhard Schütz; I contributed about 80% of the work.

22

Acknowledgment I thank Wilfried Altzinger for supporting my dissertation project. However, I am even more indebted to Jakob Kapeller. Jakob brought me to the Johannes Kepler University in Linz, where I conducted most of the research on which this dissertation is based, and he was a constant source of inspiration on how to be a good researcher. Colleagues at the Vienna Institute for International Economic Studies have also provided valuable feedback. Most of all, I thank my family whose incredible support throughout my life has enabled me to do what I love doing: policy-relevant research on our economies and societies – geared towards improving the life of the many. Vienna, March 2018

33

Contents Declaration

2

Acknowledgment

3

Abstract

5

Introduction to the dissertation 8 1. Dissertation in summary 10 2. Did fiscal consolidation cause the double-dip recession in the euro area? 11 3. The performativity of potential output: Pro-cyclicality and path dependency in coordinating European fiscal policies 13 4. The NAIRU determinants: What’s structural about unemployment in Europe? 17 References 20 Did fiscal consolidation cause the double-dip recession in the euro area? 1. Introduction 2. Which factors determine the size of fiscal multipliers? 2.1 Multipliers in crisis times 2.2 Estimates on the size of GDP losses due to fiscal consolidation in the euro area 3. Econometric strategy 4. Baseline results and discussion 5. Robustness checks 5.1 The role of outliers 5.2 Variations in the country group 5.3 Including additional control variables 6. Conclusions References Appendix 1

23 23 24 25 26 27 28 33 33 33 36 38 39 42

The performativity of potential output: Pro-cyclicality and path dependency in coordinating European fiscal policies

43 43 45 48 50 50 51

1. Introduction 2. A tool and its context: income inequality, debt, current account imbalances and the Eurozone crisis 3. The European Commission’s potential output model and its use in European fiscal policy-making 4. The performativity of the potential output model and pro-cyclical feedback loops in Europe: an overview 4.1 The performative impact of the potential output model 4.2 How the potential output model influences actual economic outcomes 5. The pro-cyclicality of NAIRU and potential output estimates: impacts on macroeconomic developments and fiscal policy-making 5.1 Pre-crisis years in the euro area: the ‘optimist loop’ 5.2 Post-crisis years: the ‘pessimist loop’ 6. Model performativity and debt trajectories in Europe: the self-defeating nature of the Stability and Growth Pact 7. Conclusions References

53 53 55 58 62 64

The NAIRU determinants: What’s structural about unemployment in Europe? 1. Introduction 2. The European Commission’s NAIRU approach: estimation and application 3. The determinants of (structural) unemployment in European countries: literature review 4. Basic economic strategy and data 5. Econometric baseline results 6. Robustness checks 7. Discussion: the NAIRU in theory, empirics and policy 8. Conclusions References

68 69 70 74 78 81 85 89 91 91

Conclusions of the dissertation 1. The context-sensitivity of ‘the’ fiscal multiplier and research limitations 2. The EU’s fiscal regulation framework and the pro-cyclicality of European fiscal policies 3. Conclusions related to institutional reform discussions in Europe 4. Future research on European fiscal policies References

94 94 95 97 99 100

44

Abstract (English) This cumulative dissertation consists of three essays, which all deal with issues relevant to fiscal policy coordination in Europe. The three essays provide answers to a set of interrelated research questions. What are the underlying estimation problems when it comes to evaluating the fiscal effort1 of individual Eurozone member countries? How do the problems that arise in the process of estimating the fiscal effort of individual countries influence fiscal policy coordination across the euro area? How has fiscal policy in the euro area affected macroeconomic developments with a particular focus on the empirical evidence since the start of the Eurozone Crisis? The first essay analyzes the short-run impact of fiscal consolidation measures on economic activity in the euro area during the Euro Crisis. It presents new econometric estimates on the link between cumulative GDP growth and fiscal consolidation measures during 2011-2013. The main empirical finding is that the depth of the economic crisis in the euro area's economies is closely related to the harshness of fiscal consolidation. Cumulative multiplier estimates are found to vary in a range from 1.4 to 2.1, depending on the data source used to identify the intensity of fiscal consolidation. Given these multiplier values, a reasonable approximation of the size of the output losses due to fiscal consolidation in the euro area during 2011-2013 is in the range of 5.5% to 8.4% of GDP. Against the background of the prevailing macroeconomic and institutional circumstances, fiscal consolidation is argued to be the cause of the double dip recession. The second paper investigates the performative impact of the European Commission’s model for estimating ‘potential output’, which is used as a yardstick for measuring the ‘structural budget balance’ of EU countries and, hence, is crucial for coordinating European fiscal policies. In pre-crisis years, potential output estimates promoted the build-up of private debt, housing bubbles and macroeconomic imbalances. After the financial crisis, these model estimates were revised downwards, which increased fiscal consolidation pressures. By focusing on the euro area’s economies during 1999-2014, we show how the model’s estimates influence actual economic outcomes. We identify two major economic impacts of the potential output model. First, the political implications of the model led to pro-cyclical

1 The fiscal effort intends to measure the effect of government policy on the fiscal balance, i.e. it is an indicator for changes in the fiscal balance that are neither due to the effects of the business cycle on revenues and expenditures nor due to budgetary one-off effects.

55

feedback loops, reinforcing prevailing economic developments. Second, the model has contributed to national lock-ins on path dependent debt trajectories, fueling ‘structural polarization’ between core and periphery countries. The third essay analyzes the determinants of the European Commission’s estimates of the non-accelerating inflation rate of unemployment (NAIRU) for 14 European countries during 1985-2012. Our main finding is that the NAIRU is not a good proxy for ’structural unemployment’: Labor market institutions – employment protection legislation, union density, tax wedge, minimum wages – underperform in explaining the NAIRU, while cyclical variables – capital accumulation and boom-bust patterns in housing markets – play an important role. This is policy-relevant since the NAIRU is used to compute potential output and structural budget balances and, hence, has a direct impact on scope and evaluation of fiscal policies in Europe. Keywords: Fiscal policy, Eurozone, public debt, unemployment, NAIRU, potential output. JEL-Codes: C54, E24, E61, E62, E63. Abstract (Deutsch) Diese kumulative Dissertation besteht aus drei Essays, die alle auf Themen fokussieren, welche für die fiskalpolitische Koordinierung in Europa von besonderer Relevanz sind. Die drei Essays liefern Antworten auf ein Bündel von zusammenhängenden Forschungsfragen. Was sind die Schätzprobleme in Bezug auf die Evaluierung fiskalischer Anstrengungen 2 einzelner Euro-Mitgliedsländer? Wie beeinflussen jene Probleme, die bei der Schätzung der fiskalischen Anstrengung einzelner Länder auftreten, die fiskalpolitische Koordinierung in der Eurozone? Was waren die Effekte der Fiskalpolitik auf die makroökonomischen Entwicklungen in der Eurozone mit besonderem Fokus auf die empirische Evidenz nach dem Ausbruch der Eurozonenkrise? Der

erste

Essay

analysiert

die

kurzfristigen

Effekte

der

fiskalischen

Konsolidierungsmaßnahmen auf die ökonomische Aktivität in der Eurozone während der Eurozonenkrise. Dabei werden neue ökonomische Schätzungen zum Zusammenhang 2 Die fiskalische Anstrengung soll den Effekt der Politik einer Regierung auf den fiskalischen Saldo messen; es geht also darum, einen Indikator für Veränderungen im fiskalischen Saldo zu ermitteln, der sowohl für die Effekte konjunktureller Schwankungen auf staatliche Einnahmen und Ausgaben als auch für budgetäre Einmaleffekte korrigiert ist.

66

zwischen dem kumulativen BIP-Wachstum und den fiskalischen Konsolidierungsmaßnahmen im Zeitfenster der Jahre 2011-2013 präsentiert. Das zentrale empirische Ergebnis lautet, dass die Tiefe der ökonomischen Krise in den Eurozonenländern in engem Zusammenhang mit der Schärfe der fiskalischen Konsolidierung steht. Die Schätzergebnisse zu den kumulativen durchschnittlichen Fiskalmultiplikatoren in der Eurozone liegen in einer Bandbreite von 1,4 bis 2,1 – in Abhängigkeit von der Datenquelle, die herangezogen wird, um die Intensität der fiskalischen Konsolidierung zu messen. Auf der Basis dieser Multiplikatorschätzungen liegt eine näherungsweise Schätzung für die Höhe der Outputverluste, die aus der fiskalischen Konsolidierung in der Eurozone im Zeitfenster 2011-2013 resultierten, in einer Bandbreite von 5,5% bis 8,4% des BIP. Vor dem Hintergrund der vorherrschenden makroökonomischen und

institutionellen

Rahmenbedingungen

kann

die

fiskalische

Konsolidierung

vertretbarerweise als Ursache der Double-Dip-Rezession identifiziert werden. Das zweite Papier untersucht den performativen Einfluss des Modells der Europäischen Kommission für Schätzungen des „Potenzialoutput“, die als Maßstab für die Messung des „strukturellen Budgetsaldos“ der EU-Länder dienen, und deshalb von zentraler Bedeutung für die Koordinierung der europäischen Fiskalpolitik sind. In den Vorkrisenjahren beförderten die Potenzialoutput-Schätzungen den Anstieg der Privatverschuldung sowie das Entstehen von Vermögenspreisblasen und makroökonomischer Ungleichgewichte. Nach dem Ausbruch der Finanzkrise revidierte die Europäische Kommission die Modellschätzungen drastisch nach unten, was den fiskalischen Konsolidierungsdruck erhöhte. Indem der Essay auf die Erfahrungen der Eurozonenländer im Zeitraum 1999-2014 fokussiert, lässt sich zeigen, dass die Modellschätzungen die tatsächlichen makroökonomischen Outcomes beeinflussen. Wir identifizieren zwei zentrale Auswirkungen des Potenzialoutput-Modells. Erstens führten die politischen Implikationen des Modells zu einer prozyklischen Feedbackschleife, welche die vorherrschenden ökonomischen Entwicklungen verstärkte. Zweitens hat das Modell zu nationalen Lock-ins in der pfadabhängigen Schuldenentwicklung beigetragen, was einer „strukturellen Polarisierung“ zwischen Kern- und Peripherieländern Vorschub leistet. Der dritte Essay analysiert die Determinanten der von der Europäischen Kommission produzierten Modellschätzungen zur non-accelerating inflation rate of unemployment (NAIRU) für 14 europäische Länder im Zeitraum 1985-2012. Das zentrale Ergebnise lautet, dass die NAIRU keine gute Näherungsgröße für die „strukturelle Arbeitslosigkeit“ darstellt: Arbeitsmarktinstitutionen

(z.B.

Arbeitsschutzbestimmungen,

gewerkschaftlicher 77

Organisationsgrad, Lohnkeil, Mindestlöhne – sind großteils keine statistisch signifikanten Determinanten der NAIRU, während zyklische Variablen – Kapitalakkumulation und BoomBust-Muster auf Häusermärkten – eine wichtige Rolle spielen. Dieses Ergebnis ist politikrelevant, weil die NAIRU-Schätzungen verwendet werden, um den Potentialoutput und strukturelle Budgetsalden zu messen, und dementsprechend einen direkten Einfluss auf die Evaluierung der Fiskalpolitiken in Europa haben. Schlüsselwörter:

Fiskalpolitik,

Eurozone,

Staatsschulden,

Arbeitslosigkeit,

NAIRU,

Potenzialoutput. JEL-Codes: C54, E24, E61, E62, E63.

Introduction to the dissertation The macroeconomic consequences of the crisis that started in 2007/2008 have revitalized the interest of macroeconomists in conducting research on how fiscal policy affects macroeconomic outcomes (e.g. Parker 2011; Ramey 2011; DeLong and Summers 2012; Blanchard and Leigh 2013; Gechert 2015; Leeper et al. 2017) – an interest that had largely vanished in the decades before the financial crisis as fiscal policy was “a backwater compared to research on monetary policy” (Ramey 2011, p. 673). The overriding goal of this dissertation is to contribute to the literature on the macroeconomic effects of fiscal policy by focusing on Europe. The dissertation provides answers to three interrelated research questions. First, what are the underlying estimation problems when it comes to evaluating the fiscal effort of euro area countries, where the fiscal effort intends to measure the effect of government policies on the fiscal balance? Second, how do the problems that characterize the estimation of the fiscal effort in individual member countries influence fiscal policy coordination across the euro area? Third, how has fiscal policy in the euro area affected macroeconomic developments with a particular focus on the empirical evidence since the start of the Eurozone Crisis? The Eurozone’s special institutional architecture is a proper starting point for the analysis of the sources of macroeconomic fragility and policy coordination problems: while the European Central Bank conducts monetary policy for the whole euro area, the existing institutions lack 88

a common fiscal and political union, which stands in contrast to the arrangements in monetary unions such as the United Kingdom and the US (e.g. Wyplosz 2006; Eichengreen 2008; Mody 2015; Farhi and Werning 2017; Martin and Phillipon 2017). With the introduction of the Euro, member states of the European Monetary Union basically gave up control over their national monetary policy, thereby abandoning a vital national instrument for macroeconomic stabilization. As the European Central Bank (ECB) seized control over monetary policy in the whole euro area and oriented its interest rates decisions towards the average of the euro area, competitive devaluations were no longer possible for individual euro area countries (e.g. Arestis and Sawyer 2011; Shambaugh 2012). However, while the euro area comes with genuine supranational capacities in the area of monetary policy, this is not the case for fiscal policy: a euro area finance ministry does not exist; there is no euro area budget from which fiscal transfers between countries and regions could be channeled for macroeconomic stabilization purposes (e.g. Lane 2012; Juncker et al. 2015); and as the issuance of government bonds is not pooled across euro area countries, single countries de facto issue debt in a currency over which they have no control (De Grauwe 2012). It has been shown that this special institutional setup makes euro area countries vulnerable to abrupt changes in financial market sentiments, which serves to increase the fragility of the euro area as a whole; this characteristic became obvious as (during earlier stages of the Eurozone crisis) the consequences of the ECB’s inability to play the role of credible lender of last resort contributed to contagion in government bond markets (Saka et al. 2015; Iversen et al. 2016). The ‘Great Recession’ has shown that none of the large currency areas (the US, the UK and the Eurozone) was able to stabilize its economy by monetary policy measures alone (e.g. Khatiwada 2009; Cottarelli et al. 2014; Blinder and Zandi 2015). Central banks that are fundamentally constrained by the zero lower bound of nominal interest rates (which has arguably been the case for the ECB since 2012; see, e.g., Coeure 2012) can prove unable to reach their price stability target as long as fiscal policy does not play along in supporting economic recovery in times of economic slack. Both theoretical and empirical research findings of recent years suggest that fiscal policy can have large effects on economic activity when the central bank is constrained – due to the zero lower bound – in its ability to stimulate economic activity by ‘conventional monetary policy’ (e.g. Christiano et al. 2011; Woodford 2011; Eggertsson and Krugman 2012; Gechert and Rannenberg 2014; Farhi and Werning 2016; Wren-Lewis 2016). In particular, the experiences in the euro area since 2010 have shown that countries with more room for fiscal policy maneuvering (due to running low fiscal 99

deficits or even fiscal surpluses and a stable or declining ratio of public-debt-to-GDP) might be unwilling to use the existing space for expansionary fiscal policies, while countries with less fiscal space (due to substantial fiscal deficits and/or a large burden of public debt) would want to employ additional expansionary fiscal efforts to overcome their macroeconomic troubles despite being constrained by the fiscal regulation framework in place. From a policymaking perspective, this situation may be seen as problematic since the prevailing rules and instruments in the EU’s fiscal regulation framework simply do not allow for directly managing the aggregate fiscal policy stance within the monetary union, which may lead to a counterproductive policy mix across countries and suboptimal macroeconomic outcomes (European Commission 2016). This dissertation contributes to the academic literature on the macroeconomics of fiscal policy. It does so by developing new ways of thinking about the sources of Europe’s fiscal policy coordination problems in a regulatory framework that heavily relies on the European Commission’s model-based estimates of the ‘structural component’ of headline fiscal deficits, in which the underlying estimations of the ‘structural deficit’ are used to evaluate fiscal efforts in member countries. Against the background of the prevailing institutional and macroeconomic circumstances in the euro area, the dissertation also analyzes the impact of fiscal policies on macroeconomic developments with a particular focus on the years of the Eurozone crisis from 2010 onwards.

1. Dissertation in summary This cumulative dissertation contributes to the academic study on the macroeconomics of fiscal policy coordination in Europe. In what follows, I summarize the three core chapters of this dissertation, which are based on peer-reviewed journal publications. For each of the chapters, I will provide a short general introduction. I will then discuss the main research questions, the theoretical background and the main concepts under study, before focusing on important elements of the empirical research strategy. Finally, I will sketch out the results as well as the main contributions of the respective paper to the existing literature.

1010

2. Did fiscal consolidation cause the double-dip recession in the euro area? As a consequence of the crisis that started in 2007/2008, government debt levels in European economies rose rapidly. Sovereigns absorbed private-sector losses, and cyclical fiscal deficits increased steeply, as financial distress led to deep recessions, which caused declines in tax revenues and increases in unemployment benefits. The accumulation of sovereign debt since the start of the crisis was mainly a consequence of automatic fiscal stabilizers, government support to the financial sector and fiscal stimulus measures designed to counteract the economic downturn (e.g. Lane 2012; Shambaugh 2012). The spectre of economic catastrophe had triggered a temporary consensus for internationally coordinated stimulus in 2008/2009. A policy switch towards fiscal consolidation followed from 2010 onwards (e.g. Blyth 2013; Farrell and Quiggin 2017; Skidelsky and Fraccaroli 2017). Key policy institutions have justified this move by a need to immediately bring down public sector debt and fiscal deficits (e.g. ECB 2010; IMF 2010; European Commission 2010). What was the short-run impact of fiscal consolidation on economic activity in the euro area during the Euro Crisis with a particular focus on the years 2011-2013? And how did the macroeconomic and institutional circumstances shape the growth effects of fiscal consolidation in the euro area? Chapter 2 of my dissertation provides answers to those research questions by proposing a novel econometric estimation strategy. The central theoretical concept that underlies the research approach is the fiscal multiplier. The fiscal multiplier concept describes a cause-effect relationship between an exogenous change in the fiscal deficit, due to discretionary fiscal measures (cause), and a change in economic activity (effect). In fiscal multiplier theory, the basic assumption is that discretionary fiscal measures have direct effects on aggregate demand, called first-round effects. For example, a cut in government spending will lead to a direct fall in aggregate demand by causing a decrease in the overall level of spending in the economy. However, these first-round effects will also have an impact on demand in the private sector, on demand in the foreign sector and on supply decisions by households and companies, leading to second-round effects and higher-round effects, which are the result of reactions of single agents in the economy to discretionary fiscal action (e.g. Buiter 1977; Gechert 2014). All attempts to estimate the effects of fiscal (or monetary) policy on economic activity (in most studies measured in terms of real GDP) are characterized by the problem that real-world 1111

settings in macroeconomics generally do not resemble a randomized controlled experiment, in which a set of countries with very similar characteristics (in terms of institutions) would have to be randomly assigned to two groups: while the treatment group is treated with fiscal consolidation, the control group does not receive any treatment, holding all other factors between the two groups constant (e.g. Angrist and Pischke 2009; Jorda and Taylor 2016). Obviously, the real world generally looks strikingly different: countries may have quite different sets of institutions; furthermore, governments’ primary motivation to implement certain fiscal policy measures has often been to respond to macroeconomic developments (such as a downswing in the economy; e.g. Guajardo et al. 2014). In general, it is also difficult to correctly identify exogenous variation in fiscal policy (e.g. Nakamura and Steinsson 2017). Although actual economic developments in Europe do not resemble the ideal of a randomized controlled experiment, the time period of the Eurozone Crisis may be seen as a natural experiment,3 which allows for using a cross-sectional identification strategy. During the time period 2011-2013, virtually all Eurozone countries (which share important common institutions; e.g. De Grauwe 2012) implemented fiscal consolidation measures motivated by a desire to cut fiscal deficits – with marked variation in the size of these measures. Inspired by earlier contributions of Blanchard and Leigh (2013) as well as De Grauwe and Ji (2013), chapter 2 exploits the exogenous variation in the intensity of fiscal consolidation across the euro area’s economies to obtain econometric estimates on the average size of fiscal multipliers in the euro area. The empirical research strategy is guided by theoretical reasoning about the size of fiscal multiplier, as the existing academic literature suggests that numerous factors have an impact on the size of fiscal multipliers: monetary policy accommodation, the composition of fiscal consolidation (spending-based vs. tax-based), the initial level of public indebtedness, the exchange-rate regime, the openness of the economy, spillover effects with other economies, and the international business environment (e.g. Ramey 2011; Gechert and Rannenberg 2014). I use accepted tools from the empirical literature to identify exogenous vs. 3 Krugman (2013), among others, explains why the Eurozone Crisis may be seen as a natural experiment: “Really big changes in fiscal stance come rarely, and are usually associated with wars, when other things like rationing also tend to happen. And the coincidence of big fiscal shifts with constrained monetary policy is a once-in-three-generations story. But Europe after 2009 provided something that, while not a perfect natural experiment, was much closer than anything we’re likely to see for a long time. Austerity mania, enforced on countries that had no freedom of maneuver because they were on the euro, led to drastic fiscal tightening in some but not all euro area economies; these big shifts gave us a pretty good view, certainly by historical standards, of what such shifts do.“ De Grauwe and Ji (2013) as well as Blanchard and Leigh (2013) discuss that many Eurozone countries started with their fiscal consolidation efforts at a very similar point in time, although the harshness of the consolidation treatment varied considerably across countries.

1212

endogenous changes in fiscal policy, while I also discuss the problem of choosing the right counterfactual. The main contribution to the existing literature is that I present new econometric estimates on the link between cumulative real GDP growth and fiscal consolidation measures in the euro area. First, I look at a variety of data sources to measure fiscal consolidation in the crosssectional dimension – specifically, I calculate changes in the structural budget balance, an approach that Blanchard and Leigh (2013) have proposed in a seminal paper; but I also use data sources that follow the ‘narrative approach’ by identifying size and timing of fiscal consolidation measures from budgets and policy documents in the spirit of Romer and Romer (2010). I compile a data set for fiscal consolidation measures after the financial crisis for more than 30 countries, which includes all Eurozone countries. In total, I use six different data sources to identify size and timing of fiscal consolidation measures. By doing so, I am able to estimate a range of plausible cumulative fiscal multipliers for the euro area over the time period studied and to check whether results are broadly consistent across different approaches to measuring fiscal consolidation. Second, my econometric baseline estimates are used to assess cumulative GDP losses in the euro area during 2011-2013, which I discuss against the background of existing estimates in the literature. Third, I focus on providing an integrated discussion on the role of the institutional and macroeconomic circumstances in the euro area with regard to the determinants of the size of fiscal multipliers. What are the main findings in chapter 2 of this dissertation? The empirical evidence suggests that the depth of the economic crisis in the euro area during 2011-2013 was closely associated with the harshness of fiscal consolidation. Cumulative average multiplier estimates range from 1.4 to 2.1 – depending on which data source one uses to identify exogenous variation regarding the intensity of fiscal consolidation measures. My econometric findings are consistent with reviewing the fiscal multiplier literature, as key conditions for fiscal multipliers higher than 1.0 were arguably fulfilled in large parts of the euro area.

3. The performativity of potential output: Pro-cyclicality and path dependency in coordinating European fiscal policies While chapter 2 presents an econometric study dedicated to the first main topic of this dissertation – analyzing the effects of fiscal consolidation measures on economic activity in 1313

the euro area –, chapter 3 is concerned with the second main topic: the role of model-based assessments of the fiscal effort in individual member countries. The basic starting point is that the euro area lacks a fiscal and political union. While the instruments of fiscal policy are still in the hands of individual member states, the EU’s fiscal regulation framework restricts national fiscal policies with a set of rules, budget procedures and institutions that allow for surveillance and control of fiscal policies. The main fiscal objective of the regulation framework is to make sure that governments do not run large fiscal deficits that may undermine the sustainability of public debt (e.g. Claeys et al. 2016). In assessing the sustainability of public finances in EU member countries, the EU’s fiscal regulation framework relies heavily on measures of the so-called ‘structural budget balance’. Since its reforms in 2005 and 2011, the Stability and Growth Pact has made use of the structural balance as a central control indicator for medium-term fiscal conduct (ECFIN 2013). Furthermore, rules in the Fiscal Compact set deficit limits in terms of the structural budget balance (Fiscal Compact 2012). The size of the structural budget balance directly depends on estimations derived from the potential output model (e.g. Klär 2013; Tereanu et al. 2014). The Commission employs its potential output model for estimating the ‘output gap’ – the difference between actual GDP and potential output –, which is used as an indicator for the cyclical position of an economy. Output gap estimates then translate into the Commission’s judgments on how much of the fiscal deficit (or surplus) in a particular EU country is ‘structural’, i.e. not attributable to the cyclical ups and downs in the economy. If the structural deficit is estimated to be high, the fiscal scope of member countries shrinks, as they are obliged to implement fiscal consolidation measures in order to bring down ‘excessive deficits’. What are the estimation problems in evaluating fiscal efforts in euro area countries? And how do the problems that characterize the estimation of the fiscal effort influence policy coordination across the Eurozone? Chapter 3 of the dissertation – which was co-authored with Jakob Kapeller – addresses these research questions by developing new ways of thinking about the sources of Europe’s fiscal policy coordination problems. As our vantage point, we take the claim that economic models are performative as they “do not merely record a reality […] but contribute powerfully to shaping, simply by measuring, the reality” (Callon 1998, p. 23). A series of non-rival and complementary approaches to studying and analyzing the influence of economic models are available, where the performativity of economic models represents one main strand of research (Hirschman and Berman 2014). In this context, chapter 1414

3 aims at gaining a deeper understanding of “the role of economics in the cognitive infrastructure of policymaking” with particular focus on European fiscal policies by analyzing the institutionally established potential output model as an exemplary “economic policy device for seeing and deciding” (Hirschman and Berman 2014, p. 779). The ‘potential output’ concept is at the heart of the Commission’s approach to measuring the fiscal effort of the Eurozone’s member countries. Potential output is defined as the level of output in an economy at which all production factors are employed at ‘non-inflationary levels’. In this context, it is crucial to understand that one cannot observe how much an economy should be able to produce before inflation kicks in. In order to come up with empirical estimations of an unobservable theoretical concept, the Commission estimates a Cobb-Douglas production function, where potential output is modeled as a function of the production factors labor and capital, and total factor productivity serves as a proxy for technological progress (Havik et al. 2014). While the chosen production function is employed as a calculation vehicle for integrating empirical data on the respective variables, the essential underlying economic question – “Which components of GDP growth and changes in the unemployment rate are to be seen as ‘structural’ rather than cyclical?” are delegated to a statistical de-trending procedure. The decomposition of cyclical and structural components of the crucial variables in the model relies on a Kalman filter model. In chapter 3, we specifically focus on the Commission’s Kalman-filter estimation of the NAIRU – the unemployment rate at which (wage) inflation does not accelerate. The NAIRU directly affects the size of the ‘structural deficit’ within the EU’s fiscal regulation framework as an increase in the NAIRU causes a reduction in potential output leading to less fiscal policy scope (and vice versa for a decrease in the NAIRU). The intuition is that as NAIRU estimates rise, a correspondingly smaller part of the actual unemployment rate is seen to be of cyclical nature; hence, the creation of additional demand by expansionary fiscal policies would lead to an increase in inflation without reducing the ‘structural component’ in unemployment. While the alleged performativity of economic models (Callon 1998; MacKenzie 2006; Braun 2016) has been studied extensively in microeconomic contexts, especially in financial markets (e.g. Beunza and Stark 2004; Lockwood 2015), the scholarly literature has so far largely remained silent on the performative impact of macroeconomic models on overall macroeconomic performance (e.g. Braun 2014; Braun 2015). In chapter 3 of this dissertation, we contribute to closing this research gap. While a large literature deals with the problems of estimating the business cycle position of an economy by assessing the difference between 1515

actual GDP and model-based measures of potential output (e.g. Orphanides and van Norden 2002; Mishkin 2007; Summers 2016; Coibion et al. 2017), our paper is the first to directly analyze the impact of potential output estimates on fiscal policy practices and macroeconomic outcomes with a particular focus on Europe. Our research design integrates a diverse set of research methods, as we employ both ’conventional’ regression techniques (to investigate revisions in potential output and factors that are driving the Commission’s model estimates) and – rather new – complexity economics methods inspired by the approach developed in Cristelli et al. (2015).

Our main findings in chapter 3 are as follows. The Commission’s potential output model has produced pro-cyclical estimates in pre-crisis and post-crisis years. As these estimates influence assessments about whether Eurozone countries meet budgetary targets in the EU’s fiscal regulation framework, these pro-cyclical potential output estimates contributed to procyclical fiscal policies. During the boom, downward revisions in potential output gave more fiscal room to booming economies due to more favorable assessments of what’s ‘structural’ about the fiscal balance. We argue that pro-cyclical potential output estimates amplified the build-up of private debt, housing bubbles and macroeconomic imbalances in the run-up to the financial crisis. When the crisis hit and European economies entered the recession, potential output estimates were, however, revised downwards, where the countries most affected by the crisis suffered the largest downward revisions in potential output. Pro-cyclical revisions during the crisis increased fiscal consolidation pressures: as the Commission drastically revised the NAIRU upwards – implying assessments of high ‘structural unemployment’ –, the respective countries’ ‘excessive structural deficits’ were seen to be large. Because of the institutionalization of the structural deficit as a central control indicator in the EU’s fiscal regulation framework, governments were under increased pressure to curb the deficit by cutting government spending and raising taxes. Consolidation measures, in turn, served to aggravate the macroeconomic troubles in the crisis-stricken countries, because fiscal multipliers were substantial (see chapter 2 of this dissertation), which systematically subjected the Eurozone’s most fragile countries to a self-defeating cycle of fiscal consolidation measures and deteriorating macroeconomic conditions. Furthermore, the potential output model’s estimates have amplified the structural divergence between export-led creditorcountries (in the core of the Eurozone) and strongly indebted countries (in the periphery) by providing political and fiscal leeway to those already successful, while delegitimizing already stressed periphery countries via model-induced deteriorations in NAIRU, potential output and 1616

structural deficits.

4. The NAIRU determinants: What’s structural about unemployment in Europe? Chapter 4 of the dissertation (co-authored with Jakob Kapeller and Bernhard Schütz) provides further in-depth analysis of the pitfalls of coordinating European fiscal policies in a regulatory framework that heavily relies on the European Commission’s model-based estimates of the ‘structural component’ of headline fiscal deficits. Our starting point is the observation that actual unemployment rates across Europe decreased markedly during and in the aftermath of the global financial crisis. In particular, the Eurozone saw unemployment rise from 7.6% in 2008 to 12.0% in 2013, before unemployment fell slightly to 10.9% in 2015. The developments on European labor markets have led to growing research efforts geared towards estimating the determinants of the evolution of unemployment rates in European countries (e.g. Arpaia et al. 2014; ECB 2015). Does the European Commission produce reliable empirical estimates of the non-accelerating inflation rate of unemployment (NAIRU) that can be used as a proxy for ‘structural unemployment’, i.e. unemployment existing independently of all temporary and seasonal fluctuations? Which econometric determinants perform well in explaining variations in NAIRU estimates in a large panel of EU countries? Chapter 4 tackles these research questions by relating them to the central topic of this dissertation, namely fiscal policy coordination in Europe. The question how the NAIRU has evolved in European countries before and after the crisis is highly relevant, since policy-makers use them frequently to assess the ‘structural component’ of unemployment, which cannot be lowered by expansionary demand-side policies (e.g. Canton et al. 2014; Orlandi 2012). Furthermore, the NAIRU serves as a central theoretical concept in modern macroeconomics, which has over time been identified with Friedman’s (1968) idea of a ‘natural rate of unemployment’ (e.g. Ball and Mankiw 2002) that exists independently from cyclical and seasonal fluctuations in the economy. However, the NAIRU is essentially unobservable – chapter 3 of this dissertation also elaborates on this point. In many empirical applications, the NAIRU is therefore treated as an unobservable stochastic variable (e.g. Staiger et al. 1997; Laubach 2001) estimated by a variety of econometric models and statistical techniques, which are used to separate the cyclical component of the actual unemployment rate from its trend. In chapter 4 of this 1717

dissertation, we concern ourselves with the Commission’s estimation approach of the NAIRU. The Commission employs a multi-variate Kalman filter model (Planas and Rossi 2015), which does not explicitly specify the structural factors that underlie ‘structural unemployment’. We study whether theoretical arguments on the determinants of the NAIRU actually perform well in explaining the Commission’s empirical NAIRU estimates, which are based on statistical de-trending. By doing so, we assess the robustness and plausibility of commonly used NAIRU estimates. We show that this exercise is highly relevant for fiscal policy-making in Europe, since the Commission uses the NAIRU as a proxy for ‘structural unemployment’ in estimating potential output, which is then translated into judgments on the fiscal effort of individual member countries within the EU’s fiscal regulation framework (see also chapter 3). With high ‘structural unemployment’, the ‘structural deficit’ is estimated to be large (and vice versa). As a consequence, high NAIRU estimates put fiscal consolidation pressure on the respective government, because it is legally required to curb ‘excessive deficits’. Our analysis of the econometric determinants of the Commission’s NAIRU estimates for 14 European OECD countries over the time period 1985-2012 relates to a large literature on the determinants of unemployment (e.g. OECD 1994; Siebert 1997; IMF 2003; Nickell et al. 2005; Stockhammer and Klär 2011). The focus in the empirical panel data literature is to explain broad movements in unemployment across OECD countries by shifts in labor market institutions such as trade union density, employment protection legislation, unemployment benefit replacement rate, tax wedge, active labor market policies, minimum wages etc. As some studies had found no “meaningful relationship between [the] OECD measure of labor market deregulation and shifts in the NAIRU'' (Baker et al. 2005, p. 107), researchers began to include additional control variables representing alternative explanations for the evolution of (structural) unemployment. Blanchard and Wolfers (2000), for instance, control for 'macroeconomic shocks' such as changes in the long-term interest rate, deviations from the trend in total factor productivity growth and shifts in labor demand, as they emphasize the link between these macro shock variables and labor market institutions. Stockhammer and Klär (2011) regard investment as the most crucial variable in explaining unemployment; hence, they include measures of capital accumulation in their regressions. Bassanini and Duval (2006), among others, include a terms of trade shock variable in their regressions, since a change in the terms of trade is assumed to affect domestic unemployment. Finally, Orlandi

1818

(2012) introduces another essential control variable, as he considers a proxy for boom-bustpatterns in housing markets. Our paper contributes to the existing academic literature by analyzing the role of standard labor market variables in explaining the evolution of the European Commission’s NAIRU estimates, while also controlling for a comprehensive set of variables capturing alternative hypotheses with regard to the determinants of the NAIRU. Although it may seem to be an obvious strategy to econometrically compare actual NAIRU estimates with their supposed theoretical determinants, most current and past research focuses on the empirics of actually observed unemployment (e.g. Nickell 1997; Blanchard 2006; Baccaro and Rei 2007; Stockhammer and Klär 2011) instead of statistically-detrended NAIRU values. We also go beyond past work by including data on the period after the financial crisis of 2007/2008 and by providing an integrated discussion on the relevance of NAIRU estimates for policymaking. What are the main findings of chapter 4? Our results raise some skepticism with regard to the adequacy of the Commission’s NAIRU estimates. We find that the NAIRU, as estimated by the Commission, is not a good proxy for ’structural unemployment’. According to our econometric results, most indicators of labor market institutions – employment protection legislation, union density, tax wedge and minimum wage – do not perform well as explanatory variables; either is their sign inconsistent with the expectation from standard theory, they are statistically insignificant, or their significance is sensitive to the model specification. The point that NAIRU estimates are not simply driven by (changes in) labor market institutions is underscored by the finding that cyclical factors – especially capital accumulation and boom-bust patterns in housing markets – are important econometric determinants of the Commission’s NAIRU estimates. Therefore, we find that the empirics of the NAIRU are in conflict with the underlying theoretical framework, which posits that the NAIRU should be stripped off all cyclical influences. Our findings are highly relevant for fiscal policy-making in Europe, because biased NAIRU estimates – that are (partly) driven by cyclical factors – may misinform fiscal policy-makers regarding the actual scope for fiscal policies.

