Intraday activity patterns and market microstructure effects in Indian capital markets: an empirical investigation using high frequency data

Abstract

This study analyzes the intraday activity (volatility, volume and trade frequency) patterns and market microstructure effects in the Indian capital market using a large dataset. Intraday 5-minutes volume and volatility are modeled to determine the effect of cross-listing of stocks on opening volatility and volume, the effect of weekend and holidays on the period immediately preceding and succeeding the extended trading halt and the volume and volatile behaviour during the last trading hour on maturing dates of derivative contracts. We also investigate the price impact of block (large sized) trades and the price behaviour surrounding such trades to access their information and liquidity effects. French and roll (1986) found that the variance of stock returns from open-to-close of trading is five times larger than the variance of close-to-open returns, and on an hourly basis, the variance is at least 20times the variance during non-trading period. Further, within the trading period, the intraday return volatility and volume depend on the level of liquidity, discreteness, institutional settings (Abhyankar et al., 1997), over-night trading halt (Brock & Kleidon, 1992), prevalence of private versus public information (Admati & pfleiderer, 1988), and impact of time zones (Guillaume et al., 1995) among others market with diverse microstructures may produce similar volatility dynamics at lower frequency level, but at higher frequencies, they may display very different characteristics, (Anderson 7 Bollerslev, 1998a). The pattern for intraday volatility of stock retunes-J shaped and the pattern for intraday trade volume and trade frequency is U-shaped in the Indian market. We explore if the intraday pattern depends on any stock specific characteristics like size of the company, and day trading level (percentage and absolute value). The shape of intraday volatility pattern is the same across all criteria, but there is a negative relationship between the level of variation in intraday volatility and the size of the company, the liquidity level, the price level, and the day trading level for the stock. Patterns for intraday trade volume and trade frequency are not dependent on the volatility level of the stock. However, the intraday volume and frequency pattern gets flatter for stocks having less liquidity or for which day trading proportion is lower. Behaviour of day traders may be responsible for the U-shaped patterns of intraday volume and trade frequency. The intraday periodicity (serial auto-coalitions) of the volume and volatility in India shows a U shaped pattern covering 67 lags (equivalent to 1 day) indicating that the intraday pattern is consistent across days. For modeling intraday volatility, we use Fourier Flexible Form (FFF) regression suggested by Gallant (1981). This model incorporates an intraday as well as a daily volatility competent. We use generalized methods of moments (GMM) regression to model the intraday volume behavoiur with dummies representing certain time-specific phenomena. Cross-listed stocks exhibit a significant increase in volatility during the initial 15 minutes of trading at market opening but thereafter trade at lower volatility than other stocks. At market opening following holiday and weekends we notice an increase in volatility during the first 15 minutes as compared to other days, but there is not much increase in the trading volume during the same period. In contrast, market close (last 15 minutes) before holidays and weekends experience and increased volume but reduced volatility as compared to other trading days. On F&O maturity is high during the last 30 minutes of trading the volatility increases by 80% (as compared to the volatility at the same period in other days) in between 3:00 to 3:15 pm and by 22% in the last 15 minutes (3:15 to 3:30 pm). Similar effects are also observed in volume. We expected these result because the settlement price of F & O contracts in NSE is determined based on the volume weighted average price of all trades executed from 3:00 to 3:30 pm of the maturity day. We analyze the permanent (information effect) and temporary (liquidity effect) impact of block trades transacted in the normal market window of NSE. Overall, the permanent price impact is more for block purchases are more information than block sales, which may be motivated by the liquidity need. Unlike in other markets, we observe that the temporary impact is greater than the permanent impact in case of block purchase. We classify the block trades as All-or-None (AON) and Not-Aon trades depending on the number of transaction through which a block order is executed. AON trades are simultaneous block purchase and sales being executed through the same transaction. They can be assumed to be pre-negotiated deals traded in the normal market. As expected, the price impact is higher for Not-AON trades as compared to AON trades. The market may be discounting the fact that AON trades. The market may be discounting the fact that AON trades, which are simultaneous purchase and sale of large blocks, may be motivated by factors other than arrival of multiple block trades increases market confidence on the information. The permanent price impact is higher for days where there are more than one block trades increase market confidence on the information. The permanent price impact is higher for days where there are more than one block trades of similar nature than for days with only one block trade our major findings are that the magnitude of intraday variation of volume and volatility is higher in the National Stock Exchange(NSE) of India as compared to other markets abroad. For Nikkei-225 and S and P 500, the volatility at intraday peak is around two times the volatility at intraday low, but for NSE, the peak level volatility is 4 -5 times the intraday low level. Such an intraday pattern is a manifestation of the high proportion of trading driven by day-traders in India. This is also evident by the flattening of the intraday volume pattern, in case of less liquid stocks, having low day trading interest. Substantial and persistent intraday variations also induce high periodicity in the volume and volatility in the Indian market. Like other markets, block purchases at NSE are more informative than block sales and the information content of AON (pre-negotiated) block trades is less than that of Non-AON trades. Permanent price impact is higher for days with multiple block trades as compared to days with a single block trade. In case of Not-AON block trades, we find that the prices start increasing 8 minutes before block purchases but prices start falling only 1 minute before block sales. This indicates that there is front running in case of block purchase but not in the case of block sales. We also observe that in the case of block sales, the price revert quickly leaving very small permanent price impact.