Export Performance and Economic Growth in East Asian Economies – Application of Cointegration and Vector Error Correction Model

East Asian Economies are considered to be most successful economies in the world. Following the footsteps of other East Asian economies such as Japan and South Korea, China also shifted towards export-led growth strategy in 80s. This study analyzes the effect of export performance on economic growth of three major East Asian economies i.e. Japan, South Korea, and China. This study has conducted the econometric analysis of macro data under multivariate framework for the period 1980-2012. In order to examine the causal relationship between exports and economic growth, the study has applied time series techniques such as Augmented Dickey-Fuller (ADF) and PhillipsPerron (PP) unit root tests to check stationarity of variables, Johansen cointegration test for long run relationship, vector error correction model (VECM) for short run dynamics and for estimating speed of adjustment towards long run equilibrium. The analysis also made use of techniques Impulse Response Function (IRF) and Variance Decomposition Analysis (VDA) to investigate the interrelationships within the system. The estimated results suggested that all variables were cointegrated for East Asian economies. The study concluded that export-led growth (ELG) was only long run phenomenon in China and South Korea. The results for Japan supported growth led exports (GLE) particularly for short run.


Introduction and Background
After Second World War, Japan focused on industrialization and expansion of exports which led to rapid economic growth of the economy.Later on, Japan's export-led growth model was adopted by four Asian tigers or first tier of newly industrialized economies (NIEs) namely Hong Kong, South Korea, Singapore and Taiwan in the 1960s.After the success of four Asian tigers since 1970, second tier of newly industrialized economies (NIEs) of Southeast Asia namely Indonesia, Malaysia, Thailand and Philippines replicated this strategy.Finally China and India gradually followed this strategy.Hence, rapidly growing economies of Asian region have widely followed export-led growth strategy as an effective tool for development (Page, 1994;Kokko, 2002;Chow, 2012).
East Asia is considered as most successful sub region of Asia.China is the largest country in terms of geographic and demographic features.China's population is ten times more than Japan's population and twenty seven times that of South Korea.Japan and South Korea are located just off the coast of mainland China.In 2012, China, Japan and South Korea together constituted about 20 percent of world economic output in nominal terms.Moreover, China was also the largest trading partner of both Japan and South Korea in the same year.Industrialization was the main reason behind the economic development of these economies (Berglee, 2012;O'Reilly, 2012;Park & Patrick, 2013).The compound growth rates of exports, imports and trade for East Asian economies have been reported in Table 1.Table indicates remarkable exports performance of China and South Korea during 1981-2012.However, Japan's export performance was low during 80s and further declined during later decades.In case of GDP, only China was able to secure double digit growth throughout the study period.Growth rates of exports and GDP are relatively lower for Japan, which is due to Japan being a mature developed economy as well as slowdown in developed world.The purpose of this study is to empirically investigate export-growth relationship for major East Asian economies namely China, Japan and South Korea by using time series data analysis including structural breaks under multivariate framework.The methodology used in this study is little improvement over previous studies as it includes structural breaks, diagnostic tests and forecasting methods such as impulse response function (IRF) and variance decomposition analysis (VDA).The first section includes introductory part and brief background of East Asian economies.The next section reports review of literature.Further, methodology and empirical results have been given.Final section presents conclusions and also contains comparison of our results with previous studies.

Review of Literature
The literature reviewed indicates that comprehensive studies based on rigorous statistical analysis comprising forecasting methods and identifying structural breaks are lacking (Table 2).Therefore, the present study has made an attempt to analyze the relationship between exports and economic growth by adopting above mentioned methods.The evidence for export-led growth hypothesis is inconclusive as the results provided by these studies are not unanimous.Hence, in the light of above facts, the study takes into account these issues in further investigation.Three periods (1913-37, 1952-73, 1973-90)

Model, Database and Econometric Strategy
The aggregate production function used in the study can be expressed as: Where Y represents real gross domestic product and K, L, X, M represent capital, labour, exports and imports respectively.This model has been used to examine the All the variables are taken in their natural logarithms to avoid the problem of heteroskedasticity (Gujarati 1995).For the application of multivariate econometric techniques, the above stated model can be expressed in the following linear logarithmic form: The prefix 'LN' stands for natural logarithm.The study takes into account dummies for Asian Financial Crisis (1997) and Global Economic Crisis (2008).To examine the export-led growth hypothesis cointegration and VECM based causality tests have been used.

Unit Root Results
The results in Table 3 give the summary of ADF and PP tests for East Asian economies.Augmented Dickey-Fuller (ADF) test and Phillip-Perron (PP) test (including constant with trend) for five variables namely LNGDPPC, LNEXP, LNIMP, LNGCF &LNLAB have been applied to check whether series are stationary or not.The results revealed the presence of unit root for all series at levels.After differencing, all series were found to be stationary except the series LNLAB for Japan.The variable integrated of order two or I(2) was dropped from the model in case of Japan.

