Incredible Story Of Transaction Cost Measurement

On ly

The Incredible Story of Transaction Cost Management: A Personal Recollection

I

ft F

Dr a

or

th Au SUMMER 2008

IIJ-JOT-WAGNER.indd 1

We are building a 2,300-stock portfolio including a very large number of industry representatives to create a well tracking portfolio without holding all 5,000-plus securities in the index. I’m sitting on the trading desk, phoning the brokers with lists of company names which we’re trading. (Remember, it’s 1984; the FAX machine is high-tech.) I come across a company, quite famous at the time, even to me, named Green Tree Acceptance. The company financed manufactured housing, later called mobile homes, but then known as “trailers.” Clearly, whoever was in charge of assigning industry codes didn’t know that, for he had classified Green Tree as a Forestry company. (Green Tree = Forestry; makes sense, right?) I am dumbstruck. Not by the fact that Green Tree was misclassified, but by the fact that it didn’t matter that I knew nothing— absolutely nothing—about the thousands of companies I was buying. Not only that, I was buying on the basis of erroneous information! How can it be that demonstrable dummies, such as me, can sell index funds on the basis of highly competitive performance? I knew from previous measurement at Wells Fargo and Wilshire that almost all analyses of quantitative selection models or financial analysts’ picks showed value in the decisions, usually in the range of 50bp. Okay, so analysts’ models add 50bp, but performance records show portfolio managers underperform

vi

t’s 1984. The summer Olympics are in Los Angeles, the winter Olympics in Sarajevo. Wayne Gretzky Is in his prime as the Oilers win the Stanley Cup; the Celtics and the Lakers stand out in the NBA finals. Roger Clemens is a rookie. The best news from 1984 is what didn’t happen: 1984 doesn’t look anywhere near as grim George Orwell’s apocalyptic vision. Personally, I am very busy this day, April 24, in 1984.1 Wilshire Asset Management is putting into place a $1.3 billion Wilshire 5000 Index Fund for the Minnesota State Board of Investment. We had convinced the SBI that indexing could compete with active management performance, that indexing technology had progressed to the point where it was feasible to mirror the Wilshire 5000 Index in a real portfolio, and that Wilshire could do it best due to extensive work in indexing. We had tutored 40 organizations in running index funds with our software, but we had never actually run any portfolio ourselves. We were anxious to demonstrate our proficiency to the SBI. One of our best cards is to show innovative trading savvy to keep trading costs low. We devise several trading techniques to disguise our footprints and reduce impact, but those techniques are not the subject here. This narrative is about how the hidden costs of trading were discovered.

or

is a founder of Plexus Group, Inc., in Venice, CA. [email protected]

Re

WAYNE H. WAGNER

ew

WAYNE H. WAGNER

THE JOURNAL OF T RADING

1

5/20/08 5:22:30 PM

or

Dr a

ft F

or

Re

vi

ew

THE INCREDIBLE DISAPPEARING ACT OF HAROLD DEMSETZ

Before going forward, let’s trace back a little bit further. How many readers are familiar with the name Harold Demsetz? When I asked that question to my Montréal audience, I wasn’t surprised to receive blank looks. (Harold wouldn’t be surprised either!) Harold might not be renowned as the father of Best Execution as a discipline rather than a legal concept, but he was first to measure it. The quiet birth of what we now know as Transaction Cost Measurement started in 1968 with an article by Harold Demsetz.2 Considering the paucity of available data and the general lack of interest in the topic, Harold’s work was very illuminating. I have a personal story concerning Harold. My wife and I were attending a cocktail party in the early 1990s at the home of a UCLA professor. I walked up and introduced myself to a stranger—and it’s Harold Demsetz! I enthused about this seminal work which had so strongly inf luenced my career. With a wry look on his face, he said, “As far as I know, you’re the only person who ever read it.”

