FIRM-LEVEL BUSINESS CYCLE CORRELATION IN THE EU: SOME EVIDENCE FROM THE CZECH REPUBLIC AND SLOVAKIA Ladislava Issever Grochová 1, Petr Rozmahel 2 1 Mendelova univerzita v Brně, Provozně ekonomická fakulta, Zemědělská 1, 615 00 Brno Email:ladislava.grochova@mendelu.cz 2 Mendelova univerzita v Brně, Provozně ekonomická fakulta, Zemědělská 1, 615 00 Brno, Email:petr.rozmahel@mendelu.cz Abstract: Following recent trends in the Optimum Currency Area literature the paper attempts to provide some microeconomic foundations for explanation of aggregate economic activity co-movements in the European Union. In particular, the aim of the paper is to evaluate the correlation of firm-level economic activity fluctuations in the Czech Republic and Slovakia with the Euro area business cycle. The paper asks whether Slovakia, being a member of the Euro area, has more correlated firm-level business cycles with the Euro area than the Czech Republic. In addition, the correlations of individual NACE sections in both countries with the Euro area are examined in the paper. Evaluating the correlation density and the transformed density, using the Fisher z-transformation, the results indicate that Slovakia has more synchronized firm-level business cycles with that of the Euro area than the Czech Republic. Keywords: business cycle, correlation, Euro area, European integration, Fisher z-transformation, JEL classification: E32, F15, F45 1. Introduction The Endogenous Optimum Currency Area Theory suggests factors of business cycle fluctuations in the common monetary union. The more integrated area, the more similar cycles the member countries have. The literature has attempted to explain the co-movements mainly on national or regional level so far. Recently, there is an obvious effort to provide deeper microeconomic foundations for explaining aggregate fluctuations in the European Union and other integrating areas across the world in the literature. Gabaix (2011) produced an influencing and widely accepted study, in which he proposes a simple origin of aggregate shocks. The author points out that existing research has focused on aggregate shocks to explain aggregate movements in output, arguing that individual firm-level shocks average out in the aggregate. Following the studies by Simon (1955), Gabaix (1999) and Luttmer (2007) he empirically documents that this argument breaks down if the distribution of firm sizes is fat-tailed. Providing micro-foundation for aggregate shocks Gabaix concludes that a large part of aggregate fluctuations arises from idiosyncratic shocks to individual firms. This finding is in line with Acemoglu et al. (2012) who argue that idiosyncratic shocks to firms can manifest themselves in aggregate business cycle fluctuations when the linkages propagate microeconomic shocks across firms leading to positive endogenous co-movement. Giovanni, Levchenko and Mejean (2014) ask the question whether firm-level dynamics have an impact on aggregate fluctuations. In particular the authors extract the macroeconomic, sectoral, and firm-specific components of a firm s sales to a given destinations when explaining volatility of aggregate sales. Analysing a large sample of French firms data for the period 1990-2007 they find the firm-specific component accounts for an important part of the fluctuations of the aggregate sales growth. The authors interpret this as evidence for the relevance of firm-level shocks for aggregate fluctuations. Also Fidrmuc and Scharler (2014) examine the relationship between firm-specific shocks and aggregate economic fluctuations. Using a sample of 25,000 German firms and banks in 2005 and -142-
2007, they find that the movements in output e.g. business cycles are synchronized across sectors while specific shocks are uncorrelated. In our paper, we try to contribute to general effort, which is apparent in recent literature, for providing some microeconomic foundations for a rationale of aggregate economic activity co-movements in the European Union. In particular, the aim of the paper is to evaluate the correlation of firm-level economic activity fluctuations in the Czech Republic and Slovakia with the Euro area business cycle. The paper asks whether Slovakia, being a member of the Euro area, has more correlated firm-level business cycles with the Euro area than the Czech Republic. In addition, the correlations of individual NACE sections in both countries with the Euro area are examined in the paper. As this paper presents a part of larger research, we conduct in this area, it focuses mainly on descriptive statistics. 