CURRENT ACCOUNT IN THE PERIPHERIAL MEMBERS OF THE EUROPEAN UNION - PANEL ANALYSIS

U ovom radu se istražuje povezanost salda tekućeg računa perifernih članica evrozone (Italija, Španija, Portugalija, Irska i Grčka) sa glavnim determinantama koje ga opredeljuju u periodu 20002017. godine. Istraživanje je sprovedeno metodologijom panela. Saldo tekućeg računa je zavisna varijabla, a set izabranih relevantnih objašnjavajućih makroekonomskih veličina predstavlja regresore panela. Pošto su panel vremenske serije posedovale jedinični koren, sprovedeni su testovi kointegracije. S obzirom da su rezultati ovih testova potvrdili postojanje kointegracije između vremenskih serija panela, pristupilo se oceni kointegracione regresione jednačine. Za ovu ocenu primenjeni su FMOLS i DOLS estimatori. Ocenjeni koeficijenti kointegracionih regresora pomoću FMOLS estimatora su statistički signifikantni za celo razdoblje 2000-2017. godine. Jedini izuzetak na nivou polaze jednačine je FMOLS ocena parametra uz doznake, koja nije statistički signifikantna. U slučaju primene DOLS estimatora, nekoliko ocenjenih parametara nije pokazalo signifikantnost. Znaci ispred ocena parametara uglavnom se slažu sa teorijskim pristupom. Ocenjeni parametri pokazuju dugoročnu usklađenost kretanja salda tekućeg računa i izabranih makroekonomskih regresora. Dobijeni nalazi su potvrili perzistentnost salda tekućeg računa perifernih članica evrozone.


Introduction
The introduction of the euro was followed by the period of deficit expansion in the peripheral eurozone members.In this paper, the core of the eurozone will include Germany, France, the Netherlands, Belgium, Austria and Finland, and the group of peripheral members will include Italy, Spain, Portugal, Ireland and Greece.Greece and Portugal had a large current account deficit at the time of the introduction of the euro, Spain had a smaller deficit, while Ireland had a balance equilibrium.After the introduction of the euro, the current account balance deteriorated in the all peripheral eurozone members, except for Portugal, where it remained at the level of entry into the eurozone.This is documented in numerous analyzes (for example, Berger and Nitsch, 2010; Jaumotte and Sodsriwiboon, 2010).These studies point out that a large inflow of foreign capital spurred domestic demand, which caused an increase in imports and widening of the current account deficit.The increase in global liquidity facilitated the borrowing of peripheral members.The imbalance of the public sector in Greece and Portugal before the outbreak of the global economic crisis in 2008 was the main cause of the current account deficit growth.
The persistence of the current account deficit for the eurozone members is explained in different ways in reference literature.Reduced competitiveness due to higher labor costs and nonproductive investments are most often referred to as the reasons for this phenomenon.Blanchard and Giavazzi (2002) consider that the correlation between national savings and investment in the period after the introduction of the euro significantly dropped.Although the neoclassical economics emphasizes that an automatic rebalancing of the current account balance should occur on the integrated capital market, this has not happened in the case of peripheral members of the eurozone.Cesaroni and De Santis (2015) analyzed dualism in the movement of the current account deficit between peripheral members and the core of the eurozone, focusing on the role of the European financial integration process.The authors of this paper concluded that financial integration had an important role in explaining the dynamics of the current account in the European Union countries.They also presented evidence of the adverse impact of financial integration on the peripheral eurozone members.The negative correlation between financial integration and current account, according to the findings of these authors, increased after the introduction of the euro.
