Regional and global integration of Asian stock markets

Research in International Business and Finance 50 (2019) 357–368

Contents lists available at ScienceDirect

Research in International Business and Finance journal homepage: www.elsevier.com/locate/ribaf

Full length Article

Regional and global integration of Asian stock markets ⁎

Wahbeeah Mohtia, , Andreia Dionísiob, Isabel Vieiraa, Paulo Ferreirab,c

T

a

Universidade de Évora (Departamento de Gestão), Largo dos Colegiais, 2, 7000-903, Évora, Portugal CEFAGE-UE, IIFA, Universidade de Évora, Largo dos Colegiais, 2, 7000, Évora, Portugal c Escola Superior Agrária de Elvas, Instituto Politécnico de Portalegre, Portugal b

A R T IC LE I N F O

ABS TRA CT

JEL classification: G0 G1

This paper assesses the levels of regional and global stock market integration of emerging and frontier Asian countries. The long run relationships established amongst markets are investigated using Gregory and Hansen’s cointegration tests and Detrended Cross Correlation coefficients. The results of the empirical analysis indicate that all considered emerging markets display some evidence of both global and regional integration. In the case of frontier markets, however, this is true solely for Pakistan and, to a lesser extent, for Vietnam. These results are of interest, inter alia, to international investors interested in expanding the geographical scope of portfolio diversification strategies.

Keywords: Emerging markets Frontier markets Stock market integration Detrended cross correlation analysis (DCCA) Gregory and Hansen cointegration test

1. Introduction The integration of stock markets has distinct implications and its empirical evaluation may be of use in various contexts. Financial integration affects stock prices’ comovements across countries (Bekaert et al., 2002) and highly integrated stock markets exhibit statistically significant long-term relationships (Cheng, 2000). Obstfeld’s (1994) theoretical model indicated that international risk sharing improves resources’ allocation and promotes economic growth. The empirical assessment by Korajczyk (1996) concurs in showing that the level of stock market integration is positively related to economic growth. Market integration is also relevant for portfolio investors. According to the ‘modern portfolio theory’, diversification increases the return for a given level of risk, or decreases the risk for a given return. As different stock markets can be exposed to distinct factors, international diversification can enhance the advantages of domestic diversification (e.g. Grubel, 1968), provided that domestic and foreign markets are less than perfectly correlated (Masih and Masih, 1999). It is thus not surprising that the benefits from international portfolio diversification have decreased with market integration and the globalization of financial markets (see e.g. Cheung and Mak, 1992; Gilmore and McManus, 2002; Bessler and Yang, 2003; Kearney and Lucey, 2004; Alagidede, 2008). Such developments enhance the interest in potentially more segmented markets, such as frontier markets (see Speidell and Krohne, 2007; Jayasuriya and Shambora, 2009) and thus the utility of empirical assessments of their level of integration with more traditional ones. Our work therefore contributes to the financial integration literature by both extending the geographical scope of existing empirical studies and adopting a robust methodological approach that takes into account statistical specificities of the utilized data often disregarded in this type of analysis. First, we study frontier markets, namely Pakistan, Bangladesh, Sri Lanka and Vietnam, which have been relatively less investigated, probably due to reasons related to data availability. Assessment of such

⁎

Corresponding author. E-mail addresses: [email protected] (W. Mohti), [email protected] (A. Dionísio), [email protected] (I. Vieira), [email protected] (P. Ferreira). https://doi.org/10.1016/j.ribaf.2019.06.003 Received 13 November 2018; Received in revised form 28 May 2019; Accepted 3 June 2019 Available online 05 June 2019 0275-5319/ © 2019 Elsevier B.V. All rights reserved.