2. Empirical strategy In order to assess firm-level business cycle similarities with the Euro area GDP we compute correlations of nominal revenues of the Czech and Slovak firms with nominal GDP of Euro area. The time-series that are examined are of annual frequency and comprise the time span between 2003 and 2011, which in fact represents 9-year window out of which the correlation coefficients are calculated. The correlations are studied both for all firms within each state together and divided into sections relative to the aggregated NACE classification (see Tab. 1). The sample consists of 105,221 firms in the Czech Republic and 30,006 firms in Slovakia. Generalizing the sample to the whole population, we use Fisher z-transformation (Fisher, 1915; see eq. 1) to maintain the variance of correlation relatively constant for all values of the population correlation coefficient (hereafter transformed correlation). In other words we remove the constraint that the correlation coefficient has to lie between -1 and 1. As a result we obtain approximately normally distributed values reducing the aspect of diminishing variance as correlation gets closer to 1. Specifically, the transformation that we apply is given by: 1 1 ) z ln (1) 2 1 where ρ is the (untransformed) correlation coefficient and z is the transformed correlation coefficient. The firm-level data (nominal revenues) is obtained from Amadeus database, while the macroeconomic variable (nominal GDP) is downloaded from Eurostat. 3. Firm-level business cycle correlation 3.1 NACE aggregated sections The official European industry standard classification system NACE is used in the paper. In its complete form it contains sections in a range of A U consisting of 6 digit code. In our paper we aggregated the NACE section into total of 10 sections. Such aggregation of 10 or 11 sections (when setting aside Manufacturing industry) is commonly used for instance in analysis provided by the Eurostat. The final set of aggregated NACE sections we use in the paper is listed in the Table 1. The codes 1 to 10 instead of letters or abbreviations are used for each aggregated section in further text. -143-
Tab. 1. Aggregated NACE sections Code of aggregated NACE codes Aggregated sections sections 1 A Agriculture, forestry and fishing 2 B-E, C Industry (except construction), Manufacturing 3 F Construction 4 G-I Wholesale and retail trade, transport, accommodation and food service activities 5 J Information and communication 6 K Financial and insurance activities 7 L Real estate activities 8 M-N Professional, scientific and technical activities; administrative and support service activities 9 O-Q Public administration, defence, education, human health and social work activities 10 R-U Arts, entertainment and recreation; other service activities; activities of household and extra-territorial organizations and bodies Source: Eurostat The Gross Value Added GVA percentage shares of aggregated NACE sections in case of the Czech Republic and Slovakia are presented in Table 2. The values of GVA shares indicate similar industrial structures of both economies. Regarding slightly different natural conditions Slovakia has a bit larger portion of agriculture sector comparing to the Czech Republic. Still, both countries are focused mainly on service and industry sectors. Industry and manufacturing make 31% in total GVA in the Czech Republic, which is 5 percentage points larger share than in case of Slovakia. Other aggregated NACE groups make similar GVA shares, in both countries. The relative differences are in general less than 2.5 percentage points. NACE aggregated section Tab. 2. GVA percentage shares of the aggregated NACE sections (2011) 1 2 3 4 5 6 7 8 9 10 Czech Republic 2.40% 30.90% 6.20% 18.20% 5.10% 4.70% 8.80% 6.60% 14.90% 2.30% Slovakia 3.40% 26.80% 8.80% 21.60% 4.50% 3.70% 6.90% 7.20% 13.80% 3.30% Source: Eurostat Accordingly, the GVA shares of the aggregated NACE sections provide evidence of similar industrial structures of both countries, which is given by historical context of more than 70 years common history. Having been almost purely rural country, Slovakia experienced significant industrial development since establishment of Czechoslovakia in 1918. Considering the Euro adoption in 2009, Slovakia seems to be even faster developing country in terms of its European integration dynamics. The following sub-chapter provides some evidence of the level of association of the Czech and Slovak firms with the Euro area business cycle. 