This is especially true in the United States: a recent report by the Tabb Group suggests that the proportion of buy-side groups in the US that send their orders to brokers by telephone will fall to 20% by 2007.
When a firm partakes in algorithmic trading, its trades are sent through the most efficient quantitative computer-generated models, which automatically generate the timing and size of orders based upon goals specified by the parameters of the algorithm. DMA involves the automated and direct routing of a securities order to a venue for execution, thereby entirely bypassing any intervention by a third party.
“It is instructive to think of DMA as being the pipe, and algorithms as being the intelligence behind the pipe,” notes Marcus Hooper, an electronic trading specialist at Bear Stearns. “These strategies complement each other perfectly, in a natural symbiotic relationship, as algorithms are heavily dependent upon DMA processors.” DMA is the technical side, while the algorithms generally give the added value.
Marcus Hooper: the method by which algorithms have been sold is not ideal
DMA places higher responsibility upon the client – if he sends a large order, this could heavily impact the price of the stock, as the whole order is shown in the market. Seema Arora, head of sales and marketing, global execution services, EMEA, at JPMorgan Securities, explains that an algorithm would be intelligent enough to gently work in a large order of this kind, such that the likelihood of an adverse footprint being left in the market would be sharply reduced.
Advantages of the new
Cost is a major factor in influencing fund managers to move towards advanced electronic trading technologies. Clare Vincent-Silk, a consultant at Investit, reveals that the approximate commission on a DMA trade is 5 basis points, and 7bp for an algorithmic trade. By contrast, a traditional telephone order via a broker costs around 15bp. There is increased regulatory pressure upon buy-side dealers to show both reduction in costs and best execution in trading.
The anonymity of algorithmic trading and DMA is also attractive. For instance, should a hedge fund trading a large block of shares wish to trade without the knowledge of the sales trader or anyone else in the market knowing, these advanced electronic trading technologies allow trades to be sent down under the broker’s ID. This anonymity prevents information leakage, one of the major drawbacks that large buy-side firms have complained of when passing their trades through brokers.
Buy-side scepticism
Given this attractiveness, why then is there buy-side scepticism surrounding algorithms? One answer may be that, in using algorithms and DMA, some of the risk is transferred from the broker-dealer to the buy-side dealer, leaving the latter to make the difficult choices of which exchange, which strategy and which algorithms to use. The sophistication and specialist knowledge this requires may be lacking among certain buy-side dealers.
Bear Stearns’ Mr Hooper argues that the way in which algorithms have been sold to the buy-side has not been ideal. “Algorithms were initially perceived as all-encompassing instruments, something that they patently are not at the present time, and this realisation has inevitably led to some buy-side scepticism,” he explains. “For example, the vast majority of algorithms are built around simple benchmarks for execution such as VWAP and TWAP, whereas many asset managers do not wish to use these benchmarks. There is a lacuna between what the technology is capable of delivering, and the needs of the buy-side business.”
The advantages that even simple algorithmic trading tools can deliver have been rather obscured by this miscommunication. If a trader receives a very large order to buy certain stock, he can use an algorithm quietly to consume liquidity over the three to five days that it will take to conduct the execution, thereby in all probability adding value to his trading process. As long as the algorithm does not impact the market and create an adverse price movement or momentum, it can be used alongside the active trading strategy to reduce the time horizon for the completion of the order.
Advanced technology of this type is therefore not being used as a substitute for human experience. Traders use algorithmic machines as a guide, along with historical profiles and background data. “My suspicion is that, in the future, a very large proportion of orders will have something to do with algorithms,” says Mr Hooper. “For the difficult trades, algorithms are not the entire solution and will not replace the traders – instead they work alongside the knowledge and experience of the trader.”
Buy-side vs sell-side?
An increasing number of investment banks have been supplying advanced electronic trading tools to buy-side clients, among them JPMorgan, Credit Suisse First Boston and UBS. JPMorgan’s Ms Arora says, in June 2005, the bank closed a deal to buy the privately-held trading technology firm Neovest, with an initial roll-out in the US. “We plan to work on the product infrastructure of the product, and then to roll it out at the end of the year in Europe, and in Asia in 2006,” she says.
Several contentious issues have arisen between buy-side and sell-side. There has been a perception in certain quarters that sell-side vendors are holding on to the best algorithms for their own proprietary desks, rather than making them available to buy-side clients, a perception that Ms Arora rejects. “Those algorithms that we test and use, we naturally give out to our clients. There would simply be no advantage to us in doing otherwise,” she points out.
Ms Vincent-Silk of Investit notes that some consternation has been caused among sell-side players by the actions of buy-side vendors in charging the sell-side for embedding sell-side algorithms into their software.
“This not only irks the sell-side, but means that players on the buy-side all use the same algorithms,” she says, suggesting as a solution the purchase of algorithmic software from companies such as Apama, which can be tailored to the needs of individual fund managers.
New frontiers
The disparity between take-up of advanced electronic tools in Europe and the US owes much to differing market conditions. In the US, the faster uptake of algorithmic strategies follows from the greater direct buy-side interaction in North American trading culture than in its European counterpart. Furthermore, there is greater fragmentation in the way that the markets are structured in the US than in Europe, so mechanisms such as algorithms that constitute an automatic tool for liquidity consolidation will add value.
The problems in Europe may be even more fundamental. Research by Investit has shown that 43% of UK-based fund management companies surveyed do not even have any FIX capability. Thus, for some on the buy-side, implementing FIX may be a greater priority than getting to grips with new and complicated trading tools. This said, DMA and algorithmic trading are filtering over to Europe, primarily at the moment to large companies that have a number of offices scattered across the globe.
Another challenge in the coming years will be to expand DMA or algorithmic trading beyond their initial equity market, and there is already quite an active DMA offering in the futures market. As fixed income is not an order-driven market, it is unlikely that DMA or algorithmic trading will develop in this space.
However, in order to advance electronic trading, the immediate focus should be upon transforming the trading algorithms from high-capacity statistical engines – that nevertheless consume liquidity well and closely match benchmarks – into true alpha-generating processes. This will involve building more of a predictive model within the algorithm.