This paper examines the determinants of the current account balance in peripheral eurozone members.The paper contributes to the existing literature in three ways.First, it contains an empirical analysis of the determinants of the current account balance in the peripheral eurozone members throughout the period 2000-2017, as well as in two subperiods -one until the outbreak of the crisis (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007) and another after the crisis (2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017).Second, the paper uses the panel regression analysis, which includes numerous regressors and estimates their impact on the current account balance.In the end, in order to assess the role of management and quality of institutions in relation to the current account imbalance, the analysis also includes a regressor representing qualitative variables.
The paper is organized in the following way.The second part presents stylized facts concerning the imbalance of the current account in the peripheral eurozone members.The third part describes the sources of data and the time frame of the analysis, while the fourth part draws on the empirical research strategy.In the fifth part, empirical results and discussion are exposed, along with the robustness test.
The capital flows within the eurozone enabled the financing of public debt in Greece, the purchase of real estate in Spain and Ireland, while both models were applied in Portugal and Italy.This structure of capital flows shows that investors believed that the securities of the peripheral eurozone members are close substitutes for the securities issued by the countries of the eurozone core.According to Figure 2, after the introduction of the euro until the outbreak of the global economic crisis, the persistence of the current account deficit in the peripheral eurozone members increased in 2008, with the slightest deterioration in Italy, and the highest in Greece.Of course, it is impossible to claim that the introduction of the euro contributed to the increase of this deficit, but other factors also affected it.The deterioration of the current account balance was also recorded in the countries outside the euro area -Denmark, Sweden and the United Kingdom.After the outbreak of the financial crisis in 2008, there was a reduction in the current account deficit of the peripheral eurozone countries and countries outside this zone (Atoyan et al., 2013).
The capital flows were significantly influenced by the global financial crisis 2007-2008.The peripheral members were not immediately affected by this crisis, but in 2010 there was a sudden breakdown of private capital flows, which intensified public debts crisis in some countries.The credit rating of Ireland, Greece and Portugal worsened, with a rise in their public debt in relation to Germany's public debt.The adjustment of their current account was not possible by means of depreciation, because of their status as the eurozone members.As a result, these countries continued to record the current account deficit, but in smaller proportions.Thanks to the large
Schmitz and von Hagen (2011) su istraživali uzroke neravnoteža tekućeg računa unutar evrozone i došli do zaključka da su razlike u per capita dohotku između periferije i jezgra bile ključni činilac koji je doveo do privlačenja stranog kapitala i do širenja deficita tekućeg računa.Takođe su ocenili da su razlike u dohotku pojačale uticaj na neravnotežu tekućeg računa unutar evrozone posle uvođenja evra.Na neravnoteže sa partnerima izvan EU uticala je, po mišljenju ovih autora, apresijacija evra u Kovačević R. Tekući račun perifernih članica Evropske unije -panel analiza Bankarstvo, 2018, vol.47, br. 4 central bank borrowings from the European Central Bank (ECB) through a network known as the "Eurosystem Target2", these countries have compensated for the reduced inflow of private capital.In this way they avoided the balance of payments crisis, but they had to approach a gradual rebalancing of the current account.The increase of export competitiveness of these countries requires the lowering of their wages against the German's.The mitigating circumstance is that through the Target2 system they can adjust their current account gradually, in the context of the public debt crisis.

