- Email: [email protected]
Financial integration in East Asia: Evidence from panel unit root and panel cointegration tests
Journal of Asian Economics 20 (2009) 314–326
Contents lists available at ScienceDirect
Journal of Asian Economics
Financial integration in East Asia: Evidence from panel unit root and panel cointegration tests Cyriac Guillaumin * CEPN, University of Paris 13, 99 Avenue Jean-Baptiste Cle´ment, 93430 Villetaneuse, France
A R T I C L E I N F O
A B S T R A C T
JEL classification: C33 F32 F36 F41
The aim of this paper is to investigate the degree of financial integration for selected East Asian countries from 1988 to 2006 using the recently developed panel unit root and panel cointegration techniques. Investment and savings rates are found to be nonstationary and not to be cointegrated in panels. We estimate modified Feldstein–Horioka equations and our results reveal a high degree of financial integration. When we homogenize our data, results show that high-income countries have stronger financial integration than middleincome countries. Finally, we proceed to stability tests in order to test if there is a crisis effect and we find that financial integration is stronger in the post-crisis period. ß 2009 Elsevier Inc. All rights reserved.
Keywords: Financial integration Feldstein–Horioka puzzle Nonstationary panel data Panel cointegration East Asia
1. Introduction In their paper, Feldstein and Horioka (1980) measure the degree of financial integration using the correlation of domestic saving with investment rates. Their results show a high correlation during 1960–1974. This result is contradictory with the stylized fact of this period; indeed, capital mobility appears to be increasing ever since the beginning of the 1960s (Beitone, Gilles, & Parodi 2006; Flandreau & Rivie`re, 1999). The Feldstein–Horioka approach is subject to two critiques that focus either on the econometric and statistical caveats – responsible for a bias towards a significant, positive and close to one coefficient – or on the economic and historical circumstances that explain the discrepancy between an expected coefficient close to zero and its effective value. The goal of this paper is to investigate the degree of financial integration in selected East Asian countries from 1988 to 2006 using the recently developed panel unit root and panel cointegration techniques. Ho (2002), Kim, Oh, and Jeong (2005) and Be´reau (2007) have used these techniques but only Kim et al. (2005) focused on Asian countries. Kim et al. (2005) study capital mobility in 10 countries (Korea, Indonesia, Japan, Malaysia, Myanmar, Pakistan, Philippines, Singapore, Sri Lanka and Thailand) between 1960 and 1998. They find weaker financial integration during 1960–1979 than 1980–1998. The evolution of financial integration can be explained by financial liberalization of these economies in the 1980s and 1990s. But the study of Kim et al. (2005) does not integrate the last 10 years. Moreover, the second-generation panel unit root tests based on the hypothesis of interdependence between individuals are not used by any of the earlier studies. We employ first-generation panel unit root tests along the lines of Levin and Lin (1992), Im, Pesaran, and Shin (2003), Maddala and Wu (1999) and Hadri (2000), and second-generation panel unit root tests along the lines of Pesaran (2003). We
* Tel.: +33 1 49 40 32 55; fax: +33 1 49 40 33 34. E-mail address: [email protected]. 1049-0078/$ – see front matter ß 2009 Elsevier Inc. All rights reserved. doi:10.1016/j.asieco.2009.02.002
C. Guillaumin / Journal of Asian Economics 20 (2009) 314–326
315
also test for the presence of a long term relationship between investment and saving rates using Pedroni’s cointegration test (1997, 1999). We estimate three types of modified Feldstein–Horioka equations: between, within and pooling. These different estimations show a high degree of financial integration. We try to homogenize our data and estimate a new relation for highincome and middle-income countries. Our results show that high-income countries have stronger financial integration than middle-income countries. Finally, we proceed to stability tests in order to test if there is a crisis effect. Our results support the idea that financial integration is stronger in the post-crisis period. The paper is structured as follows: Section 2 provides an overview of the Feldstein–Horioka puzzle. Section 3 describes the methodology. Section 4 provides a survey of the evolution of financial integration in East Asia. In Section 5 we present and discuss results. Section 6 offers concluding remarks. 2. Feldstein–Horioka puzzle: an econometric solution? 2.1. The Feldstein–Horioka paradox In their pioneering study, Feldstein and Horioka (1980) examined the cross-sectional correlation between saving and investment by testing the following equation: I S ¼aþb þ ei;t Y i;t Y i;t
(1)
where I/Y represents the ratio of investment over Gross Domestic Product (GDP), i.e. the investment rate. S/Y represents the ratio of saving over GDP, i.e. the saving rate. Indexes i and t respectively stand for the considered country and year. ei,t represents other factors explaining investment. This term is a random walk, with zero mean (E(e) = 0) and variance s2. Feldstein and Horioka work on average data in order to prevent a bias: I¯ Y
i
S¯ ¼aþb þ ei Y i
with
I¯ Y
i
¼
974 1 1X I ; T t¼1960 Y i;t
S Y
! ¼ i
974 1 1X S T t¼1960 Y i;t
and
T ¼ 15:
0
(1 )
Coefficient b indicates the degree of financial integration: in case of weak financial integration, b is equal to 1 and to 0 in case of a strong financial integration. Feldstein and Horioka (1980) test Eq. (10 ) for 16 OECD countries during 1960–1974. The result however pointed to a quite low degree of financial integration despite a large volume of international capital movements (Beitone et al., 2006; Flandreau & Rivie`re, 1999). Since the 1970s, capital mobility has been increasing but it is hard to measure this phenomena. Still, some papers (Davanne, 1998; Obstfeld & Taylor, 2002) indicate that the daily financial transactions were equal to US$ 1,500 billion in 1999, i.e., 50 times more than daily commercial trade. In the 1970s, this figure was in the range of US$ 10–20 billion. One explanation, among others, is the emergence of financial markets at the end of the 1970s. World stock market capitalization was US$ 1400 billion in 1975 and US$ 17000 billion in 1997 (Proble`mes e´conomiques, 1997). Explaining the Feldstein– Horioka puzzle in the literature has been a matter of either econometrics or of economic analysis. 2.2. Solving the Feldstein–Horioka enigma The economic arguments are based on historical circumstances that influence capital mobility (Flandreau & Rivie`re, 1999; Obstfeld & Taylor, 2002). Therefore, financial integration can be high or low according to the precise periods and to the process of financial liberalization. Besides, these arguments try to explain Feldstein–Horioka enigma by the relationship between investment and saving with variables that can influence the monetary conditions in the studied area. Economic and historical caveats have been studied by many authors: Bayoumi (1989), Frankel (1992), and Armstrong, Balasubramanyam, and Salisu (1996) are examples. None of these papers accepts perfect capital mobility. Bayoumi (1989) explains the strong relationship between investment and saving due to fiscal and monetary policy. For example, in case of a lack of saving, the government can modify its fiscal policy (decreases in taxes or increases in interest rates) and create a new link between investment and saving. In fact, l’effet taille can modify the relationship between investment and saving: a big country, in economic terms, can influence the world interest rate. So an increase in domestic savings will lead to a decrease in interest rates and to an increase in investment. In addition, Bayoumi (1989), according to Summers (1989), notes that as saving and investment are procyclical variables, their correlation can result from a common response to permanent shocks. Moreover, others variables can influence financial integration, such as the interest rate or risk premium (Frankel, 1992). To correct for this bias, a regional approach is better than a national approach, proposed by Sinn (1992), Bayoumi and Rose (1993), Armstrong et al. (1996), Iwamoto and van Wincoop (2000) and van Wincoop (2001).1 The goal of this approach is to study financially-integrated countries. The econometric and statistical critiques focus on statistical properties of variables used in estimating Eq. (1). If saving and investment are not stationary and not cointegrated, Eq. (1) is a spurious regression. The interpretation of b is impossible.
1
See, for example, He´ricourt (2005) or He´ricourt and Maurel (2005) for a synthesis.
C. Guillaumin / Journal of Asian Economics 20 (2009) 314–326
316
A plausible long-term investment-savings relationship is tested by Coakley, Kulasi, and Smith (1996). This relationship can explain the ambiguous results of Feldstein and Horioka (1980) because the estimate of the b parameter leads to a figure close to 1. But this figure does not reflect a lack of financial integration. Coakley et al. (1996) show that a cointegration relationship prevents an accurate measure of the degree of financial integration. To correct for this problem, times series and panel data econometric techniques are employed. Some papers try to mix these two approaches (Banerjee & Zanghieri, 2003).2 2.3. Times series econometrics Sinn (1992) and Jansen (1996), for example, use times series econometrics to solve the Feldstein–Horioka enigma. They integrate nonstationary data but also plausible cointegration relationships. The authors introduce current account data (CA) as well. Economically, if there is a cointegration relationship between investment and saving, there is a long-term solvability constraint of the current account. By definition, CA = S I. With this long-term solvability constraint, the current account is equilibrated. Hence, E(CA) = 0, where E is the mean. From this, CA = S I is a stationary condition, implying that investment and saving are cointegrated. VAR or VECM models (in case of cointegration) investigate the degree of financial integration. These approaches are very different from Coakley et al. (1996) where cointegrated variables lead to spurious results. 2.4. Panel data econometrics Krol (1996) was the first author to use panel data in this context. This method partially solves the Feldstein–Horioka paradox. Krol (1996) studies the degree of financial integration for the same OECD countries studied by Feldstein–Horioka (1980), using panel data. He analyses jointly individual and time dimensions of saving and investment data. The use of panel data allows for the introduction of fixed and temporal effects.3 Thus, Krol (1996) tests Feldstein–Horioka using the following equation: I S ¼mþb þ ui ;t Y i ;t Y i ;t
(2)
with ui,t = ai + lt + ei,t, where ai is an individual effect (fixed or random), lt is the temporal effect (for all panel data) and ei,t is a random term, with zero mean and variance s2. His results show perfect capital mobility between countries. Jansen (1996) contests these results due to the presence of Luxembourg.4 But Ho (2002), using the two-stage least squares approach proposed by Kao and Chiang (2001), shows that Luxembourg does not affect the results. Nevertheless, panel data econometrics are insufficient to capture effectively financial integration as well as macroeconomics phenomena (Araujo and alii, 2004; Hurlin & Mignon, 2005). We argue that it is necessary to take into account new panel data econometrics with nonstationary and cointegrated data. 2.5. Nonstationary panel data econometrics In this paper, we investigate the degree of financial integration of selected East Asian countries from 1988 to 2006 using the Feldstein–Horioka (1980) approach but with new panel data econometric tools. The economic and historical circumstances influencing relevant variables have been the object of a large number of papers. While the work of Bayoumi and Rose (1993), Armstrong et al. (1996), van Wincoop (2001) or Iwamoto and van Wincoop (2000) adopt interesting approaches, we do not have sufficient data for Asian countries to undertake them. These difficulties are increased with the definition of saving and investment rates at a regional level. Considering the above econometric critiques, we note that a times-series approach seems appropriate but that panel data econometrics have two advantages5: First, panel data have two dimensions, individual and temporal. This double dimension allows us to study simultaneously the dynamics and heterogeneity of individuals. Second, the double dimension allows us to eliminate a difficulty in the times series: the weakness of unit root and cointegration tests with a limited dataset.6 3. Methodology 3.1. Panel unit root tests In this section, we present our methodology. Our presentation is based on Araujo, Brun and Combes (2004) and Hurlin and Mignon (2005).
2 3 4 5 6
See Be´reau (2007) for a synthesis of this literature. Coiteux and Olivier (2000), Jansen (2000) and Corbin (2001) investigate financial integration following the Krol (1996) approach. See Be´reau (2007) for details about Luxembourg. See, for example, Baltagi (1995), Hsiao (2003) or Araujo and alii (2004). See Salanie´ (1999) for a discussion on this subject.
