WIXLIB

assets. However the initiation and termination of funds within a group may increase theopportunities for cross trades.Consideration of several actual mutual fund groups may be useful. Longleaffunds manage more than 4.5 billion of assets and had more than 700 million of securitypurchases in 2001. Yet, Longleaf funds typically effect only a few cross trades each year.However, the closing of one of Longleaf’s four funds in 2001 allowed for the increaseduse of cross trading.11 Longleaf’s opportunities to cross trade are limited by the fact thatall of its funds are value funds that seek to acquire securities that are assessed to beundervalued.A more important source of cross trading opportunities arises fromopportunities to trade among different managers within a fund group. If a fund group hasdifferent types of managers or managers with differing opinions, there are increasedopportunities for cross trading. American Century has more than 85 billion undermanagement. Cross trading opportunities have resulted from a shift of investors’ assetsfrom growth to value managers within the group. The American Century Small CapValue Fund’s assets grew from 225 million at the end of 2001 to 1,286 million at themiddle of June 2002, a growth of 1,061 million. Assuming that 1% of this asset growthwas accomplished through cross trades and that a cross trade saves 1% of the transactionvalue round trip, the resulting savings are ( 1,061 million X 0.01 X 0.01) ) 106,100.1211The Longleaf funds operated four funds during 2001 including the Partners, Small-Cap and Realty funds.The Realty fund was closed. The Longleaf Partners Funds annual report at December 31, 2001 states:“Purchases in the Partners and Small-Cap Fund include securities purchases from the Realty Fund inconjunction with its liquidation. The securities were purchased from the Realty Fund at the last sales priceon the security’s primary exchange, without brokerage commission in accord with procedures previouslyadopted by the Boards of Trustees of the Funds pursuant to Rule 17a-7 of the investment Company Act of1940.“12One half of the savings accrues to the Small Cap fund and one-half to the fund with the cash outflow.4

As the amount of assets under management increases, the potential savings fromcross trading grows at an even faster rate. One reason for this is that larger funds trade inlarger blocks, which have greater market impact. Another reason is that larger fundshave more managers leading to more disagreements as to whether a given stock should beincluded in a portfolio. Fidelity Investments manages more than 150 mutual funds.Conversations with industry executives indicate that perhaps 8% to 10% of equity tradesare cross trades at Fidelity. Fidelity has an ERISA exemption, but this exemption is not asignificant factor in Fidelity cross trades.13 Consider just one Fidelity Fund, the FidelityAggressive Growth Fund, which had assets of 5,324 million and a turnover rate of 118%in 2001.14 Assume that all trades are equity trades, that cross trades save 1% round trip,and that 8% of trades are cross trades. Then the savings for this fund in 2001 due to crosstrades was ( 5,324 million X 1.18 X 0.08 X 0.01 ) 5.026 million, not counting anysavings due to purchases made as a result of fund growth. Even though the FidelityMagellan Fund had a turnover rate of only 15%, because its assets total almost 72billion, the savings from cross trades would likely have more than ( 72,000 million X0.15 X 0.08 X 0.01 ) 8.6 million.Clearly, cross trades represent economicallysignificant sums for Fidelity Investments.These considerations lead to the following observations that are applicable toERISA plans. The larger the ERISA plan the greater the potential savings from cross13PTE 94-43.14For a mutual fund, the portfolio turnover rate is an annualized rate found by dividing the lesser ofpurchases or sales by the average of portfolio assets. Suppose that a fund is growing so that purchases ofassets are greater than sales. In this case, a turnover rate of 1.0 indicates that sales of assets equaledaverage fund assets and purchases of assets equaled sales of assets plus any purchases of assets due to fundgrowth.5

trading and the potential savings grows at a much faster rate than the rate of asset growth.The main considerations that lead to this result are:1. Market impact and delay costs. Larger investment size leads to greater marketimpact costs. If 50,000 shares are traded each day, an order for 20,00 shares cantypically be executed quickly with little market impact while an order for 200,000shares will take more time to execute and incur more market impact. An order for400,000 shares is typically more than twice as difficult and takes more than twiceas long to execute as an order for 200,000 shares.2. Larger funds may involve more ERISA plans. The likelihood that there will beinflows for one plan and outflows from another at the same time is much greaterif there are five plans than if there are only two plans.3. Larger funds have more investment managers.Investment managers maydisagree about the merits of a particular investment. The greater the number ofinvestment managers the greater the likelihood that one manager will want to buya security that another manager is selling. This observation is most applicable toplans that employ multiple external investment managers.3. Transactions costsTransactions costs are all of the costs associated with the management ofinvestments, including the time involved in making investment decisions.15 Transactioncosts are classified as explicit or implicit. Explicit costs are itemized separately and paidfor like any other expenses. The most common explicit costs are commissions and15For additional information see Schwartz and Whitcomb (1988).6

