What you don't know can hurt you

In the iconic television series, the Odd Couple, Oscar and Felix were concerned that there might have been a malevolent spirit inside of their air conditioner. So, they called on an occult expert, who, after examining the air conditioner, recommended an exorcism. Unfortunately, his price was too high, so they asked for instructions on how do it themselves. His reply was classic:

"We have an expression in the spirit game: What you don't know can hurt you very much..."

While funny, and good advice when dealing with evil spirits, this commentary is about trading costs. In this context, my version is:

"The orders that you don't measure, can hurt you very much..."

Simply put, most buyside firms only analyze the executions that they receive from their broker dealers, but not all of the unfilled orders sent on their behalf. These unfilled, unmeasured orders, however, provide needed insight into both opportunity cost and market impact.

A story from my own experience running a quantitative trading unit can illustrate the importance of opportunity cost. An analyst, after analyzing our trade data, concluded that we should trade more passively. After patiently reviewing the results, I asked a simple question: "How did your analysis account for the unfilled quantity on the passive orders?" The analyst looked at me uneasily, and I went on. "If you don't, then you introduce large selection bias, since you would miss the situation where trading passively resulted in missing trading opportunities." He then told me that he had not, but would do so and get back to me. I told him that, for a first pass, he could price the unexecuted quantity at the far side of the NBBO at the time those orders were cancelled, even though there might not be enough available liquidity at that price. Once he did so, the conclusion changed; when the opportunity cost was factored into the equation, aggressive orders looked more favorable on average.

This is not particularly surprising, considering that a fundamental principle of economics is that markets tend to trade in equilibrium. This means, all other things being equal, that, for on exchange trading, where the explicit costs on passive orders are roughly one half of a cent per share cheaper than aggressive orders (due to the difference between the rebate and the access fee), one would expect the gross price performance of aggressive orders to be roughly the same half cent superior to passive orders. This is important for buy side firms to be aware of, since it would indicate that brokers, who are not passing through fees to their clients, should overwhelmingly execute the orders they send to exchanges[1] aggressively. Since this is not always the case, it is important to ask why...

To answer that question, however, buy side firms need to receive all of the orders sent on their behalf. Unfortunately, many buy side OMS systems are configured to only receive filled (or partially filled) orders from brokers. This makes the job of transaction cost analysis difficult (and potentially misleading), as there is no way to know the difference between opportunity costs and timing consequence.

To illustrate, I will provide an example with several scenarios, but first, let's define some terms:

Explicit costs - the costs incurred by paying access fees or collecting rebates from market centers

Impact cost - the price move (typically measured by comparing the midpoint of the National Best Bid or Offer (NBBO) from immediately prior to placing an order to some time period after order placement.

Opportunity Cost - the price move that occurs when a trader attempts to execute an order which is not filled or a trader makes an active decision to not trade.

Timing Consequence - the price move that occurs when traders are not active in the market; in many cases, this is due to constraints (such as % of VWAP or other participation metrics) imposed upon the broker.

An example:

Fund A sends an order to purchase 100,000 shares of stock XYZ to their broker at 11am. The below facts are consistent throughout all four scenarios that follow:

• ·XYZ typically trades roughly 1 million shares per day (so the order is 10% of ADV)

  ·XYZ has an average bid asked spread of 1.2 cents

  ·XYZ arrival price was 50.11 (immediately prior to sending the order, XYZ was bid at 50.10 and offered at 50.12)

  ·1 hour post order sending, XYZ was bid at 50.19 and offered at 50.20 with a volume weighted average price for the hour of 50.16

  ·4 hours post order sending, XYZ was bid at 50.30 and offered at 50.31, and the volume weighted average price for the 4 hour period of 50.22 (and the 3 hour period VWAP after the first hour was 50.24)

  ·At the close of the day, XYZ ended at a price of 50.20 and the VWAP for the 5 hour period was 50.22

AND

·The Broker has a "natural" seller of XYZ at the time of order receipt of 25,000 shares, which gets crossed with this order at 50.11

Now consider the following scenarios:

• 1.The broker, since this was a VWAP order, waits for the rest of the first hour to trade in the market, since the client constrained the algorithm based on % of volume. They start trading with a mix of orders and execute at an average price for the balance of the 75,000 shares at 50.24, for an average price of 50.2075.