Research in International Business and Finance 50 (2019) 357–368

W. Mohti, et al.

countries adds to current knowledge on financial integration. We also consider Asian emerging economies (India, China, Korea, Taiwan, Malaysia, Indonesia, the Philippines, and Thailand) to be able to examine integration at the regional and at the global levels, and thus to observe how the behavior of frontier markets compares to that of emerging ones in both respects. Secondly, to assess long run relationships of Asian stock markets with regional and global markets (using as proxies the Japanese and the US stock markets, respectively), we perform Gregory and Hansen’s (1996) cointegration tests and compute Detrended CrossCorrelation Analysis’ coefficients (ρDCCA). The former is a linear approach and the latter has the advantage of capturing possible (non) linearities between variables. DCCA also presents the advantages of being adequate regardless of the stationarity of the examined time series and to allow quantification of correlations at distinct time scales. Finally, to enhance the informative value of our analysis, we investigate how financial integration evolved throughout the considered time frame by separating the sample into four subsamples and examining the changes in DCCA correlations across subsequent sub periods. The study concludes that Asian emerging markets are regionally and globally integrated. In the case of frontier markets, evidence of integration emerged solely for Pakistan and, in a lesser degree, for Vietnam. Bangladesh and Sri Lanka displayed no statistically significant relationships with either the regional or the global benchmarks. The remainder of the study is organized as follows: Section 2 reviews empirical assessments of stock market integration. Section 3 presents the data and the adopted methodologies. Section 4 describes the empirical results, and Section 5 concludes. 2. Literature review Financial researchers have followed various strategies to examine the level of financial integration across markets. According to Baele et al. (2004), financial integration may be assessed from three perspectives: quantity, news, and price based approaches. Quantity measures include saving-investment correlations, initially proposed by Feldstein and Horioka (1979), or consumption correlation measures, due to Obstfeld (1993). Under this approach, stock markets are investigated in terms of asset flows between countries (Adam et al., 2002). News-based measures distinguish between information and market imperfections, such as frictions and barriers. If financial markets are completely integrated and portfolios are well diversified, the impact of new domestic information on a particular market should be similar to that of global news. Price based measures draw from the “law of one price,” according to which if financial markets are perfectly integrated, assets with the same risk are priced identically in different markets. Researchers used interest rate parity tests such as evaluations of data compliance with the covered interest parity, the uncovered interest parity, and the real interest parity (for details see: Cuestas et al., 2015; Filipozzi and Staehr, 2012; Cheung et al., 2006). Within this approach, which we follow in our own analysis, there are also studies based on stock markets’ cointegration and correlations (see Neaime, 2012; Al Nasser and Hajilee, 2016). Academic curiosity over stock market integration was enhanced by the progressive increase in interconnectedness across stock markets and also by the interest of investors in new opportunities to improve portfolios’ risk adjusted returns. Thus, although initial analyses were mainly focused on developed markets (e.g. Kasa, 1992; Richards, 1995; Engsted and Tanggaard, 2004; Rua and Nunes, 2009), later ones turned to emerging markets (e.g. Goetzmann et al., 2005; Bekaert and Harvey, 2014). The geographical scope of empirical assessments is broad – see, for instance, Syriopoulos (2007); Égert and Kočenda (2007); Wang and Moore (2008); Horvath and Petrovski (2013); Kiviaho et al. (2014), and Guidi and Ugur (2014) on European markets, or Yu and Hassan (2008); Alkulaib et al. (2009), and Aloui and Hkiri (2014) on Middle Eastern and North African markets. Researchers interested in the Asian region have so far mainly assessed the more developed and emerging markets, concluding that there are some potential diversification benefits to be explored in such areas, but also producing a variety of mixed results (Cheung and Mak, 1992; Sharma and Wongbangpo, 2002; Yang et al., 2003; Wong et al., 2004; Kim et al., 2006). For instance, studies examining comovements within Asia, and between Asian and the global market (see Choudhry et al., 2007), reported significant longterm links amongst markets in Thailand, Malaysia, Indonesia, Hong Kong, Singapore, the Philippines, South Korea, Taiwan, and Japan. Mukherjee and Bose (2008) concluded that, from 2005 onwards, the Indian stock market was influenced by those of the US, Japan, Hong Kong, South Korea and Singapore, and on its turn, also influenced other Asian markets. Some assessments have suggested that Asian stock markets were more integrated with regional than with global markets. One example is the study developed by Lee and Jeong (2016) who investigated the integration of markets of the Association of Southeast Asian (Indonesia, Malaysia, the Philippines, Singapore and Thailand) with China and the US, considered respectively as regional and global benchmarks. In the same context, and employing a variety of methodologies, Yu et al. (2010) concluded that previously weak relationships were enhanced after 2007. Gupta and Guidi (2012) reported the existence of short (but not long) run relationships of the Indian stock market and three Asian developed markets. The authors employed cointegration (proposed by Engle and Granger, Johansen, and Gregory and Hansen), Granger causality tests, and dynamic conditional correlation (DCC) GARCH models. They concluded that links between India and the other assessed markets tend to rise in times of crisis, diminishing the benefits investors could obtain by including emerging markets in their portfolios. Similar conclusions were reached by Lean and Teng (2013), who compared Malaysian stock market integration with markets in the US, Japan, China and India. Their study produced evidence of stronger links with the US and India, leading to the conclusion that Malaysian investors could benefit more from diversification into the Chinese and the Japanese markets. Recently, some evidence was put forward on Asian frontier markets. For instance, Mensi et al. (2017) studied the so-called BRICs (Brazil, Russia, India and China) and South Asian (Pakistan, Bangladesh and Sri Lanka) frontier markets’ links with developed markets (in the US, the UK and Japan). The results indicated that comovements, and thus the potential diversification benefits, between these countries changed with time. Focusing solely on South Asian countries, Sharma and Bodla (2011) reported that the 358

Research in International Business and Finance 50 (2019) 357–368

W. Mohti, et al.

Indian stock market Granger-caused those of Pakistan and Sri Lanka. The reviewed literature indicates that, although there are already various assessments of Asian emerging and developed stock markets’ integration with regional and global ones, Asian frontier markets have received relatively less attention. Mensi et al. (2017) are an exception but, although they have focused on Asian frontier markets, they ignored the increasingly relevant case of Vietnam, a rapidly growing economy. Previous research was developed using a variety of techniques. A significant stream of literature was based on cointegration and correlations. The logic behind these methodologies is that the growth of correlation over time depicts increasing integration. Traditional cointegration based studies, including the approaches proposed by Engle-Granger (1987) and Johansen (1988), are not robust in the presence of structural breaks. If they exist, more advanced methods, such as the one proposed by Gregory and Hansen (1996), should be used instead. Various other techniques were also applied in previous assessments. Lean and Teng (2013) used DCCMGARCH models, Shi and Shang (2013) employed DCCA, a method also adopted by Ferreira (2017) to study stock market integration between Portuguese and Brazilian stock markets. In what follows, we use two different methods to investigate Asian stock markets’ integration with global and regional benchmarks: Gregory and Hansen’s (1996) cointegration tests are used to analyze long run relationships; DCCA, a method that is robust regardless of the (non) stationarity of the underlying data, is used to study power law cross correlation between markets. 3. Data description and methodology 3.1. Data description This study assesses stock market integration in Asia, considering eight emerging markets (India, China, Korea, Taiwan, Malaysia, Indonesia, the Philippines, and Thailand) and four frontier markets (Pakistan, Bangladesh, Sri Lanka and Vietnam). Morgan Stanley Capital International (MSCI) classifies these markets as emerging and frontier based on their market size, number of listed stocks, level of domestic stability and several macroeconomic characteristics. We evaluate Asian markets’ links with global and regional benchmarks. The US Dow Jones and the Japanese stock indices are used as proxies for, respectively, the global and the regional markets (Japan is the biggest stock market in terms of market capitalization in Asia). Daily closing prices for the indices were collected from DataStream, from December 2009 to April 2017 (comprising a total of 1934 observations). The justification for the relatively short period of time considered in the analysis is the unavailability of data for the Bangladesh stock market before 2009. To simplify the explanation of empirical results, the following blocs are constructed: Frontier market bloc: Pakistan1, Bangladesh, Sri Lanka, Vietnam; Emerging market bloc: India, China, Korea, Taiwan, Malaysia, Indonesia, the Philippines, Thailand. 3.2. Methodology The non-stationary property of individual series is investigated with unit root tests. There are various tests to this end but, since structural breaks are a possibility in the considered time series, we use the one proposed by Zivot and Andrew (2002), encompassing two break dynamic models: innovational outlier, which assumes that breaks occur slowly; and the additive outlier, based on the assumption that breaks occur instantly. When structural breaks exist, the cointegration analysis should be performed taking such possibility into consideration. Traditional cointegration tests, as those proposed by Engle and Granger (1987) or Johansen (1988), although frequently used, have a few limitations – for instance, not allowing for structural shifts in the cointegrating vectors, or for any regime changes – and thus may produce biased results. We use the residual based cointegration test proposed by Gregory and Hansen (1996). It joins both Engle and Granger’s (1987), and Zivot and Andrew’s (2002) tests, allowing for structural breaks in estimating the cointegrating relationships amongst variables. The Gregory and Hansen’s test considers ‘no cointegration’ as its null hypothesis along with three alternative models of structural changes on unknown dates: model C – level shift (allowing for changes solely in the intercept); model C/T – level shift with trend (allowing for slope vector shifts as well); and model C/S – regime shift (allowing for changes in the intercept and in the cointegrating slope coefficients). We apply all the models to gauge the difference between them and to increase the robustness of the empirical analysis. In order to perform the financial integration tests, we use DCCA, a method proposed by Podobnik and Stanley (2008) to quantify power law cross-correlation between two series, regardless of their stationarity. This technique is an extension of the Detrended Fluctuation Analysis (DFA) initially proposed by Peng et al. (1994). DFA was originally used in other research fields (such as meteorology, geophysics or criminology), but it is now also widely used in finance (for more details see: Ausloos and Ivanova, 2000; Wang et al., 2010; Ferreira and Dionísio, 2016). The main advantage of DFA is avoiding the spurious detection of long-range dependence in non-stationary data. In this study, we do not directly apply DFA. We use the DFA exponent values to calculate DCCA correlation coefficients. DFA is used to assess time dependence in a single time series while DCCA permits the investigation of cross correlation between two series, allowing the study of both linear and non-linear relationships. The procedure to calculate DCCA coefficients is developed 1 MSCI reclassified Pakistan as an emerging market in May 2017, but as it was, under the most part of the sample, in the frontier bloc, we kept it in that group.

