The Effect of Global Financial Crisis on Trade Elasticities: Evidence from BRIICS Countries and Turkey

Transcription

1 The Effect of Global Financial Crisis on Trade Elasticities: Evidence from BRIICS Countries and Turkey Natalya Ketenci Yeditee University, Istanbul Abstract: The effect of the global financial crisis on the international trade atterns of develoed countries has been one of the main focuses of recent studies. However, the deendence level of world trade on emerging markets increases every day. Therefore, it is imortant to study the level of the negative effect of the crisis on emerging economies and the level of their recovery otential. This aer emirically studies the effects of the financial crisis on changes in the trade elasticities of BRIICS (Brazil, Russia, India, Indonesia, China and South Africa) countries and Turkey. The imerfect substitute model (Goldstein and Khan 985) for the exort and imort demand functions is used. The autoregressive distributed lag (ARDL) aroach to cointegration is alied to test the cointegration relationshis between exorts and imorts and their determinants and in order to estimate the exort and imort elasticities in the countries under examination. The emirical results rovide enough evidence to conclude that changes in the exchange rate did not lay significant role in exort and imort demand functions before the global financial crisis and after. However, foreign and domestic incomes are found highly significant and elastic in exort and imort demand functions, resectively. It is found as well that the global financial crisis had increasing effect on exort and imort resonsiveness to foreign and domestic incomes resectively, excet for Turkey and Brazil in the exort demand function and South Africa in the imort demand function. Keywords: financial markets; international trade; emerging markets. JEL Classification Codes: F4, F4. Introduction BRIC (Brazil, Russia, India and China) is a grou of countries that are considered to be the biggest emerging economies with the highest growth rates. Due to their fast growth, it is believed that these countries may be among the most dominant countries in the world by 2050 (Goldman Sachs 2007). Indonesia and South Africa (BRIICS) were added to this grou by the Organization for Economic Co-oeration and Develoment (OECD) due to Indonesia s high level of oulation growth among middle income countries in South-East Asia, and due to South Africa s highest level of develoment comared to other African countries. Figure and Figure 2 show trade atterns in the considered countries. All estimated countries have had tendencies of continuous growth in trade esecially since 2000 with the extreme case of China. At the same time, it can be seen that all of the estimated countries have had shar declines in exorts as well as in imorts in 2009 with the following recovering in 200. The develoment of the considered emerging countries was characterised by unsteady growth of GDP in the 990s, and by significant declines in the cases of Russia and Indonesia. All Natalya Ketenci, Deartment of Economics, Yeditee University, Kayisdagi, 34755, Istanbul, Turkey. Tel: Fax: nketenci@yeditee.edu.tr

2 2 BRIICS countries have followed accelerating ositive growth since 2000, with the excetion of Turkey, which had a decline in its real GDP in 200 with subsequent recovery. However, it can be seen from Figure 3 that the growth of all BRIICS countries significantly slowed down in 2009, being affected by the global financial crisis with the extreme case of Russia, where a decline in real GDP was observed. In terms of the growth of real GDP, the countries that were least affected by the global financial crisis were China and Indonesia, while the country that was the most affected was India. After substantial slowdowns, all economies had substantial recovery in the following year, 200. A great deal of attention in the literature is sent on the contagion effect of the global crisis on the financial markets of emerging economies. Aloui et al. (20) in their aer on the effect of the global financial crisis in BRIC countries emloyed a multivariate coula aroach. They demonstrated that the financial markets of countries that are highly deendent on commodity rices, Brazil and Russia, are more heavily deendent on the United States comared to such countries as China and India, which are more deendent on the exort rices of finished roducts. Dooley and Hutchison (2009) in their study using the decouling-recouling hyothesis evaluated the transmission of the U.S. crisis to emerging markets including the BRIICS and Turkey examles, excet for India and Indonesia. They found that the equity markets of emerging economies aeared to have been isolated from the U.S. financial markets for the eriod starting from the date when the first signals of the crisis aeared in the U.S. until the summer of However, starting from the summer of 2008, the financial markets in emerging economies were found to be highly correlated to the deteriorating economic conditions of the U.S. Thus studies on the financial transmission of the crisis rovide evidence of the moderate resonsiveness of emerging financial markets to the signals of the crisis in the United States. However, due to the short time san not enough studies have been comleted on change in trade tendencies in resonse to the global crisis in the world, including emerging markets. For examle, McKibbin and Stoeckel (2009) studied the otential imact of the financial global crisis on the world in 5 countries and regions including develoed as well as develoing countries by modelling the crisis as a combination of shocks to a set of changes in an economy. They found that financial crisis caused trade rotectionism in terms of increased tariffs and suort for domestic industries, which can lead to the deterioration of the domestic and trade artners GDPs. At the same time, the authors found that financial rotectionism emerged as well, enforcing the decline in international trade flows. Chor and Manova (200) in their study showed how credit conditions during the global financial crisis affected world trade flow. They found that high interbank rates and tight credit conditions were imortant channels of the transmission of financial crisis on trade flows. This study seeks to clarify emirically the consequences of the global financial crisis in the trade sector of major develoing countries. The focus of this study is on the trade atterns of the develoing countries BRIICS and Turkey. This study estimates the effect of the financial crisis by measuring trade elasticities in exort and imort demand functions for two different eriods on a quarterly basis, 989Q-2007Q2 and 989Q-200Q4. It is known that first signs of the financial crisis took lace in August 2007 in the U.S., followed by a global contagion effect that emerged in the second half of 2008 in many countries. To measure the trade elasticities of the develoing countries two eriods were chosen, the re-crisis eriod and the full eriod including the global financial crisis and its contagion effect, to be able to cature the changes in trade elasticities that may have haened before and after the contagion effect started. The financial crisis that sread in the second half of 2008 generally may be defined as