1919

References Angrist, J.; Pischke, J. (2009): Mostly Harmless Econometrics. An Empiricist’s Companion, Princeton: Princeton University Press. Arestis, P.; Sawyer M. (2011): The Problems of the Economic and Monetary Union: Is There Any Escape?, Journal of Economic Analysis, 1(1), 1–14. Arpaia, A.; Kiss, A.; Turrini, A. (2014): Is unemployment structural or cyclical? Main features of job matching in the EU after the crisis, European Economy – Occasional Papers 527. Baccaro, L.; Rei, D. (2007): Institutional Determinants of Unemployment in OECD Countries: Does the Deregulatory View Hold Water?, International Organization, 61(3), 527-569. Baker, D.; Glyn, A.; Howell, D; Schmitt, J. (2005): Labor market institutions and unemployment: a critical assessment of the cross-country evidence. In: Howell, D. (ed.) (2005): Fighting Unemployment. The Limits for Free Market Orthodoxy, Oxford: Oxford University Press. Ball, L.; Mankiw, N. (2002): The NAIRU in Theory and Practice, Journal of Economic Perspectives, 16(4), 115–136. Bassanini, A.; Duval, R. (2006): The determinants of unemployment across OECD countries: Reassessing the role of policies and institutions, OECD Economic Studies, 2006 (1), 7–86. Beunza, D.; Stark, D. (2004): Tools of the Trade: The Socio-Technology of Arbitrage in a Wall Street Trading Room, Industrial and Corporate Change, 13, 369-400. Blanchard, O.; Wolfers, J. (2000): The Role of Shocks and Institutions in the Rise of European Unemployment: The Aggregate Evidence, Economic Journal, 110 (462), C1–33. Blanchard, O. (2006): European unemployment: the evolution of facts and ideas, Economic Policy, 21(45), 5–59. Blanchard, O.; Leigh, D. (2013): Growth Forecast Errors and Fiscal Multipliers, American Economic Review: Papers and Proceedings, 103(3), 117-120. Blyth, M. (2013): Austerity. The History of a Dangerous Idea, New York: Oxford University Press. Blinder, A.; Zandi, M. (2015): The Financial Crisis: Lessons for the next one, Report by the Center on Budget and Policy Priorities, October 2015. Braun, B. (2014): Why Models Matter. The Making and Unmaking of Governability in Macroeconomic Discourse, Journal of Critical Globalisation Studies, 7, 48-79. Braun, B. (2015): Governing the future: the European Central Bank’s expectation management during the Great Moderation, Economy and Society, 44(3), 367-391. Braun, B. (2016): From performativity to political economy: index investing, ETFs and asset manager capitalism, New Political Economy, 21(3), 257-273. Buiter, W. (1977): “Crowding out” and the effectiveness of fiscal policy, Journal of Public Economics, 7(3), 309-328. Callon, M. (1998): Introduction: The Embeddedness of Economic Markets in Economics, in: Callon, M. (ed.) (1998): The Laws of Markets, Oxford: Blackwell Publishing, 1-59. Canton, E.; Grillo, I.; Monteagudo, J.; Pierini, F.; Turini, A. (2014): The Role of Structural Reform for Adjustment and Growth. ECFIN Economic Briefs 34/2014, European Commission. Christiano, L.; Eichenbaum, M.; Rebelo, S. (2011): When Is the Government Spending Multiplier Large?, Journal of Political Economy, 119(1), 78-121. Claeys, G.; Darvas, Z.; Leandro, A. (2016): A Proposal to revive the European Fiscal Regulation Framework, Bruegel Policy Contribution 2016/07. Coeure, B. (2012): Central banks and the challenges of the Zero Lower Bound, Speech by Benoit Coeure at the “Meeting on the Financial Crisis”, Miami (February 19th 2012), https://www.ecb.europa.eu/press/key/date/2012/html/sp120219.en.html [last download on November 20th 2017]. Coibion, O.; Gorodnichenko, Y.; Ulate, M. (2017): The cyclical sensitivity of potential output, NBER Working Paper No. 23580. Cottarelli, C.; Gerson, P.; Senhadji, A. (2014): Post-crisis fiscal policy, Cambridge: MIT Press. Cristelli, M.; Tacchela, A.; Pietronero, L. (2015): The Heterogenous Dynamics of Economic Complexity, PLoS ONE, 10(2), e0117174. De Grauwe, P. (2012): The Governance of a Fragile Eurozone, Australian Economic Review, 45(3), 255–268. De Grauwe, P.; Ji, Y. (2013): From Panic-Driven Austerity to Symmetric Macroeconomic Policies in the Eurozone, Journal of Common Market Studies, 51, 31–41.

2020

DeLong, B.; Summers, L. (2012): Fiscal Policy in a Depressed Economy. Brookings Papers on Economic Activity, 44 (1), 233–297. ECB (2010): Monthly Bulletin 2010, ECB publication June 2010, Frankfurt: European Central Bank. ECB (2015). Inflation and unemployment in Europe. Conference Proceedings (21-23 May 2015. Sintra, Portugal), Frankfurt: European Central Bank. ECFIN (2013). Building a Strengthened Fiscal Framework in the European Union: A Guide to the Stability and Growth Pact, European Economy – Occasional Papers 150, Brussels: European Commission. Eggertsson, G.; Krugman, P. (2012): Debt, Deleveraging, and the Liquidity Trap: A Fisher-Minsky-Koo Approach, The Quarterly Journal of Economics, 127(3), 1469–1513. Eichengreen, B. (2008): Sui Generis EMU, European Economy - Economic Papers 303. European Commission (2010): Quarterly Report on the Euro Area, 9(2), Brussels: European Commission. European Commission (2016): Towards a positive fiscal stance in the euro area, COM (2016) 727 final, Brussels: European Commission. Farhi, E.; Werning, I: (2016): Fiscal Multipliers: Liquidity Traps and Currency Unions, Handbook of Macroeconomics, 2, 2417-2492. Farhi, E.; Werning, I. (2017): Fiscal Unions, American Economic Review, 107(12), 3788-3834. Farrell, H.; Quiggin, J. (2017): Consensus, Dissensus, and Economic Ideas: Economic Crisis and the Rise and Fall of Keynesianism, International Studies Quarterly, 61(2), 269-283. Fiscal Compact (2012). Treaty on Stability, Coordination and Governance in the Economic and Monetary Union, T/SCG/EN 1, Brussels: European Commission. Friedman, M. (1968). The role of monetary policy, American Economic Review, 58(1), 1–17. Gechert, S. (2014): On the measurement, theory and estimation of fiscal multipliers, Dissertation, Chemnitz: Technische Universität Chemnitz. Gechert, S.; Rannenberg, A. (2014): Are Fiscal Multipliers Regime-Dependent? A Meta Regression Analysis, IMK Working Paper 139. Gechert, S. (2015): What fiscal policy is most effective? A meta-regression analysis, Oxford Economic Papers, 67(3), 553-580. Guajardo, J.; Leigh, D.; Pescatori, A. (2014): Expansionary Austerity? International Evidence, Journal of the European Economic Association, 12(4), 949–968. Havik, K.; Morrow, K.; Orlandi, F.; Planas, C.; Raciborski, R.; Roeger, W.; Rossi, A.; Thum-Thysen, A.; Vandermeulen, V. (2014): The Production Function Methodology for Calculating Potential Growth Rates and Output Gaps, European Economy - Economic Papers 535. Hirschman, D.; Berman, E. (2014): Do economists make policies? On the political effects of economics, Socio-Economic Review, 12, 779811. IMF (2003): Unemployment and labor market institutions: Why reforms pay off, IMF World Economic Outlook, April 2003(1), 129–150. IMF (2010): World Economic Outlook Update. „Restoring Confidence without harming recovery“, Update to the World Economic Outlook in April 2010, http://www.imf.org/external/pubs/ft/weo/2010/update/02/ [last download on November 20th 2017]. Iversen, T.; Soskice, D.; Hope, D. (2016): The Eurozone and Political Economic Institutions, Annual Review of Political Science, 19, 163185. Jorda, O.; Taylor, A. (2016): The Time for Austerity: Estimating the Average Treatment Effect of Fiscal Policy, The Economic Journal, 126, 219-255. Juncker, J.; Tusk, D.; Dijsselbloem, J.; Draghi, M.; Schulz, M. (2015): Completing Europe’s Economic and Monetary Union, The Five Presidents’ Report (Chapter 4), Brussels, 22 June 2015, 13-16. Khatiwada, S. (2009): Stimulus Packages to counter global economic crisis: A review, International Institute for Labour Studies Discussion Paper 196/2009. Klär, E. (2013): Potential economic variables and actual economic policies in Europe, Intereconomics: Review of European Economic Policy, 48, 33-40. Krugman, P. (2013): Multiplier Evidence and Multiplier Denial, The Conscience of a Liberal – New York Times Blog (December 2nd 2015), https://krugman.blogs.nytimes.com/2015/12/02/multiplier-evidence-and-multiplier-denial/ [last download on November 20th 2017]. Lane, P. (2012): The European Sovereign Debt Crisis, Journal of Economic Perspectives, 26(3), 49-68. Laubach, T. (2001): Measuring The NAIRU: Evidence From Seven Economies, The Review of Economics and Statistics, 83(2), 218–231. Leeper, E.; Traum, N.; Walker, T. (2017): Clearing Up the Fiscal Multiplier Morass, American Economic Review, 107(8), 2409-2454.

2121

Lockwood, E. (2015): Predicting the unpredictable: Value-at-risk, performativity, and the politics of financial uncertainty, Review of International Political Economy, 22, 719-756. MacKenzie, D. (2006): An Engine, Not a Camera. How Financial Models Shape Markets, Cambridge: MIT Press. Martin, P.; Philippon, T. (2017): Inspecting the Mechanism: Leverage and the Great Recession in the Eurozone, American Economic Review, 107(7), 1904-1937. Mishkin, F. (2007): Estimating Potential Output, Speech at the Conference on Price Measurement for Monetary Policy, Federal Reserve Bank of Dallas, Texas, May 24th 2007. Mody, A. (2015): Living Dangerously Without A Fiscal Union, Bruegel Working Paper 2015/03. Nakamura, E.; Steinsson, J. (2017): Identification in Macroeconomics, NBER Working Paper 23968. Nickell, S. (1997): Unemployment and Labor Market Rigidities: Europe versus North America, Journal of Economic Perspectives, 11(3), 55–74. Orphanides, A.; van Norden, S. (2002): The unreliability of output-gap estimates in real time, Review of Economics and Statistics, 84(4), 569-583. Parker, J. (2011): On Measuring the Effects of Fiscal Policy in Recessions, Journal of Economic Literature, 49(3), 703-718. Planas, C.; Rossi, A. (2015). Program GAP. Technical Description and User-manual (Version 4.4), JRC Scientific and Technical Reports April 2015, Joint Research Centre Ispra (Italy). OECD (1994): The OECD Jobs Study. Facts, Analysis, Strategies. OECD study, Paris: Organisation for Economic Co-operation and Development. Orlandi, F. (2012): Structural unemployment and its determinants in the EU countries, European Economy – Economic Papers 455, Brussels: European Commission. Ramey, V. (2011): Can Government Purchases Stimulate the Economy?, Journal of Economic Literature, 49(3), 673-685. Romer, C.; Romer, D. (2010): The Macroeconomic Effects of Tax Changes: Estimates Based on a New Measure of Fiscal Shocks, American Economic Review, 100(3), 763–801. Saka, A.; Fuertes, A.; Kalotychou, E. (2015): ECB Policy and Eurozone Fragility: Was De Grauwe Right?, Journal of International Money and Finance, 54, 168-185. Shambaugh, J. (2012): The Euro’s Three Crises, Brookings Papers on Economic Activity, 43, 157-231. Siebert, H. (1997). Labor Market Rigidities: At the Root of Unemployment in Europe, Journal of Economic Perspectives, 11(3), 37–54. Skidelsky, R.; Fraccaroli, N. (2017): Austerity Vs Stimulus: The Political Future of Economic Recovery, London: Palgrave Macmillan Staiger, D.; Stock, J.; Watson, M. (1997): The NAIRU, Unemployment and Monetary Policy, Journal of Economic Perspectives, 11(1), 33– 49. Stockhammer, E.; Klär, E. (2011): Capital accumulation, labour market institutions and unemployment in the medium run, Cambridge Journal of Economics, 35(2), 437–457. Summers, L. (2016): The Age of Secular Stagnation: What it is and what to do about it, Foreign Affairs, 95(2), 2-9. Tereanu, A.; Tuladhar, A.; Simeone, A. (2014): Structural Balance Targeting and Output Gap Uncertainty, IMF Working Papers 14/107. Woodford, M. (2011): Simple Analytics of the Government Spending Multiplier, American Economic Journal: Macroeconomics, 3(1), 1-35. Wren-Lewis, S. (2016): A general theory of austerity, Blatavnik School of Government Working Paper No. 2016/014. Wyplosz, C. (2006): European Monetary Union: the dark sides of a major success, Economic Policy, 21(46), 207–261.

2222

Review of Keynesian Economics, Vol. 5 No. 3, Autumn 2017, pp. 439–458

Did fiscal consolidation cause the double-dip recession in the euro area? Philipp Heimberger*

Vienna Institute for International Economic Studies (WIIW) and Institute for Comprehensive Analysis of the Economy, Johannes Kepler University Linz, Austria

This paper analyses the short-run effects of fiscal consolidation measures on economic activity in the euro area during the euro crisis. It presents new econometric estimates on the link between cumulative GDP growth and fiscal austerity measures during 2011– 2013. The main empirical finding is that the depth of the economic crisis in the euro area’s economies is closely related to the harshness of fiscal austerity. Cumulative multiplier estimates are found to vary in a range from 1.4 to 2.1, depending on the data source used to identify the intensity of fiscal consolidation. Given these multiplier values, a reasonable approximation of the size of the output losses due to fiscal austerity in the euro area during 2011–2013 is in the range of 5.5 to 8.4 percent of GDP. Against the background of the prevailing macroeconomic and institutional circumstances, fiscal consolidation is argued to be the cause of the double-dip recession. Keywords: fiscal policy, fiscal multiplier, fiscal consolidation, austerity, growth, eurozone JEL codes: E61, E62, E63

1

INTRODUCTION

Since 2010–2011, fiscal consolidation has been a central feature of crisis management in the euro area. Fiscal consolidation measures are defined as cuts in government spending and/or tax increases, motivated by policymakers’ desire to cut the fiscal deficit. What was the short-run impact of fiscal austerity on economic activity in the euro area? This paper analyses the ‘short-run’ growth effects of fiscal consolidation measures in the sense that it provides estimates on cumulative multipliers for the 3-year period 2011–2013, while it does not estimate the long-run (hysteresis) effects of the austerity policies under study. The research goal of this paper is to contribute to explaining the role of fiscal policy in the euro area’s double-dip recession, which started after the third quarter of 2011 and developed into a prolonged recession in 2012 and 2013 (CEPR 2015). The main contributions of the analysis are as follows. First, we present new econometric estimates on the link between cumulative real GDP growth and fiscal consolidation measures in the euro area. The econometric baseline results are used to obtain estimates on the size of cumulative GDP losses in the euro area during 2011–2013, * Email: [email protected]. The Institute for New Economic Thinking supported this work under Grant INO1500015. The author thanks Jakob Kapeller, Mario Holzner, Robert Stehrer, participants at the FMM Conference 2015 in Berlin, as well as two anonymous referees for helpful comments. © 2017 The Author

Journal compilation © 2017 Edward Elgar Publishing Ltd The Lypiatts, 15 Lansdown Road, Cheltenham, Glos GL50 2JA, UK and The William Pratt from House,Elgar 9 Dewey Court,atNorthampton MA07:50:16AM 01060-3815, USA Downloaded Online 07/13/2017

23

via Wirtschaftsuniversitaet Wien Vienna

440

Review of Keynesian Economics, Vol. 5 No. 3

which are then related to already existing estimates. Second, the paper focuses on the time period of the double-dip recession (2011–2013), which has so far received little attention in the macroeconometric literature. Third, this paper looks at a variety of data sources to measure fiscal consolidation. Specifically, we use changes in the structural budget balance, an approach that Blanchard and Leigh (2013) have proposed in a seminal paper; but we also consider the ‘narrative record’ from budgets and policy documents in the spirit of Romer and Romer (2010) to identify size and timing of fiscal consolidation measures. This multi-data-sources approach offers the advantage of allowing for an evaluation of whether the econometric results on the effects of fiscal austerity are consistent across different approaches to identifying fiscal consolidation. The fourth contribution of this paper is to provide an integrated discussion on the role of the institutional and macroeconomic circumstances in the euro area with regard to the determinants of the size of fiscal multipliers. What are the main findings? First, the empirical evidence points to a strong negative correlation between cumulative GDP growth and fiscal consolidation measures in the euro area’s economies during 2011–2013. The depth of the economic crisis was closely associated with the harshness of fiscal consolidation. Cumulative multiplier estimates range from 1.4 to 2.1, depending on the data source used to measure the extent of fiscal austerity. This econometric finding is consistent with reviewing the fiscal multiplier literature, emphasizing key conditions for fiscal multipliers higher than 1.0 that were fulfilled in large parts of the euro area. Second, using the econometric results as an approximation for the size of fiscal multipliers during 2011–2013 leads to a range of cumulative output losses due to fiscal austerity from about 5.5 to 8.4 percent of GDP. Under the macroeconomic and institutional circumstances prevailing in the euro area over the time period studied, fiscal consolidation is the cause of the double-dip recession. The remainder of this paper is structured as follows. Section 2 reviews the literature on fiscal multipliers. Section 3 describes the econometric strategy for analysing the link between cumulative GDP growth and fiscal consolidation measures. Section 4 presents the baseline econometric results and relates them to existing estimates from the literature on the size of GDP losses from fiscal consolidation in the euro area. Section 5 provides several robustness checks, as we account for the role of outliers, vary the country group and control for additional variables. Section 6 summarizes and concludes. 2

WHICH FACTORS DETERMINE THE SIZE OF FISCAL MULTIPLIERS?

The fiscal multiplier is typically defined as the ratio of a change in real GDP to an exogenous change in the fiscal balance (see for example Batini et al. 2014). Several studies demonstrate that multiplier values reported in the literature vary substantially (for example, Hemming et al. 2002; Fatas and Mihov 2009; Gechert and Rannenberg 2014; Alesina et al. 2015). The literature suggests that numerous factors affect the size of multipliers: monetary policy accommodation, the composition of fiscal consolidation (spending-based vs tax-based), the initial level of public indebtedness, the exchange-rate regime, the openness of the economy, spillover effects with other economies, and the international business environment (for example, Ramey 2011; Arestis 2012; Barrell et al. 2012; Ilzetzki et al. 2013). Gechert and Rannenberg (2014) conduct a meta-regression analysis of 98 empirical studies to discover whether fiscal multipliers vary with the business cycle. They find that multipliers increase by 0.6 to 0.8 units during an economic downturn and report that spending multipliers are markedly higher than tax multipliers, especially during recessions. During ‘normal’ © 2017 The Author

Journal compilation © 2017 Edward Elgar Publishing Ltd

Downloaded from Elgar Online at 07/13/2017 07:50:16AM24 via Wirtschaftsuniversitaet Wien Vienna

Did fiscal consolidation cause the double-dip recession in the euro area?

441

economic times and during booms, fiscal multipliers are not only lower than in downturns; they also vary less across different fiscal instruments. Several multiplier studies from recent years report that multipliers are substantially higher when economic resources are underutilized (for example, DeLong and Summers 2012; Charles et al. 2015; Fazzari et al. 2015; Qazizada and Stockhammer 2015; Jorda and Taylor 2016). 2.1

Multipliers in crisis times

In what follows, we focus on the literature on the size of multipliers in crisis times. The case of severe restrictions in conventional monetary policy effectiveness due to the zero lower bound of nominal interest rates (ZLB) has gained relevance since the outbreak of the financial crisis in 2008. This is especially the case in the mainstream New Keynesian literature, where it is argued that fiscal multipliers are substantially higher than 1.0 if central banks are constrained by the ZLB in their ability to stimulate the economy with interest rate cuts (for example, Christiano et al. 2011; Woodford 2011). Another research strand in the multiplier literature investigates how characteristics of financial crises and their aftermaths might influence fiscal policy effectiveness. For example, Corsetti et al. (2012) report that fiscal multipliers are significantly above 2.0 during times of financial crisis. Eggertsson and Krugman (2012) show that in a New Keynesian model of debt-driven slumps, where agents in the private sector are forced into rapid deleveraging, the result is a multiplier in excess of 1. Koo (2013) argues that fiscal multipliers are markedly higher than 1.0 as long as the private sector is collectively minimizing debt after an asset price bubble has burst, because the deleveraging acts as a drag on aggregate demand. The arguments presented above have implications for the research questions on the effects of fiscal consolidation measures on output, because conditions for multipliers in excess of 1.0 were actually fulfilled in the euro area during 2011–2013. The ECB was severely constrained in its ability to stimulate the economy by cutting interest rates because of the ZLB (for example, Coeure 2012). In large parts of the European Monetary Union, the private sector was in the process of deleveraging (see Koo 2015, pp. 219–229), and therefore not in a position to borrow – even at very low interest rates – which impaired the effectiveness of monetary policy. Furthermore, the monetary union is a fixed exchange-rate regime, in which individual member countries do not have control over the currency in which they issue debt (De Grauwe 2012). Therefore, currency devaluations were not available to stressed countries in order to increase price competitiveness vis-à-vis main trading partners and stimulate the economy via an increase in exports. Also, the initial position of euro-area economies in 2010–2011 was characterized by significant economic slack. The IMF estimated in real time that all euro-area countries except Malta had negative output gaps (to varying degrees) over the years 2010–2012 (IMF 2011). Negative output gaps are widely accepted as a standard indication that there are demand-side problems and that in principle it would be possible to increase production and to decrease unemployment by demand-side measures without creating any inflationary pressures. However, such standard output-gap measures tend to severely underestimate the extent of resource underutilization during a recession (for example, Klär 2013; Palumbo 2015). These points underscore that the business cycle positions of euro-area economies were incompatible with the expectation that fiscal consolidation measures would have a positive impact on growth and employment, based on the flawed idea of ‘expansionary fiscal contractions’ put forward by Alberto Alesina and others (Dellepiane-Avellaneda 2015). © 2017 The Author



442


2.2

Estimates on the size of GDP losses due to fiscal consolidation in the euro area

How does the existing literature estimate the size of GDP losses due to fiscal consolidation in the euro area during 2011–2013? In the context of this question, it has to be recognized that a comparison of existing studies is complicated by the fact that not all of the relevant papers cover estimates for the whole period 2011–2013. Furthermore, they do not all use the same data on the intensity of fiscal consolidation measures. In order to improve comparability of the studies, Table 1 – which summarizes the most relevant existing estimates on the size of GDP losses in the euro area – also depicts the size of cumulative multipliers. The European Commission (2012b) assesses the impact of fiscal consolidation as the deviation from a baseline scenario without fiscal consolidation. Using simulations with its dynamic stochastic general equilibrium (DSGE) model QUEST, it is estimated that the short-run multiplier of fiscal consolidation is low (around 0.25). Assuming that fiscal plans are fully credible and that monetary policy helps to cushion the contractive effects of fiscal adjustment, the negative impact of fiscal adjustment in 2012 and 2013 is estimated to be very limited (cumulatively 0.5 percent of GDP over 2012–2013; see Table 1). However, the study argues that the effects of fiscal consolidation depend on the reception of fiscal measures in the private sector and on monetary policy accommodation (European Commission 2012b, pp. 46–47). In particular, this means that when both a binding ZLB and the possibility that private-sector agents doubt whether the government is credibly committed to implementing fiscal consolidation measures are introduced into the model, multipliers and GDP losses are markedly higher, as the cumulative multiplier during 2012–2013 increases from 0.25 to 0.8 and the cumulative GDP loss surges from 0.5 to 1.6 percent. Rannenberg et al. (2015) point out that the assessment of the effects of fiscal consolidation on economic activity in the euro area in European Commission (2012b) does not adequately take into account the restrictions imposed on monetary policy by the ZLB, the tightening of liquidity constraints for households as a result of the financial crisis, and that it has not properly allowed for the possibility that households do not anticipate that cuts in government spending imply higher future private consumption because of lower future tax burdens. They employ two DSGE models – one is the New Area Wide Model from the European Central Bank (ECB), the other the European Commission’s QUEST III model – for their simulations, in which they constrain the response of monetary policy, account for liquidity constraints of households, and introduce a Table 1 Estimates of cumulative losses from fiscal consolidation in the euro area during 2011–2013 (in % of GDP)

European Commission (2012b, pp. 45–46) Rannenberg et al. (2015, p. 21) In ’t Veld (2013, p. 10)a Holland and Portes (2012, p. F8) Gechert et al. (2015, p. 6)

2011

2012

2013

Cumulative multiplier

–

0.3 / 0.9

0.5 / 1.6

0.3 / 0.8

– 0.7 0.5 / 1.5 4.3

– 2.0 1.0 / 3.1 6.4

5.2 3.2 1.7 / 4.0 7.7

1.3 2.1b 0.4 / 0.9b 1.9b

Notes: a. Results refer to the core euro area excluding Germany. b. The multiplier estimate was not presented explicitly in the respective paper; it was, therefore, calculated implicitly from the estimated GDP loss due to fiscal austerity and the fiscal impulse data used in the study. Sources: Own illustration, based on the sources cited in the table. © 2017 The Author




443

financial accelerator. They find that in the presence of both the financial accelerator and an increased share of liquidity-constrained households, the cumulative multiplier over the 2011–2013 period equals 1.3. If one considers that the cumulative fiscal consolidation restriction for the euro area during 2011–2013 is 4.0 percent of GDP (see Rannenberg et al. 2015, p. 8), a multiplier of 1.3 implies that fiscal consolidation caused cumulative output losses of about 5.2 percent of GDP. In’t Veld (2013), who also uses the European Commission’s QUEST model, finds that the negative growth effects of fiscal consolidation can be markedly larger when all countries consolidate simultaneously. For the core euro area excluding Germany, he reports a cumulative GDP loss of 3.2 percent from 2011 to 2013. Furthermore, In’t Veld (ibid.) emphasizes that output reductions due to fiscal austerity vary significantly across euro-area countries. For Greece and Portugal, he finds cumulative losses due to austerity that amount to 8.0 and 6.9 percent of GDP, respectively. In comparison, the 3.9 percent loss estimated for Germany is also substantial, but certainly markedly smaller (ibid., pp. 10–11). Holland and Portes (2012) use the National Institute Global Econometric Model (NiGEM), a large-scale macroeconometric model, to assess the economic impact of fiscal consolidation plans for the period 2011–2013. They calculate output losses for two major scenarios. In the first scenario, they assume that interest rates are flexible and not bound at zero, and that liquidity constraints in the private sector are not higher than the long-run average. This scenario yields a cumulative GDP loss due to fiscal consolidation of 1.7 percent of GDP. In scenario 2, however, they ‘allow for an impaired interest rate channel and heightened liquidity constraints – assumptions we consider more realistic under current conditions’ (ibid., p. F8), which yields a markedly larger GDP loss of 4 percent. Gechert et al. (2015) build on the meta-regression analysis by Gechert and Rannenberg (2014) and find that the fiscal consolidation in the euro area reduced GDP by 4.3 percent relative to a baseline scenario without fiscal adjustment in 2011, with the deviation from the baseline increasing to 7.7 percent in 2013. The next section will present the econometric strategy of this paper. Based on the literature review on the size of fiscal multipliers, the main hypothesis is that fiscal consolidation measures and cumulative real GDP growth will be negatively associated, and strongly so when main conditions for multipliers higher than 1.0 are fulfilled. 3

ECONOMETRIC STRATEGY

To investigate whether GDP growth in the euro area has been systematically related to the extent of fiscal consolidation, we use the following econometric approach. We regress the cumulative growth in real GDP during 2011–2013 on a fiscal variable that is supposed to capture exogenous changes in the fiscal balance. The baseline equation estimated is: ΔYi;2011:2013 ¼ α þ βΔFi;2011:2013 þ εi;2011:2013 ; where ΔYi;2011:2013 denotes cumulative growth of real GDP (Y) in economy i during the time period 2011–2013, ΔFi;2011:2013 captures the exogenous change in the fiscal balance in economy i during the time period 2011–2013, and εi;2011:2013 is the error term. How do we measure ΔFi;2011:2013 ? This question is highly important because of an endogeneity problem. Ups and downs in economic activity cause changes in the fiscal balance that are the result of automatic stabilizers; for example, a downswing in economic © 2017 The Author



444


growth will lead to a fall in tax revenues and an increase in unemployment-related government spending, without any actual change in fiscal policy. Such a development would both affect the explanatory variable ΔFi;2011:2013 and the error term in the same direction. In practice, ‘using the change in the overall fiscal balance to measure changes in fiscal policy would bias estimates toward finding expansionary effects of fiscal consolidation on economic activity’ (Guajardo et al. 2011, p. 6), because the fiscal balance improves (worsens) due to the effects of automatic stabilizers that are triggered by an improvement (deterioration) in economic activity. In the macroeconometric literature, one finds two major approaches that try to overcome this endogeneity problem.1 The first can be called the ‘conventional approach’ (for example, Yang et al. 2015), which looks at changes in cyclically adjusted fiscal data. The basic idea is to correct the headline fiscal balance for the effects of the business cycle on government revenues and expenditures. The IMF and the European Commission do so by estimating the fiscal balance at which the output gap – the difference between actual and potential output – would be zero. After correcting for the cyclical component of the fiscal balance, they also account for so-called budgetary one-off effects, for example costs related to bailing-out financial institutions, which yield the structural budget balance (Fedelino et al. 2009; Mourre et al. 2014). The intensity of fiscal consolidation can then be calculated by looking at changes in the structural budget balance – a strategy proposed by Blanchard and Leigh (2013). A typical criticism in the literature is that changes in the structural budget balance might not only reflect the policymakers’ desire to cut the fiscal deficit, which is due to problems related to estimating the fiscal balance at which the output gap would be zero (for example, Carnot and de Castro 2015). Therefore, the contribution of this paper is to look at other data sources as well, as we also follow the second major strategy in the macroeconometric literature for overcoming the endogeneity problem, which is called the ‘narrative approach.’ Inspired by Romer and Romer (2010), ‘narrative’ data sources identify size and timing of fiscal policy measures from budgets, budget documents and policy papers by accounting for policymakers’ motivations for implementing the respective measures. Taking a variety of data sources into account in order to identify the intensity of fiscal consolidation is an important contribution to the existing literature, because we can check whether the econometric findings for the euro area are robust to using different identification strategies. Table 2 lists the data sources used in this paper. It depicts details on the relevant time period for which data was available during 2011–2013 and shows the number of euro-area countries for which data could be included. Regarding the ‘conventional approach,’ we obtain data from European Commission (2015) and IMF (2015), respectively. Data from the ‘narrative approach’ is based on European Commission (2015), OECD (2012), and Gainsbury et al. (2011), respectively. From the six different data sources depicted in Table 2, we obtain cross-sectional data on variations in the intensity of fiscal consolidation that can be accessed in Appendix 1. 4

BASELINE RESULTS AND DISCUSSION

The baseline ordinary least squares (OLS) regressions used in this paper to obtain multiplier estimates build upon the seminal contributions by Blanchard and Leigh (2013) 1. Other approaches exist, but are not discussed here; see, for example, Blanchard and Perotti (2002). © 2017 The Author




445

Table 2 Data sources used to identify fiscal consolidation measures in the euro area (2011–2013)

‘Conventional approach’ IMF (2015) European Commission (2015) European Commission (2015) ‘Narrative approach’ European Commission (2015) OECD (2012) Gainsbury et al. (2011)

Data

Time period

EA countries

Structural budget balance in % of potential output Structural budget balance in % of potential output Primary structural budget balance in % of potential output

2011–2013

16

2011–2013

18

2011–2013

18

Discretionary fiscal measures in % of nominal GDP Fiscal consolidation measures in % of nominal GDP Fiscal consolidation measures in % of GDP per head

2011–2013

18

2011–2013

15

2011

6

Sources: Own illustration, based on the sources cited in the table.

and De Grauwe and Ji (2013). The strength of the OLS approach chosen is that it delivers estimates for the size of multipliers that are more straightforward and easier to interpret than the results from more sophisticated econometric strategies in the recent multiplier literature (for example, Qazizada and Stockhammer 2015; Yang et al. 2015; Jorda and Taylor 2016), but nonetheless robust, as will be demonstrated by the robustness checks in Section 5. There are three major reasons for focusing on the period 2011–2013. First, this paper is especially interested in analysing the effects of fiscal consolidation measures when it comes to explaining the euro area’s double-dip recession (CEPR 2015), which other authors in the macroeconometric literature have so far been unable to study in sufficient depth. Second, although some countries – such as Ireland and Latvia – had started to implement consolidation measures before the year 2011, the simultaneous turn to fiscal austerity in large parts of the euro area was most pronounced during 2011–2013 (for example, European Commission 2012b). Therefore, the time period chosen provides an ideal possibility to exploit the variation in the intensity of fiscal consolidation across the euro area’s economies to obtain econometric estimates on the size of fiscal multipliers. The third and more technical reason is data availability. ‘Conventional approach data’ on fiscal consolidation measures for the years 2009–2013 are only available from the IMF (2015), and ‘narrative approach data’ for this longer time period are not available at all from the data sources depicted in Table 2. Hence, we would not be able to use multiple data sources to identify the intensity of fiscal consolidation measures if we were to focus on the period 2009–2013. Using more than one data source is, however, central to the empirical approach of this paper. The presentation of the baseline results in this section focuses on the euro area.2 However, robustness checks in the next section will also look at the empirical evidence 2. The EA18 country group includes Belgium, Germany, Estonia, Ireland, Greece, Spain, France, Italy, Cyprus, Latvia, Luxembourg, Malta, the Netherlands, Austria, Portugal, Slovenia, Slovakia, and Finland. © 2017 The Author



446


for other country groups in order to investigate whether the experiences of euro-area countries were similar to those of non-euro-area countries. It might be argued that cross-sectional evidence on the link between fiscal consolidation measures and GDP growth strongly depends on the role of outliers. That is why the robustness checks in Section 5 will show that the results are not unduly influenced by outlier observations. Another objection might be that additional variables affect both the intensity of fiscal austerity and real GDP growth. The subsequent robustness analysis will, however, demonstrate that the β coefficient of fiscal consolidation is not unduly affected when we control for additional variables that might have both influenced real GDP growth and fiscal consolidation over the time period studied. Table 3 reports the baseline results from the OLS estimation. Using changes in the structural budget balance as estimated in IMF (2015) in order to identify fiscal consolidations, we find a strong negative correlation between cumulative real GDP growth and fiscal consolidation measures. The β coefficient is –1.85, implying that an increase of 1 percentage point in fiscal consolidation during 2011–2013 was associated with a cumulative decline in real GDP during 2011–2013 of about 1.85 percentage points. Figure 1 illustrates the statistically significant relationship with a scatterplot for each of the six data sources depicted in Table 3.3 Plotting the data suggests that those euro-area countries that implemented more intense fiscal consolidations suffered more pronounced declines in real GDP from 2011 to 2013; vice versa, countries which did not adjust (that much) performed markedly better in terms of real GDP. The estimation results based on data from European Commission (2015) are similar when we identify fiscal consolidation measures by changes in the structural budget Table 3

OLS baseline results for the euro area β

t-value β

α

Number of countries

R2

‘Conventional approach’ data Structural budget balance / IMF Structural budget balance / EC Primary structural budget balance / EC

–1.854 –2.075 –2.089

–5.683*** –5.075*** –3.626***

7.327 7.470 7.936

18 18 18

0.586 0.557 0.573

‘Narrative approach data’ European Commission (2015) OECD (2012) Gainsbury et al. (2011)

–1.382 –1.906 –1.647

–5.183*** –2.927** –6.353***

8.007 6.735 3.733

18 15 6

0.756 0.604 0.833

Sources: Author’s calculations, based on the data sources mentioned in the table. Notes: Dependent variable: cumulative real GDP growth 2011–2013. Note that for the specification using data by Gainsbury et al. (2011) we only had fiscal consolidation data for the year 2011. Following De Grauwe and Ji (2013), we use the cumulative growth in real GDP over 2011–2012 as the dependent variable, which we regress on the narrative-based variable obtained from Gainsbury et al. (2011). T-values are heteroskedasticity-robust (White). ***, **, * denote statistical significance at the 1%, 5%, and 10% level, respectively. Structural budget balance data for Cyprus and Estonia were not available in IMF (2015). Missing values were filled with structural budget balance data from European Commission (2015).