transaction costs associated with brokers protecting themselves against knowledgeable traders. (Which, as demonstrated above, we clearly weren’t.) In a world of fixed commissions and only active managers, how could we get around the problem of incurring active trading costs in a heretofore inconceivable non-active portfolio? We devised some passive trading techniques that we called teddy-bear trading to turn our indifference to individual stocks selections into an advantage. We wrote this up in 1975 for the Financial Analysts Journal, which I believe was the second article on transaction costs published.3 We went on to other things, while progress was made elsewhere. In the late 1980s, commercial transaction cost measurement services were offered by Gil Beebower at A.G. Becker and Abel Noser. I knew them as friends, but to me the services seemed to be missing something. The objective seemed to be to cast a suspicious eye on broker behavior. To me, there seemed to be much more to understand: What role did the trader himself play in the process? How about the portfolio manager? Around 1990, Plexus Group was in need of new product, and our attention turned again to the still lightly plowed fields of transaction cost measurement. It seemed that there was more that could be done to help investment managers reduce transaction costs. In all truth, this was an uphill fight, as suggested by the Demsetz quote. Portfolio mangers, chief investment officers, and head traders all echoed the thought best expressed to me by a Canadian investment manager: “I pick stocks. The rest is just plumbing.” Later, I would add: “leaky plumbing.” For a while I was infatuated about quotes on plumbing. Here is the best, from the eminently quotable John Gardner:

On ly

by something like 50bp before fees. Could the leakage be transaction costs? Not by the then-current estimates of impact plus commission costs, which were in the range of 25bp. If the value added by investment selection is 50bp and implementation costs take away 25bp, the net value-added should be positive, right? But it wasn’t. (And still isn’t!) Where did the value go? Well, that wasn’t my biggest problem at the time, so I set it aside for many years. Although I thought about it occasionally, it wasn’t until the late 1980s that I got serious about this conundrum.

th

SOME MORE ANCIENT HISTORY

Au

Let’s move on to 1972, to the fabled Management Sciences Department of Wells Fargo Bank. An investor with a background from The University of Chicago, the hotbed of early quantitative work in finance, was searching for someone who could create a real-world application of the then newly minted Sharpe model. Of course, we now know this as an index fund, although the term wasn’t invented until later. This first attempt was crude and largely off-base, but workable, except for one thing: We couldn’t afford to pay the normal 2

We must learn to honor excellence in every socially accepted human activity, however humble the activity, and to scorn shoddiness, however exalted the activity. An excellent plumber is infinitely more admirable than an incompetent philosopher. The society that scorns excellence in plumbing because plumbing is a humble activity and tolerates shoddiness in philosophy because it is an exalted activity will have neither good plumbing nor good philosophy. Neither its pipes nor its theories will hold water.

THE INCREDIBLE STORY OF T RANSACTION COST M ANAGEMENT: A P ERSONAL R ECOLLECTION

IIJ-JOT-WAGNER.indd 2

SUMMER 2008

5/20/08 5:22:30 PM

Now we’ve come full circle to the basic conundrum: can costs explain the difference between portfolio managers’ positive potential value and the negative realization? Three events added momentum to our new pursuit:

Au

th

or

Dr a

On ly

ew

vi

ft F

or

Fortunately we had good friends in the industry who showed interest in anything that would help them improve performance. I am forever indebted to Ted Aronson of (then) Aronson and Fogler, and Mark Edwards, who was running active quantitative portfolios for Minnesota SBI and ultimately became an inspiring partner at Plexus. Importantly, no one had ever looked at this level of transaction data: the order from the portfolio manager, and the process by which it transformed into changes in portfolio holdings. We had the total cost from Perold’s insight. We could isolate the commissions and impact, and they were in line with expectations. But there was something more, a missing piece: the Perold cost was three times as large as commissions plus impact. It didn’t take much searching of the data to discover where the hidden costs were, and to identify two sources of the problem. First, traders were focused on monitoring their potentially disloyal friends, the brokers, and on minimizing the impact of their trading on the market. Meanwhile, in front of their noses—but completely invisible—were other, insidious, egregious, hidden costs of trading:

Secondl, broker misbehavior might have been a problem, but it wasn’t the problem. The problem was one of institutional trading size compared with trader motivation, combined with a whole lot of temporizing, hedging, and second guessing of the trading decisions. The traders’ self-standard of excellence—minimizing impact—was disconnected from the primary motive: the maximization of investor wealth. It turned out the traders were right when they complained that the trade cost analyses of the day were simplistic and naïve. The data showed many levels of complexity, the “complex tapestry” of skills, handoffs, and tradeoffs as described in the CFA-Institute’s Best Execution Taskforce Report:

Re

• A new model: In 1988, Andre Perold defined implementation shortfall as the true all-in cost of transacting, calling on Treynor’s thoughts of 20 years before. • A new data source: Seth Merrin’s first functional, and soon to be indispensable, Order Management System. For the first time, data was available to trace the fundamental questions: who did what, when, why, and how? • An urgent motivation: Plexus Group’s relationship with Instinet’s Crossing Network had come to a close. We needed a new product to generate revenue!

walk” theory of price movements didn’t apply when there was latent, unfilled demand (or supply). Until that committed but uncompleted trading interest was filled, there would be constant one-sided pressure on the stock price. The result: delay hurt much more often than it helped. • This effect can be measured by simple arithmetic, when you know the manager’s intentions. Many terms have been applied to this cost, but we called it “delay” cost or “opportunity” cost. It is almost koan-like: the cost of not trading. • By measurement then, and even now, these egregious costs comprise 60-75% of the total costs. Yet nobody, myself included, even suspected them, although it seems so obvious now.

• The traders focused on minimizing impact, but that was only part of the story. • If you fail to trade while the stock is trending, you will ultimately trade at worse prices. The “random

SUMMER 2008

IIJ-JOT-WAGNER.indd 3

• Managers would slowly feed orders to the desk, imposing ever-shifting price limits and quantities; • The size of the institutional orders often exceeded the practical limits of available liquidity; • There was little attempt to prioritize orders, so traders could not tell which orders were the most sensitive; • Traders would attempt to time the market, usually unsuccessfully, holding back when they didn’t like the tenor of the market. All these factors contributed to the invisible costs, which were two to three times the explicit costs. Invisible, yes, but also egregious: if the price eroded in a stock to be sold because of manager hesitancy, trader reluctance, or market conditions, real losses were incurred by the investor, both on sales and purchases.

THE JOURNAL OF T RADING

3

5/20/08 5:22:30 PM

TODAY: WHERE ARE WE NOW?

Dr a

On ly

ew

vi

Re

ft F

I’d like to share with you my observations from 40-plus, on-and-off years of observing markets, transacting, and transaction cost management (TCM). TCM today has deep penetration into the practice of investment management and trading. In fact, we’re at a level of understanding and applicability inconceivable just a few years ago.

Au

th

or

1. The hidden costs are the major costs, totaling two-thirds to three-quarters of the total cost of implementing institutional investment decisions. The source of hidden cost is the inability to create liquidity upon command, i.e., when it is needed the most. The iceberg is the key insight to the process. Trading cost, we now know, is certainly not the commission and not just impact. It is the slippage between the costless value of the investment idea and the performance realization. Perold’s idea reigns supreme. 2. TCM is most effective as a feedback and selfevaluation process. It is an exercise in effective business management under conditions of uncertainty. The key insight is that it is a process, and 4

