The Gateway to Algorithmic and Automated Trading

Morgan's Mission

Published in Automated Trader Magazine Issue 09 Q2 2008

As one of the world’s largest fund managers, JPMorgan Asset Management’s trading operations increasingly resemble those of an agency broker. Daemon Bear, Head of Equity Trading, and Kristian West, Head of Trading Strategy, who recently joined from Barclays Capital, explain the tools and processes employed by the firm in pursuit of best execution.

“Communication with portfolio managers is constant and will take place on a trade by trade basis.”

"Communication with portfolio managers is constant and will take place on a trade by trade basis."

How would you describe the structure and responsibilities of JPMorgan Asset Management's equity trading desk in London?

Bear: We support a wide range of fund products that vary according to investment style (e.g. quant-driven or optimised), client (retail and institutional) and geography. We have to be multi-faceted in our approach to trading because we're servicing many clients on a daily basis.

Kristian and I are responsible for the equity trading team, including equity derivatives, which are separate from the fixed income and foreign exchange desks. In London, our team of 17 traders have trading responsibilities split by geography and speciality, e.g. single stock, list, portfolio, including a dedicated liquidity desk. The London desk trades from 7am in the morning until 9pm or 10pm at night, but we also trade from desks in New York, Tokyo, Hong Kong and Taiwan. We're constantly reviewing this structure and the opportunities for creating centres of excellence.

What are the key priorities that define JPMorgan Asset Management's approach to trade execution?

West: Our objective is to identify the best means of execution for every order. We use TCA (transaction cost analysis) on a daily basis to monitor our performance and to determine how we make trading decisions going forward. We'd like to automate the decision-making process but first need to settle on common parameters. For example, if different traders are using the same execution strategy, but with different parameters, the TCA's fairly meaningless if you can only view by strategy. However, if you have a common strategy whereby a certain profile of order is executed in a certain way, there is meaning in those numbers.

Bear: One of the key priorities is to understand the clients' needs. Our client is the portfolio manager so we need to understand the strategy he's following and the benchmark he's managing against before we apply our principles to the transaction. Communication with portfolio managers is constant and will take place on a trade by trade basis. There is quite a split between the retail and institutional products so it pays to get a feel of exactly where the fund manager's coming from. Rather than putting a precise figure on his expected return on a trade, the portfolio manager will give us an indication of what he expects the short-term alpha to be. This will allow us to take a view on the urgency of the transaction.

It's a two-way dialogue. If the portfolio manager has a good understanding of how a particular order can be executed, it can help him make decisions on portfolio structure, e.g. on stock selection or weighting. We're very keen for them to be aware of the potential estimated trade costs of the positions they're putting on.

What is the range of different execution strategies you might apply to trades required by different portfolio managers?

West: Execution depends on the access channels available to us, the tools we use internally and the tools available through our relationships. As well as direct market access, we might deploy an execution algorithm or strategy supplied by a third party, which might access dark liquidity pools at the other end, or visible liquidity on exchange. We have access to brokers' portfolio desks and single stock traders that can commit capital or balance sheet to trades. Overall, we probably have more avenues to execution than many of our sell-side counterparts. Making use of these channels in a systematic way is increasingly important to us, while also maintaining a relationship with all business areas.

"Overall, we probably have more avenues to execution than many of our sell-side counterparts. Making use of these channels in a systematic way is increasingly important to us …"

Bear: The tools are there to enable us to improve our performance. We don't use them as box-ticking exercises for marketing purposes. We couldn't tell you what percentage of trades is executed via algorithms and we don't consider it a number we need to monitor. The important numbers are the absolute cost of the trade and the cost against our chosen benchmark. In terms of continually improving our decision making between execution strategies, we look at our historical performance as well as the estimated cost of doing a transaction through different methods.

How do you identify opportunities to reduce transaction costs?

West: All our 17 traders have different roles, so we look at performance individually and as a team. We also look at the team and individual performance of the portfolio management group from a trading perspective. We take on board factors such as trading styles, market cap, region, etc., to see if there are patterns that might change how we trade. We'll pick out individual transactions to try to identify why a trade performed in a certain way. Was it a one-off? Was the market trending away from us? Should we have executed differently? Any changes are then fed back into the execution strategy for the team.

What role does TCA and other data play in the execution management process?

