Influence of Foreign Institutional Investments (FIIs) on the Indian Stock Market: An Insight by VAR Models

Abstract

The study examines the influence of foreign institutional investments (FIIs) on the Indian equity market and its role in integration with the United States (US) equity market. Different vector autoregression (VAR) models have been employed for sub-periods created by the structural breaks. Despite global recessionary conditions, both purchase and sales of FIIs have steadily increased. The FII inflows and outflows are significantly influenced by the domestic equity market. The exchange rate has no effect on FII inflows; however, outflows are influenced by its change. The US equity market has no influence on FIIs’ inflows but has marginal impact on the outflows.

JEL Classification: C58, C54, G23, G18, G15

Keywords

Foreign institutional investments (FIIs)structural change VAR

Introduction

The world economy has gradually started recovering due to buoyant economic activities in the emerging economies. However, the developed economies are still suffering from compounded factors, including large fiscal deficits, unemployment, inflation and high debts; all these have resulted in very slow economic growth. The sub-prime crisis created an environment of uncertainty and risk. The rising oil and agriculture prices, fuelling inflationary pressures has slowed down global economic recovery. However, the Indian growth story during the period of study is remarkable as its economy has exhibited resilience despite compounded factors, including persistent worldwide recessionary conditions, growing current account deficit and inflationary pressures. The other challenges faced by the Indian economy are volatility in foreign institutional investments (FIIs) flows, slowdown in exports resulting in widening balance of payment (BoP) due to shrinking global demands, increasing oil and commodities prices and existence of alternative attractive markets.

The Government of India (GoI), in 1991–1992, initiated gradual structural, economic reforms and trade liberalisation process in order to bring about substantial economic growth, integrate with global economies and provide market access for attracting foreign investments by removing restrictions and regulations. Due to the growing BoP crisis, the High Level Committee on the Balance of Payments (1993), headed by Mr C. Rangarajan, recommended to shift the composition of external flows to non-debt-creating flows. Further, by moving away from regulatory regime, foreign institutional investors were allowed to invest in both debt and equity markets, and it started in shares and debentures.

The FIIs, in the form of foreign portfolio investments (FPIs), help in enhancing trading volume and market capitalisation. Thus, they also improve functioning of the secondary market by providing an array of attractive investment opportunities of variety of assets having diversified risk, returns and liquidity profiles. Further, FIIs, in general, may lower cost of capital, provide access to cheap global credits, supplement domestic savings and investments and help in capital market reforms. However, FIIs may increase inflation and create asset bubbles, thereby bringing financial instability and volatility in the stock market due to sudden reversal of its inflows. According to Dr Subbarao, former Governor of Reserve Bank of India (RBI):

Capital flows aid growth by providing external capital to sustain an excess of investment over domestic savings. By affording the opportunity of using the world market, an open capital account permits both savers and investors to diversify their portfolio to maximize returns and minimize risks. (RBI Annual report, 2010)

The FIIs follow policies and guidelines of the RBI and Security Exchange Board of India (SEBI), which have changed from time to time due to dynamic domestic and global environment. The guidelines under SEBI (FII) Regulation, 1995 provide its linkage with government policy framework for investment limits in specific sectors. The policy framework has evolved since 1992 till today. The GoI took steady and cautious approach for gradual liberalisation of quantitative restrictions (QRs) by focusing on policy relaxations on investment limits, eligibility criterion for investment and liberalisation of investment instruments for FIIs. Under the Foreign Exchange Regulation Act (FERA), the FIIs registered with RBI should obtain permission to buy, sell and realise capital gains on investments which are made by initial corpus remitted to India, so as to invest in any recognised stock exchanges through designated bank. The FERA was replaced in 2000 by Foreign Exchange Management Act (FEMA), 1999, which now controls foreign exchange-related transactions for FIIs approved by the RBI.

The two routes for FIIs are: first, 70:30 route, wherein 70 per cent of equity and equity-related investments is permissible and balance 30 per cent is for debt; and the second route is 100 per cent debt security investment route. However, our focus is on the normal equity FII route. Furthermore, to provide flexibility to FII composition, Section 15(2) of SEBI FII Regulation, pertaining to restrictions of 70:30 investments in equity and debt, has been removed from October 2008. The FIIs are now allowed to invest in all types of securities, including government securities. They can invest up to 24 per cent of paid-up capital of the company under portfolio investment route.

The FPIs comprise global depository receipt (GDR)/American depository receipt (ADR), FIIs and offshore funds and others. The share of FIIs in FPI was 95.97 per cent during 2003–2004 and it declined to 46.05 per cent in 2006–2007. This significant fall in FIIs was due to meltdown of commodity and equity markets in May–July 2006, fall in Asian markets and tightening of capital control in Thailand.

In 1991–1992, the total foreign investments comprising of foreign direct investment (FDI) and FPI were $1.33 billion. The FPI increased from mere $4 million in 1991–1992 to $38.24 billion in 1994–1995 (Table 1). However, there were outflows of –$0.61 billion in 1998–1999—post-Asian crisis when Thai Bhat was deregulated on 2 July 1997. There was gradual recovery during 1999–2002. The fall in the technology and information technology (IT) stocks caused the bubble burst in April 2001, resulting in the decline of FPIs from $20.21 billion in 2001–2002 to $ 9.79 billion in 2002–2003.

Table 1.

FDIs and FPIs from 1991 to 2011

Year	FDI	FPI	Total
	b$	b$	b$
1991–1992	1.29	0.04	1.33
1992–1993	3.15	2.44	5.59
1993–1994	5.86	35.67	41.53
1994–1995	13.14	38.24	51.38
1995–1996	21.44	27.48	48.92
1996–1997	28.21	33.12	61.33
1997–1998	35.57	18.28	53.85
1998–1999	24.62	-0.61	24.01
1999–2000	21.55	30.26	51.81
2000–2001	40.29	27.6	67.89
2001–2002	61.3	20.21	81.51
2002–2003	50.35	9.79	60.14
2003–2004	43.22	113.77	156.99
2004–2005	60.51	93.15	153.66
2005–2006	89.61	124.92	214.53
2006–2007	228.26	70.03	298.29
2007–2008	348.35	272.71	621.06
2008–2009	351.8	–138.55	213.25
2009–2010	371.82	323.75	695.57

Source: RBI, Table 31: Foreign Investment Inflows, page 53, published in the ‘Handbook of Statistics on Indian Security Market (2009), Security Exchange Board of India’.

The time frame from 2003 to January 2008, known as bull-run period, is characterised by the revival of private foreign capital flows to the emerging market economies due to progressive liberalisation process, flexible exchange rate and strong economic growth. Net inflow during 2007–2008 was $272.71 billion, an increase of 290 per cent from $70.03 billion in 2006–2007.

