Process-driven -VS- Charismatic Trading
phlx.com
This paper contrasts the performance of a process-driven, or na‹ve approach to trading, with a charismatic, or beat-the-dealer approach to trading. Briefly, we find that a na‹ve approach to trading is substantially less costly than alternatives. The results suggest the opportunity for a new financial product that would enable liquidity motivated traders to trade at the Volume Weighted Average Price for a security over a particular time frame. Our findings suggest that institutional traders can exploit the optionality inherent in the trading process.
Chapter I
Introduction
Many investors believe that securities prices are informationally efficient. That is, stock picking and market timing raise costs without any corresponding increase in returns or reductions in risk. Those investors economize on information gathering costs and trading costs by purchasing index funds. However, there is currently no really good alternative for investors who believe that trading processes are relatively efficient.
There are many investors who do not believe it is possible to ?game? the trading process; that is take systematic advantage of stock price incongruities. Further they recognize there may be significant costs associated with a market-maker?s perception that they may be trying to ?game? the system. Currently, there is no trading process geared to the beliefs of those who are willing to be passive traders. They can try to synthesize a na‹ve trading methodology by breaking down large orders into smaller ones and executing them more or less randomly over the course of the trading day (or days), but such investors cannot simply enter an order that announces that they are pure liquidity traders who should not pay anything for the adverse selection prospect that the market-maker perceives.
This paper examines the desirability of being able to enter an order specifying that the price an investor is willing to pay is the volume weighted average price (VWAP - see Appendix 1 for computational details). Hypothetically, an investor should be able to contact a broker before trading begins and enter an order to buy or sell a block of stock at the VWAP for that day. The challenge for whomever must fill the order is trading in small enough quantities and frequently enough over the course of the day to achieve a VWAP fill at a small expected profit.
We term na‹ve trading, such as VWAP trading, process driven trading. We characterize trading that attempts to ?game? the trading activity as charismatic trading. Process-driven trading relies on fixed rules such as those used to approach the VWAP. Charismatic trading, unlike process-driven trading, relies on the trader?s genius, intuition and information to identify buy and sell opportunities.
The contra-party of the charismatic trader has to extract sufficient trading concessions to compensate for the possibility that information or intuitions are correct. Failure to extract extraordinary concessions from a charismatic trader exposes the contra-party to the economic devastation of a systematic adverse selection in the mix of trades in any particular session. One way for the investor to avoid being mistaken for a charismatic trader, thereby being assessed a premium to execute, is to announce to the world that no information or intuitions have influenced the trade, in effect, to become a process-driven trader.
Although it is possible to identify an infinite number of trading processes, this paper focuses on two processes that have a wide following: (1) trading at the volume weighted average price of the day (VWAP) , and (2) trading at the closing price on the day of the trade. The day?s VWAP provides a reasonable proxy for the price sought by uninformed passive traders because it incorporates price and volume information from the transactions executed during the measurement period. The VWAP for the day?s trading in a security has been proposed as a benchmark for estimating execution costs of listed securities . The close on the day of the trade is often sought by mutual funds and others that price their portfolios at the close of the day for purposes of valuing their investors? positions.
To compare the results from trading rules to charismatic trades, we evaluate the trading costs of a group of large institutional investors. By examining trade confirmations for these investors, we can estimate the impact of costs related to trading risk, defined here as intra-day price volatility, on seven of the largest domestic institutional investors during the third quarter of 1996 .
Note that we do not consider the opportunity costs of delayed trades. These are often measured as the difference between a security?s price at the time a decision to transact was made and the price at which a transaction actually occurred. We believe that this measure fails to capture the true economic nature of the net costs of a delayed or abandoned trade. Specifically, we contend that among the trading operations of institutional investment managers (?buy side? as opposed to ?street side? traders), a trade now competes with itself at a later time. Increasingly, buy side traders employ hedging techniques for the purpose of minimizing the adverse impact of a trending market. The simplest and most intuitive of these strategies are dollar neutral strategies that engage in selling securities to raise cash to buy securities.
Implicit in such strategies is the option to delay, alter, or cancel an order to trade. If new information arrives prompting the investor to cancel a trade after an original delay and before execution, then the delay actually produces a benefit.
This immensely complicates the measurement of opportunity cost. For instance, it is easy to measure the cost of a complete transaction program. Suppose the prevailing price at the time of the decision to buy was $30 and the average trade price was $31. The opportunity cost is this $1 per share less any market impact cost. This tells us something about the cost of a complete transaction. However, when there are delays in trading, some trade programs may be abandoned owing to new information. In evaluating trading costs for an investment organization, the benefit of having the option to abandon a trading program that itself was going to be stretched out due to possible market impact costs ought to be netted against the opportunity costs. Collecting the data to measure these net opportunity costs requires clear documentation of the trading decisions and portfolio management decisions that follow an order to trade from conception to execution. Not all of these decisions and their underlying rationale are preserved in a form amenable to analysis, in part, because there is often some overlap between the discretion of the institutional trader and the portfolio manager with respect to the treatment of an order. Thus our focus remains on market impact costs.