3 3 a decline in foreign investments, changes in foreign debt servicing burdens, a reduction in trade credits and a global decline in total exenditures. The aer is structured as follows. The next section highlights the main features of the re- and ost-crisis trade atterns of BRIICS countries and Turkey. Section 3 exlains the methodology, alied exort and imort demand functions and outlines the testing strategy. Section 4 resents and discusses the main emirical results. Finally, Section 5 gives the concluding remarks for this study. 2. Methodology To examine to what extent movements in the balance of trade are exlained by change in relative rices, income and exchange rate the imerfect substitute model (Goldstein and Khan 985) was emloyed for the exort and imort demand functions, where it is assumed that foreign and domestic roducts are imerfect substitutes. Xit = f(pxit,pt*,yt*) () Where t denotes the time eriod of estimation, Xit is the total exort of ith country, Pxit is the exort rice of ith country in the national currency, Pt* denotes the foreign rice deflator in the national currency of the estimated country, and Yt* is foreign real GDP exressed in the national currency of the estimated country. The total exort in the equation can be measured as total nominal exorts deflated by exort rice index. However there is the lack of data on exort rice index on bilateral basis. Therefore, as an alternative, exort values (or inayments) are used to determine the currency and income changes. If we divide the right-hand side of equation () by foreign rices Pt*, due to the linearity of demand functions the exort demand is not going to change (Goldstein and Khan 985). Therefore, the logarithmic form of the exort demand function may be exressed in the following form: LnXit = c0 + c Ln(Pxit/Pt*) + c2 Ln(Yt*) + εt (2) Where LnXit is the natural logarithm of the total exort value of ith country, Ln(Pxit/Pt*) is the natural log of relative exort rices of the estimated country relatively to foreign country and Ln(Yt*) is the natural logarithm of the foreign income. Finally εt is the error term. Due to the difficulty in obtaining the imort and exort rices of the estimated countries, equation 2 has to be modified. The modified aroach used in the literature is to secify relations between exort and imort values and the real exchange rate. Studies such as those by Bahmani-Oskooee and Economidou (2005), Bahmani-Oskooee and Ratha (2008), Irandoust et al. (2006), Hsing (2008), Kwack et al. (2007), Kumar (2008), Bahmani-Oskooee et al. (203), Huchet-Bourdon and Bahmani-Oskooee (203) and others used real exchange rates in their studies to calculate the exchange rate elasticity. Therefore, the alternative log-linear form of the exort demand function can be written as follows: LnXit = α0 + α Ln(Et) + α2 Ln(Yt*) + εt (3)

4 4 Where Et is the real exchange rate calculated by the following formula: where ER is the nominal exchange rate reresented in foreign currency er unit of domestic currency. As a roxy for domestic and foreign rices a GDP deflator is used (for similar studies, see Irandoust et al. [2006] and Kwack et al. [2007]). Yt* is the real GDP of the foreign trade artner. For every estimated country, a set of nine countries is chosen as a reresentative of the foreign trade artner. Countries in every set are selected according to the highest time-varying bilateral trade shares between the estimated country and its trade artners. 2 It is exected that the coefficient of relative exort rice α in equation 3 being negatively related to exort value as an increase in domestic rices will decrease the demand for exort while foreign rice increase will raise the demand for exort. Income elasticity α2 may get different signs. It will get a ositive sign if an increase in the foreign income raises demand for home country exort. However, if foreign goods and services are highly cometitive with home country exort foreign income in this case can have negative effect on the exort value from the home country. The standard form of the imort demand function can be exressed by the following equation: Mit = f(pmit,pt,yt) (4) Where Mit is the imort of ith country, Pmit is the imort rice of ith country in the national currency, Pt denotes domestic rice deflator and Yt is the domestic real GDP. There is a lack of data on imort rice index on bilateral basis, similar to exort demand equation. Therefore, imort values (or outayments) are used to determine the currency and income changes in equation 4. Following the extraction of exort demand function the right-hand side of equation (4) can be divided by domestic rices Pt. As a result, the imort demand function is taking the following form: LnMit = γ0 + γ Ln(Pmit/Pt) + γ 2 Ln(Yt) + ut (5) Where LnMit is the natural logarithm of the total imort value for ith country, Ln(Pmit/Pt) is the natural logarithm of relative imort rices, Ln(Yt) is the natural logarithm of the domestic income. Finally ut is the error term. The log-linear form of the imort demand function corrected for imort rices will take the following form: LnMit = β0 + β Ln(Et) + β2 Ln(Yt) + ut (6) 2 The following countries were selected as roxy for foreign trade artner: Turkey Germany, China, Russia, United States, Italy, France, Sain, the United Kingdom, the Netherlands; Brazil the United States, China, Argentina, Germany, Jaan, Italy, France, the United Kingdom, the Netherlands; Russia China, Germany, the United States, France, Italy, Jaan, the United Kingdom, the Netherlands, Turkey; India China, the United Arab Emirates, the United States, Australia, Germany, Switzerland, Korea, Jaan, the United Kingdom; Indonesia Singaore, China, Jaan, United States, Malaysia, Korea, Thailand, Australia, Germany; China Jaan, Korea, the United States, Hong Kong, Germany, Australia, Malaysia, Russia, Thailand: South Africa China, Germany, the United States, Jaan, the United Kingdom, India, France, Italy, Netherlands.

