Common Baltic-Nordic business cycles: Correlation- versus Markov-switching approaches

Ribaudo, Giorgio Article Common Baltic-Nordic business cycles: Correlationversus Markov-switching approaches Contemporary Economics Provided in Cooperation with: VIZJA University, Warsaw Suggested Citation: Ribaudo, Giorgio (2019) : Common Baltic-Nordic business cycles: Correlationversus Markov-switching approaches, Contemporary Economics, ISSN 2300-8814, University of Economics and Human Sciences in Warsaw, Warsaw, Vol. 13, Iss. 4, pp. 427-445, https://doi.org/10.5709/ce.1897-9254.324 This Version is available at: https://hdl.handle.net/10419/297495 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ www.ce.vizja.pl 427 This work is licensed under a Creative Commons Attribution 4.0 International License. With strong historical ties, and economic linkages that have continued to grow after the fall of the Soviet Union, the Baltic and Nordic regions form a unique economic space. How interconnected are these regions, both to each other and to the rest of the world? Greater connections can help forecast future economic linkages—and also help assess the strength of the Euro as a common currency. This study applies two methods of business-cycle analysis (cross-correlations and Markovswitching approaches) to seven countries in these regions. Both methods find evidence of a single Baltic common cycle for both output and consumption, while a Nordic cycle exists only for output, and there is no single common Baltic-Nordic cycle. Tests of correlation and concordance show there to be relatively strong connections with Germany, the U.S., and Russia—with Nordic-Baltic linkages also quire strong—but that the specific results vary by the method used 1. Introduction As Northern European countries that have been members of the European Union for more than a decade, the Baltic nations of Estonia, Latvia, and Lithuania have enjoyed growing trade with their Western neighbors and increasing global economic integration. At the same time, they maintain trade and financial linkages with Russia. Besides the major economic powers to the east and in the EU core, however, the Baltic republics have long enjoyed strong historical, cultural, and economic ties with the “Nordic” countries along the Baltic Sea. This common history goes back at least as far as the maritime Hanseatic League the expanded to the Baltic territories in the 13th century, followed by incorporation of parts into the Swedish empire in the 16th century. In the modern, post-Soviet era, foreign investment has flowed inward, in particular from Sweden (and Finland, in Estonia’s case). It is possible that these economic ties, in particular, may be in some ways stronger than linkages with Germany, Russia, or even the global economy. This may have implications for the stability Eurozone. As Mundell (1961) pointed out, common currency areas must exhibit similar economic behavior for Common Baltic-Nordic business cycles: Correlationversus Markov-switching approaches ABSTRACT F44 KEY WORDS: JEL Classification: Baltic region, Nordic region, business cycles, cross-correlations, Markov-switching Northeastern Illinois University, United States of America Correspondence concerning this article should be addressed to: Scott W. Hegerty, Department of Economics, Northeastern Illinois University, 5500 N St Louis Ave, Chicago, IL 60625. E-mail: s-heger[email protected] Scott W. Hegerty Primary submission: 30.05.2017 | Final acceptance: 23.07.2018 428 Scott W. Hegerty 10.5709/ce.1897-9254.324DOI: CONTEMPORARY ECONOMICS Vol. 13 Issue 4 427-4452019 a single monetary policy to be effective. Strong linkages between Euro and non-Euro countries might make such a policy unfeasible. It is this proposition that this paper seeks to test. Applying two statistical methods to isolate the business cycles of the countries of the region and their major neighbors, we then look for common cycles in the Baltics, the Nordic region, and the combined group. We then examine the degree of cyclical co-movements between individual countries and their respective regions, and each region each other and with major partners. This allows us to compare which linkages are the strongest. Overall, we find that the Baltics enjoy common output and consumption cycles, while the four Nordic countries share only a common output cycle, and there is no single joint “Baltic-Nordic” output or consumption cycle. While interconnections are often strongest with Germany or the United States (as a proxy for global factors), specific findings differ depending on the test that is used. A number of studies in the