January 14, 2002

Jonathan G. Katz, Secretary
Securities and Exchange Commission
450 Fifth Street, N.W.
Washington, D.C. 20549-0609

Re: Actively-Managed Exchange-Traded Funds
File No. S7-20-01

Dear Mr. Katz:

Nuveen Investments1 is writing in response to the Commission's request for comment on issues relating to actively-managed exchange-traded funds (ETFs).2 Nuveen is pleased that the Commission is soliciting public comment on the issues that may be raised by these products. This approach will allow the Commission to consider actively-managed ETFs from a variety of perspectives to determine how to regulate them, consistent with the protection of investors and the Commission's interest in encouraging the development of new financial products that can help individual investors manage risks and meet their diverse investment objectives.

To establish a frame of reference for some of our most important comments on the Actively-Managed Exchange-Traded Funds Concept Release, we would like to begin by addressing (1) the Commission's distinction between active investment management and the indexing of portfolios [Section IV. A.] and a related issue, (2) the construction and use of equity indexes. The latter topic is not specifically addressed in the Concept Release, but we address it with reference to the question about problems encountered by index ETFs raised in the second paragraph of Section IV.C.

Indexing versus Active Management

Investment management is conducted in a large number of ways and under an even larger number of labels. At one extreme might be the committed indexer who adjusts a portfolio compulsively at the moment of index change to track a benchmark index as closely as possible. Given the precise rules by which most benchmark indexes are modified, it should be possible (before reported fund expenses, brokerage commissions and other transaction-related out-of-pocket costs) for a committed indexer to track a benchmark index almost perfectly. This is generally accomplished through entry of market-on-close orders or their equivalent every time an index change is scheduled. At investment management's other extreme is the possibly mythical active manager who changes the character of a portfolio frequently in response to even modest changes in information about a company, industry or investment theme. We doubt that either of these extremes is highly effective in achieving favorable investment results for the beneficial owners of a portfolio because they both cause the fund to pay high transaction costs. The investment technique "space" between these extremes is not a smooth continuum, but it accommodates most successful investment managers.

One investment technique generally seen as close to the index end of the scale is that of the passive investment manager who will use a benchmark index as a benchmark but not feel compelled to modify the portfolio to track the index precisely. A passive manager may attempt to transact either before an index change is announced or after most investors have implemented the change. If possible, he will try to transact when his portfolio changes can provide liquidity to a less patient investor on the other side of the trade. This manager might even achieve negative transaction costs3 by being a liquidity provider.

Another attractive indexing technique is the use of what we call a Silent Index. This can be a unique index or it can be an index with most of its rules and stock universe similar to those of a popular benchmark index, but a different set of rules for timing index transitions, modifications and re-balancings. A simple example of a Silent Index Fund might be a fund based on index rules similar to the Russell 2000, but using an undisclosed index reconstitution date some months away from the summer chaos of the Russell re-balancing. Such an index fund might easily add 100 basis points or more to the average return of the Russell benchmark simply by timing its reconstitution and re-balancing differently from the Russell standard.

The Commission has granted approval for use of a technique called representative sampling4 in the management of indexed ETFs, subject only to the undertaking of the fund advisor to achieve a tracking error no greater than 5% annually. Representative sampling works quite well and can lead to very close tracking of a broad-based benchmark index that is naturally compliant with Regulated Investment Company (RIC) diversification requirements and has a large number of components. With representative sampling, this type of fund can reduce costs and track its index closely without using all the securities in the index. Examples of this approach include most Russell and Wilshire index funds which ordinarily do not hold all the stocks in their indexes.

When representative sampling is applied to an index that is too concentrated to meet the RIC fund diversification requirements, the result is often a fund that fails to track the index by any reasonable standard. The resulting fund does not look or act very much like an index fund - at least not an index fund based on that fund's nominal underlying index. In fact, it looks a great deal like an actively-managed portfolio process using the perception of indexing and an index brand name to gather assets under an index ETF exemptive order.

In contrast to its permissive stance on the use of representative sampling in non-RIC-compliant index funds, the Commission rejects the index fund classification for funds which provide either leveraged exposure or inverse exposure to an index, even if such funds would track the underlying index objective (i.e., multiple or inverse) quite closely in an ETF implementation. The Commission suggests in the Concept Release that these should be considered actively-managed funds. Given the way most investors and investment managers view the relationship between index funds and actively-managed funds, this classification is as inappropriate as according an index fund label to funds which apply representative sampling to non-RIC-compliant indexes.

If there has to be a classification principle to separate index ETFs and active ETFs, it should be achievable tracking error in the ETF format. Using tracking error as the distinction between indexing and active management is widely accepted. The iShares website glossary (www.ishares.com) even defines tracking error as "active manager risk." On this principle, leveraged and inverse index tracking funds that track their objective closely are index funds and representative sampling funds that do not track their non-RIC-compliant indexes very well are actively-managed funds.

Most actively-managed funds are based on methodologies which use substantial quantitative inputs and stock screening software. The characteristics of most active portfolios are constrained by risk measurement and control tools designed to attain an investment objective based on performance relative to a benchmark. That does not sound as different from indexing as the active vs. indexing debate usually suggests. The legendary "stock picker" who buys McDonald's stock shortly after its IPO because he likes their hamburgers is not strictly mythical, but his total portfolio will usually be managed within a set of risk tolerances. The active portfolio manager has a great deal of flexibility in theory, but the daily operating rules under which most funds are managed usually constrain the fund's departure from its benchmark - usually one of the indexes most widely used by indexers. Some labeled index portfolios, particularly those based on representative sampling of a non-RIC-compliant index, offer more investment freedom and flexibility than the typical active management strategy.

Index Construction and Fund Performance

In granting ETF Exemptive Orders for index funds, the Commission has consistently required (1) that the index be maintained by an independent party and (2) that changes in the index be made public at the same time they are revealed to the manager of the fund. This policy is contrary to the best interests of index fund shareholders. To the extent that an index is devised as a custom template for a fund, the fund's transaction costs might be substantially reduced if the index changes were not public knowledge until after the fund had transacted. As noted above, we estimate that simply changing the date of reconstitution used by the fund's custom index will improve the average return of a portfolio otherwise based on Russell 2000 index rules by over 100 basis points annually relative to the actual Russell 2000. Investors in index funds should not have to incur the added transaction costs associated with the public reconstitution and/or re-balancing of a benchmark index. After all, one of the most important of the original ideas behind indexing was to reduce a fund's transaction costs. (Gastineau (2002), a copy of which accompanies this filing, discusses the origins of indexing and the benchmark index trap that ensnares most contemporary index funds in high transaction costs.)

We do not believe the Commission should delay authorizing actively-managed exchange-traded funds while it re-evaluates its attitudes toward indexes and indexing. The most constructive process would be to let all types of managers develop ETFs with appropriate investment strategy disclosure replacing existing rules on index construction and index publication. All funds - active or indexed - might be encouraged to estimate tracking error relative to one or more public benchmark indexes of their choice even if the index serving as an index fund's template is not a public benchmark. Such a benchmark could be the return of an index, the negative of such return and/or a multiple of such return. Funds estimating a tracking error below 1 - 2% annually (after appropriate leverage, directional and volatility adjustments) could be called index funds, because the Commission and the fund industry probably will continue to use such labels. If the experience of a year or two of operation suggests that the fund is unlikely to meet its tracking error objective or if tracking is consistently negative (before fees and expenses), the index fund label would be removed. If a Silent Index was used as the fund template, the prospectus would have to disclose the index creation and dissemination process and some modification of the "index fund" appellation might be required. If an active manager did not want to be constrained by an investment policy linked to any index because, for example, market timing or sector or style shifts were contemplated, the fund could be described as unbenchmarked and the flexible policy (or the salient principles of the strategy) would be disclosed.

We have organized most of the balance of our comments to address critical active ETF issues in what we believe is a systematic way. We do not address all subjects on which the Commission has requested comments. We offer a number of suggestions pertinent to the Commission's procedures that have implications beyond the approval of active ETFs. We comment only on issues where we feel we can offer a useful perspective.

1. Is active management of portfolios desirable?

This question did not appear on the Commission's list in this form, but it is a question that dedicated indexers raise from time to time and it is implicit when anyone sets out to justify the appropriateness of actively-managed ETFs as requested in Section IV.C. Fortunately, most dedicated indexers recognize that if everyone was an index investor, there would be no mechanism for securities price discovery, i.e., for the rational pricing of one security relative to others. Some market participants need to analyze fundamental price and value relationships and act on their analysis for the markets to work properly. While a group like internet stocks can briefly soar far above any rational value, the process does work in the aggregate, and it requires "stock pickers" and bond analysts to make it work.

Every investor should have the right to choose the active portfolio management process over indexing and to take whatever risks and search for whatever returns are available to him. Investors and managers should be able to use active management unless there is a clear public interest case against it.

As some conclusions of the research listed below indicate, some active management techniques seem to work in the investor's interest.5

2. Are active ETFs better than active conventional funds?

This is the principal basis on which the Commission should evaluate the public interest stake in permitting the development and issuance of actively-managed ETFs (or any ETFs, for that matter). The appropriate reference in the Concept Release is to Section IV.C. The active ETF, from the point of view of the investor and from the point of view of the securities markets, is a better fund product for many active management strategies than a conventional actively-managed fund. The active ETF shows essentially the same advantages over conventional active funds which index ETFs have over conventional index funds [Section II.C.] including several advantages that are most significant when a fund is actively-managed:

  • The absence of redundant shareholder accounting at the fund level can reduce investors' costs.
    An increasing share of all types of fund positions is held in traditional brokerage accounts where there is no incremental charge and very little incremental cost associated with carrying such positions. The overall efficiency of the trading, clearing and custody process is enhanced by the ETF structure. The fund is not burdened with the cost of producing and distributing a separate account statement for each shareholder. In the case of index ETFs the efficiency of the DTCC clearance, settlement and custody process has invariably resulted in a reduction in the expense ratio relative to comparable conventional index funds. We expect a similar relationship of active ETF costs to active conventional fund costs. Active funds have higher expense ratios than index funds partly because they often have higher shareholder turnover and more complex and costly account statements delivered to shareholders. Savings on shareholder accounting costs will probably be slightly greater per dollar of fund assets for actively-managed ETFs than for index ETFs.

  • The same tax efficiencies which index ETFs enjoy should be reflected in improved tax efficiency in actively-managed ETFs that use redemption in-kind.

  • Intra-day trading (available in the ETF format) increases shareholder flexibility relative to conventional fund share purchases and sales at net asset value (NAV). A small number of conventional funds with limited intra-day pricing currently exist, but they typically have much higher expense ratios and their costs of providing intra-day pricing are much higher than the evidence of index ETF trading costs suggests that the auction market will impose on active ETFs.

  • The ETF structure separates and implicitly encourages investor analysis of marketing and management services offered and methods of payment for those services.
    This feature might be called transparency of explicit distribution and management costs. Such costs may be hard to separate for conventional mutual funds. They are usually easier to break out for an ETF.

  • Additional data on fund holdings and trades will permit better analysis of manager performance.
    In some respects, the most important development that will accompany the introduction of active exchange-traded funds is the enhancement of information that will permit fund analysts and fund services to evaluate fund managers and investment processes using more extensive information on fund holdings and internal fund transactions.