The sub-prime mortgage crisis occurred during 2007–2008, resulting in worldwide recessionary conditions consequently, due to weak sentiments of investors; the global securities suffered maximum loss from December 2008 to early 2009, which resulted in net outflows of FPI of –$138.55 billion during 2008–2009. Thus, there have been outflows of FPI during two years: 1998–1999 and 2008–2009. However, attractive domestic market conditions facilitated net inflow of FIIs of FPIs $323.75 billion during 2009–2010 (Table 1). The gross purchase of debt and equity FIIs increased from $ 131.97 billion in 2008–2009 to $179.50 billion in 2009–2010, an increase of 36.02 per cent and combined gross sales increased from $141.81 billion to $149.25 billion, an increase of 5.26 per cent during same period. The cumulative investment by FIIs (at acquisition cost) increased by 51.22 per cent; from $59.08 billion to $89.34 billion for March 2009 to March 2010.¹

Net FII inflow in the equity segment was $23.37 billion during 2009–2010, a total reversal from the outflow of –$10.24 billion during 2008–2009, which is an increase of 128.2 per cent from the preceding year. This remarkable positive inflow in 2009–2010 is consistent with the growth in the emerging economy in particular for the growth of BRIC countries. The time frame from 2003 to 2008 was a period of consistent inflow of FII in the equity market with a minor dip in 2006–2007. There is a further increase by 3.68 per cent of net investment in equity touching $24.23 billion in 2010–2011 from 2009–2010 level. FII stakes in the different sectors of NSE-listed companies according to March 2011 statistics, relating to percentage share of main sectors, indicates that Finance (14.38 per cent), Information Technology (13.04 per cent), Banking (10.86 per cent), FMCG (10.07 per cent), and Media & Entertainment (6.28 per cent) are the most attractive for investment.²

The number of FIIs registered with RBI was 3 in 1993. It started increasing from 2003 onwards. There was a sudden jump to 1,319 in 2007–2008, and it peaked to 1,626 in 2008–2009. Although the Sensex performed better in 2009–2010, the number of new FIIs registered declined to 87. A total number of registered FIIs increased marginally from 1,713 in March 2010 to 1,722 by March 2011. The US, UK, Mauritius and Hong Kong were major countries which are taking the FII route to invest in the Indian capital market. The large FII inflows appreciated Rupee against dollar affecting Indian exports.³

The international capital asset pricing model (CAPM), by Frankel (1982), gives a utility-maximisation model for international asset diversification, showing that the portfolio risk may be reduced by keeping foreign assets having negative correlation of their returns with home country’s assets returns. It was observed that the emerging markets have been growing faster than the advanced economies and are also considered safe and attractive investment destination. The inflows data for the first half of 2010 indicates that the emerging markets are leading the economic recovery process and may remain major destinations for equity investments. Though the investment pattern for first half of 2010 is uneven, India, Japan, Indonesia and the Philippines have shown a year-on-year increase in investments; however, inflows into Brazil and South Africa have been lower. It was seen by Poshakwale and Thapa (2010) that the rapid growth in the flow of foreign equity portfolio investment leads to greater integration of Indian equity markets with global equity markets. The FIIs have both positive and negative impact on the domestic economy, triggering significant influence in broadly three areas: stock market, exchange rate and foreign exchange reserves. It increases savings of low- and middle-income developing countries (Menkhoff, 2003; Mody, Taylor & Kim, 2001) and enhances market depth and breadth (Singh & Paliwal, 2010).

The study is organised as follows. The next section reviews related literature and gives the stable hypothesis, followed by a section that provides data, variables, theoretical framework and methodology. The next section gives data analysis and empirical findings after the structural breaks; the penultimate section provides a summary of major findings; and the last section highlights conclusions.

Literature Review

Chakarbarti (2001) studied the importance of FII flows in India and its relationship with other economic variables from May 1993 to June 2001. It was found that even though the flows are highly correlated with the equity returns, they are more likely the effect than the cause of returns. The FIIs have no informational disadvantage compared to the local investors and Asian crisis changed determinants of FII inflows resulting in domestic equity returns to be the sole drivers of flows. Kohli (2001) investigated the trend of capital inflows and their impact on some key macroeconomic variables. It was observed that inflows appreciate real exchange rate and increase the money supply. Mukherjee, Bose and Coondoo’s (2002) study was an extension of Chakarbarti’s (2001) study. They found that:

FII flows are caused by returns in the domestic equity market and not conversely.

Return on equity is the single-most important factor influencing FII inflows.

FII sales and FII net inflows are significantly affected by Indian equity market performance, but FII purchase is non-responsive to market performance.

FII investors are not using Indian equity market for diversification of their investments.

Returns from exchange rate variation and Indian economy fundamentals seem to have an influence on FII decision but these are weak.

Daily FII flows are highly autocorrelated.

Batra (2003) analysed the trading behaviour of FIIs and the impact of trading biases upon stock market stability. The FIIs were found to be positive feedback investors and trend chasers at aggregate level on a daily basis. But no evidence of positive feedback trading on monthly basis was found. There were no joint dynamics between long-horizon return and net equity purchase. The foreign investors were found to have a tendency to herd on equity market, even though it may not happen the same day. There was an excessive sell-side herding during financial crisis, although, on average, the extent of herding⁴ on the either side of market during crisis was lower than during immediate preceding period. Batra (2004), while studying stock return volatility patterns on monthly data, used asymmetric generalised autoregressive conditional heteroscedasticity (GARCH) model augmented by structural change analysis. This helped in the identification of sudden shift in stock price volatility and nature of events which caused shift in volatility. It was concluded that the period around BoP crisis and subsequent reforms was the most volatile phase. Major policy changes resulted in sudden shift in stock return volatility, which is a consequence of domestic political and economic events rather than global happenings. Bose and Coondoo (2004) examined quantitative impact of FII regulatory policy reforms on its investment flow using intervention analysis technique based on multivariate GARCH regression model. Ten policy interventions during 1999–2004 were examined for their possible significant influence on FII flows and their sensitivity to the stock returns. It was found that liberalisation of policies have the desired expansionary effect in increasing mean level of FII flows; however, some restrictive measures to control FII flows do not have significant negative impact on net inflows.

Badani and Tripathi (2009) employed autoregressive integrated moving average (ARIMA) model to examine relationship between FIIs and the Indian stock market. It was found that the past FIIs have significant impact on the current Sensex and NSE Index, but not much impact of current FIIs on the current indices was observed. A significant finding of the study was that the FIIs in India need well-calibrated policy response, whereas the daily movement of stock market can be better explained by factors other than FIIs. Bhaduri and Samuel (2009) employed logistic smooth transition method to estimate correlation and pace of integration in international equity market. It was found that the pace of integration between Indian and world market was insignificant. Sehgal and Tripathi (2009), in their study on investment strategies of FIIs in the Indian equity market, examined whether FIIs adopt positive feedback⁵ and herding strategies. They found that FIIs exhibit return chasing behaviour while using monthly data, and are using this strategy for daily data as they do not react instantaneously but wait for market information to crystallise. Further, FIIs display a strong herding behaviour, which is much stronger at the aggregate level than at individual stock level. This may be because FIIs are more cognisant of corporate fundamentals at the individual stock level. Mishra, Das and Pradhan (2010), in their study focusing on foreign investments and real economic growth in India, using vector autoregression (VAR) framework, observed that bidirectional causality runs from net FII flows to real economic growth. Economic growth is determined and influenced by the volume of portfolio investments. Mukherjee and Roy (2011) identified determinants of investment decision of mutual funds and compared it with that of FIIs. It was found that mutual funds influence the decision of FIIs in case of investment in equity and FIIs do opposite of mutual funds. Both track international interest rates. Lakshman, Basu and Vaidyanathan (2013) observed presence of market-wide herding and examined whether institutional investors are responsible for herding. They studied the impact of index return and volatility as well impact of FII inflows and mutual funds on herding.

This study is motivated due to a lack of research using high-frequency daily data, which are divided into sub-periods due to structural breaks. For the first time, VAR models comprising of different endogenous variables are employed to comprehensively understand emerging statistical and economic relationships and causation between them and related policy implications.

Objectives

To study the influence of FIIs on Sensex returns and its role in integration with the US equity market.

To examine macroeconomic determinants influencing relationship between the Sensex and FIIs.

Research Questions

Are there short-term and long-run relationships between the Sensex and FIIs? Does there exist a robust VAR to explain it?

What is the causal relationship of purchase, sales and net FII flows with Sensex?

What is the impact of structural change on the relationship between the Sensex and FIIs and other indices?

Do FIIs help in the integration of Sensex with the US equity market?

What is the relationship between the Sensex, S&P 500 and exchange rate?

Data Selection, Variables and Methodology

The daily data relating to the variables, namely, Sensex, purchase and sales of FIIs, S&P 500 and exchange rate, have been taken from www.bseindia.com, www.sebi.com, Business Beacon and uk.finance.com.

