Intra-Daily Patterns of Nifty 50 Returns before and after Rolling Settlement on the National Stock Exchange

 

Haritika Arora

Assistant Professor, C K D Institute of Management and Technology, Opp. Model Town, Near Railway Station, G.T. Road, Amritsar

 

ABSTRACT:

This paper examines Intraday market return patterns for Nifty 50 on the Indian stock market for the search of time and seasonal anomalies for Pre-Rolling settlement period and Post rolling settlement period. In light of all the consequences, it has been concluded from this study that there exists a significant time of the day effect exposed in two effects - the open jump effect and persistent end of session effect. The first 20 minutes of opening is significant for all days of the week, additionally, for Wednesday first 35 minutes is evident for pre-rolling settlement period. But in post rolling settlement period, only the first 20 minutes are found to have significant returns for all trading days. Further, the persistent end of session effect is seen for all trading days except for Monday (for Pre-Rolling settlement period) and Wednesday (for Post-Rolling settlement period).

 

 

 

1. INTRODUCTION:

Under the rolling settlement, securities transactions settle within many days after the transaction is struck. The new settlement system T+2 in Indian Securities Market effectively requires traders to deliver either cash or stock for every transaction undertaken earlier. The introduction of rolling settlement was aimed to both reduce risk (by shortening the time between trading and settlement), and to increase the efficiency of the settlement system. (Kyriacou and Mase, 2000).

 

This study tried to explore Intraday dynamics of Nifty 50 returns before and after introduction of T+2 rolling settlement in a hope to investigate some persistent and some diminishing intraday anomalous pattern for the two periods. High frequency data have been used for this study to explore such Intraday patterns.

 

High Frequency data series are complex in nature, occur at unequal time intervals, have distinguished intraday periodicity in the data with distinctive short-lived volatility dynamics (Goodhart and Hara; 1997, Andersen; 2000). Analysis done with high frequency data helps to understand the micro-structural market dynamic of various financial markets. Low frequency data (daily, weekly or monthly data sets) lacks the ability to capture micro-structural properties of the financial market.

 

2. LITERATURE REVIEW:

Substantial evidence has been accumulated on asset price behaviour, including stock returns, commodity prices and exchange rates using daily, weekly or monthly data for Indian markets. However, there is no substantial evidence for the time-of-the-day effect for Indian Stock Markets. Therefore, this study has attempted to study the intraday return patterns before and after the introduction of rolling settlement.

 

Literature demonstrate the presence of Intraday patterns in returns, volatility, trading volume, bid-ask spread, liquidity for financial market. Alabed and Al-khouri (2008) investigated the cross-market Intraday liquidity behaviors and exhibited highest liquidity levels when the market opens and the market closes and it follows U shaped pattern. Corroborating the similar results, Strawinski And Slepaczuk (2008) found that in the Warsaw Stock Exchange there was the existence of the strong open jump effect for all days except Wednesday and positive day effect for Monday and observe positive, persistent and significant open jump effect and the end of session effect from Monday, Thursday and Friday. Reasons behind such significant returns at opening and closing of the trading session was explored by Admati and Pfleiderer (1988). They observed such volume and price variability at particular time are due to the interacting strategic decisions of discretionary liquidity traders, which will prefer to trade when the market is traded bulky, which provoke them to trade at the same time. This further draw attention of informed traders to take advantage more from their private information when trade is done by noise traders.

 

Major time of day effect anomalies explored are opening sessions and closing sessions effects from some literature. Wood, Mc Inish and Ord (1985) and Mc Inish and Wood (1992) show that trading activity is elevated at the open; it turn down to a low point at midday and then increase at the close. Abhyankar et al (1997) show that the trading volume presents a double-humped pattern reaching, but intraday bid ask spread seems constant throughout the trading day, but volume seems low at 13:30 hours. However, the U shaped pattern was observed for the heavily traded stocks. Correspondingly, Al Suhaibani and Kryzanowski (2000)  also find that the number and volume of transactions show evidence of a U-shaped pattern during each within-day session. Similar patterns were explored by Bildik(2001) in Turkish stock markets is that the return, volume and volatility of the stock prices and bid-ask spreads all follow a U-shaped or more precisely a W-shaped pattern over the trading day at the Istanbul Stock Exchange, since there are two separate trading sessions in a day due to lunch break. But L-shape pattern was observed separately for the two sessions of trading for all days of the week on the Turkish stock market.

