Algorithmic trading is all the rage in India right now, and across the market the view is unanimous: the only way is up. The question is how just fast it will grow.
By the start of this year, algo accounted for some 24 percent of cash equities turnover in India and about 30 percent of equity derivatives. According to figures from the Bombay Stock Exchange, by far the smaller of the two Indian exchanges that dominate equities trading, the share for equity derivatives has already jumped to 45 percent since then.
"Algorithms and high frequency trading are the hottest topics in the market - algorithmic trading and HFT itself, and now the regulations around it," says Chetan Pandya, head of IT for one of India's largest local brokers, Kotak Securities. "This is what the majority of players in the market are focused on today."
India has the building blocks in place for a ramp-up. Co-location has been available from both the Bombay Stock Exchange and its bigger competitor, the National Stock Exchange, for 18 months. Both exchanges, and market observers, say their trading platforms can handle HFT. Direct market access is available. Smart order routing between the two exchanges has also been operating since August 2010.
The Indian regulator, the Securities and Exchange Board of India (SEBI), produced guidelines for algorithmic trading in March which brokers, exchanges and market watchers hail as a sensible response. The new rules, they say, recognise that algorithmic trading is a natural development and are aimed at preventing problems but not blocking growth.
"All the dynamics point to an increase in automated and algo trading in the next few years," according to Celent senior capital markets analyst Anshuman Jaswal, who is based in India. He expects levels of about 40-45 percent across cash equities and derivatives within the next year or so.
Global data centre provider Equinix, which specialises in huge high-performance data centres, is taking a detailed look at India. "We've had a lot of interest from clients," said Stewart Orrell, Equinix director of global financial services. "We're certainly seeing demand."
India market insiders say a site of the size and scale that Equinix builds around the world would have a big impact on India's attractiveness to foreign algo players, making it much easier for them to enter the market.
Even without such a site, India has appeal for foreign algo players because it has several different trading venues, unlike many other Asian markets. That makes for attractive competition, especially in the equity and equity derivatives market. The Bombay Stock Exchange (BSE) is fighting hard to chip away at the dominance of the National Stock Exchange (NSE), which accounts for at least three-quarters of equity trading volumes. The Multi Commodity Exchange (MCX), which dominates in commodities and in currency futures, is keen to win regulatory approval for equities trading, and in March became the first Indian exchange to list in one of India's largest and most successful recent IPOs. And a newly reinvigorated Delhi Stock Exchange has partnered with the LSE's MillenniumIT to take a new low-latency platform, which could further shake up the market.
SEBI has also sent a wake-up call to some of the dozen or so dormant regional exchanges in India, warning that if they do not reach minimum standards of liquidity they will face being wound up. That could trigger even livelier competition as those exchanges stir, consider how to attract liquidity, and ponder investment in technology and perhaps consolidation.
India is probably one of the most progressive markets outside Europe and the US from an algorithmic trading point of view, according to Citi's head of non-dollar electronic sales, Mani Singh, who is based in New York. "It has built out a significant infrastructure and has co-lo speeds of less than two milliseconds; it's a vertically integrated exchange, which is multi-asset in nature, making it attractive to a broad range of participants," he said.
Small orders - lots and lots of them
A relatively high volume of smaller orders twinned with tight spreads help make India appealing, and also challenging, for algorithmic trading. Brokers say spread profiles have stayed relatively stable despite shrinking market turnover in India over the last few months.
Ravi Apte, CTO of the NSE, says he expects the volume of order messages on his exchange could double in the next couple of years from current levels of about 350-375 million messages a day, with peaks of about 450 million. The BSE says three years ago order-to-trade ratios were running at about three or four to one; now they are at 20 to 30 in cash equities, and 50 in derivatives.
Even a cooling Indian economy and the resulting market fallout is seen as a potential boon.
Growth slowed to an annual 6.5 percent for the year to March 2012 from 8.4 percent a year before, knocking Indian markets. World Federation of Stock Exchanges data show NSE electronic order book share trading value slid nearly 30 percent in the year to December 2011, and market cap in dollar terms was down nearly 40 percent. But the pressure of falling prices and turnover may have spurred growth in algorithmic trading.
"Because of the economic slowdown, the trend [towards algo] has actually picked up pace in the market," said Pandya of Kotak Securities from Mumbai. Investors see algo trading as a way to cut costs during tough times.
