Correlation Dynamics in Equity Markets

Evidence from India

Executive Summary

Equity market integration has wider notion in finance literature. Markets are said to be
highly integrated only if irrespective of the market, assets with similar risk have identical
expected return. Albeit this, understanding the correlation structure and dynamics of the
equity markets of the world is the first step in getting the bigger picture of market
integration. Without a good correlation structure, other aspects of market integration are
not theoretically reflective. Keeping that in mind this study aimed at analyzing the
correlation structure of the Indian equity markets with that of world markets. This paper
used daily data from 1st July 1997 to 18th August 2006 of the following 11 world indices:
NASDAQ Composite (USA), S&P 500 (USA), FTSE 100 (UK) and DAX 30 (Germany)
are classified as developed markets, whereas KLSE Composite (Malaysia), Jakarta
Composite (Indonesia), Straits Times (Singapore), Seoul Composite (South Korea),
Nikkei (Japan), Taiwan Weighted Index (Taiwan) and the S&P CNX Nifty (India) are
considered as Asian markets.

The following three generic correlation measures are derived. All markets considered the
entire 11 markets specified, Asian markets considered only the 7 markets classified,
developed markets considered only the 4 markets classified. Further to get deeper insight
on the individual correlation structure between S&P CNX Nifty with world markets two
other measures are derived. S&P CNX Nifty-Asian considered S&P CNX Nifty with other
6 Asian markets and S&P CNX Nifty-Developed considered S&P CNX Nifty with the 4
developed markets. The following two methods are used to derive the correlation
structure: i) unconditional correlation estimate and ii) dynamic time varying correlation
estimate using a DCC-MVGARCH of Engle and Sheppard (2001). Both these estimates
are exhibiting a poor correlation with an average correlation is below 30 percent. We
used BSE 100 Index and estimated the unconditional correlation as a robustness check.
The results found to be in similar pattern displayed by S&P CNX Nifty Index. The
highest correlation is resulted for 4 developing countries specified with around 60
percent. The individual correlation structures between S&P CNX Nifty with other
markets are fairly lower than other estimates.

In addition a Logistic Smooth Transition Regression (LSTR) model is implemented for

the derived correlation series to identify potential regime shift in the correlation dynamics
and to categorize the phase of integration across these markets. The LSTR results for the
conditional time varying correlation of S&P CNX Nifty-Asian and S&P CNX Nifty-
Developed shows that there is a significant regime shift in the year 2000 and there is a
considerable increase in integration in the second regime. This indicates that the S&P
CNX Nifty index is moving towards a better integration with other world markets but not
at a very noteworthy phase. The high volatility in recent years faced by the Indian equity
markets can be attributed to this low level of correlation and market integration with
other world markets as it provides space for the global funds to diversify risk.
 Correlation Dynamics in Equity Markets
Evidence from India

S. Raja Sethu Durai

Saumitra N Bhaduri

Madras School of Economics, Gandhi Mandapam Road, Chennai 600 025, India

1. Introduction

Equity market integration plays a very significant role in shaping the fortunes of any
developing nation. The foreseeable benefits apply not only to the realm of financial
markets but for economic growth and development itself: First, in a fully integrated
capital market all risk factors trade at the same price. The law of one price will apply to
all securities. This per se should have a positive effect on the functioning of financial
markets and indirectly on the performance of investments. Second, greater integration
would mean a free or relatively freer access to foreign financial markets. This better
access would provide many firms a broader source for fund raising. Third, more
internationally diversified stock and bond portfolios should as a consequence shift the
frontier of efficient portfolios upwards and therefore for each given risk the average
portfolio return should increase. It would create enormous opportunities for domestic and
international investors to diversify their portfolios across the globe. Fourth, equity market
integration is of considerable significance to issuers and investors as it plays a critical
role in channeling funds. Stock markets tend to be very efficient in the allocation of
capital to its highest-value users. Such integrated markets could also help to increase
savings and investment, which are essential for economic development. An equity
market, by allowing diversification across a variety of assets, helps reduce the risk the
investors must bear, thus reducing the cost of capital, which in turn spurs investment and
economic growth.