359

Research in International Business and Finance 50 (2019) 357–368

W. Mohti, et al.

in the following steps, considering that xxk and yyk are two time series with k = 1… t observations, generated and integrated to t t produce xx (t ) = ∑k = 1 xxk and yy (t ) = ∑k = 1 yyk . Both series are divided into boxes of equal length n. Afterwards, these are divided into non-overlapping parts N – n and the method of ordinary least squares is used to obtain the local trend ( xx^k and yy^k ) for each box. Next, we compute the detrended series by taking the difference between the original values and their trends. The covariance of the residuals in each box is calculated as: 2 f DCCA =

i+n 1 ∑ (xxk − xx^k ) (yyk − yy^k ) n − 1 k=1

(1)

The detrended covariance is calculated by summing up all N - n boxes of size n:

F 2 DCCA (n) =

1 N−n

N −n

∑

2 f DCCA

(2)

i=1

In order to find the relationship between the DCCA fluctuation function and n, this whole process is repeated for boxes of different lengths. This provides the long range cross correlation FDCCA (n) given by FDCCA (n) ˜nλ. The information provided by λ is the following: if, λ = ½, this indicates that no long range cross correlation between the series yh and yh’ exists; λ > ½ depicts persistent cross-correlation; λ < ½, reflects anti persistent cross correlation between the series, which means that both series are overturning the direction of recent moves. DCCA provides information regarding cross-correlation between two series but not on their intensity. Zebende (2011) thus proposed a correlation coefficient test based on DCCA and the respective hypothesis test, arguing that the DCCA correlation coefficient provides robust results by identifying the seasonal components in both negative and positive cross correlations. The DCCA correlation coefficient is given by:

ρDCCA =

2 FDCCA ˙ yyi FDFAxxi FDFA

(3)

The correlation coefficient (ρDCCA) is comprised between -1 and +1. If ρDCCA = 0, there is no cross-correlation; values of -1 and +1 indicate the existence of negative and positive cross-correlation, respectively. To test the significance of correlation coefficients we follow the procedure proposed by Podobnik et al. (2011). With the objective of analyzing how financial integration evolved over time, we compute ΔρDCCA (n) and assess the behavior of financial integration over two consecutive periods (see Da Silva et al., 2016):

ΔρDCCA = ρDCCA t − ρDCCA t − 1 If the variation of ρDCCA is positive, integration has increased over time, the reverse occurring if it is negative. Recently, Guedes et al. (2018a,b) proposed a statistical test allowing the assessment of the significance of Δ ρDCCA (n) , under the following hypotheses:

H0: Δ ρDCCA (n) = 0 H1: Δ ρDCCA (n) ≠ 0 Rejection of the null hypothesis indicates that correlation changes between time periods are statistically significant. Positive and significant ΔρDCCA values signal correlation increases and thus increased levels of integration, whereas negative significant values indicate the opposite, i.e. more segmented markets. 4. Empirical results and discussion We use the data in natural logarithms to assess the stationary properties of the individual series and integration between indices (see Chien et al., 2015). The analysis of market integration starts with the assessment of the series’ stationarity, considering the possibility of structural breaks, and using Innovational (IO) and Additive outlier (AO) models. Table 1, reports the results for break point unit root tests for the IO model (break dates are shown for the individual series). The results obtained with the AO model are not quantitatively different from those reported. The indices of Sri Lanka and Vietnam are stationary. The remaining indices are non-stationary and integrated of order one. After unit root testing, Gregory and Hansen’s cointegration tests are performed (except on Sri Lanka and Vietnam data, which are stationary) to assess long run relationships between the series. Table 2 presents pair wise cointegration tests’ results for emerging markets long run comovements with the regional market (Japan). We considered three models for Gregory and Hansen’s cointegration tests. Recall that the C and the C/T models allow for changes solely in the intercept, and in the intercept and slope, respectively; while the C/S model accommodates changes in the intercept and in the cointegrating slope coefficients. Considering model C, only the Zt statistics for Korea are significant (at the 10% level). The other emerging markets do not display significant results. With the C/T model, Indonesia and Thailand have significant cointegration test statistics for ADF/Zt and Zα, suggesting the existence of long run cointegration with the regional market. For the C/S model, only the Philippines’ values are significant, also indicating the existence of long run links with the regional market. These results suggest that emerging markets are not fully regionally integrated. 360

Research in International Business and Finance 50 (2019) 357–368

W. Mohti, et al.

Table 1 Break point Unit Root Tests - Innovational Outlier Model. Variables

Level

First Difference

Country

T-stat

k

Break Date

T-stat

k

Break Date

India Indonesia China Korea Malaysia Philippines Taiwan Thailand Pakistan Sri Lanka Bangladesh Vietnam US Japan