  • Research into the incidence and investor impact of fund trading costs leads to the conclusion that the in-kind creation and redemption process with all share purchases and sales transactions priced consistently is fairer to all investors than cash fund share purchases and sales at NAV.6 Furthermore, frequent cash purchases and sales of fund shares increases trading frequency inside the fund and tends to transfer wealth from buy and hold investors to active fund traders.7
    The clear superiority - in terms of fairness to buy and hold fund shareholders - of the largely in-kind creation and redemption of fund shares that is characteristic of index ETFs suggests that this process (and variations of the process) should be encouraged in actively-managed funds. Research into this aspect of fund transaction costs has increased in recent years.

    In short, just as the indexed ETF is generally superior to the conventional index fund in terms of cost and fair treatment of shareholders, the active ETF can be a substantially better fund product from the investor's perspective than an active conventional fund under many circumstances. Regardless of the precise format which an active ETF follows, whether it be the Self-Indexing Fund format (explained below) which we propose to use or a wide open, virtually unrestricted active management format, there is certain to be much more information available from active ETFs than from active conventional funds on the cost of buying and holding the fund, on the contents of the fund, on the investment process and on the wisdom or lack of wisdom behind the manager's decisions.

    3. Transparency Cost in Actively-Managed ETFs

    Most investors and fund managers agree that there can be a transparency cost or loss of information value by the holder and manager of an actively-managed ETF. This transparency cost arises because there will inevitably be a greater degree of portfolio transparency in an actively-managed ETF than in a conventional actively-managed fund. This issue is raised in the last paragraph of Section IV.D.3. In fact, greater transparency and a corresponding loss in the value of information the fund's research process develops will be facts of life for some actively-managed ETFs. If transparency is a particular problem with a fund's investment strategy, that fund can still organize as a conventional fund rather than as an ETF.

    Some information on the character of the portfolio held by an actively-managed ETF will have to be revealed to specialists and market makers (and consequently to all interested investors) to encourage dealers to trade in the shares. The Commission is virtually certain to insist that a proxy for the intra-day value of any ETF be disseminated during the trading day. The share value changes reflected in this intra-day information will provide astute analysts with a great deal of information about the content of the portfolio. Future funds may provide less portfolio and pricing information than we describe here, but the level of disclosure required of actively-managed ETFs probably will always be closer to that required of today's equity index ETFs than to that required of today's conventional active funds.

    It is difficult to think of many cases today or in the recent past when a high level of secrecy as to recent changes in the contents of a mutual fund portfolio has been necessary to protect actively-managed fund shareholders from predatory traders who would front run an active portfolio manager's purchases and sales. It is also hard to develop a credible argument that there should be no more knowledge of a fund's holdings than the information provided today by most conventional mutual fund managers. In most cases, a great deal more information could be made public without harm to the fund's shareholders. The Commission seems inclined to require increased disclosure by conventional funds and will soon face decisions on disclosure requirements for actively-managed ETFs. This topic will be and should be the subject of considerable debate. We have already commented on it in the context of necessary changes in the Commission's policy on publication of changes in an index ETF's template index. We will have more to say about it in the context of active ETFs later in these comments.

    We turn now to the investor information advantages of actively-managed ETFs.

    4. How will the increase in information about actively-managed exchange-traded funds help investors?

    These comments relate primarily to Section IV.B.1. A highly desirable feature of the ETF is that a regularly updated picture of the fund will provide more information for fund analysts and fund evaluation services. Even an individual investor with relatively little investment experience can project some of a transparent portfolio's investment characteristics: What will cause it to go up? What will cause it to go down? How will it react to macroeconomic or market events relative to other funds with similar or slightly different features?

    Because their portfolios are more transparent, ETFs can be subjected to many new kinds of analysis by ETF specialty analysts. Traditional fund analysis has not had up-to-date portfolio information to work with and it can only go so far in analyzing the sensitivity of fund share price behavior to a variety of economic and financial market phenomena. ETFs lend themselves to detailed sensitivity analysis - and much more. ETFs can be the subject of intense analysis of such things as the efficiency with which the manager or advisor trades the fund, the inherent efficiency or inefficiency of the fund's underlying index (if it is an index fund) or the relative performance of the stocks an active manager buys and those she sells. It is useful to look briefly at some examples of new approaches to fund evaluation and performance comparisons that can proceed from greater disclosure of fund holdings.

    Fund manager quality is extremely important to investors. In managing an index fund, for example, there are important differences in performance based on the way the manager uses the flexibility available to time the changes that must be made in the portfolio. Over time, the better analysts and fund advisory services will develop a number of measures of management quality for ETFs that are not usable on today's conventional funds. Among the variables measured will be fund trading costs, embedded index trading costs and comparisons of the performance of the stocks an active manager purchases with the performance of the stocks she sells - after the latter leave the portfolio. The increased transparency of fund construction, transaction costs and portfolio management efficiency should revolutionize fund analysis.

    An important criterion for evaluation of the usefulness of a fund to a specific investor is its lack of correlation and corresponding risk offset when used with other investments, including other ETFs, that an investor might hold. For example, a fixed income fund or a foreign equity fund might provide useful diversification for a portfolio that is invested primarily in large capitalization domestic equities, whether the equities are held directly or through another fund. One basic piece of information we would expect an advisory service to provide would be a frequently updated table of past and forecast correlations between and among the various indexed and active portfolios used as the basis for exchange-traded funds.

    A host of other fund features are amenable to quantitative or qualitative evaluation using some traditional investment tools and some tools that will be purpose-built to use the data provided by ETFs. Applying new tools to more extensive data will help to answer questions few have thought to ask about conventional funds because there has been no hope of getting a meaningful answer.

    To summarize, fund analysts should be able to develop answers to questions about:

    For active ETFs

    For all ETFs

    With improved evaluation of manager effectiveness and efficiency, it will be harder for ineffective managers to hide and there will be more objective, up-to-the-minute evaluations available on these funds than have ever been available on active conventional funds.

    5. How Should the Commission approach the approval of actively-managed ETFs?

    These comments relate largely to the Commission's approval process rather than to specific issues of regulatory relief. While, as noted, we believe the notion of a bright line distinction between indexing and active management is not useful, the Commission has already authorized a variety of actively-managed ETFs by permitting the representative sampling technique to be used with non-RIC-compliant indexes to create some actively-managed funds that do not track the nominal index very well. The Commission might decide to reconsider the characterization of these products as index funds, but that is a minor issue. It would be simpler to ignore these funds - which have not been very successful anyway - and move on to active funds. If the Commission feels compelled to act on the non-RIC-compliant index ETFs, it might require these funds to describe themselves as active funds.

    It is important that the Commission change its position that independence of the index provider from the manager of the fund is necessary to implement an effective index management program. As long as all parties take currently required steps to prevent inappropriate use of fund or index information in other trading, the structure and relationship of the fund advisor and the index provider is irrelevant for non-benchmark indexes.

    As Gastineau (2002) argues, the use of benchmark indexes as fund templates may provide a useful marketing tool, but it is often contrary to the financial interests of fund investors. On average, a fund based on a "Silent Index" (the index changes are not published until after the fund makes its portfolio changes) should outperform an otherwise comparable benchmark index fund. Publication of the Silent Index information after the fund has traded does not invite others to trade in competition with the fund and with other index portfolios using the same benchmark. The Commission should address this index disclosure issue as it acts on active ETF management.

    The Commission need not worry about defining or identifying the instruments or concepts behind the labels (mutual fund, ETF, index fund, index, etc.). Existing disclosure requirements and the labeling rules embedded in existing exemptive orders for index ETFs provide an appropriate framework for use of the nomenclature. We do not encounter any confusion among investors and their advisors. The ETF issuers, investment advisors and the press have done an excellent job of investor education on product differences. [Section IV.C., paragraph 2].

    The Commission should adopt a policy of encouraging innovation that not only facilitates but goes beyond actively-managed ETFs. Policies needing Commission attention that are specific to ETF innovation but with wider implications include:

    The most important change the Commission can make in its handling of exemptive requests and its processing of ETF and other product filings is to create an operating framework and an atmosphere in which the Commission staff and financial product innovators work together to bring the most useful possible financial instruments to investors. This new framework should emphasize economics more than precedents. The framework should be designed to avoid the tragedy of complicating or simplifying a product to push it through an acceptable precedential pigeonhole - a process that has crippled many products that "won" approval in bloated or attenuated form.

    An improved filing approval process would help prevent situations where a sound product design is delayed, permitting an inferior product to be launched in the interim. The threat of prompt approval of an inferior product usually forces the better product out of the market. Its designer will inevitably be under competitive pressure to accept the structure that can get the fastest approval. Once a niche is occupied, it is difficult to replace the occupant with an improved product that cannot be launched until several years after the initial product appears.

    The Commission cannot, should not and need not abandon or weaken its efforts to protect investors in order to facilitate innovations that will improve market efficiency, reduce investors' risks and costs and help both individual and institutional investors obtain and process the information they need to make better investment decisions. Modifying the policies listed above and radically changing the procedural and structural mandate for staff action does not pose a risk to investors. These changes should benefit investors at least as much as most recent Commission initiatives.

    We address the specific issues of active ETF fund operation and regulation under several headings. Except as specifically noted, we would advocate a high degree of freedom to innovate in the development of active ETFs.

    How Might Actively-Managed ETFs Work?

    This heading and the two subsequent headings address a variety of operational issues raised in Section IV.B. There are two basic forms which true active management of an ETF can take: one is the Self-Indexing Fund technique that we would also use for a Silent Index ETF, and the other is what we will call a "wide open" active ETF. The Self-Indexing Fund is relatively straightforward, but the latter offers wider choices of features that raise some significant regulatory issues. None of the regulatory issues seems unusual in scope or difficult of resolution.

    All of our subsequent comments are based on one important underlying premise: There is no reason anyone other than the portfolio manager/trader who places an order and the broker who executes it needs to know what specific changes are planned for any fund until after the fund has implemented the changes. These parties, of course, are already constrained by existing rules in their use of this information. If such rules are not adequate for active ETFs, they are probably not adequate for conventional funds.

    The Self-Indexing Fund - Daily Portfolio Disclosure (U.S. Patent Applied for)

    We use the term Self-Indexing Fund because the fund develops, through its published net asset values, an index which we call the Self-Indexing Fund Index. We recognize that this is a slightly different creature than what the Commission is used to seeing in indexes, but we believe it is useful to have this name on the index to protect certain intellectual property rights which the fund will license to permit the issuance and trading of index and fund derivatives bearing the fund's name. Income from such license fees would offset some portion of fund expenses.

    The Self-Indexing Fund manager creates a Self-Indexing Fund index by purchasing an initial portfolio or by accepting an initial deposit meeting her specifications and calculating a net asset value for the fund shares. The basis for the initial deposit and for subsequent changes in the fund may be a formally calculated but unpublished Silent Index or it can be a portfolio basket developed using any of the traditional techniques of active portfolio managers or new techniques devised by the fund manager or by others. Even if the fund is based on a Silent Index designed by an entity other than the fund's advisor, the principal index of interest to investors in the fund and its derivatives will be the Self-Indexing Fund Index which will track any Silent Index used in the fund's management very closely. The template Silent Index will be of public interest only to measure how closely the fund's manager tracks it. Going forward, what matters are the implementation of any index in the fund and the fund's net asset value.