transfer taxes and fees. The most common implicit transaction costs are the bid-askspread, market impact costs, delay, and opportunity costs. Table 1 provides an estimateof each of these costs for equities.16 Note that commissions are the smallest componentof execution costs. Delay costs are the largest cost. Market impact costs are alsosubstantial totaling 50% more than commissions. The costs of executing the averagesmall cap order are more than four times as high as that of the average large cap order.17A study of 31 stocks whose weighting in the Toronto Stock Exchange 300 index waschanged due to a redefinition of public float showed that the price impact of the changeon these 31 stocks was 2.34 percent.18 Because of market impact costs, it is common forinstitutions to break up large orders for execution. Define a trade package as orders onthe same side of the market by a single investment manager. One study of 37 largeinstitutions found that more than one-half of the dollar value of executed trade packagesrequired four or more days for completion. This finding provides strong evidence that16Taking a different approach, Hasbrouck and Schwartz (1988) reason that execution costs due to bid-askbounce (the movement of stock prices between the bid and the ask), market impact, and the like,temporarily push trading prices away from their true value. Hence, comparison of volatility over differentperiods can be used to investigate the impact of transaction costs. These authors calculate the ratio of thevariance of 12-hour returns to 24 times the variance of half-hour returns. Because these intervals cover thesame calendar period, we know that the actual volatility is the same over these two measurement intervals.Therefore, the value of the ratio should be 1.0. Any temporary deviations provide a measure of executioncosts. Using this methodology these authors report average execution costs of 0.148% of the transactionsamount for NYSE-listed stocks. The comparable costs reported by Hasbrouck and Schwartz for AmericanStock Exchange and more liquid NASDAQ stocks are 0.123% and 0.438%, respectively. If the assumptionthat the entire difference in the variance ratios is due to transaction costs does not hold, the Hasbrouck andSchwartz measure will overestimate execution costs. Stabilization activities of the specialist, the presenceof limit orders, or the sequential arrival of information causing price adjustment delays might also affectshort-term price volatility.17In this example large (small) cap (capitalization) is more (less) than one billion dollars.18Kaul, Mehrotra and Morck (2000).7

many institutional orders could not be executed without substantial market impact.19There is evidence that execution costs for institutions have increased recently.20Table 2 compares equity execution costs for seven countries. In every countrycommissions represent only a fraction of total execution costs. Trading costs can also belarge for debt instruments. One study reports that trading costs for bonds are about 0.27for each 100 of face value.21 Small institutions pay more than large institutions andsmall trades cost more than large trades.22In the next two parts of this section, direct and indirect transaction costs areconsidered, in turn. Then, issues related to transaction costs—price improvement, ECNs,patient trading, and monitoring transactions costs—are considered.Implicit transaction costs: The bid-ask spreadThe bid-ask spread was the focus of much of the early work dealing withtransaction costs and the earliest investigations of spreads focused on order processingcosts. These include the costs of the dealer's time, paper-work costs, transfer taxes, andother expenses incurred by the dealer in providing immediacy. In the process of makinga market the dealer may have money invested in the asset, incurring opportunity costs19Chan and Lakonishok (1995).20On June 24, 2001, the New York Stock Exchange lowered its minimum price increment or tick size from 0.125 to 0.0625. Examining institutional orders, this change decreased total execution costs for smallorders of less than 1,000 shares, did not change execution costs for orders of from 1,000 shares to 9,999shares, and increased costs for order of more than 10,000 shares. Institutional orders of more than 100,000shares cost one-third more to execute (Jones and Lipson, 2001). These authors measure execution costs asthe proportional difference between the volume-weighted average execution price of trades executed as partof the order and the price prevailing at the time the order was released to the trading desk.21Schultz (2001).22Schultz (2001).8

that contribute to order processing costs.Another cost of providing immediacy isinventory holding costs, dealer losses resulting from a positive or negative inventoryposition.23More recently, a third cost incurred by dealers who provide immediacy hasreceived attention, namely, asymmetric information costs.24 Some investors are likelybetter informed than the dealer. When these informed investors buy from the dealer,subsequently, prices are likely to rise. And when informed investors sell to the dealer,subsequently, prices are likely to decline. Therefore, in dealing with informed investorsthe dealer is buying just before price decreases and selling just before price increases.25A buy transaction arriving after a long time interval has a lower price impact than a buytransaction arriving right after a previous buy. Because informed traders typically seek totrade quickly before their information becomes generally known, liquidity suppliers infera higher likelihood of traders being informed if the trades are close together.2623Inventory holding costs are the costs incurred or anticipated by a dealer as a result of inventory positionsacquired in the process of market making. An example of the inventory holding cost model is provided byHo and Stoll (1981) who propose that after a dealer purchase bid prices are lowered to discourageadditional sales and ask prices are also lower to encourage purchases from the dealer. Thus, the new bidand ask prices must reflect prices that make the dealer indifferent between a transaction at the bid and atransaction at the ask.24Asymmetric information costs are the costs incurred or anticipated by a dealer as a result of trading withcounterparties who are motivated by the possession of superior insight into the appropriate equilibriumprice of the asset. Of course, the loss to the less knowledgeable party is the gain to the party with superiorinsight. A trader with knowledge that current prices are not at true value would is motivated to tradeimmediately rather than provide liquidity to others.25If this were not the case the market maker’s counterparties would not truly be informed. For examples ofasymmetric information models see Copeland and Galai (1983) and Glosten and Milgrom (1985).26Dufour and Engle (2000). There is evidence that order processing costs have little intra- or inter-dayvariation, but adverse selection costs are high at the beginning and end of the trading day relative to themiddle of the day. Adverse selection costs are also higher on Mondays than on other days of the week foractively traded firms.9