A high degree of integration is not without its limitations. One constant argument
however is that these limitations of integration do not begin to have an impact until very
high rates of integration is achieved. Issues such as vulnerability to foreign price
fluctuations, drain of domestic funds and the argument that excessive integration could be
self-defeating are valid only when the degree of correlation between markets is very high.
Consequently the obvious advantages are what most emerging economies focus on in
their drive towards greater integration.

In literature measuring market integration has been done broadly through three ways.
First, testing the segmentation of the equity markets via the international CAPM. It
typically assumes that all the world’s capital markets are perfectly integrated and
therefore the asset risk can be related purely with the covariance of the local returns with
the world market portfolio. Second, a significant number of studies have examined the
integration through increasing correlations and cointegration in their returns over time.
Third, time varying estimates rectifies the weakness in the above-mentioned methods that
misses the important element of time variation in equity risk premia.

Market integration is something more than the correlation structure across the markets.
Understanding the correlation structure is the first step towards understanding the wider
notion of market integration. Without a good correlation structure, other aspects of
market integration are not theoretically reflective. This study is an attempt to analyze the
correlation structure and to test the equity market integration between the Indian equity
market with some of the major World markets including the Asian markets.

The main contribution of this study over some of the previous studies is in two fold. First,
this study uses DCC-MVGARCH model to estimate the dynamic correlation among the
equity market of a developing country (India) with the World and Asian markets along
with simple unconditional correlation. Second, a Logistic Smooth Transition Regression
(LSTR) method is used to estimate not only the extent of correlation between returns but
the also the pace of integration. The advantage of logistic trend models is that they can
indicate the speed at which markets are getting integrated, information that cannot be
attained through conventional correlation analysis. This paper is organized as follows.
Section 2 gives a brief overview of the literature on equity market integration. Section 3
narrates the methodology to estimate the unconditional correlation, conditional
correlation through DCC-MVGARCH model and explains the logistic smooth transition
regression. Section 4 discusses the results and the final section concludes with a
summary.

2. Literature

A rigorous test for equity market integration has had an interesting past, with varied
conclusion being made on the back of a wide range of studies. One of the most striking
features of financial integration is the extent of literature that exists in the topic. The
body of literature can be classified based on the approach adopted by the author both in
terms of econometric method as well as theoretical underpinning of the transmission
mechanism. Since the seminal work of Grubel (1968), which expounded the benefits of
international portfolio diversification, the relationship among national stock markets has
been analyzed in a series of studies such as Granger and Morgenstern (1970), Ripley
(1973), Lessard (1974,1976) and Panton, Lessig and Joy (1976) among others.

Other work in the field includes Eun and Shim (1989) VAR models to measure
transmission of stock movements, providing evidence of co-movements between the US
market and other world equity markets, Koutmos and Booth (1995) studied the
asymmetric volatility transmissions in international stock markets using an Exponential
GARCH model. In recent times Chelley-Steeley (2005) used a bivariate model along
with logistic smooth transition regression to establish how rapidly the countries of
Eastern Europe are moving away from market segmentation. Kearney and Poti (2006)
examined the correlation dynamics for European equity markets using an asymmetric
DCC-MVGARCH specification and found evidence in favor of structural break at the
beginning of the process of monetary integration in Euro-zone.

A small body of literature exists in the Indian context, which predominantly depends on
the bivariate and multivariate cointegration analysis. The study by Kumar and
Mukhopadhyay (2002) uses a two-stage GARCH model and an ARMA-GARCH model
to captures the mechanism by which NASDAQ daytime returns impacts not only the
mean but also conditional volatility of Nifty overnight returns. Ignatius (1992) compared
returns on the BSE Sensex with those on the NYSE S&P 500 Index and found no
evidence of integration. Agarwal (2000) concluded that there is a lot of scope for the
Indian stock market to integrate with the world market after having found a correlation
coefficient of 0.01 between India and developed markets. By using Granger causality
relationship and the pair wise, multiple and fractional cointegration, Wong, Agarwal and
Du (2005) have found that the Indian stock market is integrated with the matured markets
of the World. Nath and Verma (2003) tested for cointegration between the Nifty, STI and
Taiex and found no evidence in favor of cointegration.

