The core-periphery pattern of Europe: a re-examination using
a Structural VAR
Tamara Doeve
S2384027
MSc Economics
Faculty of Economics and Business Economics, University of Groningen
EBM877A20
January 19th, 2018
Supervisor: Sebastiaan Pool
Abstract
This paper revisits the question of the core-periphery pattern in the European Monetary Union. The paper from Bayoumi and Eichengreen (1993) is taken as a point of departure for the analysis. This paper uses a bivariate structural VAR model, where the aggregate demand and supply shocks are identified from the VAR residuals using a long-run restriction proposed by Blanchard and Quah (1989). The results show that the countries within the European Monetary Union are not integrated enough to be able to cope with idiosyncratic shocks as efficiently as countries with an independent monetary union. The results do not show a clear core-periphery pattern as identified by Bayoumi and Eichengreen (1993).
Keywords: Business cycle synchronization, European Monetary Union, Structural VAR
2 1. Introduction
Nearly 16 years have passed since the introduction of the Euro and nearly 24 years since the completion of the single market. The recent Sovereign Debt Crisis showed the danger of countries not possessing an independent monetary policy. European countries faced the collapse of financial institutions, high government debt and rapidly rising bond yield spreads in government securities. The credibility of European Monetary Unions’ (EMU) sovereigns was challenged, as the Eurozone member states of Greece, Spain, Ireland, Portugal and Cyprus were unable to repay or refinance their government debt, or bail out their insolvent banks as the European Central Bank (ECB) could not serve as a lender of last resort. A period of uncertainty on the solvency of the sovereigns followed, pursued closely with doubts about the sustainability of a common currency. This led to a renewed interest in the theories surrounding a common currency and the role of the ECB in coping with asymmetric shocks. Concerns about the efficiency of the ECB in dealing with idiosyncratic shocks were pointed out even before the EMU was established (Bayoumi & Eichengreen, 1993). The first discussions of area’s adopting common currency date back to Mundell (1961), where he introduced his idea of an Optimal Currency Area (OCA). He emphasized that an OCA is favourable if disturbances are distributed symmetrically across countries. A specified set of criteria could help countries to handle idiosyncratic shocks, as these countries do not have an independent monetary policy themselves. After the attempt of Mundell (1961) to define parameters of a successful currency union, many researchers seek to enlarge that set of criteria (McKinnon, 1963; Kenen, 1969).
3 In light of the attention attracted to the sustainability of the EMU, this paper analyses the presence of idiosyncratic shocks in the EMU and their ability to adjust to these shocks. Correlations of supply and demand shocks between the EMU members are examined, as well as the impulse responses and speed of adjustment of those countries to exogenous supply and demand shocks. In practice the supply and demand shocks are recovered from a bivariate VAR model. By imposing that demand shocks cannot have a permanent effect on output, one can identify the different shocks from the VAR residuals. This restriction can be justified with a New-Keynesian model of aggregate supply and demand curves. This paper finds that the speed of adjustment after an idiosyncratic shock is faster for countries with an independent monetary policy. Which indicates that factors of production are not mobile enough to help countries without an independent monetary policy to accommodate to shocks as fast as countries with an independent monetary policy. The results do not show a clear core-periphery pattern as identified by Bayoumi and Eichengreen (1993). This could be due to misspecification of the model, as the supply shocks give conflicting results compared to previous literature. Cooley and Dwyer (1998) note that it is difficult to proprerly identify supply shocks in a bivariat structural VAR, as potential ommitted variables are in the demand and supply shocks (innovations).
4 The remainder of this paper is structured as follows: Section 2 gives an overview of how the literature of OCA has developed, and in detail for the EU. Section 3 provides an outline for the methodology needed to recover individual demand and supply disturbances. Section 4 presents the data collected by means of descriptive statistics. Section 5 discusses the results of the structural VAR and checks the sensitivity of the results. Section 6 concludes this paper, and discusses the limitations and recommendations for future research.
2. Literature Review
2.1. Dynamics of a common currency
The central question driving the literature on OCA’s regards the costs and benefits of sharing a common currency. The foremost benefits of an OCA are gaining free factor mobility and a stable exchange rate. Where factor mobility refers to the ability to move factors of production, labour or capital between countries either within industries or across industries. Free factor mobility allows the factors of production to move to different industries where they can be more productive. A stable exchange rate is beneficial for a country to avoid currency fluctuations as this can cause more uncertainty for firms heavily engaged in trade. The uncertainty of exchange rate fluctuations can reduce the incentive for firms to invest in their export capacity. Accordingly, the higher the trade intensity, the greater the efficiency gains of adopting a single currency (De Grauwe, 2016).
5 and for country B to devaluate internally. Hence, country B should lower its wages and risk high unemployment to stabilize the balance-of-payments, which is very costly for a country (De Grauwe, 2016). In addition, dissimilarity between inflation rates is an important factor that can complicate the use of a common monetary policy. Large differences in inflation rates among regions affects the competitiveness levels of those regions. Regions that encounter high inflation rates are losing competitiveness as their products become more expensive. Whereas regions that face relative low inflation rates experience an increase in competitiveness. The larger the difference in inflation rates in regions, the harder it is for the central bank to set monetary policy that benefits both regions (Fidrmuc, 2004). There is a ‘one size does not fit all’ problem, so the question rises, when is a monetary union optimal?
2.2. Development of OCA Theory
The first discussions date back to Mundell (1961), McKinnon (1963) and Kenen (1969). In 1961, Mundell introduced his theory of optimum currency areas and provides a set of criteria to determine whether countries or regions are suited to a common currency. He outlined that regions wanting to adopt a common currency should be exposed to symmetric shocks, or possess mechanisms for the absorption of asymmetric shocks. In particular, countries that join a common currency area should have an adequate degree of international factor mobility. Labour and capital should be sufficiently mobile to maintain similar rates of employment and inflation in different regions after an asymmetric shock. Next to this wages should be flexible, especially in the short run, because it helps to facilitate the adjustment process following a shock. A country with an independent monetary policy can compensate for its rigid prices and wages by means of exchange rate adjustments, unfortunately, a member of a currency union does not have this tool. An OCA is optimal if the prevailing exchange rate regime can maintain a stable external balance without causing unemployment or demand induced wage inflation. Hereby is Mundell (1961) one of the first to jointly consider three objectives; the maintenance of full employment, the maintenance of balanced international payments and the maintenance of a stable internal average price level in a single currency area. Before his work, possible conflicts between preserving full employment and a stable balance-of-payments have been well discussed (Meade, 1951).
6 various industries. A high degree of factor mobility among industries helps to preserve the three objectives of Mundell (1961) and decrease the costs of a loss of independent monetary policy. Kenen (1969) adds the relevance of product diversification in coping with idiosyncratic shocks. A highly diversified economy is likely better able to compensate for a negative shock in one sector by a positive shock in another industry. He also stresses the usefulness of fiscal transfers in unions with diversified business cycles. Countries that face a negative demand shock receive less taxes and income, this can be offset with countries that counter a positive demand shock when the currency area is fiscally integrated. These criteria all mention the importance of business cycle correlations and many other studies add criteria for the OCA (Kenen, 2000).1
A revival in the empirical testing of OCA theory preceded the introduction of the monetary union in Europe. In defining an OCA, subsequent research focuses on determining the optimality criteria for forming a currency area by expanding the conditions proposed by Mundell (1961). In light of Mundell’s framework, the incidence of disturbances across regions is critical determinant of the design of currency areas. Therefore, it is important to know if the business cycles in EU are converging, diverging or are synchronized sufficiently. Because, the more synchronized the business cycles are, the less the cost are of losing an independent monetary policy. Two views have been put forward on this issue.
