ABSTRACT
This study investigates the relationship between FII investments and the Indian stock
market performance during November 2003 and January 2007 by using forecasting
ARIMA model. The study shows that past FII investment have significant impacts on
current BSE Sensex & NSE Index; but there is no significant impact of current FII
investment on current BSE Sensex & NSE Index. An important implication of these
findings is that the FII investments in India deserve a well- calibrated policy response
while the daily movement of stock market in India should be better explained by other
factors than FIIs.
1. INTRODUCTION
The emerging economies have been witnessing an unprecedented surge in capital inflows. The last three
years in India have witnessed virtual bull-run in terms of rising inflow of investments from the FIIs. The
numbers of registered FIIs have also shown an increasing trend..This huge increase in the number of FIIs
shows their continuing interest in investing in Indian stock market. The financial theory is that FII
investments broadens the base of portfolio diversification and cause a long-term increase in the stock
prices by reducing the equilibrium rate of return. It was however found that the FII behaviour played a
significant role during East Asian Crisis. So understanding the relationship between FII investment and
Indian Stock Market is very important as it may have important policy implications. Warther(1995) in his
‘price pressure hypothesis’ suggests that the increase in share prices associated with foreign investment
flow is caused by temporary liquidity (i.e. excess demand) and predicts that this change in share price is
subsequently reversed. Merton (1997) shows that the expected return in the market with unrestricted
investor base is higher than restricted investor base.
Entry of foreign investors in the stock market broadens the investors’ base, which increases diversification
and risk sharing, lowering the risk premium for country specific volatility. Agarwal (1997) Chakrabarti
(2001) and Trivedi and Nair (2003) found that the equity return has a significant and positive impact on
the FII. But, given the huge volume of investments, foreign investors could play a role of market makers
and book their profits i.e. they can buy their financial assets when the price are declining, thereby jacking-
up the assets price and sell when the assets price are increasing .Rai and Bhanumurty (2003) studied the
determinants of foreign institutional investment in India during the period 1994-2002, by using monthly
data and found that the equity returns is the main driving force for FII investment and is significant at all
levels. Gordon and Gupta (2003) also examined causation running from FII inflows to return in BSE and
conclude that FIIs act as market makers and book profits by investing when prices are low and selling
when they are high. Keeping in view, the present study attempts to examine the relationship between FII
investment and the Indian stock market performance. It aims to investigate whether average monthly BSE
Sensex and NSE Index are dependent on the current and past FII net inflows. It also seeks to assess
whether FII net inflows are dependent on the current and past market BSE Sensex & NSE Index.
For the purposes of this study, monthly net FII investment data and monthly average BSE Sensex and
NSE Index have been used. The FII investment monthly data have been collected from November 2003
Methodology: The multiple regression analysis has been used in the study. Before any regression
analysis can be applied to time series data, it is essential to find that these data are random or the error
terms are free from auto correlation. The most popular test to ascertain the presence or absence of auto
correlated error terms is the Durbin Watson d-statistics. For the formal test of significance, if there is no
serial correlation, the DW statistics will be around 2. The DW statistics will fall below 2 if there is positive
serial correlation. If there is negative correlation the statistics will lie somewhat in between 2 and 4. The
study uses the d-statistics.
ARIMA Model: The Acronym ARIMA stands for “Auto Regressive – Integrated Moving Average” which is
a class of linear models capable of representing stationary as well as non-stationary time series. It does
not involve independent variables in its constructions. Rather it makes use of the information in the series
itself to generate forecasts. It relies heavily on auto correlation patterns of the data. In order to determine
whether a series follow a purely auto regressive process or purely a moving average process, the Box-
Jenkins methodology comes into picture .ARIMA (Auto Regressive Integrated Moving Average) models
are generalizations of the simple AR model that use three tools for modeling the serial correlation in the
disturbance. The first tool is the auto regressive or AR term. An autoregressive model of order P, AR (P)
has the form:
U t P1U t 1 P 2 U t 2 P p U t p t
Where Ut is the time series and t is an uncorrelated random error term with zero mean and constant
variance. A moving average forecasting model uses lagged values of the forecast error to improve the
current forecast. An MA (q) has the form:
U t a 1 t 1 2 t 2 q
where a is constant The auto regressive and moving average specifications can be combined to form an
ARMA (p, q) specification.