C. Guillaumin / Journal of Asian Economics 20 (2009) 314–326
317
3.2. First-generation panel unit root tests Four first-generation panel unit root tests (discussed above) are used in our paper. All tests are based on the following Augmented Dickey–Fuller regression:
Dxit ¼ ai þ bi t þ ri xit1 þ eit
(3)
The error term is eit and is IID. ai and bi allow for fixed and unit specific time trends for each i. The null hypothesis of a unit root is: H0 : ri ¼ 0; 8 i
(4)
The alternative hypothesis is different according to the test adopted. Levin and Lin (1992) consider an alternative hypothesis where the autoregressive coefficient is homogeneous across countries: Ha : ri ¼ r < 0; 8 i
(5)
They suppose that inter-individual heterogeneity is captured by the fixed effects. Im et al. (2003) extend the Levin and Lin (1992) framework to allow heterogeneity in the value of the autoregressive coefficient. The alternative hypothesis is then specified as: Ha : ri < 0; i ¼ 1; :::; N 1
and
ri ¼ 0; i ¼ N 1 þ 1; :::; N
(6)
In this test, we have individual i = 1, . . ., N1 where xit is stationary and individual i = N1 + 1, . . ., N, where xit is nonstationary. The t-statistic is an average and can be calculated as: t¯ ¼
N 1X t N 1 i
(7)
where ti is the individual ADF t-statistics for the unit root test. Maddala and Wu (1999), relying on Fischer (1932), suggest combining the p-values of the test statistic for a unit root in each cross-sectional unit. The statistic of Maddala and Wu is given by: N X MW ¼ 2 lnð pi Þ
(8)
i¼1
where pi is the p-value of the test statistic of the ADF test statistic in each country. Finally, in the case of the Hadri (2000) approach, the null hypothesis is the stationarity of the series instead of nonstationarity. The framework is the one dealt with in KPSS7 for times series. The model may now be expressed as: xit ¼ r it þ bi t þ eit
(9)
where rit is a random walk given by: r it ¼ r it1 þ uit where uit is white noise. We may assume that E(uit) = 0 and 2 u
H0 : s ¼ 0
(10) Eðu2it Þ
¼s
2 uit
> 0. The null hypothesis is: (11)
3.3. Second-generation panel unit root tests Second-generation panel unit root tests are based on the interdependence between individuals. Numerous second-generation panel unit root tests are proposed in the literature,8 e.g., Choi (2001), Pesaran (2003), Phillips and Sul (2003). Most of them are based on the Bai and Ng (2001), Bai and Ng (2004) methodology, which posit models with common factors. This test considers two separate unit root tests on common and individual components of the series. Other tests are based on a unique test of the unit root of the series. The difference is based on the decomposition of the series. We have decided to present only the Pesaran (2003) test, which focusing on the series xit corrected by the individual average of xit1 and of first differences Dxit1. Pesaran (2003) obtains a Cross Sectionally Augmented Dickey–Fuller (CADF) test. This model, inspired by the test developed in Im et al. (2003), is expressed as:
Dxit ¼ ai þ ri xit1 þ eit 7 8
Kwiatkowski et al. (1992). See Hurlin and Mignon (2005) for details.
(12)
C. Guillaumin / Journal of Asian Economics 20 (2009) 314–326
318
where eit = giut + uit. ut is a common factor and uit is white noise. The CADF model is given by, without autocorrelation of uit:
Dxit ¼ ai þ ri xit þ ci x¯ t1 þ di Dx¯ t1 þ vit
(13)
The Pesaran statistic, Cross-Sectionally Augmented IPS (CIPS) is given by: CIPSðN; TÞ ¼
N 1X t ðN; TÞ N i¼1 i
(14)
where ti constitutes statistics coming from each CADF model for each individual i of the panel. 3.4. Panel cointegration tests The next part of the process involves testing whether there is a statistically acceptable cointegration relationship between the variables of interest. This is achieved by applying the tests developed by Pedroni (1997,1999), which develop seven tests for cointegration in a panel context. Four of the tests are within-dimension statistics and three are between-dimension statistics. The four within–dimension statistics are based on pooling the autoregressive coefficients across the different countries for the unit root tests on the estimated residuals, whereas the three between-dimension statistics are based on estimators that simply average the individual estimated coefficients for each country. For all seven tests, the null hypothesis is no cointegration. The Pedroni cointegration tests are based on estimating the static cointegration regression given by: yit ¼ ai þ bi t þ b1i x1it þ b2i x2it þ ::: þ bMi xMit þ eit
(15)
where i = 1, . . ., N, t = 1, . . ., T and m = 1, . . ., M. ri is the autoregressive term of the residuals given by:
eˆ it ¼ ri eˆ it1 þ uit
(16)
The specification of the alternative hypothesis for the within-dimension (pooled) estimation is: Ha:ri = r < 1, 8i, whereas for the between-dimension (heterogeneous) estimation, the alternative hypothesis is given by Ha:ri < 1,8i. Pedroni’s procedure requires five steps. After these steps, we can construct the statistics for all the seven tests.9 4. Financial integration in East Asia: a brief review The weight of Asia in world trade has been increasing for many years (Gue´rin & Sa, 2006). This increase is essentially intrazone (Bajou, Lefeuvre, Mamandz & Melka, 2006; Guillaumin, 2009). Although the financial links developed intensified after the financial crisis of 1997, they remain weaker than the real-sector links. Even after 1997, financial markets continue to prefer banks in order to finance growth (Eichengreen, 2006; Gyntelberg, Ma & Remolona, 2006), despite the fact that a onepillar financial system dominated by banks is problematic because Asian banks have had a lack of governance10 and have been criticised for ‘‘crony capitalism,’’ that is, conflict of interest issues due to the interaction of private and public interests.11 Two initiatives were developed to increase the share of financial markets: the Asian Bond Market Initiative and the Asian Bond Funds. The Asian Bond Market Initiative (ABMI) of the ASEAN+312 countries aims to develop robust bond markets in the region; avoid the double mismatch of maturity and currency that exacerbated the East Asian crisis; recycle Asian savings to Asian economies; and increase regional financial integration. The Asian Bond Funds (ABF) of the Executives’ Meeting of East Asia-Pacific Central Banks (EMEAP) aims to develop demand for bond denominated in local currency (Gue´rin & Sa, 2006). Nevertheless, there is a lack of pecuniary resources for these two initiatives (Gue´rin & Sa, 2006; Takeuchi, 2006). For example, the ABF has US$ 1 billion invested in US bonds by Asian countries except for Japan. According to Eichengreen (2004, 2006), the slow development of financial market is explained by: - Regional characteristics: no diversity in financial institutions; - Weak supervision and regulation. There is no banking competition, rating agency and no compensation; - Macroeconomics policies are not in favour of investment: high interest rates, and volatility and exchange-rate risks are problematic.
9
See Hurlin and Mignon (2007) for details. Ben Gamra and Plihon (2007) show that a weak quality of institutions for emerging countries increases the probability of a banking crisis. 11 Menkhoff and Suwanaporn (2007) remind that a significant number of reform have been made, and especially for Thailand, on banking regulation and supervision. Even if Thailand is also a country essentially where growth is financing by banks, financial market share increases. 12 Asean+3 is made of the following countries: Indonesia, Malaysia, Singapore, Thailand, Philippines, Brunei, Vietnam, Myanmar, Laos, Cambodia, Timor and China, Japan, Korea. 10
C. Guillaumin / Journal of Asian Economics 20 (2009) 314–326
319
Table 1 Investment and saving rates (in percent of GDP). Average on the period 1988–2006. Countries
I/Y
S/Y
(S I)/Y
China South Korea Hong Kong SAR Indonesia Japan Malaysia Philippines Singapore Thailand
32.56 33.37 26.64 24.55 28.23 31.34 20.61 32.25 31.94
38.58 35.33 32.30 29.92 29.97 40.75 17.07 46.44 33.67
6.01 1.96 5.66 5.37 1.74 9.41 3.54 14.19 1.73
East Asia and Pacific European Monetary Union USA
30.91 20.85 18.04
35.75 22.26 16.36
4.84 1.41 1.68
Notes: East Asia and Pacific contains the following countries: Australia, Brunei Darussalam, the Kingdom of Cambodia, China, Fiji, French Polynesia, Guam, Hong Kong SAR, Indonesia, Japan, Kiribati, North Korea, South Korea, Laos, Macau SAR, Malaysia, Marshal Islands, Micronesia, Mongolia, Burma, New Caledonia, New Zealand, American Samoa, North Marina Islands, Palau, Papua New Guinea, Philippines, Samoa, Singapore, Salomon Islands, Thailand, Timor, the Kingdom of Tonga, Vanuatu and Vietnam. Source: World Development Indicators, World Bank.
Takeushi (2006) notes that there is considerable disparity across financial markets in the region with respect to capital mobility, taxes and liquidity. Institutional issues are important factors and constrain the system (Bae, Bailey & Yun, 2006; Eichengreen & Luengnaruemitchai, 2004). Possible solutions that have been explored and tabled are many, including the development of public bond markets (Gyntelberg and alii, 2006; Sundaresan, 2006) and creating different rating agencies (Park & Rhee, 2006). 5. Results 5.1. Data We investigate the degree of financial integration for 9 East Asian countries: China, Korea, Hong Kong SAR, Indonesia, Japan, Malaysia, Philippines, Singapore and Thailand. This choice is justified by economic (Kim, 2001; Moneta & Ru¨ffer, 2009) and commercial (Bajou et al., 2006; Guillaumin, 2009) links between these countries. Taiwan, Vietnam, Laos, Myanmar and Cambodia are not considered in the estimations because we do not have data for them. We consider the series of investment and saving rates (that is, as a percentage of GDP). Data come from World Development Indicators of the World Bank. The starting date for these series is 1988 (until 2006) measured at an annual frequency. We begin our investigation in 1988 because the process of financial liberalization started at the beginning of the 1980s in Asia but the year 1988 is generally held by the literature to be when integration began its acceleration (Fukasaku & Martineau, 1999). Many authors have worked on the degree of financial integration in East Asia using Feldstein–Horioka measure – e.g., Sinha (2002) and Kim, Kim, and Wang (2007) using times series, and Isaksson (2001) and Kim et al. (2005) with panel data – but these papers tend to begin with data from the 1960s until the Asian financial crisis.13 Given the tend in financial integration, years after the financial crisis are of great importance. Moreover, we are able to use new tools in panel data econometrics.14 Table 1 shows investment and saving rates in these countries. From 1988 to 2006, the average saving rate was 34% with a high variance: Singapore has a saving rate of 47% and the Philippines, only 17%. The average investment rate is 29% with a lower variance, from 33% (Korea) to 20% (the Philippines). We can see that all countries, except the Philippines, have higher saving rates than their investment rates. These countries have, therefore, a current account surplus while the United States has a deficit. For East Asian countries overall this surplus of saving explains in part the ‘‘miraculous’’ growth in the 1980–1990s (Ito, 2001; Stiglitz and Yusuf, 2001). Fig. 1 shows the relationship between investment and saving rates for each of the 9 countries.15 We can observe a degree of heterogeneity in the evolution of saving and investment rates. Clearly, the financial crisis had an asymmetric impact on the East Asian economies. For all countries, except China, we observe a decrease in investment and saving rates in 1997–1998. For Korea, Malaysia and Thailand, it is a collapse. 5.2. Unit root and cointegration tests results 5.2.1. Unit root tests Tables 2 and 3 summarize panel unit root tests (first- and second-generation).16 According to these tables, the unit root hypothesis is accepted in a large number of models at a 1% level of significance. There are some exceptions. For
13 14 15 16
See, for example, Bautista and Maveyraud-Tricoire (2008) for a synthesis. Besides, only Kim et al. (2005) and Kim et al. (2007) concern East Asian countries. Appendix A presents the relationship between investment and saving rates as proposed by Feldstein and Horioka (1980). Appendix B presents results of panel unit root tests according to the level of income.
C. Guillaumin / Journal of Asian Economics 20 (2009) 314–326
320
Fig. 1. Investment (IR) and saving (SR) rates (percent of GDP). Source: World Bank, World Development Indicators.
example, the null hypothesis of a unit root is rejected, at the 10% level, for saving rates with the Lin, Levin and Chu test (models with and without trend). Models with trend do not accept unit root test for saving (5% level) and investment (10% level) rates for the Maddala Wu test. Finally, the Im, Pesaran and Shin test rejects a unit root for saving rate at 10% (model with trend).
Table 2 First generation panel unit root tests. Variables
Model
LLC
IPS
MW
Hadri
IR
Model without trend Model with trend
0.20*** (0.4189) 2.22*** (0.0130)
0.42*** (0.6660) 1.23*** (0.1087)
13.26*** (0.7754) 30.04*** (0.0372)
3.54*** (0.0002) 5.39*** (0.0000)
SR
Model without trend Model with trend
1.52*** (0.0633) 2.19*** (0.0140)
1.06*** (0.1433) 1.51*** (0.0655)
24.36*** (0.1434) 27.09*** (0.0773)
4.37*** (0.0000) 5.36*** (0.0000)
Notes: IR and SR for Investment rate and Saving rate (in percent of GDP). Unit root hypothesis is accepted at *** 1%, **5%, *10% levels. For Hadri’s test (2000), unit root hypothesis is the alternative hypothesis. LLC, IPS and MW for Levin, Lin, and Chu (2002), Im et al. (2003) and Maddala and Wu (1999) tests results. Values in parenthesis are p-value. All tests are made with fixed effects.