The Bombay Stock Exchange (BSE) was established in 1875, whereas NSE came into existence much later in 1992. Sensex or Bombay Stock Exchange Sensitive Index, comprising of 30 stocks from 11 sectors and launched in 1986, has the first-mover advantage. The global investors started making inroads in the Indian stock market in the early 1990s by benchmarking against it. Sensex is an established brand and its recall has helped the global investors to habitually continue using it to track Indian stocks movements. Nifty, introduced in 1995, is a well-diversified 50 stock index accounting for 23 sectors of the economy. It is broader based representing a larger number of sectors. Both indices are indicators of market movement. Nifty has low brand recall, but is widely traded in the derivative segment. The market capitalisation of Sensex for 2015 is slightly more than Nifty, around $1.62 trillion. Sensex is the index which the world tracks, whereas Nifty is traders favourite; however, technical analyst still follow Sensex for estimating turning points, trends and reversals. Sensex with lesser number of stocks is a relatively easier index to track. Thus, due to these factors, Sensex has been preferred as a proxy for the Indian capital market over Nifty for this study.

Data Series

FIIN = FII net inflow, which is the difference between daily FII purchase (FIIP) and FII sales (FIIS) in the equity market.

FIIP = Daily purchase of FIIs.

FIIS = Daily sales of FIIs.

FIIR is ratio of FIIP and FIIS:

FIIR = $\frac{F I I P}{F I I S}$ ; if FIIR > 1, it indicates inflow of FIIs and FIIR < 1 indicates outflow of FIIs.

EXR = Daily ₹–$ exchange rate.

SENSEX = Daily closing price of BSE 30.

S&P 500 = Daily closing price of US benchmark market index.

The following variables giving daily returns have been derived from the above time series:

GlSensex = log(Sensex/Sensex(–1)), GlSP500 = log(SP500/SP500(–1)), Glexr = log(exr/exr(–1)), Glfiip = log(fiip/fiip(–1)), Glfiis = log(fiis/fiis(–1)) and Glfiir = log(fiir/fiir(–1)).

Methodology

We evaluate the existence of possible relationship between net capital flow, Sensex return, S&P return and change of exchange rate by different using VAR models. This is done by starting with simple VAR models with few variables in the identified sub-periods and bringing in more variables into the system, which helps in identification of a robust VAR model. The descriptive statistics are first discussed, followed by the usage of statistical tests for confirming the presence of structural breaks. Tests for non-stationarity are followed by discussion about the existence of cointegration, Granger causality test, generalised impulse response analysis and variance decomposition analysis.

Test for Stationarity

Non-stationarity of time series is tested to avoid the presence of ‘spurious regression’ (Granger & Newbold, 1974). The classical regression model assumes that both dependent and independent variables in the regression model should be stationary. The presence of unit root or non-stationarity is tested using different tests, including augmented Dickey–Fuller (ADF) test (1979, 1981), Phillips–Perron (PP) test (1988) and Kwiatkowski–Phillips–Schmidt–Shin (KPSS) test (1992). Subsequently, Dickey–Fuller generalised least squares (Dickey-Fuller test statistic using a Generalized Least Squares), Elliot, Rothenberg, and Stock test and Ng–Perron test are also used in order to confirm stationarity of time series.

Vector Autoregressive (VAR) Model

In order to examine the relationship between macroeconomic variables, the VAR model, popularised by Sims (1980), has been used. The VAR model includes all variables in the system and tries to determine its variation due to its past values as well as lagged values of other variables.

An unrestricted VAR assumes that the variables are related to both their own lagged values as well as lagged values of other variables. In this study, estimated VAR are of reduced form since they only use lagged values of variables on the right-hand side, indicating non-existence of simultaneity in the system. Number of variables included in the system depends upon theoretical considerations and decision about lag length is based upon statistical tests. For an unrestricted VAR, it is necessary that the same numbers of lags of all of the variables are used in all equations. The optimal lag length is for minimum value of multivariate information criteria based on Akaike information criterion (AIC), Schwarz criterion (SC) and Hannan–Quinn (HQ) criterion. It is essential that all variables included in the VAR should be stationary so as to conduct joint significance test on the lags of identified variables. The significance of all lags of each variable within the VAR framework is examined jointly by F-tests, which will establish the joint significance of all lags of the individual variables (Brooks & Tsolacos, 1999).

Generalised Impulse Response Analysis

The impulse response function captures and measures the time profile of effect of a one-time shock to one of the innovations to both future and current values of endogenous variables of a dynamic system.

The conventional method of impulse response function (Sims, 1980) is sensitive to the sequence of the variables of the VAR (Lutkepohl, 1991). Inverse of Cholesky factor of the residual covariance matrix was used by Cholesky to orthogonalise these impulses. This imposes an ordering of variables in the VAR and attributes all effects to the variable which is first in the VAR system. The responses change significantly in case of reversing the order, which is its limitation. In order to overcome this problem for linear multivariate model, Pesaran and Shin (1998) developed generalised impulses which are orthogonal set of innovations that are independent of VAR ordering. The impulse responses can be uniquely estimated; also, they do not change due to the ordering of the variables in the VAR and they take into consideration the past patterns of correlation between different shocks.

Variance Decomposition

The impulse response function helps to find the effect of a shock to one endogenous variable imparted on other variables in the VAR, whereas the variance decomposition breaks the endogenous variable’s variations into shocks component to the VAR. We have decomposed the forecast error variance so as to determine the proportion of the movement in the different time series that are consequence of its own shock rather than shocks to other variables.

Data Analysis and Empirical Findings

Descriptive Statistics for Sub-Time Frames

The time frame from 1 January 1999 to 31 December 2010 has been divided into four sub-periods, T₁, T₂, T₃ and T₄, on the basis of global events influencing Sensex and FII flows. These sub-divisions are based on statistical analysis. The non-trading days have been adequately accounted for in calculating daily returns.

Table 2.

Four Sub-periods due to Structural Breaks

	Events	From	To	Sample Points	Number of Observations
1	Post-Asian crisis & dot-com bubble burst,T₁	1 January 1999	31 July 2003	1,147	1,147
2	Period of consolidation & reforms, T₂	1 August 2003	15 January 2008	2,266	1,119
3	Sub-prime crisis, T₃	16 January 2008	9 March 2009	2,544	278
4	Post-sub-prime crisis, T₄	10 March 2009	31 December 2010	2,984	440

Source: Authors’ compilation.

The four mutually exclusive time frames are T₁, T₂, T₃ and T₄ (Table 2).

The Asian crisis triggered after 2 July 1997 when the Bank of Thailand announced managed float of baht which amounted to spreading of recession. However, the process of slow economy recovery started post-Asian currency crisis during 1 January 1999 to 31 July 2003. This period of recovery, on the contrary, also had a dot-com bubble burst, which resulted in the meltdown of technology and IT stocks after April 2001. The Sensex touched highest of 5933.56 on 11 February 2000 and the lowest of 2600.12 on 21 September 2001 (Table 3). Maximum purchase of FIIs was $286.39 million and maximum outflow was $192.98 million.

The descriptive statistics of time series variables indicates that Sensex, NETFII, FIIP and FIIS are positively skewed, whereas S&P 500 and exchange rate are marginally negatively skewed. Sensex, S&P 500 and EXR are platykurtic, whereas FIIP, FIIS and NETFII are leptokurtic. The Jarque-Bera (JB) statistics indicate that all time series are not normally distributed.

The JB test is a goodness-of-fit test for normality. If the data comes from a normal distribution, JB statistic follows asymptotically χ² distribution with 2 degree of freedom. The null hypothesis is a joint hypothesis of the skewness and kurtosis being zero and 3 respectively. For small samples, the χ² approximation is sensitive and it often rejects the null hypothesis when, in fact, it is true. But the number of daily observations for sub-period from 1 January 1999 to 31 July 2003 is 1,147, which is sufficiently large, and as the p-value of JB statistics is 0, therefore it is not normal distribution. The Kolmogrov–Smirnov (KS) test (p-value < 0.0000) and Anderson–Darling test also confirm that the distributions significantly differ from normal.