U t a P1U t 1 P 2U t 2 P p U t p t 1 6 1 2 t 2 q t q
ARMA model use combinations of past values and past errors and offer a potential for fitting models that
could not be adequately fitted by using an AR or MA model separately when the time series have to be
differenced to make it stationary, the model is called ARIMA instead of ARMA. Q statistics is often issued,
as a test of weather the series is white noise.
The regression Model attempts to find whether currents months’ Net FII investment is dependent on
current and past BSE Sensex and on current and past NSE Index. The Table 1 shows that the one month
lag value, two- month lag and three- month lag values of BSE Sense& NSE index, and one month, two -
month lag value of net FII do not significantly affects the current months investment pattern of FIIs. The `t’
value of these variables is not significant. The regression model also examines whether current month’s
BSE sensex & NSE index is caused by present and past FII investments. It is clear from the Table-2 that
past FII investment have quite impact on current BSE sensex & NSE index. It is also found from the
analysis that one-month lag and two-month lag values of BSE&NSE sensex significantly affect current
BSE sensex at 1% and 5% level of significance. Now it is essential to find out that whether these results
are in conformity to the ARIMA findings. It is evident from all the tables that the Durbin Watson `d’
statistics for all the models are around 2. Hence it can be conclude that the error terms in each of the
models are not serially correlated. The correlogram test is done with Autocorrelation Function and Partial
Autocorrelation Function. Analysis of ACF and PACFs is important because correlograms for stationary
The auto correlation of the series of BSE sensex, NSE index and FII investment are presented in the
Table 3 through Table 5. The Table 5 reveals that the FII series is stationary and requires no differencing.
Table 3 and Table 4 indicate that there is very high auto correlation between current and past BSE
sensex as well as NSE market index. These findings are also in conformity with the regression results of
significant (at 1% level) impact of past BSE Sensex and NSE Index on present BSE Sensex and NSE
Index respectively. Table 4 and Table 5 shows that BSE sensex and NSE index series is non-stationary
and differencing and log transformation is needed to make the series stationary. After differencing once,
the ACF and PACF for BSE sensex and NSE index show that the series have become stationary which is
depicted in the Table 6, Table 7. The Table 8 and Table 9 showed the presence of mixed ARMA process.
Similarly the Table 10 and Table 11 after log transformation indicated mixed ARMA process. The
reliability of the ACFs and PACFs can also be checked through the findings of various ARIMA model.
The Table 11 shows the summary results of ARIMA model. The AIC in Table 8 indicates that BSE sensex
follows the ARIMA (1, 0, 0) when regressed on Net FII investment since the AIC value of –94.422 are
closest to zero. The Table 11 reveals that NSE index follow the ARIMA (1, 0, 0) when regressed on FII
investments, since the AIC value of –94.439 is closest to zero. But ARIMA output shows a warning that
the order of the process may not be correctly estimated. This indicates that current BSE sensex depends
on the just preceding months’ BSE sensex as well as on FII investment, which is confirmed with our
multiple regression analysis. The Table 10 reveals that the ARIMA (1,0, 0) on the AIC value is 59.914
which is minimum and Table 11 depicts ARIMA (1, 0, 0) in the AIC value’s 59.419 is minimum. This result
is also confirmed by regression results. Table-11 indicate that FII also follows autoregressive process of
order 1 and moving average process of order 0 and 1 respectively when regressed on BSE and NSE
index as given by the minimization of AIC values. The immediate next minimum AIC is for ARIMA (2, 0,
0). These results also reinforce the findings of regression models 1and 2.
4. CONCLUSION
The present study examined the impact of Net FII investment on the Indian stock market represented by
BSE sensex and NSE index. The study shows that past FII investment makes significant impact on
current BSE Sensex & NSE Index; but there is no significant impact of current FII investment on current
BSE Sensex & NSE Index. An important implication of these findings is that the FII investments in India
deserve a well- calibrated policy response while the daily movement of stock market in India should be
better explained by other factors than FIIs.