C. Guillaumin / Journal of Asian Economics 20 (2009) 314–326
321
Table 3 Second generation panel unit root tests. Variables
IR SR
Models without trend
Models with trend
CIPS1
CIPS2
CIPS3
CIPS1
CIPS2
CIPS3
2.02*** (0.20) 1.93*** (0.277)
2.21*** (0.07) 1.36*** (0.859)
1.30*** (0.894) 1.51*** (0.733)
2.33*** (0.435) 2.58*** (0.185)
2.93 (0.626) 1.51*** (0.985)
2.69 (0.107) 0.571*** (1.000)
Notes: IR and SR for Investment rate and Saving rate (in percent of GDP). Unit root hypothesis is accepted at ***1%, **5%, *10% levels. CIPSi with i = 1, 2, 3 corresponds to the Pesaran statistics test (2003) for lag i. Values in parenthesis are p-value. All tests are made with fixed effects.
Table 4 Panel cointegration test. Totala
Sample
Middle-income **
Panel v-stat Panel rho-stat Panel pp-stat Panel adf Group rho-stat Group pp-stat Group adf
***
Banking
***
3.78 2.27** 2.18** 2.55*** 0.18 2.18 1.34
1.87 1.32* 1.56* 1.06 0.64 1.63 2.53
High-income
Financial Market
**
4.17 1.44* 1.56** 0.78 1.33 1.26 1.07
2.17 1.30* 0.99 1.04 0.99 1.28 0.62
1.01 0.78 2.99*** 0.15 1.70** 1.28 1.99**
Notes: the hypothesis of no cointegation is rejected at ***1%, **5%, *10%. The four within-dimension statistics are based on pooling the autoregressive coefficients across the different countries for the unit root tests on the estimated residuals. Whereas the three between-dimension statistics are based on estimators that simply average the individual estimated coefficients for each country. a All countries in the panel. Table 5 FDI flows. Inflows
Outflows
1990–1996 China Singapore Hong Kong Malaysia Indonesia South Korea
1997–2006 20.7 5.9 4.6 3.6 2.7 2.3
China Hong Kong Singapore South Korea Indonesia Thailand
1990–1996 50.9 25.5 13.6 5.7 4.4 3.6
Hong Kong Singapore Taiwan China South Korea Malaysia
1997–2006 14.9 3.6 3.0 2.3 2.2 1.4
Hong Kong Singapore Taiwan South Korea China Malaysia
25.8 7.4 5.5 3.9 3.0 1.7
Note: billion US dollars. Source: UNCTAD FDI/TNC database.
5.2.2. Cointegration test The results of the cointegration analysis tests over the period 1988–2006 are presented in Table 4. We also test cointegration for sub-groups of countries in the panel as recommended by Banerjee and Zanghieri (2003). We study countries according to income level using the classification of the World Bank (gross national income). According to this classification, China, Malaysia, Philippines and Thailand are countries with middle-income; Korea, Hong Kong SAR, Japan and Singapore are countries with high-income; and Indonesia is low-income. We also try to homogenize countries by the type of financing in the economy: banking or financial market economies. We use stock market capitalization (in percent of GDP), domestic banking credit (in percent of GDP) and domestic banking credit to private sector (in percent of GDP).17 Our results are ambiguous, that is, we cannot conclude that the series are cointegrated. Sometimes there is a cointegrated relationship, for example, for middle-income economies, but not for the seven tests proposed by Pedroni (1997,1999). The fact that the series are not cointegrated can be deemed evidence that international capital flows are strong between these countries. Another piece of evidence might be intra-regional FDI flows. Guillaumin (2009) shows that FDI flows are high between these countries and have been higher since the financial crisis (Table 5). For the rest of the paper, we consider that series are not cointegrated. 5.3. The relationship between investment and saving 5.3.1. Methodology We propose to test the following equation:
D
I S ¼ ai ;t þ bi ;t D þ ei;t Y i ;t Y i ;t
17
The Annex presents the relationship between saving and investment rates for these sub-groups.
(17)
C. Guillaumin / Journal of Asian Economics 20 (2009) 314–326
322
With pooling, between and within models. Our presentation is based on Jansen (2000) and Be´reau (2007): I S D ¼ c þ bD þ ei;t Y i;t Y i;t I¯
D
Y
i
(18)
S¯ ¼ Db þ ei Y i
8 > I > > >
i;t
(19)
9 X S ¼ c þ bD þ d 1 þ ei;t Y i;t i¼1 i i
(20)
9 X > > > > di ¼ 0 : i¼1
with (18), (19) and (20) Pooling, Between and Within models. The pooling model is tested with OLS method on computing average data. This model supposes homogeneity across individuals. The between model is constructed as an average of saving and investment rates. We also use the OLS method. It does not take into account heterogeneity between countries but it eliminates the procyclical characters of saving and investment rates. Finally, the within model takes into account individual heterogeneity (temporal and individual). This way supposes fixed effects or random effects. Eq. (20) is a better way to take into account temporal information, as well as being a good way to P test strong or weak homogeneity of panel individuals. 9i¼1 di 1i , Eq. (20), is the sum of individuals fixed effects for each of the 9 countries of the panel. Results from pooling and within estimation give us the degree of homogeneity. 5.3.2. Results Table 6 shows the different estimates of the b coefficient. Our results show perfect capital mobility (perfect degree of financial integration) with the within model. The pooling and between models show strong but not perfect capital mobility. But the estimate of b is relatively weak. Besides, the pooling and within model estimates reveal heterogeneity between countries. Fixed coefficients in the within model show that our countries in the sample do no have not the same preferences in terms of saving and investment rates. We try to homogenize our panel by estimating a new relationship between saving and investment rates according to income level. For this classification, only the pooling and within models are estimated. The between model, in this case, is calculated as an average for four countries. Our results for the between model are less significant. Our results show that: (i) countries at a high-income level are more financially integrated than countries in the middle group; and (ii) sub-groups appear to be homogeneous because results of the pooling and within models are the same for each of them. 5.3.3. A financial crisis effect? In this subsection, we gauge whether the 1997 financial crisis had an effect on the financial integration process. That is, we want to know if there was an increase or a decrease of financial integration after 1997. We know that the causality
Table 6 Estimate of b coefficient. Model
b
All countries Pooling Between Within
0.213*** 0.390*** 0.069***
Middle-income countries Pooling Within
0.319** 0.380**
High-income countries Pooling Within
0.216* 0.202*
* ** ***
Significance at the 10% level. Significance at the 5% level. Significance at the 1% level.
C. Guillaumin / Journal of Asian Economics 20 (2009) 314–326
323
Table 7 Estimate of b coefficient. Periods 1988–1996 1998–2006 *
Pooling ***
0.53 0.28***
Within 0.07 0.00
Significance at the 10% level. **Significance at the 5% level. *** Significance at the 1% level.
between economic growth and financial liberalization may not be clear (e.g., Obstfeld, 1994), but the impact on the financial integration process is an open question. We proceed in two steps. First, we estimate the following model: 8 9 2X 005 X I S > > > D ¼ c þ bD þ di 1 i þ dt 1t þ ei;t > > Y Y > i;t i;t > t¼1988 i¼1 > > > i¼1 > > > 2X 005 > > > > dt ¼ 0 :
(21)
t¼1988
This model is a modified form of Eq. (20), except that we introduce, with fixed effects, a temporal effect. Here, 11997 = 1 for 1997 and 0 for all other years (11988 = 11989 = . . . =11996 = 11998 = . . . =12005 = 0). We decide to re-impose a fixed effect because we know that there is heterogeneity between the series. Our results show that 1997 is significant and has a higher effect than the average. In a second step, we consider two sub-periods (1988–1996 and 1998–2006). Table 7 shows the results of the estimation using the pooling and within models. With the pooling model, the coefficient is 0.53 for 1988–1996 and 0.28 for 1998–2006. With the within model, these coefficients are, respectively, 0.07 and 0.00 and they are insignificant in all cases. Therefore, our results suggest that the degree of financial integration is higher in the post-crisis period. How to explain this phenomenon? It is well acknowledged that the capital account liberalization policies and development of telecommunications and information technology have increased capital movement across countries in Asia. This phenomenon can be also explained by East Asian countries willing to develop regional financial market and to increase financial cooperation in order to avoid further financial crises. Also, for the within model, we note that there is no significant difference before and after the 1997 crisis. 6. Summary and conclusions This paper studies the degree of financial integration of nine East Asian countries from 1988 to 2006. We employ new panel unit root and panel cointegration techniques to estimate three types of modified Feldstein–Horioka equations: between, within and pooling. These different estimations show a high degree of financial integration. We try to homogenize our data and estimate a new relationship for high-income and middle-income countries. Our results show that high-income countries have stronger financial integration links than middle-income countries. Finally, we undertake stability tests in order to gauge if there is a crisis effect. We find that financial integration is stronger in the post-crisis period. Acknowledgements I am grateful to Claude Chambon, Virginie Coudert and Michael Plummer for their helpful comments. I am also grateful to Samuel Be´ji for research assistance. I thank participants of the CEPN lunch seminar and of the 18th conference of the ACAES Asian Economic Integration in a Global Context for highly profitable discussions and comments. I also thank the anonymous referee for helpful comments. All remaining errors are mine.
Appendix A. The relationship between saving and investment Fig. A1 shows the relationship between investment and saving rates from 1988 to 2006 for some East Asian countries. The figure on the right is made with the average of investment and saving rates (IR_AVE; SR_AVE), as proposed by Feldstein and Horioka (1980). Fig. A2 gives the same relationship, over the same period, according to the level of income (middle or high). Fig. A3 shows the relationship according to the type of financing in the economy.
324
C. Guillaumin / Journal of Asian Economics 20 (2009) 314–326
Fig. A1. The relationship between investment and saving rates. Source: World Bank, World Development Indicators.
Fig. A2. The relationship between investment and saving rates. (H: high-income countries; M: middle-income countries). Source: World Bank, World Development Indicators.
Fig. A3. The relationship between investment and saving rates. (Note: BANK: banking economy; FMARKET: financial market economy). Source: World Bank, World Development Indicators.
C. Guillaumin / Journal of Asian Economics 20 (2009) 314–326
325
Appendix B. Income level panel unit root tests Tables B1–B4. Table B1 First generation panel unit root rests: middle-income countries. Variables
Model
LLC
IPS ***
MW ***
Hadri ***
IR
Model without trend Model with trend
1.15 (0.1240) 1.08*** (0.1388)
0.45 (0.3257) 0.34*** (0.3635)
8.65 (0.3727) 9.77*** (0.2811)
2.33*** (0.0097) 3.36*** (0.0004)
SR
Model without trend Model with trend
1.50*** (0.0659) 1.48*** (0.0695)
1.30*** (0.0953) 0.27*** (0.3914)
13.18*** (0.1058) 8.44*** (0.3916)
2.57*** (0.0050) 3.09*** (0.0010)
Notes: IR and SR for Investment rate and Saving rate (in percent of GDP). Unit root hypothesis is accepted at ***1%, **5%, *10% levels. For Hadri’s test (2000), unit root hypothesis is the alternative hypothesis. LLC, IPS and MW for Levin et al. (2002), Im et al. (2003) and Maddala and Wu (1999) tests results. Values in parenthesis are p-value. All tests are made with fixed effects.
Table B2 Second generation panel unit root tests: middle-income countries. Variables
IR SR
Model without trend
Model with trend
CIPS1
CIPS2
CIPS3
CIPS1
CIPS2
CIPS3
1.48*** (0.685) 0.945*** (0.936
1.07*** (0.90) 0.429*** (0.994)
0.90*** (0.946) 0.203*** (0.998)
1.85*** (0.80) 0.794*** (0.998)
1.54*** (0.928) 0.002*** (1.000)
1.28*** (0.976) 0.121*** (1.000)
Notes: IR and SR for Investment rate and Saving rate (in percent of GDP). Unit root hypothesis is accepted at ***1%, **5%, *10% levels. CIPSi with i = 1, 2, 3 corresponds to the Pesaran statistics test (2003) for lag i. Values in parenthesis are p-value. All tests are made with fixed effects.