5 5 Where Et is the real exchange rate calculated by the following formula: where ER is the nominal exchange rate reresented in domestic currency er foreign currency. Y is the domestic outut. It is assumed that the relative imort rices coefficient β will be related negatively to the imort quantity as according to the demand theory increase in the imort rice will reduce the imort demand while increase in domestic rices will raise demand for imort. However, income elasticity β2 can have different signs as in the case of the exort demand function. If there are no alternatives for imorted goods in the domestic roduction, income will have a ositive effect on the imort volume. However, if there are a lot of imort substitutes in the domestic roduction, an increase in the domestic income can lead to a decrease in the imort demand. The focus of the analysis is to study the long-run relationshi and dynamic interactions among the variables in the exort and imort demand functions. To incororate the short-run dynamics, the autoregressive distributed lag (ARDL) aroach to cointegration is alied. The ARDL aroach involves two stes for estimating the long-run relationshi (Pesaran et al. 200). The first ste is to examine the existence of long-run relationshi among all variables in an equation and the second ste is to estimate the long-run and short-run coefficients of the same equation. The second ste determines the aroriate lag lengths for the indeendent variables and is alied only in the case if cointegration relationshis are found in the first ste. In errorcorrection models, the long-run multiliers and short-run dynamic coefficients imrove the exort demand function as follows: log X t 0 log X ti 2log Et i 3log Y * i i0 i0 ti log. (7) X t 2 log Et 3 log Y * t t The error correction model for the imort demand function is as follows: log M t 0 log M ti 2 log Et i 3 log Yt i i0 i0 i log M t 2 log Et 3 log Yt ut. (8) Equations (7) and (8) may be transformed to following equations in order to accommodate the one lagged error correction term: log X t 0 log X ti 2log Et i 3log Y * i i0 i0 ti

6 6 EC t t (9) log M t 0 log M ti 2 log Et i 3 log Yt i i0 i0 i EC t u t (0) The ARDL aroach is used to establish whether the deendent and indeendent variables in each model are cointegrated. The null of no cointegration H 0 in the exort 0 : 2 3 H : : 2 3 demand model is tested against the alternative hyothesis of 0. In the imort demand function the null of no cointegration H 0 is tested against the alternative hyothesis of H 0. : 2 3 The Walt-tye (F-test) coefficient restriction test is conducted, which entails testing the above null hyotheses H 0 and H. Pesaran et al. (200) comuted two sets of asymtotic critical values for testing cointegration relationshis existence. The first set assumes variables to be I(0), the lower bound critical value (LCB) and the other I(), uer bound critical value (UCB). If the F-statistic is above the UCB, the null hyothesis of no cointegration can be rejected irresective of the orders of integration for the time series. Conversely, if the test falls below the LCB, the null hyothesis cannot be rejected. Finally, if the statistic falls between these two sets of critical values, the result is inconclusive. Since the results of the F-test are sensitive to lag lengths, we aly various lag lengths in the model. However, as Pesaran and Pesaran (997, 305) argue that variables in regression that are in first differences are of no direct interest to the bounds cointegration test. Thus, a result that suorts cointegration at least at one lag structure rovides evidence for the existence of a longrun relationshi. Alternatively, Kremers et al. (992) and Banerjee et al. (998) have demonstrated that in an ECM, significant lagged error-correction term is a relatively more efficient way of establishing cointegration. So, the error correction term can be used when the F-test is inconclusive. 3. Emirical Results 3. Cointegration Test In order to ascertain whether the tested variables are stationary, the ARDL cointegration test was emloyed. Based on the cointegration test results reresented in Table, the strong evidence of the cointegrating relationshi was found in exort demand functions in all countries excet India and South Africa. On the other hand, weak evidence for cointegration was found for the cases of Russia and Indonesia with a 0% significance level. Testing Imort demand functions, the existence of cointegration can be confirmed with and 5% significance levels in all cases excet Brazil, where cointegration was confirmed with a 0% significance level, while