More than 80 of the NSE's 1,200 or so members have taken space in its co-location site, while the BSE's site has about 25 users. The scale of India's high-speed infrastructure would take a dramatic leap if Equinix goes ahead with a site in India. "If you look at the NSE, they have about 200 cabinets in their site for co-location, and the BSE has about 80 to 100," said Equinix's Orrell. "When we build a site, the smallest site we tend to look at is about 1,000 cabinets."
The exchanges, meanwhile, are pushing their infrastructure hard to shrink latencies. "We're still in a millisecond world, not a microsecond world," said BSE adviser Jim Shapiro. "India is still a relatively slow market, but it's fast enough for some high frequency strategies. And over the next two or three years you're going to see continuing convergence of Indian latency towards global levels." The development of algo trading in India through the asset classes is following the pattern already established in mature markets, though India has some of its own special plot twists. The usual trend of cash equities first, equity derivatives next, has been reversed in India. India's transaction tax, which applies to both sides of cash equity transactions, accounts for about 75 percent of trading costs. That factor has driven equities volumes, including algorithmic trading volumes, towards equity derivatives - the tax is lower on futures and options.
The NSE is at number three in the World Federation of Stock Exchanges ranking for number of equity derivatives traded in 2011, with a growth rate of 37 percent from the year before - a striking increase against overall world growth of 9.4 percent. The NSE is in second place in the world for volume of stock index options traded, and third in both single stock futures and stock index futures.
The transaction tax is cited by many as a potential roadblock for growth in high frequency trading. Strategies based on small spreads and heavy volumes are less attractive in equities because of the tax, unless the frequency relates to order placement rather than actual trades. "My feedback from international players I've spoken to about India is that it is an intriguing but complicated market, and one of the complications is the transaction tax," said Aite Group analyst Danielle Tierney, who is based in Boston.
Governments are usually unwilling to abandon tax revenues in difficult economic times, so many market players do not expect the tax to disappear anytime soon. But what could happen is a change in the structure of the tax. It does not apply to interest rate, currency or commodity derivatives now, which some say is neither logical nor fair. A smaller tax on a broader base of asset classes is an alternative which is thought to be under consideration at India's finance ministry, though opponents fear an Indian transaction tax in globalised markets such as commodities could force trading offshore and reduce liquidity in domestic Indian markets.
Another hurdle for algo traders in India is split settlement across exchanges. Smart order routing is in place between the NSE and the BSE, but each exchange does separate settlement. That means higher post-trade costs and higher capital utilisation, as brokers using smart order routing are forced to put collateral in each location. Any change on that front is likely to require regulatory muscle, to push either interoperability or clearing house merger. The busy regulator, SEBI, is also looking at this area.
NSE CTO Apte said that unlike some other locations with central counter-parties across multiple trading facilities, exchanges in India do real-time risk management, so clearing and settlement are preceded by checks on margins. Changes to settlement could affect the ability to do real-time risk management, so any change would have to be done carefully. "There hasn't been an adequate dialogue about how to deal with this practically yet," he said.
The rival Bombay Exchange believes centralised or interoperable clearing would be good for India as an investment venue, and does not believe it poses any problems on additional latency for risk management. "In India both exchanges are required to have pre-trade risk management, and it does increase latency a bit," said the BSE's Shapiro. "It's not a big deal, but it's one of the reasons Indian numbers on speed are a little lower compared to countries that don't have pre-trade risk management. We feel pretty comfortable with that, and with algorithmic trade growing, pre-trade risk management is becoming more and more part of the ecosystem."
Despite the advent of Smart Order Routing, the Bombay Stock Exchange has struggled to win market share. The additional costs of split settlement may be a factor. "If you want to take advantage of the fact that there are two exchanges to use SOR, you have to weigh that against the additional cost of trading two tickets and clearing two destinations," Shapiro said. Brokers are making decisions about whether to use SOR based on order size. "Beyond a certain order size, the price improvement benefits of SOR clearly outweigh the additional post trade costs of having two tickets. This is increasingly borne out by data provided by major firms like Goldman, Credit Suisse and others who have found significant price improvement of 3-4 bps from using SOR in India."
In cash equities, the NSE takes about 80 percent of transactions, and 90 percent or more of equity derivatives. BSE efforts to increase liquidity have included introduction of a 'maker/taker' model, with transaction rebates for market makers providing liquidity. But brokers say it is still battling to capture more liquidity. That suggests a big struggle ahead for the revitalised Delhi exchange, which has not traded for about a decade, when it re-launches electronic trading. The exchange has not given a date but the re-launch is expected during 2012. Citi's Singh: "In this environment where we are seeing low volumes it's going to be extremely competitive and challenging, but that being said, having a choice can certainly be a factor for stimulating growth. If it's competitively priced and it adds value, you will see order flow migrating to their venue."