Literature Review
The literature examining the dynamics of the current account balance in peripheral eurozone members is abundant, and mainly refers to the period before the outbreak of the 2008 crisis.The research of the current account developments cites a number of reasons for the imbalance in the eurozone.One group of authors considers that the answer to this question lies in the field of financing and concludes that a large capital inflow into peripheral countries is the main cause of large current account imbalances.Others believe that the loss of external competitiveness is the leading cause of imbalance.
Chinn and Ito (2008) especially emphasized the role of financial development and institutions in the analysis of current account determinants.As an indicator of financial development, the coefficient of private loans/GDP on the sample of 70 developing countries and 19 developed countries was used in the period 1971-2004.They identified the negative impact of this indicator on the current account of both developed and developing countries.
Jaumotte and Sodsriwiboon (2010) applied a panel analysis on the sample of 49 developed countries (including 27 EU members) and emerging markets for the period 1973-2008.These authors found that financial openness reduces the aggregate current account balance, but did not observe a separate impact of financial integration on the core and periphery countries.
Unlike the previous findings, Kentsch (2010) concluded that the current account imbalances within the euro area were mainly due to discrepancies in export performance between the periphery and the core.This is particularly relevant for the period after 2003, when imports were developing with a similar dynamics across the eurozone.The conclusion of this study is that the current account imbalances were created more due to the loss of export competitiveness than the excessive inflow of capital due to financial integration.Ahearne et al. (2007) and Schmitz and von Hagen (2011) used the data on per capita income as an approximation of financial depth, and in the case of the EU 15, they concluded that the differences in income were followed by net flows, even before the introduction of the euro.Their econometric research has confirmed the existence of differences in the financial depth of the European capital markets, and that capital moved from a country in which it is an abundant factor of production to the countries in which it was a rare factor.
Belke and Dreger (2013) compared the effects of per capita income and competitiveness diferences on the current account, using panel cointegration for 11 eurozone members in different subperiods.By applying this method, these authors concluded that the differences in competitiveness were the main factor of the external imbalance of the euro area.In a separate regression for Greece, Spain and Portugal, the authors pointed to the declining role of per capita income (this variable was used as a substitute for catching up) for these countries over time.In order to reduce the current account deficit of these countries, depreciation of the exchange rate is recommended.
Sanchez and Varoudakis (2013) examined the impact of selected macroeconomic factors on the current account imbalance in the euro area.These authors applied a vector autoregressive model to the panel data for the period 1975-2011, and did not observe the impact of financial integration.Based on the findings of this model, the authors concluded that demand shocks had more influence on the current account dynamics of the peripheral eurozone members than on the core members, while competitiveness had a lesser role in peripheral members, and a more prominent role in the core countries.
Schmitz and von Hagen (2011) investigated the causes of the current account imbalance odnosu na valute glavnih trgovinskih partnera.
Kovačević R. Tekući račun perifernih članica Evropske unije -panel analiza Bankarstvo, 2018, vol.47, br. 4 within the eurozone and concluded that the per capita income differences between the periphery and the core were the key factor that led to attracting foreign capital and expanding the current account deficit.They also estimated that differences in income increased the impact on the current account imbalance within the eurozone following the introduction of the euro.According to the opinion of these authors, the imbalances with partners outside the EU were influenced by the appreciation of the euro relative to the currencies of the main trading partners.
Holinski, Kool, and Muysken (2012) found that most of the surpluses of the northern eurozone's members are the result of the fiscal consolidation, greater competitiveness, a more stable relationship between private savings and investments, and greater inflows of net factor incomes from abroad.The findings by these authors show that peripheral members were faced with a decline in private savings after 1992, which led to an increase in indebtedness and deterioration in the trade balance.Peripheral members were faced with an increase in the net outflow of factor income due to an increase in external indebtedness.However, the authors believe that the pro-cyclical common monetary policy of the eurozone before the outbreak of the crisis contributed to the widening of the current account deficit of these countries.
Alcidi and Gros (2011) examined the reasons why the peripheral eurozone members faced difficulties in adjusting the current account.They concluded that the relatively closed economies with low savings rates are boosting fiscal multipliers, which makes that fiscal consolidation negatively affect economic growth.Lower growth rates, on their part, make it difficult to adjust in the private sector.
Berger and Nitsch (2010) investigated bilateral trade balances in European countries over the period 1948-2008 and found that the current account imbalances became persistent after the introduction of the euro.These authors also came to the conclusion that countries with a relatively higher fiscal deficit and a less flexible labor and product market systematically show lower trade surpluses than other countries.
The dominant impact of the financial sector development, deregulation and particularly mass borrowings during the eurozone financial crisis on the current accounts of members was specifically highlighted by Pisani-Ferry (2013) and Gibson et al (2014).
In contrast to this view, Schnabl and Wollmershäuser (2013) argue that the effects of the real exchange rate are not particularly robust, the main role in generating the current account imbalance being attributed to fiscal policy.A similar view of the role of fiscal policy is found in Merler and Pisani-Ferry (2012).
Baxa and Olešňaník (2014) estimated that the introduction of the euro affected the expansion of the current account deficit in periphery eurozone members.These authors also found that the role of fiscal policy in the dynamics of the current account deficit was changed with the introduction of the euro, and that the effect of twin deficits occurred in many countries.This survey documented the significant role of credit growth granted to the private sector in the creation of a persistent current account deficit.
Comunale and Hessel (2014) showed that the fluctuations in domestic demand are the main driver of the current account dynamics, while the changes in the competitiveness play a minor role.Therefore, these authors propose that more attention should be paid to credit growth and macro prudential policy, in addition to the ongoing review of competitiveness and structural reforms.
Unger (2016) also focuses on the credit factor.Based on the empirical examination of the relationship between the domestic loans developments and the current account balance, this author showed that bank loans granted to the non-financial private sector are a significant determinant of the current account.

