Automation, Competition, and the BSE
by Ajay Shah and Susan Thomas
This appeared in Business Standard, 20 March 1996. They also had an editorial alongside on the same subject.
India's equity market used to be dominated by the BSE. This used to be a market with trading by open outcry, without designated market makers, without any computerisation. The quality of this market was widely considered to be poor, in dimensions of transparency, liquidity and market efficiency.
One of the consequences of the great scam of 1992 was the urgency with which the finance ministry and SEBI sought radical reform in the functioning of the equity market. This was approached via the twin tracks of evolution (requiring that the BSE evolve towards ``screen-based trading''), and revolution (the creation of the NSE).
What does ``Screen Based Trading'' mean?
A consensus quickly emerged that ``screen based trading'' should be utilised, but screen-based trading can mean many different things. For example, the New York Stock Exchange (NYSE) relies on a human being (called the specialist) to do all order matching, and computers are only used -- in conjunction with old-fashioned telephones -- to communicate orders to the specialist.
A radical idea going around in the design of exchanges worldwide is the open electronic limit order book market. A limit order is just a buy or a sell order for a specified quantity at a specified price. In this kind of exchange, there are no human market makers. Orders are placed on screen, and the computer continually matches buy orders which have low prices against sell orders which have high prices.
The distance between the cheapest unmatched buy order and the costliest unmatched sell order is called the bid-ask spread. This is essentially the difference in price observed between buying and selling one market lot. Tight bid-ask spreads are synonymous with a highly liquid market. An example of this is Reliance, which has a spread on the NSE of around 5 paisa, versus Tata Chem, which has a spread on the NSE of around 50 paisa. For a comparison, spreads on the NYSE, where the average stock price is $40, are typically around 12.5 to 25 cents.
Both the BSE and the NSE avoided the NYSE and NASDAQ approach in exchange design. The NSE chose to implement a pure limit order book market. BSE's BOLT is also an automated exchange, in the sense that computers do order matching, except for a slight priority accorded to ``jobbers'', who offer two-way quotes.
Intuitively, it appears that the best execution would be obtained when all orders in the economy converge on one market. This should generate a deep, liquid market with tight bid-ask spreads. When multiple markets compete for orders, this generates fragmentation. This has caused much debate worldwide, with a fear that the quality of markets available to investors worsens when a competing market arises. Arbitrage between markets is the critical opposing force -- in a world with a large pool of arbitrageurs whose activities bind prices of all markets together, the best execution for an order is ensured, regardless of where it is placed.
While fragmentation raises these difficulties, competition offers inherent advantages by spurring exchanges in terms of quality of service, in dimensions such as clearing and settlement, surveillance, nationwide expansion, etc. Progress at the BSE has undoubtedly been driven by competition from the NSE, and vice versa.
Thus India's equity market has been shaped by the twin forces of exchange automation, and competition. How has the market changed as a consequence?
The NSE started equity trading on 4 November 1994, and BOLT has been considered stable as of 1 June 1995 or so. Thus in a period of around a year, we have gone from being one of the most primitive markets in the world to one of the most modern. How has this changed the quality of the market? We evaluate this question in a recent paper How competition and automation have changed the Bombay Stock Exchange, by comparing the behaviour of 1900 stocks across the periods before (May 94 to Oct 94) and after (Jun 95 to Nov 95) the transition of the market.
We find that liquidity has improved significantly in the second period compared to the first.
From the viewpoint of the country as a whole, the total trading volume was higher in the second period by a factor of Rs.100 crore per day. The NSE trades on all weekdays, and this helps generate liquidity on many days for which the BSE is closed.
On the BSE, liquidity has been enhanced at the security level, especially for smaller, thinly traded stocks. Among stocks with trading frequency below 95%, the change in period 2 has two opposing components : trading frequency is higher by around 17 percentage points owing to BOLT, and when NSE trades a given stock, it has an effect of -5 percentage points.
- Market Efficiency
Market efficiency is the simple idea that prices are not forecastable, because if they were, then thousands of profit-seekers would latch onto, and thus eliminate the forecastability. The small degree of forecastability that is observed in prices is caused by the cost of execution -- a combination of brokerage fees, and the bid-ask spread. When profits from forecasting are smaller than these transactions costs, people will not exploit them, and hence they will not be eliminated.
Competition between the BSE and the NSE has led to a sharp drop in brokerage fees, and the afore-mentioned improvements in liquidity have led to tighter spreads. Hence we would expect enhanced market efficiency, which does indeed prove to be the case. The degree of forecastability of prices is smaller today than what it used to be prior to BOLT and the NSE. On a related track, we find that the reaction time of the market to bad news is now faster than it used to be.
At first sight, the variance of BSE daily returns has dropped sharply, but this primarily reflects the new method used to calculate closing prices. Earlier, the closing price was the price seen on the last transaction. Under BOLT, the closing price is the average of the last 20 market lots. On balance, it appears that BSE volatility has risen slightly in the second period as compared with the first period. When a stock trades on the NSE, this has a slight negative impact on the BSE volatility of that stock.
This increase in volatility is not cause for alarm. Small increases in volatility are known to accompany automation the world over, and are consistent with enhanced market efficiency. Perhaps prices earlier were more sluggish in reacting to information; higher market efficiency hence implies higher volatility.
The reduced BSE volatility when the NSE starts trading a stock is probably associated with better surveillance on the NSE.
There are numerous allegations that the high-price and low-price reported by the BSE used to be artificially exaggerated. Are high and low prices reported under BOLT more accurate?
Based on economic theory, it is known that the high-low spread is proportional to volatility. In our examination of volatility, we know that volatility is slightly higher in period 2 as compared with period 1. If the accuracy of data is unchanged, then the high-low spread should also be slightly higher in period 2 as compared with period 1.
Instead, we find that the high-low spread has dropped sharply in period 2 as compared with period 1. This is consistent with the view that the high and low prices used to be exaggerated prior to BOLT, and the data released today is more accurate.
The full story of the BSE and the NSE has not yet played out, and there is every reason to expect that the quality of India's equity market will substantially improve in the coming few years. Our analysis shows that a clear payoff to institutional reform is already identifiable. This experience also hints at the opportunity available for improving the quality of the debt, commodity and foreign exchange markets.
Back up to Ajay Shah's media page