Huw Gronow, Director, Equities Trading, and Mark Nebelung, Managing Director of Principal Global Investors, make the case that TCA should be part of pre-, during, and post-trade analysis.
Transaction Cost Analysis (TCA) has evolved significantly with the advent of technology in trading, and thus the ability to capture incrementally higher quality data. Historically the preserve of compliance departments was to examine explicit costs only as a way of governing portfolio turnover; this evolution provides institutional asset managers with several opportunities: the ability to quantitatively assess the value of the trading desk, the tools to form implementation strategies to improve prioritisation to reduce trading costs, and therefore improve alpha returns to portfolios.
Cost analysis models, methods and techniques have blossomed in the environment, propagated not only by technological advancements, but also in the explosion of data available in modern computerised equity trading.
The benefits of applying cost analysis to the execution function are manifold. It empowers the traders to make informed decisions on strategy choice, risk transfer, urgency of execution and ultimately to manage the optimisation of predicted market impact and opportunity costs.
Although maturing, the TCA industry still has some way to go to fully evolve, and that is largely a function of a characteristically dynamic market environment and non-standardised reporting of trades and market data (the so-called “consolidated tape” issue). Moreover, with the advent and increase in ultra-low latency high-frequency short term alpha market participants (“HFT”), which now account for the majority of trading activity in US exchanges and who dominate the market, the exponential increase in orders being withdrawn before execution (with ratios of cancelled to executed trades regularly as high as 75:1) means that there must be an implied effect on market impact which is as yet unquantified, yet empirically must be real. Finally, fragmentation of equity markets, both in the US and Europe, provide a real and new challenge in terms of true price discovery and this must also by extension be reflected in the post-trade arena.
Nevertheless, waiting for the imperfections and inefficiencies in market data to be ironed out (and they will surely be in time, whether by the industry or by regulatory intervention) means the opportunity to control trading costs is wasted. You cannot manage what you don’t measure. Therefore, with the practitioner’s understanding allied to sound analytical principles, it is very straightforward, while avoiding the usual statistical traps of unsound inferences and false positives/negatives, to progress from an anecdotal approach to a more evidence-based process very quickly.
On the trading desk, the ability to leap forward from being a clerical adjunct of the investment process to presenting empirical evidence of implementation cost control and therefore trading strategy enhancement is presented through this new avalanche of post trade data, which of course then becomes tomorrow’s pre-trade data. The benefit of being able to enrich one’s analysis through a systematic and consistent harvest of one’s own trading data through FIX tags is well documented. The head of trading then arrives at a straight choice: is this data and its analysis solely the preserve of the execution function, or can the investment process, as a whole, benefit from extending its usage? We aim to demonstrate that both execution and portfolio construction functions can reap significant dividends in terms of enhanced performance.
Portfolio managers’ involvement in transaction cost analysis tends to be a post-trade affair at many firms, on a quarterly or perhaps monthly basis, that inspires about as much excitement as a trip to the dentist. It may be viewed as purely an execution or trading issue and independent of the investment decision making process. However, there is one key reason why portfolio managers should care about transaction costs: improved portfolio performance. The retort might be that this is the traders’ area of expertise coupled with a feeling of helplessness on how they could possibly factor transaction costs in. The answer lies in including pre-trade transaction costs estimates to adjust (reduce) your expected alpha signal with some reasonable estimate of implementation costs. You can now make investment decisions based on realisable expected alphas rather than purely theoretical ones.
A key characteristic of many investment processes that make some use of a quantitative alpha signal process is that you always have more stocks (on a stock count basis) in the small and micro-cap end of the investable universe. There are simply more stocks that rank well. This is also the same part of the universe where liquidity is the lowest and implementation shortfall is the highest. If you don’t properly penalise the alpha signals with some form of estimated transaction cost, your realized alpha can be more than eroded by the implementation costs.
Proving the Point
To illustrate the impact of including transaction cost estimates in the pre-trade portfolio construction decision making process, consider the following two simulations. Both are based on exactly the same starting portfolio, alpha signals and portfolio construction constraints. The only difference is that in the TCs Reflected simulation, transaction costs were included as a penalty to alpha in the optimisation objective function whereas in the TCs Ignored simulation, pre-trade transaction cost estimates were ignored. The simulations were for a Global Growth strategy using MSCI World Growth as the benchmark, running from January 1999 through the end of June 2012 (13.5 years) with weekly rebalancing. They were based on purely objective (quantitative) alpha signals and portfolio construction (optimisation) with no judgment overlay. Transaction cost estimates were based on ITG’s ACE Neutral transaction cost model. Starting AUM was $150 million. Post-transaction cost returns reflect the impact of the transaction cost estimates for each trade.
Despite relatively conservative assumptions relating to strategy size ($150 million poses relatively few liquidity constraints) and transaction cost model (ACE Neutral is a relatively passive cost model with lower cost estimates than a more aggressive trading strategy), the portfolio reflecting transaction costs as part of the pre-trade portfolio construction outperformed the one where they weren’t by 0.86% per annum. Figure 1 illustrates the cumulative growth of $1 between the two portfolios.
At the end of the time period, the TCs Reflected portfolio had grown to $2.94 vs. $2.63 for the TCs Ignored portfolio, an additional 30% return on initial capital. The turnover of the TCs Reflected portfolio was modestly higher, averaging 69% p.a., compared to 67% p.a. for the TCs Ignored portfolio.
Annualised transaction costs for the TCs Reflected portfolio was slightly higher at 0.64% vs. 0.62% for the TCs Ignored portfolio. Tracking error and volatility of the two portfolios is very similar. The net effect of higher excess returns (after transaction costs) and similar risk profile (tracking error) was a 34% improvement in the information ratio when transaction costs were reflected as part of the portfolio construction.
It’s hard to think of many (any?) portfolio managers that wouldn’t seize an opportunity to add an additional 0.86% per annum in excess return. Transaction cost estimates will materially alter the most attractive stocks to add to a portfolio at a given point in time and the cumulative impact on performance is significant. In order to maximise realised portfolio performance, portfolio managers need to reflect some form of implementation cost-adjusted alpha signals such that the expected returns of illiquid stocks are appropriately adjusted for expected costs of buying or selling them in current market conditions.
In addition to portfolio performance improvements, portfolio managers considering pre-trade implementation cost estimates have a better basis to judge whether to reconsider a transaction if current market implementation costs are deviating significantly from the initial estimates. By having a common understanding of implementation costs between both portfolio managers and traders, communication is enhanced pre-, during and post-trade. Where the trading function was previously simply a transaction execution function, it now becomes part of the integrated investment decision making process.