Table B3 First generation panel unit root tests: high-income countries. Variables
Model
LLC
IPS
MW
Hadri
IR
Model without trend Model with trend
0.79*** (0.7865) 1.96*** (0.0250)
1.14*** (0.8738) 1.42*** (0.0781)
2.81*** (0.9455) 18.33*** (0.0189)
2.34*** (0.0097) 4.42*** (0.0000)
SR
Model without trend Model with trend
0.51*** (0.3024) 1.19*** (0.1156)
0.07*** (0.8738) 1.11*** (0.1323)
7.95*** (0.9455) 12.43*** (0.1329)
2.68*** (0.0097) 3.72*** (0.0024)
Notes: IR and SR for Investment rate and Saving rate (in percent of GDP). Unit root hypothesis is accepted at ***1%, **5%, *10% levels. For Hadri’s test (2000), unit root hypothesis is the alternative hypothesis. LLC, IPS and MW for Levin et al. (2002), Im et al. (2003) and Maddala and Wu (1999) tests results. Values in parenthesis are p-value. All tests are made with fixed effects.
Table B4 Second generation panel unit root tests: high-income countries. Variables
Model without trend CIPS1
IR SR
Model with trend CIPS2
***
1.84 (0.413) 2.64 (0.038)
CIPS3 ***
1.46 (0.698) 1.48*** (0.689)
CIPS1 ***
1.34 (0.774) 0.104*** (0.999)
3.36 (0.016) 2.22*** (0.547)
CIPS2
CIPS3 ***
2.37 (0.431) 0.87*** (0.997)
1.82*** (0.820) 1.18*** (0.985)
Notes: IR and SR for Investment rate and Saving rate (in percent of GDP). Unit root hypothesis is accepted at ***1%, **5%, *10% levels. CIPSi with i = 1, 2, 3 corresponds to the Pesaran statistics test (2003) for lag i. Values in parenthesis are p-value. All tests are made with fixed effects.
References Araujo, C., Brun, J.-F. & Combes, J.-L. (2004). Econome´trie. Bre´al. Armstrong, H. W., Balasubramanyam, V. N., & Salisu, M. A. (1996). Domestic savings, intra-national and intra-European Union capital flows, 1971–1991. European Economic Review, 40(6), 1229–1235. Bae, K.-E., Bailey, W. & Yun, Y. S. (2006). Determinants of bond holdings by foreign investors (BIS papers 30). Bank for International Settlements. Baltagi, B. H. (1995). Econometric analysis of panel data. Willey. Bai, J. & Ng, S. (2001). A PANIC attack on unit root and cointegration (Boston College Working Paper in Economics, 469). The Johns Hopkins University. Bai, J., & Ng, S. (2004). A PANIC attack on unit root and cointegration. Econometrica, 72(4), 1127–1177. Bajou, C., Lefeuvre, E., Mamandz, S. & Melka, J. (2006). Relations commerciales en Asie. Recherche e´conomique, Ixis CIB. Banerjee, A. & Zanghieri, P. (2003). A new look at the Feldstein–Horioka puzzle using an integrated panel (Cepii working paper 2003/22). Bautista, C., & Maveyraud-Tricoire, S. (2008). Saving-investment relationship, financial crisis and structural changes in East Asian countries. Economie Internationale, 111(3), 81–100. Bayoumi, T. (1989). Saving-investment correlations: Immobile capital, government policy or endogeneous behaviour? (IMF working paper 89-66).
326
C. Guillaumin / Journal of Asian Economics 20 (2009) 314–326
Bayoumi, T., & Rose, A. (1993). Domestic savings and intra-national capital flows. European Economic Review, 37(6), 1197–1202. Beitone, A., Gilles, P. & Parodi, M. (2006). Histoire des faits e´conomiques et sociaux de 1945 a` nos jours, Dalloz. Ben Gamra, S., & Plihon, D. (2007). Qualite´ des institutions, libe´ralisation et crises bancaires: le cas des pays e´mergents. Economie Internationale, 112(4), 5–28. Be´reau, S. (2007). Une mesure macroe´conome´trique ‘‘a` la Feldstein–Horioka’’ du degre´ d’inte´gration financie`re en Europe’’. Economie Internationale, 110(2), 63– 106. Choi, I. (2001). Unit root test for panel data. Journal of International Money and Finance, 20(2), 249–272. Coakley, J., Kulasi, F., & Smith, R. (1996). Current account solvency and the Feldstein–Horioka puzzle. Economic Journal, 106(436), 620–627. Coiteux, M., & Olivier, S. (2000). The saving retention coefficient in the long run and in the short run: Evidence from panel data. Journal of International Money and Finance, 19(4), 535–548. Corbin, A. (2001). Specific effect in the Feldstein–Horioka paradox: A panel data analysis. Economics Letters, 72(3), 297–302. Davanne, O. (1998). Instabilite´ du syste`me financier international (Conseil d’analyse e´conomique, La Documentation franc¸aise). Eichengreen, B. (2004). Financial development in Asia: The way forward. mimeo. Eichengreen, B. (2006). The development of Asian bond markets (BIS Papers 30). Bank for International Settlements. Eichengreen, B. & Luengnaruemitchai, P. (2004). Why doesn’t Asia have bigger bond markets? (NBER working paper 10576). Feldstein, M., & Horioka, C. (1980). Domestic saving and international capital flows. Economic Journal, 90(358), 314–329. Fischer, R. A. (1932). Statistical methods for research workers. Edinburg: Oliver & Boyd. Flandreau, M., & Rivie`re, C. (1999). La Grande ‘‘Retransformation’’? Inte´gration financie`re internationale et controˆle des capitaux, 1880–1996. Economie Internationale, 78(2), 11–58. Frankel, J. (1992). Measuring international capital mobility: A review. American Economic Review, 82(2), 197–202. Fukasaku, K., & Martineau, D. (1999). Coope´ration mone´taire en Asie de l’Est: l’apport des tests de causalite´ et de la cointe´gration. Economie et Pre´vision, 140–141, 105–116. Gue´rin, J. & Sa, S. (2006). Les progre`s de l’inte´gration mone´taire et financie`re en Asie (Revue de la stabilite´ financie`re, 8, pp. 117–135). Banque de France. Guillaumin, C. (2009). (A)syme´trie et convergence des chocs macroe´conomiques en Asie de l’Est: une approche dynamique. Economie Internationale, 114(2), 29– 68. Gyntelberg, J., Ma, G. & Remolona, E. (2006). Developing corporate bond markets in Asia (BIS papers 26). Bank for International Settlements. Hadri, K. (2000). Testing for stationarity in Heterogeneous Panel Data. Econometrics Journal, 3(2), 148–161. He´ricourt, J. (2005). Essai sur la pratique de la politique mone´taire sous diffe´rents re´gimes mone´taires: de la zone euro a` la Turquie. The`se de Doctorat. Universite´ Paris IPanthe´on Sorbonne. He´ricourt, J. & Maurel, M. (2005). A new look at Feldstein-Horioka puzzle: An ‘‘European Regional’’ perspective. cahiers de la MSE 05070. Ho, T.-W. (2002). The Feldstein–Horioka puzzle revisited. Journal of International Money and Finance, 21(4), 555–564. Hsiao, C. (2003). Analysis of panel data. Cambridge University Press. Hurlin, C., & Mignon, V. (2005). Une synthe`se des tests de racine unitaire sur donne´es de panel. Economie et Pre´vision, 169, 253–294. Hurlin, C., & Mignon, V. (2007). Une synthe`se des tests de cointe´gration sur donne´es de panel. Economie et Pre´vision, 180–181, 241–265. Im, K., Pesaran, M., & Shin, Y. (2003). Testing for unit roots in heterogeneous panels. Journal of Econometrics, 115(1), 53–74. Isaksson, A. (2001). Financial liberalisation, foreign aid, and capital mobility: Evidence from 90 developing countries. Journal of International Financial Markets, 11(3–4), 309–338. Ito, T. (2001). Growth, crisis, and the future of economic recovery in East Asia. In J. Stiglitz & S. Yusuf (Eds.), The east Asian miracle. Oxford University Press. Iwamoto, Y., & van Wincoop, E. (2000). Do borders matter? Evidence from Japanese International Capital flows. International Economic Review, 41(1), 241–269. Jansen, W. (1996). Estimating saving-investment correlations: Evidence for OECD countries based on an error correction model. Journal of International Money and Finance, 15(5), 749–781. Jansen, W. (2000). International capital mobility: Evidence from panel data. Journal of International Money and Finance, 19(4), 507–511. Kao, C., & Chiang, M.-H. (2001). On the estimation and inference of a cointegrated regression in panel data. In Baltagi, B. (Ed.). Advances in Economics. vol. 15 (pp.179–222). Kim, S. (2001). The Saving-investment correlation puzzle is still a puzzle. Journal of International Money and Finance, 20(7), 1017–1034. Kim, S., Kim, H. S., & Wang, Y. (2007). Saving, investment and international capital mobility in East Asia. Japan and the World Economy, 19(2), 279–291. Kim, H., Oh, K.-Y., & Jeong, C.-W. (2005). Panel cointegration results on international capital mobility in Asian countries. Journal of International Money and Finance, 24(1), 71–82. Krol, R. (1996). International capital mobility: Evidence from panel data. Journal of International Money and Finance, 15(3), 467–474. Kwiatkowski, D., Phillips, P., Schmidt, P., & Shin, Y. (1992). Testing the null hypothesis of stationarity against the alternative of unit root. Journal of Econometrics, 54(1–3), 159–178. Levin, A. & Lin, C. F. (1992). Unit root tests in panel data: Asymptotic and finite sample properties (discussion paper, 56). Department of Economics, University of California at San Diego. Levin, A., Lin, C. F., & Chu, J. (2002). Unit root tests in panel data: Asymptotic and finite sample properties. Journal of Econometrics, 108(1), 1–24. Maddala, G., & Wu, S. (1999). A comparative study of unit root tests and a new simple test. Oxford Bulletin of Economics and Statistics, 61(special issue), 631–652. Menkhoff, L., & Suwanaporn, C. (2007). 10 Years after the crisis: Thailand’s financial system reform. Journal of Asian Economics, 18(1), 4–20. Moneta, F., & Ru¨ffer, R. (2009). Business cycle synchronisation in East Asia. Journal of Asian Economics, 20(1), 1–12. Obstfeld, M. (1994). Are industrial country consumption risks globally diversified (NBER working paper 4308). Obstfeld, M. & Taylor, A. M. (2002). Globalization and capital markets (NBER working paper 8846). Park, D., & Rhee, C. (2006). Building infrastructure for Asian bond markets: Settlement and credit rating. In D. Arner, J.-H. Park, P. Lejot, & Q. Liu (Eds.), Asia’s Debt Capital Markets. Springer. Pedroni, P. (1997). Panel cointegration, asymptotic and finite sample properties of pooled time series with an application to the PPP hypothesis: New results (Working paper in Economics). Indiana University. Pedroni, P. (1999). Critical values for cointegration tests in heterogeneous panels with multiple regressors. Oxford Bulletin of Economics and Statistics, 61(Special issue), 653–670. Pesaran, H. M. (2003). A simple panel unit root test in the presence of cross section dependence. Mimeo. University of Southern California. Phillips, P., & Sul, D. (2003). Dynamic panel estimation and homogeneity testing under cross section dependence. Econometrics Journal, 6(1), 217–259. Proble`mes e´conomiques (1997). Le syste`me financier international. Nume´ro double 2541–2542. Salanie´, B. (1999). Guide pratique des se´ries non stationnaires. Economie et Pre´vision, 137, 119–141. Sinha, D. (2002). Saving-investment relationships for Japan and other Asian countries. Japan and the World Economy, 14(1), 1–23. Sinn, S. (1992). Saving-Investment correlations and capital mobility: On the evidence from annual data. Economic Journal, 102(414), 1162–1170. Stiglitz, J., & Yusuf, S. (2001). Rethinking the East Asian Miracle. Oxford University Press. Summers, L. (1989). Tax Policy and International Competitiveness (NBER working paper 2007). Sundaresan, S. (2006). Developing multiple layers of financial intermediation: The complementary roles of corporate bond markets and banks (BIS papers 26). Bank for International Settlements. Takeuchi, A (2006). Identifying impediments to cross-border bond investment and issuance in Asian countries (BIS papers 30). Bank for International Settlements. van Wincoop, E. (2001). Intranational versus international saving-investment comovements. In G. Hess & E. van Wincoop (Eds.), Intranational Macroeconomics. Cambridge University Press.