Though research on India has evolved constantly both in terms of econometric techniques
and the focus of study, nevertheless, a gap still exists in terms of literature on India. The
existing literature though provides valuable insights into the extent of integration; none of
the studies have really focused on the underlying dynamics of the process of integration
over time. The liberalization effort has often been assumed to be instantaneous process
rather than a gradual one. This study, in contrast, attempts to measure the pace of
integration, which at its core works on the premise that integration follows a traceable
path over time. This potentially serves as a critical indicator of the strength of the
integration process, and provides an insight into the longevity of the same.

3. Methodology

3.1 Unconditional Correlation Estimates

The unconditional correlations are derived from the conventional second moments on
asset returns. This method is widely used in the literature because of its simplicity. The
estimates are computed from the cross products of the standardized daily log-return Rit
deviations from their monthly sample means and sum them to obtain monthly non-
overlapping correlation estimates for each pair of indices i and j.
 ∑ (R
n
i ,t − n − Ri ,t ) ( R j ,t − n − R j ,t )
ci , j ,t = 0
(1)
∑ (R i ,t − n − Ri ,t ) ∑0 ( R j ,t − n − R j ,t )
n 2 n 2
0

From the above pair wise correlation we can get the equally weighted average correlation
across the market indices as follows
n
1 n 1
UCCORRt = ∑ ∑ c i , j ,t (2)
i −1 n j =1 n
Where n is the number of market indices.

3.2 Conditional Correlation Estimates

The main shortcomings of the above mentioned unconditional correlation estimates are
two fold. First, the average of squares and cross products are consistent estimators of the
second moments of the return distributions, albeit its consistency ad hoc representation of
the volatility and correlation process, it might be biased in small samples. Second,
aggregating the daily data to get monthly estimates of correlation will result in potential
small sample problem. To overcome these deficiencies in the unconditional correlation
estimates we apply the recently developed DCC-MVGARCH model of Engle (2002) and
Engle and Sheppard (2001).

This class of MVGARCH models differs from other specification as it was designed to
allow for two-stage estimation. In the first stage the univariate GARCH model is
estimated for each series in examination and the residual series are obtained and in the
second stage these residual along with the standard deviation obtained from the first stage
are used to estimate the dynamic correlation. In this study we followed Engle and
Sheppard (2001) methodology to estimate the dynamic correlation among the market
indices. The returns from n indices are conditionally multivariate normal with zero
expected value and covariance matrix Ht.
Rt ℑt −1 ~ N (0, H t ) and H t ≡ Dt Ct Dt (3)

where Dt is nxn diagonal matrix of time varying standard deviations from univariate
GARCH models with √hit on the ith diagonal, and Ct is the time varying correlation
matrix. The elements of Dt are described as hit that takes the form
 P Q
hit = ω i + ∑ α ip Rit2− p + ∑ β iq hit − q (4)
p =1 q =1

For i = 1, 2 … with usual GARCH restrictions for non-negativity and stationarity being
imposed. The subscripts p and q are the lag length of the each series, with this Engle and
Sheppard (2001) derived the dynamic correlation structure as follows:
Qt = (1 − α − β ) Q + α (ε t −1ε t' −1 ) + β Qt −1

Ct = Qt*−1Qt Qt*−1

where Q is the unconditional covariance of the standardized residuals resulting from the
first stage estimation and Qt* is the diagonal matrix consists of square root of diagonal
qijt
elements of the Qt. So Ct will be the correlation matrix that takes the form ρijt = .
qii q jj

3.3 Logistic Smooth Transition Regression

The logistic smooth transition function has a long tradition in the statistical modeling of
changing regimes, being introduced by Bacon and Watts (1971) and extended to time
series regression and autoregressive models by Lin and Teräsvirta (1994) and Granger
and Teräsvirta (1993). The LSTR model identifies any fundamental change as a single
structural break that leads to a smooth transition between two regimes as opposed to an
instantaneous shift in the underlying relationship. In this paper, we employ a variation of
the simple, nonlinear, smooth transition logistic trend model suggested by Granger and
Teräsvirta (1993). The smooth transition model is applied to the equity market
correlations generated from the unconditional and condition correlation estimates. The
Logistic Smooth Transition Regressive model is generally defined as follows,
ρij ,t = α + βSt (γ ,τ ) + ν t (5)

