The Effect of Membership in European Monetary
Union on Macroeconomic Performance During
Eurozone Crisis – Lessons for Poland.
Iga Zadrzywilska (10313206) June 2016
Specialization: Economics Field: Monetary Policy Number of credits thesis: 12 ECTS Supervisor: Ron van Maurik
Abstract
This paper measures the effect of membership European Monetary Union (EMU) on macroeconomic performance by comparing the relative volatilities of key indicators in members of EMU and other European countries during Eurozone crisis and recessions. The results empirically confirm predictions of Mundell (1961) by showing that member states experienced (1) significantly larger output fluctuations during recession; (2) significantly smaller output fluctuations during the crisis; (3) losses due to convergence to common currency; (4) generally more stable inflation. The analysis also shows that EMU is not an Optimal Currency Area.
JEL: E12, E32, E42, E52, E58, E63.
P a g e | 2 Statement of originality
This document is written by Student Iga Ewa Zadrzywilska who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
P a g e | 3 CONTENTS ABSTRACT ... 1 1. INTRODUCTION ... 4 2. THEORETICAL FRAMEWORK ... 7 2.1. MONETARY UNION ... 7
2.2. OPTIMAL CURRENCY AREA ... 8
2.3. EUROPEAN MONETARY UNION AND OPTIMAL CURRENCY AREA ... 8
2.4. POLISH ECONOMIC RESILIENCE ... 9
2.5. MACROECONOMIC VOLATILITY ... 10 3. METHODOLOGY ... 11 4. EMPIRICAL ANALYSIS ... 16 4.1. DATA COLLECTION ... 16 4.2. DATA PROCESSING ... 17 4.3. SPECIFICATIONS TESTS ... 17
4.4. RESULTS AND DISCUSSION ... 20
4.5. ROBUSTNESS CHECK ... 29
5. CONCLUSIONS ... 32
APPENDIX 1. BUSINESS CYCLES IN EURO AREA AND EUROPE ... 34
APPENDIX 2. EMU MEMBERS AND ACCESSION DATES ... 36
APPENDIX 3. DATA SOURCES AND LONG DESCRIPTION ... 37
P a g e | 4 1. Introduction
Poland joined the European Union (EU) in May 2004 and under the Treaty of Accession, as all member states of the EU, it is required to adopt the euro and join the European Monetary Union (hereafter EMU). Prior to the euro adoption, each candidate state must meet certain conditions known as 'convergence criteria' or Maastricht criteria1. At the time of accession, Poland did not meet the convergence criteria, mainly due to excessive government deficit. In the opinion of European Commission (hereafter EC), as of June 2016, Poland fulfils the criteria on price stability, public finances and the convergence of long-term interest rates. It does not, however, fulfil the exchange rate criterion, because the polish zloty is not participating in the Exchange Rate Mechanism2 (EC, 2016a). Moreover, polish legislation concerning independence of the central bank, the prohibition of monetary financing and the central bank integration into the European System of Central Banks at the time of euro adoption is not in compliance with the Treaty (EC, 2016a).
According to EC (2015a), currently Poland does not have a target date to adopt the euro, but aims to do so as soon as possible. The Monetary Policy Council of Poland (2016), on the other hand, claims that polish participation in the ERM II and the euro area should only be considered in a situation that would allow for the maximization of benefits offered by currency integration and the minimization of associated risks. In theory, Poland could follow the precedent set by EC in case of Denmark and United Kingdom and negotiate opt-out arrangements. Until further actions are taken, or convergence criteria are fulfilled, Poland has the status of a “Member State with a derogation” (hereafter non-member) (EC, 2015a).
Joining a monetary union has considerable macroeconomic consequences for the accessing state, mainly due to fixing the exchange rate against common currency and abandoning autonomous monetary policy (Mishkin, Matthews, & Giuliodori, 2013). These changes give
1 (1) Price stability measured by consumer price inflation rates should be not more than 1.5 percentage points
above the rate of the three best performing Member States; (2) Soundness of public finances, measured by government deficit as % of GDP, which should not be higher that 3%; (3) Sustainability of public finances, measured by government debt as % of GDP, which should not be higher than 60%; (4) Durability of convergence, measured by long-term interest rate, which should be not more than 2 percentage points above the rate of the three best performing Member States in terms of price stability; (5) Exchange rate stability, measured by deviation from a central rate, reflected by participation in ERM II for at least 2 years without severe tensions (EC, 2015b).
2The Exchange Rate Mechanism (ERM II) is the framework for convergence of exchange rates between
Eurozone and candidate countries that requires the candidate country to fix its currency within policy margins (EC, 2014).
P a g e | 5 rise to both additional benefits and costs – namely higher stability of long-term interest rates (due to lower exchange risk premia) and inflation (due to improved expectations management); but also increased macroeconomic volatility in the events of asymmetric shocks (as a result of exchange rate buffer loss) (Gradziewicz, & Makarski, 2008). Mundell (1961), in an attempt to systematize the costs and benefits associated with monetary unions, coined the theory of Optimal Currency Area (OCA). When OCA conditions are fulfilled (namely the occurrence of asymmetric shocks is limited and stabilizing mechanisms are operating effectively), member states should incur no welfare losses resulting from the loss of autonomous monetary policy (Mundell, 1961).
According to standard macroeconomic theory, monetary policy is neutral in the long run. Hence, a change of the regime does not affect the long-run economic, but rather the volatility of economy and the business cycle (Gradziewicz, & Makarski, 2008). Considering the above arguments, euro adoption is likely associated with an increase of domestic output volatility and a decrease of inflation volatility, particularly pronounced during economic shocks.
The last available report on polish accession to EMU published by National Bank of Poland (NBP) in 2004 argued that the floating exchange rate should be considered both as an instrument stabilizing economic fluctuations, but also as a source of disturbances, magnifying those fluctuations (Borowski, et al., 2004). The authors further claimed that postponing the accession would worsen polish competitiveness relative to members.
In reality, Poland has been the only EU country to avoid recession, sustain stable GDP growth around 3% and remain internationally competitive in the aftermath of Eurozone crisis (EC, 2016b). Many economists insist that Poland’s economic autonomy insulated it from external shocks that have been affecting every other European country since 2009. In particular, they credit polish floating currency (the zloty), sharply weakened at the impact of crisis, as well as tight monetary policy (Sobczyk, 2012).
In light of this seeming contradiction and building on Mundell’s theory, I investigate the relationship between polish resilience to Eurozone crisis and its autonomous monetary policy. Using generalized least squares regression (GLS) on panel data for 28 European countries (EMU members and non-members), I estimate the extent to which membership in EMU (via change of monetary and exchange rate policy regime) can explain changes in past
P a g e | 6 economic volatility. The main focus is on output volatility, while controlling for GDP components. In line with Mundell’s assertion that presence of asymmetric shocks impedes optimality of the currency union, I additionally control for the effects associated with crisis and recessions. Gradzewicz and Makarski (2008) use a similar approach to build a model with an asymmetric monetary policy shock for Poland and Eurozone. They estimate that following euro adoption in Poland such shock would be associated with higher domestic output volatility and lower inflation volatility. They conclude, however, that the overall welfare loss due to the shock in absence of autonomous monetary policy is not large (Gradzewicz, & Makarski, 2008).