7 7 in the case of China the hyothesis of no cointegration was acceted. Therefore, continuing with further estimations, India and South Africa in exort demand function and China in imort demand function cannot be included. 3.2 Cointegration Coefficient Estimates The stationarity of the linear combination of a grou of non-stationary series is defined by the cointegration test. In order to find the long-run equilibrium relationshi among variables, the linear combination of the non-stationary time series has to be stationary. The long-run cointegrating coefficients are estimated by using ARDL rocedure, where the aroriate autoregressive order was chosen by using the Schwarz criterion (SC), and resented in Table 2. The coefficients α and β reresent long-run elasticities of real exchange rate for exort and imort demand functions on the basis of equations 3 and 6, resectively. The coefficients α2 reresent the long-run elasticities of foreign income for the exort demand function (equation 3), while the coefficients β2 illustrate the long run elasticities of domestic income for the imort demand function (equation 6). It is assumed in the aer that the real exchange rate coefficients of exort and imort, resectively, are related negatively to trade flows. An increase in relative foreign rices may lead to an increase in exort demand. On the other hand, an increase in exort rices leads to a decline in exort demand (see equation 2). Whereas in the case of imort demand function a raise in foreign rices leads to a decline in imort demand, while an increase in domestic rices leads to an increase in imort demand (see equation 5). The results of long-run coefficient estimations are resented in Table 2, where India and South Africa are not included due to the lack of cointegration relationshis in the exort demand function. From Table 2 it can be seen that in the exort demand function exchange rate elasticities of Turkey, Brazil, Indonesia and China roduced the exected negative sign and only in the case of Russia was the real exchange rate elasticity estimated with ositive sign for the considered eriods. In all cases of the exort, demand function exchange rate elasticities aeared to be inelastic in addition to being very close to zero. However, the majority of exchange rate estimates did not show significance, which illustrates that the real exchange rate does not influence the exort demand in the considered develoing countries in the long run. Insignificant change in the values of the exchange rate elasticities can be observed when different estimation eriods are comared. Thus in the cases of Brazil, Russia and Indonesia exchange rate elasticities almost did not show any changes in the eriod of the global financial crisis comared to the re-crisis eriod In the case of Turkey, the exchange rate elasticity of exorts declined and aeared to be significant in the eriod covering the crisis, thus illustrating the decline of the exort resonsiveness to rices. In the case of China, however, the exchange rate elasticity increased in the full eriod; nevertheless, the elasticity value is so small and insignificant that it is still illustrates the low resonsiveness of exorts to the real exchange changes in the long run. The results of the estimations are consistent with some results in the literature. For examle, in the case of Turkey, Ozkale and Karaman (2006) concluded that rice is inelastic and the sign of the real exchange rate is negative for the exort demand function for goods trade. While Aydin et al. (2004), on the other hand, found that the exchange rate is inelastic for goods but a ositive in sign in Turkey. Hossain (2009) found as well that the long-run relative rice elasticity of the demand for exorts is significantly lower than that in Indonesia. Vieira and Haddad (20) found in the case of Brazil that the trade weighted real exchange rate elasticity of manufactured exort is inelastic with exected negative sign. Algieri (2004) found

8 8 that in case of Russia the relative rices elasticity of exorts is significant and elastic with exected negative sign, contrary to the results of the resent study. However, the exorts of Russia in Algieri (2004) did not include oil, gas or its roduct. The inclusion of oil and energy roducts in exorts roduced the inelasticity of exorts to relative rices, indicating that the demand for energy roducts are inelastic to change in rices. On the other hand, the real exchange rate elasticity in the Chinese exort demand function in Cheung et al. (2009) was found with significant and highly elastic with negative sign. However, Cheung et al. (2009) in their study use the CPI-deflated exchange rate, which may be a weaker measure comare to the GDP deflator, and may roduce different results. Thus the exort estimation results show that changes in the real exchange rates do not affect exorts in the long run considering the re-crisis eriods and the eriod that saw the global financial crisis. The long-run income elasticities α2 and β2 of exort and imort, resectively, are exected to have a ositive sign demonstrating increase in exort value as a result of growing foreign incomes. Resectively, an increase in domestic incomes is exected to increase the demand for imorts, giving ositive sign to elasticity. Estimations of the exort demand function rovide enough evidence to assume a ositive relationshi between income and exort demand in all of the considered countries with high significance levels in the majority of cases. In the cases of Turkey and Brazil, the long-run income elasticities of exort demand function are elastic and significant with ositive sign. The results illustrate that the income elasticities are higher in recrisis eriods than in the eriod that exerienced the financial crisis. Thus it can be concluded that the general trend of high exort resonsiveness to income slightly declined as a result of the global financial crisis. However, in the cases of Russia, Indonesia and China, the long-run exort resonsiveness to foreign incomes increased in the eriod which exerienced the financial crisis with a high significance level only in the case of Indonesia. The statistical data show that Indonesia was one of the first to recover from the global crisis countries out of the considered countries. Indonesia has the highest growth rate of exorts value in 200 comare to If in 2009, all of the considered countries had significant declines in exort trade, in 200 the exorts values of Turkey and Russia were lower comare to 2008, while in Indonesia the exorts value were 5% higher than in 200. In second and third lace were India and China, where the growth rate was and 0%, resectively. The results of the estimations are consistent with those of the literature. For examle, Algieri (2004) found that the world income long-run elasticity of exorts is elastic in the case of Russia. Hossain (2009) in its study found evidence that long-run income elasticity for Indonesia s exorts is significantly greater than one, which is consistent with the resent study. These results are similar to the outcomes of Cheung et al. (2009) that roduce high and statistically significant income elasticity of exorts. The results illustrate that growing incomes of trading artners roortionally increase exort demands for Russian, Indonesian and Chinese goods. Accordingly, we have enough evidence to conclude that it is rimarily the foreign income that affects the exort demand in the long run in BRIICS countries and Turkey. It is found that while the tendency of exort resonsiveness to foreign income decreased in the cases of Turkey and Brazil in the eriod when the financial crisis is included, in the cases of Russia, Indonesia and China there was an increasing tendency in exort resonsiveness to foreign income. The trading artners of Turkey and Brazil had slight changes for imort substituted goods, while in the cases of Russia, Indonesia and China trade artners that had a tendency to increase imorts from these countries after the financial crises was included. The tendency of increased imorts may illustrate the comarative advantage of trading goods comare to local ones, while the global