Chow Test Results
Time series plotting of the variables indicated the presence of structural breaks.Hence, Chow breakpoint test has been used to identify and confirm structural break in dataset.This test analyzes the null hypothesis of no structural break.
The results in

VAR Lag Order Selection Criteria
The next step involves investigation of the long run relationship among variables.Before applying Johansen cointegration procedure appropriate lag length must be set.In Table 5, the results of VAR lag order selection criteria have been presented.Schwarz information criterion was adopted to estimate cointegration and unrestricted VAR.

Cointegration Results
To analyze long run relationship Johansen cointegration procedure has been employed.The results of both trace and max eigen value tests indicated that there was long run equilibrium relationship among variables namely LNGDPPC, LNEXP, LNIMP, LNGCF &amp; LNLAB.Therefore, the results of both tests confirmed the existence of long run relationship in all the selected countries (Table 6).The normalized cointegrating equation for GDP per capita has been given in Table 7.The equation shows that in long run exports affect positively GDP per capita in China.The impact was also significant.Studies like Shirazi and Manap (2005), Aktar (2008), andHusein (2010) also found positive association between exports and economic growth on the basis of normalized cointegrating equations.However, gross capital formation was found to be positive and significant too.Both imports and labour were found to be negative.However, former was significant and later was found to be insignificant in the equation.The equation for Japan indicates that imports growth and gross capital formation have positive and significant effect on GDPPC.Although the variable exports hold positive sign but it was found to be insignificant.For South Korea, the normalized cointegrating equation indicated that in the long run all variables except imports have positive and significant impact on GDPPC.

VECM Short Run Causality Results
VECM results comprise the estimate of the speed of adjustment coefficients and short run properties of series.Table 8 reports the short run causality results obtained from VECM.For China, the coefficients of error correction terms (ECT) with GDPPC and exports as dependent variable were negative but former was statistically significant at 5% level of significance indicating there is convergence from short dynamics towards long run equilibrium.The adjustment coefficient was found to be 0.33 percent implying that speed of adjustment was 33 percent towards long run equilibrium in case of disequilibrium situation.However, short run coefficients of first difference of LNEXP lagged one period for GDP per capita as dependent variable and first difference of LNGDPPC lagged one period for exports equation were found to be statistically insignificant which indicates the absence of short run causality in any direction.
In case of Japan, the results exhibit that coefficient of error correction term (ECT) was not significant in any of two cases for GDPPC and exports.However, the sign was negative (correct) for exports equation.Further, short run coefficient of first difference of LNGDPPC lagged one period for exports equation was found to be positively significant which indicated unidirectional short run causality from GDPPC to exports or growth led exports.The short run coefficient of first difference of LNEXP lagged one period for GDPPC equation was found to be negatively significant.Dummy variable was found to be negative in both cases.
The results for South Korea depicted that coefficient of error correction term was not significant for GDPPC equation however the sign was negative (correct) whereas for exports equation, the error correction term was significant but the sign was positive (incorrect).Thus, the results indicated lack of significant adjustments towards long run equilibrium in any disequilibrium situation.Further, short run coefficients of first difference of LNEXP lagged one period for GDPPC equation and first difference of LNGDPPC lagged one period for export equation were negative and insignificant.Thus, the results indicated the absence of short run causality between these two variables.Thus results are similar to those reported in the study by Lawrence and Weinstein (1999).Dummy variable for Asian crisis 1997 was found to be negative and statistically significant implying the negative impact of crisis on Korean economy.However, dummy variable for Global crisis 2008 was also found to be negative but statistically insignificant.

Diagnostic Tests
The models were tested for normality, heteroskedasticity and serial correlation.
Diagnostic tests were carried out on the data revealed that models were well specified (Table 10).Diagnostic tests also indicated that the residuals were normally distributed, homoskedastic and serially uncorrelated.However for South Korea, Jarque-Bera normality test depicted non normality.Mcdonald (2014) noted that deviation from normality in case of parametric tests is not very sensitive.Wooldridge (2012) pointed out that non-normality of errors is not a serious problem with large sample size.Ghasemi and Zahediasl (2012) also suggested that with large enough sample sizes (> 30 or 40), the violation of the normality assumption should not cause major problems this implies that we can use parametric procedures even when the data are not normally distributed.

