A Brief History Of Implementation Shortfall

 
Implementation Shortfall.jpg

As the world mourns the recent passing of the legendary physicist and cosmologist, Professor Stephen Hawking, and recall his well known book “A Brief History of Time”, we thought we would summarize one of the key items in the best execution universe. Here is a brief history of implementation shortfall.

First of all, what is “Implementation Shortfall”? Andre Perold coined the phrase and first proposed the concept in his seminal paper on the subject in 1988. Implementation Shortfall (IS) is defined as the difference in price between the time a portfolio manager makes an investment decision and the actual price achieved. Another component is the opportunity cost of any quantity unexecuted during the implementation. Perold’s paper discusses the differences between paper trading where the execution price is assumed to equate the current market level, and the reality of execution in the implementation of the decision and its resulting shortfall. It is of huge relevance to a portfolio manager since the implementation cost can have a substantial impact on the performance of the strategy and needs thorough consideration and ongoing review.

Timeline:

– 1968: The first paper written on transaction costs is believed to be by Harold Demsetz: “The Cost of Transaction, Quarterly Journal of Economics, February 1968.

– 1988: Andre F. Perold: “The implementation shortfall: paper vs. reality”, Journal of Portfolio Management, 14 (Spring), 4-9.

– 1991: Collins, B. and Fabozzi: “A Methodology for Measuring Transaction Costs”, Financial Analysts Journal, March / April, 27-46

– 1993: Wagner, W., and H. Edwards “Best Execution,” Financial Analyst Journal, Vol. 49, No. 1, Jan/Feb 1993, 65-71.

– 2000: Our own Robert Almgren wrote a paper with Neil Chriss: “”Optimal execution of portfolio transactions,” Journal of Risk (3) 2, 5-39.

– Early 2000s: The emergence of the first “Implementation Shortfall” execution algorithms in equities.

– 2011: Quantitative Brokers launches the first “Implementation Shortfall” execution algorithm for futures markets (Bolt).

– 2015: Quantitative Brokers launches the first “Implementation Shortfall” execution algorithm for on the run cash treasury markets (Bolt).

At QB, we don’t use “Implementation Shortfall” to describe our flagship algorithm Bolt. Since we don’t know the time the investment decision was made, or have any responsibility for the process before it arrives at QB, we can only measure our performance from the “Arrival Price”, the bid / offer mid point at the time we received the order. Therefore we describe Bolt as an Arrival Price (AP) algorithm. However, the concept is the same, the technical difference is just between the original time most relevant for a portfolio manager and the earliest time Bolt can be measured from. The name Implementation Shortfall has stuck widely in equity trading but in essence these “IS” algorithms are arrival price algorithms.

Over the years there have been proponents and detractors of this approach as a measurement of execution costs and a benchmark for execution algorithms. There is no perfect solution to the challenges of measuring implicit costs and assessing trader, broker, or algorithm performance, but overall this one is pretty good. The crucial point to compare this with other benchmarks is that this is the one most relevant to the portfolio manager. If the market’s volume weighted average price (VWAP) is used as a benchmark it is only half the story – the point that’s missed is that someone had to decide over what time frame to trade in order to compare the execution performance to the VWAP over that time. That decision (usually an arbitrary one made in an attempt to minimize market impact) is a crucial part of the execution process, but that itself escapes assessment if only VWAP slippage is measured. The other issue with VWAP as well as TWAP, market open or close prices is that they are benchmarks that can be influenced by the trade itself. Zero slippage to VWAP but being 100 percent of market volume does not necessarily mean it is a good execution. Zero slippage to the closing price while being part of an adverse high / low on the close is also not a good execution. AP does provide an overall measure that incorporates all aspects of the execution. Where VWAP can play an important role is in conjunction with the AP benchmark – looking at both gives context on market direction during the execution and how the execution compared with the market over that time.

Where some have become frustrated with IS / AP, is the challenge of understanding two aspects that make up realized implicit trading costs; 1) market drift and 2) market impact. Optimizing between these two is often referred to as “the trader’s dilemma”, i.e. you can trade faster with more impact to minimize market drift away, or trade slower to minimize market impact but risk market drift away. However, you will never know what the market would have done if you had not traded. You can try to extrapolate the price trajectory forward based on the momentum it showed in the minutes before the trade, but markets often change direction. You can apply some correction from the overall market, using the beta for the asset, but in practice that removes only a small part of the randomness. There is still the essential difficulty that you are part of the market, and other people may be choosing to trade at the same times and for the same reasons you are. Thus market impact and drift are not so different. It is hard to know what is good and what is bad arrival price slippage, but knowing it does allow for comparison across different execution methods, and if a reliable conclusion can be drawn, switching to the better method yields performance improvements for the strategy. It also enables portfolio managers to compare their cost assumptions with reality and understand their own implementation shortfall so that they may improve their models.

We believe that IS / AP has a long future ahead. We expect to see more adoption of its use, following QB’s pioneering efforts to bring it to futures and fixed income. With more data being gathered from the investment and execution processes, and improving techniques in data science, better metrics may also emerge. We are supportive of new approaches and further debate in this area. Ultimately, any improvement is in the best interests of the portfolio managers and the investors they serve.

Quantitative Brokers
New York
March 28, 2018

 
Guest User