9 9 crisis has a negative effect on the cometitiveness level of local roduction. However, the results illustrate that the trading artners of Turkey and Brazil refer an imort substitution olicy during crises, which significantly decreased the value of the exorts of these countries. These results are suorted by statistical data 3 demonstrating a 2% decline in exort values in 200 comared to 2008, while in Brazil exort values increased in 200 only by 2% comared to The estimations of the imort demand function do not include the case of China due to the absence of cointegration relationshis between variables. The estimates of the long-run exchange rate coefficients roduced an exected negative sign only in the case of Russia, while in all other cases the long-run exchange rate elasticity aeared to be ositive. In all of the estimated countries the long-run exchange rate elasticity was found inelastic. In the cases of Brazil, India and Indonesia, the exchange rate elasticities were found inelastic, nearly close to zero, and they were not found to be significant in the imort demand function. Estimates of the long-run exchange rate elasticities of Turkey and South Africa were found to be significant and inelastic with ositive sign. The dereciation of domestic currency leads to a slight increase in imorts indicating signs of the ossible resence of a J curve. The assumtion of existence of the J curve effect in the cases of Turkey and South Africa are verified by results obtained on the exchange rate elasticities of exorts. The dereciation of a currency making exorts cheaer to foreign buyers therefore exorts increase and imorts decrease. However, in the short run, such reasons as existing contracts, the inelasticity of exorts or imorts, the absence of alternative, do not allow exorts or imorts to change significantly. In these cases dereciation is followed by an increase in imort values and decrease in exort values. In this study, increases in imorts and decreases in exorts following dereciation in the cases of Turkey and South Africa are reflected by longrun coefficients as well, without the indication of balance of trade imrovement in the long run. However, it is imortant to note that the current study is carried out on the basis of quarterly data, where the long-run term still may be short enough to illustrate the balance of trade imrovement. Similar results are found in the literature as well. Ogus and Sohrabji (2009) found that the exchange rate has a negative effect on Turkish exorts; however, they found that the exchange rate has negative effects on imorts as well. Aydın et al. (2004) found that real dereciation will not increase exorts significantly; however, in their study they found that dereciation will decrease the volume of imorts significantly. Narayan and Narayan (2003) found relative rices of elasticity of demand in South Africa inelastic as well; however, with negative sign. The values of the long-run exchange rate elasticities in the cases of Turkey and South Africa were found to be similar in the estimated re and ost crisis eriods, roviding additional evidence of the exchange rate insignificance in the long run for the considered countries. In the case of Russia, long-run exchange rate coefficients were found significant with exected negative sign indicating that the dereciation or areciation of the Russian ruble leads to a decrease or increase in imorts, resectively. However, the inelasticity of the exchange rate indicates that changes in imorts that take lace due to the real exchange rate fluctuations are not major. On the other hand, it can be seen that the real exchange rate aeared to be more inelastic and insignificant in the eriod which covered the global financial crisis. All coefficient estimates of income for imort demand function were found to be elastic with ositive sign. In most of the estimated countries long-run income elasticities were found statistically significant. The ositive sign of income elasticity shows that with an increase in 3 OECD statistics.

10 0 income, the estimated countries have higher references for imorted goods than for domestic ones. In all of the estimated countries, excet South Africa, the values of long-run income elasticities demonstrate increase in the eriod which covered the global financial crisis. This indicates that the global crisis did not deteriorate demand for imorts in the considered develoing countries; conversely it shows an increasing tendency in demand growth for imort in resonse to growing domestic incomes. The eriod , which demonstrates an increase in long-run income elasticities, was characterised by shar declines in domestic incomes in all of the considered countries at the end of 2008 and at the beginning of 2009 (see Figure 3). Therefore, increased income elasticities may be interreted as a rising tendency in imort decline in resonse to declining domestic real incomes during the global financial crisis. Estimates of the long-run income elasticities of South Africa reveal a decline in the eriod covering the crisis indicating a slight decline in the imort demand resonse to income changes. In general, there is enough evidence to conclude that the real exchange rate does not significantly affect the exort and imort demands in the long run in the estimated develoing countries. On another hand, exort demand is highly deendent on foreign income. In Turkey and Brazil, the exort demand resonse to foreign income changes declined in the eriod covering the global crisis. This indicates that the global financial crisis slightly directed the trading artners of Turkey and Brazil towards imort substituting olicies, or towards cheaer roducers; however, these changes were not major. In Indonesia, the resonse of exort demand to changes in the foreign incomes increased, indicating that as a result of an effect of the financial crisis, an increase or decrease in foreign income led to a higher increase or decrease in exort demand, resectively. The estimations rovide enough evidence of high deendence on the imort demand function on the domestic income in the long run. The estimations illustrate that the imort demand became more sensitive to changes in domestic income after the effect of the global crisis in Turkey, Brazil, Russia, India and Indonesia, while the level of deendence of imorts on domestic incomes slightly declined in South Africa. 3.3 Error Correction Model The vector error correction model is designed for cointegrated series. The vector error correction model secifies the short-run adjustment dynamics for long-run equilibrium deviations. The results of the short-run coefficient estimates associated with the long-run relationshis obtained from the ECM version of the ARDL model are resented in Table 3. The ECM coefficient is suosed to be significant with negative sign indicating the seed of the adjustment of variables to the long-run equilibrium. Error correction terms for the exort and for the imort demand functions, resectively, were found negative and statistically significant in the case of Indonesia in the first eriod of the exort demand function and in Turkey, Brazil and Indonesia in the second eriod. Estimating the imort demand function error correction terms were found negative and statistically significant in the cases of Turkey, Russia and South Africa in the first eriod and in Turkey, India, Indonesia and South Africa in the second eriod. These results ensure once more that stable long-run relationshis among the variables in the model of current account balances exist in all considered countries, as noted by Kremers et al. (992) and Bannerjee et al. (998). The magnitude of the error correction term in the exort demand function is between and -0.08, deending on the estimated country in the first eriod, and between and in the second eriod. Therefore, it imlies that disequilibria in the exort demand function was corrected by aroximately 2-% every quarter (resective to country) before the global financial crisis. This means that a steady state equilibrium in the exort demand function can be reached between 2 and 3 years, resective to country in the re crisis eriod. However, in