Data and Research Methodology
The empirical analysis of the current account dynamics of periphery eurozone members in this paper is based on the annual time data series for five countries (countries listed in the introduction).The main sources of data are the World Bank's World Development Indicators, and the IMF's World Economic Outlook Database (WEO).Data are normalized to the GDP.The current account to GDP ratio is the dependent variable in the model.The regressors are selected according to the reference literature suggestions.

Data Description
The explanatory variables in the regression model are: Current account persistence (CA)(-1).There are numerous evidence in the reference literature of the CA persistence.Therefore, we will include the current account with the lag, the mark (CA)(-1), as a percentage of GDP (dependent variable with a lag) as an independent variable in the analysis.If countries have a CA deficit for a long time, it will be more difficult for them to implement its adjustment.
Fiscal balance (FIS).This variable represents the budget deficit/surplus as a % of GDP.The fiscal balance affects the aggregate demand, and therefore the demand for imported products.The fiscal deficit increases the import demand, thereby adversely affecting the current account balance.The budget deficit, through higher interest rates, attracts additional foreign capital needed for its financing, which leads to the appreciation of the domestic currency.On the other hand, the rise in indebtedness in the current period increases the burden of debt servicing in the future.This burden can lead to an increase in taxation rates in the future.This means that the interest rates thereby stimulate the growth of private savings instead of stimulating investment growth.
Foreign direct investment (FDI).This variable includes the net inflow of foreign direct investment as a percentage of GDP.If foreign funds are used for investing and raising productivity in the sector of tradable goods, the increase in net inflows can positively affect the CA.If CA deficit is financed by an increased capital inflow, this can lead to an appreciation of the national currency and a reduction in export motivation, which negatively affects the current account balance.The worsening of the CA balance can occur in the case of green field investments, as they can encourage imports.In the analysis we use the FDI series with a time lag.
The real effective exchange rate (REER).Data are for REER 2010 = 100.If the index of foreign exchange rate is above 100, this means that the currency has appreciated, leading to a deterioration in the country's export competitiveness, and thus deterioration in its current account.In our analysis, the REER is an independent variable in the logarithmic form with a lag, for the purpose of avoiding the problem of endogeneity.In front of the regression coefficient with this variable, a negative sign is expected, because the appreciation destimulates exports and increases import, leading to a deterioration in the CA.
Saving investment gap (SIG) is obtained as a difference in savings and investments as a % of GDP.If savings are greater than investments, this has a positive effect on CA.If savings are lower than investments, such a situation causes an increase in the CA deficit.In our sample of periphery eurozone countries, Ireland is the only country in which domestic savings exceeded investments throughout the observed period, while in other countries a variable balance is recorded.Higher domestic savings rates reduce the current domestic consumption, which positively affects CA, and leads to a country's lower external indebtedness.
Starting level of net foreign assets (NFA).NFA is observed in relation to GDP (both variables are denominated in domestic currency).A larger amount of NFA gives the country the opportunity to import more and thus create a current account deficit.If this is realized in practice, the ratio between NFA and CA is negative.This kind of development occurs mainly in the initial state of the NFA.In time, income grows on the basis of the NFAs, which causes lower imports, resulting in the NFA acting positively on the current account.According to reference literature, the last effect is considered to outweigh the initial one in an open economy, which means that in such a case a positive relationship between NFA and CA is expected.In the analysis, NFA is taken with a lag of one period, in order to avoid the problem of endogeneity with CA.
Development of the financial system.Domestic credit (DC) is used as a proxy for this phenomenon.To measure the financial depth in this paper we will use the ratio of loans granted to the private sector and GDP (DC as a % GDP).The development of the financial system in reference literature is believed to encourage mediation, which can lead to a deterioration of the current account (Mendoza et al, 2009).A negative sign is expected in front of this variable.

Metodologija istraživanja
U ovom radu se testira da li izabrane objašnjavajuće makroekonomske varijable utiču na saldo tekućeg računa perifernih članica EU.U istraživanju je primenjen regresioni panel model.Ocenjeni su koeficijenti sledeće regresione jednačine: Dummy (DUMMY).The crisis years for all the countries surveyed cover the period from 2009 to 2017.Since all peripheral eurozone members had to follow the current account adjustment path, this artificial variable aims to show that countries were affected by the financial crisis.In terms of connection with CA, a positive sign is expected in front of this variable.
Private remittances (REM).This explanatory variable includes the received remittances as a % of GDP.The received remittances play an important role in the balance of payments of the periphery eurozone members.
Regulatory quality (REGULAT).This variable shows the government's ability to formulate and implement sound policies and regulations that should improve the position of the private sector.This variable is in this paper used to check the robustness of the results.The data for 2001 are not available, while the data for other years are available.It is considered that these variables basically influence the current account positively, as they represent an assessment of the quality of institutions.The estimates for this variable range from -2.5 (low performance) to 2.5 (strong management performance).Higher quality institutions increase investment returns, which leads to a worsening of the current account.Therefore, a positive sign is expected in front of this variable.
Business cycle.As an approximation of the business cycle, the output gap (GAP) is used in this paper.A positive business cycle results in higher income, which allows higher consumption, and hence greater imports of goods and services.In this way, there is a negative impact on the current account.This indicator is under the influence of the financial crisis.This variable is in this paper calculated as the current GDP minus the potential GDP as a percentage of the potential GDP.It is expected that in this regression there will be a negative sign in front of this variable.