Since analysis of the daily data are often contradicted with weekly data, the weekly data of 229 opening day on Mondays have also been analysed, but, according to JB test and KS test, the data are still non-normal.

Table 3.

Descriptive Statistics of Variables for T₁ (1 January 1999–31 July 2003)

	SENSEX	S&P 500	EXR	FIIP	FIIS	NETFII
Mean	3801.140	1194.396	46.12230	47.06899	39.74816	7.350283
Median	3560.320	1234.450	46.74000	39.42208	34.86133	5.265621
Maximum	5933.560	1527.460	49.06000	286.3935	192.9812	261.8516
Minimum	2600.120	776.7600	42.39100	0.682835	0.309853	–111.5368
Std Dev.	707.8888	201.3728	2.140593	30.65335	23.90271	25.84790
Skewness	0.811084	–0.317788	–0.377228	2.035447	1.762453	1.430576
Kurtosis	2.756780	1.878516	1.666916	10.80060	8.185358	16.43959
Jarque-Bera	128.5876	79.41459	112.1344	3700.103	1878.829	9023.491
Probability	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000
Observations	1147	1147	1147	1147	1147	1147

Source: Authors’ compilation.

The second time frame from 1 August 2003 to 15 January 2008 focuses on the gradual reform process which helped to come out from the recession caused by the dot-com bubble burst. The recession was followed by the process of consolidation and economic growth due to the gradual reform process in regulatory environment resulting in a revival of private capital flows; though in the latter period, it also started giving indications of global sub-prime crisis, causing sudden outflows of FIIs. The descriptive statistics of the second subsample, from 1 August 2003 to 15 January 2008, indicate that the Sensex touched an all-time high at 20,873 and minimum at 3,741. This period witnessed high volatility of FIIs. The inflow of FIIs touched its peak of $2,323 million due to attractiveness of the Indian economy, but it had a record outflow of $ 2515.63 million. All time series, besides EXR, are positively skewed. Sensex and S&P 500 are platykurtic, whereas all other time series are leptokurtic. The JB statistics indicate non-normality of all time series.

The period 16 January 2008 to 9 March 2009 relates to the sub-prime mortgage crisis and failure of the global banking system. The global securities suffered maximum loss during this period, which consequently affected our economy. The Sensex showed high volatility by dropping from maximum of 19,868 and touching minimum of 8,160, indicating bear phase. The FIIs outflow touched maximum of $2,520 million and net inflow moved in the band of $967 million to $861 million. Sensex and S&P 500 are negatively skewed and remaining time series are positively skewed. Sensex, S&P 500 and EXR are platykurtic, whereas FIIs are leptokurtic. The JB test confirms that all time series are not normally distributed.

The post-sub-prime crisis period, from 10 March 2009 to 31 December 2010, shows robustness of Indian banking regulatory system and resilience of the emerging economy. The Sensex touched the highest peak of 21,029 points. The FII inflows reached maximum of $2,664 million. The maximum outflow in a day was $1,610 million. All variables are positively skewed, except Sensex and S&P 500. The distributions time series are leptokurtic for all but S&P 500, which is almost normally distributed. The JB statistics indicate non-normality of distributions of all time series.

It was observed from equality of mean test, by using Levene’s test, post hoc Tukey honest significant difference (HSD) test and analysis of variance (ANOVA), that for the four subsamples, NETFII and SENSEX returns for T₁, T₂, T₃ and T₄ have distinct distributions. Thus, each sub-period has statistically different distribution of these variables characterised by distinct set of population parameters.

Structural Breaks

The Chow breakpoint test (1960) independently fits equation for each subsample and checks existence of any significant difference in the estimated equations. The presence of significant difference in these estimated equations indicates structural change in the relationship. The Chow breakpoint test checks existence of structural break in all parameters of the equation. However, in case of a linear equation, testing structural break in the subset of parameters will be sufficient.

Why structural break study is necessary? According to Katarina Juselius (2006, p. 26): ‘Since the inferences from the VAR model are valid provided the parameters are constant, it is frequently the case that one has to split the sample period into subsamples representing constant parameter regimes.’ Structural breaks are present due to unpredictable shift in financial time series, which may result in unreliable model and unstable parameter regime. Structural breaks are also indicative of sudden change in the volatility of the stock market, which may lead to financial crisis. Further, using VAR models without considering existing structural breaks will lead to unreliable inferences. The large time period from 1 January 1999 to 31 March 2010 has distinct unpredictable happenings, namely, Asian crisis, gradual reform process after the dot-com bubble burst and sub-prime mortgage crisis leading to global recession, followed by resilience of the Indian economy. Therefore, the analysis has been carried out separately for the four sub-periods created by the structural breaks.

The structural breaks are tested for the complete data, from 1 January 1999 to 31 December 2010, having 2,984 observations for the data series, NETFII, BSER, GEXR, SP500R, or testing for its subset by running regression of NETFII on SENSEXR.

The stability of two parameters is tested by the Chow test. The null hypothesis and alternative hypothesis are:

H₀: The parameters are constant or stable across the four samples.

H₁: The parameters do not remain constant across the four samples.

Thus, from the first version of Chow break test (Table 4), it is seen that H₀ is rejected as p-value is less than α at 1 per cent. Similarly, the p-value for the second and third versions of test, based on χ² test, rejects the null hypothesis, indicating existence of structural breaks at observations 1,147, 2,266 and 2,544, or it is also an evidence of instability of parameters. Thus, the parameters do not remain constant over the whole range. Determination of reliable relationships between parameters is possible by conducting independent analysis for these four mutually exclusive partitions.

Table 4.

Chow Breakpoint Test

Chow Breakpoint Test: 1,147, 2,266, 2,544
Null Hypothesis: No breaks at specified breakpoints
Equation Sample: 2 2,984
Statistics	Value	Prob. of Statistics	p-value
F-statistic	62.22455	Prob. F(6,2975)	0.00000
Log likelihood ratio	352.65870	Prob. Chi-Square(6)	0.00000
Wald Statistic	373.34730	Prob. Chi-Square(6)	0.00000

Source: Authors’ compilation.

The Chow test for confirmation of breakpoints for complete set of variables: net inflow of FIIs, Sensex return, S&P 500 return and growth in exchange rate is conducted next. The regression equation of the variable and its output is:

₁₂₃

Interpretation

As R² < d ➩ it is a non-spurious regression (0.101 < 1.323) (Table 5). SENSEXR and GEXR are significant at 5 per cent but S&P500 is insignificant. Second, Adj R² = 0.10, which means the model can hardly explain 10 per cent of variation in the dependent variable, NETFII, by joint variation of the three independent variables, SENSEXR, SP500R and GEXR.

The p-value of F-statistics = 0.0 < 0.5; it means the model has joint explanatory power of three independent variables. This indicates goodness-of-fit. The Durbin–Watson (DW) statistics = 1.32361 < D_L, as for n = 200, number of explanatory variables k = 3, α = 5 per cent, D_L = 1.643, but for n = 2,298, the value of D_L will be much higher than 1.643. This implies reject H₀: ρ = 0 ➩ existence of autocorrelation.

As value of variance inflation factor (VIF) is less than 10, it indicates absence of multicollinearity, and thus the estimate of βs will be precise.

Further values of coefficients (βs) of parameters are not providing true picture of the state of affairs. This may be due to structural breaks occurring during the long period of observations.

The Chow test in Table 6 confirms instability of parameters over the complete timeline as p-values for the three statistics are 0.0 < α = 1 per cent. Thus, the test strongly indicates existence of structural breaks in the time-series data of FII flows (in similar way for the Sensex data).

Table 5.