3.2 Firm-level correlations density The level of synchronization can be deduced from Fig. 1, in which densities of firm-level business cycle correlations are plotted. In both countries a similar pattern of business cycle synchronicity is -144-
clearly visible. The majority of firms revenues evolve in line with the Euro area business cycle having positive elevated correlation coefficient. In particular, the last decile includes correlations greater than 0.93 in the Czech Republic and even only correlations equal to 1 in Slovakia, while the first decile comprises a larger span of correlation coefficient values below -0.75 in the Czech Republic and -0.93 in Slovakia. The skewed pattern is more visible for Slovakian data where skewness reaches the value of -0.62 compared to -0.54 in the Czech Republic. Fig. 1. Density of the Czech (a) and Slovak (b) firm-level business cycle correlations with the Euro area cycle (NACE total) As Camacho et al. (2006) state, the correlation across units is not the average of the correlations between each pair of units, we need to transform the data with the Fisher z-transformations to obtain a statistics with a known distribution for the correlation. The whole population could be still characterized with a higher level of synchronization in business cycles. Accordingly, Slovakia, being a member of Euro area, has more synchronized firm-level business cycles with that of the Euro area. Fig. 2. Density of the Czech (a) and Slovak (b) firm-level business cycle correlations transformed with the Euro area cycle (NACE total) -145-
More specifically, we focus on the aggregated NACE sections to study the effect of the structure of economy on business cycle synchronization, which is shown in Tab. 2 and 3 and graphical results are in the Attachment in Fig. 3 and 4. In both countries analysed the sector which is the most pro-cyclical in terms of average correlation, is those of Agriculture, forestry and fishing, Industry (except construction) and manufacturing, and Public administration, defence, education, human health and social work activities. This is caused by extremely high number of synchronized firm-level business cycles. 50% of firms have the correlation coefficient at least 0.7 in the sector of Wholesale and retail trade, transport, accommodation and food service activities in the Czech Republic, Slovaks add also the sectors of Industry (except construction) and manufacturing, and Information and communication. Extremely high number of synchronized firm-level business cycles can be found in the sector of Agriculture, forestry and fishing. In contrast, the relatively greatest extent of firms with least synchronized sectors are Financial and insurance activities, and Real estate activities. Tab. 2. Summary statistics of the Czech firm-level business cycle correlations with the Euro area cycle NACE section Obs Mean Std. Dev. Median Skewness Kurtosis Total 105221 0.2397 0.6156 0.3917-0.5429 1.9834 1 2946 0.4398 0.5559 0.6709-1.0963 3.0310 2 16455 0.3374 0.5860 0.5402-0.7752 2.3526 3 8903 0.2465 0.5907 0.5169-0.5890 2.1129 4 34901 0.2198 0.6331 0.7148-0.4757 1.8614 5 3441 0.2425 0.6294 0.6495-0.5444 1.9472 6 286 0.0494 0.6432 0.3970-0.1197 1.5951 7 19047 0.1769 0.6009 0.3624-0.4451 1.9710 8 15462 0.2093 0.6237 0.5242-0.4780 1.8857 9 1908 0.3624 0.5871 0.2979-0.7741 2.3136 10 1872 0.1937 0.6403 0.4104-0.4205 1.7865 In Slovakia, in general, there are relatively more firms with pro-cyclical and less firms with counter-cyclical behaviour. Specifically, almost a half of all firms have the correlation coefficient greater than 0.5. In this perspective, a large amount of cycles which are least synchronized are the cycles of firms with activities in Real estate. A similar pattern is evident in the sector of Public administration, defence, education, human health and social work activities. Nevertheless, right in this sector there is also a considerable amount of firms that are almost perfectly correlated with the Euro area GDP. This shifts this sector among those with the highest mean correlation. Tab. 3. Summary statistics of the Slovak firm-level business cycle correlations with the Euro area cycle NACE section Obs Mean Std. Dev. Median Skewness Kurtosis Total 30006 0.2631 0.6933 0.4893-0.6208 1.9609 1 979 0.3261 0.6081 0.5329-0.8015 2.3670 2 4533 0.3552 0.6320 0.7012-0.8740 2.4503 3 2376 0.2346 0.6638 0.5705-0.6084 2.0066 4 11669 0.2595 0.6977 0.8056-0.5992 1.9231 5 1212 0.2830 0.7231 0.7005-0.6319 1.8835-146-