not an outcome. As trade directives move down the chain of knowledge, valuable feedback information now moves in the opposite direction and alters future behavior. 3. We have learned the importance of data. Long ago, Pythagoras said: “Without proof there is only the poverty of arguable opinion.” Similarly, without data—accurate, timely, and highly specified data—all we can practice is conjecture and self-justification. Yet the data we capture today is increasingly inadequate to the vastly more complicated task of evaluating complex, conditional, and rapidly changing algorithmic trading and dark pool searches. 4. Structural impediments to buyer meeting seller with multiple middlemen inf lated costs by at least a factor of two. Many of those middlemen could be found on the f loor of the exchanges, where we could see the energy of financial markets personified. Nostalgic, picturesque—and expensive. Very expensive as it turns out, and in ways no one suspected. This is one occasion where regulation clearly worked to the benefit of investors: the SEC and other regulatory bodies were right to force down the commission and the spread. Squeezing out the middlemen’s “vigorish” sparked the move to electronic trading venues, which can do the job a million times faster, with lower error rates and much less information leakage.

or

What a thrill! We were standing on the edge of knowledge, observing from a completely new perspective, seeing what no one had seen. Clearly, there was a business proposition here—an opportunity to enhance returns by tightening up the plumbing. All Plexus needed were customers. No matter how striking our findings were to me, it often failed to evoke a similar response in the plumbinghaters we hoped would become customers. Not only did we need to convince them that we had found a better way, we had to persuade them that their previous knowledge—and by a quick mental deduction, their skills—were incomplete and obsolete. I have a Dilbert cartoon in my scrapbook where Dogbert describes himself in the first frame as “the quantifier of unquantifiable things.” In the second frame, he says, “I declare you to be worth $85.” In the third frame you see Dogbert walking out of the office followed by a scream of epithets. Dogbert shrugs: “No one likes to be quantified.” That is a fairly accurate description of our early commercial encounters. Still, eventually the value proposition prevailed.

At the height of the internet/biotech frenzy in 2001, Plexus Group recorded one-way costs at 142bp for USD large capitalization stocks. By 2005, these costs dropped a remarkable 64%, to 51 basis points. They have stayed in that neighborhood ever since. 5. Lower transaction costs empower lower threshold strategies. Many good ideas that would not be profitable in a high-cost environment can now be executed. Many a worthless old slag heap can now be re-mined for gold worth $1,000 an ounce. 6. There are three big cost factors that determine what a trade is going to cost: the size of the trade relative to the trading volume; the small capitalization or thin trading supply of the company; and price momentum due to a conf luence of trading interest on one side.

THE INCREDIBLE STORY OF T RANSACTION COST M ANAGEMENT: A P ERSONAL R ECOLLECTION

IIJ-JOT-WAGNER.indd 4

SUMMER 2008

5/20/08 5:22:30 PM

7. How do we think about the determinants of transaction costs? Here’s my take on the market dynamics that determine the outer bounds of transaction costs:

HOW DOES THIS AFFECT INVESTORS AND FIDUCIARIES?

Dr a

75bp 20bp 50bp 145bp

Au

th

or

At the height of the Internet frenzy, we measured transaction costs peaking at 142bp each way for buying and selling. Coupled with a 100% turnover rate, transaction costs could be twice the total fees and expenses shown above. Even in today’s more efficiently operated markets, transaction costs can still be the largest drawdown of investable funds. So, as a fiduciary, charged with acting in the best interests of the fund holders, where would place your emphasis? I suggest that the interests of the beneficial owners are best served by following the following logic: 1) Effective oversight means knowing what to look for; which questions to ask. 2) All costs are negative performance, therefore anything that reduces costs without harming

SUMMER 2008

IIJ-JOT-WAGNER.indd 5

On ly

ew

TCA has come a long way, but some important opportunities remain. I can think of several key issues that will confront transaction cost researchers in the future.

