Trade Openness and Output Comovement in EU Emerging Countries: A Panel Data Approach

The relationship between openness and economic growth has been a subject of much controversy in international macroeconomics literature. On the one hand, international trade plays a major role in the development of an economy by creating employment, and better living standards. Moreover, trade and financial openness during normal economic conditions have a positive impact on economic growth through productivity and technological innovations. On the other hand, the trade and financial linkages play a significant role in international transmission of common or idiosyncratic shocks (i.e. shocks that are country-specific). The recent global economic crisis showed a significant increased comovement of world’s economies and particularly of European Union countries. A significant debate in the related literature refers to the question whether output comovement across different regions increased as a result of financial globalization during the last decades.

The correlation between financial integration and the synchronization of economic activity depends on whether financial shocks to the banking sector or collateral shocks to firms dominate. If the economies experience a higher financially integration and the firms operating in certain countries are hit by negative shocks to their collateral, domestic and foreign banks reduce the lending in affected countries and increase lending in the non-affected ones, amplifying this way the divergence of economic growth between these economies. However, if a negative shock is affecting the banking sector in a specific country with internationally operating banks, they will draw out capitals from all countries, transmitting the domestic banking shock internationally and generating a synchronized fall in output of all countries.

Preserving financial stability is essential in order to prevent synchronized GDP growth collapses regionally. The transmission mechanism of financial shocks on economic growth synchronization during normal periods is substantially different than during crisis. If during normal periods the capital flows are channelized towards emerging markets which offer greater yields determining output divergence between regions with strong financial linkages, during crisis the financial channels favours the propagation of shocks between financially linked regions contributing to output fall synchronization.

Therefore the goal of this article is twofold. Firstly it provides empirical evidence of the impact of two important measures of external openness, namely trade and financial linkages on economic growth in seven emerging economies from the Central and East Europe (CEE). Second, it depicts the role of trade and financial linkages in international transmission of common or idiosyncratic shocks, demonstrating that during crisis times they may significantly contribute to the synchronized output fall of emerging economies.

The paper is structured as follows: the next section presents the previous empirical works on how financial and trade linkages impact on economic growth. Section 3 describes data and methodology. Section 4 reports the panel estimates on the effect of financial and trade openness on economic growth. Section 5 and 6 analyse the role of financial and trade linkages in the transmission of shocks on output comovements. Section 7 concludes.

Related Literature

Most of the theoretical literature suggests that the degree of openness impacts economic growth in a positive manner. Therefore, Romer (1993), Grossman and Helpman (1991) argue that countries that are open catch more easily with developed economies. Also, Chang et al. (2005) state that open economies allocate their resources in a more efficient manner compared to closed ones. Frankel and Rose (1998) showed that the level of trade linkages on long term is significantly positive correlated with the degree of output synchronization.

There are also opinions stating that the effects of openness on economic growth are doubtful. For instance, Rodrik and Rodriguez (2001) state that the effects of openness on economic growth may have even negative impact. They carried out a very critical  analysis of some empirical studied and concluded that in many cases the openness indicators were not appropriate measures or that the methodological choices the authors made had a strong impact on the results.

However, one major shortcoming that relates to most empirical studies is that very little attention has been paid to analysing the way international trade and financial openness may impact economic growth in emerging economies, and especially in case of Central and Eastern European countries (CEE countries). The way the market liberalisation and the increasing access of CEE countries to EU may or not have effects on economic growth is an issue well worth investigating.

Large economies originating shocks have significant effects on economic activity in emerging markets. The trade and financial linkages play a significant role in the transmission of these shocks. Claessens et al. (2012) show that financial shocks spread mostly through financial channels while, as Auerbach and Gorodnichenko (2013), Beetsma et al. (2006) argued that for fiscal policy shocks trade channels are more important that the other ones. Di Giovanni and Shambaugh (2008) argue that monetary policy shocks affect economic activity in foreign countries mostly through the interest rate channel.

Kalemli-Ozcan et al. (2013) analyse the impact of financial globalization on business cycle commovement using a proprietary database on banks’ international exposure for industrialized countries for the time span between 1978 and 2009. Following the same methodology the present article runs a series of regressions, on quarterly data, using different measures for comovements between each CEE country and the EU in order to test the effects of financial and commercial linkages on output synchronization.

Data and Methodology

Data

The sample comprises seven Central and Eastern European economies: Czech Republic, Slovakia, Poland, Hungary, Bulgaria, Lithuania and Romania. To these add the European Union, which is the main trading and financial partner of these economies. Data for each country regarding real GDP, intra-EU trade volumes and foreign direct investments were obtained from Eurostat database at annual and quarterly frequencies. The ratio of inward FDI stock to GDP comes from the UNCTAD 2008 database. The data sample for the annual analysis initially covers the interval between 1991 and 2012, and it is afterwards reduced to more recent periods due to data limitations for additional variables which are included step by step in the empirical study; for the same reason Lithuania is excluded in the annual analysis.