The pure focus of this paper is on measurable economic variables and not on immeasurable utility. Therefore, I avoid social and political considerations associated with European integration. It is crucial to underline that the methodology is solely focused on historical data. Eurozone crisis has proved to be exceptional in its extend and impact, and as such it should not be considered as a simple “shock”, but rather in the context of its hysteresis. Lucas (1976) in his critique argues that using econometric models estimated with past data to evaluate the response of the economy to policy change is most often incorrect and misleading. He argues that the way, in which expectations are formed changes together with the behavior of forecasted variables. As a result, the real effects of the policy change depend greatly on publics expectations about that policy (Lucas, 1976). The behavior of expectations is also excluded from this analysis. Regardless of these limitations, the methodology can provide valuable and up-to-date perspective for accession debate in Poland.
The rest of the paper proceeds as follows. Next section lays out the theoretical framework for assessing the costs and benefits of a monetary union; as well as measuring and evaluating macroeconomic volatility. Then, I build and discuss two measures of estimating the impact of membership in EMU on macroeconomic volatility – first, the two-way analysis of standard deviations of variables; then the regression model with GDP components. Next, I present and discuss the moments and the regression output. The last section concludes.
P a g e | 7 2. Theoretical framework
2.1. Monetary union
As Mishkin et al. (2013) define it, monetary union (MU) requires two or more countries to adopt a common currency managed by a common central bank. These countries irreversibly and irrevocably fix their exchange rates towards the common currency, they take part in conduct of monetary policy and governance and share the revenues from printing money (Mishkin, et al., 2013). Following are the main benefits of a monetary union:
• increased price transparency – one currency unit makes international price comparison easier, thus increasing competition and efficiency and resulting in reduction of price differentials (Mishkin, et al., 2013);
• reduction of foreign transaction costs – banks exchange commission charges are abolished (Mishkin, et al., 2013);
• elimination of exchange rate uncertainty – lower long-term interest rates reduce the fluctuations of future exchange rates (Mishkin, et al., 2013);
• trade stimulation between members of MU (Borowski, et al., 2004);
• lower volatility of inflation in member states with low anti-inflationary reputation – anchoring the inflation expectations to common MP reduces inflation uncertainty (Gradziewicz, & Makarski, 2008);
• increased stock of reserve currency – reduces costs of international financial transactions (Mishkin, et al., 2013).
Daras and Hagemejer (2009) estimate that following the accession, in the longer run, polish GDP will increase by 7.5% in comparison to benchmark scenario. The growth will mainly stem from increased inflow of foreign direct investments that will intensify capital accumulation up to 12.6% above the benchmark. The authors also predict that with falling production costs, imports and exports volume will increase by almost 13% as compared to benchmark scenario; and that consumption will go up by 3.7% (Daras, & Hagemejer, 2009). Borowski et al. (2004) estimate an additional permanent 0.2% increase of GDP due to reduction of foreign transaction costs. Overall welfare gains are estimated at 2% of GDP each year after accession (Daras, & Hagemejer, 2009).
On the other hand, main economic costs of monetary union are associated with the loss of monetary policy instruments (control over domestic exchange rate, short-term interest rates
P a g e | 8 and amount of money supplied) (Borowski, et al., 2004) and increased volatility of output (Gradziewicz, & Makarski, 2008). Indirect costs of accession arise during the process of convergence – namely adjusting the level of inflation (Borowski, et al., 2004). The specific costs associated with membership in EMU are elaborated on in the next sections.
2.2. Optimal Currency Area
According to Mundell’s (1961) theory, in an optimal MU members do not incur welfare losses after abandoning autonomous monetary policy. Since country-specific shocks cannot be addressed by the common central bank, member state hit by such a shock has to deal with it on its own. With the autonomous monetary policy, the affected country can lower its short-term interest rates and devaluate domestic currency to stimulate demand and recover from the shock promptly. With common monetary policy, countries are unable to influence their interest and exchange rates. Thus, the adjustment can be very costly if shocks are large and frequent. Mundell concludes, that in OCA asymmetric shocks between member states’ economies should be limited.
If the markets are efficient, member state can still rely on stabilizing mechanisms to restore the equilibrium in the events of asymmetric shocks. These mechanisms include flexible wage and labor markets as well as automatic transfers. In the latter case, in the event of the shock, tax revenues are redistributed as unemployment benefits. Such redistribution does not require budget centralization (Mundell, 1961).
2.3. European Monetary Union and Optimal Currency Area
Majority of existing literature argues that EMU is not an OCA. For example, Mishkin et al. (2013) pay attention to empirical evidence from 1990s to conclude that business cycles among European countries were unsynchronized and the shocks affecting them were important, but uncorrelated. Presence and significance of asymmetric shocks violates the first OCA requirement. The authors underline later evidence that once the Union has been formed, international trade between members has been intensifying and their business cycles came closer together. As a result, some ex post optimality is observed, but asymmetric shocks still occur. Other empirical studies show that EMU member states exhibit poor labor and wage flexibility (Gradziewicz, & Makarski, 2008). The former is mainly due to differences of culture, language, traditions, as well as social attachments and adjustment lags, while the latter results from high degree of unionization in European countries (Borowski, et al., 2004).
P a g e | 9 Furthermore, lack of fiscal discipline in EMU is an endeavor that dates back to before the creation of common currency. Convergence criteria of Maastricht Treaty, designed to ensure macroeconomic convergence of candidate states and to prevent negative spill-overs on low-inflation members have been subject to multiple exceptions. Moreover, lack of centralized system (common budget) makes it hard to execute regulations and control compliance. As a result, EMU faces significant moral hazard and asymmetric information problems that further increase divergence between states (Mishkin, et al., 2013). In response to Eurozone crisis, EU has initiated Stability and Growth Pact (SGP) that constraints national budgets to prevent excessive debt growth. As a fiscal constraint, however, SGP limits the effectiveness of automatic stabilizers and puts additional burden on indebted economies. Lastly, integration in Eurozone has been driven by political, rather than economic needs. As a result, Euro system resembles the biblical colosse aux pieds d'argile, rather than the optimal union it was meant to be.
2.4. Polish economic resilience
According to the European Observation Network for Territorial Development and Cohesion, Poland was the only European country fully resilient during the crisis and recessions (Bristow, et al., 2014). Polish GDP growth remains robust and stable (around 3%) and unemployment increase is offset by income and employment growth (EC, 2016a). OCA theory can, to some extent, explain why Poland did better than other countries in the region, assuming that the loss of autonomous monetary policy significantly reduces resilience. Bristow, et al. (2014), however, argue that polish exceptional economic performance resulted mainly from diversified structure, international openness, significant export rate and inflow of European funds invested in modernization that facilitated efficiency increase, while preserving low labor costs. In light of these arguments, contrary to the theory presented earlier, European integration, rather than isolation, is credited for polish resilience.
The benefits of European integration are hard to isolate and quantify. Resilience, however, defined as the ability of a region to avoid a fall in economic activity or return to its pre-shock peak levels of GDP and employment (Bristow, et al., 2014), should be associated with lower volatility of GDP and unemployment. Thus, in order to address the question of macroeconomic costs of membership in EMU, one should take a closer look at the volatility of key economic indicators.