3. Throughout the study, statistical inference is reported based on heteroskedasticity-robust standard errors. © 2017 The Author




EST

LVA MLT

10

GER

LUX

0

Structural budget balance / EC Cumulative real GDP growth 2011−2013

Cumulative real GDP growth 2011−2013

Structural budget balance / IMF 20

SVK IRL

BEL

AUT FRA NLD FIN SVN ESP ITA PRT CYP

−10

GRC

−20 0

5 10 Fiscal consolidation 2011−2013

20

EST

0

AUT BEL

−10

MLT

0 −10

CYP

PRT GRC

−20 0

GRC

0

20 10 0

EST

FIN

−10

SVK FRA

NLD BEL SVN

ESP

ITA PRT GRC

−20 −5

15

LVA MLT

LUX

SVK AUT

BEL

GER FIN

NLD

FRA IRL SVN

ITA

−10

ESP CYP

PRT

−20

GRC

0


10



IRL

0


5 10 15 20 Fiscal consolidation 2011−2013

25

Gainsbury et al. (2011)

LUX GER AUT

PRT

−20

OECD (2012) EST

ESP

ITA CYP


20 10

IRL

Narrative approach data / EC

LVA

SVK AUT LUX GER NLD BEL IRL FRA ESP FIN ITA SVN

SVK

GER FRA NLD

FIN SVN

15



EST

10

LVA MLT LUX

10

Primary structural budget balance / EC 20

447

0

IRL

GER ITA

ESP PRT

−10

GRC

−20 0

5 10 Fiscal consolidation 2011

Source: Own illustration. For more details on the econometric results, see Table 3; for more information regarding the data sources, see Table 2.

Figure 1

Plotting cumulative real GDP growth against fiscal consolidation measures

balance (β coefficient –2.08) and by changes in the primary structural budget balance (which excludes interest payments; β coefficient –2.09), respectively. Using data on the intensity of fiscal consolidation that was obtained from budgets and other relevant documents (‘narrative approach’), we again find a negative, statistically significant relationship between the cumulative growth in real GDP and fiscal consolidation measures during 2011–2013. OLS estimates based on fiscal consolidation numbers reported in OECD (2012)4 deliver a β coefficient of –1.91. Looking at data on discretionary fiscal measures from European Commission (2015), we find that a 1 percentage point increase in fiscal consolidation was associated with a cumulative decline in real GDP by 1.38 percentage points. Obtaining consolidation data from Gainsbury et al. (2011), we once more find a statistically significant negative 4. In the OECD (2012) specification, data for 15 euro area countries were available: the EA18 country group excluding data for Cyprus, Latvia, and Malta. © 2017 The Author



448


association between real GDP growth and austerity measures in the euro-area countries under study (β coefficient of –1.65). The natural interpretation of these econometric findings is that they provide evidence for multipliers that were, on average, substantially higher than 1.0. In Section 2, we have already discussed estimates on the size of cumulative output losses in the euro area during 2011–2013 (see Table 1). How can we use our econometric baseline results in order to contribute to the existing literature? The European Commission estimates that fiscal consolidation in the euro area cumulated to 4.0 percent of GDP between 2011 and 2013 (see European Commission 2012b, pp. 45–46). Looking at the β coefficients from Table 3 as an approximation of the size of cumulative multipliers in the euro area leads to a range of cumulative output losses due to fiscal consolidation from about 5.5 to 8.4 percent of GDP during 2011–2013 – in comparison to the unknown baseline scenario without fiscal austerity measures (see Figure 2). The advantages of these calculations are that they require fewer assumptions and that they are way simpler than building a large macroeconomic model, as Rannenberg et al. (2015) and other researchers have done. But still, these simple calculations can be used as a reasonable approximation of the size of GDP losses in the euro area, which are due to fiscal austerity. As Figure 2 illustrates, the 5.5 to 8.4 percent numbers are in the upper part of the range of estimates from the existing literature. Using the multiplier based on primary structural budget balance data from the European Commission (2015) yields a GDP loss caused by fiscal consolidation of about

Estimates of cumulative losses from fiscal consolidation in % of GDP

10 European Commission (2015), structural balance

9

OECD (2012)

8 7

IMF (2015) Gainsbury et al. (2011)

6 5 4 3

Gechert et al. (2015)

5.5% to 8.4%

European Commission (2015), primary structural balance

Rannenberg et al. (2015)

EC(2015), discretionary fiscal measures

Holland and Portes (2012) In’t Veld (2013)

2 1 European Commission (2012) 0 Each estimate of the size of GDP losses represents a unit on the x-axis. The horizontal step from one estimate to the next is always the same. Estimates were ordered on the x-axis starting from the lowest up to the highest.

Sources: Own illustration. Bold labeling indicates that the estimates are based on the author’s own calculations. The other estimates were obtained from the existing literature. See Table 2 for details on the data sources used to identify fiscal consolidation measures. See Table 1 for a table summary of the estimates from the existing literature.

Figure 2 Mapping the size of cumulative GDP losses due to fiscal consolidation in the euro area (2011–2013) © 2017 The Author




449

8.4 percent of euro-area GDP, which constitutes the upper limit of the range of estimates. 5

ROBUSTNESS CHECKS

In this section, we perform several tests to assess the robustness of the baseline results reported in the previous section. 5.1

The role of outliers

Our first step in the robustness analysis is to analyse the role of outliers. Since critics might object that the baseline results are driven by data for Greece, which implemented the most intense fiscal austerity measures of all countries, we exclude Greece from our sample. Using data from IMF (2015), the R2 declines from 0.59 to 0.35 and the β coefficient is now statistically significant at the 5 percent level (see Table 4). The size of the β coefficient is even larger (–2.05 compared to –1.85). We then test the sensitivity of the baseline results to outliers formally by applying three accepted estimation strategies designed to resist the influence of outliers. First, we re-estimate the baseline specification using robust regression, which downweighs observations with larger absolute residuals by making use of iterative weighted least squares. Robust regression is less fragile to the influence of outlier observations than OLS; the procedure is a check of whether outliers are influencing the baseline OLS results (see Blanchard and Leigh 2013, p. 9). The robust regression estimate of β (–1.84) is very similar to the OLS estimate (–1.85). The second variation in the estimation technique is implemented via quantile regression, which is also supposed to make the estimates less affected by the role of outlier observations. 5 The quantile regression estimate of β (–1.80) is again very similar to our OLS estimate. The third variation in the estimation technique was introduced as follows. We investigate the role of outlier observations by using Cook’s distance method; the approach was to discard observations with Cook’s distance greater than 4/N, where N is the sample size (eighteen countries in case of the EA18). In our euro-area sample, Cook’s distance is smaller than 4/ N for all euro-area countries; therefore, our Cook’s distance estimates are identical to the OLS estimates.6 5.2

Variations in the country group

The second step of our robustness checks is to vary the country group in order to shed light on whether the experiences of euro-area countries are similar to those of non-euro-area countries. Table 4 reports regression results not only for the EA18,

5. Quantile regression minimizes the sum of the absolute residuals about the median, rather than the sum of the squares of the residuals about the mean as in OLS (see Blanchard and Leigh 2013, p. 10). 6. Results of the exact same robustness checks, based on OLS estimates from the European Commission (2015) ‘conventional approach’ data, support the finding that the robust regression, quantile regression, and Cook’s distance estimates of β are very similar to the OLS estimate, and that they are all statistically significant. Results are available on request from the author. © 2017 The Author



450


Table 4

Robustness checks: the role of outliers and variations in the country group β

t-value β

α

Number of countries

R2

EA18 (OLS) OLS excl. Greece

–1.854 –5.683*** –2.049 –2.495**

7.327 7.803

18 17

0.586 0.351

Robust regression Quantile regression Cook’s distance

–1.835 –5.667*** –1.799 –2.627** –1.854 –5.683***

7.098 7.506 7.327

18 18 18

0.585 0.583 0.586

EU27 (OLS) OLS excl. Greece Advanced European

–1.549 –3.100*** –1.133 –1.649 –1.620 –4.300***

7.184 6.156 6.834

27 26 23

0.404 0.134 0.454


–1.531 –2.986*** –1.826 –2.026* –1.133 –1.649

6.989 7.897 6.156

27 27 26

0.403 0.390 0.134

Advanced Economies (OLS) OLS excl. Greece Liquidity trap No liquidity trap

–1.590 –1.326 –1.594 –0.279

–4.727*** –3.088*** –5.002*** –0.438

7.718 7.270 7.075 8.291

36 35 29 7

0.452 0.228 0.469 0.044


–1.588 –4.720*** –1.831 –3.374*** –1.326 –3.088***

7.666 7.973 7.270

36 36 35

0.452 0.439 0.228

12.393

35

0.063

12.077 11.192 12.848

35 35 34

0.060 0.001 0.115

Emerging Market Economies (OLS) –0.807 –1.309 Robust regression Quantile regression Cook’s distance

–0.662 –0.950 –1.355 –1.515 –1.174 –2.018*

Notes: Dependent variable: cumulative real GDP growth 2011–2013. T-values are heteroskedasticity-robust (White). Fiscal consolidation is measured as the change in the structural budget balance. ***, **, * denote statistical significance at the 1%, 5%, and 10% level, respectively. Structural budget balance data for Cyprus and Estonia was not available. Missing values were filled with structural budget balance data from European Commission (2015). The country sample in the specification ‘Advanced European’ is the EU27 excluding Romania, Hungary, Bulgaria, and Poland. In the ‘Liquidity trap’ specification, we excluded Australia, Iceland, Israel, Korea, New Zealand, Norway, and Taiwan; these countries comprise the ‘no liquidity trap’ country group. Sources: Data on fiscal consolidation and real GDP: IMF (2015); author’s calculations.

but also for the EU27,7 a group of advanced economies (including European and nonEuropean economies)8 and emerging market economies.9 7. The EU27 consists of Austria, Belgium, Bulgaria, Cyprus, the Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, the Netherlands, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, Sweden, and the United Kingdom. 8. The advanced country group consists of 36 countries: Australia, Austria, Belgium, Canada, Cyprus, the Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hong Kong SAR, Iceland, Ireland, Israel, Italy, Japan, Korea, Latvia, Lithuania, Luxembourg, Malta, the Netherlands, New Zealand, Norway, Portugal, Singapore, Slovakia, Slovenia, Spain, Sweden, Switzerland, Taiwan, the United Kingdom, and the United States. 9. This emerging markets group consists of 35 countries: Argentina, Bosnia, Brazil, Bulgaria, Chile, Colombia, Croatia, the Dominican Republic, Ecuador, Egypt, Georgia, Guyana, Hungary, © 2017 The Author


Downloaded from Elgar Online at 07/13/2017 07:50:16AM via Wirtschaftsuniversitaet Wien Vienna

34


451

For many of these additional economies, the conditions for multipliers in excess of 1.0 discussed while reviewing the fiscal multiplier literature (such as the ZLB constraint and slack in the economy) are arguably less relevant than in the euro area, which leads us to expect a smaller absolute value of β for the EA27, the advanced economies sample, and the emerging markets country group – compared to the EA18, respectively. We find that the β coefficient of fiscal consolidation is strongly negative and statistically significant in the EA18, EU27, and advanced economies specification, respectively; however, β is markedly more negative for the EA18 (–1.9) than for the EU27 (–1.5) and advanced economies country group (–1.6). Furthermore, statistical significance for the EU27 has declined; the quantile regression and Cook’s distance estimate point to the role of outliers influencing the EU27 OLS estimates. It is also notable that excluding Greece from the OLS estimation has more impact on the results for the EU27 and advanced economies group than on the EA18. In the advanced economies specification, we also test for the possible role of constraints in monetary policy. We do so by estimating a separate specification in which we only include economies that were, arguably, in a liquidity trap during this period.10 In this specification of 29 advanced economies, the estimate of β is –1.59 and strongly significant; in the – admittedly small – group of 7 no-liquidity-trap advanced economies, however, β is –0.28 and lacks significance. When we repeat the analysis for the group of 35 emerging market economies for which the IMF (2015) provided structural budget balance data, we find a β coefficient of –0.8. The fiscal consolidation coefficient in the emerging markets specification lacks significance, which also does not change when we perform robustness checks by implementing more robust estimation procedures. This finding points to the importance of accounting for the conditions of fiscal multipliers higher than 1.0, which were less important in emerging market economies during 2011–2013 than in the euro area and other parts of the global economy. Differences in the size of the fiscal multiplier across country groups might be explainable – to a non-negligible extent – by differences in the monetary policy regime. For virtually none of the emerging market economies in our sample, the central bank’s main nominal policy interest rate reached 1 percent or less during 2011–2013.11 In stark contrast, 24 of the EU27 countries did face such a liquidity trap situation at some point over the same time period.12 Additionally, Table 5 reports evidence on the link between cumulative real GDP growth and fiscal consolidation measures before the financial crisis for comparable 3-year periods (2005–2007 and 2002–2004) in a sample of 15 euro-area countries13

India, Indonesia, Jordan, Lebanon, Malaysia, Mauritius, Mexico, Morocco, Panama, Paraguay, Peru, the Philippines, Poland, Romania, Russia, Serbia, South Africa, Thailand, Turkey, Ukraine, and Uruguay. 10. The term ‘liquidity trap’ describes a situation characterized by the central bank’s inability to use interest rate cuts in order to induce investors to lend money. Consistent with Blanchard and Leigh (2013), we define our set of liquidity trap economies as those economies for which the central bank’s main nominal policy interest rate reached 1 percent or less during 2011–2013. 11. Bulgaria is the only notable exception. 12. The three exceptions are: Hungary, Poland, and Romania. 13. The 15 euro-area countries group in Table 5 consists of Austria, Belgium, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, Malta, the Netherlands, Portugal, Slovakia, Slovenia, and Spain. © 2017 The Author



452


Table 5

Robustness check regarding precrisis years

EA15 2005–2007 EA15 2002–2004 Advanced economies 2005–2007 Advanced economies 2002–2004

β

t-value β

α

Number of countries

R2

–1.245 –0.183 –0.275 0.308

–1.670 –0.205 –0.352 0.829

11.624 7.257 13.563 9.626

15 15 31 31

0.289 0.008 0.009 0.016

Notes: T-values are heteroskedasticity-robust (White). The fiscal variable is measured as the change in the structural budget balance. ***, **, * denote statistical significance at the 1%, 5%, and 10% level, respectively. Source: Data on fiscal consolidation and real GDP growth: IMF (2015); author’s calculations.

and 31 advanced economies.14 We find for both country groups that the β coefficient of fiscal consolidation is much less negative than during 2011–2013; it also lacks statistical significance in all of the precrisis specifications, which is in line with our hypothesis that conditions for fiscal multipliers higher than 1.0 mattered during 2011–2013. 5.3

Including additional control variables

The next step of the robustness checks is to introduce additional control variables, which could potentially explain both the intensity of fiscal consolidation and the evolution of real GDP. The omission of such potentially relevant control variables could bias the analysis towards overestimating the size of the negative β coefficient. When it comes to including additional controls, we estimate the following equation: ΔYi;2011:2013 ¼ α þ βΔFi;2011:2013 þ γXi;t þ εi;2011:2013 ; where ΔYi;2011:2013 and ΔFi;2011:2013 are defined as in the specification introduced in Section 3, and Xi;t represents additional control variables in economy i at time t, where t refers either to the initial year 2010 before the time period 2011–2013 or to the precrisis year 2007 – depending on the respective additional control variable, which will be described below in more detail. Additional controls are introduced into the robustness check specifications one at a time. Which additional variables does this paper control for, and what are the reasons for including them? First, economists who are suspicious of multipliers higher than 1.0 – as estimated by this paper in Section 4 – might argue that it is no surprise that economic growth turned out to be so weak in large parts of the euro area, given that government debt levels were high to start with in 2010. Although Herndon et al. (2014) have conclusively demonstrated the flaws and untenable nature of the infamous Reinhart and Rogoff (2010) finding that ‘across both advanced countries and emerging markets, high debt/GDP levels (90 percent and above) are associated with notably lower growth outcomes’ (ibid., p. 22), it might still be claimed that ‘[t]he circumstances which help 14. The 31 advanced economies from the country group in Table 5 consists of: Australia, Austria, Belgium, Czech Republic, Denmark, Finland, France, Germany, Greece, Hong Kong SAR, Iceland, Ireland, Israel, Italy, Japan, Korea, Lithuania, Luxembourg, Malta, the Netherlands, New Zealand, Norway, Portugal, Singapore, Slovakia, Slovenia, Spain, Sweden, Switzerland, Taiwan, the United Kingdom, and the United States. © 2017 The Author




Table 6

453

Robustness checks: additional control variables

Initial debt-to-GDP ratio Initial structural budget balance Initial fiscal balance Sovereign CDS spread Bank CDS spread Precrisis current-account balance Precrisis stock of net foreign liabilities Precrisis household debt-toincome

R2

β

t-value β

γ

–1.441 –1.684 –2.028 –2.067 –1.921 –1.988 –2.057

–3.813*** –3.110*** –3.087*** –2.156** –1.979* –6.033*** –4.724***

–0.059 0.203 –0.183 0.004 0.004 –0.191 –0.027

–1.434 0.285 –1.799* 0.282 0.227 –0.923 –0.853

18 18 18 17 10 18 18

0.622 0.588 0.607 0.585 0.861 0.626 0.613

–1.561

–4.851***

0.023

1.365

12

0.799

t-value Number of γ countries

Notes: Dependent variable: cumulative real GDP growth 2011–2013. The fiscal variable is measured as the change in the structural budget balance. γ refers to the coefficient of the control variable. T-values are heteroskedasticity-robust (White). ***, **, and * denotes statistical significance at the 1%, 5%, and 10% level, respectively. Constant term included in specification, but the estimate is not reported. The additional controls appear in the specifications one at a time. Source: Data on fiscal consolidation and GDP growth: IMF (2015).

to reduce the short-term costs [of fiscal consolidations] include when ... the fiscal starting position is particularly precarious and thus confidence in the sustainability of public finances is rather low’ (ECB 2010, p. 84). In order to anticipate the argument that the baseline OLS results from Section 4 are picking up the effects of public debt problems rather than the effects of fiscal consolidation measures, the robustness checks consider the role of the initial sovereign debt situation in the euro area in 2010. As can be seen from Table 6, the baseline results are robust to controlling for the initial (end-2010) government-debt-to-GDP ratio, for the initial (end-2010) fiscal-balance-toGDP ratio, and for the initial (end-2010) structural-budget-balance-to-potential-output ratio. The β coefficient of fiscal consolidation stays strongly negative and statistically significant at the 1 percent level. This suggests that the initial level of public debt does not unduly affect the multiplier estimates for the euro area found in this paper. We also control for the sovereign credit default swap (CDS) spread in the first quarter of 2011, as it can be argued that CDS spreads take potential future debt problems as perceived by financial market actors into account.15 Again, the baseline results do not change much. We then control for the initial bank CDS spread in the first quarter of 2011 in order to check whether the OLS results are picking up the effects of stress in the financial sector.16 It has also been argued that the build-up of current-account imbalances before the crisis has negatively impacted on the economic performance in countries that accumulated considerable current-account deficits. Sustained losses in competitiveness and the associated build-up of indebtedness are claimed to have contributed to the weak growth performance during the euro crisis, after capital inflows 15. Data refer to average 5-year sovereign CDS spreads; they were obtained from the companion data set to Blanchard and Leigh (2013). 16. Data refer to average 5-year bank CDS spreads; they were, again, obtained from Blanchard and Leigh (2013). © 2017 The Author



454


to deficit countries had abruptly stopped (for example, European Commission 2012a). To investigate the role of external imbalances, which might have triggered both fiscal consolidation and headwinds to economic growth, we control for the precrisis (2007) current-account-deficit-to-GDP ratio and again find that the link between GDP growth and fiscal consolidation is robust. Results are also similar when we control for the precrisis (2007) stock of net foreign liabilities.17 Finally, we control for the role of household debt. We do so because there are legitimate concerns that large household debt overhangs during a crisis have negative effects on GDP growth (for example, Keen 2013; Koo 2013; Mian et al. 2013), which could also have impacted on the relationship between fiscal consolidation measures and economic performance. Therefore, we re-estimate the baseline equation while controlling for the precrisis (2007) level of the household debt-to-disposableincome ratio.18 We again find that our estimate of the fiscal consolidation coefficient remains largely unchanged. In a nutshell, the robustness analysis suggests that the β coefficient of fiscal consolidation is neither unduly affected by the role of outliers nor by additional variables that might have both influenced cumulative real GDP growth and fiscal consolidation intensity over the time period studied.19 What’s more, results from variations in the country group support the hypothesis that conditions for multipliers higher than 1.0 in the euro area’s economies mattered. Hence, the robustness checks support the finding that multipliers in the euro area were, on average, substantially higher than 1.0 during 2011–2013. 6

CONCLUSIONS

This paper has investigated the short-run effects of fiscal consolidation measures on economic activity in the euro area, with particular focus on the years 2011–2013. The econometric evidence on the link between cumulative real GDP growth and fiscal consolidation measures points to a strong negative association, as the depth of the economic crisis over 2011–2013 in the euro area’s economies is closely related to the harshness of fiscal austerity. This finding is in line with previous studies from the recent empirical literature, which report that fiscal adjustments are typically contractionary, and strongly so in a slump (Batini et al. 2012; Zezza 2012; De Cos and Moral-Benito 2013; Guajardo et al. 2014; Qazizada and Stockhammer 2015; Yang et al. 2015; Jorda and Taylor 2016; Stockhammer et al. 2016). The evidence we find also supports our hypothesis that one has to expect highly contractionary effects of fiscal consolidation on GDP growth when major conditions for multipliers higher than 1.0 – related to considerable economic slack and constraints in monetary policy effectiveness – are met. 17. Data for stock of net foreign liabilities (in percentage of nominal GDP) are from the updated and extended version of the data set constructed by Lane and Milesi-Ferretti (2007). 18. Data on household debt-to-disposable-income ratios are from the OECD (Household accounts, downloaded on 17 May 2015). Due to data constraints, we could only include twelve euro-area countries: Belgium, Germany, Ireland, Greece, Spain, France, Italy, the Netherlands, Austria, Portugal, Slovenia, and Finland. 19. Results for the same robustness checks in terms of including addition control variables, but based on data from European Commission (2015), support this finding. Results are available on request from the author. © 2017 The Author




455

Cumulative multiplier estimates for the euro area during 2011–2013 are found to vary in a range from 1.4 to 2.1, depending on which data source one uses to measure the extent of fiscal austerity. Based on these multiplier values, the paper calculates that an approximation of the size of output losses from fiscal consolidation in the euro area over the time period studied is in the range of 5.5 to 8.4 percent of GDP. It is therefore reasonable to state that – against the background of the prevailing institutional and macroeconomic circumstances – the cause of the double-dip recession in the euro area, which started after the third quarter of 2011, is fiscal austerity. Critics might argue that some GDP loss from fiscal austerity was inevitable in the euro area, as fiscal deficits in stressed euro-area countries had to be reduced. However, this argument downplays the importance of the austerity measures’ timing and speed, which were crucial because circumstances in the euro area were very unfavorable over the time period studied, considering that the economic recovery was anything but complete and that policy options for offsetting the contractionary effects of fiscal austerity were severely constrained. Fiscal consolidation measures aggravated macroeconomic troubles via the demand side and triggered a debt-deflationary spiral, characterized by very low inflation, rising real debt burdens and further increases in public debt-to-GDP ratios (for example, Mastromatteo and Rossi 2015) – especially in the euro area’s periphery countries. Front-loading fiscal austerity in the euro area has proven to be self-defeating. REFERENCES Alesina, A., O. Barbiero, C. Favero, F. Giavazzi, and M. Paradisi (2015), ‘Austerity in 2009– 2013,’ Economic Policy, 30(83), 385–437. Arestis, P. (2012), ‘Fiscal policy: a strong macroeconomic role,’ Review of Keynesian Economics, Inaugural Issue, 93–108. Barrell, R., D. Holland, and I. Hurst (2012), ‘Fiscal multipliers and prospects for consolidation,’ OECD Journal: Economic Studies, 2012(1), 71–102. Batini, N., G. Callegari, and G. Melina (2012), ‘Successful austerity in the United States, Europe and Japan,’ IMF Working Papers 12/190. Batini, N., L. Eyraud, L. Forni, and A. Weber (2014), ‘Fiscal multipliers: size, determinants, and use in macroeconomic projections,’ IMF Technical Notes and Manuals, September. Blanchard, O. and D. Leigh (2013), ‘Growth forecast errors and fiscal multipliers,’ IMF Working Papers 13/1. Blanchard, O. and R. Perotti (2002), ‘An empirical characterization of the dynamic effects of changes in government spending and taxes on output,’ The Quarterly Journal of Economics, 117(4), 1329–1368. Carnot, N. and F. de Castro (2015), ‘The discretionary fiscal effort: an assessment of fiscal policy and its output effect,’ European Economy – Economic Papers 543. CEPR (2015), ‘Euro area business cycle dating committee: Euro area out of recession, in unusually weak expansion,’ Center for Economic Policy Research Publication, 1 October, available at: http://cepr.org/Data/Dating/Dating-Committee-Findings-01-Oct-2015.pdf? _ga=1.193623789.1555867426.1461338148 (accessed 22 April 2016). Charles, S., T. Dallery, and J. Marie (2015), ‘Why the Keynesian multiplier increases during hard times: a theoretical explanation based on rentiers’ saving behaviour,’ Metroeconomica, 66(3), 451–473. Christiano, L., M. Eichenbaum, and S. Rebelo (2011), ‘When is the government spending multiplier large?’ Journal of Political Economy, 119(1), 78–121. Coeure, B. (2012), ‘Central banks and the challenges of the zero lower bound,’ Intervention by Benoit Coeure, Member of the Executive Board of the ECB, at the ‘Meeting on the Financial

© 2017 The Author



456


Crisis,’ hosted by the Initiative on Global Markets, at the University of Chicago Booth School of Business, Miami, 19 February, available at: https://www.ecb.europa.eu/press/ key/date/2012/html/sp120219.en.html (accessed 22 April 2016). Corsetti, G., G. Müller, and A. Meier (2012), ‘What determines government spending multipliers?’ IMF Working Papers 12/150. De Cos, P. and E. Moral-Benito (2013), ‘Fiscal consolidations and economic growth,’ Fiscal Studies, 34(4), 491–515. De Grauwe, P. (2012), ‘The governance of a fragile eurozone,’ Australian Economic Review, 45(3), 255–268. De Grauwe, P. and Y. Ji (2013), ‘From panic-driven austerity to symmetric macroeconomic policies in the eurozone,’ Journal of Common Market Studies, 51, 31–41. Dellepiane-Avellaneda, S. (2015), ‘The political power of economic ideas: the case of “Expansionary Fiscal Contractions”,’ The British Journal of Politics and International Relations, 17(3), 391–418. DeLong, B. and L. Summers (2012), ‘Fiscal policy in a depressed economy,’ Brookings Papers on Economic Activity, 44(1/Spring), 233–297. Eggertsson, G. and P. Krugman (2012), ‘Debt, deleveraging, and the liquidity trap: a Fisher– Minsky–Koo approach,’ The Quarterly Journal of Economics, 127(3), 1469–1513. ECB (European Central Bank) (2010), ‘Monthly bulletin June 2010,’ ECB publication, June. European Commission (2012a), ‘Alert mechanism report 2013,’ Report from the Commission to the European Parliament, the Council, the European Central Bank, the European Economic and Social Committee, the Committee of the Regions and the European Investment Bank. European Commission (2012b), ‘European economic forecast – Spring 2012,’ European Economy – 1/2012. European Commission (2015), ‘European economic forecast – Spring 2015,’ European Economy – 2/2015. Fatas, A. and I. Mihov (2009), ‘Why fiscal stimulus is likely to work,’ International Finance, 12(1), 57–73. Fazzari, S., J. Morley, and I. Panovska (2015), ‘State-dependent effects of fiscal policy,’ Studies in Nonlinear Dynamics and Econometrics, 19(3), 285–315. Fedelino, A., A. Ivanova, and M. Horton (2009), ‘Computing cyclically adjusted balances and automatic stabilizers,’ IMF Technical Notes and Manuals, November. Gainsbury, S., A. Whiffin, and R. Birkett (2011), ‘Financial pain in Europe,’ Financial Times, 17 October, available at: http://www.ft.com/intl/cms/s/0/feb598a8-f8e8-11e0-a5f700144feab49a.html (accessed 15 May 2016). Gechert, S. and A. Rannenberg (2014), ‘Are fiscal multipliers regime-dependent? A meta regression analysis,’ IMK Working Paper 139/2014. Gechert, S., A. Hallett, and A. Rannenberg (2015), ‘Fiscal multipliers in downturns and the effects of Eurozone fiscal consolidation,’ Center for Economic Policy Research Policy Insight 79. Guajardo, J., D. Leigh, and A. Pescatori (2011), ‘Expansionary austerity: new international evidence,’ IMF Working Papers 11/158. Guajardo, J., D. Leigh, and A. Pescatori (2014), ‘Expansionary austerity? International evidence,’ Journal of the European Economic Association, 12(4), 949–968. Hemming, R., S. Mahfouz, and M. Kell (2002), ‘The effectiveness of fiscal policy in stimulating economic activity: a review of the literature,’ IMF Working Papers 02/208. Herndon, T., M. Ash, and R. Pollin (2014), ‘Does high public debt consistently stifle economic growth? A critique of Reinhart and Rogoff,’ Cambridge Journal of Economics, 38(2), 257–279. Holland, D. and J. Portes (2012), ‘Self-defeating austerity?’ National Institute Economic Review, 222, F4–F10. Ilzetzki, E., E. Mendoza, and C. Vegh (2013), ‘How big (small?) are fiscal multipliers?’ Journal of Monetary Economics, 60(2), 239–254. IMF (International Monetary Fund) (2011), ‘Tensions from the two-speed recovery: unemployment, commodities, and capital flows,’ World Economic Outlook, April. © 2017 The Author




457

IMF (International Monetary Fund) (2015), ‘Uneven growth: short- and long-term factors,’ World Economic Outlook, April. In ’t Veld, J. (2013), ‘Fiscal consolidations and spillovers in the Euro area periphery and core,’ European Economy – Economic Papers 506. Jorda, O. and A. Taylor (2016), ‘The time for austerity: estimating the average treatment effect of fiscal policy,’ The Economic Journal, 126(February), 219–255. Keen, S. (2013), ‘A monetary Minsky model of the Great Moderation and the Great Recession,’ Journal of Economic Behavior & Organization, 86(February), 221–235. Klär, E. (2013), ‘Potential economic variables and actual economic policies in Europe,’ Intereconomics, 48(1), 33–40. Koo, R. (2013), ‘Balance sheet recession as the “other-half” of macroeconomics,’ European Journal of Economics and Economic Policies: Intervention, 10(2), 136–157. Koo, R. (2015), The Escape From Balance Sheet Recession and the QE Trap, Singapore: John Wiley. Lane, P. and G. Milesi-Ferretti (2007), ‘The external wealth of nations mark II: revised and extended estimates of foreign assets and liabilities, 1970–2004,’ Journal of International Economics, 73(2), 223–250. Mastromatteo, G. and S. Rossi (2015), ‘The economics of deflation in the euro area: a critique of fiscal austerity,’ Review of Keynesian Economics, 3(3), 336–350. Mian, A., K. Rao, and A. Sufi (2013), ‘Household balance sheets, consumption, and the economic slump,’ The Quarterly Journal of Economics, 128(4), 1687–1726. Mourre, G., C. Astarita, and S. Princen (2014), ‘Adjusting the budget balance for the business cycle: the EU methodology,’ European Economy – Economic Papers 536. OECD (2012), ‘Restoring public finances, 2012 update,’ OECD Publication, November. Palumbo, A. (2015), ‘Studying growth in the modern classical approach: theoretical and empirical implications for the analysis of potential output,’ Review of Political Economy, 27(3), 282–307. Qazizada, W. and E. Stockhammer (2015), ‘Government spending multipliers in contraction and expansion,’ International Review of Applied Economics, 29(2), 238–258. Ramey, V. (2011), ‘Can government purchases stimulate the economy?’ Journal of Economic Literature, 49(3), 673–685. Rannenberg, A., C. Schoder, and J. Strasky (2015), ‘The macroeconomic effects of the Euro Area’s fiscal consolidation 2011–2013: a simulation-based approach,’ Central Bank of Ireland Research Technical Paper 03/RT/2015. Reinhart, C. and K. Rogoff (2010), ‘Growth in a time of debt,’ NBER Working Papers 15639. Romer, C. and D. Romer (2010), ‘The macroeconomic effects of tax changes: estimates based on a new measure of fiscal shocks,’ American Economic Review, 100(3), 763–801. Stockhammer, E., W. Qazizada, and S. Gechert (2016), ‘Demand effects of fiscal policy since 2008,’ Kingston University London Economics Discussion Papers 2016-8. Woodford, M. (2011), ‘Simple analytics of the government spending multiplier,’ American Economic Journal: Macroeconomics, 3(January), 1–35. Yang, W., J. Fidrmuc, and S. Ghosh (2015), ‘Macroeconomic effects of fiscal adjustment: a tale of two approaches,’ Journal of International Money and Finance, 57(October), 31–60. Zezza, G. (2012), ‘The impact of fiscal austerity in the Eurozone,’ Review of Keynesian Economics, Inaugural Issue, 37–54.