Table 1: Regression Analysis of Net FII Investment on Current and Past BSE Sensex, NSE Index
and Past Net FII Investments
Predictor BSEt BSEt- BSEt- BSEt- FIIt-1 FIIt-2 NSE t NSE t- NSE t- NSE FI It- FI It-
Variable 1 2 3 1 2 t-3 1 2
Unstand -.318 4.366 - .26 .07 - 1656.873 1.917 11.777 - .099 - 1656.873
ardised 3.857 .115 14.715 2.346 .091
B
Unstan 3.636 5.204 3.942 2.525 .283 .221 2493.504 10.100 15.284 13.919 8.060 .269 .228 2539.340
dardised
S.E.B
Stand -.147 2.100 - .144 .071 - .260 1.649 -2.158 .366 .100 -
ardised 1.995 .117 .093
T .664 -.087 .839 -.979 .103 .249 -.521 .190 .771 - .291 .368 - .688
1.057 .399
Multiple .412 .429
R
2
R .170 .184
Adj. R2 1.970 .002
DW .922 1.9
41
F 1.013
Table 6: Auto Correlation: BSE Sensex Transformations: Log transformation and differencing (1)
Lag Auto Correlation Standard Error Box-Ljung Prob.
1 0.335 0.160 4.381 0.036
2 0.220 0.158 6.323 0.042
3 0.181 0.155 7.688 0.053
4 - 0.090 0.153 8.035 0.090
5 - 0.289 0.151 11.712 0.039
6 - 0.091 0.148 12.091 0.060
7 - 0.252 0.146 15.084 0.035
8 - 0.133 0.143 15.944 0.043
9 - 0.102 0.140 16.476 0.058
10 - 0.058 0.138 16.651 0.082
11 - 0.033 0.135 16.711 0.117
12 0.155 0.132 18.075 0.113
13 - 0.029 0.130 18.124 0.153
14 - 0.002 0.127 18.124 0.201
15 - 0.011 0.124 18.133 0.256
16 - 0.112 0.121 18.990 0.269
Table7: Auto Correlation: NSE Index Log Transformation and Difference (1)
Lag Auto Correlation Standard Error Box-Ljung Prob.
1 0.300 0.160 3.509 0.061
2 0.229 0.158 5.612 0.060
3 0.139 0.155 6.416 0.093
4 - 0.054 0.153 6.540 0.162
5 - 0.351 0.151 11.982 0.035
6 - 0.112 0.148 12.552 0.051
7 - 0.317 0.146 17.301 0.016
8 - 0.102 0.143 17.811 0.023
9 - 0.118 0.140 18.516 0.030
10 0.052 0.138 18.656 0.045
11 0.029 0.135 18.703 0.067
12 0.105 0.132 19.333 0.081
13 0.011 0.130 19.340 0.113
14 0.054 0.127 19.519 0.146
15 - 0.104 0.124 20.226 0.163
16 - 0.074 0.121 20.597 0.195
Table 10: Summary of ARIMA (p, d, q) ModelsFor Net FII Investment Regression BSE Sensex
ARIMA Standard AIC ARIMA SE AIC
(P, d, q) Error (P, d, q)
1, 0, 0 0.603 59.914 1, 1, 3 0.672 68.862
1, 0, 1 0.611 61.748 1, 1, 4 0.635 68.052
1, 0, 2 0.612 63.193 2, 0, 0 0.609 61.567
1, 0, 4 0.573 63.205 2, 0, 1 0.620 63.711
1, 0, 5 0.568 64.571 2, 0, 3 0.566 64.390
1, 0, 6 0.604 67.790 2, 0, 5 0.593 67.373
1,0,9 0.603 72.464 2, 1, 0 0.644 64.740
1, 1, 0 0.649 64.707 2, 1, 3 0.631 68.760
1, 1, 2 0.662 66.901 2, 1, 5 0.677 73.537
Table 11: Summary of ARIMA (p, d, q) Models For Net FII Investment Regressed on NSE Index
ARIMA Standard AIC ARIMA SE AIC
(p, d, q) Error (p, d, q)
1, 0, 0 0.599 59.419 2, 0, 1 0.616 63.242
1, 0, 1 0.606 61.285 2, 0, 4 0.577 64.617
1, 0, 3 0.599 63.574 1, 1, 0 0.644 64.307
1, 0, 4 0.569 62.677 1, 1, 2 0.641 65.740
1, 0, 5 0.583 64.759 1, 1, 5 0.618 69.798
1, 0, 7 0.598 68.587 2, 1, 0 0.633 63.709
1,0,8 0.592 71.648 2, 1, 1 0.649 65.832
2, 0, 0 0.604 61.109 2, 1, 3 0.667 69.513