P a g e | 10 2.5. Macroeconomic volatility
According to Keynesian business cycle theory, economies experience short-run fluctuations in economic activity caused by shocks to aggregate demand and supply. In particular, when GDP growth declines, consumption growth decreases (by a smaller amount), investment growth decreases (by a larger amount), and unemployment increases (Mankiw, 2013). Central banks have the ability to shift the aggregate demand curve and offset these shocks to maintain output and employment at their natural levels (Mankiw, 2013). However, common policy of ECB responds only to shocks affecting the entire Eurozone. Thus, in the event of an asymmetric shock, a vulnerable country has to deal with the shock on its own (Mundell, 1961). In other words, the state loses its ability to smooth business cycle fluctuations with monetary policy. As a results, it is exposed to additional and costly fluctuations that increase macroeconomic volatility (Mishkin, et al., 2013).
Cariolle (2012) reviews existing literature on the subject and concludes that macroeconomic volatility hinders long-term economic growth and welfare, mainly via decreasing long-term levels of consumption, investment and factor productivity, as well as diverting capital from the most productive sectors. The impact of volatility on growth is determined by factors such as the size of the population, the degree of diversification of the economy, the capacity for operating a countercyclical economic policy, the existence of well-developed financial institutions and institutional quality (Cariolle, 2012). Hnatkovska and Loayza (2005) also show that “crisis”3 level of volatility has significantly larger negative effect on growth than “normal” volatility. Gradziewicz and Makarski (2008) pay attention to scare empirical evidence that changes in business cycle behavior have any long-run effects on economy and focus solely on changes in volatility and cyclicality of main macroeconomic variables.
P a g e | 11 3. Methodology
Several polish studies on costs and benefits of accessing Eurozone investigate relative change in volatility and cyclicality of key macroeconomic variables (output and its components, unemployment, inflation and long-term interest rates) associated with adoption of the common monetary policy (e.g. Gradziewicz, & Makarski, 2008; Daras, & Hagemejer, 2008).
Since movements in aggregate output are of key interest in this analysis, I will consider composition of GDP as in Keynesian aggregate demand function, where gross domestic product is an aggregate measure of production equal to the sum of the gross values added of all resident institutional units engaged in production (Mankiw, 2013):
𝐺𝐷𝑃 = 𝑌 = 𝐶 + 𝐼 + 𝐺 + 𝑋 − 𝑀
The total quantity demanded of an economy (Y) is equal to the sum of the final4 uses of goods and services by consumers5 (C), businesses6 (I), government (G) and foreign spending on domestic production (E), less domestic spending on foreign production (M) (Mankiw, 2013). Thus, I include volatilities7 of above components as control variables in estimating output volatility.
Alonso and Furceri (2008) pay attention to the importance of time span used in cross-country regressions. The aim of this analysis is to conclude whether or not membership in EMU is associated with increased volatility of output; and whether or not this relationship changed during crisis and recessions. Thus, developments before creation of EMU in 1999 are not of interest. To account for impact of crisis, I use the following sequence of events in the Eurozone crisis proposed by Mishkin et al. (2013):
(1) initiation of the crisis – period characterized by exacerbating fiscal imbalances since the creation of EMU (1999);
4 All uses except intermediate consumption, measured in purchasers' prices.
5 𝐶
,-= 𝑎 + 𝑚𝑝𝑐 ∙ (𝑌 − 𝑇); consumption equals autonomous consumer expenditure (essential consumption expenditure independent of income) plus marginal propensity to consume (change in consumption expenditure resulting from an additional euro of disposable income) times disposable income (income less taxes).
6 𝐼
,- = 𝐼(𝑟); where investment decreases with interest rates (r).
P a g e | 12 (2) banking crisis – characterized by policy of austerity and bailouts, followed by
recovery in some countries (2008-2009);
(3) sovereign debt crisis – continuing downturn in GIIPS8 countries (2011);
Accordingly, I include a period dummy Crisis that assumes value 1 for years 2008, 2009 and 2011 to control for shocks associated with financial and sovereign debt crisis.
Furthermore, to control for economic slowdowns experienced by most countries in EU, I control for times of recession9, which begin just after the economy reaches a peak of activity and end when the economy reaches its trough. According to OECD (2016) data, Eurozone and other European states followed roughly the same path of business cycle between 1999 and 2014 and experienced recessions in 2001, 2002, 2008 and 2012.10 Period dummy Recession assumes value 1 in these years.
Lastly, in order to account for impact of membership in EMU on volatility of output, I include dummy Member indicating membership from the year of accession on11. To test the hypothesis that there is statistically significant difference between volatilities of variables for members and non-members, the dummy enters the regression as an interaction term with all the other variables.
Longitudinal time-series can be used to model economic behaviors of countries observed over time while controlling for unobservable differences across those countries (Torres-Reyna, 2007). While analyzing panel data, however, one should be aware of problems arising from correlation between countries. In this study, countries may share some regional characteristics. Furthermore, there is a considerable degree of internationalization of policies due to centralization in European Parliament and European Central Bank.
8 Greece, Italy, Ireland, Portugal, Spain.
9 Defined as a significant decline in the level of economic activity, spread across the economy of the euro area,
usually visible in two or more consecutive quarters of negative growth in GDP, employment and other measures of aggregate economic activity for the euro area as a whole (CEPR, n.d.)
10 See Appendix 1 for detailed description of business cycles.
P a g e | 13 Panel regression model is given by the following equation12:
𝑌,- = 𝛼 + 𝛽𝑋′,-+ 𝛾𝑍′,+ 𝑐,+ 𝑢
,-𝛼: intercept
𝛽 and 𝛾: column vector of parameters
𝑋′,-: row vector of time-varying explanatory variables
𝑍′,: row vector of time-invariant explanatory variables 𝑐,: country-specific effect
𝑢,-: idiosyncratic error term
In a balanced panel each individual i is observed in all time periods t.
Two most common techniques used in analyzing panel data differentiate between fixed and random effects in regression. When differences across the countries have some influence on volatility of output, random effects model is preferred13. More specifically, it allows for inclusion of time-invariant variables (like dummy Member that does not vary once the country has accessed EMU), which would have otherwise got absorbed in the intercept in the fixed effects model (Torres-Reyna, 2007). Additionally, RE has the advantage of allowing for generalization of inferences beyond the modeled sample.
Using time-invariant variables (Member) as explanatory variables requires that each countries’ error term is uncorrelated with other explanatory variables14:
𝐸 𝑐, 𝑋,, 𝑧, = 0
Letting: 𝑣,- = 𝑐, + 𝑢,-, random effects model can be written as:
𝑌,- = 𝛼 + 𝛽𝑋′,-+ 𝛾𝑍′,+ 𝑣
,-and can be estimated using generalized least squares (GLS) regression, which allows for heteroscedasticity and possible serial correlation (Stock, & Watson, 2003). When considering
12 Model is linear in parameters 𝛼, 𝛽, 𝛾, effect error 𝑐
, and error 𝑢,-,the observations are independent across countries but not necessarily across time, idiosyncratic error term 𝑢,- is assumed uncorrelated with the explanatory variables of all past, current and future time periods of the same individual (no time lags possible) (Schmidheiny, 2015).
13 See Section 4.3 for specification tests.
14 The variance of effect can be either homoscedastic (with constant variance of the country-specific effect)
P a g e | 14 raw GDP data, it is likely that the error in predicting the GDP of a rich country will probably be larger than that of a poor country. Therefore, cluster-robust standard errors should be used to correct for heteroscedasticity.
