New kid on the street: CNQ -the Canadian Trading and Quotation System
Jeff Sanford
From the April 10-23, 2006 issue of Canadian Business magazine
The old trading floor of the Toronto Stock Exchange still exists behind a pink-and-grey limestone facade at the very heart of the financial district. The site of millions of transactions that have shaped the economic development of the nation, the floor is lorded over, appropriately, by six huge murals by artist Charles Comfort that depict the original six Canadian industries: oil, forestry, mining, smelting, farming, and pulp and paper. Since the TSX went completely electronic in 1997, the old floor has sat abandoned, save for the odd special event. But there was a whiff of nostalgia in the air one recent March evening--if not a touch of irony--when members of Toronto's trading community gathered at 234 Bay St. to listen to an announcement about the arrival of a new stock exchange in Canada--a startup that might alter the future of the TSX itself.
As the traders took their seats, John MacNaughton, former head of the Canadian Pension Plan Investment Board, strode to the podium. He was named chairman of the Canadian Trading and Quotation System Inc., or CNQ, just a week earlier, and was appearing that night to unveil Pure Trading, the first open "central limit order book" market system to arrive in Canada since the adoption of rules in 2000 that allow so-called alternative trading systems. With the new all-electronic service, traders will be able to buy and sell TSX-listed stocks using existing ticker symbols, just like the TSX, save for one big difference: lower fees. According to MacNaughton, the CNQ will charge, on average, 80% less than the TSX, which rakes in $100 million a year in fees to Bay Street.
The new kid on the Street arrives at a delicate time for the TSX. Canada's incumbent stock exchange is in a tough competitive environment. International competition is flowing over the border--the NYSE and the London Stock Exchange are wooing listings here--and a wave of consolidation is set to sweep the industry. The simple fact is, there are more exchanges than necessary these days and the smart money is betting that a shakeout is at hand. News of exchange mergers, bids and acquisitions has been bubbling up almost weekly. The TSX says that its goal is to expand its business to preserve a made-in-Canada exchange group that can compete regionally with the impending emergence of supranationals. But that's going to be harder to do with a domestic competitor set to siphon off market share and revenue. Can the TSX survive a war on the domestic and global fronts? The CNQ plans to cater to the new breed of algorithmic trader that is rapidly populating Bay Street. It has spawned an expensive technological arms race with the TSX. The fight is on.
Prior to the CNQ presentation, Ian Bandeen, co-founder and CEO of the company, gives two reporters a rundown. Bandeen is brash, opinionated and seems stimulated by the prospect of the coming tussle with the TSX. The secret weapon will be CNQ's super-fast order gateway and trading engine--the two key pieces of technology an exchange needs to route orders to the electronic "floor" and match trades. The strongest plank in the CNQ platform is this tech, which is designed to appeal to the new breed of trader on Bay Street.
The past few years have seen the rise of so-called algorithmic traders--"algo" traders for short--on both Bay and Wall streets. These are traders who use computer programs specially designed to help big institutions reduce transaction costs. One of the biggest costs institutions face when trading is market impact--the effect a big order has on a stock's price. Consider what happens if a pension fund were to send a buy order for 100,000 shares. The price of the stock would rise as news of the order spreads and other traders begin to trade in front to benefit from the increase in demand. There have been attempts to deal with that problem in the past--by setting up so-called dark-liquidity crossing services, for instance, where big orders are matched anonymously without anyone seeing the central order book--but they've never really taken off. Algorithmic trading has arrived as a new way to tackle impact costs.
Program trading works by slicing up the order into small pieces, reading liquidity conditions on the exchange, and then executing hundreds of orders in milliseconds at exactly the right time. "By taking a large block of stock and then breaking it up into smaller pieces, you can hide the intention of the seller or buyer and anonymously and efficiently filter that order into the market to minimize market impact costs," says Kyle Zasky, president of EdgeTrade Inc., a manufacturer of trading algorithms. "Ultimately, it's a very efficient way to go about your business if you're an institution."