literature examine business-cycle comovements and their underlying determinants. Many of these were conducted before the onset of the 2008 global financial crisis. Some, like Imbs (2004), focus on the determinants of integration, such as increasing financial linkages among countries. Others conduct empirical tests for specific regions. While Western Europe receives a large share of attention, those studies that examine CEE countries sometimes exclude the Baltics. Artis, Fidrmuc and Scharler (2008), for example, include only Estonia alongside five Central European countries over the period from 1995-2004, calculating contemporaneous correlations among business cycles. After the 2004 accession of 10 CEE countries to the European Union, the number of studies examining these countries’ degree of interconnection to the “core” EU members increased. Darvas and Szapáry (2008), for example, use a dynamic factor model to capture a “common factor” in the region’s business cycles and calculate correlations among business-cycle pairs. This study finds little synchronization between the Baltic group and Western Europe. Fadejeva and Melihovs (2008), on the other hand, calculate a common factor for Baltic and European growth rates. Hegerty (2010), mentioned below, finds strong evidence of a Baltic economic region, using data that end in early 2008. Other literature goes into further detail. (See, for example, Babetskii, 2005; Benczúr & Rátfai, 2005; Fidrmuc & Korhonen, 2006; Frankel & Rose, 1999; Hakura, 2009; Horvath & Ratfai, 2004; Inagaki, 2006; Furceri & Karras, 2008. For more on international spillovers, see Bayoumi & Swiston, 2009; Buch, Doepke, & Pierdzioch, 2005; Rafiq, 2011). While these previous studies often arrive at mixed results, we expect that additional years’ worth of data, spanning the preand post-crisis periods, will allow us to approach the issue with additional clarity. This study makes use of two methods of determining business cycles using time series data. The first, which filters the data using a variety of methods (the best-known of which is the Hodrick-Prescott filter), was introduced by Backus, Kehoe and Kydland (1994). Hegerty (2010) applies this method using quarterly GDP data and concludes that the three Baltic countries form an integrated region within which connections are stronger than they are with partner countries. He does not, however, include any Nordic countries in his analysis. In fact, there are few specific analyses of the Nordic region using this type of method. The use of the Hodrick-Prescott filter has been criticized for its suboptimal statistical properties (See Hamilton, 2017, for example, who notes that this filter can generate “spurious” dynamic relations that are unrelated to the underlying time series). This has led to the increasing use of alternative filtering methods, such as those of Baxter and King (1999) and of Christiano and Fitzgerald (1999). In addition, Markovswitching models, which incorporate a change in state between a high-growth and a low-growth regime, have been growing in popularity. Introduced by Hamilton (1989), this method has been applied to the CEE region to some extent. The Baltic and Nordic regions, however, are often omitted. Jiménez-Rodríguez, Morales-Zumaquero and Égert (2013) examine five CEE and eight Western European countries from 1995 to 2011, finding a high degree of synchronization in this subregion. Di Giorgio (2016) examines seven CEE countries from 1993 to 2014, including Latvia and Lithuania, but not Estonia, because the country joined the Euro during the study period. The study finds that Central Europe demonstrates the highest degree of business-cycle synchronization, particularly in the www.ce.vizja.pl 429 Common Baltic-Nordic Business Cycles: CorrelationVersus Markov-Switching Approaches This work is licensed under a Creative Commons Attribution 4.0 International License. cases of Hungary and Poland, and more often during recessionary periods. While Aastveit, Jore and Ravazzolo (2016) focus entirely on the case of Finland, the rest of the Nordic region does not receive similar attention in the literature. The purpose of this study, then, is to focus entirely on this under-analyzed part of the world, both for individual countries and as common regions. We find diverging patterns that show each country to exhibit its own unique characteristics, while at the same time uncovering common cycles that are interlinked both across Europe and with the world. This paper proceeds as follows: Section II explains the econometric methodology. Section III presents the results. Section IV concludes. 