    The Self-Indexing ETF shares are offered to investors and the initial and subsequent intra-day values and net asset values (NAVs) of the ETF shares are published as a Self-Indexing Fund Index that can be used like any other equity index except as a template for another portfolio.8 The pattern of daily values of the Self-Indexing Fund Index is determined by changes in the NAV and the IDPV (Intra-day Proxy Value) of the fund shares. The fund manager changes the composition of the fund index by selling some of the components of the portfolio and/or buying new components for the portfolio.

    The operations of the fund would proceed very much like the operations of the currently available equity index ETFs with many similarities and a few significant differences. The fund manager would disclose the Self-Indexing Fund portfolio daily and post the next day's creation and redemption baskets for public dissemination shortly after the market closes each business day. The fund would create additional fund shares by issuing them in multiples of pre-set numbers of fund shares in exchange for deposits of portfolio components in multiples of the daily creation basket and would redeem fund shares by accepting multiples of a preset number of fund shares in exchange for portfolio components in the form of like multiples of the daily redemption basket. Values are balanced by small cash payments to or from the fund in each case.

    The creation and redemption process is quite similar to the like process for existing equity index ETFs. The principal difference is that the Authorized Participant (the market maker or investor that creates or redeems fund shares) would notify the fund's distributor of plans to create or redeem fund shares earlier in the trading day than notice is currently required for most equity index ETFs. The earlier notice would permit the manager of the Self-Indexing Fund to defer selling stock she will need to deliver as part of the posted redemption basket and to sell incoming shares from creation baskets as she is selling shares in the same stocks to change the fund portfolio.

    The fund management process ultimately makes changes in the raw index value; that is, in the per share NAV or IDPV, and, consequently, in the Self-Indexing Fund Index. To state it another way, the daily closing value of the fund is the principal daily raw index value.9 Intra-day index calculations are identical to cap-weighted index calculations and the cap weights are the portfolio weights used in the NAV calculations. If the fund holds a cash equivalent balance or is leveraged, the index will include a cash equivalent component or will be effectively leveraged. In short, the Self-Indexing Fund Index fully reflects the fund's investment characteristics.

    In contrast to the existing exchange-traded equity index funds that attempt to create a portfolio to track the performance of an underlying benchmark index, the Self-Indexing Fund can use indexed or active stock selection to create a portfolio that serves as the fund portfolio on an ongoing basis, but also is the heart of the index which the fund's positions and its daily net asset value calculations create.. If the portfolio manager is attuned to the principles of fund index construction, the index will be inherently fund friendly to its own fund.

    The fund index can be used as a basis of comparison with other funds and their managers and with other indexes; that is, as a variety of benchmark. It can serve as the basis of a family of derivatives which can be based either directly on the fund (physically settled) or on the index created by the fund (cash settled).10 Through the creation of its index and disclosure of changes in the Self-Indexing Fund and/or index, the Self-Indexing Fund will report its portfolio and index changes within a few hours after the market closes on each business day when a portfolio change occurs.

    The Self-Indexing Fund offers the same kind of intra-day value proxy calculation that index ETFs currently provide and the Self-Indexing Fund's creation basket can serve as a hedging portfolio for the Self-Indexing Fund shares during the trading day. In these respects, the Self-Indexing Fund is designed to look and work as much like the current index ETFs as possible, while providing a vehicle for more transaction cost efficient indexing and for active portfolio management.

    The Self-Indexing Fund offers continuous opportunities for creation of additional fund shares and redemption of existing fund shares in Creation Unit Aggregations. Such Creation Unit Aggregations will approximate the composition of the fund portfolio as of the opening of the market on the current day, at least as closely as an index-replicating ETF's creation basket approximates the fund's template index. The creation basket will consist of the stocks held by the Self-Indexing Fund in very close proportion to the holdings of the Self-Indexing Fund in each stock. For example, if the Self-Indexing Fund holds a 3.1% position in IBM, the Creation Unit will have about a 3.1% position in IBM.11

    In the active Self-Indexing Fund, a fund manager rather than an index publisher creates the index by establishing a portfolio. Whereas, an index publisher might list and weight some stocks, price them at today's market close and list a few index maintenance rules, the manager of the active Self-Indexing Fund, and, therefore, the creator of the Self-Indexing Fund Index is first and foremost the manager of the fund. She thinks and acts in terms of the fund portfolio and the index takes care of itself. The fund's updated portfolio is disclosed in full detail shortly after the market closes each business day. The fund's investment objective and the manager's style define the "index" rules. The fund will be created and redeemed in kind like existing open-end index ETFs, so it will not have to hold material cash reserves to cover redemptions.

    We expect most managers of active Self-Indexing Funds to make occasional significant transactions (with no specific or standard frequency other than a frequency that reflects the manager's style) that will materially change the portfolio and, in consequence, change the index. There is nothing inherent in the portfolio management/index creation process that prevents the manager from making a sequence of gradual changes in the portfolio to move from one security to another over a period of days or weeks. Of course, the fact that portfolio changes will be known daily over that period encourages concentration of transactions on both the buy side and sell side over a relatively small number of days, with the transition from one position to another typically taking place on a single day, unless or until the Self-Indexing Fund becomes very large.

    We would expect asset growth following a high degree of investment success (consistent top quartile performance) to place a limit on further growth of some actively-managed Self-Indexing Funds by reducing their attractiveness to certain investors who are concerned about the fund's performance at a higher asset level. To protect investors, the Self-Indexing Fund's directors may restrict future creations, possibly permitting the shares to go to a premium over NAV in the secondary market. Apart from continued secondary market trading in the ETF, this is essentially what occurs when a conventional open-end fund restricts new purchases, except that all investors are always treated equally in terms of buying new shares in the Self-Indexing Fund case. There might be limited creations to replace shares redeemed and to help the specialist make an orderly market, but there would be no limitations on redemption other than the need to accumulate enough fund shares to make up a standardized fund share Creation Unit Aggregation for redemption.

    To insure that all readers of these comments understand the management and position disclosure process of the Self-Indexing Fund and the hedging information and choices available to market makers, it is necessary to understand the daily accounting process of a fund and the calculation of the fund's daily NAV. The daily net asset value of most U.S. funds is calculated using closing market prices which become available shortly after 4:00 p.m. Eastern time in U.S. markets. Because a portfolio manager may be trading up to and past the 4:00 p.m. nominal close of most equity markets, the manager (and more importantly the fund accountants) may not know until well after 4:00 p.m. exactly what positions have been purchased and what positions have been sold over the course of the trading day and in after-market sessions. The need to provide a net asset value calculation promptly for dissemination to the fund's investors through newspapers and various electronic media necessitated adoption of a procedure whereby the fund portfolio, adjusted for the previous day's transactions, and today's dividends, interest payments, certain accruals and corporate actions, is the basis for today's NAV calculations. Transactions which have taken place on a given day, e.g., the purchase of one stock and/or sale of another, are not reflected in that day's NAV calculation.

    A fund manager who purchases Company A's stock ("stock A") today at an average price below the closing price of stock A and sells Company B's stock ("stock B") at an average price above the closing price of stock B will have added value to the fund and the NAV would be higher if these position changes were incorporated in the NAV calculation. Given the impracticality of including this information in an NAV calculation on the trade date if some of the trades are unreported and even unexecuted, the entire day's trades are incorporated into the portfolio that will be used for the following day's NAV calculation.

    Differences in the prices at which purchases and sales are made relative to the prices used in the NAV calculation are not a cause for concern. It would be very unusual for the NAV calculation of a reasonably diversified fund to be materially affected by the one-day delay in incorporating these transactions. The effect of such transactions can be either positive or negative for the next day's NAV depending upon the prices at which the securities were purchased and sold and the prices for the respective securities used in the trade date NAV calculation.

    Exchange-traded funds have a valuable and well-established tradition of protecting ongoing shareholders from the effect of shareholders entering and leaving these funds - both in terms of costs incurred by the fund portfolio in connection with purchases and sales of portfolio shares and tax costs of exit by selling fund shareholders. To maintain this important tradition, all or nearly all purchases and sales of Self-Indexing Fund shares that involve a transaction with the fund may be in-kind creation deposits and full or partially in-kind redemptions.

    In contrast to some types of funds where extensive use of specialized creation and redemption baskets may occur, most creations and redemptions of Self-Indexing Fund shares may be for standard, identical creation and redemption baskets. There is certainly room for exceptions, but the exceptions are not likely to be particularly important and will consist primarily of situations where a specific creating or redeeming dealer is not able to purchase or sell a specific security for legal or regulatory reasons or where the portfolio manager elects to sell some portfolio shares the Self-Indexing Fund is carrying at a loss for tax purposes rather than include them in the redemption basket. A dealer will be able to create or redeem as many Self-Indexing Fund shares (in multiples of the Creation Unit Aggregation) as it wishes in exchange for each day's posted creation and redemption baskets. Subject only to the possible pre-announced removal of some positions from the redemption basket, both of these baskets usually will reflect the exact contents of the portfolio at the start of the day. Some minor differences due to rounding may be present, but they will not have a material effect on the ability of the Authorized Participant to use the securities in the creation basket as a hedging portfolio.12 All creations and redemptions will be priced consistently with the fund's NAV. 13Subject to the possibility of limiting creations if the fund gets quite large, the Authorized Participants will be able to create or redeem as many fund shares as they wish in the form of each day's posted creation and redemption baskets. There will be no restrictions on redemptions. Apart from rounding and the removal of some positions (for tax purposes) before the redemption basket is posted, these baskets will reflect nearly the exact contents of the fund portfolio at the start of the day for which they are effective.

    This brief description is not intended to provide an exhaustive discussion or description of the portfolio management or accounting processes. Managers will have different ways of using the traditional in-kind creation/redemption process that has become a defining characteristic of most ETFs. Similarly, managers will develop different mechanical ways of incorporating the effect of a day's creations and/or redemptions into their stock purchase and sale plans

    Because any changes in the composition of the fund determine the course of the index, changes in the composition of the fund/index will be announced after the market close on any day when changes are made. An investor wishing to track the fund precisely will have to do so through ownership of fund shares or their derivatives. Although someone can easily compare the performance of another portfolio to the index, the only ways to be sure of tracking it precisely (overnight) will be by holding the fund shares, some of their derivatives or cash-settled derivatives based on the Self-Indexing Fund Index.

    The Self-Indexing Fund Index has a uniquely interesting family of derivatives. The principal option contract will be options on the fund shares - like the highly successful options on the NASDAQ 100 index shares, the most active option contract traded in the United States for the year 2001. Cash-settled index options are not likely to be popular, particularly if share prices of individual funds are relatively high, minimizing the fund options' relative transaction costs. Futures contracts should be physically settled, if the Congressionally mandated rapprochement between the SEC and the CFTC on the subject of single stock futures leads to introduction of that product in the Spring of 2002. The most logical candidates for the first single stock futures contracts are futures on ETF shares. Self-Indexing Fund shares might even be the very first choice for such physical settlement, because they would track both index futures and physically-settled futures very precisely.