We find that simple process-driven trading strategies available from contemporary electronic trading facilities can add value as tools in the trading arsenal. After looking at actual transaction costs (for some combination of process-driven and charismatic trades), we evaluate the trading profit potential in charismatic strategies that are triggered by the day?s first favorable tic supported by institutional volume after 10:00 AM and before noon. The na‹ve strategy we model participates after the first favorable tic in successive institutional sized trades to fill an order. We demonstrate that for this strategy to succeed the trader must be correct in his/her opinion of the day?s price trend at least 38 percent more often than s/he is wrong for buy transactions and 89 percent more often than s/he is wrong for sell transactions.
Chapter ll
Trading Motivations: Information and Liquidity
At any time, traders are held hostage to the prices offered in the market. If they are what Jack Treynor calls information traders , they seek to have their order filled as quickly as possible because the value of the information they process decays rapidly during the period when news is disseminated. Note that orders to buy and sell securities may, themselves, be construed by market participants to be information depending on the size of position held or divested and the reputation for knowledge of the principal[s].
Liquidity traders, on the other hand, are sensitive to price. They trade on the basis of fundamental analysis at opportune instances. During the trading day, it is the liquidity traders who provide shares to and cash out the information traders. This is not to say that liquidity traders represent charitable institutions - for their willingness to transact, they expect to extract a premium sufficient to offset their losses when information trader(s) actually have accurate information.
Information traders go to great lengths to masquerade as liquidity traders in order to avoid the larger price concessions market makers try to levy on information traders. Thus, information traders gravitate towards trading facilities offering anonymity. Because trading at the VWAP signals nothing to the market place about the existence of proprietary information, it provides an advantage to the information trader: there is no reason for the contra-party to adjust the price in anticipation of being ?taken? by the information trader.
Chapter Ill
Trading Costs
The ideal transaction cost measure specifies the difference between the actual price at which an order is filled (including commission costs) and the price at which a na‹ve trader can expect to execute the trades that fill the order . A na‹ve investor has no prior information or opinion about market prices during the time the order is being filled on the trading desk. Costs associated with trading securities are of the same order of magnitude as investment management fees. The challenge to fiduciaries for natural owners of securities lies in finding cost-efficient trading strategies that do not compromise investment performance.
The VWAP has appeal both as a trading benchmark and as a trading process because it represents an attainable price available to the na‹ve trader. The computation is similar to that which brokers use to price orders filled by executing multiple trades. When orders are filled from multiple trades, the price confirmed to the customer is the volume weighted average of the prices for the specific trades executed to fill the order. Also, when brokers fill orders in a single security on the same day by executing multiple trades for multiple clients, brokers place the trades used to execute these orders in an average price account. The price confirmed to the client(s) represents the volume weighted average price of the trades residing in this average price account.
To be fair, VWAP is not universally regarded as the premier method for measuring market impact costs. Keim and Madharan point out that the measure can be gained; for large trades in illiquid stocks, the VWAP and the transaction price should be virtually identical. But this ?game? can only be successful if the portfolio manager is only being measured on transaction costs. To the extent that success in this game contributes to poor overall portfolio performance, investment managers will choose not to play it. Lastly, note that the cost of gaming the VWAP typically exceeds the economic benefits that can be derived.
Chapter IV
Stock Trading
NYSE specialists and their counterparts in other markets seek to impose order on the markets by making stabilizing trades for their own accounts, acting as the contra side of small orders as the institutional trader of last resort. These market makers represent the promise of their exchanges to provide orderly markets, particularly in congestion. It is this promise that makes possible institutional investment in the stock market under the ?prudent man rule? codified in Federal and State law. To help maintain orderly markets, floor brokers and the ?upstairs? market makers have reason to cooperate with the specialist and his counterparts on other exchanges. Cao, Choe and Hatheway note that the specialist operates as a dealer, giving up some profitable opportunities in return for the specialist franchise and the informational advantages attendant to the franchise. The specialist quotes bids and offers for stock based on knowledge of the limit order book (which he is responsible for maintaining), the mood of the crowd, and his own inventory position and risk-taking objectives. The specialist has a fiduciary responsibility to represent the limit order book ahead of his own trades.
The NYSE monitors the performance of specialist firms to determine the extent to which they fulfill their obligations. Cao, Choe and Hatheway report the success of the specialists in this endeavor.