2. Methodology For this study, we use quarterly data from the International Financial Statistics (IFS) for four Nordic and three Baltic countries, as well as the “partner” countries of the United States (a proxy for the global economy), Germany, and Russia. The data span from 1995q1 to 2014q4. Real GDP is calculated from nominal using the GDP deflator, and real consumption using the Consumer Price Index. Domestic currencies are converted to Euros where necessary, and in one case (Latvian consumption) the Euro transition created an “outlier” data point that was smoothed by taking the average of the two surrounding values. The natural logarithms of these series are then seasonally adjusted using the Census-X12 method. In all, we have 20 time series. Using these series, we first select an appropriate filter to remove the cycles from the trend series. In this “traditional” approach, we choose from three: those of Hodrick and Prescott (1997), Baxter and King (1999), and Christiano and Fitzgerald (1999), which will be referred to as HP, BK, and CF, respectively. We extract each type of filtered cycle from our output and consumption series, and for each, we compare their timeseries properties. Since, for the most part, the members of each trio are highly correlated with one another, we choose the Christiano-Fitzgerald method because recent studies in the literature prefer this method over the Hodrick-Prescott filter, and because it retains more observations at the beginning and the end of the resulting series than does the Baxter-King filter. With one output and one consumption cycle for each of our ten countries, we next test for and extract their common component(s) for the Baltic and Nordic regional groups. We also do this for the combined Nordic-Baltic region. We generate combined regional series using Principal Components Analysis (PCA), which extracts the component(s) of the cycles that contribute the majority of their common variance. Following this well-known procedure, we create principal components in those cases where the resulting eigenvalue is greater than one, labeling each a regional output or consumption cycle. We then calculate cross-correlation functions (CCFs) for various pairs of individual countries and regions. Because 10 countries, plus numerous regional cycles, allows for an impossibly large number of pairs, we take only a cursory look at connections between individual countries and their respective region. The majority of our analysis is between regions and other regions or with each of the three partner economies. Following Hegerty (2010), the CCF is calculated as follows, with a maximum of –k = 4 lags and k = 4 leads: (1) Here, contemporaneous correlation occurs where k = 0. If k > 0, X in a previous period is correlated with Y in the current period, or X “leads” Y. Likewise, if k < 0, X in the current period is correlated with Y in a previous period, so X “lags” Y. These can be tabulated over the range k = [-4, 4] for multiple pairs, to see which contemporaneous correlations are strongest, and for which country pair. This will allow us to test the strength of Baltic-Nordic linkages versus those with other partner countries. Our second method makes use of the Markov-Switching (MSM) approach of Hamilton (1989). Here, an economy can “switch” between two states: growth and contraction. The probability of such a change is calculated as: (2) We can estimate these probabilities based on an underlying AR(1) function of log changes in output or consumption, and then graph these probabilities as a measure of each cycle. We then test for linkages by calculating concordances, or the fraction of the time peri- ( )( ) ( ) ( ) ∑∑ ∑ −− −− = + + +2 2YYXX YYXX ktt ktt kt ρ 430 Scott W. Hegerty 10.5709/ce.1897-9254.324DOI: CONTEMPORARY ECONOMICS Vol. 13 Issue 4 427-4452019 Baltic Y Eigenvalues Loadings PC1 Baltic C Eigenvalues Loadings PC1 PC1 1.279 EE 0.572 PC1 1.296 EE 0.578 PC2 0.681 LT 0.570 PC2 0.584 LT 0.572 PC3 0.569 LV 0.590 PC3 0.500 LV 0.582 Nordic Y Eigenvalues Loadings PC1 Nordic C Eigenvalues Loadings PC1 PC2 PC1 1.296 DK 0.514 PC1 1.206 DK 0.589 -0.150 PC2 0.903 FI 0.549 PC2 1.051 FI -0.518 0.431 PC3 0.751 NO 0.409 PC3 0.799 NO 0.597 0.284 PC4 0.665 SW 0.517 PC4 0.714 SW 0.168 0.844 NordicBaltic Y Eigenvalues Loadings PC1 PC2 NordicBaltic C Eigenvalues Loadings PC1 PC2 PC1 1.443 EE 0.422 -0.262 PC1 1.432 EE 0.464 -0.104 PC2 1.075 LT 0.384 -0.416 PC2 1.066 LT 0.439 -0.149 PC3 0.876 LV 0.394 -0.436 PC3 0.959 LV 0.446 -0.073 PC4 0.750 DK 0.318 0.545 PC4 0.753 DK 0.418 0.037 PC5 0.678 FI 0.410 0.303 PC5 0.657 FI -0.292 0.129 PC6 0.591 NO 0.347 -0.023 PC6 0.562 NO 