vi

1. Most of the costs occur according to a sort of 80:20 rule: 20% of the trades create 80% of the costs. It seems to me, however, that it is probably not possible to avoid the runaway trading situations and to fill our trades before the price shoots up (or down). Here are my thoughts on the matter: Except for f low trading, the reason a manager decides to trade is because something has changed—i.e., a signal that calls out a response. It could be news, earnings surprise. relative change in value, changing of a rating, data errors in a quantitative database, etc., etc. But if nothing changes, the portfolio manager sees no motivation to trade. OK, but our heroic portfolio manager isn’t the only signal reader in the market; others will react as well, and the result is a herd effect. Thus the price shoots up because we, the collective, the herd, “we” are suddenly anxious to trade. As long as there are surprises in the market, there will be runaway trading cost situations. The silly solution to this problem would be to trade when there’s no immediate reason to trade, but would you invest with a portfolio manager who trades for no reason? 2. The same problem comes up when it’s time to focus our attention on the TCA reports. Our attention is naturally drawn to the blowups, but the trade desk alone cannot contain the damage. True cost containment comes from dealing with the run-ofthe-mill trading situations, not the conf lagrations. As with fighting brush fires, containment comes in fighting the many sparks before they blow up. Once the fire is out of control, the damage cannot

Re

ft F

If we add together the components of cost for an average U.S. mutual fund, we typically come up with numbers like this: Management fee Administration expenses 12(b)–1 fees Total fees and expenses

WHAT WE STILL DON’T KNOW

or

a. The minimum cost is the structural cost of producing the trading service. In a minimal cost market, the lowest cost producer becomes the dominant player. b. The maximum cost is expectational, and lies in the realm of the portfolio manager, who views the trade in terms of what can be gained for his portfolio: how much am I willing to pay to capture the potential of this investment idea? c. Liquidity failure occurs when the other side of the market disappears, often in a f lash. Once frightened, liquidity can stay away for a long time; consider the sub-prime meltdown or the Russian bond crisis for LTCM. Keynes said this best: “The market can stay irrational longer than you can stay solvent.”

performance potential is beneficial. 3) Consider Jack Bogle’s suggestion that the easiest way to be in the first quartile of performance is to be in the fourth quartile of costs. 4) Look for the low-hanging fruit to reduce the drag of transaction costs. Am I suggesting that fiduciaries should become TCA experts? Empathetically, no. But the people to whom you delegate your fiduciary responsibility should be.

THE JOURNAL OF T RADING

5

5/20/08 5:22:30 PM

Re

vi

ew

On ly

here: I cannot document that gains in trading efficiency result in thrillingly better portfolio returns. We have two data sources we can look at: the Standard & Poor’s SPIVA (S&P50 Index vs. Active) data, and the ITG/Plexus Group client database. The SPIVA data shows no trend: I divided the eight years of available data in half. In the first four years, from 2000 to 2003, the S&P500 index outperformed 54% of large-cap mutual funds. For the 2004-2007 period, the index outperformed 63% of the managers. Certainly no indication of improvement there. The Plexus data is more encouraging. In 2002 the average costs were 140bp, the costless returns were 110bp, and the net value added was a negative 30 basis points. By 2007, average costs were down to 40bp, which matched the costless return of 40bp, with a net value added of zero. Zero value added? Isn’t that wonderful? I would describe myself as unenthralled, but unappalled. The cheeriest conclusion I can reach is that this indicates market efficiency, in the “no free lunch” sense. Perhaps the best that can be expected in a hotly competitive market for investment ideas is that the search costs will be covered on average. This conclusion would fit nicely into a robust definition of market efficiency, incorporating the wedge costs of search and transaction costs. Some darker possibilities include:

Dr a

ft F

or

be contained. As we experience in California, all wildfires extinguish when they reach the ocean. So it is with trading situations; thus the disasters du jour will probably always be there. The trade desk can make heroic efforts to mitigate the damages, but the winds blow too strong. I can see two possibilities for mitigating the damage, one before the fact and one after. Both rest in the domain of the portfolio manager, not the trade desk. Before the fact, the solution is to consider the risk. As we’ve recently found out, quantitative techniques such as VAR (value at risk) all too often drastically understate the risk. Gaussian statistics don’t work during a regime change: they can’t predict 10 sigma changes. Perhaps the best solution is to induce terrible nightmares for the portfolio manager and motivate him or her to clear-cut the potential fuel that could lead to a wildfire disaster. The after-the-fact solution is to refuse to join the herd: wait six weeks and then decide if you still want to own the stock. This tactic doesn’t avoid the loss, it absorbs the blow, then looks to a new day when the value of the company can be reassessed in a more balanced frame of mind. 3. Another problem is our obsession foreseeing the future through transaction cost predictors. Over any time frame greater than a few seconds, this may be the impossible dream. Predicting all-in costs is the same as predicting future prices. If there is such a person who can do that, he has a great career ahead of him in a hedge fund. Or a religious movement.

Au

th

or

The truth is that the explanatory power of a predictor over the full execution of an institutional order is very, very low. The same equations that can explain costs after-the-fact lose over 90% of their power when applied to predictions. It is impossible to forecast independently occurring demand and supply in an unstable situation. Excessive reliance on predictions is like betting on a coin f lip. THE BOTTOM LINE

Finally, the most important point: How have reductions in transaction costs affected bottom-line investment performance? I have some perplexing news

6

• Could it be a curse of optimism, that managers will adjust their idea production to the hurdle rate and will always overstate their information advantage? • Perhaps we suffer recently from a lack of opportunity—i.e., fewer stocks with skyrocketing prospects in recent years. • Is someone else, such as the hedge funds, capturing the advantage? • Or perhaps there’s just too much noise in performance data and too loose a connection to transaction costs to notice the effect. Nonetheless, we have made great progress, and it is worthwhile to conclude with a brief list of some of the major accomplishments.

THE INCREDIBLE STORY OF T RANSACTION COST M ANAGEMENT: A P ERSONAL R ECOLLECTION

IIJ-JOT-WAGNER.indd 6

SUMMER 2008

5/20/08 5:22:30 PM

ENDNOTES

ew

On ly

This lightly overblown title comes from the kindly and fertile imagination of my friend, Robert Pouliot of RCP Partners, S.A, who invited me to give a keynote speech to the Centre for Fiduciary Excellence in Montreal, Quebec on March 12, 2008. 1 I ask your forbearance for all the vertical pronouns in this exposition. The first-person voice communicates as more evenly balanced in a speech than it does in writing. 2 Demsetz, Harold, “The Cost of Transaction,” Quarterly Journal of Economics [February, 1968]. 3 Cuneo, Lawrence, and Wayne H. Wagner, “The Costs of Transacting.” Financial Analysts Journal [November/December 1975].

vi

To order reprints of this article, please contact Dewey Palmieri at dpalmieri@ iijournals.com or 212-224-3675

Re

1. We have made great progress in understanding market microstructure, and how this affects the cost of trading. 2. The science of trade cost measurement has become widely accepted, and has led to significant enhancements to the way we trade securities. 3. These new trading methods represent a challenge to the TCM process. How do we gather the data to make intelligible choices among algorithms and the operative settings that govern them? 4. The TCM revolution has won major victories, but is still being fought daily. Continuing evolution will call forth new insights and new challenges. 5. Fiduciaries are well served by the attention they pay to these costs, not only for gains in efficiency, but also for the deeper understanding of the investment process it provides. 6. Despite whatever successes we’ve had in the past, best execution is particularly important now, when the patterns and practices are undergoing continuing rapid evolution.

Au

th

or

Dr a

ft F

or

Yes, there’s much to be done, and yet we’ve come a long way and that is cause to celebrate. We have saved billions of dollars for investors, and I am grateful to have an opportunity to make my own contribution to this noble cause.

SUMMER 2008

IIJ-JOT-WAGNER.indd 7

THE JOURNAL OF T RADING

7

5/20/08 5:22:30 PM