To avoid omitted variable bias, it is necessary to specify those individual characteristics that may or may not influence the explanatory variables (Torres-Reyna, 2007). The following table combines components of aggregate demand, variables affecting these components according to Keynes (Mishkin, et al., 2013) and probable individual country characteristics that may or may not influence the GDP components in Europe:
Component Keynesian variables Individual characteristics
Consumption Marginal propensity to consume Autonomous consumption expenditure “Animal spirits”15 Taxes
Initial wealth (inversely related to marginal propensity to consume)
Geography (e.g. developed west vs. east in transition)
Fiscal regime and practices (e.g. tight or expansionary)
Current and historical political regimes and practices (e.g. populism, time inconsistency)
Planned investment
Expectations about the future (“Animal spirits”)
Interest rates
Business practices
Degree of financialization of the market Risk aversion of investors
Government spending
Manipulation of government spending
Fiscal regime and practices (e.g. tight or expansionary)
currents or historical political regimes and practices (e.g. populism, time inconsistency, loss aversion)
Net exports
Interest rates Trade openness (economic and cultural)
Exchange rate system (fixed or floating)
P a g e | 15 As previously mentioned, random effects regression requires accounting for the individual characteristics. For the purpose of this analysis, and under the following assumptions, most of these characteristics can be explained by membership in EMU:
(1) initial member states shared some common socio-economic traits and were at similar level of development;
(2) the Stability and Growth Pact (SGP) is a set of rules designed to ensure that countries in the EU pursue sound public finances and coordinate their fiscal policies (EC, n.d.); (3) EMU member states comply with the governance of ECB; non-member states strive
to fulfill convergence criteria;
(4) in the long run, accessing EMU leads to more economic integration (Borowski, et al., 2004);
(5) member states are part of fixed exchange rate system, non-member states operate under floating system (EC, 2016).
The assumptions impose limitations on the model, but are necessary for compliance with exogenity assumption of the random effects regression, so that membership in EMU can be included as an endogenous variable (Schmidheiny, 2015).
All in all, the analysis focuses on combined cross-section time-series regressions using data from 1999 to 2014 on 28 European countries. The following two regressions for output volatility are estimated using random effects model:
𝜎D
,- = 𝛽E𝜎F,-+ 𝛽G𝜎H,- + 𝛽I𝜎J,-+ 𝛽K𝜎L,-− 𝛽M𝜎N,-+ 𝛿E𝐶 + 𝛿G𝑅 (1)
𝜎D
,- = 𝛽E𝜎F,-+. . . −𝛽M𝜎N,-+ 𝛿E𝐶 + 𝛿G𝑅 + 𝛽R(𝜎F,-∙ 𝑀𝑒𝑚𝑏𝑒𝑟)+. . . −𝛽EU(𝜎N∙
𝑀𝑒𝑚𝑏𝑒𝑟),-+ 𝛿I(𝐶 ∙ 𝑀𝑒𝑚𝑏𝑒𝑟) + 𝛿K(𝑅 ∙ 𝑀𝑒𝑚𝑏𝑒𝑟) (2)
Where the index i (i = 1, …, 28) denotes the country, the index t (t = 1999, …, 2014) indicates the year, 𝜎 denotes derived volatility of output and its components, C and R stand for crisis and recession dummies respectively. Interaction terms Member indicating membership in EMU are added in the second regression.
P a g e | 16 4. Empirical analysis
4.1. Data collection
In this paper, I focus on members of European Monetary Union and non-members, defined as member states of European Union with derogation or opt-out status. The countries included in the analysis are EU18 (excluding Malta16): Austria, Belgium, Cyprus, Estonia, Finland, France, Germany, Greece, Ireland, Italy, Latvia, Lithuania, Luxemburg, Netherlands, Portugal, Slovakia, Slovenia, Spain; and the rest of EU: Bulgaria, Croatia, Czech Republic, Denmark17, Hungary, Poland, Romania, Sweden, Switzerland, United Kingdom18.
The sample has been selected based on data availability, quality and cross-country comparability. Data on real GDP and its components as well as unemployment rate are obtained from WorldBank database19. The data share a common set of coefficients for all countries and years, which contributes to higher quality and transparency in comparison. All components are expressed in current US dollars, which improves comparability but may hamper the volatility measures due to exchange rate differentials. The data set is strongly balanced, which improves the quality of panel regression.
Other macroeconomic indicators used in the analysis include inflation and long-term interest rates. For the purpose of convergence assessment, ECB complies Harmonized Index of Consumer Prices (HICP) as a measure of inflation. HICP is based on a “typical basket” of goods and services bought by European households (Mishkin, et al., 2013). By convention, as proxy for harmonized long-term interest rates, ECB uses yields on 10-year government bonds. To ensure compliance with ECB conventions, data on the last two variables are obtained from Eurostat and expressed in euros.
16 Omitted due to data incompleteness. 17 Member of ERM II since 1999.
18 EU member state at the time of publishing. 19 See Appendix 3 for detailed description.
P a g e | 17 4.2. Data processing
First of all, in order to improve interpretability of real data characterized by large values, I apply logarithmic transformations to all variables. Thus, regression model will take a log-log form, where coefficients on variables represent elasticities (Stock, & Watson, 2003). Then the volatilities are computed.
Cariolle (2012) points out that measuring economic volatility involves evaluating the deviation between the values of an economic variable and its equilibrium value (permanent state or trend). Most frequently, researchers use standard deviation of the growth rate of a variable as a measure of volatility (Cariolle, 2012). This approach is limited by a priori assumptions about behavior of variables it imposes. It provides, however, a good starting point for the analysis. Because only one standard deviation can be computed per variable in a given time period, I pool the data into three time periods: (1) before the crisis (1999-2007); (2) the financial crisis (2008-2009); and (3) sovereign debt crisis (2010-2014), while controlling for membership. The time windows were chosen in line with ECB (2011) and OECD (2016) business cycle analysis for Eurozone and European countries.
Alternatively, volatility can be computed using variables detrended with a statistical filter. Following Jetter (2013), I decompose series into their cyclical (short-term) and trend (long-term) components using Hodrick-Prescott (HP) filter with smoothness parameter (λ) adjusted for annual data, i.e. equal to 6.25 (Ravn, & Uhlig, 2002). Since the focus of this analysis is volatility in general, in order to capture deviation from trend in any direction, I compute annual volatilities of output (Y) and its components (C, I, G, X, M) by squaring annual cycle terms isolated using HP filter for each country (Jetter, 2013). Figure 1 and 2 exhibit GDP volatilities computed using the filtering method.
P a g e | 18 Figure 1. Squared log of GDP for members (HP 6.25)
Figure 2. Squared log of GDP for non-members (HP 6.25)
The main advantage of using filtering method, rather than the simple standard deviations of variables, is that it does not make any a priori assumptions about the behavior of the series and allows the trend to change over time (Cariolle, 2012). Furthermore, the standard
P a g e | 19 deviation approach produces only one volatility measure per time period considered. Thus, when estimating the statistical significance of the impact of membership on output volatility over an extended period of time, a regression using filtered variables is preferred. Additionally, the simple standard deviation analysis generalizes membership throughout the periods. As a result, it is insensitive to immediate volatility changes due to later accessions (2004 and 2005). Regression analysis solves this problem by including dummy Member that indicates membership from the year of accession on.