0.364 0.462 PC7 0.554 SW 0.360 0.424 PC7 0.466 SW 0.000 0.854 Table 1. Principal component analysis, filtered cycles EEY LTC LTY LVC LVY BK 0.962 0.967 0.964 0.969 0.960 0.967 HP 0.971 0.971 0.964 0.969 0.963 0.972 DKC DKY FIC FIY NOC NOY BK 0.961 0.956 0.822 0.942 0.918 0.952 HP 0.958 0.949 0.898 0.958 0.936 0.947 DEC DEY RUC RUY USC USY BK 0.967 0.978 0.934 0.947 0.947 0.966 HP 0.937 0.977 0.941 0.951 0.925 0.943 Table 2. Correlations between Christiano-Fitzgerald filtered cycles and alternatives www.ce.vizja.pl 431 Common Baltic-Nordic Business Cycles: CorrelationVersus Markov-Switching Approaches This work is licensed under a Creative Commons Attribution 4.0 International License. ods in which the members of each pair are simultaneously in the same state. Higher fractions mean stronger linkages. These concordances can be written as: (3) Applying these two methods, we can then test our main hypothesis. While both sets of tests show the presence of some Baltic and Nordic regional cycles, they differ when assessing the relative strength of integration among country pairs. Our results are provided below. 3. Results In the interest of space, Figure 1 depicts the CF-, HP-, and BK-filtered cycles for selected output and consumption series. All three appear to behave quite similarly, which is confirmed by the contemporaneous correlations in Table 1. Taking into account the abovementioned criticisms of the HP filter, as well as the fact that the BK filter truncates the first and last few observations, we proceed with the CF filter for the rest of this study. All Baltic and Nordic series are plotted individually in Figure 2. We see that Estonia, Latvia, and Lithuania share a common “boom” and “bust” before and after the 2008 financial crisis, but that the Nordic countries—particularly Finnish consumption and Norwegian output—appear to follow different patterns. We suspect that our visual clues will be reflected in our formal analysis. Table 2 provides the results of our Principal Components Analysis. Based on the eigenvalues, we see that there are single Baltic output and consumption cycles, as well as only one Nordic output cycle, but there are two Nordic consumption cycles. The three Baltic countries load equally on their first principal component (labeled PC1 in the PCA summary and BALY and BALC elsewhere), but Norway loads relatively less on the Nordic output principal component NORY. The Nordic PC1 for consumption (NORC1) has a rather small factor loading for Sweden and a negative loading for Finland. The Nordic consumption PC2 cycle (NORC2) has a small loading for Norway and a negative loading for Denmark. We therefore surmise that there are in fact two “Nordic” pairs rather than a single region. Following the same approach to extract common Nordic-Baltic cycles, we isolate two principal components for output and consumption. While PC1 (NBY1) Figure 1. Selected filtered business cycles.. 432 Scott W. Hegerty 10.5709/ce.1897-9254.324DOI: CONTEMPORARY ECONOMICS Vol. 13 Issue 4 427-4452019 Figure 2. Filtered country output and consumption cycles Figure 3. Multiple PCA consumption cycles. www.ce.vizja.pl 433 Common Baltic-Nordic Business Cycles: CorrelationVersus Markov-Switching Approaches This work is licensed under a Creative Commons Attribution 4.0 International License. does not present any issues for output, PC2 (NBY2) has the Baltics’ factor loadings opposite from Denmark, Sweden, and Finland; Norway’s factor loading is rather small. For the first consumption principal component (NBC1), Finland’s loading is negative and Sweden does not load at all. For NBC2, the Baltics’ factor loadings are negative and small. Sweden loads heavily on this component, as does Finland to a lesser extent. The cycles that include the Nordic countries, which in three of the four cases have two principal components, are depicted in in Figure 3. In these cases, they are clearly out of sync with one another. We can conclude that not only is there no single “Nordic” consumption cycle or any “Nordic-Baltic” cycle in any sense, there are instead three axes: the Baltics, Denmark-Norway, and Sweden-Finland. These findings help drive our regional analysis. Our final cycles for the Baltic, Nordic, and Baltic-Nordic combined series are presented in Figure 4. “Partner” business cycles are depicted in Figure 5. We see that Germany, Russia, and the United States often follow similar patterns, but that certain periods (such as the 1998 Russian default) lead to distinctive individual patterns. We expect there to be differences in the interregional connections between the Baltics, Nordic countries, and the