Multiple Regression Model

Dependent Variable: NETFII
Method: Least squares
Date: 18 May 2011 Time: 21:40
Sample (adjusted): 2 2,984
Included observations: 2,983 after adjustments
Variable	Coefficient	Std Error	t-Statistic	Prob.	Collinearity Statistics
					Tolerance	VIF
C	30.23948	2.476039	12.21284	0.0000
SENSEXR	1754.847	149.5067	11.73758	0.0000	0.905	1.105
SP500R	–190.1358	187.3934	–1.014634	0.3104	0.958	1.044
GEXR	–75.35857	7.072709	–10.65484	0.0000	0.933	1.071
R-squared	0.101531	Mean dependent var		31.42062
Adjusted R-squared	0.100627	S.D. dependent var		142.4261
S.E. of regression	135.0702	Akaike info criterion		12.65081
Sum squared resid	54348790	Schwarz criterion		12.65885
Log likelihood	–18864.68	Hannan–Quinn criterion		12.65370
F-statistic	112.2140	Durbin–Watson stat		1.323616
Prob (F-statistic)	0.000000

Source: Authors’ compilation.

Table 6.

Chow Breakpoint Test for Multiple Variables

Chow Breakpoint Test: 1,147, 2,266, 2,544
Null Hypothesis: No breaks at specified breakpoints
Varying Regressors: All equation variables
Equation Sample: 2 2,984
F-statistic	32.06337	Prob. F(12,2967)	0.00000
Log likelihood ratio	363.73020	Prob. Chi-Square(12)	0.00000
Wald Statistic	384.76040	Prob. Chi-Square(12)	0.00000

Source: Authors’ compilation.

Test for Stationarity

The non-stationarity at level is tested for LSENSEX, LSP500, LEXR, NETFII, LFIIP, LFIIS and LFIIR from 1 January 1999 to 31 July 2003, using a variety of tests. It was observed that most of the time series are non-stationary, excluding LFIIR, which gave mixed signals.

The stationarity for first difference of variables LSENSEX, LSP500, LEXR, LFIIP, LFIIS and LFIIR, by different unit root tests, confirms that all variables are integrated of order 1.

Four VAR models are developed with different combinations of variable returns. This will help in determining interrelationship between the returns of significant variables, identifying appropriate model and understanding causality relationships for post-Asian crisis period. The analysis of above-mentioned four models suggests to move towards a liberalised FII policy framework to make India an attractive destination for investments.

The following four VAR models comprising of different variables are considered for analysis:

VAR I: GLFIIP and GLSENSEX

VAR II: GLFIIP, GLSENSEX, GLSP500 and GLEXR

VAR III: NETFII and SENSEXR

VAR IV: GLSENSEX, GLFIIS, GLEXR and GLSP500

VAR Model I

VAR Lag Order Selection

The VAR model I comprises two endogenous variables, GLZFIIP and GLSENSEX. The lag order of unrestricted VAR is calculated for different information criteria (Lutkepohl, 1991; Sims, 1980). It is seen that the lag length for 4 gives minimum value for all three information criterion: AIC, SC and HQ. Thus, VAR(4) is suitable for further analysis.

The VAR estimate is given by:

GLFIIP = – 0.713482971793*GLFIIP(–1) – 0.45304718571*GLFIIP(–2) – 0.342581139703*GLFIIP(–3) – 0.292258954201*GLFIIP(–4) + 3.58908503342*GLSENSEX(–1) – 0.467850251669 *GLSENSEX(–2) – 2.18844429875*GLSENSEX(–3) – 0.481869425643*GLSENSEX(–4) + 0.0056547268182

Interpretation

About 37.53 per cent variation in the growth of FII inflows (purchase) is explained by the model, whereas hardly 2.4 per cent growth of Sensex can be explained by growth in FIIs purchase. Thus, in the short term, the growth of Sensex has a higher influence on the growth of inflows of FIIs than otherwise. The F-value indicates that the regression coefficients are significant for FIIP.

Impulse Response

Figure 1 gives the impulse response associated with unit standard deviation shock in each of the two explanatory variables.

Impact on FII inflows

Innovation in FII inflows: Considering the signs of responses, unexpected FII inflows have a mixed impact on itself and the effect of shock dies down after 10 days. On the first day, it is positive; the second day, negative; and it recovers to positive on the third day, dipping on the 4th and 5th days but peaking up on the 6th day.

Innovation in Sensex returns: Return of Sensex has a positive impact on FII inflows on the first day, but from 2nd to 5th day, it has a negative impact on FII inflows. The FII inflows remain positive from 6th to 7th day and it has subsequently no effect on FII inflows.

Figure 1.

Generalised Impulse Response

Impact on Sensex return

Innovation in FII inflows: Increase in FII inflows has a positive impact on Sensex return till five days ahead and has a marginal dip on the 6th day, but has a positive dying impact after the 7th day.

Innovation in Sensex returns: Innovation in Sensex return has a significantly positive impact one day ahead, but it suddenly falls on 2nd day and has marginal positive impact till the 6th day and has no impact on Sensex returns further.

The impulse response dies out to zero in all four cases, therefore the VAR model is stationary.

Variance Decomposition

Variance decomposition of GLFIIP

The first day’s decomposition of growth FIIP (FII inflows) is due to its own innovation. Even after the 10th day, shock to the FIIP accounts for 98.45 per cent variation in growth of FII inflows, or 98.45 per cent variation in growth FII inflows is explained by its own shock, whereas Sensex return shocks account for 1.55 per cent of variance of FII inflows. It implies that FII inflows are dependent on itself rather than Sensex returns.

Variance decomposition of GLSENSEX

About 98.51 per cent growth in Sensex return is explained by itself on the 1st day, whereas 1.48 per cent of growth in Sensex return is explained contemporaneously by inflow of FIIs; 96.83 per cent variation in Sensex return is explained by shock 10 days ahead, whereas 3.16 per cent variation in Sensex return is explained by FII purchase.

Granger Causality

Correlation does not imply causation in a meaningful way; it only provides strength of relationship between the two random variables. The correlation coefficient between the endogenous variables: growth in Sensex return (GLSENSEX) and growth in FII inflows (GLFIIP) is 0.0578.

The short-term dynamics are examined using the Granger causality for lag 4 between GLSENSEX and GLFIIP. As p-values are 0.0005 and 0.0007, which are significant at 1 per cent, we therefore reject both the null hypotheses: (a) GLSENSEX does not Granger-cause GLFIIP; and (b) GLFIIP does not Granger-cause GLSENSEX, indicating that there exists a bidirectional causality running between growth of Sensex return and growth of FII inflows.

However, it is seen that Granger causality is sensitive to the selection of order of lag. GLSENSEX Granger-causes GLFIIP but GLFIIP does not Granger-cause GLSENSEX. Thus, two-way causation does not exist between two variables for 2-lag length.

VAR Model II

Let us include more endogenous variables in the model and find its appropriateness in terms of goodness-of-fit and improvement over the previous model. The lag order selection criteria for VAR for endogenous variables, GLFIIP, GLSENSEX, GLSP500 and GLEXR, confirm minimum value for log 1 by SC.

The VAR estimate is given by Table 7.

Interpretation

Adj R² for regression of FIIP is 23 per cent, which implies 23 per cent variation in growth of inflow of FIIs is explained by the model and remaining 77 per cent remains unexplained. Further, as F_k-1,n-k = 89.999, k = 4, n – k = 1,145 – 4 = 1,141, H₀:β₁ = β₂ = β₃ = β₄ = 0, α = 5 per cent, F_3,1141 = 2.08. F-statistics for overall significance of ordinary least squares (OLS) for GLFIIP rejects the null hypothesis, as F_cal > F_tab, implying β’_is are significant. Thus, the model is correctly specified in jointly explaining variation in growth of FII inflows on the basis of explanatory variables. To test H₀:β₁ = β₂ = β₃ = β₄ = 0 by F-test, which is measure of overall significance of parameters, is similar to test significance of $R^{2} = 0 [a s F = \frac{\frac{R^{2}}{k - 1}}{\frac{(1 - R^{2})}{n - k}}]$ . Thus, OLS for GLFIIP gives good fit. Similarly, F_cal for GLSENSEX and GLEXR is 6.7206 and 5.2406 respectively, thus regression coefficients for GLSENSEX and GLEXR are significant; however, the null hypothesis for GLSP500 is accepted, implying no relationship of growth in S&P 500 with the growth of explanatory variables of FII inflows, Sensex and exchange rate. R² for VAR II is less than VAR I, but it was observed that ‘R² is an inappropriate measure when variables are trending’ (Juselius, 2006, p. 59). The F-statistics, which tests H₀ that all of the slope variables are jointly zero, indicates that the regression coefficients are jointly significant, confirming that there is significant relationship between GLFIIP and other explanatory variables.