Exhibit 1: Specialist Obligations
Demand for special handling of complex trades has led to the emergence of the ?upstairs? market. In effect, these ?alternate market makers? provide liquidity to the natural owners of large security positions by finding the other side of trades or by committing firm capital. Natural owners of securities seek out the ?upstairs? market to design and execute complex trades with unusual characteristics involving size, terms, immediacy or applicable regulations.
Trading is a competitive activity in which success can be measured either with respect to price or immediacy of execution. Traditionally, portfolio managers and their traders have measured immediacy in minutes from the time the order is placed. As portfolio managers have become more sophisticated and their portfolio requirements more complex, immediacy has come to mean the minimum number of days required to trade an institutional position without roiling the market(s) in which the securities trade (see Exhibit 2) . Data in Exhibit 2 indicate that institutional investors create multiple orders (reflected in the trade confirmations) from positions they seek to acquire or divest, imputed by consecutive trades in a stock.
Liquidity groups are formed by ranking the 8000 most active stocks by market capitalization into 80 groups of 100 stocks. Liquidity Group 1 consists of the 100 largest stocks by market capitalization Liquidity Group 2 the next hundred largest and so forth.
Transaction confirmations from seven large institutional traders during the third quarter of 1996 indicate that sells were more costly to execute than buys. Appendix 2 reports market impact of their trades by liquidity group, formed by rank ordering stocks according to market capitalization as a proxy for liquidity such that the first liquidity group contains the 100 most liquid stocks and the eightieth liquidity group contains the 100 least liquid stocks. In general, transaction costs are higher for less liquid stocks. Further examination of the confirmation data for the most liquid stocks suggests that cost is related to size of confirm and immediacy. Buy orders are smaller and traders take more time to execute them than sell orders.
Orders For Groupings of The Most Liquid Stocks
Chapter V
Institutional Trading Risk
Intra-day price volatility is an important part of trading risk. The day?s high and low prices for trades in which institutional traders can participate set the boundaries for trading risk. Scale economies in clearing and settlement prevent institutions from filling large orders with huge numbers of very small trades. For purposes of estimating the extremes of trading risk we assume that institutional orders in the 400 most liquid stocks (Liquidity Groups 1 through 4) are filled with trades of 2,500 or more shares.
When computing trading risk in Exhibit 3, we consider only trades of 2,500 or more shares. The risk that traders fear most is to buy at or near the high on a day the market rises and sell at or near the low on a day that the market falls. Investors who acquire the worst 5 percent of prices during the days when the closing price differs from the opening price have orders filled at prices that average $2.50 or more away from the day?s best 5 percent of prices. Market trend appears to have little effect on the magnitude of the variation in the high-low range for up-symbol days (days where the closing price for a stock exceeds the open for a stock) and down-symbol days (days were the open price for a stock exceeds its close).
Chapter VI
Trading Performance of Large Institutional Investors
I n aggregate, institutional investors do not obtain prices that are more favorable than the volume weighted average price of the day on which the trade takes place. To analyze the trading performance of institutional investors, we examined the trading of seven large institutional investors during the 3rd quarter of 1996. These investors initiated $151.2 billion of trades. Our examination of the trade confirmation records reveals that buy transactions execute at prices that, on average, exceed the VWAP by 7.2 basis points and sell transactions execute at prices that, on average, fall below the VWAP by 11.2 basis points . These confirmations report trade executions at prices that were less advantageous to the executing institutions than the VWAP by $135.7 million.
Professional traders have differential results even with most liquid stocks. Among the seven institutional traders the average difference between best and worst institutions with respect to aggregate market impacts for buy transactions exceeded $0.05 per share and the difference between best and worst aggregate market impacts for sell transactions exceeded $0.06 per share. Higher commissions did not appear to purchase better executions. Some of the differences in transaction costs may be related to the mix of securities traded, but the bulk of the trades, shares, and dollars were traded in the most liquid stocks.
Transaction Costs Comprise a Significant Portion of Investment Performance
Transaction costs are the price concessions required to attract the opposing side of a trade. A significant portion of these costs are the commissions incurred to pay for brokerage services. In our examination of the trading of seven large institutional investors during the 3rd quarter of 1996, we found that commission costs averaged 11.5 basis points per share and market impact costs averaged 9 basis points per share (see Exhibit 5). Most of the commission costs were incurred for transactions in listed securities . Only a small number of OTC agency trades incurred commission costs.
