Feb. 20, 2004

Jonathan G. Katz, Secretary
Securities and Exchange Commission
450 Fifth Street, NW
Washington, DC 20549-0609

File No: 33-8349

Dear Secretary Katz,

Increased disclosure of transaction costs is a worthy aim, given their evident effect on fund performance. The challenge is designing an objective and cost-effective procedure that yields a comparable metric across funds and that captures all of the costs involved with transacting in portfolio securities. The solution proposed below offers a reasonable compromise on all four dimensions, with the result being that investors get a much better picture of where their returns come from.

We frame our proposal around the "implementation shortfall," as developed by Andre Perold, "[T]he difference between performance on paper and in reality is what we call implementation shortfall (or just shortfall)." The aim is to present fund performance to investors as a combination of three contributing factors: (1) the portfolio manager's investment decisions using a benchmark/reference price; (2) the costs and adverse market impact associated with executing trades to implement these decisions (that is, the implementation shortfall); and (3) the fund's operational expenses. Current disclosure commingles investment decisions with the implementation shortfall, but separates out expenses.

Our specific proposal is that the fund's Annual Report to Shareholders and Prospectus be required to disentangle the performance effects of investment decisions from the cost of executing those decisions, according to an algorithm detailed below. Many funds provide information on returns, fees, and turnover in a Financial Highlights section of the ARS. The disclosure that we propose would fit well within such a discussion. Exhibit 1, discussed in greater detail below, depicts the disclosure we have in mind.

Framework

In its purest form, the implementation shortfall measure requires that the precise time of all investment decisions be recorded, and it requires that market prices at the time the decision is made be recorded, and finally it requires diligent maintenance of a benchmark "paper" portfolio to track the return on those decisions. Armed with such information, one can gain valuable insight into where investment decisions added value, and where the execution left investors with a return that fell short of the mark.

Ensuring that this paper portfolio is tracked faithfully and accurately would place a great burden on accounting systems. Indeed, the current fund infrastructure has no means through which such a process could be managed and automated. Hence, producing the paper benchmark portfolio required for a pure implementation shortfall measure would be prohibitively costly. Moreover, the information required is subjective and unverifiable, so validity would always be a problem.

Our compromise solution is to construct the fund's benchmark paper portfolio as if the fund executes all transactions at a price no higher than, and no lower than, the average transaction price across all market participants for the day on which the fund trades. Thus, the paper portfolio equates to the actual portfolio except that all transactions are incorporated on an as if basis - as if the trade price was the volume weighted average price (VWAP) for the day. The difference between this VWAP based 'as-if' NAV and the fund's actual NAV, which incorporates trades at their actual execution price less commissions, indicates the cost of executing investment decisions relative to comparable transactions. Exhibit 2 details the calculation of comparable-transaction execution costs.

Computing comparable-transaction execution costs requires data that fund accounting already process on a daily basis in computing NAV: the trades for the day and the prices of those trades. It also requires additional data - the VWAP price for all securities traded. Because this price is common to all funds it could cost effectively be provided by the exchanges without difficulty, much like closing prices.

Our Specific Proposal

Expense Ratio Approach?

From an accounting point of view, it would be easy to place the comparable-transaction execution cost (or just 'execution cost') on the expense ratio. However, this would involve a rather aggressive accounting position. It is important to recognize the full implication of such a treatment.

Suppose for example that fund XYZ bought a single share of stock for $11 when the VWAP for the day was $10. Traditional (GAAP) accounting would deduct $11 from cash and book the stock investment at $11, recognizing no expense on the transaction. While the execution cost is $1, it is not reflected in the fund's accounting.

Suppose instead that the $1 execution cost was deducted as an expense. Then the transaction would be booked as follows: deduct $11 from cash, offset it by booking a $10 asset and a $1 expense to shareholder's equity. Note that this necessarily implies booking the stock purchase at VWAP, rather than at the actual purchase price. Hence the "rather aggressive" accounting treatment.

Such distortion of the actual price paid is not alien to GAAP, for example current practice bakes the commission into the book value, causing a departure from the purchase price. Under this practice of capitalizing the commission, the security is booked at a price higher than the transaction price. When the security's value is later marked to market, the commission premium vanishes into a capital loss. This imparts a slight negative bias to the fund return, but there is no direct expense charged. By contrast, were the commission booked as an expense at the time of transacting, fund returns would not have the negative bias that they now do.