St (γ ,τ ) = (1 + exp(−γ (t − τT )))−1 , γ > 0 (6)

where ρij,t is the correlation between the NSE index (i) and one of the market index in
examination (j) at time t with St playing the role of a smooth transition continuous
function bounded between 0 and 1. The α and β are coefficients. The parameter τ
determines the timing of the transition midpoint and γ measures the speed of adjustment.
For γ > 0 we have S-∞ (γ;τ) = 0, S+∞ (γ;τ) = 1 and St (γ;τ) = 0.5. In the limiting case

when γ = 0, St (γ;τ) = 0 for all t and no integration takes place. The trend component has
been removed from the model, as there is no reason to expect equity market correlations
to exhibit a trend increase. If γ <0 the initial and final states of the model are reversed,
however the interpretations of the various parameters still remain the same. Further, the
model assumes ρij,t to be stationary around a mean that changes from an initial value of
‘α’ (prior to integration) to ‘α+β’. Thus, ‘α‘ is a measure of market integration in the
first regime and ‘β’ is the increase (if b is positive) or decrease (if b is negative) in
market integration in the second regime.

4. Results

This paper considered the following 11 indices. As a representative of developed

markets, NASDAQ Composite (USA), S&P 500 (USA), FTSE 100 (UK) and DAX 30
(Germany) are taken. Asian markets include S&P CNX Nifty (India), KLSE Composite
(Malaysia), Jakarta Composite (Indonesia), Straits Times (Singapore), Seoul Composite
(South Korea), Nikkei (Japan) and Taiwan Weighted Index (Taiwan). Daily data from 1st
July 1997 to 18th August 2006 are taken from NSE India website (www.nseindia.com) for
S&P CNX Nifty index and Yahoo Finance (finance.yahoo.com) for all other indices.

The study period we have selected is dictated by two facts. First, we use S&P CNX Nifty
as benchmark index for India, so we have data only from July1997. Second, we get
consistent data for all other markets in question for the specified time period only.

4.1 Unconditional Correlation Estimates

A set of unconditional correlation estimates have been derived. It includes correlation for
all the markets taken, correlation for Asian markets and correlation for developed
markets. As explained in Section 3, estimation of unconditional correlation involves three
steps. First calculating monthly mean returns of each index for 110 months from July
1997 to August 2006. Second calculating non overlapping pair wise correlation for the
 returns of each two indices using Equation1 and finally the average of these pair wise
correlations will give us the unconditional correlation estimates across all these markets,
while the stand alone average of Asian markets and the developed markets will give us
the unconditional correlation of those particular markets.

Table 1 list the average pair wise correlation structure between the markets and Table 2
provides the average unconditional correlation estimates for all the three market
classification i.e. all markets, Asian markets and developed markets.

Table 1: Average Pair wise Correlation Structure

S&P
Markets CNX S&P FTSE DAX Strait
Nifty NASDAQ 500 100 30 KLSE JTSE Times Seoul NIKKEI Taiwan
S&P CNX
Nifty 1.000
NASDAQ 0.118 1.000
S&P 500 0.103 0.882 1.000
FTSE 100 0.206 0.447 0.480 1.000
DAX 30 0.191 0.524 0.533 0.749 1.000
KLSE 0.126 0.120 0.094 0.134 0.136 1.000
JTSE 0.232 0.046 0.037 0.120 0.121 0.238 1.000
Strait Times 0.241 0.230 0.202 0.349 0.339 0.373 0.298 1.000
Seoul 0.291 0.231 0.214 0.298 0.321 0.257 0.243 0.454 1.000
NIKKEI 0.244 0.267 0.228 0.347 0.346 0.293 0.207 0.455 0.488 1.000
Taiwan 0.183 0.217 0.171 0.213 0.230 0.230 0.210 0.414 0.423 0.390 1.000

Table 2: Average Unconditional Correlation Estimate with S&P CNX Nifty

Market Average Unconditional Correlation
Estimate
All Markets 0.2824
Asian Markets 0.2994
Developed Markets 0.6026

The average unconditional correlation estimates clearly shows that the developed markets
are highly integrated with higher correlation estimate whereas the Asian markets are
integrated very lower with low correlation estimate. This also pulls down the average all
market correlation to further lower level.
To understand the correlation between S&P CNX Nifty index with the Asian markets and
other developed markets, we calculate non overlapping pair wise correlation for the
returns of S&P CNX Nifty index with other indices using Equation1 and average it to get
unconditional correlation estimates. The average unconditional correlation estimate of
S&P CNX Nifty index with Asian markets is 0.2193 where as with the developed
markets it is very lower with 0.1545.