Data Sources
The analysis in this paper considers a sample of periphery eurozone members.The regression analysis uses a panel timing series for five countries.The period from 2000 to 2017 is considered.The series has 90 observations.In some series, the data are missing for some years.The data are from the following sources: Objašnjavajuće varijable su izabrane prema nalazima iz literature.Pre nego što pristupimo ocenjivanju regresione jednačine, testiraćemo panel vremenske serije na postojanje jediničnog korena.

Research Methodology
This paper tests whether the selected explanatory macroeconomic variables affect the current account balance of the periphery member states.The study applied the panel regression model.The coefficients of the following regression equation are estimated: where the coefficients are defined in the above text.In order to check the robustness of the obtained results, an extended regression equation is estimated: Explanatory variables are selected according to the literature findings.Before we approach the estimation of the regression equation, we will test the time series for the existence of a unit root.

Panel Unit Root Tests
In order to check the stationarity of the panel time series, we will perform tests to show if there is a single root in the series.We will apply the following tests:

Levin-Lin-Chu (LLC) Test
This is one of the first panel unit root tests.This test was formulated by Levin, Lin and Chu (2002).The LLC test is based on an estimate of the following equation (the basic ADF specification) (this and the following equations in this section are listed according to EViews 8 User Guide II): where i = 1, 2, ..., N cross-section units, observed in t = 1, 2, ..., T, periods, is an exogenous variable that represents either a constant or a constant and trend.
The model assumes that α = ρ -1, where ρ i are autoregressive coefficients, and ρ i is identical for all observation units.Null and alternative hypotheses in this test can be shown as H 10 : α = 0 (there is a unit root for each individual component), or H 1A : α < 0 (there is no unit root, or each individual component is stationary).The LLC method requires the specification of the number of lags in each cross-section in the extended Dickey-Fuller (ADF) regression (Dickey and Fuller, 1979), as well as the appropriate kernel choice.Exogenous variables must also be specified during testing.

Im, Pesaran, and Shin (IPS) Test
The Im, Pesaran, and Shin (2003) test allows for an individual unit root process.The same applies to the Fisher-ADF and Fisher-PP test, so that ρ i can vary across cross-sections.This test starts from the specification of separate ADF regressions for each cross section.The starting specification is: Null hypothesis is H 20 : α i = 0, for each i, And the alternative hypothesis is: The IPS test in the framework of the null hypothesis assumes that all series are nonstationary, that is, that part of the series is stationary within an alternative hypothesis.Accordingly, this test differs from the LLC test, where under the alternative hypothesis it is assumed that all series are stationary.The IPS test statistics requires a specification of the number of lags and a specification of the deterministic component for each cross section of the ADF equation.An individual constant, or an individual constant and trend can be included.

Fisher-ADF and Fisher-PP Test
An alternative approach to the panel unit root test formulates a test that combines p-values from single root tests to test the panel unit root.This Fisher test was formulated independently by Maddala and Wu (1999) and Choi (2001).
The formula for this test is: -2 (5) The test is asymptotically distributed as chisquared with 2N degrees of freedom.
Choi showed that: where φ -1 is an inverse function of the standard normal cumulative distribution.The null and alternative hypotheses are the same as for the IPS test.The advantage of this test is that it can include unbalanced panels.
H 2 o implies that all panel time series have a unit root, and that they are non-stationary, while the alternative assumes that at least one series is stationary.That is why Fischer's test allows some series to be stationary while others are not.

Cointegration Tests
The two types of the panel cointegration are distinguished in the literature -homogeneous and heterogeneous.The variables in the model are homogeneously cointegrated if the values of the cointegration parameters are the same for each individual panel unit.Heterogeneous cointegration implies that the values of cointegration parameters are different for each individual unit.Among the cointegration tests with the assumption of independence of the panel units Kao and Pedroni test stand out, which are used to test the null hypothesis of the unstationarity of the residual of the estimated cointegration equation.We will provide a brief overview of both these tests.

