It’s hardly news to anyone working in the global government bond markets that high-frequency traders (HFT) now account for a larger share of secondary volume in US treasuries than traditional bank primary dealers. The interdealer market has grown increasingly electronic, conducted on platforms such as BrokerTec and eSpeed, and firms like Jump and Citadel dominate with firms like KCG not far behind. Bank dealers, by contrast, have cut the amount of capital they devote to trading substantially.
However, much remains unclear in US treasuries trading, not least how much business is conducted away from the screens on a bilateral basis between dealing banks and large asset managers, how much of this is done by voice and how much by clicking and trading against live prices streamed by the dealers to select customers, who might compare several sources to comply with requirements to transact at best price. How much of this activity is a prelude to the assumption of net market risk that banks may later seek to lay off electronically on BrokerTec and eSpeed, which one banker suggests to Euromoney are merely the secondary, secondary markets? No one can even guess.
Just because HFTs do more business on electronic market places does not make them good providers of a customer market-making service to real money or leveraged asset managers. In their investigation of the flash rally in US treasuries last October, US regulators concluded that while HFTs commonly act as short-term liquidity providers, buying and selling frequently in small amounts, they rarely take significant, unhedged intraday positions and are too thinly capitalized to end the day with much net exposure.
This liquidity provision is prop trading by another name and is provided primarily on the basis of immediate profitability, rather than as a service offered in the context of existing customer relationships that are intended to be profitable over time.
Trading algorithms can calculate that, in the process of filling an order, especially one requiring multiple linked buy and sell orders in correlated markets, there are times when it makes sense to pay the bid-offer spread and times when it is a better strategy to rest passively and capture spread by leaving the bid-offer open to other takers.
Wary of taking comfort
So while HFTs might show tighter bid-offer spreads than bank dealers during the normal course of business, they are even quicker to reduce the already small sizes in which they trade during periods of volatility. And even if they do continue to display tight bid-offer spreads during turbulent conditions, other fund managers should be wary of taking comfort that they will provide an exit route. The depth of order books may prove very shallow.
Much of the debate around the role of HFTs has concentrated recently on accusations of market manipulation, stuffing orders into exchanges and electronic platforms to slow order processing, then cancelling orders and finally using a speed advantage to trade profitably in the aftermath.
Yet in all this something important may be being missed. There are signs that some non-bank dealers may be at the point of building genuine customer market-making businesses, going beyond provision of largely inter-dealer liquidity to selling a service to fund managers analogous to that which bank dealers used to provide and based not just on tight spreads but also a maintenance of firm executable prices in large size.
Euromoney reports this month on how firms such as Citadel Securities in interest rate swaps and XTX Markets in spot foreign exchange are seeking more direct links with big asset manager customers rather than just providing complementary capacity to other dealing banks.
What are banks good at? Provision of credit to enable trading, for example through prime brokerage certainly, perhaps also collateral management, which is so crucial to their own treasury departments’ funding operations and also managing the rails of payments.
But is the old-style senior trader on the bank dealing desk likely to have a better intuitive capacity for capturing the information content in shifting patterns of liquidity flow and pricing in multiple markets whose correlations with each other wax and wane than the best algos? Probably not. And can any large bank’s fractured systems compete with the newcomers on speed of updating risk exposures in real time? It’s hardly even a fair question.