−3.68356 −3.46049 −3.27557 −4.38829 −3.70223 −3.34186 −3.46823 −4.29425 −3.3375 −5.38777*** −3.25813 −4.5811** −2.64545 −4.1039

0 4 1 0 1 0 0 0 1 0 2 1 0 0

2/13/2014 6/04/2012 7/09/2015 5/25/2010 9/26/2011 9/26/2011 2/06/2014 10/04/2011 1/16/2013 12/14/2009 12/13/2010 5/10/2011 11/15/2012 11/14/2012

−42.1105 −42.3942 −42.2298 −44.0593 −39.6806 −41.6604 −42.6807 −43.6871 −39.348 −37.6262 −49.5728 −39.7024 −46.5518 −45.3831

0 0 0 0 0 0 0 0 0 0 0 0 0 0

8/24/2015 12/21/2009 7/08/2015 8/19/2011 9/26/2011 6/13/2013 8/05/2011 10/06/2011 8/24/2015 10/12/2010 1/04/2010 5/25/2011 8/24/2015 3/15/2011

Critical Values: 1% level -4.94913: 5% level -4.44365: 10% -4.19363, *** indicates significance at the 1% level. k indicates the automatic lag selection. Table 2 Gregory and Hansen’s Cointegration Tests (Emerging bloc with Japan). Indices

Models

ADF

Lag

BP

Zt

Zα

BP

China

Model Model Model Model Model Model Model Model Model Model Model Model Model Model Model Model Model Model Model Model Model Model Model Model

−3.94 −4.04 −3.94 −4.30 −4.27 −4.30 −3.44 −4.91* −3.99 −3.96 −4.34 −3.99 −3.92 −4.17 −3.98 −3.77 −4.20 −4.81* −3.36 −3.92 −3.46 −3.89 −4.76* −4.28

2 2 2 0 0 0 4 4 4 1 2 1 0 1 0 2 1 1 0 0 0 1 1 2

4/22/2013 3/27/2013 4/22/2013 4/15/2014 11/26/2013 4/15/2014 3/15/2011 7/15/2015 1/7/2013 3/14/2013 6/30/2015 3/14/2013 1/10/2011 6/23/2015 2/1/2016 9/21/2011 12/10/2012 3/19/2012 7/2/2017 8/30/2011 8/7/2011 12/20/2012 7/30/2015 12/24/2012

−3.98 −3.92 −4.18 −4.31 −4.30 −4.36 −4.21 −6.16*** −5.06 −4.29 −4.64 −4.34 −4.50* −4.32 −4.25 −3.93 −4.51 −4.95** −3.54 −4.19 −3.70 −3.83 −4.99** −4.59

−30.37 −30.75 −31.35 −37.36 −36.96 −37.68 −32.49 −66.69*** −43.73* −28.27 −35.06 −28.99 −32.90 −36.11 −35.74 −26.41 −36.07 −43.75* −25.21 −34.02 −27.41 −24.64 −43.82* −33.88

4/12/2013 3/19/2013 8/11/2011 4/15/2014 11/26/2013 5/6/2014 3/16/2011 7/31/2015 1/8/2013 3/13/2013 6/30/2015 3/13/2013 1/10/2011 6/22/2015 1/10/2011 9/22/2011 12/13/2012 3/20/2012 3/7/2011 8/9/2011 8/9/2011 1/20/2011 8/26/2015 12/21/2012

India

Indonesia

Malaysia

Korea

Philippines

Taiwan

Thailand

C C/T C/S C C/T C/S C C/T C/S C C/T C/S C C/T C/S C C/T C/S C C/T C/S C C/T C/S

***, ** and * indicate significance at the 1%, 5% and 10% levels, respectively.

Table 3 presents results for tests assessing emerging markets long run comovements with the global market (US). In model C, two countries - Indonesia and Taiwan – display significant test statistics that indicate the existence of long run links. However, only the Zα statistic values for Thailand (at 5%), Malaysia and Korea (at 10%) are significant. Considering the C/T model, again Indonesia and Taiwan have significant comovements with the US, the same occurring with the C/S model for Indonesia, Malaysia and Taiwan. Korea displays statistically significant Zα values in all three models. For a few markets, such as India and China, no long run comovements with the regional or the global markets could be found. Table 4 displays Gregory and Hansen’s cointegration tests’ results for frontier markets vis-à-vis the regional and global markets. Sri Lanka and Vietnam are excluded because their time series are stationary. For the two remaining frontier markets, and in relation to Japan, only Pakistan has significant links (C/T model). In relation to the US, and taking the three models into account, again only Pakistan has significant test statistics. The results obtained so far are not informative on the intensity of the interconnections between countries. In order to complement our findings, we calculate DCCA correlation coefficients, which allow investigating the type of relationships between markets regardless of the linearity of cross correlations. Long-range correlation coefficients between pairs of markets are computed using Eq. (3) 361

Research in International Business and Finance 50 (2019) 357–368

W. Mohti, et al.

Table 3 Gregory and Hansen’s Cointegration Tests (Emerging bloc with US). Indices China

India

Indonesia

Malaysia

Korea

Philippines

Taiwan

Thailand

Model Model Model Model Model Model Model Model Model Model Model Model Model Model Model Model Model Model Model Model Model Model Model Model

C C/T C/S C C/T C/S C C/T C/S C C/T C/S C C/T C/S C C/T C/S C C/T C/S C C/T C/S

ADF

Lag

BP

Zt

Zα

BP

−3.88 4.12 −3.87 −4.04 −4.27 −4.07 −4.39* −4.46 −4.48 −4.25 −4.57 −4.84* −3.99 −4.48 −3.36 −3.85 −3.86 −3.86 −5.00** −5.26** −4.87* −3.82 −4.31 −4.53

0 2 0 0 0 0 3 4 3 1 1 2 0 3 2 1 0 3 2 2 2 1 1 3

8/29/2011 3/26/2015 7/22/2011 5/6/2014 4/14/2014 5/6/2014 7/16/2013 6/19/2015 7/15/2013 11/24/2014 11/24/2014 11/19/2014 1/10/2011 2/12/2013 10/10/2014 6/8/2012 8/4/2015 12/7/2012 8/29/2011 10/26/2011 7/22/2011 1/8/2014 9/3/2013 5/21/2013