Obviously, chopping up an order of 100,000 shares into blocks of 200 would take a human trader an uneconomical amount of time. But an algorithm can do it quickly, and the adoption of programmed trading is actually changing markets themselves by increasing the speed at which they run. "It's an incredible revolution," says Zasky. "We're not in an environment where someone calls a broker, tells them what they want, and then waits around for a few minutes to hear if it was executed. We're dealing with such small time frames that it is beyond a human's capability to react in a timely manner."
The longer-term prospect is of a market populated with algorithms programmed to seek out and exploit trading opportunities before humans even know what's going on. Consider a simple arbitrage play between two prices--an easy example would be Nortel, which trades in New York and Toronto. If a broker sees Nortel trading a few cents cheaper in Toronto than in New York (taking exchange rates into account, of course), then it's profitable to buy a bunch of Nortel on the TSX and sell them on the NYSE. Soon enough, the new demand in Toronto will raise the price there while the new supply in New York will decrease the price, and the two prices--which, theoretically, should be the same--come back into equilibrium. A human might have done that in the past, but with computers programmed to constantly track and compare thousands of data points, such arbitrage opportunities are picked up and disappear almost before a human notices. "The responsiveness of markets has definitely increased over the last 10 years, and that's going to continue," says Zasky. "We're in the millisecond world."
There are intriguing possibilities even further out on the horizon. Algorithms that concentrate on a whole portfolio, rather than a single stock, are on the way, while algos that "read" news are also in the pipeline. (Don't even bother trying to play that missed earnings target through your discount broker--it's going to be played out by the time you finish reading the press release.) In fact, the future has arrived south of the border. At the New York Stock Exchange, there have been weeks in the past year that program trading has comprised 60% to 70% of daily volume. Here in Canada, algo trading, which made up less than a percent of volume just a few years ago, now accounts for 15% of daily trading on the TSX and will likely go to 30% within three years. Says Bandeen: "Just look at the fact that you've seen message traffic go up 700% in the last two years and the number of trades has gone up 300%. In '04, you saw an average of 160,000 trades a day, in '05 you saw 220,000 a day and by the end of last year you saw 280,000 trades per day. The algos are arriving."
That has set up a race between the CNQ and the TSX. After all, program traders live and die on super-low latency periods (the wait time for a trade to be executed) and super-high throughput (the number of messages the gateway can handle). Those two stats comprise the new Holy Grail of the exchange business. Provide a millisecond advantage in transaction time and your exchange has an advantage, which is exactly how CNQ hopes to make its mark. Bandeen claims the TSX's current latency is a fat and flabby 100 to 300 milliseconds. "Until now there's been a gap in the Canadian market," says Bandeen. "That's not acceptable to foreign algo traders."
The high latency on the TSX is, to some extent, a result of the exchange going electronic early. It was one of the first back in 1977 to offer electronic trading, and was an industry pioneer again when it went completely electronic in 1997. But at the time, there was no off-the-shelf software available, and the exchange had to build a proprietary software gateway called STAMP; it is still in use here but only used by the TSX. "STAMP is cute, but it's kind of slow, and is the kiss of death to foreigners," claims Bandeen. The Pure Trading platform, on the other hand, combines X-Stream, built by the OMX Group of Scandinavia, with three gateways--one of which is FIX, the new global standard. Combining the two should result in latency times below 20 milliseconds, says Bandeen. "At the end of the day, an exchange is a commodity," he adds. "It's a platform, a utility. I'm not saying there's any brilliant intellectual paradigm or construct to what we're doing here. This is simply making trading faster and cheaper by a factor and giving that to the client."
The TSX is not taking any of this lying down. The exchange has been vigorously planning a defence, which it expects to launch this summer (the CNQ launches in July). The TSX recently put on a half-day seminar on algo trading and has been updating its technology to get latency times down, says Rik Parkhill, Toronto-based president of TSX Markets, the arm of the exchange that overlooks trading services and products. A new FIX gateway, which would run alongside of STAMP, is in testing, while capacity on the trading engine has been doubled to handle the increase in messages sent by program traders. The TSX is also dedicating more of its processing power to the most heavily traded stocks and will move its data feeds off the trading engine to free up capacity. All of these improvements will be wrapped up into a new service called TSXpress, which Parkhill says will reduce response times by 50%. "Once the TSXpress improvements are done," he adds, "I think we'll have response times that are competitive with anything in North America." |
