The effect of trade intensity on business cycle correlation synchronization in the Eurozone : a look after the crisis

(1)

The effect of trade intensity on business cycle

correlation synchronization in the Eurozone: a look after

the crisis

Kaloyan Dimitrov Supervisor: Stephanie Chan

Faculty of Economics and Business University of Amsterdam

(2)

1 Introduction

Ever since its inception, the European economic and monetary union has been strongly criticized by economists that claim it is comprised of a heterogeneous group of countries (Feldstein, 1998). The importance of such a claim stems from the prerequisites of the Optimum Currency Area Theory developed by Mundell (1961). Mundell theorizes that there are benefits and costs associated with membership in a monetary union, and that the net gain from joining the union increases with the degree of homogeneity between the member countries. Frankel and Rose (1997) suggest, however, that the Eurozone may have a saving grace. They find evidence that increased trade flows caused by a single currency would help synchronize the business cycles of the member countries thereby improving the overall homogeneity of the union.

There has been a great deal of research carried out that confirms the notion that larger trade flows affect business cycle synchronization positively. Furthermore, Darvas and Szapary (2004) find evidence to suggest that the euro area has indeed become more synchronized after its initial creation. However, there has unfortunately been a lack of research on the topic in more recent years.

In the aftermath of the Global Financial Crisis, the Eurozone has again come under scrutiny with criticism that not enough progress has been made in the direction of homogenization (Feldstein, 2011). A question has therefore surfaced: has the aforementioned relationship between trade and business cycle synchronization remained the same, or has the crisis altered this relationship by affecting the channels through which the two variables interact with each other? Empirical research is needed to provide an answer to this question.

To that effect, the paper first extracts information on the business cycles of all 17 Eurozone member states by analyzing real GDP and employment data using three different filtering techniques – the Hodrick-Prescott, Baxter-King and Christiano-Fitzgerald filters. The second step involves computing the bilateral correlations between the business cycles over two time periods: 2004-2007 and 2008-2011. Finally, to measure the effect of trade intensity on business cycle correlation, different regressions are run using business cycle correlation as a dependent variable and bilateral trade intensity as an independent variable. The six different versions of the dependent variable combined with two different proxies for bilateral trade

(4)

2

intensity produce 12 different regressions for each period. Trade intensity is instrumented using several gravity model variables in order to resolve endogeneity concerns.

The results for the period 2004-2007 confirm the findings of previous literature that bilateral trade intensity has a positive effect on business cycle synchronization. After the crisis, however, the effect between the two variables is found to not be significantly different from zero. The findings appear to be robust to the selected measure of real economic activity, the detrending technique applied, and the definition of trade intensity used.

The structure of the paper is as follows. Section 2 guides the reader through related literature. Section 3 goes into depth regarding the methodology applied in this paper. Section 4 presents and discusses the results of the research. Section 5 provides the conclusion and offers ideas for future investigation.

2 Related Literature

2.1 Measures of economic activity

The most commonly used data in literature is real GDP and industrial production. With regards to the frequency used, annual data is generally avoided in order to capture higher frequency fluctuations. However, quarterly real GDP data does not exist for all countries. For example, when working with a set of developing countries, Calderon et al. (2007) were forced to make use of annual data instead. On a positive note, real GDP is the most encompassing measure available.

An advantage of using industrial production as a measure is that monthly data is readily available. However, the drawbacks of this data relate to the fact that industrial production represents only a small component of output, and that its volatility is generally much higher than broader measures of output.

Although not the most prevalent choice, Frankel and Rose (1998) make use of employment and unemployment data. This is beneficial due to the broad nature of this data set and its general availability.

(5)

3

2.2 Methods for extracting the business cycle

The simplest method to quantify the business cycle, as discussed in literature, is to compute the first differences of the natural logarithms of the data set in question. Among many, Kose and Yi (2002) employ this technique. According to Baxter and King (1999), however, this is not an optimal method. They determine that timing relationships are altered between variables and, more importantly, that it over-weights higher frequency fluctuations.