−3.73 −3.48 −3.75 −4.03 −4.17 −4.04 −4.70** −5.43** −4.75* −4.30 −4.19 −4.75* −3.99 −4.17 −4.57 −3.49 −3.94 −3.81 −4.61** −4.85* −4.57 −3.71 −3.81 −4.54

−28.84 −38.64 −28.86 −32.54 −36.79 −32.52 −50.69*** −54.92** −52.07** −36.34* −41.44 −46.86* −33.87* −52.20** −47.90** −31.77 −31.73 −37.71 −56.55*** −68.13*** −56.56** −29.50 −35.39 −48.65**

8/12/2011 1/16/2015 7/25/2011 5/6/2014 5/27/2014 5/6/2014 6/26/2013 6/16/2015 6/26/2013 11/24/2014 11/21/2014 11/21/2014 9/12/2014 3/25/2013 7/22/2011 6/4/2012 8/26/2015 11/16/2012 8/30/2011 9/28/2011 8/30/2011 11/29/2013 11/5/2013 5/17/2013

***, ** and * indicate significance at the 1%, 5% and 10% levels, respectively. Table 4 Gregory and Hansen’s Cointegration Test (Frontier bloc). Indices

Models

Frontier Bloc with Japan Pakistan Model Model Model Bangladesh Model Model Model Frontier Bloc with US Pakistan Model Model Model Bangladesh Model Model Model

ADF

Lag

BP

Zt

Zα

BP

C C/T C/S C C/T C/S

−3.61 −5.33** −3.83 −3.89 −3.93 −4.08

4 6 2 2 2 4

3/14/2013 4/8/2015 3/6/2013 1/17/2013 1/17/2013 12/4/2012

−3.78 −5.42** −4.07 −3.67 −3.96 −3.75

−25.12 −52.97** −26.45 −27.75 −28.02 −28.79

3/18/2011 3/3/2015 2/5/2013 1/15/2013 1/15/2013 1/15/2013

C C/T C/S C C/T C/S

−5.64*** −5.33** −5.61*** −3.61 −3.93 −4.19

1 6 1 2 2 2

3/9/2015 4/8/2015 10/8/2015 3/8/2011 1/17/2013 7/22/2013

−5.63*** −5.42** −5.63*** −3.36 −3.96 −4.13

−67.95*** −52.97** −67.61*** −23.57 −28.02 −34.86

3/4/2015 3/3/2015 3/3/2015 2/14/2014 1/5/2013 12/26/2013

***, ** and * indicate significance at the 1%, 5% and 10% levels, respectively.

and their statistical significance is evaluated following the procedure proposed by Podobnik et al. (2011). ρDCCA coefficients are shown in Fig. 1, which also depicts the lower and upper critical values to establish their statistical significance. If the calculated coefficients are comprised between the critical values, the correlation is not significant. Conversely, if they lie outside those bounds, the correlation is significant. The behavior of emerging markets’ correlations with the regional market (Japan) is similar. The correlation coefficients of Indonesia, Korea, Malaysia, and Thailand are initially significant, and then display a decreasing trend, indicating that the markets are integrated in the short run but not in the long run. Coefficients for China and Taiwan are statistically significant. China displays the highest correlation coefficients in the long run. The country is geographically close to Japan, also a very relevant economy in both the Asian region and the world (Ma et al., 2013). In 2016, China was Japan’s main trading partner, with 19% of Japan’s exports directed to that country. Japan ranked third in the list of China’s main trading partners, receiving in that year imports from China valued in $129.5 billion. Taiwan also displays high and consistent correlation coefficients with the regional market. It was the fourth major trading partner of Japan (and vice-versa), with exports of approximately $19.6 billion, which amounted to 7% of Japan’s imports, and imported $40.6 billion worth of goods and services.2 These may be reasons for the relatively high level of financial integration

2

http://www.worldstopexports.com. 362

Research in International Business and Finance 50 (2019) 357–368

W. Mohti, et al.

Fig. 1. DCCA Cross correlation coefficients for Asian (Emerging and Frontier) stock markets as function of n (day). In the left hand side, correlations are with the regional market (Japan); in the right hand side, correlations are with the global market (US).

between these markets. Such results are consistent with Bracker et al. (1999) who suggested that strong bilateral trade increases the level of financial integration. Regarding the global market, results for the emerging bloc of countries indicate that all markets are integrated with the US in the long-run, although the values for China become non-significant at longer scales. China and various other Asian emerging countries 363

Research in International Business and Finance 50 (2019) 357–368

W. Mohti, et al.

are rival exporters for the US market. Voon and Yue (2003) investigated export rivalry between China, Malaysia, Singapore, Thailand and Indonesia and concluded that the first two countries (China and Malaysia) performed better than the other three. This may explain why Asian emerging stock markets tend to move in tandem with the US. In fact, Fujiwara and Takahashi (2012) pointed out that events in the US remain the major factors impacting developments in Asian stock markets. Frontier stock markets perform differently. The correlation coefficients of Bangladesh and Sri Lanka are not statistically significant in both cases. Pakistan is the only country displaying significant and high correlation coefficients with the US and Japan. Vietnam is also integrated with the regional market. The country’s correlation coefficients with the global benchmark are first significant and then non-significant but remaining very near the critical values, indicating weak integration in the long run. Asian frontier markets are small and less accessible for investors, and this may explain the weaker integration with regional and global markets. Low correlation between frontier markets with both regional and global markets also confirms the potential diversification benefits in these markets. The results of Gregory and Hansen’s cointegration tests and of the DCCA correlation coefficients are distinct. With the former method, we found no evidence of links between the Chinese and Indian markets and the regional or the global benchmarks. Yet, the DCCA produced evidence of relationships amongst such markets. Gregory and Hansen’s is an appropriate technique to investigate long run relationships between stock markets in the presence of structural breaks. However, as other cointegration tests, it uses a linear approach. Assessment with the ρDCCA produce more robust results, because it captures both linear and non-linear relations that might not be identified with other tests, as was the case in the cointegration analysis of China with both benchmarks. Furthermore, unlike Gregory and Hansen’s tests, DCCA may be used regardless of the (non) stationarity of the underlying data. In order to improve the contribution of this research to the existing literature, a further analysis attempting to gauge how the patterns of integration evolved with the passage of time is also implemented. To this end we calculate the value of ΔρDCCA considering various periods of time which result from dividing the whole sample into four subsamples: the first including the values observed from 2010 to 2011 (and also the few observations of December 2009 contained in the sample); the second and the third covering the periods from 2012 to 2013, and from 2014 to 2015, respectively; and the fourth containing the remaining observations (from 2016 and 2017). We first calculate the ΔρDCCA between the first and the last sub periods to obtain the total variation of financial integration (Fig. 2). Subsequently, we calculate this variation across the four consecutive sub periods. Results are presented in Figs. 3–5. For the whole sample (Fig. 2), the levels of financial integration have generally decreased. In the case of the emerging markets, India and Malaysia show a slight increase of correlation with the Japan, but solely in the short run (lower time scales), while in what regards the American market the same occurs with Thailand, but for longer time scales. For frontier markets, although three indices have decreasing correlations, that of Bangladesh displays the opposite effect, not only with the regional market but also (and mainly) with the global one. We consider now the correlation changes in the consecutive sub samples. Focusing on the first and the second sub periods (Fig. 3),