Impulse response function and variance decomposition analysis results
The results of impulse response function indicated that among all variable one positive shock to gross capital formation and GDP per capita results in positive response in GDP per capita.In case of exports, one positive shock to imports and gross capital formation brings positive response for exports while positive shock to GDP per capita results in negative response of exports.Variance decomposition analysis (VDA) depicted that GDPPC shock accounted for whole variance of GDPPC in first year.After 10 years, GDP per capita (63.44 percent), imports (28.76 percent), gross capital formation (6.53 percent) and exports (1.10 percent) shocks bring variability in GDP per capita.For exports, in first year exports (81.27 percent) and GDP per capita (18.72 percent) shocks account for variance of exports in case of China.This proportion predicted to change over time as after 10 years, exports (75.80 percent), GDP per capita (8.52 percent), gross capital formation (12.17 percent) and imports (3.28 percent) shocks found to be important source of export variability for China.For Japan, impulse response function indicated one positive shock to GDPPC brings entire positive response of GDPPC.No other variable was found responsible for positive response in GDPPC.Variance decomposition analysis of Japanese GDPPC showed that GDPPC is unexpurgated source of variation in its forecast error.After ten years, variation in GDPPC is accounted for by GDPPC (53.58 percent), exports (45.30percent) and rest of other variables contributed less than one percent.For exports, predominant source of variation are exports (61.47 percent) and GDPPC (38.52 percent).In the tenth year, exports (72.86 percent), imports (10.23 percent), GDPPC (9.58 percent) and gross capital formation (7.31 percent) contributed in exports variability.In case of South Korea, Variance decomposition analysis exhibited GDPPC as predominant source of variation in GDPPC, ranging from 100 percent to 91.11 percent.For exports, GDPPC (74.20 percent) and exports (25.79 percent) accounted for export variability.After ten years, exports (58.74 percent), GDPPC (31.11 percent), labour (4.87 percent) and imports (4.34 percent) and gross capital formation (0.92 percent) were the sources of variation in exports (see appendices Table A1, Table A2, and Table A3).

Conclusion and Policy Implications
In order to observe the relationship between exports and economic growth for East Asian economies during 1980-2012, this study constructed multivariate framework using the variables GDP per capita, exports, imports, gross capital formation and labour.Time series techniques such as Augmented Dickey-Fuller (ADF) and Phillips-Perron (PP) unit root tests, Johansen cointegration test, vector error correction model (VECM) were employed.The analysis also made use of forecasting techniques namely Impulse Response Function (IRF) and Variance Decomposition Analysis (VDA).The study also conducts diagnostic tests for normality, heteroskedasticity, autocorrelation using Jarque-Bera Normality test, ARCH Heteroskedasticity test and Breusch-Godfrey LM test.
The estimated results suggested that all variables were cointegrated for East Asian economies.The normalized equation shows that in long run exports affects positively GDP per capita in China and South Korea.The impact was also significant.The result exhibits that coefficient of error correction term (ECT) for GDPPC equation was significant only for China indicating significant adjustments towards long run equilibrium in any disequilibrium situation.No short run causality was found between GDPPC and exports in case of China.The short run coefficient of first difference of LNGDPPC lagged one period for exports equation was found to be positively significant for Japan.In case of South Korea, the results indicated the absence of short run causality between these two variables.Hence, export-led growth (ELG) hypothesis was not found valid for China, Japan and South Korea particularly in short run however, reverse causation i.e. growth led exports (GLE) was confirmed for Japan in short run.Thus, the study concluded that export-led growth (ELG) was only long run phenomenon in China and South Korea.The results for Japan supported growth led exports (GLE) particularly for short run.Although East Asian economies export performance remained exceptionally well.But the Asian Financial Crisis 1997 and Global Economic Crisis 2008 resulted in long term adverse effects excluding China.
The results of the study clearly highlight the importance of exports in the selected East Asian economies.In the long run exports are positively affecting GDP per capita in China and South Korea.Hence, these economies should continue to promote their exports.Japan is matured developed economy with high per capita income and has different structure of the economy.Japan has experienced growth led exports in short run and hence this economy will have to promote growth internally as it is suffering from past two decades of stagnation.
The present study gives strong support to the findings of Lin (1999), Yao (2006) for positive effect of exports on economic growth; Liu, Burridge and Sinclair (2002) for long run relationship while study contradicts Liu, Song and Romilly (1997) for short run results and Tsen (2010) who found bidirectional causality for China.For Japan, the study supported Fawson and Chang (1994) while it contradicts Zhang and

Table 1 . Growth performance of East Asian Economies (Compound Growth Rates)
Source: Calculations based on data from World Development Indicators (WDI), online database.

Table 2 (cont.) A Brief Review of the Related Economic Literature on East Asia Author Period of the study
*Note: BDC refers to bi-directional causality, ELG refers to export led growth, GLE refers to growth led export.
export-led growth (ELG) hypothesis for major East Asian economies.The study has used annual data at the 2005 constant US dollar prices from 1980 to 2012.Data on real GDP per capita(GDPPC), real exports(EXP), real imports(IMP), real gross capital formation(GCF) has been compiled from World Development Indicators (WDI) online database, World Bank, while data on total labour force(LAB) is collected from United Nation Conference on Trade and Development (UNCTAD) Statistics.

Table 3 . Results of Unit Root test for variables
Note: *, ** and *** indicate significance at the 1%, 5% and 10% respectively.Values in the parentheses show p-values.
Table4clearly indicate the absence of structural break in dataset as the null hypothesis of no structural break can't be rejected for China.For Japan and South Korea, the results affirmed the presence of structural break in dataset.Asian crisis of 1997 had severe impact on South Korea's economy.The results showed the presence of structural break for Japan and South Korea's dataset for 2008.