In Krugman’s (1993) view, trade liberalization facilitates more specialization in countries or regions according to comparative advantage theory. More regional concentration of industries can make sector specific shocks more severe and increase the likelihood of asymmetric shocks and diverging business cycles. The view as developed by Frankel and Rose in 1998, is referred to as endogenous OCA theory. They investigated the relationship between international trade patterns and international business cycle correlations and argue that both variables are endogenous. They acknowledge the possibility of regional concentration and more asynchronous business cycles, but also emphasize that entering a currency union may raise international trade linkages, increase synchronicity and therefore increase the benefits of joining a common currency area. Because an increase in trade affects the nature of the business cycles, as they might converge between countries (Frankel & Rose, 1998).
Empirical research on monetary union’s effects on trade, indicates a possible increase in trade of up to 50 percent in the EMU (Glick & Rose, 2016). According to the endogeneity hypothesis this significant trade increase should result in synchronization of EMU business cycles, there are several empirical papers that support this hypothesis (Clark & Wincoop, 2001; Cerqueira & Martins, 2009). Other research adds that not only an increase in trade integration, but also stronger spillovers across countries, increased labour mobility and the establishment of common risk sharing systems have contributed to this positive Euro effect (Gächter & Riedl, 2014). In addition, Kenen (2000) emphasizes that the ability
7 of a monetary union to transmission shocks increases. as trade intensity rises and business cycles converge, but the incidence of asymmetric shocks is not necessarily reduced. Furthermore, Hughes Hallet and Piscitelli (2001) show that a monetary union may induce convergence of business cycles, but only if there is already a sufficient symmetry in the shocks and institutional structure across the countries. Respectively, two different paradigms are put forward in the literature. A ‘pessimistic view’ by Paul Krugman (1991), and an ‘optimistic view’ first introduced by Frankel and Rose (1998). 2.3. Empirical Evidence on Business Cycles
Much empirical analysis has been and is still being undertaken to investigate which path the Eurozone will take. With limited cross-border labour mobility and the lack of a (federal)fiscal transfer mechanism, an adequate presence of asymmetric shocks across countries entail that a common monetary policy is less effective to smooth over shocks (Martin, 2001). Accordingly, symmetry across regions in the EMU is of increased importance. Early in the literature, Burns and Mitchell (1946), made a distinction between classical business cycles and deviation cycles. They define the former in terms of absolute expansions and contractions and the latter as the deviation of economic activity from a trend. Recently, most researchers focus on deviation in cycles as growth cycle recessions are more present than classical recessions. However, disagreement remains on how the trend component should be identified and estimated, as well as how to measure convergence of business cycles. To examine the business cycles in the EMU several filter such as the Hodrick-Prescott filter, the Baxter-King and the Cristiano-Fitzgerald band-pass filters and the phase average trend have been used.2
Artis and Zhang (1995) investigate the effects of the exchange rate mechanism of the European Monetary System on the international business cycle in terms of the linkage and synchronization of cyclical fluctuation between countries. They use the Hodrick-Prescott filter to obtain the cyclical component of industrial production indices. This filter separates a time-series into growth and cyclical component. They show that in the period before the exchange rate mechanism, the exchange rate mechanism countries were more linked to the business cycles of the USA. Afterwards, the linkages of these country’s business cycles shifted towards the German business cycle. Implying convergence of European business cycles as integration rises. In line with this research, Belo (2001) studies the convergences of the Euro area countries by calculating the correlations of sovereigns with the Euro area business cycle between 1960 and 1999. Concluding that three distinct groups of countries exist based on these correlations. The first group experienced a high degree of association and synchronisation with the European business cycle, including German, France and Belgium. The second group with an increase in the business cycle convergence consists of Spain, Italy, Austria, the Netherlands and Portugal. A last
8 group is formed by Finland, Ireland and Luxembourg which do not show evidence of business cycle convergence with the Euro area. Koopman and Azevedo (2007) have similar results with their multivariate stochastic cycle model and time-varying correlations.
De Haan et al (2007) emphasise the disagreements between various studies about business cycle synchronization in the EMU. In their work they summarize various studies that examine business cycles. They notice that if industrial specialization accelerates within the EMU, synchronicity might decrease as predicted by Krugman (1993). Nonetheless, they observe that country-specific fluctuations become less important within monetary unions. Accordingly, there is no consensus on the importance of synchronicity. Noteworthy is that a part of the research summarized in de Haan et al (2007) originate from the work of Blanchard and Quah (1989). They were able to develop a technique that is able to decompose permanent and temporary shocks. Bayoumi and Eichengreen (1993) adopted this technique to identify demand and supply shocks for the prospective members of the monetary union. The primary belief is that an economy is hit by two types of shocks, demand and supply shocks. These shocks are diagnosed with a bivariate VAR analysis and with the help of the restriction that the long-term impact of demand shocks on the output is zero. The authors compare the coherence of the underlying shocks and the speed of adjustment in 11 European commission member countries with US regional data and find that shocks in the EU member countries are more peculiar than in the US. Most compelling discovery is that a core and periphery pattern is observable in Europe. The core exists of Germany, France, Belgium, the Netherlands and Denmark and the periphery consist of the United Kingdom, Italy, Spain, Portugal, Ireland and Greece. In a follow up study, Campos and Macchiarelli (2016) use the same methodology, window length and sample, and add an over-identifying restriction test to examine whether the EMU weakened or strengthened the core-periphery pattern. They find that the pattern weakened and a new smaller periphery has emerged consisting of Spain, Portugal, Ireland and Greece.
9 As yearly data is used, the number of observations are relatively low, which could affect the reliability of the structural VAR. Fidrmuc and Korhonen (2003) followed the methodology of Bayoumi and Eichengreen (1993), but with quarterly data, and found that some countries that previously were considered as peripheral are highly correlated with the Euro area shocks. This supports the endogeneity theory of Frankel and Rose (1998). Henceforward, following the theory of OCA and endogeneity, it is interesting to research to what extent the asymmetric demand and supply shocks are present in the EU, specifically for the EMU, after 17 years of the Euro. This paper follows the methodology of Bayoumi and Eichengreen (1993), as there is no consensus on how convergence between time series and business cycles should be quantified (Haan, et al., 2007). The structural VAR helps to investigate the correlations of demand and supply shocks, the impulse responses to these shocks and the speed of adjustment. Therefore, it guides answering the research question whether the speed of adjustment is the same for a country with and without an independent monetary policy and whether the core- periphery pattern still exist.
3. Methodology
As the literature has pointed out the importance of correlated demand and supply shocks to reduce the costs of the EMU. This section presents the methodology used to recover these supply and demand shocks in order to investigate the symmetry of the EMU. It describes the bivariate structural VAR model, where the aggregate demand and supply shocks are recovered from the VAR residuals using a long-run restriction proposed by Blanchard and Quah (1989). A structural VAR approach is attractive to use to recover the underlying demand and supply shocks, as I am interested in the impulse responses functions after a shock rather than estimating the structural parameters itself. As VARs can be transformed to interpret the evolution of the system’s variables as a function of orthogonalized innovations in any of these variables (Cooley & Dwyer, 1998). Moreover, a benefit of this approach is that a minimum set of identifying assumptions is required to estimate the dynamic effects of economic shocks (Christiano, et al., 2006).