The Hodrick-Prescott filter is another popular technique used in literature. It is a high-pass filter which means that it focuses on fluctuations with higher frequency and removes the rest. Baxter and King (1999) point out several drawbacks of the HP filter and instead propose that the ideal detrending technique is a band-pass filter. Band-pass filters remove fluctuations with a frequency above and below certain predetermined cutoff points. Baxter and King (1999) believe this method is ideal because it takes into an account the statistical properties of the business cycle. Their method approximates the optimal band-pass filter, however both the HP and BK filter are not able to deal completely with low-frequency fluctuations and allow some to pass through. The Christiano and Fitzgerald (2003) filter is another band-pass filter. It is shown to perform better at stopping the low-frequency fluctuations compared to the HP and BK alternatives. Unlike the BK filter, it relies on the assumption that the data is a random-walk process. The authors believe their filter is suitable to use when extracting the business cycle as most economic time series can be approximated by random-walk processes. Furthermore, unlike the BK filter, no observations are lost when applying the CF filter.

Canova (1998) makes the important observation that each filtering method provides different types of information. Under some circumstances this can lead to misleading results. Burnside (1998) notes that economists must not limit their research by using too few filtering techniques, as doing so would reduce the power of their findings.

2.3 Measuring business cycle synchronization

The most common method used in literature to quantify the degree of synchronization of business cycles between countries is the Pearson correlation. This is probably because of its relative simplicity. Some of the authors who use it are Otto et al. (2001) and Calderon et al. (2007). However, researchers have also proposed more complex methods of measurement.

(6)

4

Harding and Pagan (2002) introduce their form of a concordance index. The measure works by calculating the proportion of time in which the two data series are in the same phase of the business cycle. The disadvantage to this approach is that it examines a binary variable and thus risks discarding potentially valuable information.

On the other hand, Bayoumi and Eichengreen (1997) have developed an index of business cycle asymmetries that works by measuring the standard deviation of the change in relative output between two countries. Essentially, if the outputs of two countries move together, the value for the measure will be low.

A dynamic correlation measure has been introduced by Croux et al. (2001), labelled “cohesion”. The measure is useful when trying to gauge the correlation of business cycles within a group of countries. However, when working with only two economies and a finite sample size, the method becomes imprecise.

2.4 Importance of highly synchronized business cycles

Following the work of Mundell (1961), it has become clear that a high degree of synchronization between the business cycles of member states is necessary to ensure the sound functioning of a monetary union. If, for example, member states were in different phases of their business cycles, then the common monetary policy imposed by the union would not be ideal for the individual circumstances of each member. In other words, a country in a recession would benefit from a reduction of interest rates to bolster demand, whereas a country with strong economic activity would prefer an increase in order to curb inflationary pressures and cool down their economy.

The European Union generally suffers from the notion that a handful of key members subsidize smaller, troubled nations which lack competitiveness, such as Greece. This has on multiple occasions halted efforts aimed at expanding the roles of the union. Synchronized business cycles might help alleviate these concerns and thus facilitate further fiscal integration within the European Union.

(7)

5

2.5 Determinants of business cycle synchronization

Researchers have determined several factors that have may have a significant effect on the synchronization of business cycles. For example, the coordination of fiscal policies has been found to have a positive effect (Darvas et al., 2005). Inklaar et al. (2008) reach the same conclusion in their own research. They believe this is because fiscal integration reduces idiosyncratic shocks in a system of countries.

Financial market integration is considered to have two transmission channels working in opposing directions. On the one hand, closely linked financial markets increase the effect of spillovers between countries, thus leading to higher synchronization of their business cycles. On the other hand, it also promotes a higher degree of industrial specialization as capital is more mobile, which in turn would lead to asymmetries in their output. The findings of Imbs (2004) suggest that the first channel is stronger.

The effect of monetary policy integration, however, is more ambiguous. Bergman (2004) finds that higher exchange rate volatility leads to higher business cycle correlation. He believes that exchange rates serve as an adjusting mechanism and moving to a fixed exchange rate policy is detrimental. On the other hand, Inklaar et al. (2008) find that coordinated monetary policies have a positive on correlation. Furthermore, Glick and Rose (2002) reach the conclusion that members of a currency union trade with each other nearly double as much as they otherwise would. The effect this has on business cycle correlation, however, is ambiguous.