11 the eriod covering the crisis, the general tendency of the disequilibria correction almost did not change, the steady state equilibrium was reached in the eriod between 2 and 2 years, resective to country. Only some slight changes were observed on the individual country level. Thus, in Turkey, the steady state equilibrium was reached in aroximately 6 years in the re crisis eriod, while under the effect of the global crisis this eriod declined to 2 years. In the imort demand function the equilibrium adjustment seed is higher comare to exort functions. Thus the magnitude of the error correction term is between and in the re crisis eriod and between and in the full eriod. Therefore the steady state equilibrium can be reached in the eriod between less than a year and four and half years. Particularly in Turkey the steady state equilibrium was reached in less than a year with no effect from the global crisis, while in South Africa the adjustment rocess declined from year and a half before the crisis to three years and a half under the effect of the crisis. Signs of the short-run elasticities are consistent with those of the long-run elasticities signs from Table 2. Strong suort was found for concluding that the short-run exchange rates do not lay a very imortant role in the long-run behaviour of imort and exort demands. In contrast to studies on exort and imort demand functions for services, where for examle Ketenci and Uz (200) in the examle of Turkey found that short-run exchange rate elasticities of exort and the imort of services are highly elastic comared to inelastic long-run exchange rate elasticities. In all countries, the short run exchange rate elasticities in exort as well as in imort demand functions were found highly inelastic, nearly close to zero. Thus, only 0.% of the disequilibrium of imort in Turkey is corrected by exchange rate, and only 0.0 % of the disequilibrium of exort in China is corrected by exchange rate. In contrast to studies on exort and imort demand functions for services, where, for examle, Ketenci and Uz (200) in the examle of Turkey found that short-run exchange rate elasticities of exort and imort of services are highly elastic comare to inelastic long-run exchange rate elasticities. Signs of the short-run income elasticities are consistent with signs of the long-run income elasticities in exort as well as in imort demand functions, excet for the case of Turkey, where the short-run foreign income elasticity aeared with negative sign indicating that with an increase of income, foreign countries follow imort substitution olicies. However, the shortrun income elasticity in the case of Turkey was not found significant; therefore, the conclusion cannot be certain. Estimations of the exort demand function illustrate that in all countries excet China the short-run foreign income is inelastic, demonstrating that on average about 20% of the disequilibrium in the exort was adjusted by foreign income in the re crisis eriod in the considered develoing countries. On the other hand, the global financial crisis increased the imortance of foreign income for exort demand. Thus as a result of the crisis effect between 25 and 40% of the disequilibrium in exort, resectively to a country, was adjusted by foreign income. In the case of China, foreign income was found to be highly imortant for exort demand with increasing tendency after the crisis. Thus more than 300% disequilibrium in exort was adjusted by foreign incomes in the re-crisis eriod, while under the effect of the global crisis foreign income was resonsible for adjustment of 400% of disequilibrium in exort. Estimations of the short-run income elasticities in the imort demand function rovided highly statistically significant results in all countries. In all countries, the global crisis increased the imortance level of domestic incomes in the imort demand function by increasing the value of the short-run income elasticities. The extreme case is South Africa, where before the crisis about 70% of the disequilibrium in imorts was adjusted by domestic income, while with the effect of the crisis domestic income became resonsible for more than 500% of the disequilibrium in imorts, illustrating the stee increase in the imort demand sensitivity level

12 2 to domestic incomes in South Africa. In other words, when deviations from the long-run equilibrium occur in the exort and imort demand functions of selected countries, it is rimarily the foreign and domestic incomes that adjust to restore long-run equilibrium each quarter in the exort and imort demand functions, resectively, rather than the real exchange rate.