Granger Causality

Chakarbarti (2001) also examined causality between net FII and Sensex for monthly returns; however, Granger causality focuses to detect statistically significant short-term lead–lag relationship in the pair of datasets of two variables. As Sensex responds spontaneously, examination of monthly data will fail to capture inherent exact causality (Mukherjee et al., 2002), which is a limitation of Chakarbarti’s study. The Granger causality relations are provided in Table 8.

Table 7.

VAR II Estimates

Vector Autoregression Estimates
Sample (adjusted): 3 1,147
Included observations: 1,145 after adjustments
Standard errors in ( ) & t-statistics in [ ]
	GLFIIP	GLSENSEX	GLSP500	GLEXR
GLFIIP(–1)	–0.488994	5.97E-05	0.000273	–3.45E-05
	(0.02590)	(0.00073)	(0.00060)	(5.3E-05)
	[–18.8810]	[0.08235]	[0.45716]	[–0.65682]
GLSENSEX(–1)	2.572146	0.062471	0.005448	–0.008706
	(1.05934)	(0.02965)	(0.02440)	(0.00215)
	[2.42806]	[2.10660]	[0.22323]	[–4.05113]
GLSP500(–1)	1.878052	0.165163	0.006088	–0.002661
	(1.28712)	(0.03603)	(0.02965)	(0.00261)
	[1.45912]	[4.58390]	[0.20533]	[–1.01912]
GLEXR(–1)	15.10960	–0.123246	0.039660	–0.062585
	(14.6258)	(0.40943)	(0.33693)	(0.02967)
	[1.03307]	[–0.30102]	[0.11771]	[–2.10944]
C	0.004952	0.000199	–0.000192	7.73E-05
	(0.01743)	(0.00049)	(0.00040)	(3.5E-05)
	[0.28404]	[0.40700]	[–0.47721]	[2.18559]
R-squared	0.239999	0.023038	0.000289	0.018056
Adj. R-squared	0.237332	0.019610	–0.003219	0.014611
Sum sq. resids	395.0995	0.309619	0.209675	0.001626
S.E. equation	0.588709	0.016480	0.013562	0.001194
F-statistic	89.99941	6.720620	0.082285	5.240646
Log likelihood	–1015.532	3078.731	3301.882	6083.972
Akaike AIC	1.782588	–5.368963	–5.758745	–10.61829
Schwarz SC	1.804611	–5.346941	–5.736723	–10.59627
Mean dependent	0.004338	0.000170	–0.000188	7.16E-05
S.D. dependent	0.674114	0.016644	0.013540	0.001203
Determinant resid covariance (dof adj.)	2.41E-14
Determinant resid covariance	2.36E-14
Log likelihood	11464.08
Akaike information criterion	–19.98967
Schwarz criterion	–19.90158

Source: Authors’ compilation.

Table 8.

Pair-wise Granger Causality

Lags	Pair-wise Granger Causality
	p-value
	1	2	3	4	5	6
Null Hypothesis
GLSENSEX does not Granger-cause GLFIIP	0.0181*	0.0077*	0.0004*	0.0005*	0.0011*	0.0041*
GLFIIP does not Granger-cause GLSENSEX	0.8699	0.2579	0.2652	0.0007*	0.0037*	0.0126*
GLSP500 does not Granger-cause GLFIIP	0.1219	0.1640	0.1575	0.3650	0.3281	0.2646
GLFIIP does not Granger-cause GLSP500	0.6374	0.8887	0.9819	0.9886	0.9926	0.9946
GLEXR does not Granger-cause GLFIIP	0.5121	0.7578	0.3151	0.5729	0.0149*	0.0336**
GLFIIP does not Granger-cause GLEXR	0.3621	0.6138	0.4506	0.0274	0.0430**	0.0713
GLSP500 does not Granger-cause GLSENSEX	5.E-06	1.E-05	1.E-06	4.E-06	9.E-06	3.E-05
GLSENSEX does not Granger-cause GLSP500	0.8144	0.8327	0.9700	0.9948	0.9592	0.9865
GLEXR does not Granger-cause GLSENSEX	0.7301	0.3609	0.1952	0.1926	0.1856	0.3196
GLSENSEX does not Granger-cause GLEXR	4.E-05*	0.0002*	0.0006*	0.0024*	0.0018*	0.0016*
GLEXR does not Granger-cause GLSP500	0.9303	0.2004	0.3786	0.5358	0.5970	0.7155
GLSP500 does not Granger-cause GLEXR	0.2298	0.3349	0.5337	0.6389	0.7318	0.7471

Source: Authors’ compilation.

Note:* indicates significance at 1% and ** indicates significance at 5%.

Interpretation

GLSENSEX Granger-causes GLFIIP ➔ past value for growth of Sensex returns causes growth in FII inflows but the causation does not run the other way at 1 per cent level of significance. This confirms that FII flows in India, post-Asian crisis, are mostly due to contemporaneous return of Sensex, which is similar to research done by Mukherjee et al. (2002) and Panda (2005) for a limited dataset. Badani & Tripathi (2009) study also confirms non-existence of other way causation that FII inflows cause return of SENSEX.

Another significant feature is the non-existence of causation between growth of the US market returns and growth of inflows of FII. Although S&P 500 has significant influence on Sensex returns, other-way causation does not exist, which reflects the true state of affairs.

A fall in the exchange rate of rupee against dollar represents depreciation of rupee (more Rs for a $), whereas rise in exchange rate represents an appreciation of rupee. By Table 8, it is evident that no definite conclusions can be drawn about Granger causality between exchange rate and inflow of FIIs for T1, which conforms to the study of Bhattacharya and Mukherjee (2002), who suggested that there is no causal linkage between FII inflows and exchange rate. However, Kohli (2001) found that inflows of FIIs appreciate real exchange rate. Badani (2005) found, for the monthly data from April 1993 to March 2004, existence of long-term relationship between FII inflows and exchange rate. However, the period of the study prior to 1999 has a somewhat regulated policy framework for FIIs and usage of monthly data cannot provide a true picture. From the Table 8, it is evident that the change in the exchange rate does not Granger-cause Sensex return; however, Sensex return Granger-causes change in exchange rate (which contradicts Bandai’s [2005] study where it was seen that short-term causality runs from change in exchange rate to stock returns and not vice versa for monthly data). Bhattacharya and Mukherjee (2002) also found unidirectional causality from change in exchange rate to stock returns at 10 per cent for monthly data, implying that the exchange rate movements lead the BSE Sensitive Index. Thus, the absence of causal relationship between exchange rate and inflow of FIIs indicates that the causality between exchange rate and Sensex return is not due to FII purchases but due to other factors. Thus, stock market volatility may be stabilised by focusing on domestic economic policies. Furthermore, stock market returns can neither capture changes in inflows of FIIs nor change in exchange rate. Therefore, a suitable profit-making tactical strategy may be formulated on the basis of this information. The non-existence of any causation between S&P 500 and change in exchange rate is also seen.