The explanatory variable most discussed in literature, and in fact the variable this paper focuses on, is bilateral trade intensity. Researchers have presented conflicting opinions regarding the direction of the effect. Krugman (1993) points out that increased trade leads to greater levels of specialization and thus has a negative effect on the synchronization of business cycles. However, such a claim is only true if the majority of trade is between different industries. This view is supported by economic theory through the models of Ricardo, whereby an openness to trade induces countries to specialize in that which they have a comparative advantage. If the trade that dominates is intra-industry, however, the effect on business cycle correlation is positive. The logic behind this is that if countries have developed similar industries with each country specializing in a specific part of the supply chain, then industry

(8)

6

specific shocks would impact all members. Another important transmission mechanism is through aggregate demand. Trade can intensify spillovers from partner countries caused by demand shocks. For example, a recession in a trading partner may cause a reduction of exports in the home country, thus placing downward pressure on the home economy and increasing the synchronization between the two countries. Researchers tend to focus on the net effect of these channels. For instance, Frankel and Rose (1998) find evidence to suggest a positive effect of trade on business cycle correlation. In a newer study, Gruben et al. (2002) confirm this but estimate the effect to be weaker.

For the period before the crisis, we expect to confirm past findings. It is unclear whether this will remain the case for the period after the crisis as the severity of the shock could have caused a shift in the relative strength of the different channels. There has not yet been any research that covers the period after 2008.

3 Methodology

3.1 Variables and Data

The first step in measuring the effect of trade intensity on business cycle correlation is selecting a way to measure the two concepts. Two different proxies for economic activity are used to calculate the bilateral business cycle correlations between the countries of the Eurozone. In particular, quarterly real GDP and quarterly total employment data measured in persons spanning between 2002 and 2013, sourced from the European Central Bank’s Statistical Data Warehouse. These have been chosen as they are measures that capture a wide view of the overall economy.

Natural logarithms are taken of both datasets and the cyclical component is extracted. Due to a lack of consensus among researchers regarding the optimal detrending method, three different filters are utilized in order to ensure the robustness of the results. Specifically, the Hodrick-Prescott (HP), Baxter-King (BK) and Christiano-Fitzgerald (CF) filters. Baxter and King (1999) suggest discarding three years of data (12 quarters) at the beginning and at the end of the sample when using the HP or BK filters. This is done in order to overcome shortcomings

(9)

7

of the two filters. However, due to the lack of more recent data, only two years (8 quarters) will be dropped. Lastly, a smoothing parameter of 1600 is used when applying the HP filter as suggested by Ravn and Uhlig (2002).

Finally, the Pearson correlation between the cyclical parts of real economic activity of countries 𝑖 𝑎𝑛𝑑 𝑗 is calculated over two periods: 2004-2007 and 2008-2011. The sample size of 17 countries and the 3 detrending methods provide us with 3 sets of 136 observations for each period. The countries in the sample have generally become more closely correlated in the period after the creation of the Eurozone. However, the Global Financial Crisis had a particularly negative effect on Greece and the resulting prolonged recession is expected to be reflected in the data as a divergence between Greece and the rest of the European Union after 2008. In other words, the Pearson correlations between Greece and the other economies should decrease in the second period covered by this paper. This divergence between Greece and the Euro area is illustrated by Graph 1. The economic difficulties in Greece signaled the start of the European Sovereign Debt Crisis that also affected several other countries (Ireland, Portugal, Spain and Cyprus). The effects on these countries, however, came a few years later. Due to these timing differences, the correlations between these countries and other Eurozone members should not decrease significantly between 2008 and 2011 (in contrast to Greece). However, if future research includes 2012 in the period used to calculate the correlations between countries the divergence should be more profound.