Contents lists available at ScienceDirect
Journal of Asian Economics
Financial integration in East Asia: Evidence from panel unit root and panel cointegration tests Cyriac Guillaumin * CEPN, University of Paris 13, 99 Avenue Jean-Baptiste Cle´ment, 93430 Villetaneuse, France
A R T I C L E I N F O
A B S T R A C T
JEL classification: C33 F32 F36 F41
The aim of this paper is to investigate the degree of financial integration for selected East Asian countries from 1988 to 2006 using the recently developed panel unit root and panel cointegration techniques. Investment and savings rates are found to be nonstationary and not to be cointegrated in panels. We estimate modified Feldstein–Horioka equations and our results reveal a high degree of financial integration. When we homogenize our data, results show that high-income countries have stronger financial integration than middleincome countries. Finally, we proceed to stability tests in order to test if there is a crisis effect and we find that financial integration is stronger in the post-crisis period. ß 2009 Elsevier Inc. All rights reserved.
Keywords: Financial integration Feldstein–Horioka puzzle Nonstationary panel data Panel cointegration East Asia
1. Introduction In their paper, Feldstein and Horioka (1980) measure the degree of financial integration using the correlation of domestic saving with investment rates. Their results show a high correlation during 1960–1974. This result is contradictory with the stylized fact of this period; indeed, capital mobility appears to be increasing ever since the beginning of the 1960s (Beitone, Gilles, & Parodi 2006; Flandreau & Rivie`re, 1999). The Feldstein–Horioka approach is subject to two critiques that focus either on the econometric and statistical caveats – responsible for a bias towards a significant, positive and close to one coefficient – or on the economic and historical circumstances that explain the discrepancy between an expected coefficient close to zero and its effective value. The goal of this paper is to investigate the degree of financial integration in selected East Asian countries from 1988 to 2006 using the recently developed panel unit root and panel cointegration techniques. Ho (2002), Kim, Oh, and Jeong (2005) and Be´reau (2007) have used these techniques but only Kim et al. (2005) focused on Asian countries. Kim et al. (2005) study capital mobility in 10 countries (Korea, Indonesia, Japan, Malaysia, Myanmar, Pakistan, Philippines, Singapore, Sri Lanka and Thailand) between 1960 and 1998. They find weaker financial integration during 1960–1979 than 1980–1998. The evolution of financial integration can be explained by financial liberalization of these economies in the 1980s and 1990s. But the study of Kim et al. (2005) does not integrate the last 10 years. Moreover, the second-generation panel unit root tests based on the hypothesis of interdependence between individuals are not used by any of the earlier studies. We employ first-generation panel unit root tests along the lines of Levin and Lin (1992), Im, Pesaran, and Shin (2003), Maddala and Wu (1999) and Hadri (2000), and second-generation panel unit root tests along the lines of Pesaran (2003). We
* Tel.: +33 1 49 40 32 55; fax: +33 1 49 40 33 34. E-mail address: [email protected]. 1049-0078/$ – see front matter ß 2009 Elsevier Inc. All rights reserved. doi:10.1016/j.asieco.2009.02.002
C. Guillaumin / Journal of Asian Economics 20 (2009) 314–326
315
also test for the presence of a long term relationship between investment and saving rates using Pedroni’s cointegration test (1997, 1999). We estimate three types of modified Feldstein–Horioka equations: between, within and pooling. These different estimations show a high degree of financial integration. We try to homogenize our data and estimate a new relation for highincome and middle-income countries. Our results show that high-income countries have stronger financial integration than middle-income countries. Finally, we proceed to stability tests in order to test if there is a crisis effect. Our results support the idea that financial integration is stronger in the post-crisis period. The paper is structured as follows: Section 2 provides an overview of the Feldstein–Horioka puzzle. Section 3 describes the methodology. Section 4 provides a survey of the evolution of financial integration in East Asia. In Section 5 we present and discuss results. Section 6 offers concluding remarks. 2. Feldstein–Horioka puzzle: an econometric solution? 2.1. The Feldstein–Horioka paradox In their pioneering study, Feldstein and Horioka (1980) examined the cross-sectional correlation between saving and investment by testing the following equation: I S ¼aþb þ ei;t Y i;t Y i;t
(1)
where I/Y represents the ratio of investment over Gross Domestic Product (GDP), i.e. the investment rate. S/Y represents the ratio of saving over GDP, i.e. the saving rate. Indexes i and t respectively stand for the considered country and year. ei,t represents other factors explaining investment. This term is a random walk, with zero mean (E(e) = 0) and variance s2. Feldstein and Horioka work on average data in order to prevent a bias: I¯ Y
i
S¯ ¼aþb þ ei Y i
with
I¯ Y
i
¼
974 1 1X I ; T t¼1960 Y i;t
S Y
! ¼ i
974 1 1X S T t¼1960 Y i;t
and
T ¼ 15:
0
(1 )
Coefficient b indicates the degree of financial integration: in case of weak financial integration, b is equal to 1 and to 0 in case of a strong financial integration. Feldstein and Horioka (1980) test Eq. (10 ) for 16 OECD countries during 1960–1974. The result however pointed to a quite low degree of financial integration despite a large volume of international capital movements (Beitone et al., 2006; Flandreau & Rivie`re, 1999). Since the 1970s, capital mobility has been increasing but it is hard to measure this phenomena. Still, some papers (Davanne, 1998; Obstfeld & Taylor, 2002) indicate that the daily financial transactions were equal to US$ 1,500 billion in 1999, i.e., 50 times more than daily commercial trade. In the 1970s, this figure was in the range of US$ 10–20 billion. One explanation, among others, is the emergence of financial markets at the end of the 1970s. World stock market capitalization was US$ 1400 billion in 1975 and US$ 17000 billion in 1997 (Proble`mes e´conomiques, 1997). Explaining the Feldstein– Horioka puzzle in the literature has been a matter of either econometrics or of economic analysis. 2.2. Solving the Feldstein–Horioka enigma The economic arguments are based on historical circumstances that influence capital mobility (Flandreau & Rivie`re, 1999; Obstfeld & Taylor, 2002). Therefore, financial integration can be high or low according to the precise periods and to the process of financial liberalization. Besides, these arguments try to explain Feldstein–Horioka enigma by the relationship between investment and saving with variables that can influence the monetary conditions in the studied area. Economic and historical caveats have been studied by many authors: Bayoumi (1989), Frankel (1992), and Armstrong, Balasubramanyam, and Salisu (1996) are examples. None of these papers accepts perfect capital mobility. Bayoumi (1989) explains the strong relationship between investment and saving due to fiscal and monetary policy. For example, in case of a lack of saving, the government can modify its fiscal policy (decreases in taxes or increases in interest rates) and create a new link between investment and saving. In fact, l’effet taille can modify the relationship between investment and saving: a big country, in economic terms, can influence the world interest rate. So an increase in domestic savings will lead to a decrease in interest rates and to an increase in investment. In addition, Bayoumi (1989), according to Summers (1989), notes that as saving and investment are procyclical variables, their correlation can result from a common response to permanent shocks. Moreover, others variables can influence financial integration, such as the interest rate or risk premium (Frankel, 1992). To correct for this bias, a regional approach is better than a national approach, proposed by Sinn (1992), Bayoumi and Rose (1993), Armstrong et al. (1996), Iwamoto and van Wincoop (2000) and van Wincoop (2001).1 The goal of this approach is to study financially-integrated countries. The econometric and statistical critiques focus on statistical properties of variables used in estimating Eq. (1). If saving and investment are not stationary and not cointegrated, Eq. (1) is a spurious regression. The interpretation of b is impossible.
1
See, for example, He´ricourt (2005) or He´ricourt and Maurel (2005) for a synthesis.
C. Guillaumin / Journal of Asian Economics 20 (2009) 314–326
316
A plausible long-term investment-savings relationship is tested by Coakley, Kulasi, and Smith (1996). This relationship can explain the ambiguous results of Feldstein and Horioka (1980) because the estimate of the b parameter leads to a figure close to 1. But this figure does not reflect a lack of financial integration. Coakley et al. (1996) show that a cointegration relationship prevents an accurate measure of the degree of financial integration. To correct for this problem, times series and panel data econometric techniques are employed. Some papers try to mix these two approaches (Banerjee & Zanghieri, 2003).2 2.3. Times series econometrics Sinn (1992) and Jansen (1996), for example, use times series econometrics to solve the Feldstein–Horioka enigma. They integrate nonstationary data but also plausible cointegration relationships. The authors introduce current account data (CA) as well. Economically, if there is a cointegration relationship between investment and saving, there is a long-term solvability constraint of the current account. By definition, CA = S I. With this long-term solvability constraint, the current account is equilibrated. Hence, E(CA) = 0, where E is the mean. From this, CA = S I is a stationary condition, implying that investment and saving are cointegrated. VAR or VECM models (in case of cointegration) investigate the degree of financial integration. These approaches are very different from Coakley et al. (1996) where cointegrated variables lead to spurious results. 2.4. Panel data econometrics Krol (1996) was the first author to use panel data in this context. This method partially solves the Feldstein–Horioka paradox. Krol (1996) studies the degree of financial integration for the same OECD countries studied by Feldstein–Horioka (1980), using panel data. He analyses jointly individual and time dimensions of saving and investment data. The use of panel data allows for the introduction of fixed and temporal effects.3 Thus, Krol (1996) tests Feldstein–Horioka using the following equation: I S ¼mþb þ ui ;t Y i ;t Y i ;t
(2)
with ui,t = ai + lt + ei,t, where ai is an individual effect (fixed or random), lt is the temporal effect (for all panel data) and ei,t is a random term, with zero mean and variance s2. His results show perfect capital mobility between countries. Jansen (1996) contests these results due to the presence of Luxembourg.4 But Ho (2002), using the two-stage least squares approach proposed by Kao and Chiang (2001), shows that Luxembourg does not affect the results. Nevertheless, panel data econometrics are insufficient to capture effectively financial integration as well as macroeconomics phenomena (Araujo and alii, 2004; Hurlin & Mignon, 2005). We argue that it is necessary to take into account new panel data econometrics with nonstationary and cointegrated data. 2.5. Nonstationary panel data econometrics In this paper, we investigate the degree of financial integration of selected East Asian countries from 1988 to 2006 using the Feldstein–Horioka (1980) approach but with new panel data econometric tools. The economic and historical circumstances influencing relevant variables have been the object of a large number of papers. While the work of Bayoumi and Rose (1993), Armstrong et al. (1996), van Wincoop (2001) or Iwamoto and van Wincoop (2000) adopt interesting approaches, we do not have sufficient data for Asian countries to undertake them. These difficulties are increased with the definition of saving and investment rates at a regional level. Considering the above econometric critiques, we note that a times-series approach seems appropriate but that panel data econometrics have two advantages5: First, panel data have two dimensions, individual and temporal. This double dimension allows us to study simultaneously the dynamics and heterogeneity of individuals. Second, the double dimension allows us to eliminate a difficulty in the times series: the weakness of unit root and cointegration tests with a limited dataset.6 3. Methodology 3.1. Panel unit root tests In this section, we present our methodology. Our presentation is based on Araujo, Brun and Combes (2004) and Hurlin and Mignon (2005).
2 3 4 5 6
See Be´reau (2007) for a synthesis of this literature. Coiteux and Olivier (2000), Jansen (2000) and Corbin (2001) investigate financial integration following the Krol (1996) approach. See Be´reau (2007) for details about Luxembourg. See, for example, Baltagi (1995), Hsiao (2003) or Araujo and alii (2004). See Salanie´ (1999) for a discussion on this subject.