4.3. Specifications tests
Hausman test evaluates the consistency and efficiency of an estimator. If the model is correctly specified and there is no correlation between independent variables and country-specific effects, then the coefficients estimated by fixed-effects estimator and random-effects estimators should not statistically differ. There is significant evidence to infer that the differences in coefficients are not systematic and the unique errors are not correlated with the regressors, 𝜒G(14, N = 448) = 12.43, p = .57. Thus, random effects model is preferred.
Figures 3 and 4 confirm the above conclusion. Data displays homogeneity across years, but heterogeneity across countries.
P a g e | 20 Figure 4. Heterogeneity across countries.
Breusch and Pagan Lagrandian multiplier tests for significant differences across units or panel effects. The test fails to reject the hypothesis that random effects are not appropriate, 𝜒G(1, N = 448) = 0, p = 1.00. There is significant evidence to infer that variances across
countries are non-zero and random effects model is preferred to OLS.
Pesaran test of cross-sectional independence tests whether the residuals are correlated across countries, which could lead to bias in test results (Torres-Reyna, 2007). With CD = 1.1277 (p = 0.24), there is significant evidence of cross sectional independence.
Lastly, Wooldridge test for autocorrelation in panel data detects serial correlation in macro panels that causes the standard errors of the coefficients to be smaller and R2 to be higher than they actually are. With F (1, 27) = 0.02, p = 0.89 the test provides significant evidence to conclude that data does not have first-order autocorrelation. 20
4.4. Results and discussion
Table 1 reports volatilities of variables for EMU members and non-members measured by their standard deviations in three periods. It also shows relative change of volatility
P a g e | 21 associated with membership. It emerges that output volatility is smaller in member states than non-member states before the crisis in 2007, but larger afterwards, that is during financial and sovereign debt crisis. In particular, financial crisis (period 2) output volatility in member states is almost four times as high as in period 1, but returns to initial levels in period 3. Non-members faced less than a twofold increase in output volatility during financial crisis and lower-than-initial volatility during debt crisis. The results suggest that the financial crisis volatility was, on average, more detrimental for members than non-members. Members were relatively resilient and able to return to pre-crisis levels of volatility, but the recovery was more pronounced in non-member states. The preliminary findings support the hypothesis that the loss of monetary policy tools exacerbates output losses during the crisis due to higher output volatility; but there is no significant evidence that resilience is related to membership.
Table 1. Volatilities measured as standard deviations of variables
Period 1 Period 2 Period 3
Non-member Member Change
Non-member Member Change
Non-member Member Change
∂Y 1.917 1.328 -31% 3.762 4.880 30% 1.167 1.328 14% ∂C 0.133 0.104 -22% 0.172 0.132 -23% 0.063 0.060 -5% ∂I 0.210 0.135 -36% 0.285 0.215 -25% 0.093 0.125 34% ∂G 0.155 0.108 -30% 0.149 0.120 -19% 0.069 0.070 1% ∂X 0.144 0.132 -8% 0.203 0.194 -4% 0.099 0.085 -14% ∂M 0.185 0.142 -23% 0.236 0.213 -10% 0.095 0.099 4% ∂U 0.178 0.262 47% 0.219 0.388 77% 0.120 0.149 24% HICP 19.495 2.946 -85% 2.750 2.035 -26% 1.804 1.365 -24% ∂r 2.252 3.207 42% 1.780 1.750 -2% 1.912 3.196 67%
Period 1: 1999-2006; Period 2: 2007-2009; Period 3: 2010-2014. ∂ represents growth rates of variables calculated as relative changes: DWDXDY
Y .
In general, inflation volatility in non-member countries has been systematically decreasing, yet in every period it is larger for members than members. Most profoundly, non-members experienced seven times larger volatility than non-non-members before 2007. It seems that delegation of monetary policy to ECB is associated with decrease of inflation volatility, particularly during the debt crisis (one possible explanation emerges from unconventional monetary policy). Inflation stability, in turn, is associated with more stable economic growth. Figures 5 and 6 display these findings – it is visible that HICP of member states is less variable and moves closely together for most countries, while in non-member states it is more spread out and variable.
P a g e | 22 Figure 5. Volatility of HICP for members (HP 6.25)
P a g e | 23 Considering the volatility of long-term interest rates, it appears to be larger for members than non-members in the first and third period and about equal in the second period. That finding is contradicting the theory that members of a monetary union face lower and more stable interest rates due to reduction of risk premia and lower long-term yields on government bonds (Borowski, et al., 2004). A posteriori analysis of Eurozone crisis shows that financial markets incorrectly priced government debts of EMU members at the initial stage of convergence. As a result, interest rates in GIIPS were significantly (but incorrectly) lowered in period 1. After the crisis unraveled in period 2, investors reevaluated the prices, and interest rates peaked. As a result, there were significant fluctuations of interest rates in EMU unpredicted by the theory.
Lastly, it emerges that members experienced higher unemployment volatility. The precise effect, however, is ambiguous – figure 7 exhibits unequal distribution of unemployment in EMU mainly due to large rates in GIIPS countries.
P a g e | 24 Table 2 presents cyclicality measures of each variable computed as correlation with GDP. Even though the size of measures differs for members and non-members, the sign (indicating cyclicality for positive numbers and countercyclicality for negative) is the same. The exceptions are inflation during debt crisis, which is is countercyclical for non-members and cyclical for members; and interest rates before 2007, which are countercyclical for non-members and cyclical for non-members. Furthermore, all GDP components exhibit cyclicality for both members and non-members, which stands in line with definition of GDP by expenditure approach. Lastly, there was a significant increase in (counter-)cyclicality during the financial crisis for both members and non-members, which indicates an overall increase in macroeconomic volatility.
Table 2. Cyclicality measured as correlation with GDP
Period 1 Period 2 Period 3
Non-member Member Change
Non-member Member Change
Non-member Member Change
∂Y 1 1 1 1 1 1 ∂C 0.271 0.113 -58% 0.834 0.896 7% 0.626 0.533 -15% ∂I 0.592 0.327 -45% 0.848 0.932 10% 0.661 0.699 6% ∂G 0.136 0.000 -100% 0.835 0.782 -6% 0.656 0.573 -13% ∂X 0.530 0.313 -41% 0.852 0.842 -1% 0.416 0.461 11% ∂M 0.622 0.369 -41% 0.861 0.880 2% 0.554 0.635 15% ∂U -0.379 -0.437 15% -0.822 -0.792 -4% -0.421 -0.630 50% HICP 0.368 0.217 -41% 0.415 0.251 -40% -0.218 0.040 -118% ∂r -0.255 0.051 -120% -0.240 -0.465 94% -0.252 -0.613 143%
Period 1: 1999-2006; Period 2: 2007-2009; Period 3: 2010-2014. ∂ represents growth rates of variables calculated as relative changes: DWDXDY
Y .
Table 3 reports estimates of the effect of the volatility of GDP components on GDP volatility (using the HP filter for annual data with smoothness parameter λ = 6.25) (Regression 1) and the combined effect of components and membership (Regression 2). In GLS regression, coefficients include both within- and between-country effects – in case of time-series cross-section data, coefficients represent the average effect of X over Y when X changes across time and between countries by one unit (Torres-Reyna, 2007). In regression 2, one-unit increase in dummy Member indicates the difference being and not being a member of EMU.