3.1 Identification Strategy
10 Although there is no unique way to identify different shocks affecting a given economy, economic theory can provide guidance in identifying the model. There is considerable agreement about the qualitative effects of demand and supply shocks. As inference is robust across a large subset of the identification schemes that have been considered in the literature (Christiano, et al., 1999). Ahmed (2005) uses a recursive VAR, were the investment-saving and liquidity preference-money supply (IS-LM) model is used as a guide for identification. However, for this economic question and the structural VAR it is common to impose a long-run restriction on the impact of demand shocks on output based on a New-Keynesian model of aggregate supply and demand curves.3 Here demand shocks are defined as a sudden unexpected event that temporarily increases or decreases demand for goods or services, for example a monetary shock. Accordingly, a supply shock is defined as a sudden increases or decreases in the supply of a commodities or services, for example a productivity shock.
This Aggregate Supply and Aggregate Demand (AS-AD) model distinguishes between the short-run and the long-run equilibria for the economy. In particular this framework assumes that the Long-Run Aggregate Supply curve (LRAS), which is vertical, is likely to differ from the Short-Run Aggregate Supply curve (SRAS), which is positively sloped. The aggregate supply relation captures the effects of output on the price level, this relation is derived from the behaviour of wages and prices. Wages are set by firms or by bargaining between workers and firms. Wages depend negatively on the unemployment rate and positively on the expected price level, as wages are typically set in nominal terms for some period of time. Therefore, wages do not adjust instantly as the price level changes or when prices are not as expected by wage setters. Prices are set by firms, which in turn depends on the wage and on the mark-up of prices over wages. When the expected real wage rate is equal to the effective real wage that firms are willing to pay, the labour market is in equilibrium. This relationship is in the short-run captured by SRAS and in the long-run by LRAS. The difference between the SRAS and LRAS is induced by the rigidity of wages, as nominal wages are generally set for a longer period of time. Hence, in the long-run real wages adjust perfectly to changes in prices. As rises in output lead to an increase in the price level, the AS curve is upward sloping. Shifts in the AS curve can be caused by changes in the expected price level which shifts the curve accordingly.
The aggregate demand (AD) curve captures the effect of the price level on output, which is derived from the equilibrium conditions in the goods and financial markets. The goods market is determined by the interest rate and the level of output. It is in equilibrium when output equals the demand for goods. The financial market shows the combinations of the interest rate and the level of output and is in equilibrium when the supply of money equals the demand for money. Using both markets one can derive the relation between the price level and the level of output, which becomes the AD curve. Which is downward
11 sloping, since an increase in the price level leads to a decrease in the real money stock, which increases the interest rate. This causes people to prefer to have money on the bank, hence lowers demand for goods and lowers output.
By construction, when the AS equals the AD curve, the goods market, the financial markets and the labour market are all in equilibrium. Henceforth, the AD-AS model can be used to reason the effects of a demand and supply shock to the economy. Figure 1 shows the effects of a demand and supply shock. First considering the effects of a demand shock to output and prices, implied by a shift to the right from AD to AD’ in figure 1.a. Due to this positive shock the output increases and the economy moves from E to D’. Now the price level is not as employers expected, hence they have to adjust the nominal wages to compensate for the loss of real wage of the employees. It is likely that the wage setters upward their expectations of what the price level will be in the future. As a consequence, a higher nominal wage leads firms to set higher prices until real wages are adjusted to the price change in the long-run. This is indicated in the graph as a counter-clockwise rotation of the supply curve and a move of the economy from D’ to D’’. The higher price level decreases demand and pushes output to its initial level, accordingly the long-run effect is a higher price level and the same output level.
12 Figure 1: A Demand Shock and Supply Shock
a) Demand Shock b) Supply Shock
This New-Keynesian model provides two distinct features of the original shocks affecting the economy. First, supply shocks are allowed to have a permanent effect on output and prices. However, demand shocks are allowed to have a permanent effect on prices, but its effect on output is restricted in the long-run. Second, a positive demand shock raises prices while positive supply shocks reduce the price level. The last property is used in Bayoumi and Eichengreen (1993) as an over-identifying restriction. It is not imposed pre-estimation, but it serves as a check of the aggregate demand and supply model post-estimation.
3.2 Estimation Method
The method of Blanchard and Quah (1989) is used to identify demand and supply shocks. Following Bayoumi and Eichengreen (1993), and using the New-Keynesian model, I estimate a VAR with output and prices as variables. As the New-Keynesian model assumes that demand and supply shocks are observable in output and prices. The theoretical model can be written as an infinite moving average representation of demand and supply disturbances:
13 ∑ 𝐿𝑖𝐴 𝑖𝜀𝑡 ∞ 𝑖=0 = ∑∞𝑖=0𝐿𝑖[ 𝑎11𝑖 𝑎12𝑖 𝑎21𝑖 𝑎22𝑖] [𝜀𝜀𝑑𝑡𝑠𝑡] = [∆𝑝𝑡] ∆𝑦𝑡 (2)
Where, 𝑎11𝑖 represents element 𝑎11 for the 𝑖th country, in matrix 𝐴𝑖 and 𝐿𝑖 stands for the lag operator for the 𝑖th country. The AS-AD framework can be used to identify one of the elements of the matrix 𝐴
𝑖. As the framework assumes that demand shocks will not have a permanent effect on the output level. For this model this implies that the cumulative effect of a demand shock on the change in output must be zero, as formulated by the following equation:
∑∞ 𝑎11𝑖
𝑖=0 = 0 (3)
Next to this identification, it is also assumed that supply and demand shocks are uncorrelated. Because I want to know, once the demand and supply shocks are defined, what the individual effect of an demand shock is on output and the price level. As the model is exactly identified now, there exists one best value for each parameter for a specified lag. Where the appropriate lag order is assumed to be known. Hence, the model defined by Equation (2) and (3) can be estimated by the following VAR:
𝑋𝑡 = 𝐵0+ 𝐵1𝑋𝑡−1+ 𝐵2𝑋𝑡−2+ ⋯ + 𝐵𝑛𝑋𝑡−𝑛+ 𝑒𝑡 (4)
Where 𝑋𝑡 is a vector of output growth and inflation, 𝐵𝑖 represents the impulse response functions of the shocks to 𝑋𝑡 and 𝑒𝑡 represents a vector of white noise innovations. This implies that the variables 𝑋𝑡 are covariance-stationary, hence the expected value of output growth and inflation will not depend on time. And the covariance matrix of 𝑋𝑡 depends on the time lapsed and not on the reference period t. Due to this assumption, the VAR representation, which is the reduced form of the structural model, can be inverted. Using the lag polynomial to rewrite Equation (4):
(𝐼 − 𝐵(𝐿))𝑋𝑡= 𝑒𝑡
𝑋𝑡 = (𝐼 − 𝐵(𝐿))−1𝑒𝑡 (5)
Equation (5) is the moving average representation of 𝑋𝑡. In the moving average representation the system is reparametrized to express the variables in 𝑋𝑡 as a function of current and past reduced form white noise shocks. Henceforth, 𝑒𝑡 is a vector of regressions of the lagged values of the elements of 𝑋𝑡 on the current values of each in turn. This combined with Equation (2) gives us the following relationship:
𝑒𝑡 = 𝐴0𝜀𝑡 (6)
14 [𝑐𝑜𝑣(𝑒𝑣𝑎𝑟(𝑒𝑑𝑡) 𝑐𝑜𝑣(𝑒𝑑𝑡, 𝑒𝑠𝑡)
𝑑𝑡, 𝑒𝑠𝑡) 𝑣𝑎𝑟(𝑒𝑠𝑡) ] = 𝐴0[
𝑣𝑎𝑟(𝜀𝑑𝑡) 𝑐𝑜𝑣(𝜀𝑑𝑡, 𝜀𝑠𝑡)
𝑐𝑜𝑣(𝜀𝑑𝑡, 𝜀𝑠𝑡) 𝑣𝑎𝑟(𝜀𝑠𝑡) ] 𝐴0′ (7)
Equation (6) represents the variance-covariance matrix of residuals. This matrix entails necessary information about the correlations in the model, and with the mentioned restrictions the model can be estimated. The first restriction was Equation (3), which indicates that the cumulative effect of a demand shock does not affect output growth in the long-run. The second restriction was the orthogonality restriction, which has been assumed after Equation (3) and is common practice to assume when identifying demand and supply shocks (Fidrmuc & Korhonen, 2003; Campos & Macchiarelli, 2016). The intuition behind this restriction is that shocks are exogenous forces and therefore must be treated as uncorrelated. The third and fourth restrictions are normalizations, which sets the variance of the errors equal to unity. It is convenient to use this restriction but not necessary to estimate the model. The normalizations allow to interpret the standard deviation shocks as unit shocks in ɛ𝑑𝑡 and ɛ𝑠𝑡.