Kao Cointegration Tests
Kao (1999) test follows the basic approach like the Pedroni test, but assuming that the individual units are homogeneous.In the bivariate case, the test is shown as (EViews 8 Users Guide II, pp.865-866), y it =α i +βx it +e it (9) for y it =y it-1 +u it (10) x it =x it-1 +ε it (11) for t = 1, …, T; i = 1, …, N.
Kao defined five variants of the Dickey-Fuller residual test (DFR).All five test-statistics in asymptotic conditions have a normal distribution with parameters 0 and 1.
To estimate the panel cointegration equation we will use FMOLS (Fully Modified OLS) and DOLS (Dynamic OLS) estimators (see Phillips and Hansen, 1990).

Empirical Results and Discussion
Table A in the Appendix provides estimates of the applied unit root tests in the panel for selected macroeconomic variables that affect the current account imbalance in a sample of five peripheral eurozone members.In order to check the degree of integration of the selected variables, several unit root tests (LLC, IPS, ADF-Fisher Hi squared, PP-Fisher Hi squared, Hadri test) were used.A large number of tests have been applied to avoid their weaknesses that are manifested in cases of shorter time series.All variables refer to the period 2000-2017.The test equation for all tests includes individual effects or an option that includes individual effects along with an individual trend.Tests were applied to the level of occurrence and the first difference of all variables.The obtained results are given in Table A in the annex.The stationarity for the qualitative variables and artificial variables used in the model were not checked in this study.
The data in Table A in the Appendix show presence of a unit root for most unit root tests at the level of the series, with significance of 5%.An additional check at the first differential level showed that all series are stationary, that is, the integrated order 1 ((Xt~I(1)).The exception is the variable that represents the real effective exchange rate (REER) whose stationarity of the first differential has not been confirmed by the Hadri test, as well as variable credits granted to the private sector (DC) and the output gap (GAP), where several tests did not confirm the first diference stability.Since the preconditions were met, we applied the Pedroni and Kao test to check the existence of cointegration between the current account (CA) variable and other macroeconomic variables.The null hypothesis according to which the residuals of the estimated cointegration equation are non-stationary, or the residuals obtained from the regression equation possess a unit root, is examined in both tests.
Kovačević R. Tekući račun perifernih članica Evropske unije -panel analiza Bankarstvo, 2018, vol.47, br. 4 cointegration in the panel (a unique equilibrium for all countries), while the second variant is based on the assumption that there is heterogeneous cointegration in the panel (different equilibrium across countries).Each variant of the test has two options regarding the inclusion of deterministic components in the cointegration equation: the first option implies that only an individual component is involved, while the other one includes individual effects and a trend.Table 1 shows the results of 11 Pedronian statistics, divided into two groups.The first group includes three tests, which assume heterogeneous cointegration in the panel within the alternative hypothesis.Other tests assume homogeneous cointegration in the panel.Based on 6 out of 11 tests in the option involving individual fixed effects and trends, the null hypothesis of no cointegration is rejected at a significance level of 1%.The conclusion is that, according to this test, there is a long-term compliance in the developments of the observed panel time series for a set of peripheral eurozone members.
U cilju dodatne provere robustnosti dobijenih koeficijenata, ocenjena je kointegraciona jednačina uključivanjem novih varijabli (DUMMY, DC i REGULAT).U tabeli 4. se daje uporedni prikaz ocenjenih koeficijenta proširene kointegracione jednačine, pri čemu se FMOLS i DOLS koriste kao estimatori.significant at the level of 1% (coefficients for net foreign assets and lag of one period, and the logarithmic value of foreign direct investment, also with a lag of one period, are statistically significant at the level of 5%).In the period after the outbreak of the global financial crisis (2008-2017), all the coefficients estimated with the FMOLS estimator are statistically significant at the level of 1%, and in some cases, the signs are different in relation to the entire observed period (the signs with foreign direct investment and the fiscal balance are positive and show that these variables have contributed to reducing the current account deficit).The estimated coefficient in respect of the savings-investment variable is positive and statistically significant at the level of 1%, both throughout the sample period and in the observed subperiod.The growth of savings and slower investments in the periphery members of the euro area contributed to a reduction in the current account deficit.