  2.The broker, attempting to trade exclusively passively, keeps bidding at or just below the market for an hour, without executing any shares until the first hour passes. They then become more aggressive and trade over the rest of the day, executing at an average price of 50.24 for the balance of the 75,000 shares for an average price of 50.2075.

  3.The broker executes the 75,000 remaining shares entirely using randomized, spaced out, aggressive orders using their Smart Order Router (SOR) over the day for a price of 50.225 for an average price of 50.19625.

  4.The broker keeps the balance of the order in their dark pool as they execute in the market with randomized, aggressive orders using their SOR. They execute the balance of the order for 50.175 in the first hour broken down as 25,000 executed in the market at 50.16 and 50,000 executed in their dark pool at 50.1825. This equates to an average price for the order of 50.15875

I would make the following observations about the trading in these scenarios:

First, when comparing scenarios 1 and 2, the outcome is precisely the same, but there is a major difference in attribution. In scenario 1, the fact that the offer price for XYZ moved 8 cents higher is timing consequence (and attributable more to client constraints than trading), while in scenario 2, the 8 cent movement is clearly opportunity cost as the broker chased the bid higher.

Second, when comparing scenario 3 to both scenarios 1 and 2, the broker performed better on both arrival price and VWAP metrics. It is also likely that opportunity cost was minimal and that timing consequence was less overall than in the other scenarios. Explicit costs to the broker, however, were likely higher by an amount similar to the outperformance due to the consistent paying of the access fee.

Third, when comparing scenario 4 to all others, it performs better on an arrival price basis, meaning it experienced lower costs, but underperformed on an interval VWAP basis (as the first hour VWAP of 50.16 is 1.5 cents below where the broker executed). It is likely to have an opportunity cost that was lower overall than scenario 2, but still around 6 cents per share. Depending on what metric was used, scenario 4 might be considered inferior or superior to the others, but in this case, with perfect hindsight it was superior. If similar results were achieved over many orders, however, despite being inferior to interval VWAP, it would be judged superior. On the other hand, if on the majority of occasions, the price after the dark pool fills moved adversely, then one could assume that the broker overused their dark pool.

Most of these observations required knowledge of all the orders placed to understand the trading strategy and draw conclusions. Knowledge of the orders is the only way to determine if adverse price movement is opportunity cost or timing. It is also the only way to measure the market impact that might be created by the trading strategy, as one can analyze the market impact directly correlated to the placement of displayed orders as well as from orders that take liquidity.

Note, however, that it is important to understand that there are different situations in the lifecycle of a trade. In the early stages of an order, controlling market impact is more important than when completing the order, for example. This is due to the fact that impacting the market early in an order directly increases the cost of trading the balance of that order. In the above scenarios, for example, had the broker placed large, displayed, passive orders early in the trade, it might have been a key reason why the price moved higher (i.e. created market impact).

Markets are complex, and it is important to understand that it is unfair to expect "best" execution for any individual order, since, at a specific point in time, there is no way to know all the unknowns, such as where there is dark liquidity. There is, however, certainly "better execution" which means a better process, potentially leveraging quantitative analysis to make key decisions. These include how break up an order to fit the market or how or when to trade a block, when to go to the visible markets (and whether to take liquidity or be passive), when to go to dark pools either passively or actively and how to do so and so on. Many, but not all, brokers invest heavily in the quantitative finance and technology required to make these decisions, and it is common sense that such firms will achieve better results than those that do not. The critical question, therefore, is how to measure those results. Doing so, however, must include the analysis of all the orders sent out, whether filled or not. That is the only way to make sure that those unfilled orders are not hurting performance.

[1] This simplistic analysis is reasonable when comparing displayed passive orders with aggressive orders on exchanges where the natural bid offer spread is similar to the 1 cent tick size. Comparing dark orders requires more complex analysis and securities whose spreads are sub-penny or much greater than a few cents also require different analysis.

S&P Global provides industry-leading data, software and technology platforms and managed services to tackle some of the most difficult challenges in financial markets. We help our customers better understand complicated markets, reduce risk, operate more efficiently and comply with financial regulation.

This article was published by S&P Global Market Intelligence and not by S&P Global Ratings, which is a separately managed division of S&P Global.