13 3 4. Conclusion This aer emirically examined the effects of financial crisis on changes in the trade elasticities of BRIICS (Brazil, Russia, India, Indonesia, China and South Africa) countries and Turkey. The effect of the financial crisis was estimated by measuring trade elasticities in exort and imort demand functions for two different eriods on the quarterly basis: 989Q-2007Q2 and 989Q-200Q4. The first eriod was the re-crisis eriod and second was the full eriod that covered the global financial crisis and its contagion effect. These two eriods were studied in order to cature the changes in trade elasticities haened before and after the contagion effect started. The emirical results rovide strong suort for concluding that short-run exchange rates do not lay a very imortant role in the long-run behaviour of imort and exort demands. In all of the estimated countries, excet China, the short-run foreign income was found inelastic with increasing tendency under the effect of the global crisis. The short-run income elasticities in the imort demand function were found highly statistically significant and elastic in all countries. The results indicate that in all estimated countries the global crisis increased the imortance level of domestic incomes in imort demand. The emirical results of long-run coefficients rovide enough evidence to conclude that changes in the real exchange rate do not significantly affect the exort and imort demands in the long run. On another hand, foreign and domestic incomes were found highly significant and elastic in exort and imort demand functions. In Turkey and in Brazil, the resonsiveness of exort demand to foreign income declined after the global crisis. This indicates that the global financial crisis slightly directed the trading artners of Turkey and Brazil towards imort substituting olicies or towards cheaer roducers. In Indonesia, the global financial crisis increased the sensitivity of exort demand to changes in foreign incomes. Indonesia is one of a few countries that did not negatively affected by the financial crisis. Indonesia increased its global market share and domestic sales as well. This increase in exorts is mainly attributable to resource based commodities, while there is still limited rogress in exorts of manufactured roducts 4. The emirical results illustrate the high deendence level of the imort demand function on the domestic income in the long run. Thus the imort demand became more sensitive to changes in the domestic income as a result of the global crisis effect in Turkey, Brazil, Russia, India and Indonesia, while the level of the deendence on imorts on domestic incomes slightly declined in South Africa. In general, the resonsiveness of exorts and imorts to the exchange rate in the considered emerging markets was very low and in many cases insignificant, where the global crisis did not have any effect on these relationshis. On the other hand, the crisis in most of countries increased the already high resonsiveness of exorts and imorts to foreign and domestic incomes, resectively. Taking into account that the incomes in the world imroved after the crisis and started to increase in 2009 and 200, it can be concluded that recovering from the crisis s negative effects emerging countries and their artners did not close their countries, but followed the tendency of international trade increase. Therefore, the trade olicies of emerging countries should be based mainly on foreign and domestic incomes. The further research has to include extended dataset that will be helful in estimation of the effect of new slowdown in the world s growth. 4 Trade Develoment in Indonesia, World Bank.

14 4 References Algieri, B Price and income elasticities of Russian exorts, The Euroean Journal of Comarative Economics, (2), Aydin, M.F., U. Cilak and M.E. Yucel Exort suly and imort demand models for the Turkish economy, Research Deartment, Working Paer 04/09. The Central Bank of the Reublic of Turkey. Aloui, R., M.S.B. Aïssa and D.K. Nguyen. 20. Global financial crisis, extreme interdeendences, and contagion effects: The role of economic structure? Journal of Banking & Finance, 35, Bahmani-Oskooee, M. and C. Economidou How sensitive are Britain s inayments and outayments to the value of the British ound, Journal of Economic Studies, 32, Bahmani-Oskooee, M. and A. Ratha Exchange rate sensitivity of US bilateral trade flows, Economic Systems, 32, Bahmani-Oskooee, M., H. Harvey and S.W. Hegerty The effects of exchange-rate volatility on commodity trade between the US and Brazil, The North American Journal of Economics and Finance, 25, Banerjee, A., J. J. Dolado and R. Mestre Error-correction mechanism tests for cointegration in a single equation framework, Journal of Time Series Analysis, 9, Goldstein, M. and M. Khan Income and rice effect in foreign trade, In Handbook of International Economics. Jones, R.W., Kenen, P.B. (Eds.). Amsterdam, North Holland, Cheung, Y.W., M.D. Chinn and E. Fujii China s Current Account and Exchange Rate, NBER Working Paer Cambridge. Chor, D. and K. Manova Off the Cliff and Back? Credit Conditions and International Trade During the Global Financial Crisis, NBER Working Paer 674, Cambridge. Dooley, M. and M. Hutchison Transmission of the U.S. subrime crisis to emerging markets: Evidence on the decouling-recouling hyothesis, Journal of International Money and Finance, 28, Goldman Sachs Global Economics Grou BRICs and Beyond. Study on BRIC and Next Eleven develoing countries. Hossain, A.A Structural change in the exort demand function for Indonesia: Estimation, analysis and olicy imlications, Journal of Policy Modelling, 3(2), Hsing, Y A study of the J-curve for seven selected Latin American countries, Global Economy Journal, 8(4), -4. Huchet-Bourdon, M. and M. Bahmani-Oskooee Exchange Rate Uncertainty and Trade Flows Between the United States and China, Chinese Economy, 46(2),

15 5 Irandoust, M., K. Ekblad and J. Parmler Bilateral trade flows and exchange rate sensitivity: Evidence using likelihood-based anel cointegration, Economic Systems, 30 (2), Ketenci, N. And I. Uz Trade in Services: The elasticity aroach for the case of Turkey, The International Trade Journal, 24 (3), Kremers, J. J. M., N.R. Ericsson and J.J. Dolado The ower of cointegration tests, Oxford Bulletin of Economics and Statistics, 54, Kumar, S An emirical evaluation of exort demand in China, Journal of Chinese Economic and Foreign Studies, 2 (2), Kwack, S.Y., C.Y. Ahn and D.Y. Yang Consistent estimates of world trade elasticities and an alication to the effects of Chinese Yuan (RMB) areciation, Journal of Asian Economics, 8, McKibbin W.J. and A. Stoeckel The Potential Imact of the Global Financial Crisis on World Trade, Policy Research Working Paer 534. The World Bank. Narayan, S. and P.K. Narayan Imort Demand Elasticities for Mauritius and South Africa: Evidence from two recent cointegration techniques, Discussion aers 03/09, ISSN Monash University, Deartment of Economics. Ogus, A. and N. Sohrabji Elasticities of Turkish Exorts and Imorts, Working Paer, 0906, Izmir University of Economics. Ozkale, L. and F.N. Karaman (in Turkish). Static effects of the EU-Turkey customs union, International Economics and Foreign Trade Policies. Undersecretariat of the Prime Ministry for Foreign Trade. Pesaran, M. H., Y. Shin and R.C. Smith Bounds testing aroaches to the analysis of level relationshis, Journal of Alied Economics, 6, Pesaran, M. H. and B. Pesaran Working with Microfit 4.0: Interactive Econometric Analysis. Oxford University Press, Oxford. Vieira, F.F. and E.A. Haddad. 20. A anel data investigation on the Brazilian state level exort erformance. TD Nereus, Sao Paulo.