    Self-Indexing Funds are not likely to be replacements for all actively-managed, broad-based and widely diversified funds. Apart from their natural Silent Index Fund applications, they are more likely to be specialty actively-managed funds, perhaps managed by a skilled stock picker or a featured analyst who will select favored stocks for the portfolios or a market strategist who might select sector funds and specialty funds in a portfolio-of-funds or fund-of-funds approach. Other applications might include theme funds providing a portfolio manager's perspective on specific investment concepts or stocks recommended by a firm's equity research department. As noted, the index fund version of the Self-Indexing Fund should be based on a Silent Index that does not publish its changes until after the fund implements them. A comparison of conventional indexed ETFs and indexed and active Self-Indexing Funds is summarized in the attached Exhibit 1.

    Exhibit 1 Comparison of Index ETFs and Self-Indexing Active ETF
    Feature Conventional Index ETF Improved Index ETF Self-Indexing Active ETF
    Index Typically standard benchmark Fund-Friendly Silent Index - Designed for fund use Self-Indexing Fund Index
    Source of Index Index publisher Index publisher or fund Advisor Self-Indexing Fund NAV
    Public knowledge of portfolio change Announcement of Index change Announcement of Portfolio/Index change Announcement of Portfolio/Index change
    Time of public knowledge relative to trade 1 week before trade 2 -3 hours after trade 2-3 hours after trade
    Index economics for fund Fund pays license fee directly or indirectly Fund may pay or receive fees, depending on source of index Fund receives license fees for derivative products
    Portfolio calculation algorithm Index change template Index change template Pattern chosen by manager/trader
    Can others trade on information that fund is trading? Yes Rarely, if ever Rarely, if ever
    Can fund use other than market-on-close orders without concern for tracking error? No No, with respect to template index Yes, with respect to fund index Yes
    Does manager control portfolio turnover? No No, with respect to outside index Yes, with respect to Advisor developed index Yes
    Does manager control portfolio composition? No No, with respect to outside index To a limited extent with respect to Advisor developed index. Yes

    Full Active Management - Delayed Portfolio Disclosure

    The Self-Indexing Fund is a relatively simple, though significant, extension of the traditional index ETF. Its significance is that it recognizes the fund's right to make changes in the fund/index before the change is made public. The Self-Indexing structure, nonetheless, protects specialists and other market makers from the risk of mistracking the fund/index to the extent that they plan their hedging activities to be flat in terms of exposure to the fund/index each day at the market close. It is quite possible that some specialists and market makers, recognizing the tendency of the market to snap back from the impact of a fund's transactions - particularly those made at or near the close - will ignore the opportunity to hedge through the market close and simply accept that risk in their market making on the expectation that on average they will be as well off with this overnight exposure as without it.

    With full active management, in contrast to the Self-Indexing Fund, we expect that there will be more limited disclosure of the portfolio composition from day to day. The amount and timing of portfolio disclosure may be mandated by the Commission or it may be left to the issuer of an active ETF subject to current reporting requirements for all funds . We expect considerable debate on this issue, but the disclosure process does not seem to require tight regulation. As will be clear from subsequent paragraphs in these comments, market forces can dictate appropriate portfolio disclosure. In periodic fund shareholder reports, most of the positions passing through the portfolio will be revealed, albeit on a delayed basis. In an actively-managed fund with turnover of as much as several hundred percent a year, the usefulness of even quarterly disclosure of the portfolio composition will be limited. The intra-day share value proxy calculation will continue to be disseminated, perhaps at the 15-second intervals used today, or perhaps at a different frequency chosen by the fund issuer. Very frequent publication of valuations will provide interested parties with a great deal of contemporary portfolio content information because a series of valuations at close intervals will provide good indications to a sophisticated quantitative analyst of the more significant portfolio changes and even of the relative sizes of positions. This ability to derive information on portfolio content changes from fund share value changes does not necessarily support the argument that intra-day share valuations constitute adequate portfolio disclosure. In fact, the ability of some market participants to make superior estimates of portfolio contents from intra-day value calculations suggests that some increase in explicit portfolio disclosure is necessary to level the playing field. A partial solution to this problem might be a requirement that a position must be disclosed within X trading days after the date on which the first shares are acquired or after a manager begins to sell an established position.

    To supplement whatever level of portfolio disclosure is selected, the fund or a service provider might publish a suggested hedging basket of highly liquid positions with a disclosed historic tracking error to be used as the basis for hedging transactions by specialists and other market makers. The usefulness of a hedging basket arises from the fact that less than full portfolio disclosure would make the creation basket a doubtful hedging vehicle for a market maker in the fund shares. A formal hedging basket also will automatically provide supplementary and updated information on the character of the portfolio. The hedging basket will have to be changed when there is a change in the content of the portfolio and the tracking error will be recalculated, providing some information about portfolio content and portfolio changes. The precision of this information will not even approximate the precision of information that is available for today's index ETFs or that will be available for the Self-Indexing Funds. Nonetheless, it is substantially more information than what today's actively-managed conventional mutual funds publish. This suggested minimum level of information should be enough to entice some dealers to make markets in these funds. The fund board (or the Commission) will have to decide on a tradeoff between any fund performance cost of portfolio disclosure on one hand and, on the other hand, the effect of reduced fund share liquidity and higher fund share bid/asked spreads on present and prospective fund shareholders. The manager cannot treat the fund's market makers any better than he treats anyone else with respect to portfolio transparency and he has an incentive not to abuse the market makers, lest they stop making markets in the fund's shares.

    It will take time and experimentation to get from the current equity index ETFs to an acceptable disclosure/trading spread tradeoff for this kind of full active management. Given the nature of U.S. securities laws' dependence primarily on disclosure and with necessary, but probably appropriate, deference to the prescriptive and proscriptive aspects of the Investment Company Act of 1940, the Commission should eventually become comfortable with some combination of (1) less frequent intra-day fund share value proxy calculations, (2) reduced disclosure of ETF portfolio composition relative to existing index ETFs and (3) an undertaking by the active fund manager or a service provider to publish a hedging basket with a disclosed historic tracking error.

    Barring a performance record good enough for regular and enthusiastic discussion by the press and at cocktail parties all over the world, an active manager using this limited disclosure technique is unlikely to raise as much money as a manager using the Self-Indexing Fund variation of active management because of the inevitably wider fund share trading spreads and the difficulty many shareholders will have getting comfortable with this unfamiliar and less transparent ETF structure. Regulators should come to realize that they do not need to control most kinds of transparency. If regulations and enforcement can attack the occasional fraud and the somewhat more common misrepresentations that affect market fairness in various investment products, market makers and investors will determine the acceptable level of transparency as market makers set their spreads and investors decide they are comfortable enough with the available portfolio information and the market maker's spread. In this case, as in most others, the market itself is the most effective and impartial regulator.

    Having described the Self-Indexing Fund which is a small step beyond what exists today and the full-fledged actively-managed fund which has relatively few constraints, we feel it appropriate to comment on these two structures and the strengths and weaknesses of each. As noted, investment company accounting uses the portfolio as it existed after today's trades are posted to determine the intra-day share value proxy and net asset value tomorrow. There is no need to update the composition of the portfolio or change the composition of the creation and redemption baskets at any time during the trading day. This vastly simplifies the entire process and is totally consistent with the way most conventional funds are run today. We are not aware of any funds that update their portfolios on an intra-day basis even if they do calculate more than one NAV per day. While intra-day trade posting might be theoretically possible, it seems unnecessary, considerably more costly than the present system, subject to posting and pricing errors adversely affecting some groups of shareholders and of no apparent incremental value. After appropriate consideration of all issues raised by comments on the Concept Release, we would encourage the Commission to approve carefully monitored experiments in updating and pricing open-end fund portfolios in real time. We believe the Self-Indexing Fund variant or the full active ETF model will be economically more attractive to market participants than real time pricing with the fund as each investor's counterparty, but that is certainly a testable proposition. Assuming the real time updated and priced fund would disclose holdings like a conventional fund does, it would be less transparent and less "market" regulated than any ETF.

    The Effectiveness of Arbitrage

    Members of the Commission staff frequently express concern that the arbitrage process may not have sufficient strength to operate if the smallest change is made in the basic ETF structure introduced with the SPDRs in 1993. This concern is also apparent throughout Section IV and at other points in the Concept Release. In fact, the arbitrage process is very robust and changes are unlikely to do anything more than change the boundaries of the price range within which Creation and Redemption will occur. Changes in these price boundaries will be small in most circumstances.

    Four specific inputs determine how the arbitrage pricing mechanism - which constrains the pricing of ETF shares in the market place - will operate. Three of these inputs are:

    (1) transparency of the contents of the portfolio,

    (2) transparency in pricing the portfolio on an intra-day basis, and

    (3) liquidity of the underlying securities and other instruments in the portfolio.

    Taking these three factors one at a time with respect to the Self-Indexing Fund, we see that transactions are reported daily shortly after the close of trading and the portfolio contents are fully revealed for any time period in which they affect the net asset value calculation or intra-day share pricing. In other words, the portfolio contents may not be known before the trade is made, but they will be known before the trade affects the posted value calculations.

    To make transactions transparent on an intra-day basis while calculating the net asset value in the standard way raises a question as to the composition of the appropriate hedging basket. Should the hedge be effective through today's close or through tomorrow's opening? Any change in intra-day fund tracking information will encourage hedging transactions by market makers which may interfere with the fund's transaction plans. Daily revelation of portfolio changes after the close is clearly the most appropriate mechanism. Post-close disclosure will not encourage arbitrage transactions that may be contrary to the interests of the shareholders of the fund. Arbitrage transactions stimulated by reported intra-day fund transactions might not even provide useful protection to the market makers who would be led to undertake them.

    Active ETF share values will be calculated every 15 seconds unless the Commission permits or requires a change. This is hard to beat for pricing transparency.

    The determination of individual security and portfolio liquidity standards for publicly traded securities should be left to the Advisor/portfolio manager or the Fund board. The manager is obligated to be fully aware of the liquidity constraints which the in-kind creation and redemption process of ETFs imposes on the fund and he should be able to deal with this in full knowledge of the tradeoffs - reduced liquidity means less effective arbitrage pricing and wider bid-asked spreads on the fund shares. Various divisions of the Commission have implemented liquidity standards for publicly traded securities to be used for specific purposes such as indexes underlying options contracts and even ETFs. Such standards have not been useful or consistent. We hope they are not inevitable. Restrictions on fund positions in futures, options and other instruments which cannot be readily transferred in creation and redemption baskets - beyond the restrictions already in place for conventional funds - are unnecessary for ETFs. All relevant market participants are used to conventional fund liquidity and portfolio composition rules and they do not need to make any unusual procedural changes to accommodate to these rules in managing active ETFs. Non-transferable positions, primarily futures contracts, will be disclosed much like any other fund position and they can be taken into account effectively in the calculation of intra-day fund share values and in the construction of hedging portfolios and creation and redemption baskets. The cost of establishing or closing out a futures position in the fund can be part of the cost of Creation or Redemption. To date, all management company ETFs have fairly broad authority to use futures and other instruments. To the best of our knowledge, no one is currently using equity instruments other than securities in ETFs traded in the United States. This will change eventually as a natural part of an experimental and evolutionary process. There are market mechanisms which permit transfer of such positions between Authorized Participants and the fund outside the Creation or Redemption process.14

    As noted, portfolio transparency, price transparency, and liquidity are just three of the four significant determinants of the price range over which a fund's shares will trade without stimulating an arbitrage-motivated creation or redemption. The fourth (and often the most important) determinant of the fund share price arbitrage range is:

    (4) the cost of creation and redemption.