rest of the world. Figure 4. Regional output and consumption cycles. EEY - BALY LTY - BALY LVY - BALY EEC - BALC LTC - BALC LVC - BALC 0.937 0.933 0.965 0.971 0.961 0.978 DKY - NORY FIY - NORY NOY - NORY SWY - NORY DKC - NORC1 DKC - NORC2 0.864 0.922 0.688 0.868 0.856 -0.166 FIC - NORC1 FIC - NORC2 NOC - NORC1 NOC - NORC2 SWC - NORC1 SWC - NORC2 -0.753 0.475 0.869 0.313 0.245 0.931 Table 3. Correlations between country cycles and regional cycles 434 Scott W. Hegerty 10.5709/ce.1897-9254.324DOI: CONTEMPORARY ECONOMICS Vol. 13 Issue 4 427-4452019 BALY-DEY BALY-RUY BALY-USY BALYBALC BALY -NORY BALYNBY1 BALYNBY2 BALYNBC1 BALYNBC2 -4 0.120 0.228 0.277 0.682 0.272 0.478 -0.37 0.676 -0.159 -3 0.221 0.322 0.379 0.801 0.378 0.621 -0.396 0.775 -0.194 -2 0.325 0.423 0.478 0.890 0.471 0.747 -0.425 0.860 -0.193 -1 0.414 0.510 0.566 0.947 0.532 0.832 -0.438 0.916 -0.170 00.454 0.560 0.619 0.951 0.565 0.881 -0.454 0.928 -0.116 10.422 0.566 0.619 0.895 0.515 0.824 -0.437 0.876 -0.092 20.338 0.521 0.575 0.796 0.416 0.717 -0.432 0.787 -0.068 30.208 0.443 0.497 0.663 0.285 0.570 -0.426 0.657 -0.065 40.066 0.340 0.408 0.501 0.151 0.410 -0.409 0.502 -0.063 BALC-DEC BALC-RUC BALC-USC BALCNORC1 BALCNORC2 BALCNBY1 BALCNBY2 BALCNBC1 BALCNBC2 -4 -0.395 0.417 0.101 0.461 -0.262 0.392 -0.383 0.608 -0.193 -3 -0.306 0.540 0.226 0.544 -0.268 0.544 -0.428 0.736 -0.201 -2 -0.182 0.639 0.350 0.600 -0.249 0.674 -0.452 0.837 -0.191 -1 -0.072 0.696 0.466 0.648 -0.220 0.769 -0.467 0.912 -0.166 00.027 0.724 0.546 0.676 -0.176 0.817 -0.479 0.951 -0.127 10.099 0.701 0.580 0.663 -0.153 0.803 -0.483 0.916 -0.099 20.172 0.650 0.579 0.613 -0.133 0.728 -0.481 0.841 -0.081 30.229 0.591 0.548 0.528 -0.140 0.612 -0.490 0.728 -0.093 40.245 0.516 0.493 0.418 -0.146 0.481 -0.480 0.589 -0.109 NORY-DEY NORY-RUY NORY-USY NORYNORC1 NORY – NORC2 NORYNBY1 NORYNBY2 NORYNBC1 NORYNBC2 -4 0.399 0.142 0.266 0.020 0.246 0.280 0.222 0.112 0.169 -3 0.546 0.161 0.420 0.138 0.255 0.473 0.310 0.235 0.203 -2 0.688 0.193 0.565 0.322 0.312 0.640 0.354 0.388 0.307 -1 0.794 0.227 0.711 0.457 0.345 0.767 0.399 0.492 0.381 00.811 0.211 0.806 0.608 0.438 0.888 0.451 0.569 0.530 10.699 0.156 0.802 0.639 0.433 0.780 0.359 0.562 0.553 20.524 0.034 0.711 0.639 0.386 0.674 0.282 0.517 0.536 30.306 -0.111 0.600 0.612 0.326 0.529 0.194 0.444 0.502 40.084 -0.249 0.485 0.528 0.189 0.351 0.086 0.344 0.377 Table 4. Cross-correlation functions between regional cycles and partners www.ce.vizja.pl 441 Common Baltic-Nordic Business Cycles: CorrelationVersus Markov-Switching Approaches This work is licensed under a Creative Commons Attribution 4.0 International License. BY1 BC1 NY1 NC1 NC2 EEC 0.564 0.333 0.564 0.551 0.449 EEY 0.872 0.077 0.897 0.269 0.705 LTC 0.423 0.731 0.372 0.487 0.462 LTY 0.154 0.795 0.179 0.654 0.244 LVC 0.641 0.205 0.692 0.500 0.526 LVY 0.795 0.179 0.846 0.397 0.577 DKC 0.731 0.269 0.782 0.333 0.744 DKY 0.141 0.859 0.115 0.692 0.256 FIC 0.603 0.551 0.526 0.487 0.385 FIY 0.269 0.731 0.218 0.615 0.308 NOC 0.564 0.436 0.538 0.449 0.449 NOY 0.244 0.679 0.346 0.641 0.385 SWC 0.359 0.615 0.359 0.577 0.449 SWY 0.885 0.141 0.910 0.308 0.718 Table 8. Local concordances between Markov-switching expansion phases. Bold = larger concordances BY1 BC1 NY1 NC1 NC2 BY1 10.205 0.821 0.295 0.654 BC1 10.179 0.654 0.321 NY1 10.346 0.654 NC1 10.359 NC2 1 NBY1 NBY2 NBC1 NBC2 NBC3 NBY1 10.654 0.321 0.603 0.436 NBY2 10.205 0.641 0.218 NBC1 10.308 0.731 NBC2 1 0.295 Table 9. Concordances between Markov-switching regional expansion phases. Bold = larger concordances 442 Scott W. Hegerty 10.5709/ce.1897-9254.324DOI: CONTEMPORARY ECONOMICS Vol. 13 Issue 4 427-4452019 ure 6. The 2008 recession is clearly depicted in the plots for Estonian and Latvian output; Latvian consumption declines both in 2008 and 2014. Lithuania, however, is more likely to exhibit growth “spikes” throughout the study period. The Nordic phase probabilities are depicted in Figure 7. Danish output shows a “boom” prior to 2010, with Swedish output and consumption following more of an oscillating pattern. Again, Finland follows its own unique pattern, with consumption reaching its lowest point after 2010. Table 7 presents the percentages of quarters during which each cycle is predicted to be in an expansion phase. Estonian output is shown to be high (91.0%), with Lithuanian output’s value very low (6.4%). This is likely due to the properties of the model itself rather than because of any failure of the Lithuanian economy. This gives reason to be suspicious of over-relying on one specific modeling method. Likewise, Danish consumption is in an expansion phase during 82.1% of the quarters, while output