© 2017 The Author



458


APPENDIX 1 Table A1 Data on the intensity of fiscal consolidation measures in the euro area’s economies (2011–2013) Source

Data Austria Belgium Cyprus Estonia Finland France Germany Greece Ireland Italy Latvia Luxembourg Malta Netherlands Portugal Slovakia Slovenia Spain

IMF (2015)

European Commission (2015)



OECD (2012)

Gainsbury et al. (2011)a

SBC

SBC

PSBC

DFM

FCM1

FCM2

1.995 1.074 2.911 –1.084 0.649 2.872 2.652 14.274 4.897 3.387 1.508 1.68 1.05 4.464 6.098 5.067 2.41 4.784

1.885 1.046 2.911 –1.084 0.302 2.603 2.810 12.008 3.949 2.436 1.000 2.148 1.521 3.168 5.891 5.776 2.420 4.851

1.471 0.739 3.949 –1.078 0.237 2.483 2.357 10.145 5.351 2.992 0.735 2.176 1.335 2.924 7.873 6.043 3.326 6.237

3.293 3.528 8.037 –1.534 1.874 4.854 0.321 20.597 5.547 5.651 2.293 1.398 1.979 3.289 12.029 3.794 6.170 7.450

2.6 3.4 – –3.3 1.2 3.2 2.5 7.5 1.8 5.9 – 2.6 – 1.7 9.9 4.8 2.7 4.4

– – – – – – 0.4 11.1 3.8 1.8 – – – – 5.0 – – 3.1

Note: More details on the data sources can be obtained in table form from Table 2 in this paper. SBC = structural budget balance (in % of potential output), change 2011–2013. PSBC = primary structural budget balance (in % of potential output), change 2011–2013. DFM = discretionary fiscal measures 2011–2013 (in % of nominal GDP). FCM1 = fiscal consolidation measures 2011–2013 (in % of nominal GDP). FCM2 = fiscal consolidation measures 2011 (in % of GDP per head). Regarding the calculated changes in structural budget balance data, a positive sign is interpreted as fiscal tightening and a negative sign signals fiscal loosening. a. Note that the Gainsbury et al. (2011) data are not for 2011–2013, but for 2011 only.

© 2017 The Author



REVIEW OF INTERNATIONAL POLITICAL ECONOMY, 2017 VOL. 24, NO. 5, 904–928 https://doi.org/10.1080/09692290.2017.1363797

The performativity of potential output: pro-cyclicality and path dependency in coordinating European fiscal policies Philipp Heimbergera† and Jakob Kapellera,b Downloaded by [WU Vienna University Library] at 01:17 11 September 2017

a

Institute for Comprehensive Analysis of the Economy, Johannes Kepler University, Linz, Austria; Department of Economics, Johannes Kepler University, Linz, Austria

b

ABSTRACT

This paper analyzes the performative impact of the European Commission’s model for estimating ‘potential output’, which is used as a yardstick for measuring the ‘structural budget balance’ of EU countries and, hence, is crucial for coordinating European fiscal policies. In pre-crisis years, potential output estimates promoted the build-up of private debt, housing bubbles and macroeconomic imbalances. After the financial crisis, these model estimates were revised downwards, which increased fiscal consolidation pressures. By focusing on the euro area’s economies during 1999–2014, we show how the model’s estimates influence actual economic outcomes. We identify two major economic impacts of the potential output model. First, the political implications of the model led to pro-cyclical feedback loops, reinforcing prevailing economic developments. Second, the model has contributed to national lock-ins on path dependent debt trajectories, fueling ‘structural polarization’ between core and periphery countries.

KEYWORDS Performativity; potential output; path dependency; Eurozone crisis; fiscal policy; austerity

1. Introduction This paper poses the question about the impact of macroeconomic theory and modeling on economic development in Europe in the run-up to and in the aftermath of the global financial crisis 2008/2009. In doing so, we take the claim that economic models ‘do not merely record a reality […] but contribute powerfully to shaping, simply by measuring, the reality’ (Callon 1998, 23) as our main vantage point. While the performativity of economic models (Callon 1998; MacKenzie 2003, 2006) has been studied extensively in microeconomic contexts, especially in financial markets (e.g. Beunza and Stark 2004; Lockwood 2015; MacKenzie 2005; MacKenzie and Millo 2003; MacKenzie 2011; M€ ugge 2009; Paudyn 2013; Svetlova 2012), the scholarly literature has so far largely remained silent on the performative impact of macroeconomic models on overall economic performance (Braun 2014, 2015). This gap in the literature is remarkable, given that the role of macroeconomic theory and models has been the subject of intense

CONTACT Philipp Heimberger [email protected] † Present address: Vienna Institute for International Economic Studies, Vienna, Austria. © 2017 Informa UK Limited, trading as Taylor & Francis Group

43

Downloaded by [WU Vienna University Library] at 01:17 11 September 2017

REVIEW OF INTERNATIONAL POLITICAL ECONOMY

905

academic debate, especially when it comes to explaining the global financial crisis (e.g. Cochrane 2009; Colander et al. 2009) and economic policies in Europe in the crisis aftermath (e.g. Blyth 2013; Truger 2013). Similarly, the impact of economic thought on politics in general is a classic theme in the political economy literature (e.g. Hall 1989; Mirowski and Plehwe 2009; Skidelsky 2003). In this paper, we aim to address this research gap by analyzing the performative effect of the European Commission’s model for estimating ‘potential output’ with a particular focus on the Eurozone over the period 1999–2014. In the EU’s fiscal regulation framework, estimates of ‘potential output’ play a crucial role. Potential output is defined as the (unobservable) level of output in an economy at which all production factors are employed at ‘non-inflationary levels’ (Havik et al. 2014). Estimates of potential output are derived from the European Commission’s ‘potential output model’ (Planas and Rossi 2015) – henceforth: the ‘PO-model’ – which builds on a standard production function approach (Havik et al. 2014, 9). The model estimates have important implications for the scope of fiscal policy as they are used to calculate ‘structural budget balances’, which in turn translate into country-specific fiscal policy restrictions (e.g. Kl€ar 2013; Tereanu et al. 2014). More specifically, the Commission uses its in-house model for constructing estimates of the ‘output gap’ – the difference between actual GDP and unobservable potential output – as an indicator for the cyclical position of an economy. When the output gap is positive, an economy is said to be over-heated, while a negative output gap is signaling an underutilization of economic resources. The Commission’s estimate of the output gap is translated into a judgment on how much of the fiscal deficit (or surplus) in a particular country is ‘structural’ in the sense that it cannot be attributed to variations in the business cycle. In other words, the estimates for potential output and the corresponding output gap estimates have an impact on the fiscal policy scope of member states. Since 2005, the structural balance serves as an important control indicator for fiscal conduct as implied by the Stability and Growth Pact as well as the Fiscal Compact (ECFIN 2013; Fiscal Compact 2012). Due to this institutionalization of the structural balance in the EU’s fiscal regulation framework, an increase in the structural deficit amplifies the pressure to implement fiscal consolidation, while a decrease in the structural deficit (or an increase in the structural surplus) reduces the urgency for fiscal adjustment. Against this backdrop, we analyze the PO-model not primarily as a scientific device that allows economists to assess the position of an economy in the business cycle, but rather as a conceptual foundation for an authoritative political practice that structures the room for fiscal policy maneuvering in EU countries. In doing so, we find that the model exerts a performative effect: it imposes estimates on the economies under study that are self-reinforcing both in boom and crisis times and move the economy closer towards the world represented by the PO-model. This paper is structured as follows. In Section 2, we provide a framework for summarizing economic developments in the Eurozone since the introduction of the Euro to provide an adequate background for understanding the performativity of the PO-model, which is more thoroughly introduced in Section 3. While Section 4 illustrates the performative effects of the PO-model and the mechanisms underlying the model’s influence on actual economic outcomes, Section 5 empirically illustrates the pro-cyclicality of the Commission’s model estimates and discusses their impact on macroeconomic developments. Section 6 provides an empirical analysis on the role of 44

906

P. HEIMBERGER AND J. KAPELLER

the PO-model in shaping private and public sector debt trajectories in Europe. Section 7 concludes our argument.


2. A tool and its context: income inequality, debt, current account imbalances and the Eurozone crisis In order to understand how the PO-model influenced European economic developments in recent years, we introduce a simple framework for summarizing the build-up of imbalances and fragilities before the global financial crisis of 2008/2009 as well as their prolonging during the Eurozone crisis from 2010 onwards. In doing so, we provide information on the specific economic context in which the PO-model was applied. The PO-model plays a key role in understanding the European policy-response to the crisis. While we will subsequently demonstrate how the Commission’s model estimates have influenced fiscal policy-making, we start by providing an overview on macroeconomic developments in Europe to illuminate the specific historical and economic context in which the model was applied. For illustration purposes, we focus on four stylized

Figure 1. Distribution, debt and housing prices in five countries.

Data: Wage shares were obtained from AMECO (5 November 2015), data on income shares from the World Wealth and Income Database (29 March 2015), data on private sector debt from OECD.Stat (2 December 2015) and real house prices from the Dallas Fed (International House Price Database, 2015:Q3; no data on France).

45



907

empirical facts, depicted in Figure 1 for five selected countries (France, Germany, Ireland, Italy and Spain). First, wage shares have shown a falling trend across the Eurozone from the early 1980s to the financial crisis. Second, income inequality has increased markedly over the same time period. Third, private sector debt has witnessed a significant rise in many Eurozone countries after the introduction of the Euro, with the remarkable exception being Germany, where private sector debt to GDP has been falling since the turn of the millennium. Finally, real house prices have risen in many, but not all Eurozone countries, with some economies experiencing sharp increases in house prices before the financial crisis and corresponding declines in more recent years. In order to set the stage for analyzing the role of the PO-model in European fiscal policy-making, our framework relates these stylized facts to each other in a coherent form: first, an increase in inequality in conjunction with the deregulation of financial markets drives pre-crisis economic developments in the Eurozone. The underlying mechanics are twofold: (1) falling and more unequally distributed wage shares put downward pressure on domestic demand (e.g. Cynamon and Fazzari 2016; Stockhammer 2015), and (2) increasing inequality and financial market deregulation both contributed to increasing private sector indebtedness and rising asset prices before the financial crisis (e.g. Perugini et al. 2016; Stockhammer and Wildauer 2016; Storm and Naastepad 2016). Second, in the course of the financial crisis the resulting imbalances between creditor and debtor countries triggered instability, which lead to further declines in demand and an acute recession in many Eurozone countries. The link between income concentration and private sector debt is typically explained by two complementary mechanisms. First, households confronted with stagnant or declining incomes may try to preserve their living standards – either out of mere necessity or to keep up with socially mediated consumption standards –, which increases the propensity to incur debt (Kapeller and Sch€ utz 2014). Second, rising asset prices increase the wealth but not the liquid means of those households that hold at least some assets (Stockhammer and Wildauer 2016). In this context, household debt can serve as a means to transforming the increase in asset prices into actual liquidity. In the shortterm, such an increase in household debt may compensate the impact of wage losses on aggregate demand, ‘thus providing the solution to the contradiction between the necessity of high and rising consumption levels, for the growth of the system’s actual output, and a framework of antagonistic conditions of distribution which keeps within limits the real income of the vast majority of society’ (Barba and Pivetti 2009, 113). In turn, financial innovations and the liberalization of capital flows ensure that the increase in households’ demand for credit is met by sufficient credit supply. Indeed, the emerging fragilities of individual Eurozone countries during the pre-crisis years were fueled by credit-led economic growth and large capital flows from Eurozone ‘core countries’ like Germany, France and the Netherlands to ‘periphery countries’ such as Spain, Ireland, Portugal and Greece (e.g. Baldwin et al. 2015; Hobza and Zeugner 2014; Storm and Naastepad 2016).1 As a consequence, capital inflows and private sector credit expansion triggered the build-up of major bubbles in housing markets, especially in Spain and Ireland (e.g. Drudy and Collins 2011; Ruiz et al. 2016). The periphery countries in general accumulated large current account deficits before the financial crisis; as the Eurozone’s pre-crisis current account balance was close to zero, other Eurozone countries – such as Germany, Austria and the Netherlands – had to run correspondingly large current account surpluses (e.g. Tressel et al. 2014). Hence, the financial crisis served as a trigger for the unfolding of the Eurozone crisis, which finds its essential roots in the accumulation of severe macroeconomic imbalances 46


908


between creditor and debtor countries (e.g. Stockhammer and Sotiropoulos 2014). Political constraints on fiscal and monetary sovereignty have, in turn, contributed to the inability of European countries to adequately react to the economic breakdown. While the loss of monetary sovereignty – due to the introduction of a common monetary policy for all euro area countries – is widely acknowledged (e.g. de Grauwe 2012), our paper highlights another, equally important source for the persistence of the current crisis: the profound impact of European constraints and regulations on the fiscal policy space of member countries, where assessments of the existing fiscal policy scope depend on estimates of ‘potential output’ as derived from the Commission’s PO-model. Specifically, the importance of the Commission’s PO-model derives from the fact that there are three general possibilities to compensate for a stagnation of or a decline in effective demand. First, an economy may compensate the downward pressure on domestic demand by expanding its exports. This is what has happened in Germany (e.g. Storm and Naastepad 2015b) and – to a lesser extent – in other surplus countries like Austria and the Netherlands. Second, the decrease in demand may be compensated by expansionary fiscal policy and a corresponding increase in public debt as in Greece and Portugal during pre-crisis years (e.g. Lane 2012) – a path now largely blocked by European economic governance in conjunction with the PO-model. Third, the economies concerned may develop a growth model that is driven by private sector debt accumulation. Since the adoption of the Euro, debt-led growth regimes have especially characterized large parts of the Eurozone’s periphery countries (e.g. Stockhammer and Wildauer 2016). Hence, although the Euro had raised high hopes for economic convergence (e.g. Blanchard and Giavazzi 2002), economic reality was characterized by the build-up of large macroeconomic imbalances, which did not reflect a healthy ‘catch-up process’ in the poorer countries, but rather the emergence of an unsustainable mix of debt-led and export-led growth regimes across Eurozone countries. As long as a seemingly benign financial environment masked the fragilities corresponding to the accumulation of private debt and the rising dependencies regarding the financing of current account deficits, the emergence of macroeconomic imbalances stimulated the real economy in large parts of the Eurozone. The outbreak of the financial crisis, however, revealed these fragilities. Deficit countries suffered a ‘sudden stop’ in capital inflows, followed by massive capital outflows, implying that large current account deficits had to be reduced (e.g. Giavazzi and Spaventa 2010). This reversal in capital flows triggered a fall in economic growth and increases in unemployment in the deficit countries, as over-indebted private sector actors were forced to deleverage by cutting back on their spending. Public debt and fiscal deficits rose sharply, while the public sector in many Eurozone countries socialized private sector and financial sector debt in order to avoid a breakdown of the financial system. During this process, Greece, Portugal, Ireland and Cyprus were forced to apply for being bailed out by the Troika, consisting of the European Commission, the ECB and the IMF. Financial assistance was granted on the condition that stressed countries implement drastic cuts in government spending and wages (Sapir et al. 2014). After the start of the financial crisis, downward revisions of the Commission’s model estimates of potential output have systematically affected the policy space of Eurozone countries, because these revisions implied that ‘structural fiscal deficits’ were estimated to be substantial, with the most significant restrictions coming for the countries hit hardest by the crisis (this argument will be made in detail in the subsequent sections of this paper.). In the countries that were forced to implement the harshest fiscal austerity 47



909

measures, demand was squeezed the most (e.g. de Grauwe and Ji 2013; Heimberger 2017), which improved the current account of these countries due to a sudden decrease in imports. The process of unwinding pre-crisis imbalances and reducing the private sector debt overhang thereby had a strong negative effect on the real economy (Koo 2015). This observation is consistent with historical evidence that debt deleveraging weighs heavily on aggregate demand, implying sluggish recoveries (Jorda et al. 2016) – with the most pronounced impact on countries that had previously accumulated the largest current account deficits. In this section, we have sketched the historical and economic context, in which the Commission’s PO-model has been applied. The purpose in the remaining sections is to analyze how the model has been employed as part of an authoritative political practice that – through its institutionalization in the EU’s fiscal regulation framework – has helped to structure and shape the developments described in this section.

3. The European Commission’s potential output model and its use in European fiscal policy-making In what follows, we open the ‘black box’ of the Commission’s PO-model, where ‘black box’ is understood as a device that is opaque to outsiders, because its content is regarded to be overly technical (MacKenzie 2005). The ‘unpacking’ of the model will foster our understanding about how it impacts the economic developments discussed in Section 2. In essence, the PO-model is used for judging which part of the fiscal balance in EU countries is ‘structural’, i.e. related to the true capabilities of an economy and driven neither by the business cycle nor by temporary one-off effects such as costs related to averting the break-down of the financial system (Havik et al. 2014; Mourre et al. 2014). The Commission employs a standard production function approach based on a Cobb–Douglas production function (Cobb and Douglas 1928). Estimates of the ‘structural balance’ directly depend on the Commission’s measure of the output gap (OGt ), which is derived from the PO-model and enters the corresponding formula for calculating the ‘structural balance’ as sketched in Figure 2 (Mourre et al. 2014, 9). The PO-model is the Commission’s preferred operational surveillance tool for evaluating fiscal policies as it supplies measures of potential output (POt ), which are translated into estimates of the structural balance (SBt ) by calculating the relative difference between actual output and potential output, called the ‘output gap’ (OGt ). The institutional importance of these estimates derives from the EU’s fiscal regulation framework as the Stability and Growth Pact defines countries’ medium-term budgetary objectives (MTOs) in terms of the structural balance. In case of a deviation from the MTO, a country has to correct ‘excessive structural deficits’ by improving the structural balance

Figure 2. Estimating structural budget balances.

48

910



Figure 3. Estimating potential output by employing a Cobb–Douglas production function.

by 0.5 percentage points of GDP per year (ECFIN 2013). Moreover, the Fiscal Compact makes reference to estimates of the structural deficit, since governments are legally obliged to ensure that the structural deficit does not exceed 0.5% of GDP per year – a rule which signatory states had to codify into national law, preferably as a constitutional safeguard (Fiscal Compact 2012). Hence, larger structural deficits amplify the pressure to implement fiscal consolidation measures; and vice versa. The Commission defines potential output as the level of output at which inflation remains stable. As can be seen from equation (1) in Figure 3, the PO-model employs a Cobb–Douglas production function2 to obtain estimates of unobservable potential output POt (Havik et al. 2014). While measures of the capital stock (Kt) are taken as provided in the Commission’s AMECO-database,3 the production factor labor Lt is operationalized as a statistically filtered trend of total working hours (HOURSt) offered by the active labor force (POPWt ! PARTSt), which would be employed according to the Commission’s estimates of ‘natural unemployment’ (1 – NAIRU). This argument is captured by equation (2) in Figure 3. The final ingredient of the underlying production is total factor productivity (TFPt ), which is first calculated as average output per hours worked, then corrected for ‘cyclical’ deviations by a statistical filter and eventually reinserted into the model as a proxy for ‘technological progress’. Hence, the Cobb–Douglas function primarily serves as a calculative vehicle for integrating empirical data, while the essential economic question – ‘Which components of unemployment and productivity growth are to be judged “structural” or “cyclical”?’ – is delegated to the statistical de-trending of the respective time-series on unemployment and TFP. The de-trending procedure makes use of a Kalman filter approach (Durbin and Koopman 2012; Kalman 1960), which is at the heart of the Commission’s PO-model. The basic idea behind this kind of statistical filtering is to separate the underlying time series (e.g. the unemployment rate in Spain over time) into two components: a structural component and a trend component, where the former is assumed to be determined by institutional features of an economy, while the latter captures the ups and downs of the business cycle. The structural component as estimated by the Kalman filter is interpreted as a form of ‘natural unemployment’ (Friedman 1968; Phelps 1967), namely as the ‘non-accelerating inflation rate of unemployment’ (in short: NAIRU), which is said to represent the true employment capacities underlying any given economy. The NAIRU’s central proposition is that any economy can be characterized by a clearly defined, but unobservable, rate of unemployment, at which (wage) inflation would remain stable. While in theory the NAIRU depends on the institutional characteristics of a given economy – especially labor market regulations –, in practice the Commission simply determines NAIRU values by means of statistical filtering: the ‘structural component’ obtained 49



911

from the Kalman filter is assumed to represent the NAIRU. As indicated by equation (2) in Figure 3, the NAIRU directly impacts the size of the ‘structural deficit’ as an increase in the NAIRU will cause a reduction in potential output leading to less fiscal policy scope. The intuition here is that with high NAIRU values, a smaller part of actual unemployment is considered to be cyclical, so that creating additional demand by fiscal means would subsequently lead to inflation (because actual unemployment would fall below the NAIRU). The Kalman filter is of crucial importance for the Commission’s model’s estimates, on which budgetary targets in the EU’s fiscal regulation framework are eventually based (Fioramanti 2016). Kalman filtering is a statistical technique originally developed in engineering (Kalman 1960), where its basic purpose is to allow for the refinement of noisy empirical measurements and marginally inexact theoretical predictions (e.g. in GPS-navigation or aviation). The filtering process itself is based on a recursive procedure, i.e. all estimates obtained from the underlying model – even those relating to past periods – change whenever new data (e.g. new values of the unemployment rate and wage inflation) are brought into the model (e.g. Planas and Rossi 2015). This feature explains why NAIRU estimates constantly vary: in practice, the whole time-series of NAIRU estimates and all associated forecasts change whenever new observational data is entered into the model. In the process of calculating real-time estimates, the Kalman filter furthermore assigns a crucial role to the most recent observations – a phenomenon that the statistical filtering literature calls the ‘end point bias’ (e.g. Ekinci et al. 2013; Havik et al. 2014; Kaiser and Maravall 2001). As a consequence, NAIRU estimates based on the Kalman filter exhibit a pro-cyclical bias, i.e. the NAIRU tends to fall in boom times and to increase in crisis times. Indeed, although ‘natural rate theory’ suggests that the NAIRU can be explained by only referring to the ‘institutional features’ of an economy, the Commission’s NAIRU estimates are in fact significantly determined by ‘cyclical factors’ (Heimberger et al. 2017). It follows that the Kalman filter approach as employed by the Commission leads to NAIRU estimates of structural unemployment, which exhibit an end-point bias, are influenced by cyclical factors and subjected to successive revisions.

4. The performativity of the potential output model and pro-cyclical feedback loops in Europe: an overview In this section, we develop our arguments on the performativity of the Commission’s model and describe the mechanisms that are relevant for understanding the model’s influence on actual economic outcomes. 4.1 The performative impact of the potential output model We proceed by explaining the role of the PO-model within the general economic dynamics described in Section 2. We focus on two distinct pro-cyclical feedback loops for the Eurozone during 1999–2014: first, an ‘optimist loop’ operated from the introduction of the Euro up to the financial crisis; it reinforced private debt-driven economic growth and the development of asset-price bubbles in the pre-crisis period, but also contributed to the emergence of large-scale macroeconomic imbalances. Second, a ‘pessimist loop’ has emerged in the period after the outbreak of the financial crisis, which is characterized by austerity policies, i.e. by a combination of fiscal tightening 50


912


and deflationary wage pressure geared towards increasing international competitiveness throughout the Eurozone. At its core, the performativity of the PO-model follows the same basic principle in both the boom and the bust phase: first, it provides a series of estimates for ‘natural unemployment’, i.e. the NAIRU, and ‘potential output’ to assess the ‘true position’ of a given economy within the business cycle. Second, these estimates trigger political reactions, which make actual economic outcomes sensitive to the size of and changes in these estimates. The size of NAIRU estimates matter from a static perspective: ceteris paribus, higher NAIRU-estimates imply less fiscal leeway and, hence, higher unemployment. Changes in NAIRU estimates, on the other hand, exert their influence in a dynamic context, e.g. when an increase in the NAIRU further constrains fiscal space in the face of rising unemployment. In other words, these NAIRU estimates come with political impacts that move the economy closer towards the world as represented by the Commission’s PO-model. Thereby, two mechanisms are crucial for understanding the impact of NAIRU and potential output estimates on economic policy-making. The first of these mechanisms is that the Commission’s model estimates provide allegedly exact quantitative evaluations of the ‘structural health’ of a country’s macroeconomic and fiscal developments, thereby influencing the policy-makers’ (non-)priorities in terms of policy objectives and policy measures. The second mechanism is that estimates of NAIRU and potential output affect the timing and speed of fiscal policies, which is due to their importance for assessing whether member countries meet structural deficit targets in the EU’s fiscal regulation framework. The performative effect of the PO-model contributes to positive feedback effects arising from the model’s political application and triggers a reinforcement of cyclical trends and developmental trajectories. The reinforcement of cyclical trends arises from the model’s impact on short-run fiscal policy-making as the PO-model promotes a pro-cyclical fiscal policy stance, which is most visible in the countries hit hardest by the crisis. A second instance of positive feedback is geared more towards long-term, structural developments: the pro-cyclical policies and constraints imposed by the PO-model in conjunction with the Stability and Growth Pact make it much more difficult for crisis-ridden countries to change their current path of development as they lack policy space. In other words, the PO-model reinforces the lock-in on current growth-models, which may lead to developmental traps for some countries and to prolonged prosperity for others. For the Eurozone as a whole, this constellation facilitates fragility and institutional conflicts between creditor and debtor countries. 4.2 How the potential output model influences actual economic outcomes How do the Commission’s NAIRU and potential output estimates influence actual economic outcomes? For illustration purposes, consider the example of Spain. In Autumn 2011, the Commission estimated the Spanish NAIRU for 2011 to be 16.8% as compared to an actual unemployment rate of 21.4%. In other words, the Commission’s NAIRU model estimate implied that about 80% of total unemployment was ‘natural’ or ‘structural’ and, hence, incurable by expansionary fiscal efforts. However, as explained in Section 3, the Commission’s NAIRU estimates suffer from the well-known end-point bias, which means that the most recent data points have an over-proportional impact on NAIRU estimates (Ekinci et al. 2013; Heimberger et al. 2017; Kl€ar 2013). Starting 51



913

from a NAIRU of 16.8%, the Commission calculates potential output (€1008 billion) and the output gap (¡5.0%), where the latter expresses the difference between actual GDP (€957 billion) and potential output in percent of potential output. The output gap is then used to perform the cyclical adjustment of the budget balance in accordance with the EU’s fiscal regulation framework (see Figure 2). Specifically, the Commission’s official estimates lead to a cyclically adjusted budget deficit of ¡4.2% of potential output (instead of a headline fiscal deficit of ¡6.6%), from which another 2.5% of GDP are to be subtracted to incorporate budgetary one-off effects (costs for bailing out financial institutions), resulting in a structural deficit of ¡1.7% (see Table 1). The resulting deficit thereby misses the structural deficit target of ¡0.5% within the EU’s fiscal regulation. Eventually, these outcomes put further pressure on Spain to implement fiscal consolidation measures in order to cut ‘excessive deficits’ (see Section 3). To illustrate the performative effect of the PO-model, we employ a replication of the Commission’s model to compare the path actually taken with two alternative scenarios, where we assume either that the NAIRU is 6%-points lower than the official estimate (this would correspond to a NAIRU estimate of 10.8%) or, alternatively, that the NAIRU is 6%-points higher (implying a NAIRU of 22.8%, which is higher than actual unemployment of 21.4%). In the first scenario, Table 1 indicates that potential output increases by €46.7 billion compared to the official baseline Commission estimate in Autumn 2011, which nearly leads to a doubling of the (negative) output gap and a corresponding decline of the structural deficit by 2.0 percentage points (from ¡1.7% to +0.3%). For simplicity, let us assume that this difference of 2.0 percentage points could have been used by Spain to implement expansionary fiscal policies in 2011. The current literature suggests that fiscal policy has more of an impact during crisis times. Thus, an exogenous increase in government spending by 1%-point of GDP leads to an increase in output of at least 1%-point, likely more (e.g. DeLong and Summers 2012; Heimberger 2017). As Spain’s economy was still mired in crisis back in 2011 (e.g. Koo 2015), it is thus conservative to gauge that the impact of an expansionary fiscal impulse of 2%points would have increased GDP by at least 2 percentage points, leading to a fall in unemployment. By bringing actual unemployment closer towards the assumed ‘structural’ unemployment rate of 10.8%, this example illustrates how the size of the NAIRU acts as an attractor for actual unemployment. The reverse argument also holds. In our second scenario, we assume that the NAIRU is 22.8% and, hence, higher than actual unemployment. As can be seen from Table 1, this goes along with lower estimates of potential output (declining by €47.8 billion) and the (negative) output gap (now at ¡0.3%) as well as a higher structural deficit (¡4.0%). For illustration purposes and simplicity, we assume that in 2011, Spain would Table 1. The performative impact of the Commission’s potential output model. NAIRU PO OG Spain, Year 2011 Baseline: estimate from Autumn 2011 16.8% €1007.9bn ¡5.0% First alternative: NAIRU minus 6ppts. 10.8% €1054.6bn ¡9.2% Second alternative: NAIRU plus 6ppts. 22.8% €960.1bn ¡0.3%

CAB

SB

¡4.2% ¡2.2% ¡6.5%

¡1.7% 0.3% ¡4.0%

Notes: Data: AMECO. All potential output numbers were calculated at constant prices with the base 2005 = 100. NAIRU, non-accelerating (wage) inflation rate of unemployment (in %); PO, potential output (in billion €); OG, output gap (in % of PO); CAB, cyclically adjusted budget balance (in % of PO); SB, structural balance (in % of PO). Ppts. denotes percentage points.

52


914


have had to cut the structural balance by 2.3%-points to reach the same deficit as in the baseline scenario. Under the conservative assumption that a reduction in the structural deficit by 1%-point leads to a decline in GDP by at least 1%-point (e.g. Blanchard and Leigh 2014), the negative impact of pro-cyclical fiscal austerity would have amounted to 2.3%-points of GDP, leading to an increase in actual unemployment – towards the NAIRU, which, again, acts as an attractor for actual unemployment. The analysis presented above has thus clearly illustrated how the size of NAIRU estimates contributes to shaping the fiscal policy space in Eurozone countries in a static framework of analysis. However, as has been shown, NAIRU estimates are themselves sensitive to economic trends due to their well-known end-point bias. This feature implies that the changes in NAIRU estimates also have political consequences as they constrain or widen fiscal policy space over time relative to some historical vantage point. In the post-crisis phase, the PO-model has attained a self-reinforcing inertia as it is used to justify austerity-measures, which bring forth large negative growth effects (e.g. de Grauwe and Ji 2013) and a corresponding increase in unemployment mirroring the increase in NAIRU estimates after the crisis (e.g. Kl€ar 2013). As fiscal austerity contributes to prolonging the slump, NAIRU estimates and structural deficits are pushed upwards and (additional) reductions in public spending become mandatory, which indicates the dynamic impact of the PO-model. In contrast, an initially low NAIRU estimate or successively falling estimates provide additional fiscal policy space to governments, which – if used for expansionary fiscal policies to stimulate the economy – allows for pushing down unemployment, so that the economy is again dragged in the direction of the initially low NAIRU estimate. In the following subsections, we analyze the mechanisms that are crucial for understanding the impact of NAIRU and potential output estimates on fiscal policy-making and economic developments.

5. The pro-cyclicality of NAIRU and potential output estimates: impacts on macroeconomic developments and fiscal-policy-making In this section, we analyze and empirically illustrate the positive feedback processes associated with applying the PO-model in European fiscal policy-making in more detail. 5.1 Pre-crisis years in the euro area: the ‘optimist loop’ As can be seen from Figure 4, the pre-crisis ‘optimist loop’, characterized by seemingly favorable real economic developments in large parts of the Eurozone, lead to downward revisions in ‘real-time’ NAIRU estimates, which were most pronounced in the Eurozone’s periphery countries (Kl€ar 2013; Palumbo 2015). These downward revisions in the NAIRU suggested ‘structural labor market improvements’. The resulting reaffirmation of optimistic pre-crisis beliefs about macroeconomic convergence served to justify policy inaction with respect to the build-up of private debt, housing bubbles and macroeconomic imbalances. European policy-makers and mainstream economists largely ignored these factors or interpreted them as being part of a healthy ‘catch-up process’ in the Eurozone (e.g. Blanchard and Giavazzi 2002; Giavazzi and Spaventa 2010) – an interpretation fully supported by pre-crisis estimates of the Commission’s PO-model. As potential output estimates signified an improvement in structural balances, the pre-crisis loop was also characterized by more fiscal scope. Table 2 again uses the case of Spain to empirically illustrate this point. In the run-up to the recent crisis, Spain 53



915

Figure 4. Pro-cyclical NAIRU estimates in the Eurozone.

Source: Different AMECO forecast vintages. ‘Real-time’ means that we use the NAIRU estimate from the Commission’s Spring forecast vintage of the respective year (e.g. Spring 2009 NAIRU estimate for the 2009 NAIRU), since those ‘real-time’ estimates are most relevant for shaping fiscal policy scope. Core countries consist of Belgium, Germany, France, Luxembourg, Netherlands, Austria, Finland. Periphery countries include: Ireland, Greece, Spain, Italy, Portugal.

experienced a housing boom driven by a surge in private sector debt, which led to a substantial reduction in unemployment (e.g. Ruiz et al. 2016). In Autumn 2007, the Commission’s official NAIRU estimate for the year 2006 was 8.6%, which implied a potential output of €773.6 billion, an output gap of ¡0.6% and a cyclically adjusted budget surplus of 2.1%. To illustrate the effects of NAIRU downward revisions in precrisis years, we compare the results given above with estimates obtained by using the NAIRU estimates from Spring 2005, when the EC had forecast that the Spanish NAIRU in the year 2006 would stand at 9.6% (i.e. 1 percentage point higher than the estimate published in 2007). In such a scenario, potential output would have been estimated 0.7% lower, the output gap would have turned positive (overutilization of resources instead of underutilization) and the cyclically adjusted balance would have been equal to 1.8% instead of 2.1% of potential output. In practical terms, these numbers imply that the Commission’s model estimates granted the Spanish government more fiscal policy space as the housing bubble picked up speed. Table 2. Pro-cyclical NAIRU estimates and their impact on potential output, the output gap and structural balances. NAIRU PO GDP OG CAB Pre-crisis BOOM " # " # Spain, Year 2006 (estimates from Autumn 2007) Official Commission estimates 8.6% €773.6bn €768.7bn ¡0.6% 2.1% Estimates (Spring 2005 NAIRU) 9.6% €767.8bn €768.7bn 0.1% 1.8% Notes: Data: AMECO. All potential output numbers were calculated at constant prices with the base 2000 = 100. NAIRU, non-accelerating (wage) inflation rate of unemployment (in %); PO, potential output (in billion €); GDP, real gross domestic product; OG, output gap (in % of PO); CAB, cyclically adjusted budget balance (in % of PO).