Chapter VII
Testing a Trading Rule
One way to mimic the behavior of a charismatic trader is to assume that such a trader would place orders after seeing a favorable institutional sized trade. Further, we assume that such a trader would accumulate shares from successive institutional sized orders until the order is complete. We assume an institutional order size of 12,000 shares. To model this behavior, the first institutional sized trade (of at least 2,500 shares occurring on or after 10:00 AM) priced below the prior day?s close begins the sequence of buy transactions; and the first institutional sized trade above the prior day?s close (on or after 10:00 AM) begins the sequence of sell transactions. This process allows the charismatic trader to wait for the market to respond to the opening trade. Our charismatic trader is assumed to acquire 12,000 shares by participating in successive trades of at least 1000 shares until the order is filled. Although this assumption imposes na‹ve constraints, it recognizes that traders cannot own the last sale, but only the future sales.
To analyze the results, we segment our sample of the 400 most liquid listed stocks for 17 stock days (during the period May 7 to June 8, 1998) into up symbol-days, unchanged symbol-days and down-symbol-days. Using the VWAP as a benchmark our charismatic buyer and our charismatic seller incur substantially greater market impact costs resulting from an incorrect market direction prediction than value added from a correct prediction.
Our simulated charismatic trader adds value over the VWAP of 33 basis points from buys on a down symbol days and incurs a market impact cost with respect to the VWAP of 46 basis points from sells on down symbol days (see Exhibit 8). Conversely, this same trader adds 24 basis points of value over the VWAP from sells on up symbol-days and incurs market impact costs of 45 basis points from buys on up symbol days.
Assuming trades only on days when the market closes at prices that are different from the open, our charismatic trader with equal dollar amounts of buy and sell orders for liquid stocks would have to guess the direction of a stock correctly at least 38 percent (abs(-0.4546/0.3293)-1) more frequently than guessing wrong just to break even on the buys. To break even on the sells the charismatic institutional trader would have to guess right 89 percent more frequently than guessing wrong.
Relaxing the assumption that the charismatic trader would participate in successive trades as an automaton has little impact on the asymmetric payoff structure associated with trading decisions that depend on an opinion about the direction of the market. Appendix 5 presents the results where charismatic traders are allowed a random draw from the sequence of trades following a favorable tic on institutional volume.
Exhibit 8: Execution Costs of Modeled Charismatic Trades
Chapter VIII
Conclusions
Trading costs have a significant impact on investing. The failure to manage trading costs decreases returns to investors and may incur the ire of the Department of Labor ERISA oversight officials as it has been more than 10 years since the DOL issued the memorandum clarifying a plan sponsor?s obligation to attempt to acquire ?best execution ?. In an environment where mutual fund managers over the last 10 years have had aggregate performance below the S & P 500 index funds by amounts ranging from 21 basis points to 281 basis points (with the possible exception of mid-cap growth funds), trading costs become a significant factor .
There is a marked pattern of increasing interest in controlling transaction costs. Some large institutional investment managers have adopted strategies to seek trading profits that will, in part or in full, offset trading costs. Given the significance of trading costs to investment performance, we expect the lines between trading and investing to become increasingly blurred in practice. In part this will be due to evolving notions of the value of the option implicit in the orders that hit the trading desk.
Large institutional trading desks receive orders to buy some securities and sell different securities. Typically, the sell orders generate the cash to fund the buy orders. Optionality comes from:
1. The ability to modify the list of securities and the number of shares contained in buy orders, sell orders, or both to either maintain dollar neutral positions or attempt to profit from a trending market (thus netting against opportunity costs);
2. The ability to extend the time over which an order is filled thereby changing the distribution of prices to which the order is exposed;
3. The ability to execute significant components of buy or sell orders from cross trade opportunities; and
4. The ability to execute buy orders or sell orders from principal trading accounts (subject to significant regulatory constraint among ERISA funds).
In effect, trading costs arise when there is a mismatch between the dollar volume or risk of buy orders and sell orders. Under these circumstances, trading costs become the cost of properly hedging this mismatch or buying insurance to cap losses resulting from this mismatch. Note that brokers offering to trade packages at some benchmarked price are bundling insurance and brokerage services in a single priced transaction.
Heretofore, brokers and securities dealers have captured the value of these embedded options and insurance premia. They have profited handsomely. New instruments such as the VWAP trading facility proposed by the Philadelphia Stock Exchange, and the emerging electronic trading facilities such as Instinet, ITG Posit, the Arizona Stock Exchange, Lattice and Optimark, make feasible the implementation of trading strategies that will enable institutions to realize the value of the option to trade. Thus, institutions have an opportunity to dramatically improve returns on assets for which they serve as fiduciaries. Widespread use of charismatic strategies is not likely to survive. The aggregate market impact cost of approximately 5 basis points associated with trades in listed, mostly liquid stock incurred by seven of the largest institutional investors suggests that the prescience required to predict the direction of the market for specific stocks or portfolios lies beyond mortal means. |
