Fixed Income Meets the Black Box

Oct 24, 2005
URL: http://www.wallstreetandtech.com/showArticle.jhtml?articleID=172900005

The universe of fixed income securities - with more than 3 million names in the U.S. alone - dwarfs the global equities market, which only has about 15,000 stocks from which to choose. As the dot-com bust has consolidated more than 100 bond-trading platforms to just a few entities with meaningful liquidity, the fixed income market would seem to be ripe for algorithmic trading, and in the fast-moving, deeply liquid interdealer market in government bonds, this certainly is the case. But it may be quite some time before algorithmic trading becomes commonplace for institutional asset managers and mutual fund managers, due to structural issues with the dealer-to-customer marketplace that result in a lack of transparency and navigability for automated trading patterns.

In the interdealer market, an active arbitrage business has developed between the two giants of electronic U.S. Treasury trading - eSpeed and Icap - and between these two platforms and the futures market. Opportunistic traders attempt to gain small profits by purchasing bonds on one platform and immediately selling them on the other, which requires lightning-fast connectivity and firm quotes - and it certainly doesn't hurt to have a computer with built-in parameters doing the work. According to David Rutter, CEO of electronic brokerage at London-based Icap, the vast majority of this trading is conducted by "less than 10" quantitative trading firms, most of which have been spun off of Chicago-based futures commission merchants (FCMs) to lay off risk against bond-based futures contracts traded on the Chicago Board of Trade (CBOT).

"Algorithmic trading is the fastest-growing customer segment that we have, and there has been a dramatic change in the last year - more than 50 percent of our bids and offers are now black-box-oriented," Rutter says. "[Black-box traders] have become a very important source of liquidity, which was traditionally the domain of dealers."

Rutter notes that relatively few transactions result from black-box postings, but they have helped reduce price discrepancies in the Treasury market. He adds that Icap now is experimenting with creating a volume weighted average price (VWAP) benchmark for its most active securities - a hallmark of the equities market and a harbinger of more algorithmic trading to come.

New York-based eSpeed also is considering creating a VWAP for "on the run" (i.e., most recently issued) Treasuries, officials say, and the broker is making preparations for more-automated trading. Over the past year and a half, the platform has been increasing its message-rate capacity so as to pave the way for active black-box customers.

"We have a group of people really dedicated to helping our clients operate programmatically with us," says Matt Claus, eSpeed's CTO. "Program trading, in our environment, is customers who have developed software that interacts with our API [application programming interface] without human intervention." Though he cannot provide specific figures, Claus estimates that algorithmic trading on eSpeed will grow tenfold in the next five years.

The attitude toward algorithmic trading, however, is much more conservative at dealer-to-customer venues such as TradeWeb, the predominant trading platform for government securities, and MarketAxess, the predominant corporate-debt trading platform. According to a spokesperson, TradeWeb is not actively pursuing algorithmic trading capacity.

For its part, MarketAxess is making improvements to its API to support faster trading, according to John Dean, the company's head of connectivity. But, he says, the algorithm bug probably will require 12 months to 18 months to take form in the fixed income dealer-to-customer marketplace, starting with Treasuries and moving to the recently launched credit default swap (CDS) index market, and finally on to the most-active corporate securities.

Critical Difference

The critical difference between the interdealer market and the dealer-to-customer market is comprised of equal parts immediacy, liquidity and transparency. In the interdealer space, quotes for common bonds, such as the 10-year Treasury note, are anonymous and available for instant execution, similar to the Nasdaq stock market. In the dealer-to-customer space, however, dealers provide quotes when requested by customers, and trading is not anonymous - customers risk showing their hand when they request a quote on the open market.

Although the dealers have automated pricing engines for this purpose, they retain the capability to refuse a trade request or change a quote before offering it again to a customer - which makes authoring an algorithm more challenging. Since most algorithms rely on a constant stream of current market data, the challenge increases the further one gets away from Treasuries.

In the corporate and municipal debt markets, trades are far less frequent and current pricing information often is not available. Also, the more obscure the security, the more likely it is that a single dealer may hold all the available liquidity. Although the National Association of Securities Dealers (NASD) has made improvements to the Trade Reporting and Compliance Engine (TRACE), corporate-bond trade information is delayed by 15 minutes, and even Treasury price information is not necessarily immediate and accurate, notes Harrell Smith, head of the securities and investments practice at Celent Communications.

"The limit-order algorithm works great in a highly transparent world, but it does not fit as cleanly into the fixed income world," Dean says. "In fixed income, you have no clue about where that bond is - at least not to the nth degree - so it is better to do a quote request and subject that to your pricing models. It is a technical hurdle, but it can be [overcome]."

There also are fewer reasons to use algorithms in fixed income trading. Whereas equities traders can benefit greatly from splitting up orders across multiple venues to avoid detection - one of the main drivers for algorithmic trading in the equities marketplace - there is little benefit to doing so in fixed income, where a trader does not necessarily pay more to conduct transactions of high value, and there are not many competing electronic venues offering the same security, according to Travis Bagley, head of fixed income transition at Russell Investment Group, which manages $34.2 billion in global fixed income securities.

"If you are holding a $1 million or $2 million piece of a bond that is tradable, you wouldn't incur as much impact as you would trying to trade that much value in equities," explains Bagley. "At this point, we see algorithmic trading as more theoretical. We expect some firms are out there using it, but it is not a large segment of fixed income trading, and we don't know what the tipping point would be to make it so."

Eric Goldberg, CEO of Portware, a software company that has developed algorithms for buy-side and sell-side firms in the equities and futures markets, believes there is a tipping point, but he is skeptical that it will come to pass. "Algorithmic trading is a natural extension once there is electronic trading with streaming prices on an open platform accessible through an API," Goldberg asserts. "But we don't have that yet in fixed income. A lot of the single-dealer and dealer-to-customer platforms are built for single order entry, and no one really wants it to be an open book."

Unless the market structure changes so that it resembles an exchange with an open book and live orders, with the possible exception of the most adventurous hedge funds and quantitative trading shops, it seems that the use of algorithms in the dealer-to-customer market will be limited to internal algorithms that try to find optimal price points against common benchmarks, such as the yield curve, according to Gavin Little-Gill, senior analyst at TowerGroup. "You will also see people looking at betting across markets," Little-Gill says. "They will play fixed income versus equity markets. They may build these complex models that culminate in some fairly rudimentary 'if-then' statements that trigger transactions."

On The Net

eSpeed
Icap
Chicago Board of Trade (CBOT)
Nasdaq
TradeWeb
MarketAxess
NASD
Celent Communications
Russell Investment Group
TowerGroup