54


916


In the pre-crisis phase, Spain’s government-debt-to-GDP-ratio fell from 60.9% to 35.6% (data from 1999 to 2007). To a large extent, this development reflected the (housing) boom of the Spanish economy, which increased tax revenues and reduced unemployment-related government spending. However, some economists argue that given the debt-driven boom, Spain (and some other) countries should have leaned more strongly against the wind by pushing for restrictive fiscal policies to counteract the rise in private debt and dampen the cycle (e.g. Lane 2012). Clearly, the Commission’s favorable potential output model estimates did not support this argument in precrisis times, as downward revisions in the NAIRU suggested to policy-makers that there was even more room for fiscal maneuvering than originally envisaged. Thereby, the performative impact of the model estimates did not only contribute to a reaffirmation of optimistic assessments of pre-crisis growth trajectories and economic policies (implying a non-priority for curbing unsustainable private debt dynamics); pro-cyclical NAIRU estimates were also fed into the PO-model, thereby increasing the leeway for fiscal policy-makers due to lower ‘structural deficits’. Summing up, the ‘optimist loop’ in pre-crisis years was characterized by positive feedback effects arising from the model’s political application. As the Kalman filter integrated new observations on unemployment rates – which were falling in large parts of the periphery after the introduction of the Euro as a consequence of a (mostly private sector) debt-led upswing –, the model estimates picked up this tendency and NAIRU estimates tended to be revised downwards, which fuelled optimism and increased fiscal policy space. As the PO-model did not consider private debt and current account imbalances, it reinforced general macroeconomic developments by shaping policymakers’ assessments of the ‘structural health’ of macroeconomic and fiscal developments and by a model-induced pro-cyclical bias in fiscal policy coordination within the Eurozone. The resulting ‘optimist loop’ lasted until it was broken by the financial crisis, which triggered an unwinding of the macroeconomic imbalances accumulated in precrisis years as described in Section 2. 5.2 Post-crisis years: the ‘pessimist loop’ Positive feedback effects arising from the application of the Commission’s PO-model are even more apparent in the post-crisis period, as the Eurozone crisis had a strong impact on potential output measures in European countries (e.g. Ball 2014; Palumbo 2015). In order to empirically illustrate this point, we employ the methodology proposed in Ball (2014) and extrapolate the developments in potential output estimates before the financial crisis in 2007 (PO!!) to compare these pre-crisis trends with recent potential output estimates from November 2015 (PO!).4 From the y-axis values in Figure 5, it can be seen that losses in potential output (in the year 2014) relative to precrisis trends vary markedly across European countries, ranging from 36.3% in Greece and 24.4% in Ireland to much smaller losses in ‘core countries’ such as Germany (1.1%). The y-axis values depict how much actual output in 2014 was below the extrapolated pre-crisis trend in potential output. It can be seen that the losses in actual output and potential output are almost perfectly correlated, suggesting that the countries most affected by the crisis suffered the largest downward revisions in potential output and vice versa. Downward revisions in potential output have supported the dominant narrative that ‘excessive fiscal deficits’ are at the roots of Europe’s economic crisis (e.g. Blyth 2013; 55


917

40

Greece

30

Loss in potential output in %; year 2014


Ireland Hungary Slovenia Romania Spain

Bulgaria

20

Luxembourg Finland Cyprus Czech Republic Italy Poland Denmark Belgium

10

y = 0.91x − 0.64 T−value (HAC−robust) = 35.46 (***) R_sq = 0.98

Portugal

Austria

United Kingdom France Netherlands Sweden Germany

0

Malta 0

10

20

30

40

Loss in actual output in %; year 2014

Figure 5. The close correlation of actual and potential output losses.

Data: AMECO (December 2007, November 2015); authors’ calculations. Loss in potential output = (PO!!-PO!)/PO!!. Loss in actual output = (PO!!-Y)/PO!!. PO!!… extrapolated estimate of pre-crisis PO. See Ball (2014, 150) for details on the extrapolation methodology. PO!… EC’s PO estimate (AMECO, November 2015). Y… real GDP (AMECO, November 2015). !!!denotes statistical significance at the 1% level.

Storm and Naastepad 2016). The sudden divergence in ‘natural unemployment’ as depicted in Figure 4 was taken as evidence for a sudden ‘structural shock’ that revealed ‘labor market mismatches’ between the jobs offered and the labor supply of people looking for jobs, with mismatches supposedly affecting the structural capacities of peripheral economies (e.g. European Commission 2013). This interpretation of Figure 4 provides a clear-cut theoretical justification for the turn to fiscal austerity in 2010/2011, which was apparent in the design of the Troika adjustment programs for Greece, Ireland, Portugal and Cyprus (Sapir et al. 2014), the reform of the Stability and Growth Pact in 2011 and the introduction of the Fiscal Compact in 2012. In this process, the structural deficit has gained additional importance when it comes to coordinating fiscal policies in Europe (ECFIN 2013). Via the institutionalization of structural balances in the EU’s fiscal regulation framework, downward revisions in potential output increased fiscal consolidation pressures in Europe (Tereanu et al. 2014). Table 3 illustrates this relationship for five Eurozone periphery countries and five core countries. Negative output gaps would have been 56


918


Table 3. Downward revisions in potential output estimates increased fiscal consolidation pressures: all numbers for the year 2014. Output gap Output gap!! Cyclically adjusted balance Cyclically adjusted balance!! Periphery countries Greece ¡9.1% ¡42.1% 0.8% 16.6% Ireland ¡1.1% ¡25.2% ¡3.3% 9.5% Portugal ¡3.9% ¡12.6% ¡5.2% ¡0.7% Spain ¡6.9% ¡25.2% ¡2.2% 7.7% Italy ¡4.0% ¡15.2% ¡0.9% 5.1% Core countries Austria ¡0.9% ¡7.9% ¡2.2% 1.9% Germany ¡0.4% ¡1.4% 0.5% 1.1% France ¡1.9% ¡8.3% ¡2.8% 1.0% Netherlands ¡2.7% ¡7.1% ¡0.6% 2.2% Belgium ¡1.0% ¡8.0% ¡2.5% 1.8% Data: AMECO (December 2007, November 2015); authors’ calculations. Output gap = (Y-PO!)/PO!. Output gap!! = (Y-PO!!)/PO!!. Cyclically adjusted balance = FB – e OG. Cyclically adjusted balance!! = FB – e OG!!. FB… fiscal balance (AMECO, November 2015). e … budgetary semi-elasticity (Mourre et al. 2014, 21). See Figure 5 for details on Y, PO! and PO!!.

much larger than the Commission’s official numbers provided in November 2015 if one assumes that potential output during 2010–2014 had grown at a constant average precrisis growth rate (as in Figure 5). For example, the official output gap estimate for Spain in 2014 was ¡6.9% of potential output, which corresponded to a cyclically adjusted budget deficit of ¡2.2%. However, assuming that the potential output loss computed in Figure 5 has not occurred, we find that the output gap is ¡25.2% (OG!!), which indicates a much more severe underutilization of economic resources than the Commission’s official estimate. As a consequence, Spain would exhibit a cyclically adjusted budget surplus of 7.7% of GDP. Whoever finds the OG!! estimates in Table 3 implausibly large should take the procyclical nature of pre-crisis potential output estimates into account, which underscores our point about the optimistic nature of the pre-crisis loop described in the previous subsection. Table 3 shows that this pattern holds not only for the other periphery countries, but also for the core countries, although in a less pronounced way. Without the substantial downward revisions in potential output, which vary across European countries depending on how hard the respective country was hit by the crisis (see Figure 5), fiscal consolidation pressures would have been much less severe, because model estimates would have indicated substantial cyclically adjusted budget surpluses and, hence, would have pointed to a need for fiscal expansion. As can be seen from Figure 4, ‘real-time’ NAIRU estimates in 2014 in the Eurozone’s periphery countries had more than doubled from the pre-crisis year 2007. Meanwhile, core countries – which were affected less by the crisis than the periphery – experienced a comparably small increase in NAIRU estimates (see Figure 4). As illustrated in Section 4, higher NAIRU estimates put further fiscal consolidation pressure on the countries concerned, since they lead to higher ‘structural deficits’. It becomes clear from these illustrations on the ‘pessimist loop’ in post-crisis years that the implicit imperative of the Commission’s model during a crisis is fiscal austerity. A broad literature has shown in recent years that pro-cyclical fiscal tightening has pronounced negative effects 57



919

on economic growth and employment (e.g. Blanchard and Leigh 2014; Jord"a and Taylor 2016), which aligns well with the finding that demand was squeezed the most in those European countries that implemented the harshest fiscal austerity measures (de Grauwe and Ji 2013; Heimberger 2017). Thereby, it is no coincidence, but rather an implicit consequence of the ‘end-point bias’ in Kalman filtering that downward revisions in potential output were most pronounced where the crisis stroke hardest, which systematically subjected the Eurozone’s most fragile countries to a self-defeating cycle of austerity measures.5 In sum, we have argued that the Commission’s PO-model has produced pro-cyclical estimates in pre-crisis and post-crisis years. As these estimates influence assessments about whether Eurozone countries meet budgetary targets in the EU’s fiscal regulation framework, these pro-cyclical estimates translated into pro-cyclical policies. In the precrisis ‘optimist loop’, pro-cyclical model estimates justified policy non-action regarding the build-up of private debt, asset-price bubbles and macroeconomic imbalances and provided additional scope for fiscal policies, thereby reinforcing boom-patterns in several Eurozone countries. In the post-crisis ‘pessimist loop’, downward revisions in potential output increased the pressure to implement fiscal consolidation measures via the institutionalization of structural balances in the EU’s fiscal regulation framework. The austerity-burden caused by model-induced deteriorations in structural deficits has clearly affected those periphery countries the most, which had accumulated the largest current account deficits and debt overhangs in pre-crisis years. Against this backdrop, the next section turns to an analysis on the PO-model’s impact on structural development paths in Europe.

6. Model performativity and debt trajectories in Europe: the self-defeating nature of the Stability and Growth Pact In the previous section, we described the pro-cyclical impact arising from the application of the Commission’s NAIRU and potential output estimates as authoritative guides for designing fiscal policies across Europe. We did so by highlighting the role of these model estimates in reinforcing national growth paths. In this section, we go beyond this argument by providing a more explicit consideration of private and public sector indebtedness in the context of international competition. In Section 2, we sketched three possible ways out of economic stagnation on a national level, namely to increase aggregate demand either by rising private sector debt, expansionary fiscal policies or an increase in exports. Only the first two strategies imply a – ceteris paribus – rise in a country’s aggregate debt, while the latter strategy requires other countries to accumulate additional (foreign) debt to finance their current account deficits. Against this backdrop, we argue in this section that the NAIRU and PO-model do not only amplify cyclical fluctuations and growth paths of national economies in Europe, but also influence their overall structural development. As developmental trajectories are currently diverging, the underlying spirit of the Stability and Growth Pact – to coordinate and to harmonize economic developments across Europe – is successively undermined. At its core, our argument goes as follows. While the impact of the political application of the PO-model was rather uniform across countries in pre-crisis years – reinforcing optimism by slightly varying degrees, largely independent of a country’s specific growth model –, this tide turned rather quickly in the post-crisis period. In the aftermath of the financial crisis, countries focusing on compensating deficiencies in 58


920


domestic demand via the export side faced mainly financial risks and were continually granted comparably positive assessments of their real economic development (Storm and Naastepad 2015a, 2015b, 2015c). However, those European countries that in the pre-crisis years had relied on increases in private and public sector debt to increase demand and thereby accumulated large current account deficits, were confronted with a much more intense economic downturn and a reversal of their developmental trajectories in close correspondence with the extent of their private and public sector debt overhang (Lane 2012; Shambaugh 2012). The application of the PO-model has amplified this structural divergence between export-led creditor-countries and (overly) indebted countries by providing political and fiscal leeway to those already successful, while delegitimizing already stressed periphery countries via model-induced deteriorations in ‘natural unemployment’ (NAIRU), potential output and structural deficits. In addition, due to their importance in determining MTOs and moderating excessive deficit procedures in the EU’s fiscal regulation framework, the PO-model’s estimates serve as the technocratic component for enforcing fiscal discipline in the rise of the ‘European consolidation state’ (Chapter 4 in Streeck 2016). In exploring this argument in more depth, we first provide an empirical analysis of the development paths of individual economies in a plane constructed out of national time-series for the NAIRU and the sum of private and public sector debt in percent of GDP (Figure 6) in order to assess the intensity of structural polarization in Europe. As a first step to making sense of the information contained in Figure 6, we suggest

Figure 6. Four different patterns in country-specific trajectories on a NAIRU-debt plane.

Data: OECD (private sector debt in % of GDP), AMECO (November 2015); authors’ calculations. Total debt (y-axis) is the sum of private sector and public sector debt in % of GDP.

59



921

focusing on the developmental trajectories of individual countries. In doing so, four basic types emerge: (1) countries experiencing a rough non-linearity in their developmental path when the financial crisis hit, resembling the structure of a ‘Minsky–Veblen Cycle’ (Kapeller and Sch€ utz 2014); (2) countries, which are – either very slowly or rather rapidly – ‘losing ground’, as debt-levels and NAIRU estimates rise simultaneously; (3) countries with rising debt levels, but a decreasing NAIRU, which are ‘catching up’ to the Eurozone’s core countries; and (4) a single country – Germany – exhibiting both decreasing levels of debt and a falling NAIRU, thereby signifying Germany’s position as the powerful ‘victor’ in the European race for competitiveness (Simonazzi et al. 2013; Storm and Naastepad 2015b). While this approach supplies us with an overview on the individual countries’ developmental trajectories that are reinforced by the Commission’s model, a disadvantage of this perspective is that it hardly allows for synthesizing data and interpretation across countries. In order to remedy this fact, we provide an aggregate NAIRU-debt plane for 20 EU countries, including 15 Eurozone countries. Figure 7 is based on a simple aggregation of all time-series provided in Figure 6. Its main feature is that it separates the whole plane into grids and thereby calculates the average movement per period within the respective grid and plots these averages as arrows. This setup is inspired by the complexity economics approach developed in Cristelli et al. (2015), who argue that a vector-like representation in a plane such as ours allows for a better understanding of the complex trajectories of different countries. Dividing the NAIRU-debt plane into grids not only allows for visualizing average dynamics within a pooled set of countries, but also for assessing the relative strength of diverging moves within more densely populated areas, i.e. grids characterized by many observations: while for a certain, more extreme, range of NAIRU-debt values clear patterns emerge, in more densely populated areas the dynamics across countries tend to

Figure 7. Europe on a NAIRU-debt plane: an aggregate view.

Data: OECD, AMECO (Autumn 2015); authors’ calculations. Total debt is the sum of private sector debt and public sector debt (in % of GDP).

60


922


level each other out. Regarding the sensitivity of the results in the NAIRU-debt plane, we find that they are robust with respect to variations in the number of grids applied. In this bird’s-eye view on developmental trajectories in Europe, several areas with distinct properties emerge from Figure 7, which roughly resemble the individual trajectories depicted in Figure 6. First, there is a small group of countries (Luxembourg, Netherlands and Ireland) with a pronounced financial sector, high debt-levels and varying NAIRU estimates. Second, there is a slightly larger group of countries where the NAIRU is estimated to be high, while the burden of indebtedness is also enormous (Greece, Portugal, Spain as well as, less pronounced, France and Italy). These countries seem to have fallen into an austerity-trap from which there is no clear way out. The ongoing deleveraging in the private and/or public sector leads these countries deeper into debt-deflationary territory (e.g. Koo 2013; Mastromatteo and Rossi 2015), from which the Commission’s pro-cyclical potential output estimates make it very difficult to escape, because the model’s implicit imperative in a prolonged crisis is simply more fiscal austerity (see Section 5). Third, there is a densely populated middle area, where the trajectories of individual countries largely cancel each other out. The only exception is a small ‘path of hope’ exemplified by Poland, Slovakia and (partially) the Czech republic, signaling the possibility that an increase in debt might allow for a sustainable catch-up process – but only for those countries starting with rather low levels of total debt. While there are not too many data points underlying the pattern exhibited by the second group, consisting of countries in an austerity-trap, these observations are still of high economic and political significance as the main questions – how to bring them back into the game and how to reset their developmental trajectories – remain unanswered by current policy approaches. Quite on the contrary, the policy tools currently in place further reinforce the underlying divergence, as the PO-model provides no escape route from a debt-deflationary path that causes countries to move further into the (upper) right ‘grids of despair’ in the NAIRU-debt plane (Figure 7). These countries are caught in self-defeating debt-deflation since the European regulatory innovations introduced in the aftermath of the financial crisis place strong restrictions on their political and fiscal leeway. In this context, the countries in the upper-right part of the NAIRU-debt plane are under direct disciplinary supervision regarding their debt outlooks, where this supervision is, again, based on ‘structural deficit numbers’ derived from estimates of NAIRU and potential output as provided by the model under study. According to the prevailing rules in the Stability and Growth Pact, these countries are legally obliged to bring down ‘excessive structural deficits’ and, hence, experience limited financial autonomy, which undermines the introduction of alternative policies ensuring a more sustainable economic development (e.g. the buildup of competitive industries). A possible objection to our argument regarding the austerity-promoting features of the PO-model’s estimates in crisis times is that the Commission retains some discretionary power within the EU’s fiscal regulation framework when it comes to assessing fiscal conduct, as there are several flexibility clauses and exceptions that could temporarily override structural deficit targets (e.g. Claeys et al. 2016). For example, one could argue that a large, politically strong country such as France has more leverage to withstand the Commission’s pressure to implement corrective actions than a smaller, less powerful country such as Portugal. However, although the existing fiscal regulation framework leaves some scope for discretionary flexibility of the Commission, the 61



923

PO-model’s estimates shape the EU fiscal regulation framework’s policy objectives, allowing for allegedly exact quantitative evaluations of the ‘structural health’ of a country’s fiscal situation from which political exceptions may be granted on a discretionary basis. Against this background, it becomes clear that the Commission’s NAIRU and potential output estimates play an important role in the ‘drama of democratic states being turned into debt-collecting agencies’ (Streeck 2016, 29). In doing so, the model does not only promote polarization, but also fuels political conflicts between debtor and creditor countries in Europe – for example between Germany and France, as the former is running current account surpluses while the latter has been accumulating deficits (e.g. Simonazzi et al. 2013). In other words, it might be the case that a large country such as France is more likely than smaller countries to benefit from a flexible assessment, even if the Commission’s model estimates point to the imperative of more fiscal austerity. However, the pro-cyclical model estimates increase the consolidation pressure on the French government, as its position vis-"a-vis creditor countries with more favorable NAIRU and potential output estimates is nonetheless weakened due to the model’s restraining impact on structuring the available policy scope. While it is evident that austerity policies aiming at improvements in the structural development by increasing competitiveness can never succeed in all countries at the same time, the performative impact of the Commission’s model is not only to be found in fiscal restriction: by providing pro-cyclical downward revisions of potential output estimates as well as correspondingly higher numbers on ‘excessive structural deficits’ in times of crisis, the potential output model translates an econometric problem (the ‘end-point bias’ in calculating NAIRU and TFP by means of Kalman filtering) into political momentum. This has helped policy-makers to argue that the ‘structural health’ of the economies in the Eurozone periphery had previously been overestimated, and that drastic deflationary austerity measures would be ‘without alternative’ to increase price competitiveness and ensure public debt sustainability. Accordingly, model estimates of the NAIRU and potential output also contribute to lopsided attributions of ‘blame’ for the dire macroeconomic developments in Europe (e.g. Varoufakis 2017). The decrease in political scope in debt-burdened countries makes alternative political proposals or economic visions for European economic policy more difficult to defend.

7. Conclusions This paper shows that the Commission’s model for estimating the non-accelerating inflation rate of unemployment (NAIRU) and potential output has contributed powerfully to shaping macroeconomic developments and fiscal policy-making in Europe, with particular focus on the euro area’s economies during 1999–2014. Our arguments fill a gap in the performativity literature, which has so far mostly neglected the role of macroeconomic models in economic policy-making. We have demonstrated that the potential output model provides estimates for ‘natural unemployment’, i.e. the NAIRU, and ‘potential output’, which are interpreted as assessments of an economy’s ‘true position’ in the business cycle. Against the background of their institutionalization in the EU’s fiscal regulation framework, these model estimates then lead to political reactions, which increase the sensitivity of actual economic outcomes to the size of and changes in these model estimates. In particular, when estimates of the NAIRU go up during crisis times, a country’s scope for additional government borrowing decreases. The reason 62


924


is that higher NAIRU estimates imply that less of the actual unemployment rate is cyclical; hence, expansionary fiscal policies cannot be expected to reduce unemployment significantly before inflation kicks in. As a consequence, the Commission’s NAIRU estimates constrain the policy-maker’s scope for revitalizing the economy through extra debt-financed spending within the EU’s fiscal regulation framework. In sum, we have identified two major economic impacts that are brought forth by the performativity of the potential output model. First, the Commission’s estimates were demonstrably pro-cyclical – both in the pre-crisis years from the introduction of the Euro to the financial crisis, and in the post-crisis period. The application of these estimates for macroeconomic coordination purposes in turn reinforced general economic developments not only by affecting national fiscal policies but also by reaffirming and amplifying established views on economic conditions and appropriate policies in Europe. The second self-reinforcing feedback loop lies in the model’s amplifying effects on structural development paths in Europe. While the PO-model could not account for the increasing polarization underlying the pre-crisis period as it ignores international competition and the financial system altogether, it did indeed track the harsh reversal in the developmental paths of the Eurozone’s periphery countries – and, in doing so, amplified the crisis in those countries as the public sector was forced to deleverage simultaneously with the private sector. The Commission’s potential output model blocks any promising possibility to overcoming the resulting austerity-trap, because the massive downward revisions in potential output in those countries hit hardest by the crisis have put persistent fiscal consolidation pressure on the respective governments. Counteracting the drag on aggregate demand exerted by private sector deleveraging and overcoming the divergence in structural development trajectories in Europe within the given focus on improving competitiveness eventually requires fiscal scope for public investment to foster structural improvements and innovations (e.g. Koo 2015; Mazzucato 2013). The Commission’s model, however, has systematically failed to grant the necessary policy leeway. Hence, it has proven self-defeating in the sense that it contributes to the increase in structural divergence between the Eurozone’s core and periphery countries – a phenomenon that contradicts the spirit of convergence allegedly embedded in the Stability and Growth Pact. While the practical difficulties associated with surveilling budgetary discipline gave rise to the potential output model in the first place, today it serves as a restrictive and intransparent ‘experts cage’ for confining democratic policy-making.

Notes 1. Hence, the roots of the Eurozone crisis in the years prior to the financial crisis of 2008/2009 lie not in excessive fiscal deficits and public debt, although the crisis has created severe sovereign debt problems from 2010 onwards (e.g. de Grauwe and Ji 2014; Lane 2012; Shambaugh 2012). 2. The Cobb–Douglas framework used by the Commission is well established, although many criticisms have been put forward that challenge its theoretical foundations and empirical usage (e.g. Felipe and McCombie 2014). 3. Criticisms related to measures of the capital stock are profound but beyond the scope of this paper; see Felipe and McCombie (2014) for a recent literature review. 4. The Commission’s forecast from December 2007 provides time-series data for potential output for all EU countries through 2009 (we exclude five countries for which the 2007 data could not be compared to 2015 data). We take this pre-crisis data, denote them by PO!!, and extend all time-

63


925

series beyond 2009 by means of log-linear extrapolation. Specifically, we compute the average annual change in the logarithm of PO!! during 2000–2009, and then assume that potential output has increased at a constant rate from 2010 to 2014 (see Ball 2014, 150). 5. Partly and to varying degrees across countries, revisions in potential output were also due the collapse in the growth rate of the capital stock (e.g. Darvas 2013; Kl€ar 2013; Palumbo 2015).


Acknowledgments The Institute for New Economic Thinking supported this work under Grant INO1500015. The authors thank Bernhard Sch€ utz, Claudius Gr€abner, Stephan P€ uhringer, Georg Feigl, Stefan Steinerberger, Leonhard Dobusch and Mario Holzner as well as three anonymous referees for helpful comments. All remaining errors are our own.

Disclosure statement No potential conflict of interest was reported by the authors.

Funding The Institute for New Economic Thinking [grant number INO1500015].

Notes on contributors Philipp Heimberger works as an economist at the Vienna Institute for International Economic Studies (Vienna, Austria) and at the Institute for Comprehensive Analysis of the Economy (Johannes Kepler University Linz, Austria). Jakob Kapeller is assistant professor at the Johannes Kepler University Linz (Department of Economics) and acts as head of the Institute for Comprehensive Analysis of the Economy (Linz, Austria).

References Baldwin, R., Beck, T., Benassy-Quere, A., Blanchard, O., Corsetti, G., de Grauwe, P., den Haan, W., Giavazzi, F., Gros, D., Kalemli-Ozcan, S., Micossi, S., Papaioannou, E., Pesenti, P., Pissarides, C., Tabellini, G. and Weder di Mauro, B. (2015) ‘Rebooting the Eurozone: Step 1 – agreeing a crisis narrative’, CEPR Policy Insight 85: 1–15. Ball, L. (2014) ‘Long-term damage from the Great Recession in OECD countries’, European Journal of Economics and Economic Policies 11: 149–60. Barba, A. and Pivetti, M. (2009) ‘Rising household debt: its causes and macroeconomic implications – a long-period analysis’, Cambridge Journal of Economics 33: 113–37. Beunza, D. and Stark, D. (2004) ‘Tools of the trade: the socio-technology of arbitrage in a wall street trading room’, Industrial and Corporate Change 13: 369–400. Blanchard, O. and Giavazzi, F. (2002) ‘Current account deficits in the euro area: the end of the Feldstein-Horioka puzzle?’, Brookings Papers on Economic Activity 33: 147–210. Blanchard, O. and Leigh, D. (2014) ‘Learning about fiscal multipliers from growth forecast errors’, IMF Economic Review 62: 179–201. Blyth, M. (2013) Austerity. The History of a Dangerous Idea, Oxford: Oxford University Press. Braun, B. (2014) ‘Why models matter. The making and unmaking of governability in macroeconomic discourse’, Journal of Critical Globalisation Studies 7: 48–79. Braun, B. (2015) ‘Governing the future: the European Central Bank’s expectation management during the Great Moderation’, Economy and Society 44: 367–91. Callon, M. (1998) ‘Introduction: The embeddedness of economic markets in economics’, in M. Callon (ed) The Laws of Markets, Oxford: Blackwell Publishing, pp. 1–59.

64


926


Claeys, G., Darvas, Z. and Leandro, A. (2016) ‘A proposal to revive the European Fiscal Regulation Framework’, Bruegel Policy Contribution 7: 1–20. Cobb, C. and Douglas, P. (1928) ‘A theory of production’, American Economic Review 18: 139–65. Cochrane, J. (2009) ‘How did Paul Krugman get it so wrong?’, Economic Affairs 31: 36–40. Colander, D., Goldberg, M., Haas, A., Juselius, K., Kirman, A., Lux, T. and Sloth, B. (2009) ‘The financial crisis and the systemic failure of the economics profession’, Critical Review 21: 249–67. Cristelli, M., Tacchela, A. and Pietronero, L. (2015) ‘The heterogenous dynamics of economic complexity’, PLoS ONE 10: e0117174. Cynamon, B. and Fazzari S. (2016) ‘Inequality, the great recession and slow recovery’, Cambridge Journal of Economics 40: 373–99. Darvas, Z. (2013) ‘Mind the gap! And the way structural budget balances are calculated’, Bruegel Blog Post, 18 October; accessed at http://bruegel.org/2013/10/mind-the-gap-and-the-way-structural-bud get-balances-are-calculated/, 12 May 2016. de Grauwe, P. (2012) ‘The governance of a fragile Eurozone’, Australian Economic Review 45: 255–68. de Grauwe, P. and Ji, Y. (2013) ‘From panic-driven austerity to symmetric macroeconomic policies in the Eurozone’, Journal of Common Market Studies 51: 31–41. de Grauwe, P. and Ji, Y. (2014) ‘How much fiscal discipline in a monetary union?’, Journal of Macroeconomics 39: 348–60. DeLong, B. and Summers, L. (2012) ‘Fiscal policy in a depressed economy’, Brookings Papers on Economic Activity 44: 233–97. Drudy, P. and Collins, M. (2011) ‘Ireland: from boom to austerity’, Cambridge Journal of Regions, Economy and Society 4: 339–54. Durbin, J. and Koopman, S. J. (2012) Time Series Analysis by State Space Methods, Oxford: Oxford University Press. ECFIN (2013) ‘Building a strengthened fiscal framework in the European Union: a guide to the stability and growth pact’, European Economy Occasional Papers 150: 1–35. European Commission (2013) ‘Labour market developments in Europe 2013’, European Economy 6: 1– 133. Ekinci, F., Kabas, G. and Sunel, E. (2013) ‘End-point bias in trend-cycle decompositions: an application to the real exchange rates of Turkey’, Central Bank of Turkey Working Paper 13: 1–23. Felipe, J. and McCombie, J. (2014) ‘The aggregate production function: ‘Not Even Wrong’, Review of Political Economy 26: 60–84. Fioramanti, M. (2016) ‘Potential output, output gap, and fiscal stance: is the EC estimation of the NAWRU too sensitive to be reliable?’, MPRA Paper 73762: 1–16. Fiscal Compact (2012) Treaty on Stability, Coordination and Governance in the Economic and Monetary Union, T/SCG/EN 1, Brussels: European Commission. Friedman, M. (1968) ‘The role of monetary policy’, American Economic Review 58: 1–17. Giavazzi, F. and Spaventa, L. (2010) ‘Why the current account matters in a monetary union: lessons from the financial crisis in the Euro Area’, CEPR Discussion Paper 8008: 1–17. Hall, P. A. (ed) (1989) The Political Power of Economic Ideas: Keynesianism Across Nations, Princeton, NJ: Princeton University Press. Havik, K., Morrow, K. M., Orlandi, F., Planas, C., Raciborski, R., Roeger, W., Rossi, A., Thum-Thysen, A. and Vandermeulen, V. (2014) ‘The production function methodology for calculating potential growth rates and output gaps’, European Economy Economic Papers 535: 1–112. Heimberger, P., Kapeller, J. and Sch€ utz, B. (2017) ‘The NAIRU determinants: what's structural about unemployment in Europe?’, Journal of Policy Modeling. accessed at https://doi.org/10.1016/j.jpol mod.2017.04.003, 10 August 2017. Heimberger, P. (2017) ‘Did fiscal consolidation cause the double-dip recession in the euro area?’, Review of Keynesian Economics 5: 439–58. Hobza, A. and Zeugner, S. (2014) ‘The ‘imbalanced balance’ and its unraveling: current accounts and bilateral financial flows in the euro area’, European Economy – Economic papers 520: 1–25. Jorda, O. and Taylor, A. (2016) ‘The time for austerity: estimating the average treatment effect of fiscal policy’, Economic Journal 126: 219–55. Jorda, O., Schularick, M. and Taylor, A. (2016) ‘Sovereigns versus banks: credit, crises, and consequences’, Journal of the European Economic Association 14: 45–79. Kaiser, R. and Maravall A. (2001) ‘Some basic limitations of the Hodrick-Prescott filter, measuring business cycles in economic time series’, Lecture Notes in Statistics 154: 87–115.

65



927

Kalman, R. (1960) ‘A new approach to linear filtering and prediction problems’, Journal of Basic Engineering 82: 35–45. Kapeller, J. and Sch€ utz, B. (2014) ‘Debt, boom, bust: a theory of Minksy-Veblen cycles’, Journal of Post Keynesian Economics 36: 781–813. Kl€ar, E. (2013) ‘Potential economic variables and actual economic policies in Europe’, Intereconomics 48: 33–40. Koo, R. (2013) ‘Balance sheet recession as the “Other Half” of macroeconomics’, European Journal of Economics and Economic Policies 10: 136–57. Koo, R. (2015) The Escape From Balance Sheet Recession and the QE Trap, Singapore: John Wiley & Sons. Lane, P. (2012) ‘The European sovereign debt crisis’, Journal of Economic Perspectives 26: 49–68. Lockwood, E. (2015) ‘Predicting the unpredictable: value-at-risk, performativity, and the politics of financial uncertainty’, Review of International Political Economy 22: 719–56. MacKenzie, D. (2003) ‘An equation and its worlds: bricolage, exemplars, disunity and performativity in financial economics’, Social Studies of Science 33: 831–68. MacKenzie, D. and Millo, Y. (2003) ‘Constructing a market, performing theory: the historical sociology of a financial derivatives exchange’, American Journal of Sociology 109: 107–45. MacKenzie, D. (2005) ‘Opening the black boxes of global finance’, Review of International Political Economy 12: 555–76. MacKenzie, D. (2006) An Engine, Not a Camera. How Financial Models Shape Markets. Cambridge, MA: MIT Press. MacKenzie, D. (2011) ‘The big, bad wolf and the rational market: portfolio insurance, the 1987 crash and the performativity of economics’, Economy and Society 33: 303–34. Mastromatteo, G. and Rossi, S. (2015) ‘The economics of deflation in the euro area: a critique of fiscal austerity’. Review of Keynesian Economics 3: 336–50. Mazzucato, M. (2013) The Entrepreneurial State: Debunking Public vs. Private Myths in Risk and Innovation, London: Anthem Press. Mirowski, P. and Plehwe, D. (eds.) (2009) The Road from Mont Pelerin. The Making of the Neoliberal Thought Collective, Cambridge: Harvard University Press. Mourre, G., Astarita, C. and Princen, S. (2014) ‘Adjusting the budget balance for the business cycle: the EU methodology’, European Economy – Economic Papers 536: 1–32. M€ ugge, D. (2009) ‘Tales of tails and dogs: derivatives and financialization in contemporary capitalism’, Review of International Political Economy 16: 514–26. Palumbo, A. (2015) ‘Studying growth in the modern classical approach: theoretical and empirical implications for the analysis of potential output’, Review of Political Economy 27: 282–307. Paudyn, B. (2013) ‘Credit rating agencies and the sovereign debt crisis: performing the politics of creditworthiness through risk and uncertainty’, Review of International Political Economy 20(4): 788– 818. Perugini, C., H€ olscher, J. and Collie, S. (2016) ‘Inequality, credit and financial crises’, Cambridge Journal of Economics 40: 227–57. Phelps, E. (1967) ‘Phillips curves, expectations of inflation and optimal unemployment over time’, Economica 34: 254–81. Planas, C. and Rossi, A. (2015) ‘Program GAP. Technical description and user-manual (Version 4.4)’, Joint Research Centre Ispra Scientific and Technical Reports 4: 1–46. Ruiz, J., Stupariu, P. and Vilarino, A. (2016) ‘The crisis of Spanish savings banks’, Cambridge Journal of Economics 40: 1455–77. Sapir, A., Wolff, G., de Sousa, C. and Terzi, A. (2014) The Troika and Financial Assistance in the Euro Area: Successes and Failures, (committee study on the request of the Economic and Monetary Affairs Committee, February 2014). Brussels: European Parliament. Shambaugh, J. (2012) ‘The Euro’s three crises’, Brookings Papers on Economic Activity 43: 157–231. Simonazzi, A., Ginzburg, A. and Nocella, G. (2013) ‘Economic relations between Germany and Southern Europe’, Cambridge Journal of Economics 37: 653–75. Skidelsky, R. (2003) John Maynard Keynes 1883 –1946. Economist, Philosopher, Statesman, London: MacMillan. Stockhammer, E. and Sotiropoulos, D. (2014) ‘Rebalancing the Euro area: the costs of internal devaluation’, Review of Political Economy 26: 210–33.