is in this phase only 7.7% of the time. While we treat these results with caution, we are able to arrive at some interesting conclusions, particularly involving the common Nordic-Baltic series. The second principal components of both output and consumption growth rates, in which Norway, Denmark, and Sweden appear to form a common grouping, are in expansion a larger percentage of the time than is the case for the other principal components. The regional cycles, created using our Markovswitching models, are depicted in Figure 8. There are clear “booms” and “busts” surrounding the 2008 crisis; this is particularly clear for the case of Baltic consumpBY1 BC1 NY1 NC1 NC2 DEY 0.769 0.205 0.821 0.679 0.218 RUY 0.718 0.359 0.692 0.577 0.423 USY 0.731 0.244 0.782 0.667 0.359 DEC 0.269 0.782 0.269 0.333 0.692 RUC 0.769 0.308 0.718 0.500 0.397 USC 0.436 0.487 0.538 0.654 0.526 Table 10. Concordances between Markov-switching regional expansion phases. Bold = larger concordances CF-Filtered Cycles Markov-Switching Approach Output Consumption Output Consump-tion Baltic USY, BALC, NBY1 RUC, NBY1, NBC1 DEY, RUC, NY1 DEC Nordic (1) DEY, USY, NBY1 BALC, NBY1, NBC1 DEY, USY, NC1 DEY, USY Nordic (2) USC, NBY1, NBC1 DEC Nordic-Baltic (1) USY, NBC1 RUC NBY1 NBC3 Nordic-Baltic (2) DEY USC NBY2 NBC2 Table 11. Summary of especially high correlations. www.ce.vizja.pl 443 Common Baltic-Nordic Business Cycles: CorrelationVersus Markov-Switching Approaches This work is licensed under a Creative Commons Attribution 4.0 International License. tion. Baltic output also shows strong fluctuations during the earlyand mid-2000s. Nordic cycles exhibit additional fluctuations, particularly after 2010 for output and during the early 2000s for consumption (NC1). As might be surmised given the earlier results, the phase probabilities for the common cycles (provided in Figure 9) show limited evidence for common Nordic-Baltic cycles. NBY1 oscillates more than would be expected, and NBY2 registers a drop around 2005 that precedes the actual crisis. Only NBC1 (which captures all countries except Finland in its factor loadings) shows the pre-crisis “boom” for the entire region. This provides further evidence that examining a common Baltic cycle, and perhaps one or two common Nordic cycles, is more appropriate for future analyses. Partner phase probabilities, depicted in Figure 10, show the expected patterns, with particularly clear cycles for Russian output and Russian and U.S. output and consumption. Russia, in addition, shows contractionary periods around the time of the 1998 default and the economic sanctions following the 2014 invasion of Ukraine. Are these expansionary and contractionary periods linked to those in the Nordic and Baltic regions? We examine this question in Tables 8, 9, and 10. First, in Table 8 (which shows the highest proportions in the darkest text) we depict concordances between individual countries and their respective regional cycles. We find that of the three Baltic countries, Estonian output is most strongly linked to the Baltic cycle (with concordances during 87.2% of quarters)—but that it is even more closely linked to the Nordic cycle (89.7%). The same is true, to a lesser extent, for Latvian output (with 79.5% concordance with the Baltic cycle, and 84.6% with the Nordic cycle). Lithuanian output, on the other hand, is more closely tied to Baltic consumption. Perhaps historic links, particularly regarding Lithuania’s ties to Poland and Estonia’s connections to Finland, might help explain these differences. Of the Nordic countries, only Sweden exhibits a strong connection to its own regional cycle, perhaps because it is the main driver of the principal component. Concordances among the regional cycles are presented in Table 9. The Baltic and Nordic output cycles are highly correlated, with a concordance rate of 82.1 percent. Baltic and Nordic consumption (NC1) are also in the same phase probability during 65.4 percent of quarters. The second Nordic consumption cycle (which is primarily represented by Finland, and to a lesser extent, Sweden) has the same proportion of concordant quarters vis-à-vis Baltic and Nordic output. The other concordances, however, are quite low. Among the combined Nordic-Baltic sample, we see that output and consumption are more closely connected with their own principal components—i.e., NBY1 and NBY2 have a high concordance, as do NBC1 and NBC3. To assess the relative strength of the economic connections between the Nordic and Baltic regions and their major neighbors, we examine the business-cycle concordances that are presented in Table 10. While Baltic output