There are several variables through which the degree of trade openness can be measure and they are divided into two categories: trade volume measures and trade restrictions measures. Since the last one is more difficult to quantify as it includes taxes on international trade, measures of trade barriers, etc. it is preferable to use a volume trade measure quantified by the ratio of exports plus imports to GDP which is usually known as trade share. As far as the financial openness is concerned, the ratio of FDI stock to GDP is used instead of the ratio of FDI flows to GDP due to the fact that spillovers may not be related to the current FDI inflows but to the ones which have accumulated along years. However, due to the data availability limitations in case of CEE countries, for the quarterly frequency analysis the flows of FDI to GDP ratio are used.

The main explanatory variables are the ratio of inward FDI stock to GDP, the ratio of trade to GDP to which add some other control variables, such as external demand and real effective exchange rate, which actually may play a significant role in emerging economies as CEE countries. The dependent variables are the economic growth and some different measures of the degree of synchronization between CEE countries and EU, which are described in the next sections.

The impact of financial and trade linkages on output synchronization between CEE countries and EU is assessed using a balanced panel of seven pairs consisting of each CEE country included in the analysis and EU over the period Q1 2004 — Q3 2013. The panel estimates assess how the evolution of output synchronization is affected via trade and financial channels by common or idiosyncratic shocks.

Methodology

In order to assess the impact of trade openness on economic growth, static and dynamic panel data estimation methods are used, having the advantage that they allow exploiting time series but also cross sectional dimensions of data. The following equation is estimated:

〖∆GDP〗 _(i,t)=c_it+α_1∙trade_(i,t)+α_2∙fdi_(i,t)+α_3∙control_(i,t)+ε_(i,t)             (1)

H=[β Ì‚_FE-β Ì‚_RE ]^’∙[Var(β Ì‚_FE-β Ì‚_RE ) ]^(-1)∙[β Ì‚_FE-β Ì‚_RE ]                                               (2)

where β ̂_FE and β ̂_RE are the estimators corresponding to the fixed and random effects models, the statistics having the null hypothesis that the random effects estimator is more efficient. A large value of the Hausman statistic indicates the rejection of the null hypothesis. The random effects model may also be evaluated individually by running the Breusch Pagan Lagrange multiplier test with the null hypothesis that the variances across entities are zero. The possible time-invariant country fixed effects validated by the Hausman test are afterwards removed by using the first differenced Generalized Method of Moments (GMM) developed by Arrelano and Bond (1991).
 
Regarding the analysis of the effects that trade and financial linkages may have on GDP comovement, a slightly different methodology was employed. In order to evaluate the role of financial and trade linkages in transmitting the shocks originated during the economic global crisis, different correlations measures of output growth between each analysed country and the EU were regressed on the trade and financial linkages between them.

The growth rate of each country is assumed to be determined as:

y_ (i,t)=ε_t+ε_(i,t)+∑_j▒〖ρ_(i,j,t)∙ε_(j,t) 〗                               (3)

in which y_(i,t) denotes real economic growth in country i, ε_t denotes common shocks, ε_(i,t) represents domestic idiosyncratic shocks, ε_(j,t) denotes the other regions’ specific shocks, and ρ_(i,j,t) quantifies the correlation between each country and the EU within the analysed pair. Further, the analyses focus on standard linkages between different economies represented by the financial and trade channels.

Following Kalemli-Ozcan et al. (2013) the following regressions, on quarterly data, using different measures for comovements between each CEE country and the EU were run. The general form of the regression will be as following:

Comv_(i,EU,t)=c_(i,EU)+α_0∙fin_(i,EU,t-1)+β_0∙trade_(i,EU,t-1)+crisis_t+α_1∙fin_(i,EU,t-1)∙crisis_t+
+ 〖 β〗_1∙trade_(i,EU,t-1)∙crisis_t+ε_(i,EU,t)                                                                    (4)

in which Comv_(i,EU,t)  is the economic growth rate correlation between country i and the EU in period t, fin_(i,EU,t-1) and trade_(i,EU,t-1) denote the lagged financial and trade linkages between country i and the EU and crisis_t is a dummy variable which equals 1 during the crisis period. The dummy variable account for the effect of global crisis shocks that affected both output patterns and financial and trade integration. The specification also includes country pair fixed effects c_(i,EU). This effects account for non-measurable factors such as strategic coordination of EU countries, cultural ties or other unobservable factors, all of which may have a significant effect output comovement.

The results of different version of the regression tested using the quarterly data for the sample from 2004 to 2013 are in line with those reported by Kalemli-Ozcan et al. (2013) and IMF (2013) at the global scale. The effect of finance shocks in a normal state of the economy is negative, but it becomes positive during the periods of crisis.

The sample included in the analysis, spanning the period between Q1 2004 and Q3 2013, comprises a crisis period, characterized by a major financial shock, between Q3 2008 and Q2 2009. The remaining of the time interval is considered normal period. The effect of trade and financial linkages are allowed to differ across the crisis and normal periods, since the shocks which affect the economic growth in each period are of different nature. During crisis period common shocks are more likely to manifest while during normal times the idiosyncratic shocks are more frequent. For this purpose a dummy variable which takes 1 during the crisis period and 0 elsewhere is included. Using the dummy variable it is possible to test whether the financial and trade channels work differently during normal and crisis periods.