66


928


Stockhammer, E. (2015) ‘Rising inequality as a cause of the present crisis’, Cambridge Journal of Economics 39: 935–58. Stockhammer, E. and Wildauer, R. (2016) ‘Debt-driven growth? Wealth, distribution and demand in OECD countries’, Cambridge Journal of Economics 40: 1609–34. Storm, S. and Naastepad, C. W. M. (2015a) ‘Europe’s hunger games: income distribution, cost competitiveness and crisis’, Cambridge Journal of Economics 39: 959–86. Storm, S. and Naastepad, C. W. M. (2015b) ‘Crisis and recovery in the German economy: the real lessons’ Structural Change and Economic Dynamics 32: 11–24. Storm, S. and Naastepad, C. W. M. (2015c) ‘NAIRU economics and the Eurozone crisis’ International Review of Applied Economics 29: 843–77. Storm, S. and Naastepad, C. W. M. (2016) ‘Myths, mix-ups and mishandlings: understanding the Eurozone Crisis’ International Journal of Political Economy 45: 46–71. Streeck, W. (2016) How Will Capitalism End?, London: Verso. Svetlova, E. (2012) ‘On the performative power of financial models’, Economy and Society 41: 418–34. Tereanu, A., Tuladhar, A. and Simeone, A. (2014) ‘Structural balance targeting and output gap uncertainty’, IMF Working Papers 14: 1–30. Tressel, T., Wang, S., Kang, J. and Shambaugh, J. (2014) ‘Adjustment in Euro area deficit countries: progress, challenges, and policies’, IMF Staff Discussion Note 14: 1–34. Truger, A. (2013) ‘Austerity in the euro area: the sad state of economic policy in Germany and the EU’, European Journal of Economic and Economic Policies 10: 158–74. Varoufakis, Y. (2017) Adults in the Room: My Battle With Europe’s Deep Establishment, London: Bodley Head.

67

Available online at www.sciencedirect.com

ScienceDirect Journal of Policy Modeling 39 (2017) 883–908

The NAIRU determinants: What’s structural about unemployment in Europe?! Philipp Heimberger a,∗ , Jakob Kapeller b , Bernhard Schütz c a

b

Vienna Institute for International Economic Studies and Institute for Comprehensive Analysis of the Economy, Johannes Kepler University, Linz, Austria Department of Philosophy and Theory of Science and Institute for Comprehensive Analysis of the Economy, Johannes Kepler University, Linz, Austria c Department of Economics and Institute for Comprehensive Analysis of the Economy, Johannes Kepler University, Linz, Austria Received 12 February 2017; received in revised form 20 March 2017; accepted 20 April 2017 Available online 3 May 2017

Abstract This paper analyzes the determinants of the European Commission’s estimates of the non-accelerating inflation rate of unemployment (NAIRU) for 14 European countries during 1985–2012. The NAIRU is a poor proxy for ‘structural unemployment’. Labor market institutions – employment protection legislation, union density, tax wedge, minimum wages – underperform in explaining the NAIRU, while cyclical variables – capital accumulation and boom-bust patterns in housing markets – play an important role. This finding is policy-relevant since the NAIRU is used to compute potential output and structural budget balances and, hence, has a direct impact on scope and evaluation of fiscal policies in Europe. © 2017 The Society for Policy Modeling. Published by Elsevier Inc. All rights reserved. JEL classification: C54; E24; E62 Keywords: Unemployment; Fiscal policy; Labor market institutions; Potential output; NAIRU

!

This work was supported by the Institute for New Economic Thinking (INET) under Grant INO1500015. Corresponding author. E-mail addresses: [email protected] (P. Heimberger), [email protected] (J. Kapeller), [email protected] (B. Schütz). ∗

http://dx.doi.org/10.1016/j.jpolmod.2017.04.003 0161-8938/© 2017 The Society for Policy Modeling. Published by Elsevier Inc. All rights reserved.

68

884

P. Heimberger et al. / Journal of Policy Modeling 39 (2017) 883–908

1. Introduction Unemployment rates across Europe have increased markedly during and in the aftermath of the global financial crisis of 2008/2009. In the Eurozone, unemployment increased from 7.6% in 2008 to 12.0% in 2013, falling to 10.9% in 2015. However, as the increase in unemployment has proven persistent in many European countries in the sense that actual unemployment rates are still way above pre-crisis levels, the debate about the factors driving the evolution of European unemployment is in full swing – both in the academic world as well as among policymakers (e.g. Arpaia, Kiss, & Turrini, 2014 ; European Central Bank, 2015). In the context of the debate about cyclical and structural determinants of unemployment across Europe, one of the crucial questions is how the ‘non-accelerating inflation rate of unemployment’ – or, in short, NAIRU – has developed in European countries before and after the crisis, as the NAIRU is frequently used as a proxy for ‘structural unemployment’, determined by labor market institutions (e.g. European Commission, 2013; Orlandi, 2012). The NAIRU is a major concept in modern macroeconomics. Its core proposition is that, for any economy and at any point in time, there exists some (unobserved) rate of unemployment at which inflation remains constant. Historically, the NAIRU can be seen as a direct offspring of the famous Phillips curve, which posits a negative relationship between unemployment and (changes in) inflation. However, over time the NAIRU has also been identified with the idea of a ‘natural rate of unemployment’ (Ball & Mankiw 2002), which would prevail in the absence of any cyclical fluctuations and, hence, represents structural unemployment existing independently of all temporary and seasonal fluctuations (Friedman, 1968; Phelps, 1967). The NAIRU concept has confronted empirical researchers with a troubling question, namely: How to produce reliable empirical estimates of a theoretically postulated but unobservable variable? In many of the past and current applications, this question is resolved by pragmatic approaches, which treat the NAIRU as an unobservable stochastic variable (e.g. Franz, 2005; Staiger, Stock, & Watson, 1997; Watson, 2014 ), employing a variety of econometric models and statistical techniques to estimate this variable. In this paper, we argue that this practice creates a certain tension between theory and empirical application: while theoretical accounts connected with the idea of the ‘natural rate of unemployment’ posit that structural factors determine the NAIRU, most actual empirical estimations of the NAIRU are devoid of such considerations, but take a comparably empiricist approach, which is either based on pure statistical technique – such as the Hodrick–Prescott filter (Hodrick & Prescott, 1997) – or relies on the integration of a Phillips curve relationship into a statistical de-trending and filtering process, as in typical Kalman Filter applications (e.g. Durbin & Koopman, 2012; Laubach, 2001). These methods are used to separate ‘trend’ and ‘cyclical’ components of unemployment without explicitly specifying the structural factors underlying ‘trend unemployment’, which is nonetheless interpreted as a suitable estimate for the NAIRU of the economies under study. From this perspective, the connection between theoretical account and empirical practice is established only implicitly – by effectively assuming that one’s de-trended series does indeed represent the structural factors driving unemployment and, hence, is a good proxy of the ‘true’ NAIRU values in the economy under study. In this paper, we aim to constructively exploit this tension between theory and empirical application by critically assessing the empirical plausibility of the essential underlying hypothesis that the evolution of the NAIRU is driven by structural factors. Specifically, we study whether theoretical arguments on structural unemployment are consistent with commonly used estimates of the NAIRU as an unobservable stochastic variable. In doing so, we assess the robustness and plausibility of these NAIRU estimates and the underlying assumption that these estimates indeed represent the unob-

69

885


servable posited by theory. Hence, we contribute to answering the question what commonly used NAIRU estimates actually tell us about the state of an economy. In operationalizing this aim, we focus on a specific case, namely the non-accelerating wage inflation rate of unemployment (NAWRU) as estimated by the European Commission (EC). This case is of major interest due to its high policy relevance as the NAWRU is used as a proxy for structural unemployment in calculating cyclically-adjusted budget balances (Havik et al., 2014 ), which are especially crucial for coordinating fiscal policy across euro area member states and for determining fiscal adjustment paths (e.g. ECFIN, 2013 ). With high ‘structural unemployment’, the ‘structural component’ of the fiscal deficit is estimated to be large. Hence, high NAWRU estimates increase the pressure on EU countries to implement fiscal consolidation measures, because essential fiscal targets in the EU’s fiscal regulation framework are set in terms of the structural budget balance. Although it might seem to be a rather obvious strategy to econometrically compare actual NAIRU estimates with their supposed theoretical determinants, most current and past research focuses on the empirics of actually observed unemployment (e.g. Blanchard, 2006 ; Baccaro & Rei, 2007 ; Nickell, 1997 ; Stockhammer & Klär, 2011). According to our best knowledge, there have only been three attempts so far to look at the empirical determinants of Kalman-filtered NAIRU estimates in a larger group of EU countries, where two of these studies were conducted by EC economists themselves (European Commission, 2013 ; Orlandi, 2012) and one by researchers at the OECD (Gianella, Koske, Rusticelli, & Chatal, 2008). However, our econometric analysis goes beyond this literature in various respects by including additional control variables, by considering a longer time frame – we also include data for some years after the financial crisis of 2008/2009 – and by providing several additional robustness checks. The evidence that we provide on the determinants of the EC’s NAIRU estimates should both be valuable for monetary and fiscal policy-makers and for a broader audience of researchers interested in analyzing the cyclical and structural determinants of unemployment. The remainder of this article is structured as follows: in Section 2, we provide a short introduction to the empirical estimation and political application of the EC’s NAIRU estimates. Section 3 in turn reviews past empirical literature that analyzes the determinants of European unemployment and concisely summarizes the theoretical underpinnings of these applications. In Section 4 , we develop our econometric strategy for assessing the theoretical plausibility and robustness of the EC’s NAIRU estimates. Section 5 presents the econometric baseline results and Section 6 assesses the robustness of these findings. Section 7 discusses the role of the NAIRU in theory, empirics and policy practice. Finally, Section 8 concludes our argument. 2. The European Commission’s NAIRU approach: estimation and application In accordance with common practice, the EC defines the NAIRU as the unemployment rate at which (wage) inflation remains stable (European Commission, 2014 ) and, hence, introduces the NAWRU as an alternative acronym for the NAIRU concept,1 which is identified as a proxy for structural unemployment. Moreover, when actual unemployment (ut ) is equal to the NAIRU (Nt ) – i.e. the unemployment gap (ut − Nt ) is zero –, the economy is running at potential output (Havik et al., 2014 ). The EC’s NAIRU model is based on a Kalman Filter applied to an econometric model cast into a state-space framework (Durbin & Koopman, 2012), which consists of (a) a set of assumptions 1

In the rest of this paper, the terms NAIRU and NAWRU are, therefore, used interchangeably.

70

886


about the unobservables in the model that are of statistical nature (like lag-structures and autoregressive processes), as well as (b) a theoretical component based on a Phillips curve framework. In the latter case, estimated unemployment gaps are used to explain the growth in unit labor costs within the state-space model, possibly in conjunction with a series of exogenous regressors to increase the statistical precision of the underlying Kalman Filter model (Planas & Rossi, 2015). Hence, the theoretical arguments enter the model setup only indirectly to provide additional information for judging whether the observational data or the underlying model should be given priority in the recursions that make up the Kalman Filter (that is, for calculating the ‘Kalman gain’, see: Harvey, 1990; Kalman, 1960). The two so-called measurement equations of the NAIRU model formally look as follows: ut = N t + G t

(1)

grulct = αgrulct−1 + β1 Gt + β2 Gt−1 + γZt + arulct

(2)

with !1 < 0, !1 > 0; and where ut is the actual unemployment rate; Nt is the trend component of the unemployment rate (i.e., the NAIRU); Gt is the unemployment gap (ut − Nt ); grulct is the growth rate of real unit labor costs at time t, and grulct−1 is the lagged growth rate of rulc; Zt is a vector consisting of exogenous variables (which may include changes in terms of trade and in labor productivity etc.); and arulct is the error term, which captures measurement errors in grulct . Since the Spring Forecast 2014, the EC has been using Eq. (2) for most European countries,2 which is “based on rational expectations [. . .] [, implying] that a positive unemployment gap [. . .] is associated with a fall in the growth rate of real unit labor cost” (European Commission, 2014; p. 22). The measurement equations are complemented by a set of state equations, which specify the dynamics of the unobserved components of the model (Planas & Rossi, 2015) and have the following form: $Nt = ηt−1

(3)

$ηt = aηt

(4)

Gt = φ1 Gt−1 + φ2 Gt−2 + aGt

(5)

where the change in the NAIRU ($Nt ) is modeled as a Gaussian noise process (ηt ) governed by aηt . All shocks are normally distributed white noises, which are also assumed to be independent from each other. From Eqs. (3) and (4) on the dynamics of the unobserved components, we can see that the NAIRU is specifically modeled as a second-order random walk. And Eq. (5) means that the unemployment gap (Gt ) follows a second-order auto-regressive process, which has a sample mean of zero. The assumption that the unemployment gap follows an autoregressive process with zero mean is supposed to ensure that – in the absence of shocks – the unemployment rate converges to the structural rate of unemployment. What’s more, “specifying the unemployment gap as a process that reverts to a zero mean [. . .] seems to capture Friedman’s (1968) view that the

2

As of November 2015 (Autumn 2015 Forecast by the EC), this ’New Keynesian specification’ shown in Eqs. (1) and (2) has officially been used by the EC to obtain NAIRU estimates for the following EU countries: Bulgaria, Cyprus, Czech Republic, Denmark, Estonia, Greece, Spain, Finland, France, Hungary, Ireland, Latvia, Poland, Portugal, Romania, Sweden, Slovenia, Slovakia, UK. For the other EU countries, the EC still uses ”the so-called traditional Keynesian Phillips (TKP) curve based on static or adaptive expectation assumptions [relating] a positive unemployment gap [. . .] with a fall in the change of the growth rate of nominal unit labor cost [. . .] (and vice versa)” (European Commission, 2014)).

71

887


unemployment rate cannot be kept away indefinitely from the natural rate [of unemployment]” (Laubach, 2001; p. 221). The time-path of the NAIRU is extracted from the information contained in the measurement equations by employing the Kalman filter recursions. As the true values of the unobserved components – including the unemployment gap and the NAIRU – are unknown, the Kalman filter provides an algorithm to finding estimates for the unobservables (see Durbin & Koopman, 2012; p. 85). For the purpose of this paper, it is important to note that neither the variables capturing labor market structures (such as employment protection legislation, unemployment benefits, tax wedge, trade union density etc.) nor the non-structural variables (such as capital accumulation or the longterm interest rate), which might have an impact on the labor market, are included in the model. Nevertheless, the assumption that the NAIRU does eventually represent structural aspects and rigidities in labor markets manifests itself in the EC’s treatment of the subject (Orlandi, 2012; European Commission, 2013, 2014; Lendvai, Salto, & Thum-Thysen, 2015). Whether the NAIRU is determined by structural factors is most crucial when it comes to estimating potential output, which is basically derived from a production function approach making use of empirical data in conjunction with Kalman-filtered estimates for NAIRU (as explained above) and total factor productivity (TFP), where the rationale for filtering the latter is basically to smooth out cyclical variances in productivity growth, given a measure of factor utilization. The conceptual idea behind ‘potential output’ is to denote a hypothetical level of output at which all production factors would be employed at non-inflationary levels (Havik et al., 2014). In this context, the output gap is used as an indicator for the position of an economy in the business cycle: a positive output gap is said to indicate an over-heating economy, a negative output gap signals underutilization of economic resources. Hence, if there is no discrepancy between actual output and potential output, the output gap is zero. The EC’s production function approach is based on the following Cobb–Douglas production function: YPOTt= Ltα Kt1−α TFPt

(6)

where YPOTt is potential output, Lt is the contribution of labor supply to potential output, Kt is the contribution of the capital stock to potential output, and TFPt is total factor productivity. α and (1 − α) are the constant output elasticities of labor and capital, respectively (Havik et al., 2014; p.10).3 Since our focus is on the NAIRU, we look more specifically at the estimation of the labor component Lt , which crucially depends on NAIRU estimates: Lt= POPWt∗ PARTSt∗ (1 − NAIRUt) ∗ HOURSt

(7)

where POPWt is population of working age, PARTSt is the smoothed labor force participation rate, NAIRUt is the non-accelerating wage inflation rate of unemployment and HOURSTt is the trend of average hours worked (Havik et al., 2014; p. 14). PARTSt and HOURSTt are detrended 3

The EC assumes that the output-elasticities of labor and capital are equal to 0.65 and 0.35, respectively: ”The same Cobb-Douglas specification is assumed for all countries, with the mean wage share for the EU15 over the period 1960–2003 being used as guidance for the estimate of the output elasticity of labor, which would give a value of .63 for all Member States and, by definition, .37 for the output elasticity of capital [. . .] Since these values are close to the conventional mean values of 0.65 and 0.35, the latter are imposed for all countries.” (Havik et al., 2014, p. 10)

72

888


Table 1 Estimates for Spain in 2015: changes in NAIRU estimates have an impact on potential output and structural budget balances.

UNEMP NAIRU YREAL YPOT OG SB

(1) DATA AMECO

(2) NAIRU −1pp

(3) NAIRU −2.5pp

(4) NAIRU +1pp

(5) NAIRU +2.5pp

22.3% 18.5% 1071.1 1114.8 −3.9% −2.5%

22.3% 17.5% 1071.1 1123.7 −4.7% −2.1%

22.3% 16.0% 1071.1 1136.0 −5.7% -1.6%

22.3% 19.5% 1071.1 1105.9 −3.1% −2.9%

22.3% 21.0% 1071.1 1092.4 −2.0% −3.6%

Notes. Official AMECO data (column 1) is from the Autumn 2015 forecast of the EC. Output gaps and structural budget balances are measured in % of potential output. The calculations are based on the European Commission’s potential output model for calculating structural budget balances (Havik et al., 2014; Mourre et al., 2014; Planas & Rossi, 2015). All scenarios were estimated by holding everything but the NAIRU estimate constant. UNEMP, unemployment rate; NAIRU, non-accelerating (wage) inflation rate of unemployment; YREAL, GDP at 2010 prices (in billion D ); YPOT, potential output at 2010 prices (in billion D ); OG, output gap in % of potential output; SB, structural budget balance in % of potential output.

variables; they are calculated by using the Hodrick–Prescott-Filter.4 It can be seen that potential employment is equal to the labor force – obtained as the product of POPWt and PARTSt – times (1-NAIRUt ). In other words, estimates of the NAIRU are central to constructing estimates of potential output.5 We will now use a replication of the EC’s model for estimating the NAIRU and potential output to show how changes in the NAIRU have a direct impact on the scope and evaluation of fiscal policy. The structural budget balance, which is defined as the cyclically-adjusted budget balance, corrected for one-time and temporary effects (e.g. costs related to bailing-out financial institutions), is given by: (8)

SBt= FBt− εtOGt− OEt.

where SBt is the structural budget balance; FBt is the reported fiscal balance (defined as government revenues minus government expenditures relative to nominal GDP); εt is an estimate for the budgetary semi-elasticity, measuring the reaction of the fiscal balance to the output gap (OGt ); and OEt are one-off and temporary effects (Mourre, Astarita, & Princen, 2014). Table 1 illustrates the impact of changes in the NAIRU on potential output and the structural budget balance by using Spain as an example. The EC’s official Spanish NAIRU estimate in Autumn 2015 for the year 2015 was 18.5%. In the production function methodology, this NAIRU estimate corresponds to potential output of D 1114.8 billion, an output gap of −3.9% and a structural budget balance of −2.5% (both expressed in % of potential output). Holding everything else constant and assuming that the NAIRU in 2015 would have been estimated to be 1% point lower, we find that potential output rises to D 1123.7 billion, an increase of about 0.8% relative to the official estimate. As a consequence, the negative output gap is substantially larger than in the 4

The HP filter is a univariate approach to removing the cyclical component of a time series from the trend component (Hodrick & Prescott, 1997). Regarding the basic limitations of the HP filter – with particular emphasis on the so-called ’end-point bias’ –, see, e.g., Kaiser & Maravall (2001). 5 While the standard Cobb-Douglas framework is well established, there are criticisms concerning the foundations and the usage of aggregate production functions (e.g. Felipe & McCombie, 2014). This debate, however, goes beyond the focus of this paper.

73

889


baseline scenario (−4.7% compared to −3.9%), which translates into a decrease in the structural deficit from −2.9% to −2.1% (column 2). The differences are even more pronounced when we assume the Spanish NAIRU in 2015 to be 2.5% points (pp) lower than the initial estimates (column 3). Similarly, we can illustrate that upward revisions in the NAIRU compared to the official EC estimates lead to a substantial decrease in potential output going along with an increase in the structural deficit (colums 4 and 5). In other words, the larger (smaller) the estimate of the structural component of unemployment, the larger (smaller) the structural component of the fiscal deficit. The important point is that the structural budget balance is a central control indicator in the EU’s fiscal regulation framework. Crucially, medium-term budgetary objectives (MTOs) for EU countries are defined in terms of the structural budget balance (e.g ECFIN, 2013; Tereanu, Tuladhar, & Simone, 2014). In cases where member countries deviate from their MTO, they have to conform to the rules of the Stability and Growth Pact, which require an improvement of the structural budget balance by 0.5% points of nominal GDP per year. Since the reform in 2011, the Stability and Growth Pact also stipulates that deviations from the adjustment path to the MTO are significant when the ex-post improvement in the structural budget balance has not amounted to at least 0.5% points of GDP in one year or cumulatively over two years (European Union, 2011). According to the European Fiscal Compact, which came into effect on January 1st 2013, the yearly structural deficit may not exceed 0.5% of nominal GDP. The Fiscal Compact also includes the commitment of member countries to codify its rules in national law, preferably in the form of a constitutional safeguard (Fiscal Compact, 2012). Because of this institutionalization of structural budget balances, an increase in the structural deficit translates into more fiscal consolidation pressure. Against this background, it is essential whether the NAIRU is a good proxy for structural unemployment; otherwise, its usefulness as a key measure for estimating potential employment could be called into doubt. Hence, the empirical section of this paper will econometrically investigate the determinants of the NAIRU in order to shed light on the question: what does the NAIRU, as estimated by the EC, actually (not) measure? 3. The determinants of (structural) unemployment in European countries: literature review Due to the historical rise in European unemployment from the late 1970s to the 1990s, the literature on the cross-country determinants of (structural) unemployment grew rapidly in the 1990s and in the first half of the 2000s, as researchers were trying to explain changes in observed unemployment (see Table 2). A number of influental studies emphasized the link between labor market rigidities imposed by protective labor market institutions and rising unemployment across Europe (e.g. Bassanini & Duval, 2006 ; Belot & van Ours, 2004; International Monetary Fund, 2003; Nickell, Nunziata, & Ochel, 2005; OECD, 1994; Siebert, 1997;). This view and corresponding calls for ’structural labor market reforms’ provided the dominant theoretical interpretation of increasing unemployment in Europe, supported by “a wide range of analysts and international organizations – including the EC, the Organization for Economic Cooperation and Development (OECD), and the International Monetary Fund (IMF) –, [all of whom] have argued that the causes of high unemployment can be found in labor market institutions.” (International Monetary Fund, 2003, p. 129) However, several empirical studies have shown more recently that the empirical evidence for the view that institutions are at the heart of the European unemployment problem from the 1970s to the 1990s is modest at best, since the underlying correlation lacks robustness with regard to variations in control variables, estimation techniques as well as selected countries and time periods

74

Gianella et al. (2008)

Bertola, Blau, & Kahn (2007)

Baccaro & Rei (2007)

Palacio-Vera, Aguilar, de la Cruz, & Martinez-Canete (2006) Arestis, Baddeley, & Sawyer (2007)

Bassanini & Duval (2006)

Nickell et al. (2005)

Baker et al. (2005)

Belot and van Ours (2004)

International Monetary Fund (2003)

Elmeskov, Martin, & Scarpetta (1998) Blanchard and Wolfers (2000)

Nickell (1997)

9 OECD countries (quarterly data, max. 1979–2002). Time series 18 OECD countries (1960–1998). Dynamic panel; Panel with 5-year-averages 20 OECD countries (1960–1996). Panel with 5-year averages 19 OECD countries (1978–2002). Panel (annual)

NAIRU (OECD)

Employment rate

UNEMP

UNEMP

NAIRU (OECD)

UNEMP

UNEMP

UNEMP

UNEMP

UNEMP

UNEMP

UBR, BD, UnD, EPL, COORD, ALMP TW, PMR, UBR, UnD

UBR, BD, UnD, EPL, COORD, TW

UBR, strike activity

UBR, BD, UnD, EPL, COORD, ALMP UBR, BD, UnD, EPL, COORD, TW UBR, BD, EPL, UnD, COORD, ALMP; PMR –

UBR, BD, UnD, EPL, CBC, TW, ALMP UBR, UnD, EPL, CBC, TW, ALMP, MW UBR, BD, UnD, COORD, TW, ALMP, MW UBR, EPL, UnD, COORD, TW <I, TFPS, TOTS, CBI UBR, EPL, UnD, CWB

UNEMP

20 OECD countries (1983–1994). Panel with two 6-year averages 19 OECD countries (1983–1995). Panel (annual) 20 OECD countries (1960–1996). Panel with 5-year averages 20 OECD countries (1960–1998). Dynamic panel (annual) 17 OECD countries (1960–1999). Panel with 5-year-averages 20 OECD countries (1960–1999). Panel with 5-year averages 20 OECD countries (1961–1995). Dynamic panel (annual) 21 OECD countries (1982–2003). Dynamic panel (annual) USA 1964:2–2003:1. Time series UNEMP

LMI variable

Dependent variable

Data

Table 2 Literature review: Selected empirical studies on the determinants of (structural) unemployment.

LTI

LTI, TFPS, LDS

LTI, TFPS, TOTS, LDS

ACCU

ACCU, TOTS

LTI, TFPS, LDS, TOTS, money supply LTI, TFPS, TOTS, LDS

–

–

LTI, TFPS, TOTS, CBI

LTI, TFPS, TOTS, LDS

–

–

Other controls

890 P. Heimberger et al. / Journal of Policy Modeling 39 (2017) 883–908

75

20 OECD countries (1983–2003; 1960–1999); Panel with 5-year-averages 13 EU countries (1985–2009). Panel (annual) 20 OECD countries (1961–1995). Dynamic panel (annual) 32 EU and OECD countries (1980–2009). Panel (annual) 15 EU Countries (1985–2008). Panel (annual) 19 OECD countries (1960–2000). Panel (annual) 12 OECD countries (2007–2011). Panel (annual) UNEMP

UNEMP

NAIRU (EC)

UNEMP

TFP growth rate, HBOOM – LTI, HBOOM, ACCU

EPL, UnD, UBR, CWB, TW EPL, ALMP, MW, UnD, GRR

TFP growth rate, LTI, HBOOM LTI, TFPS, LDS, TOTS, money supply TOTS, LTI, CBI

TOTS, ACCU, TFPS, LTI, LDS

UBR, BD, UD, EPL, COORD, TW UBR, EPL, TW, COORD, UnD TW, PLM, ALMP, SMI, MEI

UBR, TW, UnD, ALMP

NAIRU (EC) UNEMP

UBR, BD, UnD, EPL, TW, COORD, CBC, PMR

UNEMP

Notes: Illustration on the basis of Stockhammer & Klär (2011, p. 441), whose literature review table was complemented with a) some more information on the respective studies and with b) the relevant literature since 2007. ACCU, capital accumulation; ALMP, active labor market policy; BD, benefit duration; CBC, collective bargaining coverage; CBI, Central Bank Independence index; COORD, wage bargaining coordination; CWB, centralization of wage bargaining; EPL, employment protection legislation; HBOOM, proxy for boom-bust patterns in housing; LMI, labor market institution; LDS, labor demand shock; LTI, long-term real interest rate; MEI, Matching efficiency indicator; MW, minimum wage; PLM, passive labor market policies; PMR, product market regulation; SMI, skill mismatch indicator; TFPS, deviation of total factor productivity from its trend; TOTS, terms of trade shock; TW, tax wedge; UnD, trade union density; UBR, unemploment benefit replacement rate.

Stockhammer, Guschanski, & Köhler (2014)

Flaig & Rottmann (2013)


Avdagic & Salardi (2013)

Vergeer & Kleinknecht (2012)

Orlandi (2012)

Stockhammer & Klär (2011)

P. Heimberger et al. / Journal of Policy Modeling 39 (2017) 883–908 891

76

892


(e.g. Avdagic & Salardi, 2013; Baccaro & Rei, 2007 ; Howell, Baker, Glynn, & Schmitt, 2007 ; Stockhammer & Klär, 2011; Vergeer & Kleinknecht, 2012). The focus in the empirical panel data literature is to explain broad movements in unemployment across OECD countries by shifts in labor market institutions (LMIs) such as trade union density, employment protection legislation, unemployment benefit replacement rate, tax wedge, active labor market policies, minimum wages etc. (see Table 2). As some studies had found no “meaningful relationship between [the] OECD measure of labor market deregulation and shifts in the NAIRU” (Baker, Glyn, Howell, & Schmitt, 2005; p. 107 ), researchers began to include additional control variables representing alternative explanations for the evolution of (structural) unemployment. Blanchard and Wolfers (2000), for instance, control for ‘macroeconomic shocks’ such as changes in the long-term interest rate, deviations from the trend in total factor productivity growth and shifts in labor demand, emphasizing the link between these shocks and labor market institutions. Stockhammer and Klär (2011) regard investment as the most crucial variable in explaining unemployment; hence, they include measures of capital accumulation in their regressions. Bassanini and Duval (2006), among others, include a terms of trade shock variable in their regressions, since a change in the terms of trade is assumed to affect domestic unemployment: whenever a country’s terms of trade improve (deteriorate), this implies that for every unit of export sold, this country can purchase more (less) units of imported goods; when imports become less (more) attractive, domestic employment is affected positively (negatively). Finally, Orlandi (2012) introduces another essential control variable, as he considers a proxy for boom-bust-patterns in housing markets. This modification aims to empirically scrutinize the assertion that ‘non-structural’ factors do not affect ‘structural’ unemployment at all and, indeed, he finds that in some instances such ‘non-structural factors’ are “the main drivers of NAWRU developments” (Orlandi, 2012; p. 26). However, all major empirical studies on the econometric determinants of unemployment in OECD countries making use of panel data (see Table 2) are characterized by at least one of the following two shortcomings: first, neglecting the role of capital accumulation and investment, the impact of boom-bust patterns related to housing and other macroeconomic developments, like changes in the real interest rate and the terms of trade; second, including only few institutional labor market variables or not considering this aspect at all. Moreover, there are only three studies, which have already looked at the determinants of Kalman-filtered NAIRU estimates across several OECD countries, while all the other papers use observed (and in some cases smoothed) unemployment rates as their preferred dependent variable. The relevant papers by Orlandi (2012), the European Commission (2013) and Gianella et al. (2008), however, are also incomplete in the sense that they fail to account for the possibility of relevant alternative explanations for the evolution of NAIRU estimates. Our paper closes this gap by analyzing the role of standard labor market variables in explaining the evolution of the EC’s NAIRU estimates, while also controlling for a comprehensive set of variables capturing alternative hypotheses with regard to the determinants of the NAIRU. While the debate on the causes and evolution of European unemployment is again in full swing (e.g. Arpaia et al., 2014 ; European Central Bank, 2015), in what follows this paper provides an empirical contribution to this debate by econometrically assessing the validity of widely used NAIRU measures for ’structural unemployment’ in European countries.

77

893


4. Basic econometric strategy and data The empirical part of this paper analyzes the econometric determinants of the EC’s NAIRU estimates. For this purpose, we identified a comprehensive set of explanatory variables covering the basic theoretical and empirical rationales employed in past work and composed a corresponding time-series cross-section data set of 14countries,6 for which the complete set of the relevant data could be retrieved. We derive two main specifications from this data: First, we analyze a longterm baseline model based on data for the time period 1985–2011, which covers 11 European OECD countries. Second, we provide an alternative baseline specification focusing on a more recent period (2001–2012). Aside from data considerations – the short-term sample allows for the inclusion of 14 countries and two additional LMI variables –, this second specification is motivated by the specific temporal settings, which makes it possible to focus on (a) the euro-era and (b) the run-up and aftermath of the financial crisis. Our data set enables us to go beyond past contributions on the subject in at least three dimensions. First, we study factors explaining the EC’s NAIRU estimates, while nearly all other comparable empirical papers analyze the determinants of observed actual unemployment rates. Second, the time frame of our data set is longer than in comparable studies (Gianella et al., 2008; Orlandi, 2012; European Commission, 2013). In particular, we go beyond past work by including data on the period after the financial crisis of 2008. Third, we look at a more diverse set of potential explanatory variables as compared to past studies. Specifically, we combine data on labor market institutions as provided by the OECD with additional explanatory variables in order to account for alternative hypotheses regarding the evolution of the NAIRU. The baseline model uses the official NAIRU estimates from the EC’s Autumn 2015 forecast as the dependent variable (NAIRUi,t ). The regression equation has the following form: NAIRUi,t = βLMIi,t+ γCi,t+ δ1 FEi + δ2 FEt+ εi,t

where β represents a vector of regression coefficients related to different structural labor market indicators (LMIi,t ), while ! is a set of regression coecients covering other explanatory factors for the NAIRU used in past works (Ci,t ), which will be introduced in Table 3 below. We also introduce country-fixed effects (FEi ) to account for unmeasurable, time-invariant country-specific characteristics that may influence the NAIRU as well as period-fixed effects (FEt ) to capture time-varying shocks affecting all countries. εi,t represents the stochastic residual. By including country-fixed effects and period-fixed effects, we follow usual practices in the relevant empirical literature (e.g. Baccaro & Rei, 2007; Nickell et al., 2005; Stockhammer & Klär, 2011). Table 3 provides a detailed overview of the variables included in our data set. Our data on structural labor market indicators (LMIi,t ) comprises six standard labor market variables obtained from the OECD’s data base: employment protection legislation (EPL), expenditures on active labor market policies (ALMP),7 trade union density (UnD), unemployment benefit replacement rate (UBR and UBR2),8 tax wedge (TW) and minimum wage (MW). Variables related to alternative 6

This group of 14countries includes: Austria, Belgium, Czech Republic, Denmark, Finland, France, Germany, Ireland, Netherlands, Poland, Portugal, Slovak Republic, Spain, Sweden. Six other countries – Estonia, Greece, Italy, Luxembourg, Slovenia, United Kingdom – have been excluded from the analysis due to data limitations, which are most pronounced in the context of institutional labor market variables. 7 In this case, we use the ratio of ALMP expenditures (in % of nominal GDP, as provided by the OECD) to the unemployment rate to account for the fact that ALMP expenditures rise and decrease with current unemployment rates. 8 For the period 2001–2012, we use OECD data on net replacement rates (UBR2). However, as those data are only available from 2001, we have to use gross replacements rates for the period 1985–2011 (UBR). The OECD’s gross

78

894


Table 3 Variables and data sources.