Fig. 2. ΔρDCCA (sub period 4 – sub period 1) for Asian (Emerging and Frontier) stock markets as function of n (day). In the left hand side, correlations are with the regional market (Japan) and in the right hand side correlations are with the global market (US). The shaded area corresponds to the area between the limits of the confidence intervals proposed by Guedes et al. (2018a,b). 364

Research in International Business and Finance 50 (2019) 357–368

W. Mohti, et al.

Fig. 3. ΔρDCCA (sub period 2 – sub period 1) for Asian (Emerging and Frontier) stock markets as function of n (day). In the left hand side, correlations are with the regional market (Japan) and in the right hand side correlations are with the global market (US). The shaded area corresponds to the area between the limits of the confidence intervals proposed by Guedes et al. (2018a,b).

the behavior of emerging markets against Japan and the US is distinct: while with the regional market there was an increase in correlation for most indices, with the global market this just happens with Korea and the Philippines, and with more relevance at higher scales. The indices of Taiwan and Thailand also present some correlation increases but of smaller magnitude. The Chinese and Korean indices display a strong decrease in correlations with the US. Regarding the frontier markets, there was an increase in correlations for Bangladesh and Sri Lanka with both the regional and the global markets. This was the period immediately following the occurrence of the subprime crisis, which could indicate that during this particular time span the examined emerging markets appear to have become more segmented vis-a-vis the US, probably as a reaction to the crisis. For the following sub periods (Fig. 4) there was an increase in financial integration of emerging markets with the US, with almost all countries displaying increases in correlations, which may be related to the recovery of the American economy. Vis-a-vis Japan, increases are mostly evident in the short-run. In contrast, for frontier markets, if some increase in correlation with Japan existed in some time scales, vis-a-vis the US correlations are in all cases near zero or significantly decreasing. Finally, when comparing the last two sub periods (Fig. 5), most emerging markets have decreases in correlation with both the US and Japan, although Korea and Thailand present a slight increase in financial integration with the US. Regarding frontier markets, Bangladesh and Pakistan, there is an increase in correlation levels with both the regional and the global benchmarks. Summing up, the results of both the DCCA correlations and of their variation for the whole sample indicate that Asian emerging markets are generally more integrated with the regional and the global markets than frontier ones. A closer examination of the distinct sub samples indicates that emerging markets’ levels of integration with the US appear to have suffered a relevant negative impact, probably due to the subprime crisis, showing some signs of subsequent recuperation. Amongst frontier markets, only Pakistan, and on a smaller scale Vietnam, display integration patterns that, although weaker, are more in line with those observed for emerging markets. 5. Conclusions In this study we assessed Asian emerging and frontier stock markets’ regional and global integration using daily data from December 2009 to April 2017 and two distinct methodologies. The results of the Gregory and Hansen’s cointegration tests generally suggested that there are long run links connecting Indonesia, Korea, the Philippines, and Thailand with Japan (the regional benchmark). With the exception of China, India and the Philippines, the examined emerging markets also displayed evidence of long run connections with the global market (proxied by the US). Amongst frontier markets, only Pakistan shared similar long run relationships with Japan and the US. As the three models considered for these tests produced mixed outcomes, the analysis was complemented with DCCA, a technique that is robust regardless of the (non) stationarity of the underlying data. DCCA correlation coefficients produced more comprehensive evidence of integration, with all emerging markets displaying statistically significant regional and global links. In the frontier market bloc, the same occurred solely for Pakistan, and to a lesser 365

Research in International Business and Finance 50 (2019) 357–368

W. Mohti, et al.

Fig. 4. ΔρDCCA (sub period 3 – sub period 2) for Asian (Emerging and Frontier) stock markets as function of n (day). In the left hand side, correlations are with the regional market (Japan) and in the right hand side correlations are with the global market (US). The shaded area corresponds to the area between the limits of the confidence intervals proposed by Guedes et al. (2018a,b).

Fig. 5. ΔρDCCA (sub period 4 – sub period 3) for Asian (Emerging and Frontier) stock markets as function of n (day). In the left hand side, correlations are with the regional market (Japan) and in the right hand side correlations are with the global market (US). The shaded area corresponds to the area between the limits of the confidence intervals proposed by Guedes et al. (2018a,b).

extent for Vietnam. In these two cases, association with the global market was stronger. The reclassification of Pakistan as an emerging market in 2017, announced by MSCI in 2016, may have improved the attractiveness of Pakistan’s stock market for foreign investors and impacted its performance. The other two assessed frontier countries (Bangladesh and Sri Lanka) did not display 366