The results clearly exhibits the week equity market integration across the Asian markets
and in particular Indian equity market integration with both Asian and developed
markets. As a robustness check we used BSE 100 index instead of S&P CNX Nifty index
as a representative of Indian equity market. The results are in the similar pattern with
slightly lower correlation than the S&P CNX Nifty index. Table 3 presents the results.

Table 3: Average Unconditional Correlation Estimate with BSE 100

Market Average Unconditional Correlation
Estimate
All Markets 0.2815
Asian Markets 0.2989
Developed Markets 0.6026
BSE-Asian Markets 0.2176
BSE-Developed Markets 0.1453

4.2 Conditional Correlation Estimates

The advantage of conditional correlation estimates is that it gives a dynamic time varying
estimates. So there is no need to find the deviation of daily return from the monthly mean
to calculate the correlation for each month. This removes the disadvantages associated
with averaging. The DCC-MVGARCH model is used on the daily returns of the above
mentioned indices to derive the time varying conditional correlation estimates1. Table 4
presents with the parameter estimates of the model in two segments. First the estimates of
univariate GARCH (1, 1) models of the individual indices, followed by the DCC-
MVGARCH (1, 1) estimates. Engle and Sheppard’s (2001) test for constant correlation

1
The model is estimated in MATLAB using GARCH Tool Box
among the returns rejected the null of constant correlation in favor of a time varying
correlation matrix.
Table 4: Estimates of the DCC-MVGARCH model
Coefficient
Univariate GARCH (1, 1)
ω 0.0025** (0.0000)
S&P CNX Nifty α1 0.1044** (0.0017)
β1 0.8264** (0.0071)
ω 0.0002** (0.0000)
NASDAQ α1 0.0751** (0.0005)
β1 0.9249** (0.0004)
ω 0.0002** (0.0000)
S&P 500 α1 0.0897** (0.0007)
β1 0.9015** (0.0008)
ω 0.0002** (0.0000)
FTSE 100 α1 0.0962** (0.0003)
β1 0.8945** (0.0003)
ω 0.0006** (0.0000)
DAX 30 α1 0.0979** (0.0007)
β1 0.8911** (0.0006)
ω 0.0001** (0.0000)
KLSE α1 0.0556** (0.0012)
β1 0.9444** (0.0011)
ω 0.7778 (0.5319)
JTSE α1 0.0957 **(0.0036)
β1 0.0001 (0.0001)
ω 0.0003** (0.0000)
Strait Times α1 0.1118** (0.0014)
β1 0.8882** (0.0011)
ω 0.0003** (0.0000)
Seoul α1 0.0739** (0.0009)
β1 0.9257** (0.0008)
ω 0.0007** (0.0000)
NIKKEI α1 0.0824** (0.0004)
β1 0.8949** (0.0006)
ω 0.0004** (0.0000)
Taiwan α1 0.0704** (0.0007)
β1 0.9229** (0.0007)
DCC-MVGARCH (1, 1)
α 0.0123** (0.0000)
β 0.9620** (0.0002)
Note: Standard errors in parentheses and ** denotes significance at 1% level
The results show a significant parameter estimates for DCC-MVGARCH (1, 1) model.
Figure 1 plots the three conditional correlation estimates derived for returns of all the
markets, Asian markets and the Developed markets. The graph clearly shows that none of
these conditional correlation estimates are showing any significant trend. Over the past
decade the correlation among these markets are flat, in other words we can say the market
integration across these markets are naïve.

Figure 2 graphs the time varying conditional correlation estimates derived for S&P CNX
Nifty index with developed markets and Asian markets. The graph shows a slight upward
movement across S&P CNX Nifty and other Asian markets and the developed markets in
the recent years. The average of this conditional correlation estimates are again gives a
similar picture that of unconditional correlation estimates. Table 5 presents the results.
While considering correlation on returns as an initial aspect of equity market integration,
these results clearly shows that the Indian equity market is still in the infancy with respect
to world market integration.