When all the elements are defined, the structural VAR can be used to recover the underlying demand and supply shocks in the individual countries of the sample. The shocks are used to calculate correlation coefficients of the individual countries within the core of the EMU. The higher the correlation coefficients are the fewer idiosyncratic shocks the EMU endures. As a high degree of synchronization reduces the cost of a common currency and increases the efficiency benefits. Next to this, the underlying shocks will be used to simulate demand and supply shocks to output growth and inflation. The obtained impulse response graphs should give a representation of the effect of a shock at impact and how fast individual countries adopt to such a shock.
4. Data
4.1 Data Description
15 same quarter a year before of the implicit GDP deflator is used as a measurement for inflation. Now both variables are measured in percentages, which makes the results more convenient to interpret.
To obtain this data Eurostat has been used as the primary source for the European countries. Table A.1 in Appendix A provides more details on the data used. Quarterly data on real GDP and the implicit GDP deflator is collected from the first quarter of 1996 to the third quarter of 2017. An advantage of this publicly available database is that it consist of seasonally adjusted data for all variables. The control group consists of ten additional OECD countries Australia, Canada, Iceland, Japan, South Korea, Mexico, New Zeeland, Norway, Switzerland and the United States. Data on real GDP and the GDP deflator are taken from the OECD statistics and are seasonally adjusted.
Differently from Bayoumi and Eichengreen (1993) quarterly data on real GDP and GDP deflator are used. The beginning of 1996 has been chosen as a starting point, because that was the year that most countries started presenting quarterly data on OECD statistics and Eurostat. They use yearly data of real and nominal GDP spanning the period 1960-1988. This expansion of the data possibly gives a more detailed pattern of reactions to demand and supply shocks in the European union. On the other hand I would expect noisier results as more detailed data is used. By using seasonally adjusted data and year on year mutated quarterly data most anomalies that can occur during specific seasons should be eliminated. In addition, a control group of OECD countries is used rather than regions in the United States. Since international trade is increasing, possibly the business cycles of all individual countries are converging. So to differentiate from probably a worldwide converge due to increasing trade patterns, I use additional EU and OECD countries as a benchmark. Hence, by adding countries that do have an independent monetary policy on can observe potential differences in countries responding and adjusting to idiosyncratic shocks.
Table 1: Summary Statistics of real GDP Growth in Percentages
Variables N Mean St. Dev. Min Max
16
Variables N Mean St. Dev. Min Max
Lithuania 83 3.822 4.222 -9.939 13.766 Luxembourg 83 2.428 1.943 -2.464 7.397 Malta 67 2.287 1.343 -0.479 5.216 Netherlands 83 1.513 1.138 -1.762 3.549 Austria 83 1.404 0.805 -1.146 2.730 Portugal 83 1.422 1.434 -2.281 4.025 Slovenia 83 1.931 1.692 -2.959 4.979 Finland 83 1.589 1.422 -3.470 3.683
Average over the
EMU 2.161 2.159 -4.093 6.346 Bulgaria 67 3.246 2.508 -1.427 8.784 Denmark 83 1.382 1.049 -2.523 3.086 Hungary 83 2.486 3.155 -7.762 10.528 Poland 59 2.256 4.223 -9.295 10.272 Croatia 67 1.855 2.256 -3.983 5.618 Sweden 83 1.539 2.910 -7.994 8.916 United Kingdom 83 1.536 3.932 -9.799 11.107 Average over additional EU countries 2.043 2.862 -6.112 8.329 Australia 83 2.48 1.035 -0.401 4.516 Canada 83 1.889 1.398 -3.479 4.382 Iceland 78 3.408 2.057 -1.475 8.335 Japan 83 0.067 0.956 -3.760 1.674 South Korea 83 2.653 1.427 -2.442 6.202 Mexico 83 4.059 2.541 -1.846 10.956 New Zealand 82 2.081 0.994 -0.091 4.205 Norway 83 2.353 2.514 -3.736 9.249 Switzerland 83 1.009 0.929 -1.057 2.855 United States 83 1.801 0.891 -1.409 3.159 Average of the control group 2.180 1.474 -1.970 5.554
17 In Table 1 the summary statistics of percentage real GDP growth for EU countries are presented, as percentage growth rates enables one to do cross country analysis better. Also, percentages will be the unit used for the structural VAR, as output growth and inflation are considered. For every country we have 83 observations, except for Bulgaria, Poland, Croatia, Malta, New Zealand and Iceland. For New Zealand quarter 3 of 2017 is missing, the other countries miss data from the beginning of the data set. As the same data misses for real GDP as for the implicit GDP deflator and the VAR is estimated per country, no adjustments have to be made regarding omitted data. Large negative growth rates are experienced by Latvia, Lithuania, Poland and United Kingdom mostly at the start of the great recession. This table reveals some positive outliers in Ireland, Latvia, Lithuania, Hungary, Poland, United Kingdom and Mexico. Ireland experienced a growth rate of 21.98 percent from the fourth quarter of 2014 to the first quarter of 2015. This surge is caused by including the valuation of overseas companies in the value of the corporate sector on the Irish balance sheet (Inman, 2016).
Considering the averages of the summary statistics, the average growth rate of the EMU is as high as the average growth rate of the OECD countries, only those of the additional EU countries is slightly lower. The average growth rates of countries that later joined the EMU cause a higher average growth rates. Czech Republic, Estonia, Ireland, Latvia, Lithuania, Luxembourg, Malta and Slovenia are countries that later joined the EMU and experience higher average growth rates than the first EMU countries. This can be explained by conventional growth theory, as it predicts catching-up of emerging countries to developed countries (Barro & Sala-I-Martin, 1991). Looking at the standard deviations, the countries just named are the countries that have high standard deviations. This could indicate instability, or fast growth by catching up on the developed countries. Volatility in these countries is high compared to the low volatility observed in developed countries. As low standard deviations in output growth are generally observed in developed countries (Stock & Watson, 2003).