16 6 Figure. Exorts in billions of US dollars Turkey Brazil China India Indonesia Russian Federation South Africa Source: Calculations are made on the basis of OECD statistics Figure 2. Imorts in billions of US dollars Turkey Brazil China India Indonesia Russian Federation South Africa Source: Calculations are made on the basis of OECD statistics

17 7 Figure 3. Real GDP, Growth Rate Turkey Brazil China India Indonesia Russian Federation South Africa Source: Calculations are made on the basis of OECD statistics Table. F-statistics for testing cointegration relationshi EXPORT IMPORT Country Lags F-statistic Probability Lags F-statistic Probability Turkey 6 F(3,57)= 2.887* F(3, 57)= 4.005** 0.02 Brazil 4 F(3, 65)= 3.356** F(3, 57)= Russia F(3, 77)= F(3, 6)= 3.268** India 4 F( 3, 65)= F(3, 77)= 3.703** 0.05 Indonesia F(3, 77)= F(3, 69)= 4.026** 0.0 China 6 F(3, 57)= 4.55** F(3, 57)= South Africa 6 F(3, 57)= F(3, 57)= 8.359* Notes: Asymtotic critical value bounds are obtained from Table Critical values for the bounds test case III: unrestricted intercet and no trend for k=3 from Narayan (2005). *, ** indicate significance at 0 and 5 ercent levels, resectively.

18 8 Table 2. Cointegration Coefficient Estimates (long run) Exort Coefficients lag lag Turkey α (3,0,) (0.04) (,0,) -0.04** (0.006) α2 6.28* (3.685) 3.979*** (0.403) Brazil α (,0,0) (0.002) (,0,0) (0.0009) α * (2.957) 4.852*** (.087) Russia α (,,0) (0.08) (,,0) (0.00) α (2.859) 3.58 (.957) India α - - α2 - - Indonesia α (,0,0) (0.003) (,0,0) (0.003) α2.909*** (.3656) 2.250*** (0.3) China α (,0,) ** ( ) (,0,2) (0.008) α ** (.475) (9.706) South Africa α - - α2 - - Imort Turkey β (,0,2) 0.004** (0.002) (,,) 0.005**(0.002) β *** (0.4) 3.92*** (0.) Brazil β (2,0,3) (0.002) (2,0,4) (0.00) β (2.062) 2.428** (.98) Russia β (,0,0) -0.09*** (0.007) (,0,2) (0.03) β2.327** (0.543) (2.84) India β (,0,0) (0.0004) (,0,0) (0.003) β2 0.84** (0.079) 2.6*** (0.249) Indonesia β (3,0,) (0.004) (3,0,) 0.00 (0.004) β2.48*** (0.465) 2.33*** (0.42) China β - - β South Africa β (,0,0) 0.055** (0.023) (,0,) 0.056** (0.025) β *** (0.94) 3.09*** (0.643) Notes : *, **, *** indicate significance at 0%, 5% and % levels, resectively; standard errors for the coefficient estimate are given in arenthesis. α and β are the elasticities of exchange rates for exort and imort from equations 3 and 6, resectively. α2 and β2 are elasticities of income for exort and imort from equations 3 and 6, resectively. Standard errors are given in brackets.

19 9 Table 3. Vector Error Correction λ2 λ3 λ2 λ3 Exort Turkey *** *** ** (0.048) (0.0006) (0.262) (0.043) (0.0007) (0.227) Brazil * ** ** (0.047) (0.000) (0.72) (0.039) (0.000) (0.65) Russia ** ** 0.33** (0.07) (0.002) (0.8) (0.069) (0.002) (0.24) India Indonesia -0.08** *** -0.5** ** (0.053) (0.0003) (0.9) (0.047) (0.0003) (0.8) China South Africa (0.032) ** (0.000) 3.429** (.475) (0.023) *** (0.000) 3.989*** (.02) μ2 μ3 μ2 μ3 Imort Turkey *** 0.00** *** *** * *** (0.079) (0.0006) (0.300) (0.072) (0.002) (0.267) Brazil *** *** (0.058) (0.000) (0.535) (0.044) (0.000) (0.459) Russia -0.27** *** 0.68*** ** (0.054) (0.00) (0.088) (0.055) (0.00) (0.468) India ** -0.35*** *** (0.043) (0.0004) (0.079) (0.039) (0.0004) (0.084) Indonesia *** -0.06* *** (0.072) (0.0004) (0.548) (0.056) (0.00) (0.59) China South -0.55** 0.008** 0.696*** *** 0.004* 5.738*** Africa (0.06) (0.004) (0.228) (0.023) (0.002) (0.044) Notes: *, **, *** indicate significance at 0, 5, and ercent levels, resectively. Standard errors are in arentheses. λ, λ2, λ3 - measure the seed of adjustment of the exort of selected service categories, exchange rate and foreign income, resectively, towards the equilibrium, μ, μ2, μ3 - measure the seed of adjustment of the imort of selected service categories, exchange rate and domestic income, resectively, towards the equilibrium.