 

3. DATABASE AND RESEARCH METHODOLOGY:

The objective of this research work is to examine Intraday return patterns for two periods in the sample. In order to explore such patterns, high frequency data for Nifty 50 is purchased from professional set-up "DotEx International Limited" (100% subsidiary of National Stock Exchange) which deals with the data and info-vending products. Based on nearest figure and volume adjusted weighted average of transaction data, complete data sets is extracted out with fixed interval of 5 minute, so as to make suitable for analysis of various econometric models. Intra-daily return and volatility is best forecasted by Intraday 5-minute interval returns. (Martens, 2002)

5 minute return is computed as the first difference of the natural logarithm as follows.

 

r_t = (log p_t–log p_t-1)

where,

r_t = Market return for the period t

P_t = Price index for the period t

P_t-1 = Price index for the period t–1

log = Natural log

 

Firstly, the summary statistics have been computed using Mean, Standard Deviation, Skewness and Kurtosis of Nifty 50 returns to develop the research design. Then, Augmented Dickey and Fuller (ADF) test has been applied to check the stationarity in price series as well as differenced series. ADF test has been applied to determine unit roots that consist of regressing the first difference of the series against a constant, the series lagged one period and the differenced series at n lag lengths (Pindyck and Rubinfeld 1998). The null hypothesis for ADF test is that the series has a unit root (series is non-stationary), against the alternative that series is stationary.

 

Presence of auto-correlation (ACF) and partial auto correlation (PACF) functions in stock market returns form a part of the identification of a suitable ARIMA (Autoregressive Integrated Moving Average) model. When error terms of this developed Autoregressive Integrated Moving Average model for stock returns does not exhibit constant variance (shows the period of high volatility is followed by the period of high volatility and the period of low volatility is followed by the period of low volatility), this suggests that the residuals or error terms are conditionally heteroscedastic and can be represented by ARCH and GARCH model.

 

15 minutes returns dummy are introduced into the return equation of the model as it is thought that the 15 minutes time interval is small enough to capture the sensitivity of stock prices and a large enough period for the market to react to some news and further, it is found to be sufficient time to conduct a trade. Moreover, fifteen minutes time intervals have been used in other studies such as Harris, 1986; Abhyankar et al., 1997; Bildik, 2001 Niarchos, N. A. Alexakis, C. A., 2010 and thus there is a source for comparisons. There is an exception in first dummy interval which is taken as 20 minutes because during the period of study, market opens at 9:55 am and closes at 3:30 pm and equal division in intervals of 15 minutes was not possible. For Intraday tests, twenty two dummy variables Di to represent intervals from 9:55am-10:15am through 3:15pm-3:30pm for the period from 1st Jan 1999 to 31st March 2003(for the period before the introduction of rolling settlement) for each day and for overall sample has been taken.

 

An econometric model which is used to investigate the presence of stock return patterns, is a model in which stock returns are set to be time dependent , that is,

Rt=α₁D₁ + α₂D₂ + α₃D₃ +.......... α₂₂D₂₂ +∑ β Rt-n

where: Rt is a series of actual stock returns and D₁ ,D₂ ,D₃ ...., D₂₂ are dummy variables which refer to the stock returns of 15 minutes time intervals 1; 2; 3; . . . ; 22 during a each trading session. Rt-n represents lagged intraday returns

 

Under the Efficient Market Hypothesis, stock returns should be independent of time i.e.

α₁ =  α₂ =  α₃ = ......... = α₂₂ = 0

 

When the Intraday returns variances dependent of time, then above models were adjusted to take into account these Autoregressive Conditional Heteroscedasticity (ARCH) effects. A natural extension of an ARCH(q) model is a GARCH model, which is widely employed in practice, as it overcomes problems associated with ARCH model. GARCH model allows the conditional variance to be dependent upon previous own lags

 = ώ+ ώ₁ + ²_t-1

 

This is a GARCH(1,1) model where σ_2t is called as conditional variance. This model focuses on the time-varying variance of the conditional distributions of returns. Using the GARCH model it is possible to interpret the current fitted variance, σ_t2 , as a weighted function of a long-term average value dependent on ώ, information about volatility during the previous period (ώ₁) and the fitted variance from the model during the previous period (σ²_t-1). GARCH model can be expressed in a form that shows that it is effectively an ARMA model for the conditional variance. The advanced volatility models like GARCH capture the financial market volatility that appears in clusters and persist over the time.