Table 2: Quarterly Inflation Growth in Percentages
Variable N Mean St. Dev. Min Max
18
Variables N Mean St. Dev. Min Max
Latvia 83 3.351 5.314 -14.217 17.304 Lithuania 83 2.495 3.399 -7.050 10.617 Luxembourg 83 1.966 2.114 -2.896 7.725 Malta 67 2.270 0.684 0.013 3.996 Netherlands 83 1.531 1.033 -0.959 3.748 Austria 83 1.506 0.689 0.021 2.897 Portugal 83 1.936 1.128 -1.146 3.572 Slovenia 83 2.688 1.857 -1.603 6.261 Finland 83 1.663 1.035 -0.937 3.825 Average over EMU countries 1.967 1.747 -2.863 5.970 Bulgaria 67 3.275 3.324 -3.676 10.792 Denmark 83 1.643 1.028 -0.451 5.035 Hungary 83 3.921 1.966 -0.380 8.363 Poland 59 2.001 1.369 -1.180 4.676 Croatia 67 2.059 1.584 -0.733 6.270 Sweden 83 1.525 0.819 0.250 3.809 United Kingdom 83 1.703 0.857 -0.726 3.410 Average over additional EU countries 2.304 2.617 -0.985 6.051 Australia 83 2.229 2.120 -2.000 7.200 Canada 83 1.723 1.766 -4.200 5.400 Iceland 78 3.704 3.368 -2.900 17.800 Japan 83 -0.664 1.169 -2.500 3.100 South Korea 67 1.988 1.150 -0.400 4.400 Mexico 83 4.741 1.756 0.300 8.700 New Zealand 82 1.871 1.934 -2.300 8.400 Norway 83 2.795 4.234 -7.000 12.700 Switzerland 83 0.417 0.828 -0.800 2.600 United States 83 1.727 0.616 0.200 3.000 Average over OECD countries 2.053 1.894 -2.160 7.330
19 For inflation, data on the GDP deflator has been collected. Inflation is calculated with the GDP deflator of this quarter minus the GDP deflator a year ago of the same quarter, henceforth inflation is measured in percentages. The number of observations for each country is equal to the number of observations of real GDP growth, hence there are no omitted values considering the time-series. In the time period considered the average inflation for all countries is around the two percent. Notable is that Japan has a negative average inflation rate, as Japan is coping with deflationary pressures for decades. Considering the standard deviations of the sample, they are the highest for the same countries that experienced the most variation in output growth. The high maximum values in Latvia are experienced before the start of the financial crisis, and the most deflation during and in the aftermath of the crisis. Interesting is that, on average, the EMU countries experience the smallest standard deviation in the inflation rate. This implies that the inflation rate is the least volatile in the EMU. Whereas volatility is expected to be the highest in the EMU, because of its central monetary policy.
Both, Bayoumi and Eichengreen (1993) and Campos and Macchiarelli (2016), use Germany as a proxy for the European business cycle. Germany is an important reference country due to its large size and its strong ties in trade with many EMU countries. Nevertheless, in previous research France and Europe as a whole have also been used as yardsticks for Europe’s business cycle (Fidrmuc & Korhonen, 2003). Horvath and Rátfai (2004) emphasize that business cycles of individual countries should be harmonized with the Euro area as a whole. Other studies find evidence against using Germany’s business cycle as a yardstick for the European business cycle. Aguiar-Conraria and Soares (2011) find that France, not Germany, has been the leading European cycle in the long-run. The same conclusions are drawn by Fidrmuc and Korhonen (2003) and Ahlborn and Wortmann (2018), who argue that France follows the business cycle of Europe more closely than Germany. So different yardsticks have to be addressed when reassessing the core-periphery pattern based on demand and supply shocks. This paper uses both, Germany and France, as benchmark countries for the European Business cycle. Therefore Table 3 presents the standard deviations and correlation coefficients real GDP growth and inflation with Germany and France.
Table 3: Standard Deviations and Correlation Coefficients of real GDP Growth and Inflation
Real GDP Growth Inflation
Countries
St. Dev
Correlation
St. Dev
Correlation
Germany France Germany France
Belgium
0.799 0.714 0.913 0.620 -0.635 0.464Czech Republic
2.942 0.446 0.749 2.044 0.328 0.094Germany
0.932 1.000 0.730 0.692 1.000 -0.481Estonia
3.388 0.642 0.903 2.389 -0.375 0.78120
Countries
St. Dev. Germany France St. Dev Germany France21
Countries
St. Dev. Germany France St. Dev Germany FranceNorway
2.514 0.430 0.740 4.234 -0.666 0.539Switzerland
0.929 0.584 0.841 0.828 -0.487 0.751United States
0.891 0.572 0.836 0.616 -0.521 0.623Average of OECD
countries 1.474 0.483 0.669 1.894 -0.269 0.276
Notes: the standard deviations and correlation coefficients are calculated of real GDP growth and inflation in percentages over the whole available time-period.
Analysing the average correlation coefficients of the EMU, one can see that output growth and inflation move more similar to France than Germany. This suggests indeed a shift of EMU countries from Germany to France. Considering correlation coefficient averages of the additional EU countries and the OECD countries, they are also more correlated to the output growth and inflation of France. Interesting is that the correlation coefficients do suggest a more correlated pattern in the EMU with both, Germany and France, than the additional EU or OECD countries do with Germany and France. This implies indeed more similarity in the currency union. Overall, his table shows that it is relevant to include France as a reference group of the EMUs’ business cycle. More analysis has to be conducted to identify country groups and to reveal the correlation of business cycles in the EMU.
4.2 Data Adjustments
In order to perform a VAR estimation correctly and avoid the problem of a spurious regression the variables used must be stationary. Nonstationary variables can include a deterministic trend, where a trend always revert to an increasing or decreasing trend in the long-run, but the effects of shocks are eventually eliminated. They can also include a stochastic process, which have a permanent impact on the mean and the forecast intervals grow over time. In order to evaluate the effects of the individual demand and supply shocks appropriately the output growth and inflation should be trend and unit-root stationary. As both variables means should be stable and should not be allowed to increase or decrease in the long-run. To minimize the variance the logarithm of real GDP is used, and the logarithm of real GDP and the GDP deflator are differenced relative to the same quarter in the previous year. Differenced time series are expected to yield a stationary stochastic process, in order to test this formally the Dickey-fuller unit-root test is performed with a drift option. The null hypothesis states that there exists an unit root with a nonzero drift in the data, the alternative hypothesis that the process is a non-random walk with zero drift. If the null hypothesis cannot be rejected the data is nonstationary and/or a drift exists, than an economy would not recover after a shock in the model.
22 variance do not change over time. The Dickey-fuller test of a unit root does not reject the null of a unit root and a nonzero trend in the real GDP growth and inflation of Spain. This raises two issues for the regressions of Spain. First, it could be that one or both variables experience an increasing or decreasing trend. Secondly an unit root could exists in one or both of the process, which might be affected by demand and supply disturbances in the long run. To correct for the integrated process the series were de-trended by using the Hodrick-Prescott filter as used by Blanchard and Quah (1989). The filter separates the time series into trend and cyclical components and removes the trend from the series. The new variable for output growth of Spain did not become stationary, as the inflation rate did, a shock to the economy of Spain should result in the variables going to a stable rate.
Also a Granger causality test is used to test whether the past values of one variable are useful in forecasting another. If one variable is suitable to predict the other variable, than one variable is said to granger cause the other variable. The null hypothesis is that the coefficients on the four lags of inflation that appear in the equation for output growth are jointly zero. The alternative hypothesis is that the lagged values are unequal to zero. Table A.3 in the Appendix shows the results of the Granger causality tests of the VAR in Equation (4). Most values are significant, implying that Equation (4) is well defined for most countries.
In order to estimate a structural VAR an appropriate lag length has to be set. To test for this I have used the Akaike, Schwarz and Hannan-Quinn information criteria for the individual time series. VAR order selection criterion are performed before the structural VAR when the variables are stationary. The results are presented in Table A.4 of the Appendix, where the lag length used ranges from 1 till 12. The average is used for two reasons. Firstly, as there is no consensus in the literature on which information criterion to adopt. Secondly, symmetry in the lag length across countries was chosen in order to preserve similarity between the countries impulse responses (Bayoumi & Eichengreen, 1993).