The balance-sheet implications of putting execution cost on the expense ratio must not be overlooked. Going back to the preceding example, under current accounting, the $1 execution cost is capitalized into the book value of the asset. Since on average tomorrow's market price will be about equal to today's market price, the $11 book value will on average be marked down to about $10. Thus, the $1 execution cost that was capitalized at the time of the transaction gets quickly funneled off the balance, imparting a negative bias on fund returns in the process.

Were the execution cost booked as an expense in the first place, the stock purchased would have been initially booked at a value of $10 and the negative bias that current accounting imparts on fund returns would disappear. That is, fund returns gross of expenses would be a truer representation of the manager's investment decisions.

Our proposal

Our view is that the aggressive accounting associated with expensing execution costs is not likely to sit well within GAAP. Moreover, execution costs are relatively conceptual in nature and would therefore not well align with existing (out of pocket) expenses. In any event, the need is to provide investors with the means to evaluate how much the portfolio manager's investment decisions added value, and how much that value was eroded by implementation shortfall. The information needed to make that evaluation could be easily and effectively provided with a brief, boilerplate disclosure in the fund's Annual Report to Shareholders and Prospectus.

Exhibit 1 depicts the disclosure we have in mind, where we reproduce the 'Financial Highlights' section of the Annual Report to Shareholders for a typical fund. In the box is the proposed added disclosure. There, the fund's actual net return is broken down into three components (see exhibit 2 for details):

Return due to investment decisions:

This is the 'as if' fund return assuming that all of the fund's security transactions occurred at the market-wide average transaction price for the security that day (i.e., all trades evaluated at VWAP).

Return due to trade-execution costs:

This is the difference between the fund's actual transaction price and the market-wide average transaction price for the security that day (VWAP), summed over all of the fund's transactions.

The operating expenses during the year.

Evaluating this approach

Objectivity

While theoretically appealing and potentially applicable in an internal setting, a pure implementation shortfall measure is unworkable as a publicly disclosed measure of transaction costs. To construct the paper portfolio benchmark, one needs the exact time that investment decisions are made, and a complete listing of all such decisions. The current accounting infrastructure could not facilitate such a demand. Moreover, it is hard to see how the integrity and objectivity of such information could be relied upon. Our measure employs only actual transaction prices so there is no data input that could be considered subjective.

Cost effective

As discussed above [objectivity], a pure implementation shortfall measure places demands on the fund accounting and pricing infrastructure that could not be met without substantial and costly modifications. In particular, the decision time for every trade would have to be recorded, as well as the current market quotations at that time. By contrast, our approach requires only the quantity and price of each of the fund's trades during the day, and a universally applicable VWAP for the security. Fund accountants already must process and incorporate the quantity and price of each trade on a daily basis to compute NAV, so there is no material burden from that requirement. To compute a universally applicable VWAP for each security traded, one merely needs the daily dollar volume and share volume of the security in question. Such data are typically available from the pricing feeds of the exchanges and NASDAQ. Where they are not available, the VWAP might be proxied with the average of the low, high, open and closing prices.

Comparable

Since all funds trading a given security are held up against the same benchmark price, a universally applicable VWAP, the resulting cost assessment is comparable across funds. One concern in this regard, and it is a weakness of the VWAP measure, is that costs are understated to the extent that the fund's trades represent a substantial fraction of the day's volume. In the extreme, the VWAP-estimated trading cost for a fund that does all of the day's volume is zero, regardless of how much the trade adversely affected prices. This could mean that comparing large funds to small funds may present a biased picture. Thus, the measure compromises somewhat on comparability. However, offsetting that is the fact that a large fund is likely to move the prices more and thus have a higher VWAP-estimated trading cost.

Completeness

Relative to a pure implementation shortfall, our measure understates transaction costs. The true implementation shortfall incorporates commission and exchange fees, as does ours, a market impact on prices from trading, as does ours, but also the opportunity cost associated with trades not executed. Ours does not include an opportunity cost. However, the opportunity cost of a trade not executed is at best ambiguous and subjective. Hence we find this a reasonable compromise. There are other reasons that the VWAP-estimated cost is an understatement of funds' all-encompassing transaction costs or implementation shortfall, including multiple-day trades and the aforementioned distortion of large trades on the VWAP benchmark. However, the VWAP procedure that we propose is certainly more encompassing than simply taking commissions as a measure of trading costs.

Sincerely,

Roger M. Edelen, Ph.D.
Managing Director of Research
ReFlow Management, LLC

With collaboration from:

Alan Siegerman, CFA
Vice President of Product Development
ReFlow Management, LLC



Exhibit 1 is in Acrobat format.

Exhibit 2 is in Acrobat format.