Liquidity is all via exchanges in India - dark pools are not permitted by Indian regulations, and market players do not expect that to change any time soon.
The NSE, which was the upstart that launched electronic trading after it was formed in 1992, says it welcomes the prospect of new competition. "We changed the game; we fully understand that's what competition does. So new exchanges will come up, and everybody will deal with it as best we can," Apte said.
One area of the market, the broking industry, has been surprisingly unchanged by electronification, the fragmentation of trading venues, the introduction of Smart Order Routing and the arrival of Direct Market Access and algorithmic trading. Most of the large investment banks have had local broking operations in India focused on foreign institutional investors for the last 10 years. "Almost all of them have been here since 2002 or 2003. They are more or less Indians now," said one domestic broker. But so far there has been little consolidation among Indian brokers. Market observers say that just as the depth of the order book in India partly reflects the strong part retail investment plays in the market, the enormous and widely dispersed Indian broking industry is serving a large market of small investors.
"Even though the exchanges are all electronic, there are still a lot of individual investors," NSE's Apte said. "That number is in the tens of millions." The NSE has more than 200,000 dealing screens spread all over the country; there is at least one, and sometimes more, in almost every town with a population of 10,000 or above. Industry observers say at times of low volume, some of the army of thousands of small brokers switch attention into other areas such as wealth management while they wait for the market to pick up. Meanwhile, the bigger domestic players such as Kotak Securities and Edelweiss are competing with the big international players, offering more sophisticated services such as algorithmic trading and Smart Order Routing execution to an institutional client base.
Outside of equities and equity derivatives markets, algorithmic trading is still relatively small. Algo trading accounts for about 10 percent of the FX market, according to Celent's Jaswal. It's lower still, at 5 percent or below, in commodity derivatives. Indian regulations prevent foreigners and foreign banks from trading in those markets. But the regulations in equities markets are looser - foreign institutional investors are allowed access via India-based brokers after registration with SEBI, meeting criteria including diversified ownership and management of client money. In January the regulator made another tweak to the rules to establish a category called 'Qualified Foreign Investors', a simplified route for foreign investors to directly access equities. That may make it easier for foreign hedge funds and specialist proprietary traders to trade the Indian market remotely, though brokers are still grappling with how to set that route up.
Rising order-to-trade ratios
Inside the algo sector itself, a broad range of algorithmic styles and strategies are at work, including some high frequency trading. The BSE's Shapiro said it was difficult to put a number on the proportion of algo trading that is HFT, but it is growing. "What we see is that clearly there are new players coming into the market, and the way they're trading is high-frequency oriented. We're seeing it in the growth of the order-to-trade ratio."
India traders already using the traditional set of execution algorithms are now increasingly focusing attention on more sophisticated models with better analytics to help them grapple with their transaction and execution costs, according to Dr Richard Bentley, vice president of capital markets at Progress Software. "It's not that HFT doesn't exist in India - it certainly does," he said. "But trading very small volumes for very small margins is not possible in India to the extent it is in US and European markets, as a result of India's transaction tax."
Anihrudda Pant, founder and CEO of AlgoAnalytics, a firm based in Pune which builds algorithms for the local market, agrees about the trend towards sophistication. "Situational algorithmic trading is just beginning to happen," he said. The market is still small, but it is growing.
Trading firms already operating outside India which are looking at how they can use their existing algorithmic models in the Indian market may find the arrival of the Delhi Exchange's MillenniumIT platform a potentially interesting move, according to Equinix's Orrell, because they are already using the platform in other markets, including the LSE. "So in a sense people are ready to trade on that platform already - it's certainly a system that's well-known and already written to," he said.
Inside and outside India, the consensus is that algorithmic trading is ripe for growth there. The same forces that have driven the growth and development of algorithmic trading in mature markets are afoot in India, as traders look for ways to increase liquidity, knit together fragmented markets and drive down costs. Just a little behind the rest of the world but catching up fast, India too is seeing the rise of the machines.
Algo Rules, OK SEBI's guidelines find favour with the market
What's that, regulations that the market approves of? It's not every day that market practitioners find more to praise than criticise in the work of a regulator, but as with so many matters, India proves the exception. As Mary-Ellen Barker finds, here's a case where being one step behind the curve looks more like being ahead of the curve.
A brief document issued by Indian securities market regulator SEBI (Securities and Exchange Board of India) has laid to rest many fears about the outlook for algorithmic trading in India. The four-page circular, 'Broad Guidelines on Algorithmic Trading', has been greeted with open arms and just a trace of relief by brokers, exchanges and financial market observers.