5. Empirical results
This section presents the estimation results for the structural VAR with four lags for all countries and using quarterly data closely following Bayoumi and Eichengreen (1993). In all cases, the number of lags was set to 4, based on the average of the Akaike, Schwarz and Hannan-Quinn information criteria for the individual time series4. The first section discusses the impulse response functions for the structural VARs. Section 5.2 studies the speed of adjustment of a country after a supply or demand shock. Section
23 5.3 analyses the correlations of demand shocks in Germany and France with those in other EU countries. Section 5.4 analyses the robustness of the results regarding the quarterly data considered and the time period.
5.1 Impulse Response Graphs and Theory
In order to investigate the responses of the varies economies to demand and supply shocks, the impulse response functions are derived directly after the structural VAR. The graph shows by default the impulse responses of output growth and inflation after a positive unit shock to output growth and inflation. As output growth and inflation are measured in percentages, an unit shock corresponds to a positive one percent shock. All impulse response graphs and its confidence intervals approach zero in the long-run, proving the stationarity of the time series5.
Table 4: Impulse response functions
Demand shock on output Supply shock on output Demand shock on prices Supply shock on prices Be lg iu m Cz ec h Re p u b li c Ge rm an y Esto n ia Ire lan d
27 Demand shock on output Supply shock on output Demand shock on prices Supply shock on prices S o u th K o re a M ex ico Ne w Z ea lan d No rwa y S witze rlan d Un it ed S tate s
28 response functions stay positive or the rate becomes negative a little before going to zero in the long-run. Implying that countries experience a growth in the price level, as predicted by the AS-AD model.
The AS-AD framework predicts a positive effect on output and a negative effect on the price level after a positive supply shock in the long-run. Effects of a supply shock to output and the price level should go by definition in opposite directions. Following the AS-AD framework, the impulse response graph of a supply shock to output should be positive and going to zero in the long-run. Its effect on the inflation rate should be the opposite and be negative at impact, and returning to zero in the long-run. Generally most countries experience an opposite effect of a supply shock to output growth and the inflation rate. Thus an decrease in the level of output and an increase in the price level. However, the impulse response graphs shown in Table 4 present the opposite results. Where output growth decreases at impact and inflation increases at impact and both impulse response graphs go to zero. This finding is not in line with earlier literature that has constructed a bivariate structural VAR with long-run restrictions (Campos & Macchiarelli, 2016).
5.2 Speed of Adjustment
The impulse response functions provide information on how long it takes for the individual countries to adjust to an exogenous shock. This is interesting for the cost and benefit analysis of the EMU, as this reveals information on their ability to adjust to supply and demand shocks without the help of an independent monetary policy. When EMU countries adopt as quick to an asymmetric shock as countries with an independent monetary policy could suggests that factors of production move sufficiently between countries and industries to adjust.
Table 5: Average Number of Quarters it takes for Countries to adjust for Shocks
Demand Shock to Supply Shock to
Output Prices Output Prices
Average over EMU 16.895 18.632 17.526 17.000
Average over the additional EU countries 14.857 16.571 15.714 13.714
Average over the OECD countries 14.500 15.200 15.600 14.800
Notes: The maximum number of quarters investigated is 20.
29 response graphs, and the number of quarters it takes to get to zero. Generally, it takes a half year longer for the EMU countries to adjust to a specific shock. The EMU countries have the most implications accommodating the price level after a supply shock, which takes a year longer. This confirms previous literature, as countries with an independent monetary policy are better able to adjust to shocks than countries in a currency union (De Grauwe, 2016). Suggesting that factors of production in the EMU are not mobile enough across countries or industries to compensate for an idiosyncratic shock.
5.3 Correlation of Demand and Supply Shocks in the EU
As explained in the literature review, serious problems in the EMU could stem from countries experiencing asymmetric shocks. Therefore this section investigates the correlation of demand and supply disturbances of the EU with the demand and supply coefficients of Germany and France. By extracting the residuals of the structural VAR and multiplying these by the inverse of the matrix of contemporaneous variables the structural shocks are recovered. Figure 1 and Figure 2 show the contemporaneous correlation between demand and supply shocks of Germany and France with the individual EMU countries. The results for the correlation coefficients are presented in Table B.1 of Appendix B.
Examining Figure 1, one can see that almost all countries have a positive correlation with German’s demand disturbances. Only Malta’s demand shocks do not correlate with the one of Germany, but its supply shocks do. Correlation of supply disturbances with Germany are more dispersed, as more countries exhibit a negative correlation with Germans’ supply disturbances. Looking at Figure 2, demand shocks of the EU countries all positively correlate with the demand shocks of France. Considering the individual supply shocks, 7 countries have a negative correlation with France, whereas 8 countries experience a negative correlation with Germany. That supply shocks are more dispersed is not striking due to the restriction that demand shocks do not exert long-run effects on output. Hence, all output response functions to demand shocks converge to zero, and therefore one can expect that correlation coefficients are higher (Frenkel, et al., 1999).
30 Figure 2 and Figure 3 do not reveal a clear core-periphery pattern as suggested by previous literature (Haan, et al., 2007). The core countries, as specified by Bayoumi and Eichengreen, are Germany, France, Belgium, the Netherlands and Denmark. Considering their demand shock correlations, the country group is still well correlated. However, the country group cannot be labelled as a ‘core’ as supply shocks are not correlated to the supply shocks of Germany or France. Henceforth, the findings presented in Figure 2, Figure 3 and Table 6 regarding the supply shocks are not in line with previous literature (Campos & Macchiarelli, 2016).
31 Figure 3: Correlation of Supply and Demand Disturbances with French Supply and Demand Disturbances
Table 6 shows the changes between the correlation coefficients from demand and supply shocks before the financial crisis and correlation coefficients from after the crisis. Where previous literature has stressed concerns about Central and East European countries joining the EMU. There is mixed evidence as to the convergence of business cycles in the EU. The results in Table 6 show that Estonia’s and Hungary’s demand and supply shocks became more similar to both French and German demand and supply shocks after the crisis. This is also shown in Figure 2 and 3, as both countries have relative good correlation with demand and supply shocks with Germany and France compared to other EU countries. Fidrmuc and Korhonen (2003) has noted the same a decade ago, and argues that this is not surprising as these two countries have very extensive trade relations with the Euro area countries. The Czech Republic moved a little toward the business cycles of Germany and France, but there are mixed results for Cyprus, Latvia, Lithuania and Slovenia.
32 Table 6: Differences between the Correlation Coefficients of Demand and Supply Shocks from before the Crisis and after the Crisis
Country
Germany France
Demand Supply Demand Supply
33 Iceland 0,843 -0,201 0,158 -0,319 Japan 0,478 -0,509 -0,373 0,060 Korea -0,471 -0,043 0,253 0,229 Mexico 0,290 -0,414 -0,176 -0,227 New Zealand 0,319 -0,601 0,248 -0,444 Norway 0,567 -0,212 0,038 0,022 Switzerland -0,351 -0,072 0,081 0,723 United States -0,626 0,285 0,560 -0,392 OECD total 0,275 -1,194 0,800 0,086 Average 0,028 -0,119 0,080 0,009 Total 4,871 -4,839 0,419 5,674 Average total 0,132 -0,131 0,011 0,153
Notes: The first time period is from the first quarter of 1997 to the fourth quarter of 2007. The second time period is from the first quarter of 2008 onwards.