P a g e | 25 Table 3. GDP volatility expenditure composition including period dummies
and interaction terms (in Regression 2)
Variable Regression 1 Regression 2
Control variables σC .945*** 1.007*** (15.43) (20.73) σI .052* .047* (2.41) (2.50) σG .005 -.015 (0.10) (-.32) σX .145 .085 (3.41) (1.66) σM -.104* -.084* (-2.91) (-2.11) Recession -.000 -.000 (-0.04) (-.66) Crisis -.000 .000 (-.14) (1.52) Interaction terms Member . σC -.249 (-1.91) Member . σI .041 (1.62) Member . σG .043 (0.42) Member . σX .147* (2.03) Member . σM -.079 (-1.25) Member . Recession .001* (2.19) Member . Crisis -.001** (-2.80) No. obs. 448 448
Notes: t-statistics are in parenthesis. Robust standard errors to control for heteroscedasticity. *,**,*** - Statistically significant at 10, 5 and 1 percent level respectively.
σ denotes volatilities of variables.
Member – dummy variable that assumes the value 1 for all the EMU countries. Crisis – dummy variable that assumes the value 1 for years 2008, 2008 and 2012.
Recession – dummy variable that assumes the value 1 for years 2001, 2002, 2008 and 2012. The HP6.25 filter was used to decompose the series.
P a g e | 26 According to Mankiw (2013), consumption is the largest share of GDP (approximately 2/3) and investment volatility is the source of business cycle fluctuations due to the inventory component. In both regressions, on average, consumption and investment volatility significantly contribute to higher output volatility in all years, which confirms the theory presented by Mankiw.
The opposite is true for import volatility – on average, when imports fluctuate more, consumption tends to be more stable. Additionally, it appears that the effect of government expenditure volatility on GDP volatility is negligible.
The effect of membership on output volatility is generally significant, except for consumption, imports and government expenditure volatilities.
On average, import and export volatilities in member states has stronger positive effect on output volatility than in non-member states. This conclusion is in line with the assertion that EMU members tend to trade more with each other as their markets integrate (Borowski, et al., 2004).
It emerges that recession in member states contributes to higher output volatility than in non-member states, while the opposite is true for crisis. Specifically, during the recessions EMU members experienced on average 0.001% higher output volatility than non-members 𝜒G(1, N = 448) = 2.36, p = .12. During the crisis, output volatility was on average 0.001%
lower in member states than in non-member states 𝜒G(1, N = 448) = 5.59, p = .02. Again,
as in the two-way analysis, it appears that members of EMU experience more severe recessions and take more time to recover to peak levels of economic activity than non-members. On the other hand, contrary to the hypothesis and findings in table 1, it seems that non-members experienced larger fluctuations of output than members during the crisis. The computation of volatility as square of cyclical component of the variable is insensitive to the direction of fluctuations and treats decline and growth equally. Figure 8 shows how output growth rates were evolving in each country and indicates that higher volatility was associated with growth decline. These finding are in line with 2011 ECB report, which concludes that due to deep trough in output in 2009, GDP volatility has increased substantially in euro area during the recession but has been decreasing since (ECB, 2011).
P a g e | 27 Figure 8. Output growth rates per country
Table 4 presents specific volatilities of output per period and per country, as well as averages for members and non-members. It emerges that high average volatility of output in Eurozone in period 2 (during the crisis) is mainly explained by high volatilities of new entrants – Estonia (2011), Latvia (2014), Lithuania (2015), Slovakia (2009) and Slovenia (2007). It is worth pointing out that these countries share some regional and historical similarities with Poland, thus observing their performance can yield valid conclusions for polish economy. Borowski et al. (2004), pay attention to the fact that the convergence towards common currency (via fulfilling convergence criteria and later participation in ERM II) imposes short-run costs on the economy due mainly due to inflation adjustment. Hence, the volatility of GDP in period 2 presented in table 1 likely captures the combined effect of membership during crisis and costs of inflation adjustment. Notably, after the crisis, the volatilities dropped sharply indicating rapid economic improvements and increased resilience after adjustment has been completed. As a result, the conclusion that in general members experienced smaller output fluctuations during crisis is more valid.
P a g e | 28 Table 4. Standard deviations of output in non-member (left) and member states (right)
P a g e | 29 4.5. Robustness check
A common practice in examining possible misspecifications in regression is to modify or remove regressors in order to observe the behaviors of some “core” regression coefficient estimates (Lu, & White, 2014). In such setting, structural validity is proved by plausibility and robustness of coefficients. Lack of robustness likely indicates omitted variable bias.
Alonso and Furceri (2008) underline that when specifying panel regression, one should consider including country variables to control for unobserved heterogeneity. GLS regression coefficients absorb both within- and between- country effects. Thus, in order to control for possible misspecification due to exclusion of country dummies, I re-estimate regression 1 adding country dummies (table 5). Next, to control for possible misspecification due to inclusion of period dummies, I re-estimate regression 2 without period dummies (table 6). Both of the re-estimated results are robust – as in previous regression, average consumption and investment volatility contribute to higher output volatility in all years, the opposite is true for import volatility and it appears that the effect of government expenditure volatility on GDP volatility is negligible.
P a g e | 30 Table 5. GDP volatility expenditure composition with interaction terms,
without period dummies (Re-estimated Regression 2)
Variable Regression 2 Re-estimated Regression 2
Control variables σC 1.007*** 1.002*** (20.73) (19.66) σI .047* .047* (2.5) (2.49) σG -.015 -.015 (-.32) (-.32) σX .085 .092 (1.66) (1.68) σM -.084* -.080* (-2.11) (-1.99) Recession -.000 (-.66) Crisis .000 (1.52) Interaction terms Member . σC -.249 -.196 (-1.91) (-1.48) Member . σI .041 .045 (1.62) (1.71) Member . σG .043 .041 (.42) (0.39) Member . σX .147* .136 (2.03) (1.78) Member . σM -.079 -.097 (-1.25) (-1.52) Member . Recession .001* (2.19) Member . Crisis -.001** (-2.80) No. obs. 448 448
Notes: t-statistics are in parenthesis. Robust standard errors to control for heteroscedasticity. *,**,*** - Statistically significant at 10, 5 and 1 percent level respectively.
σ denotes volatilities of variables.
Member – dummy variable that assumes the value 1 for all the EMU countries. Crisis – dummy variable that assumes the value 1 for years 2008, 2008 and 2012.
Recession – dummy variable that assumes the value 1 for years 2001, 2002, 2008 and 2012. The HP6.25 filter was used to decompose the series.
P a g e | 31 Table 6. GDP Volatility expenditure composition including period
and country dummies (in Re-estimated Regression 1)
Variable Regression 1 Re-estimated Regression 1
Control variables σC .945*** .950*** (15.43) (11.98) σI .052* .054* (2.41) (2.81) σG .005 -.002 (.10) (-.03) σX .145 .15*** (3.41) (3.92) σM -.104* -.11** (-2.91) (-3.48) Recession -.000 -.000 (-.04) (-.04) Crisis -.000 -.000 (-.14) (-.21) No. obs. 448 448
Notes: t-statistics are in parenthesis. Robust standard errors to control for heteroscedasticity. *,**,*** - Statistically significant at 10, 5 and 1 percent level respectively.