C. Guillaumin / Journal of Asian Economics 20 (2009) 314–326
317
3.2. First-generation panel unit root tests Four first-generation panel unit root tests (discussed above) are used in our paper. All tests are based on the following Augmented Dickey–Fuller regression:
Dxit ¼ ai þ bi t þ ri xit1 þ eit
(3)
The error term is eit and is IID. ai and bi allow for fixed and unit specific time trends for each i. The null hypothesis of a unit root is: H0 : ri ¼ 0; 8 i
(4)
The alternative hypothesis is different according to the test adopted. Levin and Lin (1992) consider an alternative hypothesis where the autoregressive coefficient is homogeneous across countries: Ha : ri ¼ r < 0; 8 i
(5)
They suppose that inter-individual heterogeneity is captured by the fixed effects. Im et al. (2003) extend the Levin and Lin (1992) framework to allow heterogeneity in the value of the autoregressive coefficient. The alternative hypothesis is then specified as: Ha : ri < 0; i ¼ 1; :::; N 1
and
ri ¼ 0; i ¼ N 1 þ 1; :::; N
(6)
In this test, we have individual i = 1, . . ., N1 where xit is stationary and individual i = N1 + 1, . . ., N, where xit is nonstationary. The t-statistic is an average and can be calculated as: t¯ ¼
N 1X t N 1 i
(7)
where ti is the individual ADF t-statistics for the unit root test. Maddala and Wu (1999), relying on Fischer (1932), suggest combining the p-values of the test statistic for a unit root in each cross-sectional unit. The statistic of Maddala and Wu is given by: N X MW ¼ 2 lnð pi Þ
(8)
i¼1
where pi is the p-value of the test statistic of the ADF test statistic in each country. Finally, in the case of the Hadri (2000) approach, the null hypothesis is the stationarity of the series instead of nonstationarity. The framework is the one dealt with in KPSS7 for times series. The model may now be expressed as: xit ¼ r it þ bi t þ eit
(9)
where rit is a random walk given by: r it ¼ r it1 þ uit where uit is white noise. We may assume that E(uit) = 0 and 2 u
H0 : s ¼ 0
(10) Eðu2it Þ
¼s
2 uit
> 0. The null hypothesis is: (11)
3.3. Second-generation panel unit root tests Second-generation panel unit root tests are based on the interdependence between individuals. Numerous second-generation panel unit root tests are proposed in the literature,8 e.g., Choi (2001), Pesaran (2003), Phillips and Sul (2003). Most of them are based on the Bai and Ng (2001), Bai and Ng (2004) methodology, which posit models with common factors. This test considers two separate unit root tests on common and individual components of the series. Other tests are based on a unique test of the unit root of the series. The difference is based on the decomposition of the series. We have decided to present only the Pesaran (2003) test, which focusing on the series xit corrected by the individual average of xit1 and of first differences Dxit1. Pesaran (2003) obtains a Cross Sectionally Augmented Dickey–Fuller (CADF) test. This model, inspired by the test developed in Im et al. (2003), is expressed as:
Dxit ¼ ai þ ri xit1 þ eit 7 8
Kwiatkowski et al. (1992). See Hurlin and Mignon (2005) for details.
(12)
C. Guillaumin / Journal of Asian Economics 20 (2009) 314–326
318
where eit = giut + uit. ut is a common factor and uit is white noise. The CADF model is given by, without autocorrelation of uit:
Dxit ¼ ai þ ri xit þ ci x¯ t1 þ di Dx¯ t1 þ vit
(13)
The Pesaran statistic, Cross-Sectionally Augmented IPS (CIPS) is given by: CIPSðN; TÞ ¼
N 1X t ðN; TÞ N i¼1 i
(14)
where ti constitutes statistics coming from each CADF model for each individual i of the panel. 3.4. Panel cointegration tests The next part of the process involves testing whether there is a statistically acceptable cointegration relationship between the variables of interest. This is achieved by applying the tests developed by Pedroni (1997,1999), which develop seven tests for cointegration in a panel context. Four of the tests are within-dimension statistics and three are between-dimension statistics. The four within–dimension statistics are based on pooling the autoregressive coefficients across the different countries for the unit root tests on the estimated residuals, whereas the three between-dimension statistics are based on estimators that simply average the individual estimated coefficients for each country. For all seven tests, the null hypothesis is no cointegration. The Pedroni cointegration tests are based on estimating the static cointegration regression given by: yit ¼ ai þ bi t þ b1i x1it þ b2i x2it þ ::: þ bMi xMit þ eit
(15)
where i = 1, . . ., N, t = 1, . . ., T and m = 1, . . ., M. ri is the autoregressive term of the residuals given by:
eˆ it ¼ ri eˆ it1 þ uit
(16)
The specification of the alternative hypothesis for the within-dimension (pooled) estimation is: Ha:ri = r < 1, 8i, whereas for the between-dimension (heterogeneous) estimation, the alternative hypothesis is given by Ha:ri < 1,8i. Pedroni’s procedure requires five steps. After these steps, we can construct the statistics for all the seven tests.9 4. Financial integration in East Asia: a brief review The weight of Asia in world trade has been increasing for many years (Gue´rin & Sa, 2006). This increase is essentially intrazone (Bajou, Lefeuvre, Mamandz & Melka, 2006; Guillaumin, 2009). Although the financial links developed intensified after the financial crisis of 1997, they remain weaker than the real-sector links. Even after 1997, financial markets continue to prefer banks in order to finance growth (Eichengreen, 2006; Gyntelberg, Ma & Remolona, 2006), despite the fact that a onepillar financial system dominated by banks is problematic because Asian banks have had a lack of governance10 and have been criticised for ‘‘crony capitalism,’’ that is, conflict of interest issues due to the interaction of private and public interests.11 Two initiatives were developed to increase the share of financial markets: the Asian Bond Market Initiative and the Asian Bond Funds. The Asian Bond Market Initiative (ABMI) of the ASEAN+312 countries aims to develop robust bond markets in the region; avoid the double mismatch of maturity and currency that exacerbated the East Asian crisis; recycle Asian savings to Asian economies; and increase regional financial integration. The Asian Bond Funds (ABF) of the Executives’ Meeting of East Asia-Pacific Central Banks (EMEAP) aims to develop demand for bond denominated in local currency (Gue´rin & Sa, 2006). Nevertheless, there is a lack of pecuniary resources for these two initiatives (Gue´rin & Sa, 2006; Takeuchi, 2006). For example, the ABF has US$ 1 billion invested in US bonds by Asian countries except for Japan. According to Eichengreen (2004, 2006), the slow development of financial market is explained by: - Regional characteristics: no diversity in financial institutions; - Weak supervision and regulation. There is no banking competition, rating agency and no compensation; - Macroeconomics policies are not in favour of investment: high interest rates, and volatility and exchange-rate risks are problematic.
9
See Hurlin and Mignon (2007) for details. Ben Gamra and Plihon (2007) show that a weak quality of institutions for emerging countries increases the probability of a banking crisis. 11 Menkhoff and Suwanaporn (2007) remind that a significant number of reform have been made, and especially for Thailand, on banking regulation and supervision. Even if Thailand is also a country essentially where growth is financing by banks, financial market share increases. 12 Asean+3 is made of the following countries: Indonesia, Malaysia, Singapore, Thailand, Philippines, Brunei, Vietnam, Myanmar, Laos, Cambodia, Timor and China, Japan, Korea. 10
C. Guillaumin / Journal of Asian Economics 20 (2009) 314–326
319
Table 1 Investment and saving rates (in percent of GDP). Average on the period 1988–2006. Countries
I/Y
S/Y
(S I)/Y
China South Korea Hong Kong SAR Indonesia Japan Malaysia Philippines Singapore Thailand
32.56 33.37 26.64 24.55 28.23 31.34 20.61 32.25 31.94
38.58 35.33 32.30 29.92 29.97 40.75 17.07 46.44 33.67
6.01 1.96 5.66 5.37 1.74 9.41 3.54 14.19 1.73
East Asia and Pacific European Monetary Union USA
30.91 20.85 18.04
35.75 22.26 16.36
4.84 1.41 1.68
Notes: East Asia and Pacific contains the following countries: Australia, Brunei Darussalam, the Kingdom of Cambodia, China, Fiji, French Polynesia, Guam, Hong Kong SAR, Indonesia, Japan, Kiribati, North Korea, South Korea, Laos, Macau SAR, Malaysia, Marshal Islands, Micronesia, Mongolia, Burma, New Caledonia, New Zealand, American Samoa, North Marina Islands, Palau, Papua New Guinea, Philippines, Samoa, Singapore, Salomon Islands, Thailand, Timor, the Kingdom of Tonga, Vanuatu and Vietnam. Source: World Development Indicators, World Bank.
Takeushi (2006) notes that there is considerable disparity across financial markets in the region with respect to capital mobility, taxes and liquidity. Institutional issues are important factors and constrain the system (Bae, Bailey & Yun, 2006; Eichengreen & Luengnaruemitchai, 2004). Possible solutions that have been explored and tabled are many, including the development of public bond markets (Gyntelberg and alii, 2006; Sundaresan, 2006) and creating different rating agencies (Park & Rhee, 2006). 5. Results 5.1. Data We investigate the degree of financial integration for 9 East Asian countries: China, Korea, Hong Kong SAR, Indonesia, Japan, Malaysia, Philippines, Singapore and Thailand. This choice is justified by economic (Kim, 2001; Moneta & Ru¨ffer, 2009) and commercial (Bajou et al., 2006; Guillaumin, 2009) links between these countries. Taiwan, Vietnam, Laos, Myanmar and Cambodia are not considered in the estimations because we do not have data for them. We consider the series of investment and saving rates (that is, as a percentage of GDP). Data come from World Development Indicators of the World Bank. The starting date for these series is 1988 (until 2006) measured at an annual frequency. We begin our investigation in 1988 because the process of financial liberalization started at the beginning of the 1980s in Asia but the year 1988 is generally held by the literature to be when integration began its acceleration (Fukasaku & Martineau, 1999). Many authors have worked on the degree of financial integration in East Asia using Feldstein–Horioka measure – e.g., Sinha (2002) and Kim, Kim, and Wang (2007) using times series, and Isaksson (2001) and Kim et al. (2005) with panel data – but these papers tend to begin with data from the 1960s until the Asian financial crisis.13 Given the tend in financial integration, years after the financial crisis are of great importance. Moreover, we are able to use new tools in panel data econometrics.14 Table 1 shows investment and saving rates in these countries. From 1988 to 2006, the average saving rate was 34% with a high variance: Singapore has a saving rate of 47% and the Philippines, only 17%. The average investment rate is 29% with a lower variance, from 33% (Korea) to 20% (the Philippines). We can see that all countries, except the Philippines, have higher saving rates than their investment rates. These countries have, therefore, a current account surplus while the United States has a deficit. For East Asian countries overall this surplus of saving explains in part the ‘‘miraculous’’ growth in the 1980–1990s (Ito, 2001; Stiglitz and Yusuf, 2001). Fig. 1 shows the relationship between investment and saving rates for each of the 9 countries.15 We can observe a degree of heterogeneity in the evolution of saving and investment rates. Clearly, the financial crisis had an asymmetric impact on the East Asian economies. For all countries, except China, we observe a decrease in investment and saving rates in 1997–1998. For Korea, Malaysia and Thailand, it is a collapse. 5.2. Unit root and cointegration tests results 5.2.1. Unit root tests Tables 2 and 3 summarize panel unit root tests (first- and second-generation).16 According to these tables, the unit root hypothesis is accepted in a large number of models at a 1% level of significance. There are some exceptions. For
13 14 15 16
See, for example, Bautista and Maveyraud-Tricoire (2008) for a synthesis. Besides, only Kim et al. (2005) and Kim et al. (2007) concern East Asian countries. Appendix A presents the relationship between investment and saving rates as proposed by Feldstein and Horioka (1980). Appendix B presents results of panel unit root tests according to the level of income.
C. Guillaumin / Journal of Asian Economics 20 (2009) 314–326
320
Fig. 1. Investment (IR) and saving (SR) rates (percent of GDP). Source: World Bank, World Development Indicators.
example, the null hypothesis of a unit root is rejected, at the 10% level, for saving rates with the Lin, Levin and Chu test (models with and without trend). Models with trend do not accept unit root test for saving (5% level) and investment (10% level) rates for the Maddala Wu test. Finally, the Im, Pesaran and Shin test rejects a unit root for saving rate at 10% (model with trend).
Table 2 First generation panel unit root tests. Variables
Model
LLC
IPS
MW
Hadri
IR
Model without trend Model with trend
0.20*** (0.4189) 2.22*** (0.0130)
0.42*** (0.6660) 1.23*** (0.1087)
13.26*** (0.7754) 30.04*** (0.0372)
3.54*** (0.0002) 5.39*** (0.0000)
SR
Model without trend Model with trend
1.52*** (0.0633) 2.19*** (0.0140)
1.06*** (0.1433) 1.51*** (0.0655)
24.36*** (0.1434) 27.09*** (0.0773)
4.37*** (0.0000) 5.36*** (0.0000)
Notes: IR and SR for Investment rate and Saving rate (in percent of GDP). Unit root hypothesis is accepted at *** 1%, **5%, *10% levels. For Hadri’s test (2000), unit root hypothesis is the alternative hypothesis. LLC, IPS and MW for Levin, Lin, and Chu (2002), Im et al. (2003) and Maddala and Wu (1999) tests results. Values in parenthesis are p-value. All tests are made with fixed effects.