Figure 1: Conditional Correlation Estimates

Figure 2: Conditional Correlation Estimates between S&P CNX Nifty and Asian &
Developed Markets

Table 5: Average Conditional Correlation Estimate with S&P CNX Nifty

Market Average Conditional Correlation
Estimate
All Markets 0.2777
Asian Markets 0.2747
Developed Markets 0.6145
S&P CNX -Asian Markets 0.2321
S&P CNX -Developed Markets 0.1924

In literature there are many studies that established the emerging markets have relatively
low correlation with the developed markets. Low correlation provides a space for
international diversification opportunities and in effect provides an explanation for
flooding capital towards emerging markets. This capital flow is not only for risk
diversification but also for the returns in emerging markets that are much higher than the
developed markets. Goetzmann and Jorion (1999) found that the returns from a sample of
emerging markets are three times higher than that from a sample of developed markets.
We used Logistic Smooth Transition Regression (LSTR) model to qualify this low
correlation exhibited by S&P CNX Nifty index against the Asian and the developed
markets towards the following objectives. Firstly to validate the low correlation as an
indicator of low level of market integration and secondly to check for any possible
movement it shows towards integration.

4.3 Logistic Smooth Transition Regression

Smooth transition analysis is basically an approach to model deterministic structural
change in time series regression. Chelley-Steeley (2004) applied that for equity markets
in Asia-Pacific region to analyze market integration. She argues that first step prior to the
application of smooth transition is to check for stationarity of the correlations and
consistent levels of comovement. Since a stationary correlation would indicate no
breaking point and no change in the level of integration. We perform an augmented
Dickey Fuller (ADF) test on all the correlation series we derived above. The results are
shown Table 6 for unconditional correlation series and Table 7 for conditional time
varying correlation series.
Table 6: Unit Root Test for Unconditional Correlation Series
Correlation Series ADF Statistics
All Markets -2.582**
Asian Markets -3.100**
Developed Markets -1.968*
S&P CNX -Asian Markets -8.463**
S&P CNX -Developed Markets -5.591**
** and * indicates significance at 1% and 5% levels respectively
Table 7: Unit Root Test for Conditional Time varying Correlation Series
Correlation Series ADF Statistics
All Markets -0.437
Asian Markets -0.417
Developed Markets -0.691
S&P CNX -Asian Markets -1.327
S&P CNX -Developed Markets -0.827
** and * indicates significance at 1% and 5% levels respectively
From the above tables, it clearly indicates that for the unconditional correlation series all
the series are stationary whereas for the conditional time varying correlation series all the
series are I(1). As pointed out above running a smooth transition regression for stationary
series is meaningless. Also the idea of running a smooth transition regression is to figure
out the level of market integration by S&P CNX Nifty index towards other Asian and
developed markets. So we estimated the model depicted in equations 5 and 6 only for
S&P CNX Nifty with Asian markets and the developed markets. Table 8 provides the
results.
Table 8: Results from Smooth Transition Model
Parameters S&P CNX-Asian S&P CNX-Developed
α 0.1885 (0.002)** 0.1739 (0.002)**
β 0.0113 (0.003)** 0.0804 (0.002)**
γ 0.4433 (3.041) 0.5119 (0.452)
τ 0.3036 (0.010)** 0.2743 (0.001)**
** and * indicates significance at 1% and 5% levels respectively
A significant τ shows that the series have a regime shift in terms of market integration.
By considering the values of τ we can find the transition midpoint for S&P CNX Nifty-
Asian correlation series in mid of April 2000 and for S&P CNX Nifty-Developed
correlation series it is in January 2000. For both Asian and developed markets β is
positive, that means that there is an increase in market integration in the second regime.
Insignificant γ in both the estimates reflects the fact that the pace of integration is not
very rapid across these markets. The results clearly shows that after the year 2000, S&P
CNX Nifty index is moving towards a better integration with Asian and other developed
markets but not with significant level of speed.