Research in International Business and Finance 50 (2019) 357–368

W. Mohti, et al.

significant relationships with the regional and the global markets. Analysis of the correlation coefficients’ variation across time indicated that most markets have become more segmented. This result, obtained when considering the complete period of time covered by the study, was confirmed for each of the smaller consecutive time periods. There was at first an increase in integration with Japan and a decrease with the US, some reversion of this initial pattern in the middle of the sample, and a decrease in the last sub period. This empirical analysis may be of use for international investors interested in expanding the geographical scope of their portfolios. The results suggest that the relatively more segmented Asian frontier stock markets may offer greater potential diversification benefits than those available in emerging markets. The latter, according to our assessment, display stronger sings of integration and are thus more prone to move in tandem with the regional and the global benchmarks. Acknowledgement This paper is financed by National Funds of the FCT – Portuguese Foundation for Science and Technology within the project UID/ ECO/04007/2019″. References Adam, K., Jappelli, T., Menichini, A., Padula, M., Pagano, M., 2002. Analyse, Compare, and Apply Alternative Indicators and Monitoring Methodologies to Measure the Evolution of Capital Market Integration in the European Union. Report to the European Commission. pp. 1–95. Al Nasser, O.M., Hajilee, M., 2016. Integration of emerging stock markets with global stock markets. Res. Int. Bus. Finance 36, 1–12. Alagidede, P., 2008. African stock market integration: implications for portfolio diversification and international risk sharing. Proceedings of the African Economic Conferences. pp. 1–31. Alkulaib, Y.A., Najand, M., Mashayekh, A., 2009. Dynamic linkages among equity markets in the Middle East and North African countries. J. Multinatl. Financ. Manag. 19 (1), 43–53. Aloui, C., Hkiri, B., 2014. Co-movements of GCC emerging stock markets: new evidence from wavelet coherence analysis. Econ. Model. 36, 421–431. Ausloos, M., Ivanova, K., 2000. Introducing false EUR and false EUR exchange rates. Phys. A Stat. Mech. Appl. 286 (1-2), 353–366. Baele, L., Ferrando, A., Hördahl, P., Krylova, E., Monnet, C., 2004. Measuring Financial Integration in the Euro Area. European Central Bank Occasional Paper (No. 14). Bekaert, G., Harvey, C.R., 2014. Emerging Equity Markets in a Globalizing World. available at SSRN: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2344817. Bekaert, G., Harvey, C.R., Lumsdaine, R.L., 2002. Dating the integration of world equity markets. J. Financ. Econ. 65 (2), 203–247. Bessler, D.A., Yang, J., 2003. The structure of interdependence in international stock markets. J. Int. Money Finance 22 (2), 261–287. Bracker, K., Docking, D.S., Koch, P.D., 1999. Economic determinants of evolution in international stock market integration. J. Empir. Finance 6 (1), 1–27. Cheng, H., 2000. Cointegration Test for Equity Market Integration: The Case of the Great China Economic Area (Mainland China, Hong Kong, and Taiwan). George Washington University, Japan, and the United States. Cheung, Y.L., Mak, S.C., 1992. The international transmission of stock market fluctuation between the developed markets and the Asian—Pacific markets. Appl. Financ. Econ. 2 (1), 43–47. Cheung, Y.W., Chinn, M.D., Fujii, E., 2006. The Chinese economies in global context: the integration process and its determinants. J. Jpn. Int. Econ. 20 (1), 128–153. Chien, M.S., Lee, C.C., Hu, T.C., Hu, H.T., 2015. Dynamic Asian stock market convergence: evidence from dynamic cointegration analysis among China and ASEAN-5. Econ. Model. 51, 84–98. Choudhry, T., Lu, L., Peng, K., 2007. Common stochastic trends among Far East stock prices: effects of the Asian financial crisis. Int. Rev. Financ. Anal. 16 (3), 242–261. Cuestas, J.C., Filipozzi, F., Staehr, K., 2015. Uncovered Interest Parity in Central and Eastern Europe: Sample, Expectations and Structural Breaks. Sheffield Economics Research Paper Series (SERPS), 201501(014). Égert, B., Kočenda, E., 2007. Interdependence between Eastern and Western European stock markets: evidence from intraday data. Econ. Syst. 31 (2), 184–203. Engle, R.F., Granger, C.W., 1987. Co-integration and error correction: representation, estimation, and testing. Econometrica 251–276. Engsted, T., Tanggaard, C., 2004. The comovement of US and UK stock markets. Eur. Financ. Manag. 10 (4), 593–607. Feldstein, M.S., Horioka, C.Y., 1979. Domestic savings and international capital flows. Econ. J. 90, 314–329. Ferreira, P., 2017. Portuguese and Brazilian stock market integration: a non-linear and detrended approach. Port. Econ. J. 16 (1), 49–63. Ferreira, P., Dionísio, A., 2016. How long is the memory of the US stock market? Phys. A Stat. Mech. Appl. 451, 502–506. Filipozzi, F., Staehr, K., 2012. Uncovered interest parity in Central and Eastern Europe: convergence and the global financial crisis. Discus. Estonian Econ. Pol.: Theory Pract. Econ. Policy 20, 58–82. Fujiwara, I., Takahashi, K., 2012. Asian financial linkage: MacroFinance dissonance. Pacific Econ. Rev. 17, 136–159. Gilmore, C.G., McManus, G.M., 2002. International portfolio diversification: US and Central European equity markets. Emerg. Mark. Rev. 3 (1), 69–83. Goetzmann, W.N., Li, L., Rouwenhorst, K.G., 2005. Long-term global market correlations. J. Bus. 78 (1), 1–38. Gregory, A.W., Hansen, B.E., 1996. Residual-based tests for cointegration in models with regime shifts. J. Econom. 70 (1), 99–126. Grubel, H.G., 1968. Internationally diversified portfolios: welfare gains and capital flows. Am. Econ. Rev. 58 (5), 1299–1314. Guedes, E.F., Brito, A.A., Oliveira Filho, F.M., Fernandez, B.F., de Castro, A.P.N., da Silva Filho, A.M., Zebende, G.F., 2018a. Statistical test for ΔρDCCA crosscorrelation coefficient. Phys. A Stat. Mech. Appl. 501, 134–140. Guedes, E.F., Brito, A.A., Oliveira Filho, F.M., Fernandez, B.F., de Castro, A.P.N., da Silva Filho, A.M., Zebende, G.F., 2018b. Statistical test for ΔρDCCA: methods and data. Data Brief 18, 795–798. Guidi, F., Ugur, M., 2014. An analysis of South-Eastern European stock markets: evidence on cointegration and portfolio diversification benefits. J. Int. Financ. Mark. Inst. Money 30, 119–136. Gupta, R., Guidi, F., 2012. Cointegration relationship and time varying co-movements among Indian and Asian developed stock markets. Int. Rev. Financ. Anal. 21, 10–22. Horvath, R., Petrovski, D., 2013. International stock market integration: central and South Eastern Europe compared. Econ. Syst. 37 (1), 81–91. Jayasuriya, S.A., Shambora, W., 2009. Oops, we should have diversified!. Appl. Financ. Econ. 19 (22), 1779–1785. Johansen, S., 1988. Statistical analysis of cointegration vectors. J. Econ. Dyn. Control 12 (2-3), 231–254. Kasa, K., 1992. Common stochastic trends in international stock markets. J. Monet. Econ. 29 (1), 95–124. Kearney, C., Lucey, B.M., 2004. International equity market integration: theory, evidence and implications. Int. Rev. Financ. Anal. 13 (5), 571–583. Kim, S., Lee, J.W., Shin, K., 2006. Regional and Global Financial Integration in East Asia. Institute of Economic Research Korea University Discussion Paper Series, 0602. Kiviaho, J., Nikkinen, J., Piljak, V., Rothovius, T., 2014. The co‐movement dynamics of European frontier stock markets. Eur. Financ. Manag. 20 (3), 574–595. Korajczyk, R.A., 1996. A measure of stock market integration for developed and emerging markets. World Bank Econ. Rev. 10 (2), 267–289. Lean, H.H., Teng, K.T., 2013. Integration of world leaders and emerging powers into the Malaysian stock market: a DCC-MGARCH approach. Econ. Model. 32, 333–342. Lee, G., Jeong, J., 2016. An investigation of global and regional integration of ASEAN economic community stock market: dynamic risk decomposition approach.