Variance decomposition analysis almost confirms the given analysis: on the first day, 100 per cent variation in FII inflows is explained by variation in itself, and hardly 0.47 per cent is explained by Sensex returns 10 days ahead; and change in exchange rate and the US equity market has no influence on FII inflows. About 99.60 per cent variation in Sensex return is explained by itself on the first day, whereas 0.37 per cent variation in Sensex return is explained by FII inflows 10 days ahead; and 1.8 per cent variation in Sensex return is explained by S&P 500 returns 10 days ahead. Variance decomposition of change in exchange rate confirms unidirectional causality from Sensex return to change in exchange rate. A 1.94 per cent variation in change in exchange rate is due to Sensex return on the first day and it increases to 3.15 per cent two days ahead and maintains the same even 10 days ahead.

VAR III

It is seen that NETFII and SENSEXR are both stationary, I(0), for the time period T₁. VAR lag order selection has minimum value 1 for SC, though it is 3 for HQ and 5 for AIC. R² value in the OLS of VAR(1) of NETFII and SENSEXR is very small, signifying that the model is not a good fit. More variables need to be included. But the model is able to jointly explain variations as by F-statistics, β coefficients are significant. There exists unidirectional causality in Granger sense from Sensex returns to net FIIs at α = 1 per cent, when lag is 1, and bidirectional at α = 5 per cent. Bidirectional causality exists in case lag is 3 (for HQ) at α = 5 per cent. However, for lag 5 (for AIC), bidirectional causality exits at α = 1 per cent (Table 9).

Table 9.

Pair-wise Granger Causality

Lags	Pair-wise Granger Causality
	p-value
	1	2	3	4	5	6
Null Hypothesis
SENSEXR does not Granger-cause NETFII	7.E-08	6.E-07	2.E-09	9.E-09	1.E-09	2.E-09
NETFII does not Granger-cause SENSEXR	0.0332	0.0532	0.0322	0.0610	0.0105	0.0433

Source: Authors' compilation.

The Granger causality test has been used to study market information efficiency. The information efficiency exists in case unidirectional lagged causal relationship from an economic variable to Sensex return, or bidirectional between Sensex return and economic variables, cannot be found. It implies that market is efficient as economic variables cannot influence or be influenced by Sensex volatility. Sensex and variables movements are statistically independent of each other. The analysis indicates that bidirectional causality exists between Sensex return and net FIIs at 5 per cent level of significance. This means that market information efficiency hypothesis may be rejected for Sensex return and net FIIs. But the result is consistent with the base-broadening hypothesis which assumes positive and long-term impact of FIIs on stock price due to reduction of risk premium on account of international diversification. There is expansion of investor base to include foreign investors, which results in increased diversification, and this is followed by reduced risk which lowers the required risk premium (Clark & Berko, 1997; Merton, 1987; Warther & Shumway, 1999).

VAR IV

The VAR for stationary endogenous variables, GLSENSEX, GLFIIS, GLEXR and GLSP500, has minimum value of lag 1 by SC criterion and lag 4 by AIC and HQ criterion. The VAR for lag 4 is considered. The estimated VAR(4) has low R² value signifying poor fit, but as the F_k-1,n-k = F_16-1,1147-16 = F_15,1131 = 1.67, F-statistics is significant for OLS in the VAR of GLSENSEX (3.224094), GLFIIS (38.69679) and GLEXR (2.267202), but insignificant for SP500 (0.773234), where values within the parentheses are F-calculated. Further, simple linear regression between daily FII sales and BSE return is found to be negative and significant on the same day, whereas FII purchase has positive and significant relationship with BSE return at α = 10 per cent.

It is seen that pair-wise Granger causality exists from GLSENSEX to GLFIIS, GLSENSEX to GLEXR and GLEXR to GLFIIS. This mean that decline in returns of Sensex results in the sale of FIIs; Sensex return brings change in the exchange rates; and change in the exchange rates affects outflow of FIIs.

The variance decomposition of Sensex return shows that 100 per cent variation in it is due to itself, but returns of S&P 500 affects one day ahead and 2.8 per cent variation in S&P 500 return affects Sensex return 10 days ahead. Other variables have no influence on Sensex.

The variance decomposition of sales of FII indicates that 99.81 per cent variation is due to itself a day ahead. More that 3 per cent variation in outflow in FIIs is due to change in the exchange rate and return of both Sensex and S&P 500 10 days ahead.

The variance decomposition of change in exchange rate confirms that 97.70 per cent variation is due to itself and remaining 2.05 per cent is due to Sensex return one day ahead. Variation in Sensex return has persistent influence on change in exchange rate.

A similar analysis can be conducted for the other three sub-periods.

Summary of Major Findings

It is evident from VAR I between Sensex return and FII inflows that in the short term, Sensex return is responsible for the growth of inflows of FIIs in the post-Asian crisis period. This is also confirmed by the innovation of Sensex and FII inflows. The variance decomposition indicates that FII inflows are dependent on itself rather than on Sensex returns. Significant variation in Sensex return is explained by itself, whereas small variation in Sensex return is explained by FII purchase. Growth of Sensex return and growth of FIIs inflows have bidirectional causality.

VAR II includes more endogenous variables, namely, GLFIIP, GLSENSEX, GLSP500 and GLEXR. Past value of growth of Sensex return causes growth of FII inflows, but the causation does not run the other way. The FII flows in India during post-Asian crisis are mostly due to contemporaneous return of Sensex. The other-way causation—FII inflows being cause of return of Sensex—does not exist, which was also confirmed by Badani and Tripathi (2009). Non-existence of causation between growth of the US market returns and growth of inflows of FII is seen. However, S&P 500 has significant influence on Sensex returns, but other-way causation does not exist. No definite conclusions can be drawn about Granger causality between exchange rate and inflow of FIIs for T₁, which conforms to the study of Bhattacharya and Mukherjee (2002). It was observed that the change in the exchange rate does not Granger-cause Sensex return; however, Sensex return Granger-causes change in exchange rate. The absence of causal relationship between exchange rate and inflows of FII indicates that the causality between exchange rate and Sensex return is not due to FII purchase but due to other underlying factors. Hence, by focusing on domestic economic policies volatility in the stock market may be contained.Variance decomposition confirms that inflows of FII are explained mostly by itself and Sensex returns also contribute marginally. The US equity market has no influence on FII inflows. Variance decomposition of change in exchange rate confirms unidirectional causality from Sensex return to change in the exchange rate.

VAR III comprises of two variables: net inflows (NETFII) and Sensex return (SENSEXR). Unidirectional causality at α = 1 per cent when lag is 1 exists from SENSEX returns to net FII’s whereas at α = 5 per cent bi-directional Granger causality exists. The bidirectional causality means market information efficiency hypothesis may be rejected for Sensex return and net FIIs. This leads to strong evidence consistent with the previous research work conducted confirming with base-broadening hypothesis.

VAR IV includes variables GLSENSEX, GLFIIS, GLEXR and GLSP500 and it examines the role of FII sales or outflows. The FII sales/purchase has corresponding negative/positive relation with Sensex return. It was observed that the decline in the return of Sensex results in the sales of FIIs; Sensex return brings about change in the exchange rates; and change in the exchange rates affects outflow of FIIs. The variance decomposition of Sensex return shows that 100 per cent variation in it is due to itself, but returns of S&P 500 affects one day ahead. About 99.81 per cent variation in FII sales is due to itself a day ahead. Change in the exchange rate and returns of both Sensex & S&P 500 causes more than 3% variation in the outflow of FIIs. It is found that change in exchange rate is continuously influenced by the variation in Sensex return.