σ denotes volatilities of variables.
Crisis – dummy variable that assumes the value 1 for years 2008, 2008 and 2012.
Recession – dummy variable that assumes the value 1 for years 2001, 2002, 2008 and 2012. The HP6.25 filter was used to decompose the series.
Another threat to internal validity results from possible reverse causality between membership in EMU and GDP volatility. Existing literature does not, however, suggest that such causality exists.
Lastly, using component form of GDP inevitably imposes high correlation between variables21. High correlations can lead to multicollinearity, which in turn lowers reliability and stability of estimates of regression coefficients (variances of a coefficients are “inflated” because of linear dependence with other predictors). Multicollinearity cannot be easily resolved, but can be ignored if collinear variables are only used as control variables (which is true for GDP components in my regressions), while the variables of interest (here membership in EMU) are not affected by collinearity (Stock, & Watson, 2003). All in all, high correlations between GDP components neither affect their performance as controls, nor distort coefficients on dummy Member.
P a g e | 32 5. Conclusions
Amid Eurozone crisis, the question of EMU expansion gave place to the question of its very existence. Many economists in Poland credit polish economic autonomy for its exceptional performance during recent crisis and recession. On the other hand, broader analysis of regional macroeconomic performance in Europe suggests that polish resilience stems from ongoing integration with EU. Both views are to some extend supported by theory and one does not rule out another. More importantly, they disagree about relative size of costs and benefits of polish accession to EMU. It is therefore quite relevant for the current debate to access the validity of these arguments. Therefore, in this paper I investigate the impact of membership in EMU on macroeconomic performance between 1999 and 2014 for the set of 28 countries, including EMU members and other EU states.
The results suggest that membership had statistically significant impact on GDP volatility and thus macroeconomic performance. In particular, due to increased trade intensity within EMU, member states experienced stronger fluctuations in GDP due to export and import fluctuations. Moreover, membership during recessions is associated with 0.001% higher output volatility. Defining resilience as the ability of an economy to recover after a shock, it follows that more resilient economies exhibit lower volatilities of output (Bristow, et al., 2014). Higher volatility, on the other hand, impedes long-term economic growth and welfare (Cariolle, 2012), particularly in times of crisis and recessions (Hnatkovska, & Loayza, 2005). Accordingly, the results suggest that membership during times of recession is more detrimental to economic growth, which confirms theoretical predictions proposed by Mundell (1961) and that European Monetary Union is not and Optimal Currency Area. Specifically, the countries that retained autonomous monetary policy instruments (such as Poland) were in general able to recover faster from losses, especially in comparison with GIIPS countries.
Some interesting lessons emerge from analysis of individual output volatilities. First of all, new entrants experienced much higher volatilities during the crisis, likely due to the combined effect of crisis and costs of inflation adjustment, which proves that accession imposes additional costs on the economies in the short run (Borowski, et al., 2004). At the same time, the new entrants exhibited a large degree of resilience in later stages of integration. This observation undermines the assertion that polish resilience derives from economic autonomy.
P a g e | 33 The results also suggest that membership during crisis is associated with 0.001% lower output volatility and smaller losses. Moreover, in the longer run, delegation of monetary policy to ECB during the crisis is associated with decrease of inflation volatility, except during sovereign debt crisis. These findings confirm the characteristics of monetary union presented in section 2.1. Specifically, benefits stemming from inflation stability and increased trade intensity (Borowski, et al., 2004). The effect of membership on volatility of interest rates is ambiguous due to markets’ failure to correctly price the government bonds of GIIPS countries. The effect on unemployment variability is also ambiguous due to large outliers in GIIPS country.
In general, membership in EMU had stronger distortionary effect on macroeconomic performance of countries in consideration during recessions. On the other hand, members on average performed better during the financial and sovereign crisis. In the longer run perspective, members seem to benefit from more stable inflation. All in all, there is no significant evidence that the resilience of polish economy can be explained by retention of autonomous monetary policy. It reasonable to assume the economy benefited from the combination of factors, including diversified structure; international openness; inflow of European funds; vigorous modernization; efficiency increase; low labor costs; as well as prompt and rigorous monetary and fiscal policy response.
The analysis in this paper is limited by its sole focus on historical performance and omission of the notion of expectations, which according to Lucas critique produces incorrect inferences about the future (Lucas, 1976). In order to provide unambiguous conclusion, one should consider longer term perspective and consider all the potential sources of costs and benefits. For instance, a simulation by Borowski et al. (2004) predicts that as a result of accession, polish GDP level will be higher by 5.6–11.8% in comparison to the baseline scenario, mainly due to growth of investment. Thus, further empirical research should focus on estimating probable evolution of additional variables, such as foreign direct investments and productivity.
P a g e | 34 References
Afonso, A., & Furceri, D. (2008). Government Size, Composition, Volatility and Economic Growth. Frankfurt am Main: European Central Bank.
Borowski, J., Brzoza-Brzezina, M., Czogala, A., Fic, T., Kot, A., Jedrzejowicz, T., et al. (2004). A Report on the Costs and Benefits of Poland's Adoption of the Euro. Warsaw: National Bank of Polands.
Bristow, G., Healy, A., Norris, L., Kafkalas, G., Kakderi, C., Wink, R., et al. (2014). Economic Crisis: Resilience of Regions. Luxembourg: The European Observation Network for Territorial Development and Cohesion.
Business Cycle Dating Committee. (n.d.). Business Cycle Dating Committee Methodology. Retrieved June 2016, 14, from
http://cepr.org/content/business-cycle-dating-committee-methodology
Cariolle, J. (2012). Measuring macroeconomic volatility - Applications to export revenue data, 1970-2005. I14.
Daras, T., & Hagemejer, J. (2009). The Long Run-Effects of the Poland’s Accession to the Eurozone. Warsaw: National Bank of Poland.
European Commission. (2014). What is ERM II? Retrieved June 12, 2016, from http://ec.europa.eu/economy_finance/euro/adoption/erm2/index_en.htm
European Commission. (2015a). Introduction of the Euro in the Member States that have not yet Adopted the Common Currency. Brussels: European Commission.
European Commission. (2015b). Who can join and when? Retrieved May 09, 2016, from http://ec.europa.eu/economy_finance/euro/adoption/who_can_join/index_en.htm European Commission. (2016a). European Economic Forecast. Winter 2016. Luxemburg:
European Commission.
European Commission. (2016b). Timeline: The Evolution of EU Economic Governance in Historical Context. Retrieved May 09, 2016, from
http://ec.europa.eu/economy_finance/economic_governance/timeline/index_en.htm Feldstein, M. S. (2011). The Euro and the European Economic Conditions. 57652.
Gradziewicz, M., & Makarski, K. (2008). The Welfare Cost of Monetary Policy Loss after the Euro Adoption in Poland. Warsaw: National Bank of Poland.
Hnatkovska, V., & Loayza, N. (2005). Volatility and Growth. Managing Economic volatility and Crises, 14.
P a g e | 35 Jetter, M. (2013). Volatility and Growth: Governments are Key. Bonn: Institute for the Study
of Labour.
Lu, X., & White, H. (2014). Robustness checks and robustness tests in applied economics. Journal of Econometrics, 178, 194-206.
Lucas, R. E. (1976). Econometric Policy Evaluation: A Critique. Carnegie-Rochester Confer. Series on Public Policy, 19–46.