C. Guillaumin / Journal of Asian Economics 20 (2009) 314–326
321
Table 3 Second generation panel unit root tests. Variables
IR SR
Models without trend
Models with trend
CIPS1
CIPS2
CIPS3
CIPS1
CIPS2
CIPS3
2.02*** (0.20) 1.93*** (0.277)
2.21*** (0.07) 1.36*** (0.859)
1.30*** (0.894) 1.51*** (0.733)
2.33*** (0.435) 2.58*** (0.185)
2.93 (0.626) 1.51*** (0.985)
2.69 (0.107) 0.571*** (1.000)
Notes: IR and SR for Investment rate and Saving rate (in percent of GDP). Unit root hypothesis is accepted at ***1%, **5%, *10% levels. CIPSi with i = 1, 2, 3 corresponds to the Pesaran statistics test (2003) for lag i. Values in parenthesis are p-value. All tests are made with fixed effects.
Table 4 Panel cointegration test. Totala
Sample
Middle-income **
Panel v-stat Panel rho-stat Panel pp-stat Panel adf Group rho-stat Group pp-stat Group adf
***
Banking
***
3.78 2.27** 2.18** 2.55*** 0.18 2.18 1.34
1.87 1.32* 1.56* 1.06 0.64 1.63 2.53
High-income
Financial Market
**
4.17 1.44* 1.56** 0.78 1.33 1.26 1.07
2.17 1.30* 0.99 1.04 0.99 1.28 0.62
1.01 0.78 2.99*** 0.15 1.70** 1.28 1.99**
Notes: the hypothesis of no cointegation is rejected at ***1%, **5%, *10%. The four within-dimension statistics are based on pooling the autoregressive coefficients across the different countries for the unit root tests on the estimated residuals. Whereas the three between-dimension statistics are based on estimators that simply average the individual estimated coefficients for each country. a All countries in the panel. Table 5 FDI flows. Inflows
Outflows
1990–1996 China Singapore Hong Kong Malaysia Indonesia South Korea
1997–2006 20.7 5.9 4.6 3.6 2.7 2.3
China Hong Kong Singapore South Korea Indonesia Thailand
1990–1996 50.9 25.5 13.6 5.7 4.4 3.6
Hong Kong Singapore Taiwan China South Korea Malaysia
1997–2006 14.9 3.6 3.0 2.3 2.2 1.4
Hong Kong Singapore Taiwan South Korea China Malaysia
25.8 7.4 5.5 3.9 3.0 1.7
Note: billion US dollars. Source: UNCTAD FDI/TNC database.
5.2.2. Cointegration test The results of the cointegration analysis tests over the period 1988–2006 are presented in Table 4. We also test cointegration for sub-groups of countries in the panel as recommended by Banerjee and Zanghieri (2003). We study countries according to income level using the classification of the World Bank (gross national income). According to this classification, China, Malaysia, Philippines and Thailand are countries with middle-income; Korea, Hong Kong SAR, Japan and Singapore are countries with high-income; and Indonesia is low-income. We also try to homogenize countries by the type of financing in the economy: banking or financial market economies. We use stock market capitalization (in percent of GDP), domestic banking credit (in percent of GDP) and domestic banking credit to private sector (in percent of GDP).17 Our results are ambiguous, that is, we cannot conclude that the series are cointegrated. Sometimes there is a cointegrated relationship, for example, for middle-income economies, but not for the seven tests proposed by Pedroni (1997,1999). The fact that the series are not cointegrated can be deemed evidence that international capital flows are strong between these countries. Another piece of evidence might be intra-regional FDI flows. Guillaumin (2009) shows that FDI flows are high between these countries and have been higher since the financial crisis (Table 5). For the rest of the paper, we consider that series are not cointegrated. 5.3. The relationship between investment and saving 5.3.1. Methodology We propose to test the following equation:
D
I S ¼ ai ;t þ bi ;t D þ ei;t Y i ;t Y i ;t
17
The Annex presents the relationship between saving and investment rates for these sub-groups.
(17)
C. Guillaumin / Journal of Asian Economics 20 (2009) 314–326
322
With pooling, between and within models. Our presentation is based on Jansen (2000) and Be´reau (2007): I S D ¼ c þ bD þ ei;t Y i;t Y i;t I¯
D
Y
i
(18)
S¯ ¼ Db þ ei Y i
8 > I > > >
i;t
(19)
9 X S ¼ c þ bD þ d 1 þ ei;t Y i;t i¼1 i i
(20)
9 X > > > > di ¼ 0 : i¼1
with (18), (19) and (20) Pooling, Between and Within models. The pooling model is tested with OLS method on computing average data. This model supposes homogeneity across individuals. The between model is constructed as an average of saving and investment rates. We also use the OLS method. It does not take into account heterogeneity between countries but it eliminates the procyclical characters of saving and investment rates. Finally, the within model takes into account individual heterogeneity (temporal and individual). This way supposes fixed effects or random effects. Eq. (20) is a better way to take into account temporal information, as well as being a good way to P test strong or weak homogeneity of panel individuals. 9i¼1 di 1i , Eq. (20), is the sum of individuals fixed effects for each of the 9 countries of the panel. Results from pooling and within estimation give us the degree of homogeneity. 5.3.2. Results Table 6 shows the different estimates of the b coefficient. Our results show perfect capital mobility (perfect degree of financial integration) with the within model. The pooling and between models show strong but not perfect capital mobility. But the estimate of b is relatively weak. Besides, the pooling and within model estimates reveal heterogeneity between countries. Fixed coefficients in the within model show that our countries in the sample do no have not the same preferences in terms of saving and investment rates. We try to homogenize our panel by estimating a new relationship between saving and investment rates according to income level. For this classification, only the pooling and within models are estimated. The between model, in this case, is calculated as an average for four countries. Our results for the between model are less significant. Our results show that: (i) countries at a high-income level are more financially integrated than countries in the middle group; and (ii) sub-groups appear to be homogeneous because results of the pooling and within models are the same for each of them. 5.3.3. A financial crisis effect? In this subsection, we gauge whether the 1997 financial crisis had an effect on the financial integration process. That is, we want to know if there was an increase or a decrease of financial integration after 1997. We know that the causality
Table 6 Estimate of b coefficient. Model
b
All countries Pooling Between Within
0.213*** 0.390*** 0.069***
Middle-income countries Pooling Within
0.319** 0.380**
High-income countries Pooling Within
0.216* 0.202*
* ** ***
Significance at the 10% level. Significance at the 5% level. Significance at the 1% level.
C. Guillaumin / Journal of Asian Economics 20 (2009) 314–326
323
Table 7 Estimate of b coefficient. Periods 1988–1996 1998–2006 *
Pooling ***
0.53 0.28***
Within 0.07 0.00
Significance at the 10% level. **Significance at the 5% level. *** Significance at the 1% level.
between economic growth and financial liberalization may not be clear (e.g., Obstfeld, 1994), but the impact on the financial integration process is an open question. We proceed in two steps. First, we estimate the following model: 8 9 2X 005 X I S > > > D ¼ c þ bD þ di 1 i þ dt 1t þ ei;t > > Y Y > i;t i;t > t¼1988 i¼1 > >
(21)
t¼1988
This model is a modified form of Eq. (20), except that we introduce, with fixed effects, a temporal effect. Here, 11997 = 1 for 1997 and 0 for all other years (11988 = 11989 = . . . =11996 = 11998 = . . . =12005 = 0). We decide to re-impose a fixed effect because we know that there is heterogeneity between the series. Our results show that 1997 is significant and has a higher effect than the average. In a second step, we consider two sub-periods (1988–1996 and 1998–2006). Table 7 shows the results of the estimation using the pooling and within models. With the pooling model, the coefficient is 0.53 for 1988–1996 and 0.28 for 1998–2006. With the within model, these coefficients are, respectively, 0.07 and 0.00 and they are insignificant in all cases. Therefore, our results suggest that the degree of financial integration is higher in the post-crisis period. How to explain this phenomenon? It is well acknowledged that the capital account liberalization policies and development of telecommunications and information technology have increased capital movement across countries in Asia. This phenomenon can be also explained by East Asian countries willing to develop regional financial market and to increase financial cooperation in order to avoid further financial crises. Also, for the within model, we note that there is no significant difference before and after the 1997 crisis. 6. Summary and conclusions This paper studies the degree of financial integration of nine East Asian countries from 1988 to 2006. We employ new panel unit root and panel cointegration techniques to estimate three types of modified Feldstein–Horioka equations: between, within and pooling. These different estimations show a high degree of financial integration. We try to homogenize our data and estimate a new relationship for high-income and middle-income countries. Our results show that high-income countries have stronger financial integration links than middle-income countries. Finally, we undertake stability tests in order to gauge if there is a crisis effect. We find that financial integration is stronger in the post-crisis period. Acknowledgements I am grateful to Claude Chambon, Virginie Coudert and Michael Plummer for their helpful comments. I am also grateful to Samuel Be´ji for research assistance. I thank participants of the CEPN lunch seminar and of the 18th conference of the ACAES Asian Economic Integration in a Global Context for highly profitable discussions and comments. I also thank the anonymous referee for helpful comments. All remaining errors are mine.
Appendix A. The relationship between saving and investment Fig. A1 shows the relationship between investment and saving rates from 1988 to 2006 for some East Asian countries. The figure on the right is made with the average of investment and saving rates (IR_AVE; SR_AVE), as proposed by Feldstein and Horioka (1980). Fig. A2 gives the same relationship, over the same period, according to the level of income (middle or high). Fig. A3 shows the relationship according to the type of financing in the economy.
324
C. Guillaumin / Journal of Asian Economics 20 (2009) 314–326
Fig. A1. The relationship between investment and saving rates. Source: World Bank, World Development Indicators.
Fig. A2. The relationship between investment and saving rates. (H: high-income countries; M: middle-income countries). Source: World Bank, World Development Indicators.
Fig. A3. The relationship between investment and saving rates. (Note: BANK: banking economy; FMARKET: financial market economy). Source: World Bank, World Development Indicators.
C. Guillaumin / Journal of Asian Economics 20 (2009) 314–326
325
Appendix B. Income level panel unit root tests Tables B1–B4. Table B1 First generation panel unit root rests: middle-income countries. Variables
Model
LLC
IPS ***
MW ***
Hadri ***
IR
Model without trend Model with trend
1.15 (0.1240) 1.08*** (0.1388)
0.45 (0.3257) 0.34*** (0.3635)
8.65 (0.3727) 9.77*** (0.2811)
2.33*** (0.0097) 3.36*** (0.0004)
SR
Model without trend Model with trend
1.50*** (0.0659) 1.48*** (0.0695)
1.30*** (0.0953) 0.27*** (0.3914)
13.18*** (0.1058) 8.44*** (0.3916)
2.57*** (0.0050) 3.09*** (0.0010)
Notes: IR and SR for Investment rate and Saving rate (in percent of GDP). Unit root hypothesis is accepted at ***1%, **5%, *10% levels. For Hadri’s test (2000), unit root hypothesis is the alternative hypothesis. LLC, IPS and MW for Levin et al. (2002), Im et al. (2003) and Maddala and Wu (1999) tests results. Values in parenthesis are p-value. All tests are made with fixed effects.
Table B2 Second generation panel unit root tests: middle-income countries. Variables
IR SR
Model without trend
Model with trend
CIPS1
CIPS2
CIPS3
CIPS1
CIPS2
CIPS3
1.48*** (0.685) 0.945*** (0.936
1.07*** (0.90) 0.429*** (0.994)
0.90*** (0.946) 0.203*** (0.998)
1.85*** (0.80) 0.794*** (0.998)
1.54*** (0.928) 0.002*** (1.000)
1.28*** (0.976) 0.121*** (1.000)
Notes: IR and SR for Investment rate and Saving rate (in percent of GDP). Unit root hypothesis is accepted at ***1%, **5%, *10% levels. CIPSi with i = 1, 2, 3 corresponds to the Pesaran statistics test (2003) for lag i. Values in parenthesis are p-value. All tests are made with fixed effects.