NAIRU

Data description

Data sources

Non-accelerating wage inflation rate of unemployment

AMECO (Autumn 2015 issue)

Data on labor market institutions (LMIi,t ) EPL Strictness of employment protection, individual and collective dismissals (regular contracts) ALMP Public expenditure and participant stocks in LMP (in % of nominal GDP), divided by the unemployment rate Trade union density UnD UBR Gross unemployment benefit replacement rate UBR2 Net unemployment benefit replacement rate TW Average tax wedge (Single person at 100% of average earnings, no child) MW Real minimum wages (in 2014 constant prices at 2014 USD PPPs) Additional control variables (Ci,t ) ACCU Real gross fixed capital formation/real net capital stock (*100) HBOOM Deviation of the ratio of employment in the construction sector to total employment in all domestic industries from its mean (*100) LTI Real long-term interest rates TFP Yearly growth rate in Total Factor Productivity TOTS Yearly growth rate in terms of trade index Data for reduced form NAIRU model and different NAIRU forecasts UNEMP Unemployment rate !INFL Change in the growth rate of the harmonized consumer price index NAIRU2014 Non-accelerating wage inflation rate of unemployment NAIRU2013 Non-accelerating wage inflation rate of unemployment

OECD (December 2nd 2015) OECD (December 2nd 2015)

OECD (December 2nd 2015) OECD (December 2nd 2015) OECD (December 2nd 2015) OECD (December 2nd 2015) OECD (December 2nd 2015)

AMECO (Autumn 2015 issue) AMECO (Autumn 2015 issue)

AMECO (Autumn 2015 issue) AMECO (Autumn 2015 issue) OECD (December 22nd 2015) AMECO (Autumn 2015 issue) IMF World Economic Outlook (October 2015) AMECO (Autumn 2014 issue) AMECO (Autumn 2013 issue)

explanations of (structural) unemployment are collected in Ci,t and include the following data: first, we introduce an indicator covering changes in the capital stock (following Stockhammer & Klär, 2011). Capital accumulation (ACCU) in this sense is defined as the ratio of real gross fixed capital formation to the real net capital stock. Second, we employ a proxy for boom-bust-patterns related to the housing market (HBOOM); it is defined as the yearly deviation of the ratio of employment in the construction sector to total employment from its mean – as in Orlandi (2012). Additionally, we include the annual growth rate in total factor productivity (TFP), a variable for terms of trade shocks (TOTS) and the long-term real interest rate (LTI).

replacement rate data are only available for every second year; therefore, it was interpolated for the missing years. Two separate time series of gross replacement rates were chained. The first series ranges from 1961 to 2005 and is based on Average Production Worker wages; the second time series ranges from 2005 to 2011 and is based on Average Worker wages.

79

895


According to Nickell (1998) and other authors who emphasize the role of labor market institutions when it comes to explaining the evolution of (structural) unemployment, UnD, UBR, MW and TW are all expected to have a positive sign, i.e. to be positively associated with (structural) unemployment. The general reasoning is that labor market institutions improve the bargaining position of workers and/or reduce the willingness and capacity of unemployed workers to put downward pressure on wages, which causes labor market rigidities that lead to an increase in unemployment. In contrast, ALMP should have a negative sign, as active labor market policies are expected to increase matching efficiency and, hence, dampen labor market rigidity (e.g. Arpaia et al., 2014 ). The expected empirical effects of EPL, however, are theoretically ambiguous. On one hand, EPL will dampen job creation according to the standard model, because employers are reluctant to hire them due to the fear that they cannot be laid off easily; on the other hand, stricter EPL also increases job retention, as employers lay off fewer employees during economic downturns. Furthermore, stronger EPL could encourage investments in the training of employees as well as innovation on the firm-level (Zhou, Dekker, & Kleinknecht, 2011), thereby potentially increasing productivity. The effects of EPL are, therefore, ex ante ambiguous (Avdagic & Salardi, 2013 ). Stockhammer and Klär (2011) provide an additional perspective by emphasizing the role of capital accumulation as an explanatory factor: a decrease in investment causes unemployment to increase (and vice versa), so that ACCU is expected to have a negative sign. LTI also affects capital accumulation; it should be positively associated with unemployment, as an increase in real interest rates is likely to lead to lower aggregate demand, which increases unemployment (e.g. Baker et al., 2005). Orlandi (2012) controls for LTI, but not for ACCU; however, he introduces an additional variable (HBOOM) in his analysis to assess the impact of “severe housing boom-bust effects” (Orlandi, 2012; p. 10). Although from a textbook perspective such ‘boom-bust effects’ are of a cyclical, transitory nature and should not affect the NAIRU, Orlandi nonetheless posits a negative relationship between HBOOM and NAIRU estimates. According to Blanchard and Wolfers (2000), TFP is expected to have a negative sign, as a decline in TFP growth will cause structural unemployment to increase. Finally, TOTS is a measure for terms of trade shocks, where an improvement in the terms of trade implies that imports become relatively cheaper. Hence, the upward-pressure on wages induced by import-prices is reduced (e.g. Bassanini & Duval, 2006 ). It follows that a positive terms of trade shock is expected to lower unemployment, and vice versa. In order to identify a suitable estimation approach for running our regressions, we tested for non-stationarity by running panel unit root tests (Choi, 2001) on the country series for NAWRU, the LMI variables and the additional controls ACCU, HBUB, LTI, TFP and TOTS. For the time period 1985–2011, the null hypothesis that all country series contain a unit root can be rejected for all variables but UnD, EPL, ALMP and LTI. Against the background of these results from the panel unit root tests, we also implemented the test for co-integration proposed by Maddala and Wu (1999), where the null hypothesis is the presence of a unit root in the residuals, i.e. no co-integration amongst the variables. The Maddala–Wu test results signal that estimating our proposed model in levels is appropriate, since the test rejects the null hypothesis of no cointegration at the 1% level, implying that standard OLS and Fixed Effects estimators are consistent. To ensure robustness of the results, our estimation strategy for analyzing the econometric determinants of the EC’s NAIRU estimates is based on two different estimation strategies. In what follows, our preferred estimation approach is to use ordinary least squares (OLS) with panel-corrected standard errors (PCSE), where we include both country- and period-fixed effects. According to Beck and Katz (1995), the OLS-PCSE procedure is well suited for time-series cross-section models such as ours, where the number of years covered is not much larger than

80

896


the number of countries in the cross-sectional dimension of the data. The main reason for the superior performance of the OLS-PCSE estimation strategy – compared to the Parks estimator and other Feasible Generalized Least Squares estimators regularly used in the relevant empirical literature – is that the method proposed by Beck and Katz (1995) is well suited to adressing crosssection heteroskedasticity and autocorrelation in the residuals. Since these two properties are often characteristic of time-series cross-sectional data, the OLS-PCSE estimation strategy helps to avoid overconfidence in standard errors, which is often attributed to the empirical literature on the determinants of unemployment in Europe (Vergeer & Kleinknecht, 2012). Finally, it should be added that this estimation strategy is not an entirely new approach; in fact, using a fixed effects panel estimator in levels is a common estimation technique in recent empirical research on the determinants of (structural) unemployment (e.g. Flaig & Rottmann, 2013 ), with some authors also following the OLS-PCSE estimation and correction procedure as implemented in this paper (Avdagic & Salardi, 2013 ; Orlandi, 2012). This preferred estimation approach is complemented by using a first difference estimator applied to annual data and 5-year-data averages, respectively. In accordance with Baccaro and Rei (2007) we find that using first differences of 5-year-average-data removes the positive autocorrelation in the residuals, which is characteristic of our baseline regression results. Aside from this econometric justification, the economic rationale for using 5-year-averages has two aspects: First, it takes into account that labor market institutions only change slowly. Second, it dampens possible effects of business cycle fluctuations on (structural) unemployment, which should allow for more reliable causal interpretations. The obvious drawback from using averages, however, is a loss of information as contained in the data, which makes it especially difficult to trace short-term effects between our explanatory variables and NAIRU estimates, as well as a drastic reduction of observations, which lowers the statistical power of the tests. Against this backdrop, our preferred estimation strategy is to use annual data in levels in a time-series cross-section model with OLS-PCSE, while our alternative estimation strategy based on first-differences of either annual data or 5-year averages is used primarily as an additional tool examining the robustness of single relationships between the explanatory variables and the NAIRU estimates. 5. Econometric baseline results The econometric baseline results for 11 European OECD countries over the time period 1985–2011 are shown in Table 4 for six different models. In the first column, we regress the EC’s NAIRU estimates on four instititutional labor market indicators (EPL, ALMP, UnD, UBR); in addition, we control for TFP and TOTS. Arguably, this specification leaves ample scope for the institutional variables to explain the variation in the dependent variable. The regression coefficients represent the impact of a 1 unit increase in the respective explanatory variable on the NAIRU (in percentage points). For example, an increase in the unemployment benefit replacement rate (UBR) by 10% points increases the NAIRU by about 0.9% points. Standard errors of the fixed effects models shown in Table 4 are PCSE-corrected standard errors. As both Durbin–Watson (DW) and Breusch–Godfrey (BG) tests on autocorrelation indicate positive serial correlation in the residuals, the PCSE-procedure is a sensible tool to account for this data characteristic in our fixed effects models. In Model 1, all coefficients of the institutional variables are signed as expected in the standard literature on the determinants of structural unemployment. However, only ALMP is statistically significant at the 5% level, while UBR is weakly significant using a 90% confidence interval. The adjusted R2 indicates that the regressors are merely able to explain about 20% of the variation

81

897

P. Heimberger et al. / Journal of Policy Modeling 39 (2017) 883–908 Table 4 Results for 1985–2011. Dependent variable:NAIRU (3) OLS-PCSE

(4) OLS-PCSE

(5) FD

(6) FD

0.485 (1.782) −0.050** (0.025) 0.100 (0.091) 0.089* (0.053) 0.015 (0.088) −0.079 (0.084)

0.071 (0.060) 1.660* (0.936) −0.029** (0.013) 0.056 (0.048) 0.072*** (0.025) −0.104 (0.067) 0.008 (0.062)

−0.998*** (0.187) 0.238** (0.094) −0.134 (1.204) −0.037** (0.017) 0.092 (0.065) 0.096** (0.039) −0.229*** (0.085) −0.002 (0.071)

−1.327*** (0.233) −0.242 (0.196) 0.064 (0.063) 1.391 (0.904) −0.029** (0.013) 0.058 (0.047) 0.080*** (0.025) −0.145** (0.069) −0.006 (0.060)

−0.226*** (0.071) −0.289*** (0.075) 0.032** (0.016) 0.088 (0.274) −0.004 (0.004) 0.055*** (0.020) 0.016* (0.009) 0.001 (0.010) 0.004 (0.009) 0.064*** (0.022)

−0.721*** (0.261) −0.565* (0.288) 0.064 (0.112) 1.681** (0.726) −0.027** (0.012) 0.100 (0.060) 0.102*** (0.036) −0.417* (0.241) −0.078 (0.180) 0.116 (0.250)

11 27 297 0.195 Yes Yes 0.182

11 27 297 0.582 Yes Yes 0.448

11 27 297 0.463 Yes Yes 0.382

11 27 297 0.586 Yes Yes 0.476

11 26 286 0.323 No No 0.382

11 4 44 0.553 No No 1.900

(1) OLS-PCSE ACCU

(2) OLS-PCSE −1.509*** (0.177)

HBOOM LTI EPL ALMP UnD UBR TFP TOTS Constant Countries Time periods Observations Adjusted R2 Country FE Period FE DW test

*p < 0.1; **p < 0.05; ***p < 0.01. Notes. (1)–(4) OLS-PCSE. Standard errors in brackets () corrected for autocorrelation in residuals. Cross-section and Year Fixed Effects. (5) First difference estimator. Heteroskedasticity-robust standard errors. (6) First difference estimator, five-year-averages. Heteroskedasticity-robust standard errors. Country group in all specifications: Austria, Belgium, Denmark, Finland, France, Germany, Ireland, Netherlands, Portugal, Spain, Sweden. DW test denotes the Durbin-Watson test statistic on autocorrelation in the residuals. NAIRU, non-accelerating (wage) inflation rate of unemployment; ACCU, capital accumulation; HBOOM, housing boom/bust proxy; LTI, long-term real interest rate; EPL, employment protection legislation; ALMP, active labor market policies; UnD, trade union density; UBR, gross unemployment benefit replacement rate; TFP, total factor productivity; TOTS, terms of trade shock.

in the EC’s NAIRU estimates. In brief, the results from column 1 suggest that we ought to reject the hypothesis that NAIRU estimates can be exclusively explained by differences in labor market institutions and productivity growth. In model (2), we therefore introduce capital accumulation and the long-term real interest rate to account for alternative hypotheses aiming to explain the evolution of the EC’s NAIRU estimates. The introduction of those two additional variables leads to a tripling of the adjusted R2 , which changes to 58%. LTI is positively signed (but insignificant),

82

898


while ACCU – as expected in the relevant literature (Stockhammer & Klär, 2011) – is negatively signed and strongly significant, with the coefficient implying that an increase in the ratio of real gross fixed capital formation to the real net capital stock by 1% point lowers the NAIRU by 1.5% points. The size of the coefficients of the institutional variables in column (2) changes to varying degrees, while the estimated direction of the effects remains the same. EPL turns weakly significant, while UBR is now significant at the 1% level. In model (3), we again exclude ACCU, but instead introduce our proxy for boom-bust patterns in housing (HBOOM), which is signed as expected and highly significant, suggesting that boom (bust) patterns in housing are associated with decreases (increases) in the NAIRU. It is also notable that the coefficient of LTI in this setup is markedly larger than in column 2 and significant at the 5% level. However, as soon as we include all our regressors at once in column 4, LTI and HBOOM turn insignificant, while the coefficient of ACCU remains negative, large and highly significant, which supports the earlier finding from model 3 that capital accumulation plays an important part in explaining NAIRU estimates in our data set of European OECD countries. According to model 4, an increase in the ratio of capital formation to the capital stock by 1% point lowers the NAIRU by approximately 1.3% points, while a 10% point increase in UBR increases the NAIRU by 0.8% points. One possible issue with model 4 could be that the inclusion of fixed effects has an impact on the size and significance of the LMI coefficients. In order to investigate this issue, we also ran regressions with country-fixed effects only. We find that the LMI coefficients and their significance do not change markedly when we exclude period-fixed effects, while the coefficient of HBOOM nearly doubles to −0.45 and turns significant at the 5% level; ACCU retains its significance at the 1% level.9 In model 5, we employ a First Difference estimator to the annual data (with robust standard errors). Notably, all institutional variables are again signed as expected, but remain statistically insignificant with the exception of UBR (weakly significant) and UnD (strongly significant). In this specification, capital accumulation, the housing boom/bust proxy and the long-term interest rate have a significant effect on NAIRU estimates. Finally, model 6 follows the strategy preferred by Baccaro and Rei (2007), i.e. deploying the First Difference estimator after calculating 5-year averages for all time series. Regarding the institutional variables, model 6 finds EPL, ALMP and UBR to be signed as expected as well as statistically significant (at different levels of confidence). However, the major finding that capital accumulation and housing booms and busts are controls that ought not to be omitted when trying to explain the EC’s NAIRU estimates is also retained in this final specification. Table 5 illustrates the baseline results for the time period 2001–2012, where all model specifications with the exception of model 6 are the same as in Table 4. Looking at the institutional variables, we again find that – with very few exceptions – all LMIs are signed as expected across the different model specifications. As in the time period 1985–2011, ALMP and UBR are again the only significant LMI variables. We also support the major finding from the longer time period that ACCU plays an important part in explaining the NAIRU: in all columns, ACCU is at least significant at the 5% level. LTI has a larger coefficient and seems to play a somewhat stronger role than over 1985–2011, as it is highly significant in nearly all of the relevant models. Moreover, HBOOM is also again signed as expected and statistically significant in the majority of scenarios. Summing up, running regressions on the shorter time period of 2001–2012 – for which data avail-

9

More detailed results are available at request from the authors.

83

899

P. Heimberger et al. / Journal of Policy Modeling 39 (2017) 883–908 Table 5 Results for 2001–2012. Dependent variable: NAIRU (3) OLS-PCSE

(4) OLS-PCSE

(5) FD

(6) OLS-PCSE

−2.480* (1.401) −0.142*** (0.029) 0.134 (0.096) 0.182*** (0.038) 0.315** (0.132)

0.211*** (0.057) 0.591 (0.973) −0.064*** (0.019) 0.064 (0.056) 0.111*** (0.021) 0.032 (0.095)

−0.559*** (0.139) 0.239*** (0.062) 0.311 (0.950) −0.084*** (0.020) 0.033 (0.068) 0.124*** (0.026) 0.048 (0.104)

−0.627*** (0.186) −0.303* (0.161) 0.188*** (0.058) 0.529 (0.914) −0.064*** (0.018) 0.043 (0.057) 0.113*** (0.022) −0.017 (0.096)

−0.139** (0.070) −0.353*** (0.080) 0.031 (0.019) 0.046 (0.314) −0.006 (0.008) −0.005 (0.021) 0.016** (0.008) -0.001 (0.028)

0.110 (0.068) 0.026 (0.076)

0.051 (0.057) 0.183*** (0.058)

−0.016 (0.057) 0.180*** (0.064)

0.009 (0.055) 0.173*** (0.057)

0.008 (0.009) 0.021 (0.013) 0.020 (0.028)

−0.479** (0.224) −0.262 (0.184) 0.244*** (0.071) 0.147 (1.063) −0.079*** (0.029) 0.115* (0.066) 0.146*** (0.034) 0.042 (0.123) −0.00003 (0.0002) 0.036 (0.085) 0.124 (0.079)

14 12 168 0.479 Yes Yes 0.851

14 12 168 0.627 Yes Yes 0.884

14 12 168 0.612 Yes Yes 0.769

14 12 168 0.633 Yes Yes 0.795

14 12 154 0.339 No No 0.684

(1) OLS-PCSE ACCU

(2) OLS-PCSE −0.858*** (0.161)

HBOOM LTI EPL ALMP UnD UBR2 TW MW TFP TOTS Constant Countries Time periods Observations Adjusted R2 Country FE Period FE DW test

9 12 108 0.595 Yes Yes 0.995

*p < 0.1; **p < 0.05; ***p < 0.01. Notes. (1)–(4), (6) OLS-PCSE. Standard errors in brackets () corrected for autocorrelation in residuals. Cross-section and Year Fixed Effects. (5) First difference estimator. Heteroskedasticity-robust standard errors. Country group for specifications (1)–(5): Austria, Belgium, Czech Republic, Denmark, Finland, France, Germany, Ireland, Netherlands, Poland, Portugal, Slovak Republic, Spain, Sweden. Due to missing MW data, specification (6) excludes Austria, Denmark, Finland, Germany and Sweden. DW test denotes the Durbin-Watson test statistic on autocorrelation in the residuals. NAIRU, non-accelerating (wage) inflation rate of unemployment; ACCU, capital accumulation; HBOOM, housing boom/bust proxy; LTI, long-term real interest rate; EPL, employment protection legislation; ALMP, active labor market policies; UnD, trade union density; UBR2, net unemployment benefit replacement rate; TW, tax wedge; MW, minimum wage; TFP, total factor productivity; TOTS, terms of trade shock.

84

900


ability for LMIs has improved – supports our baseline findings from 1985–2011. This suggests that the EC’s implicit assumption that NAIRU estimates gained by de-trending the unemployment rate are a good proxy for ‘structural unemployment’ does not hold. On the contrary, most institutional variables are either statistically insignificant or their significance is sensitive to the model specification, while cyclical factors – especially capital accumulation – play a prominent role in explaining NAIRU estimates. 6. Robustness checks To assess the sensitivity of the baseline results, this section discusses several robustness checks: Specifically, we analyze the impact of variations in the country group, introduce lag specifications, consider interaction terms and, finally, implement variations in the dependent variable. The first sensitivity test consists of checking whether outlier countries drive our overall baseline results. Therefore, we varied the country group by excluding one country at a time. The results from this variation allow us to conclude that for both the long period (1985–2011) and the shorter period (2001–2012) neither the size of the coefficients of the explanatory variables nor their statistical significance are markedly affected by including or excluding single countries.10 In a second step, we investigated how the introduction of lags affects our regression results. In doing so we use specification (4) from the baseline models as a reference point, as it includes all major variables that proved to be empirically relevant in our past explorations. Table 6 depicts lag specification results for both time periods, where columns (1)–(3) refer to 1985–2011 and columns (4)–(6) depict the results for 2001–2012. In columns (1) and (4) we introduce lags for all the LMI variables to allow for the argument that institutional changes tend to affect the NAIRU with a lag, which could also have an impact on the performance of our alternative explanatory variables. However, this hypothesis is not supported by the regression results, as coefficients and standard errors of the variables ACCU, HBOOM and LTI remain largely unaffected after we introduce LMI lags, while the institutional variables either have a sign that is not in line with their standard theoretical prediction or they are statistically insignificant. We proceeded by including lags for capital accumulation, the housing boom/bust proxy and the real interest rate in columns (2) and (5) to find out whether these alternative factors impact on the NAIRU with a lag. We confirm the central role of ACCU in explaining the EC’s NAIRU estimates, although the ACCU coefficient in column 2 is less negative due to the introduction of the statistically significant ACCU lag. In columns (3) and (6) we include all possible lag terms: both for the LMI variables, and ACCU/HBOOM/LTI; in addition, we also consider lags for TOTS and TFP. The main results from the reference model in the baseline tables, however, still hold: while they underscore the importance of alternative factors – especially ACCU – in explaining the NAIRU, the econometric evidence for the role of LMI variables is at best mixed. A third sensitivity topic are interaction terms, as the econometric literature contains several papers which emphasize that LMIs should be expected to have an effect on (structural) unemployment through their interactions (e.g. Bassanini & Duval, 2006; International Monetary Fund, 2003). In a seminal paper, Blanchard and Wolfers (2000) stress the role of interactions between LMI variables and macroeconomic shocks. A major problem in this literature, however, is that “[t]he theoretical foundation for these interactions is [. . .] unspecific. For example, the IMF (2003) argues that the effects of different LMI are reinforcing, without specifying ex ante which LMI 10

Detailed regression results from varying the country group are available at request from the authors.

85

901

P. Heimberger et al. / Journal of Policy Modeling 39 (2017) 883–908 Table 6 Lagspecifications; results for 1985–2011 and 2001–2012. Dependent variable: NAIRU (1)

(2)

(3)

(4)

(5)

(6)

Time period Estimator

1985–2011 OLS-PCSE

1985–2011 OLS-PCSE

1985–2011 OLS-PCSE

2001–2012 OLS-PCSE

2001–2012 OLS-PCSE

2001–2012 OLS-PCSE

EPL

1.093 (0.725) 0.903 (0.662) 0.007 (0.016) −0.035** (0.015) −0.237*** (0.086) 0.291*** (0.089) 0.026 (0.044) 0.062 (0.051)

1.725** (0.839)

1.337* (0.792) 0.325 (0.685) −0.009 (0.015) −0.019 (0.015) −0.185** (0.093) 0.203** (0.093) 0.032 (0.045) 0.063 (0.047)

0.714 (0.804) −0.555 (1.148) −0.007 (0.021) −0.079*** (0.024) 0.016 (0.061) 0.052 (0.067)

0.332 (0.870)

0.748 (0.821) −0.704 (1.107) -0.010 (0.022) −0.072*** (0.024) 0.042 (0.061) 0.054 (0.069)

0.097*** (0.023) 0.051** (0.020) −0.151* (0.078) 0.157* (0.087) −0.637*** (0.174)

0.118*** (0.024)

EPLt-1 ALMP ALMPt-1 UnD UnDt-1 UBR UBRt-1

−0.027** (0.011)

0.056 (0.043)

0.078*** (0.025)

UBR2 UBR2t-1 TW TWt-1 ACCU

−1.468*** (0.236)

ACCUt-1 HBOOM

−0.159 (0.202)

HBOOMt-1 LTI

0.081 (0.062)

LTIt-1 TFP TFPt-1 TOTS TOTSt-1 Countries Time periods Observations

−0.140** (0.067) −0.049 (0.062) −0.036 (0.048) 0.037 (0.040) 11 26 286

−0.778*** (0.247) −0.686** (0.285) 0.211 (0.316) −0.363 (0.262) 0.107* (0.059) 0.050 (0.060) −0.221*** (0.070)

−0.006 (0.059)

11 26 286

−0.668*** (0.255) −0.524** (0.265) −0.159 (0.322) −0.166 (0.303) 0.069 (0.049) -0.032 (0.048) −0.188*** (0.069) −0.014 (0.050) −0.132*** (0.037) −0.090*** (0.033) 11 26 286

−0.209 (0.158)

0.163** (0.063)

0.064 (0.049) 0.010 (0.049) 0.108* (0.060) 0.068 (0.053) 14 11 154

−0.071*** (0.022)

0.064 (0.063)

−0.071 (0.100)

−0.492** (0.192) −0.284 (0.195) 0.137 (0.239) −0.359* (0.181) 0.173*** (0.060) 0.069 (0.054) 0.018 (0.052)

0.127** (0.056)

14 11 154

0.096*** (0.024) 0.049** (0.021) −0.168** (0.076) 0.117 (0.095) −0.578*** (0.165) −0.159 (0.182) 0.029 (0.217) -0.165 (0.172) 0.128** (0.060) 0.103* (0.055) 0.062 (0.049) 0.004 (0.051) 0.067 (0.059) 0.096* (0.057) 14 11 154

86

902


Table 6 (Continued) Dependent variable: NAIRU

Adjusted R2 Country FE Period FE DW test

(1)

(2)

(3)

(4)

(5)

(6)

0.605 Yes Yes 0.551

0.602 Yes Yes 0.479

0.625 Yes Yes 0.515

0.610 Yes Yes 1.014

0.617 Yes Yes 0.900

0.601 Yes Yes 0.963

*p < 0.1; **p < 0.05; ***p < 0.01. Notes. (1)–(6) OLS-PCSE. Standard errors in brackets () corrected for autocorrelation in residuals. Cross-section and Year Fixed Effects. Country group in specifications (1)–(3): Austria, Belgium, Denmark, Finland, France, Germany, Ireland, Netherlands, Portugal, Spain, Sweden. In specifications (4)–(6), we additionally include Czech Republic, Poland and Slovak Republic. DW test denotes the Durbin-Watson test statistic on autocorrelation in the residuals. NAIRU, non-accelerating (wage) inflation rate of unemployment; ACCU, capital accumulation; HBOOM, housing boom/bust proxy; LTI, long-term real interest rate; EPL, employment protection legislation; ALMP, active labor market policies; UnD, trade union density; UBR, gross unemployment benefit replacement rate; UBR2, net unemployment benefit replacement rate; TW, tax wedge; MW, minimum wage; TFP, total factor productivity; TOTS, terms of trade shock. t − 1 denotes the first lag of the respective variable; e.g., EPLt − 1 is the first lag of employment protection legislation.

should interact. This poses a problem for an attempt to statistically evaluate the effects of interactions: since there are numerous potential interactions, the inclined researcher is bound to find some that prove statistically significant.” (Stockhammer & Klär, 2011, p. 449). Nevertheless, we accounted for possible interactions by looking at various interaction specifications. No matter whether we include interactions between LMIs only, interactions among LMIs and the other macroeconomic controls only, or all interactions at once, the result is always that there is no systematic evidence that the effects of different LMI variables are reinforced by their interactions. This leads us to the interpretation that the data do not support the argument that LMI interaction terms are crucial for explaining the EC’s NAIRU estimates.11 As a fourth and final robustness check, we implemented variations in the dependent variable. As researchers have noted sizeable revisions in the EC’s NAIRU estimates since the outbreak of the financial crisis (Cohen-Setton & Valla, 2010; Klär, 2013), we also obtained NAIRU data from earlier forecast vintages to assess the robustness of our results with respect to a change in measuring the dependent variable. In columns (1) and (2) of Table 7, we employ the EC’s NAIRU estimates from Autumn 2014 and Autumn 2013 for the time period 1985–2011, respectively. We then proceed with another sensitivity check. In columns (3)–(6), we use the actual unemployment rate as the dependent variable. The change in the inflation rate (!INFL) was introduced as an additional control variable to capture a possible trade-off in the Phillips curve relationship between unemployment and inflation – a feature of the reduced form NAIRU models used in the empirical literature on the determinants of unemployment (e.g. Nickell, 1997; Stockhammer & Klär, 2011). We report the reduced form NAIRU model results for the time period 1985–2011 (column 3) and 2001–2012 (column 5) with country-and period-fixed effects, estimated by OLS-PCSE. Results from the First Difference estimator are shown in columns (4) and (6). In all these variations, it is evident that ACCU and HBOOM are signed as expected, and they are highly significant in all the

11

Detailed results from introducing interaction terms are available at request from the authors.

87

903

P. Heimberger et al. / Journal of Policy Modeling 39 (2017) 883–908 Table 7 Results for 1985–2011: Further robustness checks.

Dependent variable Time period Estimator

(1)

(2)

(3)

(4)

(5)

(6)

NAIRU2014 1985–2011 OLS-PCSE

NAIRU2013 1985–2011 OLS-PCSE

UNEMP 1985–2011 OLS-PCSE

UNEMP 1985–2011 FD

UNEMP 2001–2012 OLS-PCSE

UNEMP 2001–2012 FD

−1.384*** (0.240) −0.459** (0.201) 0.035 (0.072) 1.411* (0.844) −0.026** (0.012) 0.040 (0.044) 0.075*** (0.028)

−0.036 (0.051) −1.634*** (0.237) −0.587** (0.240) 0.149* (0.083) 1.727** (0.749) −0.043*** (0.012) 0.032 (0.042) 0.056** (0.028)

−0.042 (0.026) −1.188*** (0.133) −0.790*** (0.165) 0.002 (0.032) -0.355 (0.702) −0.051*** (0.009) 0.140** (0.070) −0.009 (0.016)

−0.172** (0.077) −1.619*** (0.361) −0.861*** (0.291) 0.042 (0.121) 0.776 (1.560) −0.107*** (0.040) 0.057 (0.118)

−0.095*** (0.033) −1.137*** (0.128) −0.985*** (0.172) −0.077 (0.051) 0.614 (0.621) −0.071*** (0.018) 0.108 (0.082)

−0.001 (0.199) 0.104** (0.044) -0.027 (0.092) 0.072 (0.096)

−0.037 (0.066) −0.001 (0.012) 0.103*** (0.030) 0.017 (0.036) 0.005 (0.066) 14 11 154 0.676 No No 1.384

!INFL ACCU HBOOM LTI EPL ALMP UnD UBR

−1.285*** (0.222) −0.294 (0.193) 0.073 (0.061) 1.115 (0.880) −0.031*** (0.012) 0.018 (0.044) 0.068*** (0.023)

TW UBR2 TFP TOTS

−0.168** (0.068) −0.027 (0.058)

−0.120* (0.072) −0.013 (0.060)

-0.095 (0.075) 0.042 (0.069)

11 27 297 0.587 Yes Yes 0.485

11 27 297 0.610 Yes Yes 0.484

11 27 297 0.659 Yes Yes 0.541

Constant Countries Time periods Observations Adjusted R2 Country FE Period FE DW test

0.053*** (0.019) −0.016 (0.027) 0.064 (0.054) 11 26 286 0.669 No No 1.295

14 12 168 0.652 Yes Yes 0.860

*p < 0.1; **p < 0.05; ***p < 0.01. Notes. (1)–(3) and (5): OLS-PCSE. Standard errors in brackets () corrected for autocorrelation in residuals. Cross-section and Year Fixed Effects. (4) and (6): First Difference Estimator (FD). Country group in specifications (1)–(4): Austria, Belgium, Denmark, Finland, France, Germany, Ireland, Netherlands, Portugal, Spain, Sweden. In specifications (5)–(6), we additionally include Czech Republic, Poland and Slovak Republic. DW test denotes the Durbin-Watson test statistic on autocorrelation in the residuals. NAIRU, non-accelerating (wage) inflation rate of unemployment (Autumn 2015); NAIRU2014, non-accelerating (wage) inflation rate (Autumn 2014); NAIRU2013, non-accelerating (wage) inflation rate UNEMP (Autumn 2013); unemployment rate; !INFL, change in the inflation rate; ACCU, capital accumulation; HBOOM, housing boom/bust proxy; LTI, long-term real interest rate; EPL, employment protection legislation; ALMP, active labor market policies; UnD, trade union density; UBR, gross unemployment benefit replacement rate; UBR2, net unemployment benefit replacement rate; TW, tax wedge; MW, minimum wage; TFP, total factor productivity; TOTS, terms of trade shock.