Glosten and Harris (1988), among others, have developed a methodology forestimating order processing costs and asymmetric information costs. Inventory holdingscosts, which are small, are not estimated separately, but are included in order processingcosts. For a sample of 856 equities traded on the New York Stock Exchange in 1999, thecosts as a percentage of the spread are:order process costs, 61.1%; asymmetricinformation costs, 38.9%.27Implicit transaction costs: Market impact costsThe execution of an order has the potential to move the market price. Here we aretalking about more than the change from the bid to the ask. Market impact involvesmoving the bid and/or ask either temporarily (in response to liquidity motivated trades) orpermanently (in response to information-based trades). Suppose that a market has sellerswilling to sell 5,000 shares at 50 and 5,500 shares at 50.10. An order to buy more than5,000 shares will exhaust the supply at 50 and if immediate execution is required willnecessitate executing part of the order at 50.10. Hence, this order will have marketimpact. Market impact is a cost associated with implementing an investment decision.Empirical evidence shows that for a given price change, the size of the associated trade islarger for sells than for buys. An early study reports that a decline of 0.125 is associatedwith a trade size of 451 shares while an increase of 0.125 is associated with a trade sizeof 427 shares.28 Hence, if the market impact of buys and sells differs, determining a27Van Ness, Van Ness, and Warr (2001).28Wood, McInish, and Ord (1985).10

cross-trade price by, say, taking the midpoint of the bid and ask, may be unfair to one ofthe parties.It may take several days to execute an order. There is a trade off between marketimpact costs and delay costs. Table 1 shows that delay costs are much larger thancommissions and, in fact, are almost 60% larger than commissions and market impactcombined (0.60/0.38 1.58).Opportunity costs result when the order cannot becompleted due to market price movements or other factors. Table 1 shows that for onesample the opportunity costs of the ten percent of trades not completed were twentypercent greater than commissions.Block tradesA block trade is commonly defined as a trade of 10,000 shares or more.According to one study, the price impact of block trades on Nasdaq is -0.55% for salesand 0.53% for purchases. Comparable figures for the NYSE are –0.48% for sales and0.46% for purchases.29The relationship between trade size and transaction cost is not as simple as largertrades cost more. In fact, one study reports “no evidence that larger trades are associatedwith higher indirect execution costs.”30 Larger trades have greater impact, but therelationship between trade size and market impact is not linear in trade size.31 Theinformation content of a block trade increases with trade size using earnings forecast29LaPlante and Muscarella (1997).30Block, French, and McInish (1994).31Koski and Michaely (2000).11

errors as a proxy for information asymmetry.32 Large trades have a greater price impactduring times when asymmetric information is greatest.33Buys and sells have different market impact. A price change of 1/4 is associatedwith an average trade size of more than 1,400 for a price increase and more than 1,700shares for a price decrease.34 Researchers have found that block purchases have a largerpermanent price impact than block sales.35 Similar results are reported for institutionaltrades.36 Prices go up on purchases and down on sales, but the sale prices quickly revertto their pre-trade price while the increased prices due to purchases are permanent.37 Theeffects of transactions and cumulative order activity on the price schedule increase as theaverage trading volume decreases.38The probability of information based trading is lower in high volume stocks.High volume stocks tend to have a higher incidence of information events than lowvolume stocks, leading to a higher arrival rate of informed traders. But the effects of thehigher arrival rate of informed traders are more than offset by an even higher rate ofarrival of liquidity traders.39 The majority of trades are small, but most of the cumulativestock-price change is due to medium-sized orders. Informed traders would prefer to trade32Seppi (1992).33Koski and Michaely (2000).34Wood, McInish and Ord (1985).35Kraus and Stoll (1972) and Keim and Madhavan (1996).36Chan and Lakonishok (1993).37Chan and Lakonishok (1993).38Kavajecz and Odders-White (2001).39Easley, Kiefer, O’Hara, and Paperman (1996).12

in moderate size because a very large order attracts attention and a small order does notallow the accumulation of enough shares.40 The source of these medium-sized trades thatgenerate most of the cumulative returns is institutions.41An upstairs market is a market in which larger orders are given special handling.On the NYSE brokers in the upstairs market can solicit counterparty orders, a practice notpermitted on the trading floor. On the Toronto Stock Exchange most of the orders thatare executed in the upstairs market are liquidity motivated so that they

Cross-Trading by ERISA Plan Managers. Final Report . Thomas H. McInish, Ph.D., C.F.A. Tel: 901-678-4662 . . the second fund incurs extra costs to purchase the securities and is motivated to . 8The INHAM concept is described in PTE 96-23 60 Fed. Reg. 15597 (Mar. 24, .