    If a portfolio consists of a relatively small number of highly liquid large cap domestic stocks, the arbitrage range will be extremely narrow. The addition of small cap stocks will increase the cost somewhat as trading spreads will tend to be somewhat wider on these stocks. To the extent that an index is a broad-market index such as the Wilshire 5000 where the small cap stocks are a small part of the total, they need not have a material effect on the creation/redemption spread range if the number of issues in creation and redemption baskets does not exceed a much smaller number such as 1000. As important as average capitalization is to the arbitrage trading spread, on a single unit Creation or Redemption the number of stocks in the basket and the consequent creation and redemption transaction fee may be more significant. The number of clearance and settlement regimes in which the Authorized Participant must operate will significantly affect the cost of creation and redemption of a multinational fund. If an ETF holds shares traded in a large number of markets - that is, in a wide range of countries with diverse clearance and settlement regimes - its arbitrage range will be wide because the obstacles and costs associated with a creation or redemption transaction will be large. In some cases, this clearance and settlement cost will be the most important determinant of the arbitrage range. Having said this, it is important to point out that arbitrage provides a constraint on pricing; it is very rarely the motivation behind a creation or redemption transaction.

    Creations and redemptions occur principally as a result of the specialist and other market makers managing their inventory of fund shares rather than because of arbitrage opportunities. To put the respective profit potentials for market making and arbitrage in perspective, a typical market maker might buy and sell the shares in a creation unit five or ten times between inventory adjustments with average trading profit ranging from a few pennies to a quarter of a point per share on each trading turn. An arbitrage profit of five cents per fund share would be rare in creation and redemption activity. If more specific evidence of the dominance of inventory management as the primary motivation for the creation/redemption process is needed, consider that in 2001, there have been creations and redemptions in the Nasdaq 100 Trust on many business days. Quite often creations from one market maker take place on the same day another market maker is redeeming. Both these market makers expect to earn trading profits, but they cannot both be earning arbitrage profits on closing prices with these offsetting creations and redemptions.

    If the Commission is concerned about the effectiveness of arbitrage pricing for ETFs, the single most important thing it can do is to facilitate the creation of single stock futures markets on the shares of all index and actively-managed ETFs. Among other things, an active single stock futures market will reduce the cost of short selling, tightening the ETF shares' bid-asked spreads, i.e., these contracts will facilitate ETF arbitrage.

    If an active ETF manager is given a range of choices as to what portfolio information to provide, it should be possible to predict her choices from a consideration of the probable outcome of those choices. If she provides no information beyond that provided today in the twice-yearly portfolio report and daily net asset value, there would seem to be little reason to buy that ETF. The specialist and market makers would post such wide spreads that the transaction costs for an individual investor would probably compare to loads charged by conventional load funds and no portion of the investor's cost would be paid out as a sales incentive. The investor's trading cost would be largely due to the uncertainty associated with the market making function. It is hard to imagine significant investor interest in such a fund. An active fund manager who wants to succeed with ETFs is going to have to offer more information. Rather than requiring a specific disclosure structure, it seems appropriate and in the interest of investors for the Commission to establish broad constraints on what can be said and done and what cannot, and let the winning funds be determined by market forces - which will be dominated by investor preferences.

    6. Miscellaneous Comments on Other Aspects of the Concept Release

    At the close of Section II.A.2., the Concept Release states that the revenue generated by securities lending and arbitrage gains from using futures and options may help an ETF offset expenses that would otherwise cause performance to lag behind the performance of the index "because an index does not have any expenses". These statements are not entirely correct. The principal way in which certain index funds outperform their index is by trading at times different from the times the index rules call for a change in the composition of the index. This timing shift may recapture part of the transactional (market impact) expenses which are embedded in the index. In this sense, an index, particularly a widely used index, does have expenses. The expenses just appear (or hide) in a different place. In any fund, the expense ratio does not include brokerage costs, or market impact costs associated with any re-balancing or reconstitution transaction which the fund must make to track the index. The market impact costs of index changes are very hard to measure accurately.

    Securities lending can be important in certain fixed income portfolios and in portfolios holding some non-U.S. stocks. Securities lending is unlikely to be a significant expense offset in a fund that holds domestic U.S. equity securities unless the stocks are both small in capitalization and actively borrowed by short sellers. Futures and options contracts have provided few significant arbitrage opportunities to index funds in recent years.

    In Section III, the second paragraph, the topic of disclosure of portfolio changes is addressed. If a fund issuer wishes to communicate intra-day purchases and sales to investors, the Commission should insist that the entire mechanism for each communication be transparent to all investors. To the extent that this leads to a different method of calculating each day's NAV and intra-day changes in any hedging basket provided to specialists and other market makers, we suspect that this will cause far more problems than it solves. Nonetheless, so long as investors and market makers can see what is contemplated, they can decide for themselves whether they want to use or make markets in the fund.

    At the end of that paragraph, the Release points out that an actively-managed ETF could have a less efficient arbitrage mechanism than index-based ETFs which, in turn, could lead to larger premiums or discounts in the market price of the active fund's shares. This basic notion is certainly true, particularly if there is active trading in the index fund portfolios or baskets or in a related index futures contract. However, with daily post-close transparency in both portfolio and prices and with portfolio liquidity characteristic of large cap stocks, an active ETF with a relatively small number of securities in its portfolio need not have a materially wider fund share arbitrage price range than any but the most active index funds. A fund providing intra-day trading information will probably have a less efficient arbitrage mechanism as measured by the fund share bid-asked spread.

    It is not appropriate or accurate to refer to an arbitrage-determined bid/asked spread as a cause of premiums or discounts. Arbitrage spreads simply define an arbitrage-free trading range. The price will fall somewhere in this range and the share price's position relative to the boundaries of the range may vary. True premiums, and especially true discounts, tend to be relatively persistent and they are usually larger than an arbitrage-determined spread on an ETF. In fact, premiums and discounts exist because the arbitrage process is obstructed, not because the range is wide. Premiums and discounts are discussed briefly below. Open-end ETFs do not have true premiums or discounts.15

    Index-Based ETFs versus Actively-Managed ETFs [Section IV.A.] We noted above that we would define index ETFs and actively-managed ETFs differently from the description in the first paragraph of this section. To the extent that any distinction is necessary, we noted that the distinction should probably be on the basis of tracking error relative to a designated index. The index can either be a major benchmark or the Silent Index described earlier. We see no obvious reason, as will become clearer in subsequent comments, to regulate index and actively-managed ETFs any differently.

    A Discussion of Premiums and Discounts IV.B. We could provide an extensive commentary on premiums and discounts, but we understand that the American Stock Exchange has commissioned a study which will be published shortly to deal with this tedious issue. There is no significant premium or discount on any of the existing ETFs when the prices compared are determined contemporaneously. Any appearance of a material disparity between NAV and market price of an ETF share comes almost exclusively from non-contemporaneous pricing of portfolios and the fund shares. What is needed is more convenient investor access to stock basket and fund share bids and offers and contemporaneous proxy valuations of the fund shares. An investor needs to be comfortable at the time his order is entered, that the current intra-day portfolio proxy value is very close to, or ideally between, the fund share bid and offer in the market place.

    [Section IV.B.1]. The second paragraph asks about the appropriate level of transparency in portfolio holdings to allow effective arbitrage. If the fund has a relatively clear and straightforward investment objective and per share portfolio values are disseminated at 15-second intervals, it should be possible to entice specialists and market makers to participate in the market in a fund even if 10% or even 25% of the portfolio is not revealed immediately after the close on the day purchases or sales occur. Obviously, any reduction in portfolio transparency, any difficulty in hedging or any additional cost of hedging will widen the spread; but relatively free experimentation on disclosure should be permitted.

    In the last paragraph of Section IV.B.1., the Concept Release asks whether frequent disclosure of portfolio holdings would lead to front running of the active ETF portfolio, i.e., whether investors would trade ahead of the ETF and the creators and redeemers of ETF shares. This is turning the problem upside down. Today, there is a great deal of trading ahead of index changes in pre-published benchmark indexes used as templates for existing ETFs. Once a position in the portfolio is revealed, an active ETF is not likely to increase its relative position in those shares except at relatively infrequent and unannounced intervals. To the extent that the fund is popular and growing, creation unit purchasers of fund shares may be buying more of that stock and every other stock in the deposit portfolio. Such purchases are unlikely to have a material and systematic effect on the relative price of a portfolio's positions even if the fund is well known. Certainly, the relative price effect of revealing an active fund's transactions after they are completed would be far less than the effect of revealing a composition change in a popular benchmark index before the fund trades. That is where the "front running" is occurring today - against the investors in index ETFs.

    After a new position in a stock is taken by any fund, market makers are most likely to sell it short to hedge their long equivalent positions in the portfolio stocks inside the fund shares. ETF market makers are even more likely than other market makers to trade from a long position in the fund shares and a short position in the underlying portfolio stocks. The reason for their long position in fund shares is the existence of securities lending demand for the fund shares.

    The same paragraph asks whether an investment advisor to an actively-managed ETF would face a conflict between maximizing ETF performance and facilitating arbitrage by informing the market place of the advisor's investment strategy. The advisor would have no incentive to inform the market place of the fund's "investment strategy" until transactions had been implemented. From that point on, any rush to create additional shares of the Fund or to add the recently purchased shares to another portfolio would accrue to the benefit of the earlier shareholders of the fund. This effect of copycat transactions and the convenience of buying and selling ETF shares would lead most investors to buy the ETF rather than try to duplicate the advisor's portfolio selections. Any reluctance on the part of the portfolio manager to make frequent adjustments to the portfolio would be primarily from concern over the effect of excessive turnover on the taxes and the transaction costs incurred by the portfolio and, hence, by the beneficial owners. None of this activity should have much impact on the arbitrage spread as long as portfolio changes are disclosed shortly after the close on the day of the trade.

    In Section IV.B.2, the question is raised as to whether an actively-managed ETF should be permitted to invest in illiquid securities or instruments (such as futures contracts) that could not be included in a portfolio deposit or redemption basket. Most or all of the existing index ETFs organized as management investment companies have authorization to use such instruments and we see no reason to deprive the actively-managed ETFs of similar flexibility. The non-transferable positions can be valued and incorporated in the hedging basket so that Authorized Participants can hedge effectively, even though the positions cannot be transferred directly in a creation or redemption transaction. Even if these positions cannot be transferred directly, the fund needs to buy when the hedger needs to sell. They are ideal mutual trading counterparties at the moment of Creation or Redemption. We expect innovations in portfolio and hedging basket disclosure to help reduce market makers' spreads by highlighting such natural trading counterparty relationships. If the parties can trade with one another at the price used in the NAV calculation, the issue of direct transferability to and from the fund is irrelevant. It is the responsibility of an Advisor to avoid instruments which create valuation or liquidity problems.