5. Conclusion

This study analyzed the correlation dynamics between Indian equity market with Asian
and other developed markets. For India we considered S&P CNX Nifty index as
representative index and used BSE 100 index for robustness check. Daily data for the
period from 1st July 1997 to 18th August 2006 are considered for the study with 10 other
world indices. Two methods namely unconditional correlation and conditional time
varying correlation with DCC-MVGARCH model are used to extract the correlation
across these markets. Both the measures display a poor correlation across these markets
particularly that of S&P CNX Nifty index with six Asian markets and S&P CNX Nifty
index with four developed markets separately. The average correlation for these markets
are below 30 percent with only the four developed markets show a high correlation
between them with more than 60 percent. Further a smooth transition model is applied to
understand regime shifts in the correlation series as well as the pace of market
integration. The results support a significant regime shift in the year 2000 and after that
the pace of integration across S&P CNX Nifty index with Asian markets and other
developed markets has showed a positive movement but not at very rapid pace. This is in
support of the argument that an emerging market gives space for diversification for
global funds and the high volatility in recent years faced by the Indian equity markets can
be attributed to this fact.

6. References

Agarwal, R N (2000), “Capital Market Development, Corporate Financing Pattern and

Economic Growth in India,” Institute of Economic Growth Discussion Papers
No.20
Bacon, D and D Watts (1971), “Estimating the transition between two intersecting
straight lines,” Biometrika, 58, 525–534.
Chelley-Steeley, P L (2004), “Equity Market Integration in the Asia-Pacific Region: A
Smooth Transition Analysis”, International Review of Financial Analysis, 13,
621-632.
Chelley-Steeley, P L (2005), “Modeling Equity Market Integration Using Smooth
Transition Analysis: A Study of Eastern European Stock Markets”, Journal of
International Money and Finance, 24, 818-831.
Engle, R (2002), “Dynamic Conditional Correlation: A Simple Class of Multivariate
Generalized Autoregressive Conditional Heteroskedasticity Models”, Journal of
Business & Economic Statistics, 20(3), 339-50.
Engle, R and K Sheppard (2001), “Theoretical and Empirical properties of Dynamic
Conditional Correlation Multivariate GARCH,” NBER Working Paper No. 8554.
Eun, C and S Shim (1989), “International Transmission of Stock Market Movements”,
Journal of Financial and Quantitative Analysis, 24, 241–56.
Goetzmann, W and P Jorion (1999), “Re-emerging Markets”, Journal of Financial and
Quantitative Analysis, 34, 1-32.
Granger, CWJ, and O. Morgenstern (1970), “The Predictability of Stock Market Prices”,
Heath Lexington Books, Heath & Co, Lexington.
Granger, CWJ and T Teräsvirta (1993), “Modeling Nonlinear Economic Relationships”,
Oxford University Press, Oxford.
Grubel, H G (1968), “Internationally Diversified Portfolios: Welfare Gains and Capital
Flows," American Economic Review, 58(5), 1299-1314.
Ignatius, R (1992), “The Bombay Stock Exchange: Seasonalities and Investment
Opportunities”, Indian Economic Review, 27(2), 223-227.
Kearney, C and V Poti (2006), “Correlation Dynamics in European Equity Markets”,
Research in International Business and Finance, (forthcoming)
Koutmos, G and G G Booth (1995), “Asymmetric volatility transmission in international
stock markets” Journal of International Money and Finance, 14(6), 747-762.
Kumar, K and C Mukhopadhyay (2002), “Equity Market Interlinkages: Transmission of
Volatility - A Case of US and India”, NSE Working Paper No.16.
Lessard, D R (1974), “World, National, and Industry Factors in Equity Returns”, Journal
of Finance, 29(2), 379-91.
Lessard, D R (1976), “International Diversification,” Financial Analysts Journal,
January/ February, 32-38.
Lin, C J and T Teräsvirta (1994), “Testing the constancy of regression parameters against
continuous structural change”, Journal of Econometrics, 62(2), 211-228.
Nath, G C and S Verma (2003), “Study of Common Stochastic Trend and Cointegration
in the Emerging Markets - A Case Study of India,” Singapore and Taiwan” NSE
Working Paper No. 25.
Panton, D, VP Lessig and O. Joy (1976), “Co-movement of international equity markets:
A taxonomic approach”, Journal of Financial and Quantitative Analysis, 11(3),
415-432.
Ripley, D M (1973), “Systematic Elements in the Linkage of National Stock Market
Indices”, Review of Economics and Statistics, 55(3), 356-61.
Wong W K, A Agarwal and J Du (2005), “Financial Integration for India Stock Market,
a Fractional Cointegration Approach”, National University of Singapore
Working Paper No. WP0501.