The second important variable is bilateral trade intensity between countries 𝑖 𝑎𝑛𝑑 𝑗. The benefit of using the intensity approach is that it eliminates the differences in scale between countries and thus allows for the analysis of developments in their trade patterns. In order to ensure the robustness of the results, two measures of bilateral trade intensity between countries 𝑖 𝑎𝑛𝑑 𝑗 are used. These same measures are utilized by Calderon et al. (2007):

(𝑒𝑞. 1) 𝑇𝐼_𝑖,𝑗 = 𝑡𝑖,𝑗

𝑇_𝑖 + 𝑇_𝑗 𝑎𝑛𝑑 (𝑒𝑞. 2) 𝑇𝐼𝑖,𝑗 = 𝑡_𝑖,𝑗 𝑌_𝑖 + 𝑌_𝑗

Where 𝑡_𝑖,𝑗 is the amount of trade between countries 𝑖 𝑎𝑛𝑑 𝑗, 𝑇_𝑖 and 𝑇_𝑗 represents the total trade volume (imports plus exports) of each country, and 𝑌_𝑖𝑎𝑛𝑑 𝑌_𝑗 denotes the output of each trade

(10)

8

partner. The ratios are calculated yearly between 2004 and 2011 for each country pair and then averaged out over the two periods mentioned earlier: 2004-2007 and 2008-2011. As noted by Calderon et al. (2007), exports from country 𝑖 to country 𝑗 are not necessarily equal to imports of country 𝑗 from country 𝑖. The solutions they provide are also followed in this paper. Specifically, relying on the data provided by the country with higher GDP.

Annual bilateral trade data is obtained from the Organisation for Economic Co-operation and Development’s Structural Analysis Database. The data covers the volume of goods traded among the sample countries between 2004 and 2011. As shown by Graph 2, average bilateral trade flows in the Eurozone have steadily increased between 2004 and 2008, however, the onset of the Global Financial Crisis has led to a drop between 2008 and 2009. After 2009 bilateral trade levels have begun to recover, reaching nearly pre-crisis levels in 2011. Greece is an exception to this trend due to the severity of the impacts on the economy caused by the crisis. Graph 3 shows the average bilateral trade flows between Greece and its partners in the Eurozone. This figure shows similarities to Graph 1 between 2004 and 2009, however the recovery phase after 2009 is not present.

3.2 Regression model

In order to answer the research question, the following regression will be performed:

𝐶𝑜𝑟𝑟(𝑦𝑖, 𝑦𝑗) = 𝛼 + 𝛽 ∗ ln (𝑇𝐼𝑖,𝑗) + 𝜀𝑖,𝑗

Where 𝐶𝑜𝑟𝑟(𝑦_𝑖, 𝑦_𝑗) denotes the business cycle correlation between countries 𝑖 𝑎𝑛𝑑 𝑗 over the given period, and ln (𝑇𝐼_𝑖,𝑗) is the natural logarithm of the bilateral trade intensity ratio mentioned above. There are 6 different sets of dependent and 2 sets of independent variables for each period. In total, 24 regressions (12 for each period) are run to measure the effect of bilateral trade intensity on business cycle correlation before and after the Global Financial Crisis (denoted in the regression by 𝛽). We would expect to see a positive coefficient of bilateral trade intensity before the crisis as that is what previous research concludes. However, after the crisis, the effect is ambiguous as it depends on whether there has been a change to the relative strength of the transmission channels.

(11)

9

As noted by Frankel and Rose (1998), both trade intensity and business cycle correlation are endogenous. There exists the issue of simultaneous causality between the variables. This in turn makes an ordinary least squares regression inappropriate. A technique widely used in literature is applied in order to overcome these issues. Specifically, trade intensity is instrumented using determinants of trade given by the gravity trade model. The gravity model of trade was first introduced by Tinbergen (1962) with the aim of measuring the effect of economic size and distance on bilateral trade flow volumes. Over the years, the model’s empirical strength has been shown to persist (Leamer and Levinsohn, 1995). The variables used in the model are considered to be exogenous and thus will allow us to deal with the endogeneity issues discussed above. Bilateral trade intensity is instrumented as follows:

𝑇𝐼𝑖,𝑗 = 𝛼 + 𝛽1∗ ln(𝑑𝑖𝑠𝑡𝑖 ,𝑗) + 𝛽2∗ 𝑑𝐵 + 𝛽3∗ ln(𝑦𝑖) + 𝛽4ln(𝑦𝑗) + 𝜀𝑖 ,𝑗

The chosen instruments are similar to the ones used by Fidrmuc (2008). In particular, the natural logarithm of the distance between the main cities of countries 𝑖 𝑎𝑛𝑑 𝑗, a dummy variable equal to 1 if countries 𝑖 𝑎𝑛𝑑 𝑗 share a common border, the natural logarithm of the average output of country 𝑖 for the period, and the natural logarithm of the average output of country 𝑗 for the period. The instruments are tested for relevance using an F-test on the first stage regression. All regressions are carried out using robust standard errors estimates in order to correct for the potential presence of heteroscedasticity in the data. If this is not done and heteroscedasticity is present the standard errors reported would be biased.

4 Results and Discussion

4.1 Trade Intensity defined using equation 1

The regression results are summarized in Table 1. For the 2004-2007 period, the effect of bilateral trade intensity on business cycle correlation is found to be statistically significant with an alpha of 5% in all regressions except when employment is used as a measure of real economic activity with the Baxter-King filter applied. The point estimates range from

(12)

10

0.0228815 to 0.0730078. These results are consistent with previous research findings that bilateral trade intensity has a positive effect on business cycle correlation (Gruben et al., 2002). The results can also be used to say something about the relative strength of the transmission channels through which trade affects business cycle correlation as discussed in section 2.5. The results suggest that the positive effects of the aggregate demand and intra-industry trade channels are stronger than the negative specialization-effect induced by the inter-industry channel.

For the 2008-2011 period, the point estimates range from -0.0509585 to 0.0006323. However, only two of the results are significant at the 5% level. This observed decrease in the effect of trade intensity on business cycle correlation is likely caused by the specialization-effect transmission channel becoming relatively stronger after the crisis. Even though bilateral trade does not seem to be improving business cycle synchronization after the crisis, it does not mean that the process is not continuing through other channels such as fiscal or financial market integration.

4.2 Trade Intensity defined using equation 2

The regression results using equation 2 as a measure of trade intensity are summarized in Table 2. For the 2004-2007 period, the estimated coefficients are significant at the 5% level for both measures of economic activity with all three detrending techniques. Similarly to the previous subsection, the point estimates range from 0.0249371 to 0.1141606 and thus lead to the same conclusions.

For the years between 2008 and 2011, none of the estimated coefficients are significantly different from zero at the 5% level. Once more, the effect of trade intensity has seemingly decreased implying the same conclusions as those of the previous subsection.

4.3 Validity of results

A decrease in the effect of trade intensity on business cycle correlation is observed among all of the estimates. This leads us to believe that the results are robust with respect to the measure of real economic activity, the detrending technique applied, and the definition of trade intensity used.

(13)

11

When defining bilateral trade intensity using equation 1 the choice of instrumental variables has yielded an F-statistic for the first stage regression of 73.49 for the first period and 64.42 for the second. When defining bilateral trade intensity using equation 2 the choice of instrumental variables has yielded an F-statistic for the first stage regression of 64.49 for the first period and 59.10 for the second. These values are significant and one can conclude that the variables used are strong instruments in terms of their relevance.

However, the first limitation of these results stems from the fact that the F-statistic does not tell us whether the instrumental variables used have resolved the endogeneity concerns from section 3. If the instruments are endogenous themselves, the regressions would yield biased estimates of 𝛽.

Another issue with the results is the lack of relevant data after 2013. Both the Hodrick-Prescott and Baxter-King filters would benefit from an increase in available data points. Currently, only 8 quarters worth of data is discarded instead of the 12 suggested by literature. This implies that the quality of the estimates using the HP and BK filters may be slightly impaired.

Lastly, it is important to mention that the data has not been tested for the presence of autocorrelation. If autocorrelation exists in the data, tests using the 𝑡 𝑎𝑛𝑑 𝐹 statistics would be unreliable as the standard errors estimated would be biased, even though the slope coefficients would be unbiased. As a result of the bias present in the standard errors, it is possible to incorrectly evaluate the null hypothesis.