    In the second paragraph of Section IV.C., the question is raised whether there have been any problems with index-based ETFs with respect to confusion about the differences between ETFs and mutual funds and what measures should be taken to address any potential investor confusion. We are unaware of any confusion. Investors have evaluated the differences between ETFs and conventional funds and, as fund flow statistics show, investors are coming down with increasing frequency on the side of ETFs as superior to conventional index funds. The Commission's requirements for distinguishing ETFs from mutual funds have worked satisfactorily and to the benefit of ETFs.

    We are puzzled by the question in the same paragraph about any relationship between trading in ETF shares and market volatility and about any undesirable consequences of such volatility for investors. We note the question not because we have an answer, but because we do not see any reason to expect such a relationship. The principal effect of ETFs on the underlying markets in the positions in their portfolios is to increase trading volume and reduce trading costs per share. While volume and volatility are often correlated, any causal relationship is more likely to go from volatility to volume than the other way around.

    With reference to Section IV.D.3., paragraph 3,we see no reason to change basic rules with regard to conflicts of interest from the rules in place for existing conventional funds. The relief from prohibitions against in-kind transactions between a fund and certain "affiliates" is appropriate because there is little or no scope for abuse with active as with index ETFs. As long as mandated elements of independence in the investment process are maintained, we see nothing in the existing package of exemptions that needs to be changed.

    The suggestion is made in Section IV. E.2., paragraph 1, that, "Because the advisor to an actively-managed ETF would have greater discretion to designate securities to be included in the Portfolio Deposit or Redemption Basket, a greater potential for conflicts appears to exist." We may not be interpreting this comment correctly, but this seems to be the same potential "conflict" position the portfolio manager occupies in every conventional active fund. The fund's advisor has the authority to buy or sell, and we find it hard to imagine any extra potential for conflict here when the price at which the deposit or redemption and the NAV are going to be set is determined by an independent process and Creation and Redemption transactions are effected at a common set of prices. If the issue is "buying" unallocated shares in an underwriting, it should be easy to fix if current rules do not cover the situation. Some clarification of the concern addressed here would be useful.

    With respect to Section IV.F.1., we have long felt that having an ETF share class was a positive development for every category of shareholder in a conventional index fund. Everyone benefits, even if only from the possible reduction of expenses when costs are spread over more assets.

    In contrast to any index fund, an actively-managed ETF will inherently require more portfolio disclosure than a conventional actively-managed fund. Some non-taxable investors in conventional active fund shares may legitimately argue that they are being disadvantaged by this disclosure if it comes from a change in policy after they bought shares. While it is not necessarily the Commission's role to pass on such issues, expected litigation should discourage any fund board from adding an ETF share class to an existing actively-managed fund. A far safer method would be to start a new fund consisting of a single ETF share class or to start a multi-share class, active fund from scratch with full disclosure that the publication of portfolio holdings is going to be extensive because an ETF share class will be issued. We would not be surprised to see new actively-managed funds launched with multiple ETF share classes to facilitate a variety of sales incentive structures. This is another area where the Commission should encourage well-developed experiments.

    Nuveen is grateful for the opportunity to present these comments to the Commission. If you have any questions concerning these comments, please contact the undersigned at 212-207-2006.

    Sincerely,

    Gary L. Gastineau
    Managing Director

    GLG:tjl

    Encls.

    
    
    Footnotes
    1 Nuveen Investments is the sponsor of 121 management investment companies with over $43 billion under management. Nuveen has filed with the Commission two applications for exemptive relief for Exchange-Traded Funds, one for equity index ETFs, and the other for fixed-income index ETFs. Nuveen anticipates that in the coming months it will file an exemptive application for actively-managed ETFs.
    2 SEC Release No. IC-25258 (November 8, 2001), 66 Fed. Reg. 575614 (Nov. 15, 2001) (the "Release").
    3 Transaction costs include, to use the highly respected Plexus Group's breakdown, market impact costs, costs of delay and costs of missed trades as well as brokerage commissions. To the extent an index fund achieves a better execution than the benchmark modification price by more than the amount of any explicit brokerage charges, it can achieve negative transaction costs.
    4 Using only some of the index securities in the fund rather than all and/or changing the weights of the index securities in the fund portfolio.
    5 See Baks, et al. (2001), Chevalier and Ellison (1999), Elton, et al. (1996), Goetzmann and Ibbotson (1994), Marcus (1990), Gruber (1996), Siegel, et al. (2000), Phillips and Kaplan (2001), Wermers (2001), Lo and MacKinlay (1999) and Lo (2001) for what should provide in the aggregate, a useful and balanced appraisal of the effectiveness of active stock selection techniques.
    6 See Chalmers, et al. (2001), Goetzmann, et al. (2001), and Green and Hodges (2002).
    7 See Edelen (1999), Goetzmann, et al. (2001), and Greene and Hodges (2002).
    8 Another portfolio manager could not count on tracking the Self-Indexing Fund Index because the original manager will change the portfolio and the index before the second manager knows the change was contemplated.
    9 The raw index value will be the most widely used version of the index. A total return version, adjusted for income and capital gains distributions may be calculated for comparison with other indexes and funds and, in some cases, for calculation of settlement values on certain derivatives. We note that some existing equity indexes are not adjusted for dividends and some total return indexes do not publish a separate price-change-only version.
    10 Based on precedent, there may be regulatory reluctance to embrace derivatives on an index with changes that are not disclosed until after a fund which both creates the index and is based on the index has transacted in its portfolio. Any shortcomings in the rules to prevent improper trading should be easy to fix.
    11 Whereas the current index ETFs' creation baskets are designed to replicate the index as closely as possible and the fund weights may be modified from the creation basket weights to bring the fund even closer to the index, we would expect most active managers to manage their funds as strict multiples of the creation basket. The manager would manage the creation basket, knowing that she would have no odd or fractional share positions to keep track of. If there were 10 shares of a stock in a creation basket and the fund had 90 creation units in the portfolio, the fund would have exactly 900 shares of that stock. In short, trading to make small portfolio adjustments to get the portfolio closer to the index will not be necessary.
    12 If the fund manager manages the creation basket as described in the prior note, the creation basket plus or minus some cash will be a perfect hedging portfolio.
    13 Ideally, for the protection of both ongoing investors and Authorized Participants (market makers) creations and redemptions would be cut off and the NAV would be calculated at 3:00 p.m. Eastern time. The IDPV based on the opening portfolio for the day would be disseminated until the close of trading at 4:00 or 4:15 Eastern time. Apart from Commission or staff approval for this change, the exemptions now granted for index ETFs should be adequate for operation of Active Self-Indexing Funds. The Self-Indexing Fund could operate without this shift in NAV calculation time and ability to continue to designate an IDPV value based on the prior closing portfolio, but the allocation of costs would be less equitable and the arbitrage process would be slightly less effective. This is an example of the earlier reference to the need to permit the issuer to designate product characteristics.
    14 These mechanisms include exchange of futures for physicals (EFP) transactions and simple futures trades at prices related to securities market closing prices between the securities market closing time and the futures market closing time. This issue is discussed again briefly below in our specific comments on Section IV.B.2. in the Concept Release.
    15 The oft-used example of the Malaysian WEBS demonstrates this point. The fund was not open to in-kind creations and redemptions when the fund shares failed to sell very close to an arbitrage-determined parity with the underlying portfolio.
    
    

    References:

    1. Baks, Klaas P., Andrew Metrick and Jessica Wachter, "Should Investors Avoid All Actively Managed Mutual Funds? A Study in Bayesian Performance Evaluation," The Journal of Finance, February 2001, pp. 45-85.

    2. Chalmers, John M.R., Roger M. Edelen and Gregory B. Kadlec, "On the Perils of Financial Intermediaries Setting Security Prices: The Mutual Fund Wild Card Option," The Journal of Finance, December 2001, pp. 2209 - 2236.

    3. Chevalier, Judith and Glenn Ellison, "Are Some Mutual Fund Managers Better Than Others? Cross-Sectional Patterns in Behavior and Performance," The Journal of Finance, June 1999, pp. 875-899.

    4. Edelen, Roger M., "Investor Flows and the Assessed Performance of Open-End Mutual Funds," The Journal of Financial Economics, Number 53, 1999, pp. 439 - 466.

    5. Elton, Edwin J., Martin J. Gruber, and Christopher R. Blake, "The Persistence of Risk-Adjusted Mutual Fund Performance," The Journal of Business, April 1996, pp. 133-157.

    6. Gastineau, Gary L., "Equity Index Funds Have Lost Their Way," Forthcoming in The Journal of Portfolio Management, 2002.

    7. Goetzmann, William N. and Roger G. Ibbotson, "Do Winners Repeat?" The Journal of Portfolio Management, Winter 1994, pp.9-18.

    8. Goetzmann, William N., Zoran Ivkovic and K. Geert Rouwenhorst, "Day Trading International Mutual Funds: Evidence and Policy Solution," Forthcoming in The Journal of Financial and Quantitative Analysis, 2001.

    9. Greene, Jason T. and Charles W. Hodges, "The Dilution Impact of Daily Fund Flows on Open-End Mutual Funds, Forthcoming in The Journal of Financial Economics, 2002.

    10. Gruber, Martin J., "Another Puzzle: The Growth in Actively-Managed Mutual Funds," Journal of Finance, July 1996, pp. 783 - 810.

    11. Lo, Andrew W., "Personal Indexes," The Journal of Index Issues, 3rd Quarter 2001, pp. 26 - 35.

    12. Lo, Andrew and Craig MacKinlay, A Non-Random Walk Down Wall Street, Princeton, N.J., Princeton University Press, 1999.

    13. Marcus, Alan J., "The Magellan Fund and Market Efficiency," The Journal of Portfolio Management, Fall 1990, pp. 85 - 88.

    14. Phillips, Don and Paul D. Kaplan, "The Greatest Return Stories Ever Told": Comment," The Journal of Investing, Fall 2001, pp. 94 - 95.

    15. Siegel, Laurence B., Kenneth F. Kroner, and Scott W. Clifford, "The Greatest Return Stories Ever Told," The Journal of Investing, Summer 2001, pp. 91 - 102.

    16. Wermers, Russ, "The Greatest Return Stories Ever Told": Comment," The Journal of Investing, Fall 2001, pp. 92 - 93.


    Equity Index Funds Have Lost Their Way

    (Forthcoming in the Journal of Portfolio Management)

    Gary L. Gastineau
    Managing Director
    Nuveen Investments
    10 East 50th Street, 31st Floor
    New York, N.Y. 10022

    Tel: 212-207-2006
    Fax: 212-207-2083
    Email:
    gary.gastineau@nuveen.com

    Abstract

    Today's equity index funds get more credit for low costs and high efficiency than they deserve. The original idea of a low-cost, low-turnover, diversified portfolio has been replaced by funds that deliver only on their pledges of low turnover and diversification. The concentration of trading by these funds at times when the index is changed causes investors to incur transaction costs far higher than the pioneers of indexing anticipated. Index funds based on popular equity benchmark indexes have become a much more costly way to invest than they used to be - and than they need to be.