 

4. INTRADAY PATTERN (TIME OF DAY EFFECT) BEFORE AND AFTER INTRODUCTION OF ROLLING SETTLEMENT:

This section deals with the investigation of Intraday patterns for pre- rolling settlement period(1st January 1999-31st March 2003) and post-rolling settlement (1st April 2003-31st December 2009) using 5 min return. Section 4 is further bifurcated into different sub-sections. Sub Section 4.1 deals the results of descriptive statistics of Nifty 50 for both the periods. Subsection 4.2 describes Intraday pattern (Time of Day effect) in Nifty 50 using GARCH methodology.

 

4.1  Descriptive Statistics of Intra-Daily Returns of Nifty 50:

Descriptive statistics has been computed to study the distribution pattern of the 5 minute intra-day trading returns. The analysis of mean, maximum values, minimum values, standard deviation, skewness and kurtosis has been done. Further, normality test also has been checked applying Jarque bera test. Skewness and kurtosis helps to understand the characteristics of a distribution. In all cases the logarithmic transformation of the price series was used and the returns were calculated as the difference of the logarithmic prices i.e ln(Pt-Pt-1)

 

Table 1: Descriptive Statistics of Intra-daily returns of Nifty 50 for  pre- rolling settlement period and post-rolling settlement.

	Nifty 50 (pre- rolling settlement)	Nifty 50 (post- rolling settlement)
Mean	1.38E-06	1.47E-05
Median	1.28E-05	6.10E-05
Maximum	0.053208	0.110899
Minimum	-0.076292	-0.127986
Std. Dev.	0.002124	0.002201
Skewness	-1.516766	-1.033203
Kurtosis	128.9846	254.2210
Jarque-Bera Probability	46904806  (0.000)*	2.98E+08  (0.000)*
Price Series (Augmented Dickey Fuller Test Statistics)	-2.047142 (0.2667)	-1.183316  (0.6840)
Return Series (Augmented Dickey Fuller Test Statistics)	-190.1106 (0.0001)*	-144.8324 (0.0001)*

 

Table 1 reveals the descriptive statistics of Nifty 50 for the period before and after introduction of rolling settlement. Mean returns have been seen positive for Nifty 50 for the both periods. The coefficient of Jarque-bera is significant at one percent for Nifty 50 for both the periods. It documents that the trading returns are asymmetric and do not conform to the normal distribution. Leptokurtic distribution (kurtosis>3) of all the trading returns for all trading days is evident. This fat-tailed character is consistent with earlier studies (Huisman and Huurman (2002), Higgs and Worthington (2005) and Wolak (2000) and is driven by the prevalence of extremely large spikes in the returns. Literature demonstrates that stock returns tend to follow non-normal unconditional sampling distributions which account for the excess kurtosis and fat-tailed properties of the data. In consistent with fat tailed characteristics, Kayahan et. al, (2002) confirmed in high frequency finance literature that realized volatility provides a better fit than the normal GARCH model. Stock market experiences volatility clustering and hence, GARCH-type models predict the market volatility far better fit than the other simple volatility models such as historical average, moving average, EWMA etc. Asymmetric variants have also been observed in GARCH models give better fit than the symmetric GARCH model, confirming the presence of the leverage effect.

 

According to Dickey and Fuller (1979), a time series is stationary if it’s mean, variance and auto-covariance (at various orders lags) are time invariant at different point of measurement. An augmented Dickey-Fuller test is a test for a unit root is present in an autoregressive model for a time series sample. Table 1 represents Augmented Dickey and Fuller (ADF) test statistics has for price series as well as log differenced series. Results represents of the Augmented Dickey-Fuller (ADF) test clearly rejects the hypothesis of a Unit Root at the 1% level of significance for return series but price series at level is non stationary.