Bear: Although we supply a certain data set to clients so that they can analyse and compare our execution performance (JPMorgan Asset Management subscribes to ITG Plexus and itero from GSCS Information Services), we've also extended and evolved our internal analytics. We have attempted to create an internal benchmark, a kind of synthetic P&L, attributable to the trading desk. We use an implementation shortfall benchmark for the desk as a whole which helps create the touch and feel of an agency-type trading desk and encourages a sense of responsibility for performance. However, the value of implementation shortfall is limited if used in isolation, so we apply two estimated trade costs to each transaction, then we compare performance to those estimated trade costs. The difference between the two estimates is basically the volume profile of the data set.

In recent months, the data we've been using (on transaction costs) has not been as reliable as we would like, perhaps because of the recent reporting changes resulting from MiFID. This is very much a work-in-progress and we will continue to evolve our approach. We still have the ability to analyse estimated trade costs, but we're increasingly using a very evidence-based approach that relies on previous trading performance data. We're the only traders working for our particular fund managers, so our own data sets are as reliable and as relevant to our decisions going forward as any that can be supplied by a third party.

West: With some benchmarks, there's a misconception that market impact is the main or only factor when assessing execution performance. But of course the alpha in a trade has to be considered, as does market movement. The time at which we execute a transaction can be of equal importance to how we execute. We also have to be aware of how changes in market behaviour impact execution preferences. One month might be particularly good for using algos, another might require more capital commitment, so part of our analysis is directed at making the right decision at the right time. There is always going to be a human element to how tools are utilised, but we're moving toward a more systematic approach.

Going forward, what we really want to analyse is footprint, looking at each transaction's profile so we can see how an execution strategy has impacted that security on an execution basis. The onus is on us to understand a broker's strategy inside out, so that we understand the execution quality performance of a trade by analysing why it did what it did from a TCA perspective.

"Rather than juggling 50 orders, we would much rather see our traders with ten orders on their pad to which they can give all of their time."

How do you use internal data to select and monitor the performance of brokers?

West: We have a formal quarterly review process for which we analyse strategy and broker performance. The process is fully open because we feel it's only fair that our brokerage relationships understand how we view them. We judge each broker's performance against standard criteria, including algo performance, liquidity provision and access to balance sheet. We also analyse how they handle our flow, i.e. if they make us a price in X, how do they handle that risk? As such, information is exchanged on both sides.

Bear: We do view our relationships holistically. It's not strictly a hard-and-fast, black-and-white numbers game; subjective factors are considered. Equally, it's important for us to know how we're performing via the different channels. Our risk profile with the brokers is quite key because it allows us to make decisions around the strategies we execute. Given the level of sensitivity in the current market environment, the efficiency with which we use our accessible balance sheet is vital. We need to know how our risk profile is received in the brokerage world.

Our broker evaluation process was in place pre-MiFID and has been fine-tuned over time. We've always felt the need to demonstrate our views on our brokers' execution capabilities, so the fact that they've had to outline best execution policies doesn't require us to change our approach greatly. We consider many of the parameters and guidelines in the directive as business as usual.

What changes have you made to trading operations to ensure you capture liquidity in the European equity markets post-MiFID?

West: Obviously, access to an increasingly fragmented market is very important to us. Part of the broker review process involves understanding the tools available to us. We need to be able to use the brokers' suite of products, but we also need to overlay our own suite of tools to provide additional access to liquidity because there are some liquidity pools that are not accessible to the sell side.

There's a huge divergence in the way buy-side trading desks are applying technology to their execution needs. I still think there's an appetite to tick boxes and demonstrate to your clients that you're using all of the tools out there. We need to ensure we're embracing technology in the areas that are going to suit our business. We don't necessarily want to be the first people with an all-singing, all-dancing execution process, but we do we want to be pioneers in adopting certain technologies. For example, smart order routing (SOR) until now has been owned by the sell side, but it's equally a buy-side prerogative to have those kind of tools. We want to work with our brokers in developing technology-based tools, but there will be areas in which we want to develop our own.

What technologies are you using to access liquidity, both now and in the future?

Bear: We have an internally-developed order management system which acts purely as an order processing tool. We don't intend to give it any front-end functionality. It's a lightweight processing tool that allows us to utilise other front-end tools in the market. We use a third-party execution management system from Portware in which we house all our front-end functionality, including pre-trade, real-time and post-trade analytics, all market access functionality including dark pools and ECNs, and all algorithmic trading capabilities.