As debate rages in Europe and the US about regulation of high frequency trading, the Indian regulator has produced rules which have been hailed as cautious and sensible, encouraging the gradual development of more algorithmic trading while taking measures to head off possible catastrophes. That has soothed niggling anxieties about what the guidelines would hold. "If there was a fear in the market, it's more or less out," said one Indian broker.
Indian regulators have had what Progress Software's Dr Richard Bentley describes as 'second mover advantage'. "They get to see what's happened in other markets and avoid some of those mistakes." He approves of the new guidelines, which were released in March.
"They're putting a lot of emphasis on the brokers having adequate risk systems, having the appropriate pre-trade checks. They're requiring all order flow to go via domestic broker - and that can be an international firm with a domestic brokerage - so that they can control and ensure consistency. They do seem to be emphasising the risk management and surveillance responsibilities of market participants without going over the top."
Kotak Securities IT head Chetan Pandya agrees. "The regulator has been forward-looking in its regulations while, at the same time, being conservative on risk management requirements," he said.
Brokers and exchanges say the new guidelines don't make radical changes, but codify existing rules to make clear how they apply to algorithmic trading. Built-in safeguards already included market-wide circuit breakers, individual instrument circuit breakers and throttle controls which can be applied firm by firm.
Both the NSE and BSE say they are comfortable with the new rules. "Almost all, if not all, were already being done by us," said NSE CTO Ravi Apte. "There might be some fine-tuning here and there, but that's just fine-tuning."
"BSE is fully compliant with SEBI guidelines," said BSE adviser Jim Shapiro. "As in any market with algo trading, problem trades can occur. But India has pretty robust pre-trade risk management systems in place, as well as price limits. As a result we have not experienced many serious problems. The order-to-trade ratio at the BSE has been growing with the growth of algo trading and, in particular, our market making programs. But at the moment we do not feel they have reached levels where we have capacity concerns."
Concerns in the market before publication of the guidelines were less about the spectre of restrictions such as minimum resting times or fee structures that penalise high order-to-fill ratios, and more about a widening of exchange powers over the algorithms themselves.
Brokers feared the rules could be much more stringent, and might force firms to reveal the proprietary information within their own algorithms, said Anshuman Jaswal, Celent senior capital markets analyst. "That has been the trend internationally as well, that the regulators are looking at the algorithms themselves. So that was a worry, but it has not been the case."
Every regulator's fear is the rogue algorithm that creates market chaos. The market's verdict is that SEBI has struck a good balance, taking a slow and steady but not over-cautious approach.
"I do believe SEBI is right in thinking that we want to make sure exchange systems and risk management systems are capable of handling this kind of stuff, before you let everyone loose," said Aniruddha Pant, founder and CEO of Algo Analytics, which designs algorithms.
Summary of the SEBI guidelines:
- Stock exchanges must put mechanisms in place to manage systems load and volumes so they deliver consistent response time for all brokers
- Exchanges must continuously monitor system performance, including performance of the surveillance system, and upgrade them if necessary to keep pace with the speed and volume of data that may come as a result of algorithmic trading
- Exchanges should use 'economic disincentives' to deter high order-to-trade ratios, and monitor to block any order flooding
- Exchanges should use price bands and quantity limits to manage order-level risk control
- Exchanges should put mechanisms in place to identify dysfunctional algos and shut off their order flow
- Exchanges may seek details of trading strategies used by algos for the purposes of surveillance and investigation
- Exchanges must identify algo trades and brokers using them in monthly reports to the regulator
- Brokers need exchange permission before using algorithms; exchanges need to check that those brokers' systems conform to standards for orderly trade, and that normal annual audits of brokers cover algorithmic trading risk controls
- Brokers using algorithms need to satisfy the exchange that they have risk controls in place covering price, quantity, order value, cumulative order value, automated execution and "pre-defined parameters for an automatic stoppage in the event of algo execution leading to a loop or a runaway situation"
- Brokers must to satisfy the exchange that they have proper procedures, systems and technical capability to carry out algo trading, and safeguards to prevent algorithms from misuse or unauthorised access
- Brokers must have real-time monitoring systems to identify algorithms that may not behave as expected, and inform the exchange immediately of any incidents.
- Brokers must log trading for audit purposes, recording control parameters, orders, trades and data points of trades executed by algorithms
- Brokers must inform the exchange of modifications to approved algorithms or algorithmic systems, and gain approval.