 

The basic assumption for the modeled error terms is that they are unrelated, normally distributed and their variances do not vary with the effects being modeled. If error terms, do not have constant variance, they are said to be heteroscedastic and if they are correlated, they are said to be auto-correlated. Table 2 suggests Breusch-Godfrey Serial Correlation LM Test and Heteroscedasticity :ARCH Test statistics for returns residuals which clearly depicts the presence of serial correlation and heteroscedasticity in data.

 

The heteroscedaticity in residuals of Nifty 50 can also be observed from figure 1. When the period of high volatility followed by the period of high volatility and the period of low volatility followed by the period of low volatility, this suggests that the residuals are conditionally hetroscedastic, can be represented by ARCH and GARCH model. The tests for the presence of ARCH effects and for developing of the appropriate ARCH model, Q and LM tests from Lee and King (1993) and hetroscedasticity tests from Wong and Li (1995) can be used. With the help of visual inspection of the plotted autocorrelation function, identification of the suitable ARMA (Auto-regressive moving average) model is done with the help of the auto-correlation (ACF) and partial auto correlation (PACF) functions from correlogram is done. Further, the Ljung- Box Q-statistic is used for diagnostic checking. The Q-statistic is a modification of the Box-Pierce test statistic (Box and Pierce, 1970); this was suggested for testing AR, MA and ARMA models.

 

Table 2: Breusch-Godfrey Serial Correlation LM Test and Heteroscedasticity :ARCH TEST

 

 

Figure 1: Residuals of Nifty 50 for  pre- rolling settlement period and post-rolling settlement.

 

In following subsection, Intraday patterns for the Nifty 50 is observed in pre- rolling settlement period and post-rolling settlement period using time dummies in the return equation of GARCH model. This econometric model which is used to explore the presence of Intraday stock return patterns are time dependent.

 

4.2 Intraday pattern (Time of Day effect) in S and P CNX Nifty 50 using GARCH(1,1)

The present study examined the robustness of the evidence for the time of day effect anomaly in stock return data after accounting for the impact of all the possible causes for the effect. The current section examines the presence of time of day effect for index Nifty 50 two periods. However, it is evident that there is significant variation in the return across the trading session over the study period. The literature suggests the presence of certain sets of stock market anomalies which has accumulated over a decade. Disincentive Intraday patterns have been observed with significant returns at the beginning and the end of the trading day (Strawinski and Slepaczuk, 2008; Niarchos, N. A. Alexakis, C. A., 2003).

To ease the eventual autocorrelation problem, suitable identification of ARMA model is done with the help of the auto-correlation (ACF) and the partial auto correlation (PACF) functions from correlogram. According to ARMA model, the best fit model for Nifty 50 (pre- rolling settlement) is AR(1) and AR(2). AR(3) and AR(4) model is found to be suitable for Nifty 50 (post- rolling settlement).

 

Table 3: Intra-daily effects based on five-minute return data models of Nifty 50 for the period before introduction of rolling settlement

*1%, **5% significance level

Table 4: Intra-daily effects based on five-minute returns data models of Nifty 50 for the period after introduction of rolling settlement

*1%, **5% significance level

 

Visual inspection of Table 3 reveals the significant negative returns for open session on Monday, Tuesday and Friday, but significant positive returns for Wednesday and Thursday for Nifty 50 for the period of pre-rolling settlement. The positive open session effect is continued till second 15 min interval for Wednesday. Similar negative returns on Monday, at market opening is observed in Table 4, that is, for the period of post rolling settlement. Significant negative returns on Monday, at market opening corroborates the results of Lee et al. (2012) and confirms the presence of the Monday effect which is driven by large negative returns accrued during early Monday mornings. However, Tuesday, Wednesday, Thursday and Friday have significant and positive returns in Nifty 50 for a period of post rolling settlement. Opening session effect continues till second interval, that is about 35 minutes from market opening on Wednesday for pre-rolling settlement period, but after a post rolling settlement period, only the first 20 minutes is found to have significant returns for all trading days.