By evolving our order routing capability, we can utilise our capacity more effectively. If the rule of thumb says the last 10 per cent of your order takes up 90 per cent of your time, we want to give traders as much time as possible to concentrate on that last 10 per cent of their overall book. Rather than juggling 50 orders, we would much rather see our traders with ten orders on their pad to which they can give all of their time. Yes, we're increasingly automated, but we're not slicing order flow strictly into high- and low-touch baskets, so that only low-touch orders go to venue X, for example, and for high-touch orders it a case of, "Well, you're on your own, fellas". That couldn't be further from the truth. There's actually a bridging of that divide in which some trading strategies are as useful for high-touch orders as for low-touch volume.

How do you choose what flow is executed via algorithms and how do you select algorithms from brokers?

West: For managing large volumes of orders, using algorithms is the natural choice. Within any list, there will be some securities that we regard as high touch, so we'll try to repackage a list to identify orders that should be traded via high-touch methods and those that should be traded in a low-touch, algorithmic way. Within that, we'll decide which orders should be traded via a certain strategy, based on factors such as market cap, volatility, liquidity etc.

Bear: Being identified as high touch does not necessarily preclude an order from being executed via certain types of trading strategy. We may feel that an application for capital to start a high-touch order off isn't relevant to the overall strategy for that transaction, so we might decide to send it down an algorithm and allow it to tick away to get a footprint in the market, prior to putting the overall strategy in place. All our traders have access to all the tools and strategies, regardless of whether they are trading single stocks or lists on the liquidity desk.

West: Choice of strategy can also require dialogue with portfolio managers. As well as knowing the alpha they're expecting from an order, it's useful to understand any other objectives. An order for two times average daily volume might need to be done straight away or at our discretion, so that will obviously determine the execution channel we chose. A large order might be packaged up as a list with a defined end time or it could be handled by a strategy that executes over a number of days. The important thing is that we follow a systematic approach to identifying the optimum strategy to meet our clients' needs.

What are the key differences between broker-supplied algorithms?

Bear: At this point, our views are based purely on trading experience rather than any detailed analysis of the performance of one strategy against another. Having said that, flexibility is important. Every broker has a passive VWAP algo, but the flexibility of the parameters is quite key to usage by the individual trader. Overall range is also important. We're now deploying third-generation algos on some desks, with traders looking at optimised programmes that send waves of orders out in an optimised fashion. On other desks, chipping away using a preferred VWAP strategy works very well.

"There is always going to be a human element to how tools are utilised, but we're moving toward a more
systematic approach."

There's an education element to all of this. With 17 traders, the speed at which they become comfortable with different algos will differ, not least because they're working on different orders. A particular trader might have a far greater ability to enhance his trading process via use of a specific algo than a colleague who may find the same strategy adds little value. As long as that choice is based on evidential experience we're happy for traders to select their own algos.

West: Comparing the performance of algos from broker A and broker B is very difficult unless the parameters are identical. And often they're not. Even though you're putting in the same parameters they may actually mean different things with different providers. How we use the flexibility built into the algos - and the ability to overlay different strategies - is really the value-add for us. We might execute an order using three different strategies because that's what we see as the most effective approach for that order. Increasingly, we're beginning to see brokers adopting a layered approach to some of their strategies, rather than us slicing an order up ourselves.

Is information leakage a concern and how do you deal with it?

West: I think it's a natural concern. Taking ownership and giving out less information naturally makes us feel more comfortable. The quality of tools available to us becomes more important as does our ability to overlay and combine those tools. Simply taking ownership of how we interact with the market is very important in that respect. Having said that, we like to work with brokers where providing information enhances performance.

Bear: Once we assume control and ownership responsibility, any decision has to be attributed to us. We monitor the possibility of information leakage both systematically via next-day TCA and by market observation. If there is information leakage, it's our fault or our responsibility. It is the individual's decision to expose or increase the risk of information leakage, so we constantly speak to the team regarding execution performance and how we're being treated by the brokerage world. Our first port of call will be to the trader who has made that decision to increase the risk of information leakage.

What are the key elements to improving JPMorgan Asset Management's performance in the future?

Bear: The buy side has been offered a flood of tools and technological capabilities which have been embraced over the last three to five years. In Q3-Q4 2006, we clearly identified that we have the tools and the liquidity channels that most agency desks have. What we didn't have was the final piece that made all those things click. Having someone with Kristian's experience of using those tools in an agency environment from a hard P&L perspective has been invaluable. He's the final cog that makes all the wheels turn and has added immense value to our execution performance in a matter of months.

Our use of technology and our ability to build new tools that fit our process and our business model will help us to continually reduce our market impact and cost of execution. To serve clients as effectively as possible, we must both understand how third-party tools can be used to our best advantage and build our own systems that leverage that technology to create the most efficient trading platform.