The Effects of Trade and Financial Openness on Economic Growth

Most of today’s macroeconomic literature point towards the fact that developing countries opening their economies to international trade and foreign direct investments would in time experience higher economic growth and therefore close the gap with respect to developed economies.

The CEE countries have imported a lot amount of capital and experienced high commercial and financial integration in order to catch up with developed economies. As part of the EU integration process these countries have adopted reforms and policies in order to liberalize their commercial and financial channels. Therefore the analysis of the main factors which have driven economic growth in CEE countries becomes a matter of high importance. Figure 1 depicts cross-correlations between economic growth, on one hand, and trade and financial openness, on the other hand. The relation between economic growth and the two measures of openness seems to be a positive one. High openness lead to higher economic growth, although the magnitude of the relationship could be of course conditioned by other variables as the study further demonstrates.

Sources: Eurostat Database and authors calculations.

Excluding the trade share and including another measure of the degree of openness, namely the share of FDI to GDP yields different results as model 2 shows. More specific, the constant and the dummy variable are still significant, while the coefficient of the financial variable, although positive, is statistically insignificant.

In order to obtain a more complete measure of the degree of openness, the ratio of foreign direct investments to GDP along with the trade variable is considered and then the same static panel data estimation is conducted. The foreign direct investments stock is included with one lag, taking into considerations that the spillovers related to the FDI stock impact economic activity with some delay. In this case (model 3), the coefficient of the financial variable is positive and statistically significant at 5 percent level, while the one of the commercial variable is slightly negative. The dummy variables and the constant have significant coefficients.

Notes: *significant at 1 percent level, **significant at 5 percent level, ***significant at 10 percent level.

Firstly, for the extended model static panel estimation is used, afterwards the Hausman test and Breusch Pagan test are computed. The economic growth in EU is used as a proxy for the external demand and it is included with one lag in the equation. Model 4 comprises the trade share and external demand along with the dummy variable. The coefficient of the trade share is positive, although insignificant, while external demand has a positive and significant coefficient of 0.54. Replacing the trade share with the financial variable, namely the share of FDI stock to GDP reveals a positive and statistically significant coefficient for the FDI, of almost 0.05, while external demand still has a positive, although higher, effect of 0.6. The final step consists of including both measures of the degree of openness along with the external demand as indicated by model 6. The financial variable has a positive and significant at 5 percent level coefficient of 0.06, while the trade share has a negative and insignificant coefficient.

Table 2 reports the dynamic panel data estimates obtained by using the Arellano Bond method with one period lag. The main reason for using the dynamic panel data estimator is due to the fact that according to the Hausman test in most of the cases, the fixed effects model is better suited for the previously analysed specifications, while in some cases the results are not conclusive, this being the reason for computing an additional test, namely the Breush Pagan test for assessing the potential random effects which seems to reject this hypothesis in most of the cases. Therefore, by admitting the existence of time invariant characteristics (fixed effects) one needs to consider the fact that they may be correlated with the explanatory variables. The fixed effects may be contained by the error term in equation 1 that usually consists of the unobserved country specific effects. By transforming the term by first difference the fixed effects country specific is removed because this kind of effects as previously mentioned do not vary with time. For instance equation (1) when using the Arellano Bond estimator is transformed into:

〖∆GDP〗_(i,t)=α_1∙∆trade_(i,t)+α_2∙∆fdi_(i, t)+α_3∙∆control_(i,t)+∆ε_(i,t)                              (5)

The fixed effects are contained by ε_(i,t) that consists of unobserved country specific effects, ν_i and the specific errors of the observations, u_(i,t).

ε_(i,t)=ν_i+u_(i,t)                                          (6)

When differencing for the Arellano Bond estimator the fixed effects are removed and equation (6) is transformed into:

ε_(i,t)-ε_(i,t-1)=u_(i,t)-u_(i,t-1)                        (7)

Using the Arellano Bond estimator any endogeneity problem is therefore eliminated; these endogeneity issues may appear, either from fixed effects included in the previous estimation, either from the fact that the influences between the dependent variable and the explanatory factors may run in both directions. Table 2 therefore reports the results obtained by using dynamic estimation with one period lag for all the models previously described. What is interesting to point is that, although in the static panel framework there were some specifications indicating a negative coefficient for certain measures of the degree of openness, when using the dynamic estimation, all the models indicate positive impacts of both trade share and financial openness, although in some cases (e.g. model 6) the effects are statistically insignificant. The magnitude of the commercial variable ranges between 0.013 in model 1 to 0.02 in model 3, while the share of FDI in GDP has a higher impact on economic growth reaching 0.1 in model 2. In order to check the validity and robustness of the results, the Sargan test is employed and the validity of the instruments is assessed. In all cases, the instruments are valid.