 

Significant positive returns for Monday, Wednesday and Friday for the end of trading session, but significant negative returns for Tuesday, Thursday and in complete period has been seen at the end of trading session for pre-rolling settlement period. But significant positive returns for Thursday and Friday for end of trading session and significant negative returns for Monday, Tuesday, Wednesday and in complete period is seen for period of post rolling settlement. Moreover, End of session effect starts earlier in intervals corresponding last trading interval for almost all weekdays except on Monday, which includes a maximum of significant last hour trading session for the complete period before the introduction of rolling settlement. Furthermore, after the introduction of rolling settlement, end of session effect also starts earlier in intervals corresponding last trading interval for all weekdays except on Wednesday, which includes a maximum of significant last forty five minutes of trading session. Persistence of closing session effect corroborates the results of Abhyankar et al.(1997).

 

Negative opening Monday returns following higher closing Friday returns for both the periods demonstrate the presence of Weekend Effect is consistent with intra-daily patterns reported by Harris (1986). Lakonishok and Levi (1982) suggest similar results that expected returns are higher on Fridays and lower on Mondays relative to either a trading or calendar time model.

 

Mid–session significant results are only statistically different from zero for Tuesday at 11:15am-11:30am, Thursday at 12:30pm-12:45pm, Friday at 12:15pm-12:30pm, 2:15pm-2:30pm and 11:00am-11:15am, 12:00noon-12:15pm, 1:30pm-1:45pm for a complete data period before the introduction of rolling settlement. And significant mid-session effect is seen for Thursday at 11:00am-11:15am and 1:00pm to 1:15pm, Friday at 12:00noon-12:15pm and 12:15pm -12:30pm, 1:00pm-1:15pm and 1:45pm-2:00pm for complete data period after introduction of rolling settlement. The existence of these effects can be explained on the basis of major announcement which were revealed at many different hours of the day. Otherwise, this may be due to based on traders’ habits (which changes time to time).

CONCLUSION:

This paper examines intraday patterns of returns for Nifty 50, a major stock market index of National stock exchange for the emerging Indian Stock market. Taking into consideration all the results, it has been concluded that there exists strong time of the day effect exposed in two forms - the open jump effect and persistent end of session effect. Open session effect and end of session effect is consistent with earlier studies (Cornett, Schwarz and Szakmary, 1995; Bildik,2001; Strawinski and Slepaczuk, 2008; Deev and Linnertová, 2012). Opening session effect continues till second interval, that is about 35 minutes from market opening on Wednesday for pre-rolling settlement period, but after a post rolling settlement period, only the first 20 minutes is found to have significant returns for all trading days. Further, persistent end of session effect is seen for all trading days except for Monday(for Pre-Rolling settlement period) and Wednesday(for Post-Rolling settlement period). However, various mid sessions were also found to be significant which may be due to announcement effect or based on traders’ habits (which changes time to time).

 

Future research into this field which could enable finding the answer for the more explanations for market inefficiencies for emerging economies like India. Further studies can be performed on Intraday patterns of volume, liquidity, bid-ask spread, volatility and provide significant inter linkages between them, so as to explore more legitimate justifications for Intraday patterns observed over a trading session.

 

REFERENCES:

1.        Abhyankar A., D. Ghosh, E. Levin and R.J. Limmack (1997). Bid-ask Spreads, Trading Volume and Volatility: Intraday Evidence from the London Stock Exchange. Journal of Business Finance and Accounting, 24 (3)  and  (4), 343-362.

2.        Admati, A. R.,  and  Pfleiderer, P. (1988). A theory of intraday patterns: Volume and price variability. Review of Financial studies, 1(1), 3-40.

3.        Alabed, M. F.,  and  Al-khouri, R. (2008). The pattern of intraday liquidity in emerging markets : The case of the Amman Stock Exchange. Journal of Derivatives  and  Hedge Funds Volume, 14(3/4), 265-284. doi:10.1057/jdhf.2008.18

4.        Al-Suhaibani, M.,  and  Kryzanowski, L. (2000). An exploratory analysis of the order book, and order flow and execution on the Saudi stock market. Journal of Banking  and  Finance, 24(8), 1323-1357.

5.        Andersen, T. G. (2000). Some reflections on analysis of high-frequency data. Journal of Business  and  Economic Statistics, 18(2), 146-153.

6.        Bildik R. (2001). Intra-day Seasonalities on Stock Returns: Evidence From the Turkish Stock Market. Emerging Markets Review, 2, 387-417.