After the estimation of the cointegration equation, the obtained residues were tested.The same tests were used as when checking the existence of a unit root in the panel time series.All tests confirmed the stationarity of residuals.Tests that have the option to exclude the constants (LLC, ADF Fisher and PP Fisher) have shown stationarity of residuals at the significance level of 1%, when the constant is turned off (it is methodologically corect to exclude the constant when determining residual stationarity).Based on the obtained results of these tests, it is concluded that the macroeconomic variables covered by this analysis are co-integrated.It follows that the selected explanatory variables clearly explain the average balance of the current account of the peripheral eurozone members.
The estimated time interval 2008-2017 can be used to check the robustness of the obtained estimates for the sample period 2000-2017.Although there are slight differences in terms of the statistical significance of the estimated coefficient with the application of the DOLS estimator, the obtained signs are similar for both panel series.It is characteristic that the estimated coefficients with all variables with the FMOLS estimator were statistically significant at the level of 1%, and that all of them had the expected sign.The coefficient in respect of foreign direct investment for the sample period was not statistically significant with the use of the FMOLS estimator, but with the application of the DOLS estimator this coefficient was statistically significant, with a negative sign.
In the subperiod 2008-2017, this indicator was statistically significant for the estimated values with the FMOLS estimator, while the significance according to the DOLS estimator was missing.Estimated coefficients for the real effective exchange rate were statistically significant for the whole period and in the observed subperiod, with the application of the FMOLS estimator.The negative sign of this coefficient shows that the real effective exchange rate appreciation in the periphery eurozone members diminished their competitiveness, with a negative effect on the current balance.
Table 3 shows the results of applying different estimators of regular AR(2) regression of the CA balance (autoregression model -AR).
Based on them, it can be concluded that there are persistent CA accounts in the peripheral eurozone members.The data in Table 3 are PLS, FMOLS, DOLS and RLS estimates of the cointegration equation for a current account with a lag.The coefficients with one lag are statistically significant at the level of 1%, while coefficients with two lags are not statistically significant for PLS, FMOLS and DOLS estimators.Therefore, the data at the level of one lag confirm the persistence of the current account deficit.
Column 1 of Table 4 features the estimates of the cointegration vector obtained by the FMOLS and DOLS estimators.Several additional variables (artificial variable DUMMY, domestic credits to the private sector -DC, and artificial variable REGULAT) are included in the cointegration equation, and their estimates are given in columns 2, 3, and 4.These variables are included to check the suitability of the selected model.The coefficients in column 2, which are estimated with the FMOLS estimator, are statistically significant in the range of 1% to 10% (the exception is the coefficient in respect of remittances which is not statistically significant).However, several coefficients (fiscal balance, remittances and net foreign assets) were not significant with the application of DOLS estimators.After the introduction of the variable domestic credits approved to the private sector -DC, all the coefficients estimated with the FMOLS estimator are statistically significant, apart from the coefficient in respect of the net foreign assets variable.
The It is interesting that the fiscal deficit coefficient has a negative sign, which indicates that the reduction of the fiscal deficit in the peripheral members of the eurozone in the period after 2008 did not affect the reduction of the current account deficit.The estimated coefficients for a variable that represents loans to the private sector (the DC variable) are statistically significant with the application of the FMOLS estimator.Checking the robustness of the applied model has confirmed its adequacy.