    Introduction

    Peter Bernstein's history of the development and application of the great ideas of finance, Capital Ideas (1992), makes it clear that index funds were part of a broader plan. The unifying objective seems to have been to replace the traditional trust department dog-walking and stock-picking process with portfolios that had more diversification and a more "scientific" construction. Performance and diversification were as important as lower operating costs in the minds of many early practitioners of Modern Portfolio Theory, but cost reduction was on the minds of all the early indexing advocates.

    The first indexed portfolio launched in 1971 by Wells Fargo was created for a single pension fund client. In 1973, Wells Fargo organized a co-mingled fund for trust accounts. In 1976, the funds were combined and the capitalization-weighted S&P 500 Index was used as the template for the combined portfolios. By 1977, Wells Fargo had commissioned a study of the feasibility of moving beyond the S&P 500 to the Wilshire 5000.

    It is one thing to persuade pension funds to adopt indexing. Introducing the idea to individual investors was an even more daunting proposition. The idea caught on with investors thanks to some influential advocates. For example, in the first edition of his best seller, A Random Walk Down Wall Street, in 1973, Burton Malkiel called for "A New Investment Instrument." He said, "What we need is a no-load, minimum-management-fee mutual fund that simply buys the hundreds of stocks making up the broad stock-market averages and does no trading from security to security in an attempt to catch the winners."

    Paul Samuelson (1974) set down some arguments he had been making among the investment community in this journal. He noted that, "The only honest conclusion is to agree that a loose version of the `efficient market' or `random walk' hypothesis accords with the facts of life." Samuelson questioned why no money management organization offered an unmanaged diversified fund to the public. He believed that this could be done at relatively modest cost and that the fund would probably be a better repository for his savings than most actively-managed funds.

    Less than a year later, Charles D. Ellis (1975), in one of the most widely cited papers in the literature of finance and tennis, marshaled some simple facts illustrating that the institutionalization of the equity markets in the 1960's and early 1970's had made it probable that the average institutional investment manager would typically underperform the market as measured by a representative index. The costs of trading actively-managed institutional portfolios and paying administrative expenses and management fees - combined with the increased institutional share of the market - left too little stock in the hands of nonprofessional investors to let amateurs fill the ranks of underperformers. Average active institutional investors were inevitably going to underperform the unmanaged market indexes over time.

    With the implicit or explicit support of Malkiel, Samuelson, Ellis and others, John Bogle of Vanguard launched the first broad-market index fund for retail investors in 1975. Bogle was as motivated by the desire to reduce costs then as he is today. Neither Bogle nor his supporters could have anticipated the success of indexing - or the costs which the index management and publication process would impose on today's index fund investors.

    The Role of the Index Publisher

    One rarely discussed cost-saving feature that clearly attracted the indexing pioneers was access to the work of index publishers at very low cost. An index fund manager did not need a research department or conventional Street research. Most of the information necessary to "manage" an index fund was available at low cost from index publishers. The earliest market indexes were developed as market benchmarks or performance measurement gauges. No one had index fund applications in mind. Any revenue from a license sold to an index fund was pure gravy for the publisher and the fund's implied endorsement of the index was helpful in making other licensing deals. The limited information available on early index fund licensing fees (which have typically been confidential) makes it clear that the index providers were not paid high rates to provide an index fund's database and portfolio template.1

    The index publisher inevitably performed some administrative functions for a fund manager. The index rules determined when the index and, hence, the fund portfolio would be modified. In the early years when index funds held a very small fraction of the capitalization of each stock in an index, there was little reason to be concerned about competition for a limited supply of shares when a change was made in the index. As index funds grew, investors and analysts began to pay attention to index membership effects. It became clear that index membership carried implications for a company's stock performance.2

    Low turnover of the index's constituent companies was an inherent part of the indexing strategy. Tracking a low-turnover index with a fund that was growing at a modest, but steady, pace meant that the fund was buying stock relatively consistently and selling a few positions completely from time to time when stocks were dropped from the index. Portfolio turnover and its associated costs were quite low, and tax efficiency was much better than active managers were achieving. If a stock experienced problems, its weight in the portfolio was its weight in the index, and no individual was blamed for the stock's impact on fund performance.

    The index fund pioneers had achieved a number of things at once: relatively objective portfolio selection criteria, cost savings in fund operations, the limited portfolio management direction they needed, low trading costs from low portfolio turnover and a high degree of natural tax efficiency. We have found nothing in the early indexing literature to indicate that the tax efficiency of index funds was anything other than serendipitous, suggesting that it is possible to be lucky and smart at the same time.

    No one can do more than roughly estimate the total assets of indexed portfolios in the United States today. If a manager does not announce that a portfolio is tracking an index, there is no basis for the index publisher to collect a license fee. This simple statement suggests that most estimates of indexed assets are likely to be low. Morgenson (1997), cites an S&P estimate that 8% of the U.S. stock market's capitalization was indexed, presumably as of the end of 1996. Given the S&P 500's dominant share of U.S. large-capitalization index funds and the continuing growth in the Vanguard S&P 500 fund, the SPDRs and other S&P 500-indexed portfolios through the end of 2000, it is a safe assumption that at least 10% of the shares of any stock in the S&P 500 index is now held in an S&P 500 indexed portfolio. There are higher estimates for the market share of indexing. Bernstein (1992), for example, used 30% for institutional portfolios. The precise figure is not critical for our purpose because some "passive" portfolio managers take their time adjusting portfolio composition to changes in index composition. The important point is that it is safe to assume that at least 10% of the shares of any stock in the S&P 500 are held by committed indexers who change their portfolio at approximately the same time that the index changes. These hard-core indexers are part of the problem of growing transaction costs in index funds.

    Recognition of Index Fund Transaction Costs

    Enough has been said and written about the turmoil frequently associated with constituent changes in the S&P 500 and the Russell 2000 indexes that there is no need to reprise the stories here.3Several general comments are in order, however. Most observers agree that market roiling episodes associated with index changes were at their worst in 1999 and 2000 and that some of the most serious incidents involved S&P 500 stocks with restricted floats. The appropriate way to analyze and evaluate the effect of these index changes on indexing portfolios is to view them as transactions and to try to measure their costs as we would measure the transaction costs incurred by an active portfolio manager.

    We have encountered two common reactions to discussions of the transaction costs indexed portfolios incur when they trade to match changes in the index. The first reaction is that there are no transaction costs. The argument for no transaction costs goes something like this. If an index fund manager modifies an S&P 500 portfolio with market-on-close orders on the day the index change is effective, the portfolio will track the index perfectly. Unless the fund pays a commission or makes its transactions at some time other than at the market close on the day the index changes, fund performance in terms of index tracking will not be adversely affected by any transaction costs. One weakness in this argument is that there are nearly always opportunities to transact at better prices from before the announcement of an index change up to a week or so after the index change becomes effective. As evidence that there are transaction costs associated with an index change, some indexers - Vanguard being the most widely discussed - regularly outperform the S&P 500 by trading at different times than the moment of index change.4They capture some of the transaction costs embedded in the index reconstitution process and improve on the index return for their shareholders. Some managers indicate privately that they capture up to 100 basis points per year in the Russell 2000 reconstitution. The resulting favorable tracking error adds to the fund's performance and increases the popularity of the Russell 2000. Indexers as well as active managers prefer the Russell 2000 as a small-cap benchmark because the embedded transaction costs make it easy to beat.

    Investors have widely varying attitudes towards tracking error. Some passive investors are willing to live with tracking error when their portfolio is compared to the index. To reduce transaction costs, they often evaluate blocks of stock that are bid for and offered in the market and buy or sell only when their measured transaction cost is negative. They are providing liquidity for an anxious party on the other side of the trade and generally outperform their benchmark index. This kind of patient trading can be particularly effective with smaller stocks.5For marketing and, sometimes, for legal reasons, many index fund managers are constrained in their ability to deviate from precise index replication. These managers tend to trade when the index "trades."

    The most common reaction to the notion of index fund transition/transaction costs is agreement that transaction costs are present in an index change, and that they should be measured by taking the difference between prices at the time of the index change announcement and closing prices on the day of the official index change. This is a sensible approach. The most widely used measures of transaction costs start the cost measurement clock running when the decision is made to trade. In their simplest form, these measures incorporate the net price change and any commissions in the trading cost calculation. The indexer's decision to trade is "made" when the index change is announced. While the first opportunity to trade is at the next day's opening, a better case can be made for using the closing price just before the announcement as the starting point.

    The greatest problem with either of these starting points for a transaction cost calculation is the growth of a cottage industry that attempts to anticipate changes in the indexes before they are announced.6At any time a company is scheduled to leave the S&P 500, there are probably 10 to 20 serious candidates to replace it. Some of these can be set aside or moved higher on the list for one reason or another. If a candidate is in the same industry as the departing company or in an underweighted industry, it is more likely to be chosen than another company of similar size that S&P has not used in its smaller cap indexes. If a company is in the S&P 400 (Mid-Cap) Index, it is more likely to be chosen than a similar company that S&P has not used in its indexes before. If a company is a recent IPO without earnings, it is less likely to be promoted to the S&P 500 unless its capitalization is too large to ignore.

    The effort to predict the next addition to the index can lead to transactions and price impact on the new index member's shares well ahead of the index change announcement, creating a situation that does not fit easily into a transaction cost analysis framework for any of several reasons:

    Setting aside the difficulty of estimating the transition/transaction costs with precision, there clearly are incremental transaction costs relative to other ways of reconstituting an indexed portfolio. Looking at the price changes in other ways demonstrates that there are very real trading costs associated with an S&P 500 index change.7We need to focus on the probable size of these transaction costs relative to the total expenses of a fund, accepting the principle that a rough estimate is better than no estimate at all. The problem of measuring these usefully and consistently needs more attention, including attention to the cost of trading stocks that stay in the index, but are re-weighted as a result of index changes.

    Comparing Costs of Index Funds

    We will illustrate the cost of some different indexing techniques by comparing two pairs of funds - one large-cap pair and one small-cap pair - constructed and run along radically different lines.

    
    

    Exhibit


    Hypothetical Annual Index Fund Cost Comparisons - Benchmark vs.
    Fund-Friendly Indexes
    (All Numbers Are Basis Points)

    INDEX

     
    Fund Cost Element S&P 500 Fund-Friendly "500" Russell 2000 Complementary Small-cap
           
    Fund Expense Ratio 10 bp 20 bp 25 bp 30 bp
    Fund Transaction Costs Annually (Index Efficiency) 50 - 100 Less than 25 200 - 300(3) Less than 100
    Index Membership Effect (1) ? ? ? ?
    Value Added by Fund Manager (2) ? ? ? ?
             
    Annual Range of Determinable Cost Elements 60 - 110 Less than 45 225- 325 Less than 130
    Cost of Trading Fund Shares 10 20 20 30


    1. Historically, there has been a positive effect on the performance of stocks when they join the S&P 500 and, based on activity connected with the annual reconstitution, when they join the Russell 2000. There also has been an ongoing S&P 500 membership effect, probably as a result of the growth in indexing. If the arguments we make here have any impact, the ongoing S&P 500 membership effect should be a historic artifact.