7.        Box, G. E.,  and  Pierce, D. A. (1970). Distribution of residual autocorrelations in autoregressive-integrated moving average time series models. Journal of the American statistical Association, 65(332), 1509-1526.

8.        Cornett, M. M., Schwarz, T. V.,  and  Szakmary, A. C. (1995). Seasonalities and intraday return patterns in the foreign currency futures market. Journal of Banking  and  Finance, 19(5), 843-869.

9.        Deev, O.,  and  Linnertová, D. (2012). Intraday and intraweek trade anomalies on the Czech stock market. Acta universitatis agriculturae et silviculturae mendelianae brunensis, 60, 8.

10.     Dickey, D. A.,  and  Fuller, W. A. (1979). Distribution of the estimators for autoregressive time series with a unit root. Journal of the American statistical association, 74(366a), 427-431.

11.     Goodhart, C. A.,  and  O'Hara, M. (1997). High frequency data in financial markets: Issues and applications. Journal of Empirical Finance, 4(2), 73-114.

12.     Goodhart, C. A., Hall, S. G., Henry, S. B.,  and  Pesaran, B. (1993). News effects in a high‐frequency model of the sterling‐dollar exchange rate. Journal of Applied Econometrics, 8(1), 1-13.

13.     Harris L. (1986), “A Transaction Data Survey of Weekly and Intraday Patterns on Stock Returns”, Journal of Financial Economics, 16, 99- 117.

14.     Higgs, H.,  and  Worthington, A. C. (2005). Systematic features of high-frequency volatility in Australian electricity markets: Intraday patterns, information arrival and calendar effects. The Energy Journal, 23-41.

15.     Huisman, R.,  and  Huurman, C. (2003). Fat tails in power prices (No. ERS-2003-059-F and A). ERIM Report Series Research in Management

16.     Kayahan, B., Saltoglu, T.,  and  Stengos, T. (2002). Intra-day features of realized volatility: evidence from an emerging market. International Journal of Business and Economics, 1(1), 17-24.

17.     Kyriacou, K.,  and  Mase, B. (2000). Rolling settlement and market liquidity.Applied Economics, 32(8), 1029-1036.

18.     Lakonishok, J.,  and  Levi, M. (1982). Weekend effects on stock returns: a note.The Journal of Finance, 37(3), 883-889.

19.     Lee, J. H. and King, M. L. (1993). A Locally Most Mean Powerful Based Score Test for ARCH and GARCH Regression Disturbances. Journal of Business and Economic Statistics, 11(1), 17–27.

20.     Lee, S., Kim, C. S.,  and  Kim, I. M. (2012). Testing the Monday Effect using High-frequency Intraday Returns: A Spatial Dominance Approach. The Korean Economic Review, 28(1), 69-90.

21.     Martens, M. (2002). Measuring and forecasting S and P 500 index‐futures volatility using high‐frequency data. Journal of Futures Markets, 22(6), 497-518.

22.     McInish, T. H.,  and  Wood, R. A. (1992). An analysis of intraday patterns in bid/ask spreads for NYSE stocks. the Journal of Finance, 47(2), 753-764.

23.     Niarchos, N. A.,  and  Alexakis, C. A. (2003). Intraday stock price patterns in the Greek stock exchange. Applied Financial Economics, 13(1), 13-22.

24.     Pindyck, R. S.,  and  Rubinfeld, D. L. (1998). Econometric models and economic forecasts (Vol. 4). Boston: Irwin/McGraw-Hill.

25.     Strawinski, P.,  and  Ślepaczuk, R. (2008). Analysis of high frequency data on the Warsaw stock exchange in the context of Efficient market hypothesis. Journal of Applied Economic Sciences (JAES), (5), 306-319.

26.     Wolak, F. A. (2000). Market design and price behavior in restructured electricity markets: an international comparison (pp. 127-152). Springer US.

27.     Wong, H. and Li, W. K. (1995). Portmanteau Test for Conditional Heteroscedasticity, Using Ranks of Squared Residuals. Journal of Applied Statistics, 22(1), 121–134

28.     Wood, R. A., McInish, T. H.,  and  Ord, J. K. (1985). An investigation of transactions data for NYSE stocks. The Journal of Finance, 40(3), 723-739.

 

 

 

 

 

Accepted on 27.06.2017          © A&V Publications all right reserved