5.4 Robustness
5.4.1 Annual Data
A consequence of using quarterly data is that the number of observations increase, as standard errors decrease with the number of observations, t-statistics increase with the number of observations. This affects the confidence intervals of the impulse response graphs presented in the results. As a consequence of using yearly data, the confidence interval bands of the impulse response functions could be larger. By estimating a structural VAR using two lags and the same identification strategy methodology section, the impulse response graphs are estimated. The impulse responses all go to zero in the long run, and have similar signs to the quarterly data results. The results should, however, be interpreted with caution since the confidence interval bands are wide. This is as expected, but it is too large to be able to draw reasonable conclusions from the yearly data for a demand shock on output and a supply shock on output. Because one cannot conclude with a 95% confidence level if the shock is positive of negative. Therefore demand and supply shocks with annual data will not be discussed further.
5.4.2 Time Period
34 crisis countries experienced deflationary pressures and most countries failed to reach the desired close to two percent inflation. As the sample is split, not all data is stationary. Hence, the time-series that have an unit root are adjusted by the Hodrick-Prescot filter. The results produced by the structural VAR and its impulse responses show inconclusive results, as the impulse response functions of multiple countries do not return to zero in the long run. A longer time period has to be considered in order to investigate how the financial crisis has affected the correlations of business cycles in the EMU. Moreover, these results do not help to explain why the supply shocks move in opposite directions compared to previous literature (Campos & Macchiarelli, 2016).
6. Conclusion
This paper presented a bivariate structural VAR to identify the incidence of aggregate supply and demand disturbances in the EU to analyse the responses of the individual countries to those shocks. In addition, corresponding correlations coefficients of the structural shocks with Germany and France are estimated. This methodology helps to answer the question whether the business cycles in the EMU show a certain core or periphery as suggested by Bayoumi and Eichengreen (1993).
Regarding the impulse response graphs, some clear results emerge. First, most countries react similarly to the shocks. Demand shocks go as predicted by the New-Keynesian model and do not increase the output but does increase the price level. However, the long-run results of a positive supply shock to inflation do differ from the those of previous studies and the prediction of the theoretical model. Because the responses of the shock go in opposite directions. Second, countries with an independent monetary policy are able to adjust faster to asymmetric disturbances than countries in the EMU. As the former is able to use its independent monetary policy to dampen those shocks. This indicates that the ECB is less able to stabilize after such a disturbance, which has been predicted early in the literature as a cost of a currency union (Mundell, 1961). This also indicates that the criteria for an OCA, such as factor mobility, are not present enough to accommodate idiosyncratic shocks as well as countries that are not in a currency union. Third, the enlargement of the EMU blurs a clear core periphery pattern. Contrary to previous literature, Koopman and Azevedo (2007) and Campos and Macchiarelli (2016), some countries considered as peripheral are now highly correlated with the business cycles of Germany and France. And countries previous considered as a core are in this paper less correlated to the two largest economies of the EMU. However, the increased correlation of Estonia and Hungary with Germany and France is in line with the findings of Fidrmuc and Korhonen (2003).
35 dimensional structural VAR is not able to capture the whole economy. Potential omitted variables are assumed to be in the innovations, hence de error terms. Thus potential incompleteness renders the true identification of the demand and supply shocks, and the interpretation of impulse responses. Furthermore, a problematic assumption could be that shocks in the variables are independent. One may argue that the error terms consists of all the influences and variables that are not directly included in the set of variables, hence a shock in one variable may be accompanied by a shock in another variable (Cover, et al., 2006). The results suggests that the supply shock is not well identified, as structural VARs are variant to the identifying assumptions and omitted observations (Cooley & Dwyer, 1998).
Using a structural VAR to investigate the synchronization of the EMU business cycles does not reveal information on the channels that are used to transmission those shocks. This paper shows that countries that are in the EMU need more time to adjust to an asymmetric shock than a country with an independent policy. These results suggest that the EMU is not synchronized enough to be able to transmission demand and supply shocks effectively. Hence more research should be conducted on those transmission mechanisms and their ability to absorb asymmetric shocks. The better countries are able to accommodate idiosyncratic shocks, the less costs the EMU countries have in losing their independent monetary policy. Future research might find it interesting to investigate the convergence of the EMU further when there is a larger time-series available, as the EMU seems not integrated enough yet. Then it may be interesting to include trade, or other variables of the OCA theory to the model as the granger causality test was not significant for all countries. Suggesting that it might be interesting to add or remove some constrains.
7. References
Aguiar-Conraria, L. & Soares, M., 2011. Business cycle synchronization and the Euro: A wavelet analysis. Journal of Macroeconomics, 33(3), pp. 477-489.
Ahlborn, M. & Wortmann, M., 2018. The core-periphery pattern of European business cycles: A fuzzy clustering approach. Journal of Macroeconomics, Volume 55, pp. 12-27.
Ahmed, M., 2005. How Well Does the IS-LM Model Fit in a Developing Economy: The Case of India.
The international journal of applied economics, 2(1), pp. 90-106.
Artis, M. & Zhang, W., 1995. International Business Cycles and the ERM: Is There a European Buiness Cycle?. International Journal of Finance Economics, Volume 2, pp. 1-16.
Barro, R. J. & Sala-I-Martin, X., 1991. Convergence across States and Regions. Brookings Papers on
Economic Activity, Volume 1, pp. 107-182.
Bayoumi, B. & Eichengreen, T., 1993. Shocking aspects of European Monetary Integration.
36 Belo, F., 2001. Some facts about the cyclical convergence in the euro zone.. Banco de Portugal, 7(1), pp. 37-44.
Benczur, P. & Ratfai, A., 2009. Economic Fluctuations in Central and Eastern Europe. The Facts.
Applied Economics, 42(25), pp. 32-79.
Blanchard, O. & Quah, D., 1989. The Dynamic Effects of Aggregate Demand and Supply Disturbances.
American Economic Association, 79(4), pp. 655-673.
Burns, A. & Mitchell, W., 1946. Measuring Business Cycles. New York: NBER.
Campos, N. F. & Macchiarelli, C., 2016. Core and Periphery in the European Monetary Union: Bayoumi and Eichengreen 25 years later. Economic Letters, Volume 147, pp. 127-130.
Cerqueira, P. A. & Martins, R., 2009. Measuring the determinants of business cycle synchronization using a panel approach. Economic Letters, Volume 102, pp. 106-108.
Chaban, M. & Voss, G. M., 2016. Is Canada an optimal currency area? An inflation targeting perspective. Canadian Journal of Economics, 49(2), pp. 739-771.
Christiano, L. J., Eichenbaum, M. & Evans, C. L., 1999. Monetary Policy Shocks: What have we learned and tho what end?. Handbook of Macroeconomics, Volume 1, pp. 65-148.
Christiano, L. J., Eichenbaum, M. & Vigfusson, R. J., 2006. Assessing structural Vars. NBER
Macroeconomics Annual, Volume 21, pp. 1-106.
Clark, T. E. & Wincoop, E. v., 2001. Borders and business cycles. Journal of International Economics, Volume 55, pp. 59-85.
Cooley, T. F. & Dwyer, M., 1998. Business cycle analysis without much theory. A look at structural VARs. Journal of Econometrics, Volume 83, pp. 57-88.
Cover, J. P., Enders, W. & Hueng, C. J., 2006. Using the Aggregate Demand-Aggregate Supply Model to Identify Structural demand-Side and Supply-Side Shocks: Results Using a Bivariate VAR. Journal of
Money, Credit, and banking, 38(3), pp. 777-790.
Cover, J. P. & Pecorino, P., 2003. Optimal Monetary Policy and the Correlation between prices and output. Contributions to Macroeconomics, 3(1).
De Grauwe, P., 2016. The Economics of Monetary Union. 11 ed. Oxford: Oxford University Press. Dibooglu, S. & Horvath, J., 1997. Optimym Currency Areas and European Monetary Unification.
Contemporary Economic Policy, Volume 15, pp. 37-49.