Table B3 First generation panel unit root tests: high-income countries. Variables
Model
LLC
IPS
MW
Hadri
IR
Model without trend Model with trend
0.79*** (0.7865) 1.96*** (0.0250)
1.14*** (0.8738) 1.42*** (0.0781)
2.81*** (0.9455) 18.33*** (0.0189)
2.34*** (0.0097) 4.42*** (0.0000)
SR
Model without trend Model with trend
0.51*** (0.3024) 1.19*** (0.1156)
0.07*** (0.8738) 1.11*** (0.1323)
7.95*** (0.9455) 12.43*** (0.1329)
2.68*** (0.0097) 3.72*** (0.0024)
Notes: IR and SR for Investment rate and Saving rate (in percent of GDP). Unit root hypothesis is accepted at ***1%, **5%, *10% levels. For Hadri’s test (2000), unit root hypothesis is the alternative hypothesis. LLC, IPS and MW for Levin et al. (2002), Im et al. (2003) and Maddala and Wu (1999) tests results. Values in parenthesis are p-value. All tests are made with fixed effects.
Table B4 Second generation panel unit root tests: high-income countries. Variables
Model without trend CIPS1
IR SR
Model with trend CIPS2
***
1.84 (0.413) 2.64 (0.038)
CIPS3 ***
1.46 (0.698) 1.48*** (0.689)
CIPS1 ***
1.34 (0.774) 0.104*** (0.999)
3.36 (0.016) 2.22*** (0.547)
CIPS2
CIPS3 ***
2.37 (0.431) 0.87*** (0.997)
1.82*** (0.820) 1.18*** (0.985)
Notes: IR and SR for Investment rate and Saving rate (in percent of GDP). Unit root hypothesis is accepted at ***1%, **5%, *10% levels. CIPSi with i = 1, 2, 3 corresponds to the Pesaran statistics test (2003) for lag i. Values in parenthesis are p-value. All tests are made with fixed effects.
References Araujo, C., Brun, J.-F. & Combes, J.-L. (2004). Econome´trie. Bre´al. Armstrong, H. W., Balasubramanyam, V. N., & Salisu, M. A. (1996). Domestic savings, intra-national and intra-European Union capital flows, 1971–1991. European Economic Review, 40(6), 1229–1235. Bae, K.-E., Bailey, W. & Yun, Y. S. (2006). Determinants of bond holdings by foreign investors (BIS papers 30). Bank for International Settlements. Baltagi, B. H. (1995). Econometric analysis of panel data. Willey. Bai, J. & Ng, S. (2001). A PANIC attack on unit root and cointegration (Boston College Working Paper in Economics, 469). The Johns Hopkins University. Bai, J., & Ng, S. (2004). A PANIC attack on unit root and cointegration. Econometrica, 72(4), 1127–1177. Bajou, C., Lefeuvre, E., Mamandz, S. & Melka, J. (2006). Relations commerciales en Asie. Recherche e´conomique, Ixis CIB. Banerjee, A. & Zanghieri, P. (2003). A new look at the Feldstein–Horioka puzzle using an integrated panel (Cepii working paper 2003/22). Bautista, C., & Maveyraud-Tricoire, S. (2008). Saving-investment relationship, financial crisis and structural changes in East Asian countries. Economie Internationale, 111(3), 81–100. Bayoumi, T. (1989). Saving-investment correlations: Immobile capital, government policy or endogeneous behaviour? (IMF working paper 89-66).
326
C. Guillaumin / Journal of Asian Economics 20 (2009) 314–326
Bayoumi, T., & Rose, A. (1993). Domestic savings and intra-national capital flows. European Economic Review, 37(6), 1197–1202. Beitone, A., Gilles, P. & Parodi, M. (2006). Histoire des faits e´conomiques et sociaux de 1945 a` nos jours, Dalloz. Ben Gamra, S., & Plihon, D. (2007). Qualite´ des institutions, libe´ralisation et crises bancaires: le cas des pays e´mergents. Economie Internationale, 112(4), 5–28. Be´reau, S. (2007). Une mesure macroe´conome´trique ‘‘a` la Feldstein–Horioka’’ du degre´ d’inte´gration financie`re en Europe’’. Economie Internationale, 110(2), 63– 106. Choi, I. (2001). Unit root test for panel data. Journal of International Money and Finance, 20(2), 249–272. Coakley, J., Kulasi, F., & Smith, R. (1996). Current account solvency and the Feldstein–Horioka puzzle. Economic Journal, 106(436), 620–627. Coiteux, M., & Olivier, S. (2000). The saving retention coefficient in the long run and in the short run: Evidence from panel data. Journal of International Money and Finance, 19(4), 535–548. Corbin, A. (2001). Specific effect in the Feldstein–Horioka paradox: A panel data analysis. Economics Letters, 72(3), 297–302. Davanne, O. (1998). Instabilite´ du syste`me financier international (Conseil d’analyse e´conomique, La Documentation franc¸aise). Eichengreen, B. (2004). Financial development in Asia: The way forward. mimeo. Eichengreen, B. (2006). The development of Asian bond markets (BIS Papers 30). Bank for International Settlements. Eichengreen, B. & Luengnaruemitchai, P. (2004). Why doesn’t Asia have bigger bond markets? (NBER working paper 10576). Feldstein, M., & Horioka, C. (1980). Domestic saving and international capital flows. Economic Journal, 90(358), 314–329. Fischer, R. A. (1932). Statistical methods for research workers. Edinburg: Oliver & Boyd. Flandreau, M., & Rivie`re, C. (1999). La Grande ‘‘Retransformation’’? Inte´gration financie`re internationale et controˆle des capitaux, 1880–1996. Economie Internationale, 78(2), 11–58. Frankel, J. (1992). Measuring international capital mobility: A review. American Economic Review, 82(2), 197–202. Fukasaku, K., & Martineau, D. (1999). Coope´ration mone´taire en Asie de l’Est: l’apport des tests de causalite´ et de la cointe´gration. Economie et Pre´vision, 140–141, 105–116. Gue´rin, J. & Sa, S. (2006). Les progre`s de l’inte´gration mone´taire et financie`re en Asie (Revue de la stabilite´ financie`re, 8, pp. 117–135). Banque de France. Guillaumin, C. (2009). (A)syme´trie et convergence des chocs macroe´conomiques en Asie de l’Est: une approche dynamique. Economie Internationale, 114(2), 29– 68. Gyntelberg, J., Ma, G. & Remolona, E. (2006). Developing corporate bond markets in Asia (BIS papers 26). Bank for International Settlements. Hadri, K. (2000). Testing for stationarity in Heterogeneous Panel Data. Econometrics Journal, 3(2), 148–161. He´ricourt, J. (2005). Essai sur la pratique de la politique mone´taire sous diffe´rents re´gimes mone´taires: de la zone euro a` la Turquie. The`se de Doctorat. Universite´ Paris IPanthe´on Sorbonne. He´ricourt, J. & Maurel, M. (2005). A new look at Feldstein-Horioka puzzle: An ‘‘European Regional’’ perspective. cahiers de la MSE 05070. Ho, T.-W. (2002). The Feldstein–Horioka puzzle revisited. Journal of International Money and Finance, 21(4), 555–564. Hsiao, C. (2003). Analysis of panel data. Cambridge University Press. Hurlin, C., & Mignon, V. (2005). Une synthe`se des tests de racine unitaire sur donne´es de panel. Economie et Pre´vision, 169, 253–294. Hurlin, C., & Mignon, V. (2007). Une synthe`se des tests de cointe´gration sur donne´es de panel. Economie et Pre´vision, 180–181, 241–265. Im, K., Pesaran, M., & Shin, Y. (2003). Testing for unit roots in heterogeneous panels. Journal of Econometrics, 115(1), 53–74. Isaksson, A. (2001). Financial liberalisation, foreign aid, and capital mobility: Evidence from 90 developing countries. Journal of International Financial Markets, 11(3–4), 309–338. Ito, T. (2001). Growth, crisis, and the future of economic recovery in East Asia. In J. Stiglitz & S. Yusuf (Eds.), The east Asian miracle. Oxford University Press. Iwamoto, Y., & van Wincoop, E. (2000). Do borders matter? Evidence from Japanese International Capital flows. International Economic Review, 41(1), 241–269. Jansen, W. (1996). Estimating saving-investment correlations: Evidence for OECD countries based on an error correction model. Journal of International Money and Finance, 15(5), 749–781. Jansen, W. (2000). International capital mobility: Evidence from panel data. Journal of International Money and Finance, 19(4), 507–511. Kao, C., & Chiang, M.-H. (2001). On the estimation and inference of a cointegrated regression in panel data. In Baltagi, B. (Ed.). Advances in Economics. vol. 15 (pp.179–222). Kim, S. (2001). The Saving-investment correlation puzzle is still a puzzle. Journal of International Money and Finance, 20(7), 1017–1034. Kim, S., Kim, H. S., & Wang, Y. (2007). Saving, investment and international capital mobility in East Asia. Japan and the World Economy, 19(2), 279–291. Kim, H., Oh, K.-Y., & Jeong, C.-W. (2005). Panel cointegration results on international capital mobility in Asian countries. Journal of International Money and Finance, 24(1), 71–82. Krol, R. (1996). International capital mobility: Evidence from panel data. Journal of International Money and Finance, 15(3), 467–474. Kwiatkowski, D., Phillips, P., Schmidt, P., & Shin, Y. (1992). Testing the null hypothesis of stationarity against the alternative of unit root. Journal of Econometrics, 54(1–3), 159–178. Levin, A. & Lin, C. F. (1992). Unit root tests in panel data: Asymptotic and finite sample properties (discussion paper, 56). Department of Economics, University of California at San Diego. Levin, A., Lin, C. F., & Chu, J. (2002). Unit root tests in panel data: Asymptotic and finite sample properties. Journal of Econometrics, 108(1), 1–24. Maddala, G., & Wu, S. (1999). A comparative study of unit root tests and a new simple test. Oxford Bulletin of Economics and Statistics, 61(special issue), 631–652. Menkhoff, L., & Suwanaporn, C. (2007). 10 Years after the crisis: Thailand’s financial system reform. Journal of Asian Economics, 18(1), 4–20. Moneta, F., & Ru¨ffer, R. (2009). Business cycle synchronisation in East Asia. Journal of Asian Economics, 20(1), 1–12. Obstfeld, M. (1994). Are industrial country consumption risks globally diversified (NBER working paper 4308). Obstfeld, M. & Taylor, A. M. (2002). Globalization and capital markets (NBER working paper 8846). Park, D., & Rhee, C. (2006). Building infrastructure for Asian bond markets: Settlement and credit rating. In D. Arner, J.-H. Park, P. Lejot, & Q. Liu (Eds.), Asia’s Debt Capital Markets. Springer. Pedroni, P. (1997). Panel cointegration, asymptotic and finite sample properties of pooled time series with an application to the PPP hypothesis: New results (Working paper in Economics). Indiana University. Pedroni, P. (1999). Critical values for cointegration tests in heterogeneous panels with multiple regressors. Oxford Bulletin of Economics and Statistics, 61(Special issue), 653–670. Pesaran, H. M. (2003). A simple panel unit root test in the presence of cross section dependence. Mimeo. University of Southern California. Phillips, P., & Sul, D. (2003). Dynamic panel estimation and homogeneity testing under cross section dependence. Econometrics Journal, 6(1), 217–259. Proble`mes e´conomiques (1997). Le syste`me financier international. Nume´ro double 2541–2542. Salanie´, B. (1999). Guide pratique des se´ries non stationnaires. Economie et Pre´vision, 137, 119–141. Sinha, D. (2002). Saving-investment relationships for Japan and other Asian countries. Japan and the World Economy, 14(1), 1–23. Sinn, S. (1992). Saving-Investment correlations and capital mobility: On the evidence from annual data. Economic Journal, 102(414), 1162–1170. Stiglitz, J., & Yusuf, S. (2001). Rethinking the East Asian Miracle. Oxford University Press. Summers, L. (1989). Tax Policy and International Competitiveness (NBER working paper 2007). Sundaresan, S. (2006). Developing multiple layers of financial intermediation: The complementary roles of corporate bond markets and banks (BIS papers 26). Bank for International Settlements. Takeuchi, A (2006). Identifying impediments to cross-border bond investment and issuance in Asian countries (BIS papers 30). Bank for International Settlements. van Wincoop, E. (2001). Intranational versus international saving-investment comovements. In G. Hess & E. van Wincoop (Eds.), Intranational Macroeconomics. Cambridge University Press.