Conclusion
The paper estimates the influence of the selected factors on the creation of the current account deficit in the peripheral eurozone members.There are five countries in the sample (Italy, Spain, Portugal, Ireland and Greece).The observed time period is 2000-2017.The aim of the research is to provide the robust estimates of the long-term relationship between the current account balance and the explanatory variables.Panel analysis was applied.Since the panel time series of the selected macroeconomic variables were not stationary, several tests (Kao, 1999;Pedroni, 1999) were applied with the task to determine whether there is a cointegration of the starting panel time series.The tests confirmed the existence of cointegration.The equation was then estimated using the FMOLS and DOLS estimators.On the basis of the results obtained, it can be concluded that the peripheral eurozone members are characterized by the current account persistence.Foreign direct investment plays an important role in covering this deficit.Thanks to the FDI inflow, the countries under review maintain a satisfactory export dynamics.The increase in the rate of foreign direct investment leads to a rising deficit in the primary income of balance of payments account, causing additional pressure on the current account.The negative sign in front of the coefficient on the foreign assets shows that sudden turnarounds in capital flows can endanger the sustainability of the current account deficit of these countries.A sudden breakdown of net capital inflows can lead to an increase in public debt that should compensate for a shortfall in private capital inflows.The estimated coefficients on the exchange rate variable confirm that a real effective exchange rate appreciation contributes to an increase in the current account deficit of the periphery countries sample.The loss of competitiveness on this basis in the long-term adversely affects the current account and requires structural interventions in the economy.The internal devaluation, through a reduction in wages, could enable a reallocation of resources from non-tradable to tradable sectors, thus strengthening the export competitiveness of these countries.This particularly applies to peripheral members whose unit labor costs are above the identical costs of its main trading partners.At the same time, these costs are a significant factor in imports.The limitation of this paper's analysis are the short time horizons, which diminish the quality of econometric estimates.
N(0, 1) distributed.The Hadri panel unit root test is similar to the KPSS (Kwiatkowsky, et al, 1992) unit root test.The null hypothesis is H 0 : no unit root in any of the series in the panel.Like the KPSS test, the Hadri test is based on residuals from individual OLS regressions of y it on a constant, or a constant and trend.If both the constant and trend are included, the following relationship is estimated: Tekući račun perifernih članica Evropske unije -panel analiza Bankarstvo, 2018, vol.47, br. 4 Tekući račun perifernih članica Evropske unije -panel analiza Bankarstvo, 2018, vol.47, br. 4 Za proveru robustnosti zaključaka izvedenih na osnovu Perdoni testa, koristili smo Kao i Fisher test.Kao test statistika nije pružila mogućnost da se odbaci nulta hipoteza o odsustvu kointegracije.Pošto vrednosti test statistike zavise od veličine uzorka, naš panel sa godišnjim podacima ima kraću vremensku dimenziju, što smanjuje pouzdanost zaključivanja o stacionarnosti serije.Sem toga, kriza iz 2008.godine je izazvala poremećaje u trendovima posmatranih varijabli.Za razliku od Kao testa, rezultati Johansen-Fisher(Johansen, 1991) testa kointegracije podržavaju postojanje kointegracione veze između posmatranih varijabli, kako prema test statistici traga, tako i prema testu maksimalne verodostojnosti.Ako se varira broj endogenih promenljivih kao i broj docnji prve diference ovih varijabli, varira i broj kointegracionih vektora od 1 do 2, koje ovaj test preporučuje na osnovu statistike traga.Pošto je na osnovu većine testova ustanovljeno postojanje kointegracije, potrebno je da ocenimo koeficijente kointegracione jednačine, odnosno da ocenimo dugoročnu vezu između analiziranih makroekonomskih varijabli pomoću FMOLS and DOLS modela.Ocenjeni koeficijenti daju se u tabeli 2. Tabela 1. Pedronijev test kointegracije u panelu perifernih članica evrozone derived from the Perdoni test, we used Kao and Fisher tests.The Kao test did not provide an opportunity to reject the null hypothesis of no cointegration.Since the test statistic values depend on the size of the sample, our annual panel data have a shorter time dimension, which reduces the reliability of conclusions about the time series stationarity.In addition, the 2008 crisis caused disruptions in the trends of the observed variables.In contrast to the Kao test, Johansen-Fisher(Johansen, 1991) results of the cointegration test support the existence of a co-integration relationship between the observed variables, both according to the trace test, and the maximum eigenvalue test.If the number of endogenous variables varies as well as the number lags of the first differential of these variables, the number of cointegration vectors also varies from 1 to 2, which this test recommends based on the trace statistics.Since the existence of cointegration has been established by the most tests, it is necessary to estimate the coefficients of the cointegration equation, or to estimate the long-term relationship between the analyzed Tabele 1. Pedroni cointegration test for a panel of periphery eurozone countries

Tabela 2 .
Determinante tekućeg računa perifernih članica evrozone -rezultati FMOLS and DOLS modela, ceo period 2000-2017.i potperiod 2008-2017.godine Tekući račun perifernih članica Evropske unije -panel analizaBankarstvo, 2018, vol.47, br. 4 macroeconomic variables using the FMOLS and DOLS models.The estimated coefficients are given in Table2.The specification of the cointegration equation is with fixed effects.The panel time series used the annual data.The cointegration model was estimated for the entire sample period (2000-2017), as well as for the subperiod (2008-2017).The FMOLS model with constant parameters is estimated, using only the constant as a specific trend regresor in the cross section data.For estimating the cointegration equation by means of a dynamic model of the ordinary last squares (DOLS), a constant is used as a deterministic component.The results in

Table 2
show that the observed determinants have a long-term impact on the current account of peripheral eurozone members (the GAP variable represents the impact of the business cycle).The estimated coefficients with FMOLS estimators over the entire period 2000-2017 are statistically significant at the level of 1% for all regressors, except for the remittances