Algorithmic Trading - Why Indian Stock Markets are Endangered

A few months ago, I had blogged about tight coupling in financial markets. In that entry,I had explained at a micro-level the impact of algorithmic trading. Given below is a "macro" story about how high-frequency trading in securities using computer programs played a big role in (though I would stop short of saying that they caused) a violent fluctuation in shares' and securities' prices on Wall Street last year. In less than 15 minutes, the Dow Jones Industrial Average Index plummetted and lost almost 6% of the opening value - and then recovered almost all of it in the next 15 minutes.

What happened on May 6, 2010 on Wall Street?

Major equity indices in futures as well as securities markets, already down 4% from the earlier day's close, suddenly plummetted a further 5-6% before recovering equally quickly, all in minutes. This affected almost all the 8,000 securities and ETFs in similar manner. Over 20,000 trades were reported to have been transacted at prices 60% from their prices just a few moments earlier.

Why did this happen?

At 2:32 pm, on an already volatile day, Waddell and Reed Financial Inc (not named by the joint CTFC-SEC report dated Sept 30, 2010, but named by many news reports) started a computer program to sell 75,000 E-Mini futures contracts worth close to $4.1 Bn. This was programmed to sell at any price and time, so instead of an orderly sale over a few hours, it sold this huge quantity of contracts within the space of 20 minutes, accelerating the sales as prices fell.

What actually happened during the crash?

The contagion spread to the equities market when arbitrageurs noticed the growing gap between the equities and futures prices. A significant finding is that 6 HFT (High Frequency Trading) firms (i.e., firms that extensively used algorithmic trading) remained active in the market even during the crash period of a few minutes. A blow-by-blow account follows:

• Five minutes into the crash, at 2:37 pm, data feeds from computers groaning under the huge numbers of contracts, started slowing down, leaving both, exchanges and investors uncertain about where share prices stood.
• The NASDAQ went into “self-help mode” at 2:37 pm where the transactions were not routed through NYSE's Arca electronic trading platform. CBoT and BATS exchanges (BATS at 2:49 pm) followed and also went into “self-help mode” which means that trades on NASDAQ, CBoT and BATS did not need to honour an Arca quote from NYSE.
• By 2:40, some trading and market-making firms started pulling out, due to algorithms that pause when they sense large price movements that could be due to questionable data feed. This left the market short of ready buyers and sellers. Apple, for example fell by $23 in 2 minutes, with the buy-sell spread going up to $5 instead of a few cents.
• Volumes of E-Mini contracts that normally mimic the S&P 500 surged but liquidity dried up. As a result, at 2:45:17 pm, E-Mini prices plunged 12.75 cents in half a second. This set off a circuit breaker that halted trading for 5 seconds.
• As individual stocks declined as much as 10%, ETF traders started withdrawing from the market.
• Then, at 2:46, even more strange things started happening because of the sheer speed difference between trades being put through and displayed – P&G shares were offered for purchase at prices higher than offered for sale! This is never supposed to happen in an electronic exchange.
• At 2:47, Dow reaches its nadir for the day, down 998 points or 9.2% from the opening level. Accenture, trading minutes earlier at $40, was offered at 1 cent.
• Then, at 2:49, the Dow rebounded by 300 points in 1 minute.

• There were no takers for ETF shares – iShares S&P500 Value Index Fund traded for 11 cents. But the broad recovery continued. By 2:58, indices reached the level they were at 2:30 pm.
• At 3:01, almost a half-hour to the minute since the crisis began, NASDAQ snapped out of its self-help mode and resumed routing orders to the Arca trading platform.
• At 4 pm, the DJIA closed 340 points below its previous close.

The Joint CFTC-SEC investigating committee reported that several HFT firms they interviewed had algorithms that took trading decisions based on direct proprietary data feed from the exchange directly rather than on consolidated market data, to reduce “latency” or delays measured in milliseconds. These algorithms went awry when the data feed from the exchange slowed down. Those HFT firms that did not depend on direct feeds for trading decisions got contradictory feeds that led to unease in taking decisions. Yet others that were not concerned with data latency in milliseconds simply withdrew from the markets.

The HFT firms that depended on their algorithms for trading decisions were not affected by the “self-help” declarations of NASDAQ, CBoT and BATS, and continued to rout orders to these exchanges. Therefore, the “self-help” declarations were ruled out as a cause of the volatility.

While no clear single cause was pointed out, HFTs using algorithms to trade rapidly (in one documented case, 200 trades exchanged hands 27,000 times in 14 seconds) were commonly thought of as the villains. It must be said, though, that there have been spirited defences by algorithmic trading experts, who point (among other factors) to volatility when markets are closed (ie difference between closing prices and opening prices on next day) as the real villain of the piece – on the logic that overnight differences can only be caused by humans, who are prone to panic unlike computers.

Even so, the SEC has since instituted a system of circuit breakers to arrest rollercoaster falls like the one experienced on Wall Street on May 6, 2010. This is another lesson that they have learnt by experience – instead of simply looking eastwards and learning from Indian bourses.

What can we learn from this?

But now, the stage has come to re-learn from our own wisdom. Algorithmic trading is now allowed on Indian bourses. Reports have suggested that over 40% of all trades are done by computers on NSE and BSE. Add to it the other dangerous fact - that FIIs that invest "hot money" that can fly out of the country in seconds account for over 70% of all floating stock (ie, stock that gets traded on the bourses).  See this in juxtaposition with the often displayed behaviour of FII fund managers who, like a herd of sheep, make a beeline for the two exits (NSE and BSE) for their investments at the merest sniff of danger anywhere in the world (even if it does not endanger their holdings in India), and it becomes clear that we have set up our bourses for spectacular volatility where securities' prices falling off a cliff in minutes will become sickeningly regular occurrences.