Mankiw, N. G. (2013). Macroeconomics Eighth Edition. New York: Worth Publishers. Mishkin, F., Matthews, K., & Giuliodori, M. (2013). The Economics of Money, Banking, and
Financial Markets (European Edition). Harlow: Pearson Education.
Monetary Policy Council. (2016). Monetary Policy Guidelines 2016. Warsaw: National Bank of Poland.
Mundell, R. A. (1961). A Theory of Optimum Currency Areas. American Economic Review, 657–665.
Ravn, M., & Uhlig, H. (2002). On Adjusting the Hodrick-Prescott Filter for Frequency of Observations. Review of Economics and Statistics, 371-380.
Schmidheiny, K. (2015). Panel Data: Fixed and Random Effects. Basel: Basel University. Sobczyk, M. (2012, June 5). Euro’s Popularity Hits Record Low in Poland. Retrieved May
10, 2016, from The Wall Street Journal:
http://blogs.wsj.com/emergingeurope/2012/06/05/euro’s-popularity-hits-recordlow-in-poland/
Stock, J., & Watson, M. (2003). Introduction to Econometrics Third Edition. New York: Prentice Hall.
Torres-Reyna, O. (2007). Panel Data Analysis Fixed and Random Effects Using Stata. Princeton: Princeton University.
P a g e | 36 Appendix 1. Business cycles in Eurozone and Europe
Eurozone OECD Europe Trough 1999M1 1999M3 Peak 2001M1 2000M10 Trough 2003M7 2003M6 Peak 2008M2 2008M1 Trough 2009M6 2009M5 Peak 2011M5 2011M6 Trough 2013M3 2013M2 Source: OECD, 2016.
Appendix 2. EMU member states and accession dates
Country Accession Austria January 1, 1999 Belgium January 1, 1999 Cyprus January 1, 2008 Estonia January 1, 2011 Finland January 1, 1999 France January 1, 1999 Germany January 1, 1999 Greece January 1, 2001 Ireland January 1, 1999 Italy January 1, 1999 Latvia January 1, 2014 Lithuania January 1, 2015 Luxembourg January 1, 1999 Malta January 1, 2008
The Netherlands January 1, 1999 Portugal January 1, 1999 Slovakia January 1, 2009 Slovenia January 1, 2007
Spain January 1, 1999
P a g e | 37 Appendix 3. Data sources and long description
Volatility of Output (σY or HPY2) – Square of the cyclical component of the log of GDP at purchaser's prices (in current U.S. dollars, converted from domestic currencies using single year official exchange rates) isolated using the HP filter for annual data with smoothness parameter λ = 6.25.
Volatility of Consumption (σC or HPC2) – Square of the cyclical component of the log of household final consumption expenditure (in current U.S. dollars) isolated using the HP filter for annual data with smoothness parameter λ = 6.25.
Volatility of Investment (σI or HPI2) – Square of the cyclical component of the log of gross capital formation (in current U.S. dollars) isolated using the HP filter for annual data with smoothness parameter λ = 6.25. Source: World DataBank
Volatility of Government Expenditure (σG or HPG2) – Square of the cyclical component of the log of general government final consumption expenditure (in current U.S. dollars) isolated using the HP filter for annual data with smoothness parameter λ = 6.25.
Volatility of Exports (σE or HPE2) – Square of the cyclical component of the log of exports of goods and services (in current U.S. dollars) isolated using the HP filter for annual data with smoothness parameter λ = 6.25.
Volatility of Imports (σM or HPM2) – Square of the cyclical component of the log of imports of goods and services (in current U.S. dollars) isolated using the HP filter for annual data with smoothness parameter λ = 6.25.
See the World Bank national accounts data and OECD National Accounts data files for details on the definition and construction of the variables.
Harmonized Indices of Consumer Prices (HICP) – The official measure of consumer price inflation in the euro area for the purposes of monetary policy and the assessment of inflation convergence as required under the Maastricht criteria for accession to the euro. Produced by Eurostat, using 2015 as base year.
Maastricht Criterion Bond Yields (r) - Interest rates for long-term government bonds denominated in national currencies. Produced by Eurostat, using data on central government bond yields on the secondary market, gross of tax, with a residual maturity of around 10 years.
Unemployment Rate (U) – Indicated by unemployment rate, 3 years average; long-term unemployment rate, % of active population aged 15-74 - 3 years change in p.p.; youth unemployment rate, % of active population aged 15-24 - 3 years change in p.p. Produced by Eurostat.
P a g e | 38 Table 7. Correlation matrix period 1 non-members
Y dY I dI G dG C dC X dX M dM U dU HICP r Y 1.00 dY -.36 1.00 I 1.00 -.36 1.00 dI -.30 .59 -.31 1.00 G .99 -.37 .99 -.30 1.00 dG -.20 .14 -.21 .76 -.20 1.00 C 1.00 -.35 1.00 -.28 .99 -.18 1.00 dC -.24 .27 -.25 .84 -.25 .95 -.22 1.00 X 0.99 -.36 .99 -.33 .99 -.24 .98 -.29 1.00 dX -.33 .53 -.34 .67 -.32 .58 -.32 .67 -.34 1.00 M 1.00 -.34 1.00 -.30 1.00 -.21 .99 -.26 1.00 -.31 1.00 dM -.31 .62 -.32 .79 -.31 .63 -.30 .72 -.33 .96 -.30 1.00 U -.39 .04 -.41 .04 -.40 .04 -.37 .08 -.42 .28 -.40 .20 1.00 dU .08 -.38 .08 -.16 .12 .06 .07 .07 .12 .08 .11 -.03 -.03 1.00 HICP -.37 .37 -.38 .50 -.38 .35 -.35 .44 -.41 .33 -.37 .35 .02 -.05 1.00 r -.18 -.25 -.19 -.08 -.21 .11 -.17 .15 -.22 -.05 -.20 -.11 .30 .18 .44 1.00
Table 8. Correlation matrix period 1 members
Y dY I dI G dG C dC X dX M dM U dU HICP r Y 1.00 dY -.43 1.00 I .99 -.41 1.00 dI -.18 .33 -.17 1.00 G .99 -.43 .98 -.18 1.00 dG -.14 .00 -.14 .81 -.13 1.00 C 1.00 -.42 .99 -.18 .99 .14 1.00 dC -.13 .11 -.13 .86 -.13 .95 -.13 1.00 X .94 -.43 .92 -.18 .93 -.13 .93 -.13 1.00 dX -.15 .31 -.15 .71 -.15 .71 -.14 .78 -.11 1.00 M .96 -.43 .95 -.18 .95 -.13 .95 -.13 .99 -.12 1.00 dM -.14 .37 -.14 .83 -.15 .72 -.14 .81 -.12 .91 -.12 1.00 U .19 .17 .19 .09 .17 -.01 .20 .07 .09 .15 .10 .15 1.00 dU .04 -.44 .02 .06 .05 .31 .04 .27 .06 .19 .05 .09 -.10 1.00 HICP -.25 .22 -.24 .25 -.26 .27 -.24 .29 -.27 .32 -.26 .30 .17 .01 1.00 r -.16 .05 -.18 -.21 -.17 -.27 -.15 -.25 -.24 -.27 -.25 -.30 .26 .01 .30 1.00