6 76 0.0497 0.7860 0.4447-0.1535 1.4248 7 1908 0.1965 0.7266 0.4783-0.4442 1.7041 8 5409 0.2115 0.7545 0.5414-0.4927 1.6893 9 1381 0.2922 0.5783 0.3895-0.5915 2.3807 10 463 0.2432 0.7001 0.5793-0.5515 1.8657 4. Conclusions The similarity of firm-level business cycles of Czech and Slovak firms with the Euro area GDP cycle was examined in the paper. The GVA shares of the aggregated NACE sections provide evidence of similar industrial structures of both countries, which is given by historical context of more than 70 years common history. The results show that Slovakia, being a member of the Euro area, has more synchronized firm-level business cycles with that of the Euro area. In Slovakia, the largest NACE aggregate section covering Industry and Manufacturing denotes the highest average correlation with the Euro area GDP cycle. Similarly in the Czech Republic this section reveal similar patterns. The majority of firms in respective industries in Slovakia have the correlation higher than 0.7, when examining the median values. This comprises three sections of Industry and manufacturing (2), Wholesale, retail trade, transportation and other services (4) and Information and Communication (5). These sections, in fact, cover more than 58 % of all firms in the sample. More than half of firms have higher correlation than 0.7 only in one section of Wholesale, retail trade, transportation and other services (4) in the Czech Republic. This industry represents roughly one third of the Czech firms sample. This research has received funding from the European Union s Seventh Framework Programme for research, technological development and demonstration under grant agreement No. 290647. References [1] Acemoglu, D., Carvalho, V.M., Ozdaglar, A. and A. Tahbaz-Salehi. 2012. The Network Origins of Aggregate Fluctuations. Econometrica, vol. 80, issue 5, pp. 1977 2016. [2] Camacho, M., Perez-Quirosb, G. and L. Saizc. 2006. Are European business cycles close enough to be just one? Journal of Economic Dynamics and Control, vol. 30, issues 9 10, pp. 1687 1706. [3] Fidrmuc, J. and J. Scharler. 2014. What Determines Borrowing Costs at the Firm-Level: Firm-Specific and Aggregate Information. Annual Conference 2014 (Hamburg): Evidence-based Economic Policy 100322, Verein für Socialpolitik / German Economic Association. [4] Fisher, R.A. 1915. Frequency distribution of the values of the correlation coefficient in samples of an indefinitely large population. Biometrika (Biometrika Trust), vol. 10, issue 4, pp. 507 521. [5] Gabaix, X. 2011. The Granular Origins of Aggregate Fluctuations. Econometrica, vol. 79, issue 3, pp. 733 772. [6] Gabaix, X. 1999. Zipf s Law for Cities: An Explanation. Quarterly Journal of Economics, vol. 114, pp. 739 767. [7] di Giovanni, J., Levchenko, A.A. and I. Mejean, 2014. Firms, Destinations, and Aggregate Fluctuations. Econometrica, Econometric Society, vol. 82, issue 4, pp. 1303 1340. -147-
[8] Luttmer, E.G.J. 2007. Selection, Growth, and the Size Distribution of Firms. Quarterly Journal of Economics, vol. 122, pp. 1103 1144. [9] Simon, H. 1955. On a Class of Skew Distribution Functions. Biometrika, vol. 42, pp. 425 440. Attachments Fig. 3. Densities of the Czech firm-level business cycle correlations with the Euro area cycle (NACE aggregated sections) -148-
Fig. 4. Densities of the Slovak firm-level business cycle correlations with the Euro area cycle (NACE aggregated sections) -149-
Fig. 5. Density of the Czech firm-level business cycle correlations transformed with the Euro area cycle (NACE aggregated sections) -150-
Fig. 6. Density of the Slovak firm-level business cycle correlations transformed with the Euro area cycle (NACE aggregated sections) -151-
-152-