Eichengreen, B., 1990. Is Europe and Optimum Currency Area. European Economic Integration, pp. 1-28.
Faust, J. & Leeper, E. M., 1997. When Do Long-Run identifying Restrictions Give Reliable Results?.
Business & Economic Statistics, 15(3), pp. 345-353.
Fidrmuc, J., 2004. Endogeneity of the optimum currency area criteria, intra-industry trade, and emu enlargement. Contempory Economic Policy, 22(1), pp. 1-12.
37 Fleming, J. M., 1971. On Exchange Rate Unification. The Economic Journal, 81(323), pp. 467-488. Frankel, J. A. & Rose, A. K., 1998. The endogeneity of the Optimum Currency Area criteria. The
Economic Journal, 108(449), pp. 1009-1025.
Frenkel, M., Nickel, C. & Schmidt, G., 1999. Some Shocking Aspects of EMU Enlargement. Frankfurt am Main., Deutsche Bank Research.
Fry, R. & Pagan, A., 2011. Sign Restrictions in Structural Vector Autoregressions: A Critical Review.
Journal of Economic Literature, 49(4), pp. 938-960.
Gächter, M. & Riedl, A., 2014. One money, one cycle? The EMU experience. Journal of
Macroeconomics, Volume 42, pp. 141-155.
Glick, R. & Rose, A. K., 2016. Currency Unions and Trade: A Post-EMU Reassessment. European
Economic Review, Volume 87, pp. 78-91.
Haan, J. D., Inklaar, R. & Jong-A-Pin, R., 2007. Will Business Cycles in the Euro Area converge? A Critical Survey of Empirical Research. Journal of Economic Surveys, 22(2), pp. 234-273.
Horvath, J. & Rátfai, A., 2004. Supply and demand shocks in accession countries to the Economic and Monetary Union. Journal of Comparative Economics, Volume 32, pp. 202-211.
Ingram, J. C., 1973. The case for European monetary integration. International Finance, Volume 98, pp. 1-33.
Inman, P., 2016. The Guardian. [Online]
Available at: https://www.theguardian.com/business/2016/jul/12/irish-economic-growth-revised-figures-foreign-investment-aircraft
[Accessed 3 December 2017].
Kenen, P., 1969. The Theory of Optimum Currency Areas: An Eclectic View. In: R. A. Mundell & A. K. Swoboda, eds. Monetary Problems of the International Economy. Chicago: University of Chicago Press, pp. 41-60.
Kenen, P. B., 2000. Currency Areas, Policy Domains, and the Institutionalization of Fixed Exchange
rates. Discussion Paper 467 ed. London: Centre for Economics and Political Science.
Krugman, P., 1991. Increasing Returns and Economic Geography. The Journal of Political Economy, 99(3), pp. 483-499.
Krugman, P. R., 1993. Lessons of Massachusetts for EMU. In: F. Torres & F. Giavazzi, eds. Adjustment
and growth in the European Monetary Union. Cambridge: Cambridge University Press and CEPR, pp.
241-261.
Martin, R. L., 2001. EMU versus the regions? Regional convergence and divergence in Euroland.
Journal of Economic Geography, Volume 1, pp. 51-80.
Massmann, M. & Mitchell, J., 2004. Reconsidering the Evidence: Are Euro Area Business Cycles Converging?. Journal of Business Cycle Measurement and Analysis, 1(3), pp. 275-307.
38 Mundell, R. A., 1961. A Theory of Optimym Currency Areas. The American Economic Review, Volume 51, pp. 657-665.
Nitsch, V., 2004. Have a Break, Have a ... National Currency: When Do Monetary Unions Fall Apart.
CESifo Working Paper Series No. 1113, pp. 1-26.
Stock, J. H. & Watson, M. W., 2001. Vector Autoregressions. The Journal of Economic Perspectives, 15(4), pp. 101-115.
Stock, J. H. & Watson, M. W., 2003. Has the Business Cycle Changed? Evidence and Explanations. Kansas City, Federal Reserve Bank of Kansas City, pp. 1-47.
A. Appendix
Table A.1: Variable Names and Definitions
Country Variable Source Composition Seasonally adjusted
Germany, France, Italy, Netherlands, Portugal, Belgium, Luxembourg, Spain, Ireland, Austria, Finland, Estonia, Latvia, Slovak Republic, Slovenia, Czech Republic, United Kingdom, Denmark, Poland, Hungary. GDP Eurostat National Accounts Gross Domestic Product. Current prices 2010 Yes GDP Deflator Eurostat National Accounts
Price Deflator for Gross Domestic Product. 2010=100
Yes
New Zealand, Australia, Canada, Japan, South Korea, United States, Switzerland, Norway. GDP OECD National Accounts Gross Domestic Product. Current prices 2010 Yes GDP Deflator OECD National Accounts
Price Deflator for Gross Domestic Product. 2010=100
39 Sweden GDP OECD National Accounts Gross Domestic Product. Current prices 2010 Changes to non-SA in October 2000 and back to SA in February 2007 GDP Deflator OECD National Accounts
Price Deflator for Gross Domestic Product. 2010=100
Changes to non-SA in October 2000 and back to SA in February 2007 Iceland GDP OECD National Accounts Gross Domestic Product. Current prices 2010 Changes to SA in February 2007 GDP Deflator OECD National Accounts
Price Deflator for Gross Domestic Product. 2010=100 Changes to SA in February 2007 Mexico GDP OECD National Accounts Gross Domestic Product. Current prices 2010 Changes to non-SA in October 2000 and back to SA in February 2007 GDP Deflator OECD National Accounts
Price Deflator for Gross Domestic Product. 2010=100
Changes to non-SA in October 2000 and back to SA in February 2007
Table A.2: Dicky-fuller Unit Root Test
Country
GDP growth InflationStatistic p-value Statistic p-value
40
Spain
-0.996 0.162 -1.018 0.156France
-2.180 0.016** -1.901 0.031**Italy
-2.549 0.006*** -3.670 0.000***Cyprus
-1.893 0.031** -3.691 0.000***Latvia
-1.792 0.039** -2.169 0.017**Lithuania
-2.767 0.004*** -2.989 0.002***Luxembourg
-3.184 0.001*** -3.707 0.002**Malta
-3.362 0.001*** -5.422 0.000***Netherlands
-1.806 0.037** -3.048 0.002***Austria
-2.448 0.008*** -2.888 0.003***Portugal
-1.967 0.026** -2.002 0.024**Slovenia
-2.027 0.023** -1.972 0.026**Finland
-2.329 0.011** -3.647 0.000***Bulgaria
-2.369 0.010*** -4.180 0.000***Denmark
-3.005 0.002*** -3.656 0.000***Hungary
-3.150 0.001*** -4.266 0.000***Poland
-2.807 0.003*** -3.258 0.001***Croatia
-2.076 0.021** -1.979 0.026**Sweden
-2.878 0.003*** -3.594 0.000***United Kingdom
-2.739 0.004*** -4.653 0.000***Australia
-2.629 0.005*** -2.610 0.005***Canada
-2.733 0.004*** -3.299 0.001***Iceland
-4.537 0.000*** -3.207 0.001***Japan
-3.110 0.001*** -2.214 0.015**South Korea
-3.152 0.001*** -2.948 0.002***Mexico
-2.850 0.003*** -2.394 0.010**New Zealand
-3.909 0.000*** -3.864 0.000***Norway
-2.709 0.004*** -2.855 0.003***Switzerland
-2.141 0.018** -1.751 0.041**United States
-1.992 0.025** -1.999 0.025**41 Table A.3: Granger causality tests
Inflation on Output Growth Output Growth on Inflation