is in concordance with all three partners’ GDP roughly 70 to 80 percent of the time, the value is highest vis-à-vis Germany (and lowest versus Russia). Interestingly, Baltic output is closely linked to Russian consumption, suggesting a possible role for exports in driving the Baltic cycle. Interestingly, the proportion of concordances is higher between the Baltic and Nordic cycles than between the Baltic cycle and that of any major partner. Baltic consumption is much more tightly connected to German consumption than with Russia or the United States. His differs from the CCF results, which showed the strongest link between Baltic consumption and Russia. Nordic output is tied to German GDP as well, as is the second consumption cycle. The first consumption cycle, on the other hand, is connected most strongly to German output as well as U.S. consumption. Comparing and contrasting these results with those provided via the cross-correlation functions, we conclude that the relative strength of economic integration differs for the two tests. A summary of the findings using these two approaches is provided in Table 11. The CCF method shows Baltic output to be linked to the U.S. (Global) cycle, while consumption is most correlated with Russia. Nordic GDP is connected to German output using both tests, but while the CCF method shows the same links to Germany, the MSM has the largest proportion of concordances vis-à-vis the United States. We are therefore wary when rejecting the CCF model because of its use of filtered data. In fact, we suspect (based on the phase probabilities) that the MSM only models accurate cycles in certain cases. We therefore evaluate our findings—regarding common cycles or a lack thereof, along with the relative strength of region- 444 Scott W. Hegerty 10.5709/ce.1897-9254.324DOI: CONTEMPORARY ECONOMICS Vol. 13 Issue 4 427-4452019 al synchronization—using the CCF method as well as Markov-Switching Models. 4. Conclusion While many studies have been conducted, applying a variety of statistical tests, to examine the degree of synchronization between EU business cycles and those of the bloc’s most recent entrants, relatively little has been written regarding the Baltic nations as a single economic region in this process—and even less research has been done on the Nordic economic region. This study applies two time-series methods to extract common Baltic, Nordic, and joint Baltic-Nordic output and consumption cycles, before examining connections between each country and its regional cycles, between the regional cycles themselves, and between the regional cycles and a set of neighboring economies. Using Principal Components Analysis, we find evidence of common Baltic output and consumption cycles, which is the case whether we use filtered data or growth rates. The Nordic region, however, has only a single common output cycle; consumption is often driven by a Danish-Norwegian grouping, as well as a separate Finnish (and Swedish) grouping. As a combined, seven-country “Nordic-Baltic” region, there is no single common cycle for either output or consumption. We then test for economic integration using two methods, which often provide differing results. For example, when we generate cross-correlation functions for pairs of filtered cycles, we find all three individual Baltic countries to be highly synchronized with their respective cycles. We also find that Baltic output cycles are most closely correlated to the U.S. (Global) cycle, and consumption cycles to be correlated with Russia. Our Markov-Switching models, however, show Lithuanian output to share an expansion or contraction phase with the Baltic cycle a relatively low percentage of the time, and that both Baltic output and consumption have the largest proportion of concordances vis-à-vis Germany. While Nordic output exhibits linkages to Germany via both methods, differences persist in terms of consumption. This leads us to conclude that, while recent trends in the literature have been to prefer Markov-switching models over simpler methods, these models should be treated with care. The resulting business cycles do not necessarily correspond with actual events, and the cycles, as well as the results they produce, differ compared to those given by other techniques. For this reason, we consider both CCF and MSM models simultaneously. Our results show strong evidence that the three Baltic countries represent a common economic space, which is also tied strongly to Western cycles. While the common Baltic cycle is shown to be more closely connected to the Nordic cycle than to that of Germany, the U.S., or Russia, there