367

Research in International Business and Finance 50 (2019) 357–368

W. Mohti, et al.

Emerg. Mark. Finance Trade 52 (9), 2069–2086. Ma, F., Wei, Y., Huang, D., 2013. Multifractal detrended cross-correlation analysis between the Chinese stock market and surrounding stock markets. Phys. A Stat. Mech. Appl. 392 (7), 1659–1670. Masih, A.M., Masih, R., 1999. Are Asian stock market fluctuations due mainly to intra-regional contagion effects? Evidence based on Asian emerging stock markets. Pacific-Basin Finance J. 7 (3-4), 251–282. Mensi, W., Shahzad, S.J.H., Hammoudeh, S., Zeitun, R., Rehman, M.U., 2017. Diversification potential of Asian frontier, BRIC emerging and major developed stock markets: a wavelet-based value at risk approach. Emerg. Mark. Rev. 32, 130–147. Mukherjee, P., Bose, S., 2008. Does the stock market in India move with Asia? A multivariate cointegration-vector autoregression approach. Emerg. Mark. Financ. Trade 44 (5), 5–22. Neaime, S., 2012. The global financial crisis, financial linkages and correlations in returns and volatilities in emerging MENA stock markets. Emerg. Mark. Rev. 13 (3), 268–282. Obstfeld, M., 1993. Are Industrial-country Consumption Risks Globally Diversified? National Bureau of Economic Research, pp. w4308. Obstfeld, M., 1994. Risk taking, global diversification and growth. Am. Econ. Assoc. 84, 1310–1329. Peng, C.K., Buldyrev, S.V., Havlin, S., Simons, M., Stanley, H.E., Goldberger, A.L., 1994. Mosaic organization of DNA nucleotides. Phys. Rev. E 49 (2), 1685. Podobnik, B., Stanley, H.E., 2008. Detrended cross-correlation analysis: a new method for analyzing two nonstationary time series. Phys. Rev. Lett. 100 (8), 084102. Podobnik, B., Jiang, Z.Q., Zhou, W.X., Stanley, H.E., 2011. Statistical tests for power-law cross-correlated processes. Phys. Rev. E 84 (6), 066118. Richards, A.J., 1995. Comovements in national stock market returns: evidence of predictability, but not cointegration. J. Monet. Econ. 36 (3), 631–654. Rua, A., Nunes, L.C., 2009. International comovement of stock market returns: a wavelet analysis. J. Empir. Finance 16 (4), 632–639. Sharma, G.D., Bodla, B.S., 2011. Inter-linkages among stock markets of South Asia. Asia-Pac. J. Bus. Adm. 3 (2), 132–148. Sharma, S.C., Wongbangpo, P., 2002. Long‐term trends and cycles in ASEAN stock markets. Rev. Financ. Econ. 11 (4), 299–315. Shi, W., Shang, P., 2013. Cross-sample entropy statistic as a measure of synchronism and cross-correlation of stock markets. Nonlinear Dyn. 71 (3), 539–554. Da Silva, M., Pereira, É., da Silva Filho, A., de Castro, A., Miranda, J., Zebende, G., 2016. Quantifying the contagion effect of the 2008 financial crisis between the G7 countries (by GDP nominal). Phys. A Stat. Mech. Appl. 453, 1–8. Speidell, L.S., Krohne, A., 2007. The case for frontier equity markets. J. Invest. 16 (3), 12–22. Syriopoulos, T., 2007. Dynamic linkages between emerging European and developed stock markets: Has the EMU any impact? Int. Rev. Financ. Anal. 16 (1), 41–60. Voon, J., Yue, R., 2003. China-ASEAN export rivalry in the US market. J. Asia Pac. Econ. 8 (2), 157–179. Wang, P., Moore, T., 2008. Stock market integration for the transition economies: time‐varying conditional correlation approach. Manchester Sch. 76, 116–133. Wang, Y., Wei, Y., Wu, C., 2010. Cross-correlations between Chinese A-share and B-share markets. Phys. A Stat. Mech. Appl. 389 (23), 5468–5478. Wong, W.K., Penm, J., Terrell, R.D., Ching, K.Y., 2004. The relationship between stock markets of major developed countries and Asian emerging markets. J. Appl. Math. Decis. Sci. 8 (4), 201–218. Yang, J., Kolari, J.W., Min, I., 2003. Stock market integration and financial crises: the case of Asia. Appl. Financ. Econ. 13 (7), 477–486. Yu, I.W., Fung, K.P., Tam, C.S., 2010. Assessing financial market integration in Asia–equity markets. J. Bank. Finance 34 (12), 2874–2885. Yu, J.S., Hassan, M.K., 2008. Global and regional integration of the Middle East and North African (MENA) stock markets. Q. Rev. Econ. Financ. 48 (3), 482–504. Zebende, G.F., 2011. DCCA cross-correlation coefficient: quantifying level of cross-correlation. Phys. A Stat. Mech. Appl. 390 (4), 614–618. Zivot, E., Andrews, D.W.K., 2002. Further evidence on the great crash, the oil-price shock, and the unit-root hypothesis. J. Bus. Econ. Stat. 20 (1), 25–44.

368