Conclusions

The large time frame was sub-divided in sub-periods due to existence of structural breaks confirmed by the usage of different statistical tests as inferences drawn from VAR model are valid for the constant parameter regime. Different VAR models for derived sub-periods provided interesting results indicating statistically significant relationships. It was observed that FII inflows and outflows are significantly influenced by the returns in the domestic equity market. Net inflows dependence on equity market returns indicates daily return-chasing behaviour in the short term by the foreign institutional investors. The existence of bidirectional causality for Sensex return and net FIIs means that market information efficiency hypothesis may be rejected; at the same time, it confirms base-broadening hypothesis. By the usage of generalised impulse function, variance decomposition and Granger causality test, it was found that the change in the exchange rate has no effect on the inflows of FIIs; however, outflows are influenced by the change in the exchange rate. Sensex returns bring about change in the exchange rates and change in the exchange rates affects outflow of FIIs. The US equity market has no influence on FIIs inflows but it has marginal influencing role in its outflows. The policy implication of these findings motivates to move towards more liberalised regime so as to regain investor’s confidence in Indian equity market, thereby ensuring greater value to all stakeholders.

Footnotes

Notes

References

Badani

K. N.

(2005). Dynamic relationship among stock prices, exchange and net FII flow in India. Retrieved November 10, 2010, from http://www.iiml.ac.in/confrence/abstract/5.pdf

Badani

K. N.

, & Tripathy

(2009). Behavior of Indian stock market: Evidence and explanation. Journal of International Finance and Economics, 9, 124–130.

Batra

(2003). The dynamics of foreign portfolio inflows and equity returns in India. Working Paper No. 109, ICRIER, New Delhi, India.

Batra

(2004). Stock return volatility patterns in India. Working Paper No. 124, ICRIER, New Delhi, India.

Bekaert

, Harvey

, & Lumsdaine

(2002). The dynamics of emerging market equity flows. Journal of International Money and Finance, 21(3), 295–350.

Bhaduri

S. N.

, & Samuel

A. A.

(2009). International equity market integration: The Indian conundrum. Journal of Emerging Finance, 8(1), 45–66.

Bhattacharya

Basabi

, & Mukherjee

Jaydeep

. (2002). Causal relationship between stock market and exchange rate, foreign exchange reserves and value of trade balance: A case study for India. Retrieved December 25, 2015, from http://www.igidr.ac.in/money/mfc_5/basabi.pdf

Bose

, & Coondoo

(2004, July–December). The impact of FII regulations in India: A time series intervention analysis of equity flows. Money & Finance, ICRA Bulletin, pp. 54–83. Retrieved December 25, 2015, from http://www.icra.in/Files/MoneyFinance/2004JulDeckoondoo.pdf

Brooks

, & Tsolacos

(1999). The impact of economic and financial factors on UK property performance. Journal of Property Research, 16(2), 139–152.

10.

Campbell

, & Perron

(1991). Pitfalls and opportunities: What macroeconomist should know about unit roots. In Blanchard

O. J.

and Fisher

(eds), NBRE Macroeconomic Annual, University of Chicago Press, Illinois, pp. 141–220.

11.

Chakarbarti

(2001). FII flows to India: Nature and causes, Money & Finance, 2(7), 61–81.

12.

Chow

G. C.

(1960). Tests of equality between subset of coefficient in two linear regression models. Econometrica, 28, 459–460.

13.

Clark

, & Berko

(1997). Foreign investment fluctuations and emerging market stock returns: The case of Mexico. Federal Reserve Bank of New York Working Paper, No. 24.

14.

Dickey

D. A.

, & Fuller

W. A.

(1979). Distribution of the estimators for autoregressive time series with unit root. Journal of the American Statistical Association, 74(366), 427–473.

15.

Dickey

D. A.

, & Fuller

W. A.

(1981). Likelihood Ratio Statistics for Autoregressive Time Series with a Unit Root. Econometrica, 49(4), 1057–1072.

16.

Frankel

(1982). A search of exchange risk premium: A six currency test assuming mean variance optimization. Journal of International Money and Finance, 1, 255–274.

17.

Granger

C. W. J.

(1969). Investigating causal relations by econometric models and cross-spectral methods. Econometrica, 37, 424–438.

18.

Granger

C. W. J.

, & Newbold

(1974). Spurious regression in econometrics. Journal of Econometrics, 2(2), 11–120.

19.

Juselius

(2006). The cointegrated VAR model: Methodology and applications. Oxford University Press. ISBN: 9780199285679.

20.

Kohli

(2001). Capital flows and their macroeconomic effects in India. Working Paper No. 64, ICRIER, New Delhi, India.

21.

Kwiatkowski

Denis

, Phillips

Peter

, Schmidt

Peter

, & Shin

(1992). Testing the null hypothesis of stationarity against the alternative of a unit root. Journal of Econometrics, 54, 159–178. Retrieved from http://www.deu.edu.tr/userweb/onder.hanedar/dosyalar/kpss.pdf

22.

Lakshman

M. V.

, Basu

, & Vaidyanathan

(2013). Market-wide herding and the impact of institutional investors in the Indian capital market. Journal of Emerging Finance, 12(2), 197–237.

23.

Lutkepohl

Helmut

. (1991). Introduction to multiple time series analysis. New York: Springer-Verlag.

24.

Menkhoff

, & Suwanaporn

(2003). The Rationale of Bank Lending in Pre-Crisis Thailand, (Center for Development Research ZEF, Bonn). http://www.eea-esem.com/papers/eea-esem/2003/1279/REL-040702.pdf

25.

Merton

R. C.

(1987). A simple model of capital market equilibrium with incomplete information. Journal of Finance, 42(3), 483–510.

26.

Mishra

P. K.

, Das

K. B.

, & Pradhan

B. B.

(2010). Foreign investments and real economic growth in India: A causality test. International Research Journal of Finance and Economics, 41, 212–219.

27.

Mody

, Taylor

Mark P.

, & Kim

Jung Yeon

. (2001). Modeling fundamentals of forecasting capital flows of emerging markets. International Journal of Finance and Economics, 6(3), 201–216.

28.

Mukherjee

, Bose

, & Coondoo

(2002). Foreign institutional investment in the Indian equity market: An analysis of daily flows during January 1999–May 2002. Money & Finance, April–September, pp. 21–51. Retreived December 25, 2015, from http://icra.in/Files/MoneyFinance/aprsep2002equity.pdf

29.

Mukherjee

, & Roy

(2011). The nature and determinants of investments by institutional investors in the Indian stock market. Journal of Emerging Finance, 10(3), 253–283.

30.

Panda

(2005). ‘An Empirical Analysis of the Impact of FIIs’ on Investments in the Indian Stock Market. ICFAI Journal of Applied Finance, 11(1), 53–61.

31.

Pesaran

, & Shin

(1998). Impulse response analysis in linear multivariate models. Economics Letters, 58(1), 17–29.

32.

Phillips

, & Perron

(1988). Testing for a unit root in times series regression. Biometrica, 75, 335–346.

33.

Poshakwale

, & Thapa

Chandra

. (2010). Foreign investors and global integration of emerging Indian equity market. Journal of Emerging Market Finance, 9(1), 1–24.

34.

RBI Annual report (2010). Comments of Dr. Duvvuri Subbarao, Governor, Reserve Bank of India at the High-level Conference on ‘The International Monetary System’. Swiss National Bank and IMF, Zurich, 11 May 2010. Retrieved from https://www.rbi.org.in/scripts/BS_SpeechesView.aspx?id=504

35.

Sehgal

, & Tripathi

(2009). An examination of home (bias) argument in the Indian financial markets: Domestic financial institutional investors (DFIIs) vis-a-vis foreign institutional investors (FIIs). Asian Journal of Finance & Accounting, 1(2), 163–174

36.

Sims

Chris

. (1980). Macroeconomics and reality. Econometrica, 48(1), January, 1–48.

37.

Singh

, & Paliwal

(2010). Liberalization of foreign institutional investment (FIIs) in India: Magnitude, impact assessment, policy initiatives and issues. Global Journal of International Business Research, 3(3), 22–41.

38.

Warther

V. A.

, & Shumway

(1999). The Delisting Bias in CRSP’s Nasdaq Data and Its Implications for the Size Effect. The Journal of Finance, 54(6), 2361–2379.