is no evidence that the Nordic and Baltic regions together share a common cycle. Instead, different cycles exist, one for Denmark and Norway and one for Finland, which suggest that there is only a limited “Nordic” cycle. References Aastveit, K. A., Jore, A. S., & Ravazzolo, F. (2016). Identification and real-time forecasting of Norwegian Business Cycles. International Journal of Forecasting, 32(2), 283–292. Artis, M. J., Fidrmuc, J., & Scharler, J. (2008). The transmission of business cycles: Implications for EMU enlargement. Economics of Transition, 16(3), 559-582. Babetskii, I. (2005). Trade integration and synchronization of shocks. Economics of Transition, 13(1), 105-138. Backus, D. K., Kehoe, P. J., & Kydland F. E. (1994). Dynamics of the trade balance and the terms of trade: The J-curve?, American Economic Review, 84(1), 84-103. Baxter, M., & King, R. G. (1999). Measuring business cycles: Approximate band-pass filters for economic time series. The Review of Economics and Statistics, 81(4), 575-593. Bayoumi, T., & Swiston, A. (2009). Foreign entanglements: Estimating the source and size of spillovers across industrial countries (Working Papers No. 07/182). International Monetary Fund. Retrieved from https://www.imf.org/~/media/Websites/ IMF/imported-full-text-pdf/external/pubs/ft/ wp/2007/_wp07182.ashx Benczúr, P., & Rátfai, A. (2005, January). Economic fluctuations in Central and Eastern Europe: The facts (Discussion Papers No. 4846). Centre for Economic Policy Research. Available at https://ssrn. com/abstract=717682 Buch, C. M., Doepke, J., & Pierdzioch, C. (2005). Financial openness and business cycle volatility. www.ce.vizja.pl 445 Common Baltic-Nordic Business Cycles: CorrelationVersus Markov-Switching Approaches This work is licensed under a Creative Commons Attribution 4.0 International License. Journal of International Money and Finance, 24(5), 744-765. Christiano, L. J., & Fitzgerald, T. J. (1999, July). The band pass filter (Working Paper No. 7257). National Bureau of Economic Research. Retrieved from https://www.nber.org/papers/w7257.pdf Darvas, Z. & Szapáry, G. (2008). Business cycle synchronization in the enlarged EU. Open Economies Review, 19(1), 1-19. Di Giorgio, C. (2016). Business cycle synchronization of CEECs with the Euro Area: Regime switching approach. Journal of Common Market Studies, 54(2), 284–300. Fadejeva, L., & Melihovs, A. (2008). The Baltic States and Europe: Common factors of economic activity. Baltic Journal of Economics 8(1), 75-96. Fidrmuc, J. & Korhonen, I. (2006). Meta-analysis of the business cycle correlation between the Euro Area and the CEECs. Journal of Comparative Economics, 34(3), 518-537. Frankel, J. A., & Rose A. K. (1999). The endogeneity of the optimum currency area criteria. Economic Journal, 108, 1009-1025. Furceri, D., & Karras, G. (2008). Business-cycle synchronization in the EMU. Applied Economics, 40(12), 1491–1501. Hakura, D. S. (2009). Output volatility in emerging market and developing countries: What explains the Great Moderation of 1970−2003? Czech Journal of Economics and Finance 59(3), 229-254. Hamilton, J. D. (1989). A new approach to the economic analysis of non-stationary time series and the business cycle. Econometrica, 57(2), 357–384. Hamilton, J. D. (2017). Why you should never use the Hodrick-Prescott filter (Working Paper No. No 23429). National Bureau of Economic Research. Retrieved from https://www.nber.org/papers/ w23429.pdf Hegerty, S. W. (2010). Central European business cycles: Might global (and local) linkages dominate regional ones? Eastern European Economics, 48(2), 56-73. Hodrick, R., & Prescott, E. C. (1997). Postwar U.S. business cycles: An empirical investigation. Journal of Money, Credit, and Banking, 29(1), 1–16. Horvath, J., & Ratfai, A. (2004). Supply and Demand shocks in accession countries to the Economic and Monetary Union. Journal of Comparative Economics, 32(2), 202-211. Imbs, J. (2004). Trade, finance, specialization and synchronization. Review of Economics and Statistics, 86(3), 723–734. Inagaki, K. (2006). Output correlation and EMU: Evidence from European countries. Journal of Economic Integration, 21(3), 458-473. Jiménez-Rodríguez, R., Morales-Zumaquero, A. & Égert, B. (2013). Business cycle synchronization between Euro Area and Central and Eastern European countries. Review of Development Economics, 17(2), 379–395. Mundell, R. A. (1961). A theory of optimum currency areas. American Economic Review, 51(4), 657-665. Rafiq, M. S. (2011). Understanding the interaction between international and Euro area output volatility. Bulletin of Economic Research, 63(1), 03073378.