5 Conclusion

The paper attempts to address the lack of research in recent years regarding the effect of trade intensity on business cycle synchronization. If the findings of past literature are confirmed to persist after the crisis then proponents of the union would be able to subdue criticism over the validity and effectiveness of the Eurozone.

Trade intensity is found to have a statistically significant positive effect on business cycle synchronization between 2004 and 2007. This is in line with literature, in the sense that the specialization channel is not strong enough to create a negative net effect. Between 2008

(14)

12

and 2011, however, the net effect is no longer statistically different from zero. A potential explanation for this is that the crisis has increased the relative strength of the specialization channel or has diminished the effect of the other two channels. These results are found to be robust.

In general, these findings should be taken with a grain of salt as they are not without limitations. A larger timeframe would improve the accuracy of the Baxter-King and Hodrick-Prescott filters. Secondly, if the instrumental variables are not truly exogenous, the regressions would yield biased and inconsistent results. Autocorrelation might exists in the data causing uncertainty in the conclusions relating to the significance of the effects. Furthermore, even if the results are reliable one cannot infer much about the way in which the specialization channel has become relatively stronger. Future researchers should focus on resolving these issues in order to portray a more accurate representation of the mechanisms at play.

(15)

13

6 Bibliography

Baxter, M., & King, R. G. (1999). Measuring business cycles: approximate band-pass filters for economic time series. Review of economics and statistics, 81(4), 575-593. Bayoumi, T., & Eichengreen, B. (1997). Ever closer to heaven? An optimum-currency-area

index for European countries. European economic review, 41(3), 761-770. Bergman, U. M. (2004). How similar are European business cycles?. Economic Policy

Research Unit, Department of Economics, University of Copenhagen.

Burnside, C. (1998). Detrending and business cycle facts: A comment. Journal of Monetary

Economics, 41(3), 513-532.

Calderon, C., Chong, A., & Stein, E. (2007). Trade intensity and business cycle

synchronization: Are developing countries any different?. Journal of International

Economics, 71(1), 2-21.

Canova, F. (1998). Detrending and business cycle facts. Journal of monetary economics,

41(3), 475-512.

Christiano, L. J., & Fitzgerald, T. J. (2003). The Band Pass Filter*. International Economic

Review, 44(2), 435-465.

Croux, C., Forni, M., & Reichlin, L. (2001). A measure of comovement for economic variables: theory and empirics. Review of Economics and Statistics, 83(2), 232-241. Darvas, Z., & Szapary, G. (2004). Business cycle synchronization in the enlarged EU:

Comovements in the new and old members.

Darvas, Z., Rose, A. K., & Szapáry, G. (2005). Fiscal divergence and business cycle

synchronization: irresponsibility is idiosyncratic (No. w11580). National Bureau of

Economic Research.

Feldstein, M. (1998). The political economy of the European economic and monetary union:

political sources of an economic liability (No. w6150). National Bureau of Economic

Research.

Feldstein, M. S. (2011). The euro and European economic conditions (No. w17617). National Bureau of Economic Research.

(16)

14

Fidrmuc, J. (2008). The Endogeneity of the Optimum Currency Area Criteria, Intra‐industry Trade, and EMU Enlargement. Contemporary Economic Policy, 22(1), 1-12.

Frankel, J. A., & Rose, A. K. (1997). Is EMU more justifiable ex post than ex ante?.

European Economic Review, 41(3), 753-760.

Frankel, J. A., & Rose, A. K. (1998). The endogenity of the optimum currency area criteria.

The Economic Journal, 108(449), 1009-1025.

Glick, R., & Rose, A. K. (2002). Does a currency union affect trade? The time-series evidence. European Economic Review, 46(6), 1125-1151.

Gruben, W. C., Koo, J., & Millis, E. (2002). How much does international trade affect

business cycle synchronization? (Vol. 2, No. 3). Federal Reserve Bank of Dallas.