88

904


reduced form NAIRU models. In contrast, ALMP and (partially) UBR are the only LMI variables that are consistently signed as expected and significant across all models. 7. Discussion: the NAIRU in theory, empirics and policy Our setup for analyzing the econometric determinants of the EC’s NAIRU estimates leads to a confrontation between theory and empirics: while the NAIRU is a theoretically postulated concept, which explains structural unemployment by institutional rigidities, its estimation in the particular context is largely devoid of theoretical rationales, but rather follows a Kalman-Filter approach for detrending time-series data. It is, hence, more of an econometric than an economic exercise. Against this backdrop, our results raise some skepticism with regard to the adequacy of the EC’s NAIRU estimates. However, we cannot provide a conclusive answer about whether the poor fit between NAIRU estimates and their supposed structural explanatory variables is due to principal theoretical deficiencies or rather has to be attributed to a sub-optimal performance of the underlying Kalman-filtering techniques for estimating the NAIRU. Nonetheless, our analysis allows for a closer examination of ‘what’s wrong’ with the EC’s NAIRU estimates. According to the econometric findings discussed in the previous sections, the performance of labor market institutions with regard to explaining the EC’s NAIRU estimates is moderate at best. In the specifications we tested, variables such as the tax wedge, union density, employment protection legislation and minimum wages either do not have the sign expected by standard theory or they are statistically insignificant. This finding contradicts the theoretical framework used by the EC, which assumes that the NAIRU is a good proxy for structural unemployment caused by institutional factors. Orlandi (2012) found for 13 EU countries covering the period 1985–2009 that structural labor market indicators provide a good fit for “[the] structural unemployment rate, as measured by the Commission services (i.e. the so-called NAWRU)” (Orlandi, 2012; p. 1). Similarly, Gianella et al. (2008) had reported that “the set of structural variables provides a reasonable explanation of [the OECD’s Kalman-filtered] NAIRU dynamics over the period 1978–2003” (Gianella et al., 2008; p. 1). In our empirical analysis, we went beyond these earlier studies in many respects. Most crucially, we included additional alternative explanatory factors for the NAIRU and took the years after the financial crisis into account. Our findings are in stark contrast to the assessments by Orlandi (2012) and Gianella et al. (2008). Given that institutional variables underperform in our regressions, we conclude that the NAIRU is not a good proxy for ‘structural unemployment’. This point is reemphasized by the central role that cyclical factors – such as capital accumulation and the housing boom/bust proxy – play in our regressions when it comes to explaining the EC’s NAIRU estimates. Finally, our results provide food for thought regarding more general drawbacks imposed by a ‘one-size-fits-all’ analytical approach to understanding unemployment in Europe, especially as such a framework, quite naturally, translates into a ‘one-size-fits-all’ policy approach. With regard to the analytical aspects we should ask which cyclical variables affect the NAIRU estimates of different countries, and, hence, whether NAIRU estimates might also require context-sensitive interpretations, depending on the country under study. It is remarkable that, although the economic situation of the Eurozone countries exhibits considerable variation, the policy approach suggested by the EC is, nonetheless, quite uniform: ‘structural reforms’ which aim at deregulating labor markets are thereby widely recommended, as member countries are urged to lower structural unemployment by supply side reform (Canton, Grillo, Monteagudo, Pierini, & Turini, 2014 ).

89

905


Fig. 1. Correlation of HBOOM and NAIRU in Spain and Ireland (2001–2012), respectively. Source: AMECO (Autumn 2015); authors’ calculations

We argue that a more nuanced analytical approach, departing from ‘one size fits all’, is in order. Such a new approach would have to allow for the incorporation of a more diverse set of facts, e.g. that Germany’s competitiveness is rather based on sectoral specialization and strong ‘non-price’ competitiveness than on flexible labor markets (Carlin, Glyn, & Van Reenen, 2001; Storm & Naastepad, 2015). Another example would be to consider whether Spain’s and Ireland’s NAIRU before and after the financial crisis might actually have been pro-cyclically driven by the development of their respective housing markets and the repercussions of the boom-bust-cycle in the labor markets, as indicated by the strong relationship between the housing boom/bust proxy and the NAIRU plotted in Fig. 1. By considering different structural and cyclical factors that impact on NAIRU estimates in specific countries, a nuanced approach would allow for devising more flexible, adaptive and versatile policy strategies by more effectively taking into account the economic idiosyncracies of individual countries. Since the NAIRU is used as a proxy for

90

906


‘structural unemployment’ in calculating potential output and structural budget balances in EU member countries, so that it has a direct impact on the scope and evaluation of fiscal policy (see Section 2), a framework considering the role of institutional and cyclical factors in driving NAIRU estimates would be superior to the predominant approach preferred by the EC, which implicitly assumes that Kalman-filter estimates of the NAIRU reflect ‘purely’ structural factors, stripped off any cyclical influences. Our analysis measures that both economists and policymakers have to be cautious in interpreting NAIRU estimates as measure of ‘structural unemployment’ that can unambiguously be used to assess the contribution of the production factor labor to potential output. On the contrary, our econometric findings suggest that the predominant framework for coordinating fiscal policies in the euro area may be dysfunctional, because it crucially rests on an econometric estimate of the NAIRU that is neither consistent with its key theoretical underpinning nor with its political application. Eventually, this poses the risk of using a deficient measure – the output gap – for judging what’s ‘structural’ about fiscal deficits, thereby misinforming policy-making at large. 8. Conclusions This paper has analyzed the determinants of the European Commission’s NAIRU estimates for 14 European OECD countries over the time period 1985–2012. Our main finding is that the NAIRU, as estimated by the EC, is not a good proxy for ‘structural unemployment’. Most indicators of labor market institutions – employment protection legislation, union density, tax wedge and minimum wage – do not explain much; either is their sign inconsistent with the expectation from standard theory, they are statistically insignificant, or their significance is sensitive to the model specification. Only active labor market policies and unemployment benefit replacement rates are consistently signed as expected and significant. The point that NAIRU estimates are not simply driven by institutions is underscored by the finding that cyclical factors – especially capital accumulation and boom-bust patterns in housing markets – are important determinants. This shows that the empirics of the NAIRU are in conflict with the EC’s theoretical framework, in which the NAIRU is interpreted as the structural component of the unemployment rate, independent of all cyclical factors. Our econometric findings are highly relevant for policy making in the EU. First, they point to the fact that increases in the NAIRU cannot simply be attributed to more institutional rigidities with corresponding calls for labor market deregulation to lower ‘structural unemployment’. At the same time, they indicate that the causes for a decline in the NAIRU in a specific country are not always to be found in successful labor market reforms, as downward revisions in the NAIRU might also be driven by cyclical factors. Second, our findings show that there is a considerable risk that NAIRU estimates misinform fiscal policy-making in the EU. The reason is that the NAIRU is used as a proxy for ‘structural unemployment’ in calculating output gaps as a measure for the position of an economy in the business cycle – an indicator that is then transformed into a judgement on how much of the fiscal deficit is due to structural and cyclical factors, respectively. Accordingly, flawed estimates of the NAIRU as the ‘structural unemployment rate’ can lead to miscalculations of the size of the structural deficit and inappropriate fiscal policies. References Arestis, P., Baddeley, M., & Sawyer, M. (2007). The relationship between capital stock, unemployment and wages in nine EMU countries. Bulletin of Economic Research , 59(2), 125–148.

91

907


Arpaia, A., Kiss, A., & Turrini, A. (2014). Is unemployment structural or cyclical? Main features of job matching in the EU after the crisis. European Economy—Occasional Papers, 527. Avdagic, S., & Salardi, P. (2013). Tenuous link: Labour market institutions and unemployment in advanced and new market economies. Socio-Economic Review, 4(11), 739–769. Baccaro, L., & Rei, D. (2007). Institutional determinants of unemployment in OECD countries: Does the deregulatory view hold water? International Organization, 61(3), 527–569. Baker, D., Glyn, A., Howell, D., & Schmitt, J. (2005). Labor market institutions and unemployment: A critical assessment of the cross-country evidence. In D. Howell (Ed.), Fighting unemployment. The Limits for Free Market Orthodoxy, Oxford: Oxford University Press. Ball, L., & Mankiw, N. (2002). The NAIRU in theory and practice. Journal of Economic Perspectives, 16(4), 115–136. Bassanini, A., & Duval, R. (2006). The determinants of unemployment across OECD countries: Reassessing the role of policies and institutions. OECD Economic Studies, 2006(1), 7–86. Beck, N., & Katz, J. (1995). What to do (and not to do) with time-series cross-section data. The American Political Science Review, 89(3), 634–647. Belot, M., & van Ours, J. (2004). Does the recent success of some OECD countries in lowering their unemployment rates lie in the clever design of their labor market reforms? Oxford Economic Papers, 56(4), 621–642. Bertola, G., Blau, F., & Kahn, L. (2007). Labor market institutions and demographic employment patterns. Journal of Population Economics, 20(4), 833–867. Blanchard, O. (2006). European unemployment: The evolution of facts and ideas. Economic Policy, 21(45), 5–59. Blanchard, O., & Wolfers, J. (2000). The role of shocks and institutions in the rise of European unemployment: The aggregate evidence. Economic Journal, 110(462), C1–33. Canton, E., Grillo, I., Monteagudo, J., Pierini, F., & Turini, A. (2014). The role of structural reform for adjustment and growth. ECFIN economic briefs 34/2014. European Commission. Carlin, W., Glyn, A., & Van Reenen, J. (2001). Export market performance of OECD countries: An empirical examination of the role of cost competitiveness. Economic Journal, 111(468), 128–162. Choi, I. (2001). Unit root tests for panel data. Journal of International Money and Finance, 20(2), 249–272. Cohen-Setton, J., & Valla, N. (2010). Unnoticed potential output revisions and their impact on the stimulus/austerity debate. CEPR Policy Portal Publication (17 August 2010). Durbin, J., & Koopman, S.-J. (2012). Time series analysis by state space methods (2nd edition). Oxford University Press. ECFIN. (2013). Building a strengthened fiscal framework in the European union: A guide to the stability and growth pact. European economy—occasional papers 150. European Commission. Elmeskov, J., Martin, J., & Scarpetta, S. (1998). Key lessons for labor market reforms: evidence from OECD countries’ experience. Swedish Economic Policy Review, 2(5), 205–252. European Central Bank. (2015). Inflation and unemployment in Europe. In Conference Proceedings (21–23 May 2015). European Commission. (2013). Labour market developments in Europe. European economy 6/2013. European Commission. European Commission. (2014). New estimates of Phillips curves and structural unemployment in the Euro Area. Quarterly report on the euro area 13/1. European Commission. European Union. (2011). Regulation (EU) No 1175/2011 of the European Parliament and of the Council of 16 November 2011, amending Council Regulation (EC) No 1466/97 on the strengthening of the surveillance of budgetary positions and the surveillance and coordination of economic policies. Official Journal of the European Union L 306/12. European Union. Felipe, J., & McCombie, J. (2014). The aggregate production function: ‘Not Even Wrong’. Review of Political Economy, 26(1), 60–84. Fiscal Compact. (2012). Treaty on stability, coordination and governance in the economic and monetary union. T/scg/en 1. European Commission. Flaig, G., & Rottmann, H. (2013). Labour market institutions and unemployment: An international panel data analysis. Empirica, 40(4), 635–654. Franz, W. (2005). Will the (German) NAIRU please stand up? German Economic Review, 6(2), 131–153. Friedman, M. (1968). The role of monetary policy. The American Economic Review, 58(1), 1–17. Gianella, C., Koske, I., Rusticelli, E., & Chatal, O. (2008). What drives the NAIRU? Evidence from a panel of OECD countries OECD economics department working papers 649. OECD Publishing. Harvey, A. (1990). Forecasting, structural time series models and the Kalman filter. Cambridge: Cambridge University Press.

92

908


Havik, K., Morrow, K., Orlandi, F., Planas, C., Raciborski, R., Roeger, W., et al. (2014). The production function methodology for calculating potential growth rates and output gaps. European economy—Economic papers 535, Directorate General Economic and Financial Affairs (DG ECFIN). European Commission. Hodrick, R., & Prescott, E. (1997). Postwar US business cycles: An empirical investigation. Journal of Money, Credit and Banking, 29(1), 1–16. Howell, D., Baker, Dean, Glynn, A., & Schmitt, J. (2007). Are protective labor market institutions at the root of unemployment? A critical review of the evidence. Capitalism and Society, 2(1), 1–73. International Monetary Fund. (2003). Unemployment and labor market institutions: Why reforms pay off. pp. 129–150. IMF World Economic Outlook. April 2013. Kaiser, R., & Maravall, A. (2001). Some basic limitations of the Hodrick–Prescott filter: Measuring business cycles in economic time series. Lecture Notes in Statistics, 154, 87–115. Kalman, R. (1960). A new approach to linear filtering and prediction problems. Journal of Basic Engineering, 82(1), 35–45. Klär, E. (2013). . pp. 33–40. Potential economic variables and actual economic policies in Europe Intereconomics: Review of European economic policy (48) Springer., 1. Laubach, T. (2001). Measuring the NAIRU: Evidence from seven economies. The Review of Economics and Statistics, 83(2), 218–231. Lendvai, J., Salto, M., & Thum-Thysen, A. (2015). Structural unemployment vs. NAWRU: Implications for the assessment of the cyclical position and the fiscal stance. European Economy –Economic Papers 552, Directorate General Economic and Financial Affairs (DG ECFIN). European Commission. Maddala, G., & Wu, S. (1999). A comparative study of unit root tests with panel data and a new simple test. Oxford Bulletin of Economics and Statistics, 61, 631–652. Mourre, G., Astarita, C., & Princen, S. (2014). Adjusting the budget balance for the business cycle: The EU methodology. European economy—economic papers 536, Directorate General Economic and Financial Affairs (DG ECFIN). European Commission. Nickell, S. (1997). Unemployment and labor market rigidities: Europe versus North America. Journal of Economic Perspectives, 11(3), 55–74. Nickell, S. (1998). Unemployment: Questions and some answers. Economic Journal, 108(448), 802–816. Nickell, S., Nunziata, L., & Ochel, W. (2005). Unemployment in the OECD since the 1960. What do we know? Economic Journal, 115(500), 1–27. OECD. (1994). The OECD jobs study. Facts, analysis, strategies. OECD study. OECD. Orlandi, F. (2012). Structural unemployment and its determinants in the EU countries. European Economy –Economic Papers 455, Directorate General Economic and Monetary Affairs (DG ECFIN). European Commission. Palacio-Vera, A., Aguilar, I., de la Cruz, E., & Martinez-Canete, A. (2006). Capital stock and unemployment: Searching for the missing link. Economics working paper archive 475. Levy Economics Institute. Phelps, E. (1967). Phillips curves, expectations of inflation and optimal unemployment over time. Economica, 34(135), 254–281. Planas, C., & Rossi, A. (2015). Program GAP. Technical description and user-manual (Version 4.4). JRC scientific and technical reports. Italy: Joint Research Centre Ispra. April. Siebert, H. (1997). Labor market rigidities: At the root of unemployment in Europe. Journal of Economic Perspectives, 11(3), 37–54. Staiger, D., Stock, J., & Watson, M. (1997). The NAIRU, unemployment and monetary policy. Journal of Economic Perspectives, 11(1), 33–49. Stockhammer, E., Guschanski, A., & Köhler, K. (2014). Unemployment, capital accumulation and labour market institutions in the great recession. European Journal of Economics and Economic Policies: Intervention, 11(2), 182–194. Stockhammer, E., & Klär, E. (2011). Capital accumulation, labour market institutions and unemployment in the medium run. Cambridge Journal of Economics, 35(2), 437–457. Storm, S., & Naastepad, C. (2015). Crisis and recovery in the german economy: The real lessons. Structural Change and Economic Dynamics, 32(C), 11–24. Tereanu, E., Tuladhar, A., & Simone, A. (2014). Structural balance targeting and output gap uncertainty. IMF working papers 14/107. International Monetary Fund. Vergeer, R., & Kleinknecht, A. (2012). Do flexible labor markets indeed reduce unemployment? A robustness check. Review of Social Economy, 70(4), 451–467. Watson, M. (2014). Inflation persistence, the NAIRU, and the great recession. American Economic Review, 104(5), 31–36. Zhou, H., Dekker, R., & Kleinknecht, A. (2011). Flexible labor and innovation performance: Evidence from longitudinal firm-level data. Industrial and Corporate Change, 20(3), 941–968.

93

Conclusions of the dissertation This dissertation has contributed to academic studies of fiscal policy coordination in Europe. At its core, the dissertation consists of three peer-reviewed articles. The essays make novel contributions to the existing literature regarding two main subjects of European fiscal policies. The first main research interest has been to shed new light on the role of modelbased estimations of the ‘structural deficit’ within the EU’s fiscal regulation framework, as this indicator plays an important role in shaping fiscal policy practices and influencing macroeconomic outcomes. A second major topic has been to provide econometric estimations on the link between fiscal consolidation measures and economic activity in the euro area.

1. The context-sensitivity of ‘the’ fiscal multiplier and research limitations The empirical design for estimating fiscal multipliers in the euro area used in chapter 2 has been guided by theoretical reasoning on the size of fiscal multipliers. The main finding is that average fiscal multipliers in the euro area’s economies during the Eurozone Crisis (with particular focus on the time period 2011 to 2013) were substantially higher than unity (in a range from 1.4 to 2.1, depending on the data source one uses to identify size and timing of fiscal consolidation measures). It should be reemphasized that the size of fiscal multipliers – which measure the effects of exogenous changes in fiscal policy on GDP growth – are highly dependent on various context factors such as: the position of an economy within the business cycle; the question of whether monetary policy is constrained by the zero lower bound of nominal interest rates; the exchange-rate regime; spillover effects from other economies etc. In other words: “Asking what ‘the’ government spending multiplier is […] is like asking what ‘the’ temperature is. Both vary over time and space. The really interesting intellectual questions involve the extent to which the whole set of other important factors causes the multiplier to vary”. (Carroll 2009, p. 246) In this dissertation, I have argued that there are important factors that caused multipliers in euro area countries to be relatively high during the Eurozone Crisis: the recovery from the financial crisis-related recession was everything but complete (implying the existence of substantial economic slack), deleveraging in the private sector was still ongoing, and Eurozone countries basically lacked the (monetary) policy tools to counteract the contractionary fiscal consolidation impulse. Other studies from the academic literature also find that fiscal consolidation measures in the euro area were highly 94 26

contractionary, implying substantial negative growth effects (e.g. Gechert et al. 2016; Banque de France 2017; House et al. 2017). By using cross-sectional data variation in the intensity of fiscal consolidation efforts across countries, chapter 2 has shed new light on the effectiveness of fiscal policy under special macroeconomic and institutional circumstances. My results on the size of multipliers in the euro area during the Eurozone Crisis, however, cannot be generalized across time and space: empirical estimations of ‘the’ fiscal multiplier and their interpretation – which need to be guided by theoretical arguments – have to be contextsensitive.

2. The EU’s fiscal regulation framework and the pro-cyclicality of European fiscal policies While chapter 2 of this dissertation has analyzed the growth effects of pro-cyclical fiscal policies during the Eurozone Crisis, chapter 3 has approached the question about the cyclicality and context-sensitivity of fiscal policy from a different angle. By turning to the second main topic of the dissertation – the role of model estimates in assessing the fiscal effort that is required of a country to comply with the rules in the EU’s fiscal regulation framework –, we have shown that fiscal policy coordination in Europe already had a procyclical bias before the financial crisis hit the euro area: the model estimates tend to provide more room for fiscal maneuvering when an economy is booming. I have shown that this characteristic is due to a pro-cyclical bias in statistical filtering methods that are used to obtain estimates for theoretically postulated but essentially unobservable variables (the ‘nonaccelerating inflation rate of unemployment’, which is directly linked to estimates of ‘potential output’ and the ‘output gap’). The European Commission’s model for estimating the ‘potential output’ of individual economies – defined as the level of output at which all production factors are utilized in a non-inflationary way – produces estimates that are disproportionately driven by the most recent macroeconomic observations (the so-called endpoint bias). In particular, chapter 3 has pointed out that the non-accelerating inflation rate of unemployment (NAIRU) – which serves as a proxy for ‘structural unemployment’ – which is posited to be independent of cyclical factors – in the potential output model tends to be revised downwards when the economy is booming and to be revised upwards when the economy is in crisis. The underlying revisions in model-based estimates lead to a systematic bias in the assessment of fiscal policies, as booming economies are granted more room for policy maneuvering in good times, while crisis economies lose space in direct relation to the 95 27

depth of the downswing. By means of an in-depth econometric analysis, furthermore, chapter 4 of this dissertation has shown that cyclical factors are important determinants in explaining the evolution of the Commission’s NAIRU estimates – a finding that stands in contrast to the underlying theoretical framework, according to which indicators of labor market institutions should be able to explain the NAIRU values produced by the Commission’s model. As a consequence, estimates of the ‘structural deficit’ promote pro-cyclical fiscal policies within the EU’s fiscal regulation framework that aggravate policy coordination problems within the euro area. Notably, such pro-cyclical fiscal policies run against the basic Keynesian insight that discretionary fiscal policy should neither be geared towards further inflating the boom in good times nor to deepening the crisis in bad times, but should instead be dedicated to dampening unsustainable overheating and to supporting the economy in times when economic resources are underutilized. In 1937, John Maynard Keynes wrote: “The boom, not the slump, is the right time for austerity at the Treasury.” (Keynes 1937, p. 390) The experiences with procyclical fiscal tightening since 2010 lend support to Keynes’ insight (which he already formulated back in 1932 in a letter to The Times; see MacGregor et al. 1932) that the state should not cut spending when individuals are already hoarding money during a downswing, since fiscal consolidation in bad times deprives the economy of domestic demand that would otherwise be available to increase output and lower unemployment – this Keynesian argument has also been prominently made by Krugman (2015) and Wren-Lewis (2015), among others, in contemporary debates on the effects of fiscal austerity. Careful readings of the empirical evidence have refuted influential contemporary pro-austerity arguments according to which fiscal consolidation measures – even in bad times – were bound to improve private-sector confidence and support economic recovery by means of ‘expansionary austerity’ (e.g. Blyth 2013; Dellepiane-Avellaneda 2015; Farrell and Quiggin 2017). Fiscal consolidation measures are always contractionary (Guajardo et al. 2014; Jorda and Taylor 2016), but the size of ‘the’ fiscal multiplier varies: as suggested by the econometric estimations provided in this dissertation, average multipliers were much larger than in boom times when simultaneous fiscal consolidation efforts were launched in Eurozone countries in 2011. By weakening large parts of the Eurozone’s economy and by pushing debt-to-GDP ratios upwards (due to the decrease in nominal GDP), the goal of stabilizing fiscal deficits and government-debt-ratiosto-GDP was also missed (e.g. Mazzolini and Mody 2014; De Grauwe and Ji 2014).

96 28

An alternative to the actual European practice of fiscal policy would have been to improve policy coordination in the euro area in order to support economic recovery in the crisis-ridden periphery countries – a route that was effectively undermined by the tightening of the fiscal rules after the financial crisis (e.g. Arestis and Sawyer, 2015). An alternative approach would have required the ECB to step in earlier as the lender of last resort in order to stop speculation on the government bonds of single countries (e.g. De Grauwe 2012; Saka et al., 2015). Against this background, fiscal policy could have struck a more appropriate balance between reaching deficit targets and fulfilling employment and social goals. While an alternative strategy would have required a more gradual fiscal policy approach in the Eurozone’s periphery countries, the core countries could have contributed to the euro area recovery by implementing additional stimulus measures from 2010 onwards (e.g. De Grauwe, Ji, 2013; Mody, 2015; Stiglitz, 2016). In addition to the arguments on the pro-cyclicality of European fiscal policies and their feedback effects with macroeconomic development, the results in chapter 3 indicate that estimates of NAIRU and potential output provide policy guidance that effectively undermines the spirit of the Stability and Growth Pact, which stipulates that economic convergence among member countries should be fostered. In fact, the Commission’s model estimates, which systematically deny that there is any fiscal space for crisis countries to employ countercyclical policies in order to overcome the crisis and to renew their economic structures, promote further divergence that serves to destabilize the euro area in the long-run.

3. Conclusions related to institutional reform discussions in Europe What kind of conclusions can we draw from this dissertation in terms of improving fiscal policy coordination in the euro area against the background of institutional reform discussions? From a theoretical perspective, a move towards fiscal union may be seen as desirable: a common fiscal policy, built on a sizeable Eurozone budget (amounting to several percentage points of Eurozone GDP), would allow for automatic stabilization to counteract (asymmetric) swings in the business cycle in individual member countries; fiscal integration could help to cushion macroeconomic crises and mitigate the problem of coordinating diverse national fiscal policies (e.g. Farhi and Werning 2017; Juncker et al. 2015). A combination of a fully-fledged banking union and a centralized, sizeable Eurozone budget would break the sovereign-banking nexus that has triggered doom-loops between sovereign risk and bank risk 97 29

pushing several Eurozone countries close to the brink of insolvency (e.g. Beck 2012; Mody and Sandri 2012; Breuss 2016; De Rynck 2016). Furthermore, if Eurozone countries were to issue common and safe bonds, the fragility of individual member countries in terms of speculation against their government bonds could be reduced, because the ECB would then be able to credibly fulfill its crucial role as the lender of last resort (De Grauwe 2012). Moving towards fiscal and political union, however, might not be in the political cards (e.g. European Commission 2017). One can think of several reasons why attempts to move towards fiscal and political union might not turn out to be successful in the future: policy-makers might be unwilling to use the political capital that would be necessary to establish more fiscal integration (Sandbu 2015); Eurozone core countries such as Germany and Austria might simply oppose fiscal integration as they ignore the interests of periphery countries and continue to see the enforcement of ‘fiscal discipline’ within the existing institutional frameworks as crucial for maintaining their export-led growth models (Iversen et al. 2016); politicians willing to integrate further might prove unable to convince their skeptical electorate of the desirability of institutional reforms (Franchino and Segatti 2017); and so on. Under the assumption that further substantial fiscal integration will not be possible in the foreseeable future, an obvious conclusion that can be drawn from this dissertation would be to replace or improve the European Commission’s macroeconomic model, which underlies the estimation of the ‘structural deficit’. By improving the potential output model – in particular by finding solutions to avoiding pro-cyclical outcomes –, one could allow for counter-cyclical fiscal policies at a national level in the future (within the existing rules-based regulations in the Stability and Growth Pact and the Fiscal Compact). Economic knowledge – such as knowledge on the ‘output gap’ of a particular economy –, if accepted as authoritative, can be performative, i.e. shape the behavior of policymakers who are exposed to the knowledge (e.g. Hirschman and Berman 2014; MacKenzie and Millo 2003). In this dissertation, I have argued that the European Commission’s potential output model has performative effects, as its institutionalization in the existing fiscal regulation framework biases and shapes relevant fiscal policy choices, thereby affecting macroeconomic outcomes. The existing fiscal rules in the EU have triggered and amplified suboptimal fiscal policy decisions that caused a deepening of the crisis in large parts of the euro area (from 2010 onwards). Even without a move towards fiscal and political union, however, the Eurozone could function more properly than since 2010 by raising (public) awareness of the importance of technicalities in the coordination of national fiscal policies. An alternative route 98 30

in the quest to reforming the fiscal regulation framework would be to make large changes to the existing rules by decreasing their reliance on rather intransparent and revision-prone model-based estimates of the ‘structural deficit’ (e.g. Claeys et al. 2016).

4. Future research on European fiscal policies Future research could focus on how to change the estimation of potential output within the EU’s fiscal regulation framework in order to improve macroeconomic plausibility of policyrelevant estimates (e.g. Hristov et al. 2017a; Hristov et al. 2017b; Coibion et al. 2017). Furthermore, additional research efforts are needed to increase our knowledge of “the role of economics in the cognitive infrastructure of policymaking” and of economic models as “devices for seeing and deciding” (Hirschman and Berman 2014, p. 779). Although this dissertation has provided new contributions, the role of economic models in shaping political processes remains a research area that is underdeveloped. In particular, it seems fruitful to develop a research agenda regarding the performativity of macroeconomic models in central banking on which relatively little is known (e.g. Braun 2014; Braun 2016; Christophers 2017). Regarding the second central topic of this dissertation – estimating the effects of fiscal policy on economic activity –, an obvious extension to the chosen econometric approach would be to use a different research design in order to estimate fiscal multipliers for particular countries (e.g. Auerbach and Gorodnichenko 2012; Auerbach and Gorodnichenko 2017). In general, economists still have much to learn from the fiscal policy experiences of European countries since the Eurozone Crisis. Learning from research on past crises may be seen as a (partial) guide to avoiding deep macroeconomic troubles in the future.

9931

References Auerbach, A.; Gorodnichenko, Y. (2012): Measuring the Output Responses to Fiscal Policy, American Economic Journal: Economic Policy, 4(2), 1-27. Auerbach, A.; Gorodnichenko, Y. (2017): Fiscal Multipliers in Japan, Research in Economics, 71, 411-421. Arestis, P.; Sawyer, M. (2015): The Eurozone needs a complete make-over of its fiscal policies, in: Bitzenis, A.; Karagiannis, N.; Marangos, J. (2015) (ed.): Europe in Crisis. Problems, challenges, and alternative perspectives, London: Palgrave Macmillan, 111-120. Banque de France (2017): The cost of deficiencies in euro area economic policy coordination, in: Banque de France (2017): Quarterly Selection of Articles. Banque de France Bulletin, Summer 2017, 19-31. Beck, T. (ed.) (2012): Banking Union for Europe – Risks and Challenges, London: CEPR E-book. Blyth, M. (2013): Austerity. The History of a Dangerous Idea, New York: Oxford University Press. Braun, B. (2014): Central bank agency and monetary governability in the euro area: Governing through money, trust, and expectations, Dissertation, Coventry: University of Warwick. Braun, B. (2016): From performativity to political economy: index investing, ETFs and asset manager capitalism, New Political Economy, 21(3), 257-273. Breuss, F. (2016): The European Banking Union: The Last Building Block towards a New EMU?, in: Guderzo, M.; Bosco, A. (2016): A Monetary Hope for Europe. The Euro and the Struggle for the Creation of a New Global Currency, Florence: Firenze University Press. Carroll, C. (2009): Comments and discussion on “By how much does GDP rise if the government buys more output?”, Brookings Papers on Economic Activity, 2009(2), 232-249. Christophers, B. (2017): The performativity of the yield curve, Journal of Cultural Economy, 10(1), 63-80. Claeys, G.; Darvas, Z.; Leandro, A. (2016): A Proposal to revive the European Fiscal Regulation Framework, Bruegel Policy Contribution 2016/07. Coibion, O.; Gorodnichenko, Y.; Ulate, M. (2017): The cyclical sensitivity of potential output, NBER Working Paper No. 23580. De Grauwe, P. (2012): The Governance of a Fragile Eurozone, Australian Economic Review, 45(3), 255–268. De Grauwe, P.; Ji, Y. (2013): From Panic-Driven Austerity to Symmetric Macroeconomic Policies in the Eurozone, Journal of Common Market Studies, 51(S1), 31–41. De Grauwe, P.; Ji, Y. (2014): How Much Fiscal Discipline in a Monetary Union?, Journal of Macroeconomics, 39(B), 348-360.

10032

Dellepiane-Avellaneda, S. (2015): The political power of economic ideas: the case of “Expansionary Fiscal Contractions”, The British Journal of Politics and International Relations, 17(3), 391-418. De Rynck, S. (2016): Banking on a union: the politics of changing eurozone banking supervision, Journal of European Public Policy, 23(1), 119-135. European Commission (2017): Reflection paper on the deepening of the Economic and Monetary Union, COM(2017) 291, Brussels: European Commission. Farhi, E.; Werning, I. (2017): Fiscal Unions, American Economic Review, 107(12), 3788-3834. Farrell, H.; Quiggin, J. (2017): Consensus, Dissensus, and Economic Ideas: Economic Crisis and the Rise and Fall of Keynesianism, International Studies Quarterly, 61(2), 269-283 Franchino, F.; Segatti, P. (2017): Public opinion on the Eurozone fiscal union: evidence from survey experiments in Italy, Journal of European Public Policy, forthcoming. Gechert, S.; Hughes-Hallett, A.; Rannenberg, A. (2016): Fiscal multipliers in downturns and the effects of Euro Area consolidation, Applied economics letters, 23(16), 1138-1140. Guajardo, J.; Leigh, D.; Pescatori, A. (2014): Expansionary austerity? International evidence, Journal of the European Economic Association, 12(4), 949-968. Hirschman, D.; Berman, E. (2014): Do economists make policies? On the political effects of economics, Socio-Economic Review, 12, 779-811. House, C.; Proebsting, C.; Tesar, L. (2017): Austerity in the Aftermath of the Great Recession, NBER Working Paper No. 23147. Hristov, A.; Raciborski, R.; Vandermeulen, V. (2017a): Assessment of the Plausibility of the Output Gap Estimates, European Economy – Economic Brief 023/2017, Brussels: European Commission. Hristov, A.; Planas, C.; Roeger, W.; Rossi, A. (2017b): NAWRU estimation using structural labour market indicators, European Economy – Discussion Paper 069, Brussels: European Commission. Iversen, T.; Soskice, D.; Hope, D. (2016): The Eurozone and Political Economic Institutions, Annual Review of Political Science, 19, 163-185. Jorda, O.; Taylor, A. (2016): The Time for Austerity: Estimating the Average Treatment Effect of Fiscal Policy, Economic Journal, 126, 219-255. Juncker, J.; Tusk, D.; Dijsselbloem, J.; Draghi, M.; Schulz, M. (2015): Completing Europe’s Economic and Monetary Union, The Five Presidents’ Report (Chapter 4), Brussels, 22 June 2015, 13-16. Keynes, J. (1937): How To Avoid a Slump, The Times (January 12-14th 1937), reprinted in: Moggridge, D. (1973): The Collected Writings of John Maynard Keynes, London: Macmillan, 384-395.

10133

Krugman, P. (2015): The austerity delusion, The Guardian (April 29th 2015), reprinted in: Skidelsky, R.; Fraccaroli, N. (2017): Austerity Vs Stimulus: The Political Future of Economic Recovery, London: Palgrave Macmillan, 35-52. MacGregor, D.; Pigou, A.; Keynes, J.; Layton, W.; Salter, A.; Stamp, J. (1932): Money for Productive Investment, The Times (October 17th 1932, p. 13), reprinted in: in: Skidelsky, R.; Fraccaroli, N. (2017): Austerity Vs Stimulus: The Political Future of Economic Recovery, London: Palgrave Macmillan, 11-13. MacKenzie, D.; Millo, Y. (2003): Constructing a market, performing theory: the historical sociology of a financial derivatives exchange, American Journal of Sociology, 109(1), 107-145. Mazzolini, G.; Mody, A. (2014): Austerity tales: the Netherlands and Italy, Bruegel Blog Post (October 3rd 2014), http://bruegel.org/2014/10/austerity-tales-the-netherlands-and-italy/ [last download on December 27th 2017]. Mody, A.; Sandri, D. (2012): The eurozone crisis: how banks and sovereigns came to be joined at the hip, Economic Policy, 27(70), 201-230. Mody, A. (2015): Living Dangerously Without A Fiscal Union, Bruegel Working Paper 2015/03. Saka, A.; Fuertes, A.; Kalotychou, E. (2015): ECB Policy and Eurozone Fragility: Was De Grauwe Right?, Journal of International Money and Finance, 54, 168-185. Sandbu, M. (2015): Europe’s Orphan. The Future of the Euro and the Politics of Debt, Princeton: Princeton University Press. Stiglitz, J. (2016): The Euro. How a common currency threatens the future of Europe, New York: Norton & Company. Wren-Lewis, S. (2015): We already have a simple and conventional story to explain the weak recovery, VoxEU (January 30th 2015), http://voxeu.org/article/fiscal-policy-explains-weakrecovery [last download on December 27th 2017].

10234

Dissertation

Dissertation

Suggest Documents

dissertation

Dissertation

dissertation

Dissertation

Dissertation

dissertation

Dissertation

dissertation

dissertation

DISSERTATION

DISSERTATION

DISSERTATION

Dissertation

DISSERTATION

Dissertation

Dissertation

Dissertation

Dissertation

DISSERTATION

Dissertation

DISSERTATION

DISSERTATION

dissertation

Dissertation