Harding, D., & Pagan, A. (2002). Dissecting the cycle: a methodological investigation.

Journal of monetary economics, 49(2), 365-381.

Imbs, J. (2004). Trade, finance, specialization, and synchronization. Review of Economics and

Statistics, 86(3), 723-734.

Inklaar, R., Jong-A-Pin, R., & De Haan, J. (2008). Trade and business cycle synchronization in OECD countries—A re-examination. European Economic Review, 52(4), 646-666. Kose, M., & Yi, K. M. (2002). The trade comovement problem in international

macroeconomics. FRB of New York Staff Report, (155).

Krugman, Paul, 1993, “Lessons of Massachusetts for EMU,” in F. Giavazzi and F. Torres, eds., The Transition to Economic and Monetary Union in Europe, Cambridge University Press, New York, 241-261.

Leamer, E. E., & Levinsohn, J. (1995). International trade theory: the evidence. Handbook of

international economics, 3, 1339-1394.

Mundell, R. A. (1961). A theory of optimum currency areas. The American Economic Review,

51(4), 657-665.

Otto, G., Voss, G. M., & Willard, L. (2001). Understanding OECD output correlations. Reserve Bank of Australia.

Ravn, M. O., & Uhlig, H. (2002). On adjusting the Hodrick-Prescott filter for the frequency of observations. Review of Economics and Statistics, 84(2), 371-376.

(17)

15

Graph 1: Business cycle correlation between Greece and the Eurozone aggregate (2004-2011).

Graph 2: Average bilateral trade flows in the Eurozone (2004-2011).

-1,5 -1 -0,5 0 0,5 1 1,5 2004 2005 2006 2007 2008 2009 2010 2011 0 4000000 8000000 12000000 16000000 20000000 2004 2005 2006 2007 2008 2009 2010 2011

(18)

16

Graph 3: Average bilateral trade flows between Greece and its Eurozone partners (2004-2011). 0 800000 1600000 2400000 3200000 4000000 2004 2005 2006 2007 2008 2009 2010 2011

(19)

17

Table 1: Estimates of the effect of Trade intensity (equation 1) on business cycle correlation before and after the crisis.

Measure Method 2004-2007 2008-2011 GDP Hodrick-Prescott 0.0228815* (0.0056705) -0.0253757 (0.0178740) GDP Baxter-King 0.0325386* (0.0093127) -0.0423698* (0.0210743) GDP Christiano-Fitzgerald 0.0613531* (0.0161480) -0.0509585* (0.0256662) Employment Hodrick-Prescott 0.0497556* (0.0140029) -0.0014045 (0.0256457) Employment Baxter-King 0.0371884 (0.0202698) -0.0348732 (0.0244076) Employment Christiano-Fitzgerald 0.0730078* (0.0366370) 0.0006323 (0.0317973)

* indicates that the coefficient is statistically significant at 5%. Robust standard errors given in parenthesis.

The F-statistic of the first stage regression is equal to 73.49 for the 2004-2007 period and 64.42 for the 2008-2011 period.

(20)

18

Table 2: Estimates of the effect of Trade intensity (equation 2) on business cycle correlation before and after the crisis.

Measure Method 2004-2007 2008-2011 GDP Hodrick-Prescott 0.0249371* (0.0059019) 0.0017992 (0.0145311) GDP Baxter-King 0.0378060* (0.0096239) -0.0065793 (0.0169418) GDP Christiano-Fitzgerald 0.0779813* (0.0161210) -0.0075575 (0.0205319) Employment Hodrick-Prescott 0.0588731* (0.0144177) 0.0402875 (0.0211904) Employment Baxter-King 0.0500961* (0.0198949) 0.007049 (0.0198053) Employment Christiano-Fitzgerald 0.1141606* (0.0352275) 0.0508249 (0.0272509)

* indicates that the coefficient is statistically significant at 5%. Robust standard errors given in parenthesis.

The F-statistic of the first stage regression is equal to 64.49 for the 2004-2007 period and 59.10 for the 2008-2011 period.

The effect of trade intensity on business cycle correlation synchronization in the Eurozone : a look after the crisis