    2. Aggressive index fund managers have traditionally added value by trading at a different time than the official moment of index change. By doing so, they have recovered part of the annual transaction costs. Based on the performance of Vanguard and others, the best managers seem to have recovered 10 - 20% of the transaction costs in the S&P 500 and as much as one-third of the Russell 2000 transaction costs. This transaction cost recovery relative to the formal index value is not a zero-sum game. More managers should be able to beat the benchmark index (before expenses) than underperform the index. This opportunity to recapture transaction costs will largely disappear when the fund-friendly indexes are adopted.

    3. Part of the evidence for the extremely high re-balancing cost of the Russell 2000 is the finding by Julian (2001) that the stocks graduated from the Russell 2000 to the Russell 1000 in 1990 had a return of 15.6% through the end of 2000 versus the 11.8% return on the Russell 2000 itself. It is not possible to divide this outperformance between the superior performance of the stocks that "graduated" from the 2000 to the 1000 that year and the transaction costs associated with the annual reconstitution; but there is widespread agreement that the annual cost of the reconstitution has been rising as use of the Russell 2000 as a small-cap fund template has grown.
      
      
      

      In the Exhibit, we attempt to compare hypothetical relative cost efficiencies of two pairs of exchange-traded funds. One fund in each pair is based on a benchmark index and the other on a fund-friendly index designed specifically to replace the benchmark in fund applications. In the first comparison of the S&P 500 to the "Fund-Friendly 500", the latter has enough companies to cover 85% of the U.S. market as measured by capitalization of publicly traded companies with a market value over $100 million. This works out to be just over 500 companies. It is worth noting that the S&P 500 includes only about 75% of U.S. companies by capitalization. The second comparison pits the Russell 2000 against a "Complementary" Small-Cap Fund-Friendly index which includes the remainder of the U.S. corporate equity capitalization from the largest U.S. company not included in the Fund-Friendly 500 down to the $100 million capitalization level. The two fund-friendly indexes cover all U.S. publicly traded companies with capitalizations over $100 million or more than 99% of total U.S. equity market capitalization. The new Fund-Friendly indexes are float-weighted with annual re-balancing and various rules designed to reduce turnover.

      We assume the benchmark index funds have two apparent advantages over the fund-friendly index funds. The expected fund expense ratios for the benchmark index funds are lower, and the cost of trading the fund shares/index basket (which is in large measure a function of the standardization of the portfolios) is presumed to be lower for the benchmark index funds than for the fund-friendly index funds. Focusing on the large-cap funds, the expense ratio on the S&P 500 fund is assumed to be 10 basis points (reflecting the expense ratios of 9.45 basis points on the iShares S&P 500 Fund and 12 basis points on the SPDRs) whereas the Fund-Friendly 500 is assumed, at least initially, to be a smaller fund. With fewer assets to spread expenses over, we assume an expense ratio of 20 basis points. We could assume that the manager of the new fund would match the expense ratio on the S&P 500 funds, but the manager might decide instead to stake the case for the new fund on performance after full expenses.

      The trading spread for an S&P 500 standardized basket or for comparable fund shares has recently averaged around 10 basis points. Although the Fund-Friendly 500 should track the S&P 500 very closely most of the time, we assume that the trading spread for these shares - until the new basket becomes widely accepted - will be about 20 basis points. Fund and basket trading cost differences are unlikely to be materially greater than Exhibit 1 indicates because of the inevitable close correlation of the two large-cap indexes and the larger average size of the significantly weighted companies in both of the Fund-Friendly indexes. With an aggressive specialist, the new fund shares might match the trading spread on competitive benchmark funds from the beginning.

      Certain other effects such as any index membership effect and the value added by a fund manager in timing transactions differently from those required in the index rules, we treat as unknowns. There is reason to believe that the net of these effects, in contrast to some past periods, might favor the fund-friendly indexes, but any effects should be modest in either case.

      The remaining difference between the two indexes in each index pair is in the expected internal fund transaction costs. In the case of the benchmarks, the 50 to 100 basis point estimate for the S&P 500 and the 200 to 300 basis point estimate for the Russell 2000 are rough estimates for recent annual transition/transaction costs for funds based on these indexes. Trading costs to modify and re-balance S&P 500 portfolios probably exceeded 100 basis points in 1999 and ran closer to or even below a 50 basis point annual rate for the first nine months of 2001. The actual transaction costs may average higher than the estimates if index managers underestimate the importance of market impact on both sides of an index fund internal reconstitution transaction. When a new company is added to the S&P 500, the nature, if not the precise magnitude, of the market impact on the stock of the new company is obvious. Less clear, but often quite important, is the market impact on the other side from transactions in the stock being removed from the index and in the 499 other stocks whose weights are changed. The transactions on both pieces of the trade occur at a time and in an environment where they represent one-way transactions. There is likely greater market impact in a known one-way transaction than in a randomly timed transaction. These are one-way transactions because many investors want to change their position in specific stocks in the same way at about the same time. Natural liquidity providers tend to be scarce at such times.

      The very low transaction cost estimates for the two fund-friendly indexes reflect the fact that these new indexes are designed to minimize transactions at all times and to offset as much of transaction costs as possible by trading between complementary fund portfolios at the annual fund re-balancing. An S&P 500 fund will engage in a large number of relatively small re-balancing transactions over the course of a year. Transaction costs are incurred regularly in contrast to being largely offset in an annual re-balancing that ideally consists of non-impact trading between complementary funds. In the case of the Russell 2000, the greater relative (to capitalization) popularity of the Russell 2000 than the Russell 1000 means that the market impact of transactions to re-balance the Russell 2000 is much greater than in its complementary large-cap index, the Russell 1000. Also, in contrast to the Complementary Small-cap Index, the Russell 2000 is a middle-range index in the sense that it gains and loses some sizable companies not only at the top, but also at the bottom. Rather than being representative of a capitalization range, it accepts a fixed number of companies each summer. This structure leads to a 25% to 30% annual turnover at re-balancing. Because of the weighting disparity in the use of the Russell 1000 and the Russell 2000 in indexed portfolios, this turnover is reflected in the 200 - 300 basis point estimate of embedded annual transition/transaction costs for the Russell 2000.

      The fund-friendly indexes are fully complementary. The Fund-Friendly 500 would cover the largest 85% of the capitalization range, and the Complementary Small-cap index would cover the remainder. At the bottom of the small-cap index, the impact of interaction with companies moving up from the below the $100 million capitalization range or down into it would be relatively modest, and - with a buffer capitalization range between the small-cap and large-cap indexes - turnover would be well below the annual two-way migration between the Russell 1000 and the Russell 2000. If a fund issuer were responsible for both large-cap and small-cap funds with proportionate capitalizations, or if the overall adoption of the two indexes were approximately proportional to their capitalization weights, the market impact of the re-balancing transactions involving switches between large-cap and small-cap would be negligible on the combined funds -making the transition/transaction costs even less than the estimates in the Exhibit. Even if the adoptions were not ideally weighted, it would be a long time before market impact costs approached recent levels for the S&P 500 and the Russell 2000, because one feature of the fund-friendly index rules in this application is that index changes are not announced until after a fund has traded. Given the cost-saving opportunities that development of such indexes appears to offer an index fund, product developments along this line seem inevitable. Indexers will surely want to return to the low transaction cost environment of their early years.

      The Outlook for Index Transition/Transaction Costs

      The dominant portion of the transaction costs that S&P 500 and Russell 2000 index funds experience is attributable to all the company that the funds indexed to these benchmarks have whenever they need to transact. From the time index funds were introduced until a few years ago, market impact costs from index modifications were a non-issue for index fund managers. The success of indexing and consequently anticipating and trading in competition with other index fund reconstitution trades, is the direct cause of the growing magnitude of these transaction costs. Only the end of growth in equity fund indexation or a radical change in the indexes used (and how they are used) will keep the benchmark index fund internal portfolio transaction costs illustrated in the Exhibit from continuing to grow.

      The indexing process must change. The index fund manager needs an index that has modification rules that are as independent as possible from the rules used by other indexers operating in the same general market space (e.g., large-cap, small-cap, sector or style). We will elaborate on this point and offer a new operating format for index funds to use the modified indexes most efficiently.

      Index Transparency - An Unnecessary Feature of Today's Index Funds

      The public at large has been fully aware of any changes scheduled to be made in an index fund portfolio before the index change is implemented. This is not to say, of course, that the manager of the index fund will make the change at the close of trading on the day that the index changes. Even the "manager" of the original SPDR had considerably greater flexibility than an exact timing match of index and fund changes.8Within some period, usually beginning about a week before the date of the index change, the world at large knows that a change will be made in an S&P index and that the corresponding change in the fund will be made either shortly before, at or shortly after the market close when the index change becomes official. Astute index managers have frequently (in some cases, consistently) traded early or late to avoid the crush at the market close on the day the index itself changes, so there is nothing new in making the change at a time selected by the manager. However, whether the management of a fund is based on a set of index rules or on the stock picking judgment of a fund manager, there is no reason why anyone other than the fund manager needs to know when the index change or fund change will be made or even what the change will be. This is a version of indexing not widely discussed, so it will take some getting used to.9

      Indexing was developed to create a fund that represented all or an important and consistent segment of the market without using traditional stock selection techniques that led to high turnover. The index never needed to be constructed in a fishbowl and published before the fund acted. Investors cared about the general characteristics of the fund, not its specific portfolio. Publication of index changes before fund portfolio changes puts the fund and its shareholders at a disadvantage if a large number of investors want to act in the same way at the same time. The pioneers of indexing used benchmark indexes because they were available; they could be licensed cheaply; and they seemed to be more objectively determined than some of them have turned out to be. At the time they were introduced, index funds were such a small factor in the market that no one thought about the market impact cost of making an index fund portfolio change when the index changed. Markets have changed, though, and indexing has become too popular for us to continue to operate in the old way. The principle of confidentiality of fund trading plans until after the trade is complete has served active managers well. It also works well in a new breed of index fund that attempts to eliminate unnecessary transaction costs.

      New and Improved Index Funds

      The characteristic of portfolio change that defines current index funds is that, before the index change date arrives, there is widespread knowledge that the fund will trade to match the index change. The index provider determines and publishes what the fund will have to buy and what it will have to sell. In other words, knowledge of the pending index change belongs to the world, not exclusively to the fund.

      Active fund managers go to great lengths to keep their intentions and transactions confidential until long after a trade has been executed. With the exception of some pressure to report the positions they have already taken more promptly after the trade, no one argues that an active manager should have to disclose more about trading activity. No one suggests the active manager announce plans to buy one stock to replace another before the trade has been completed. The result of the active fund manager's transaction may not be made public for months, even after some proposed changes to accelerate portfolio disclosure are implemented.

      The current index investing principle requires that the manager of a fund tracking a benchmark index not even find out about a change in the index before the change is announced to the world. The more popular and widely used the index, the more managers and scalpers will be trying to implement the same index change at about the same time. In fact, the more popular the index, the more traders who will be trying to make a similar trade before the index fund manager - perhaps attempting to reverse their position at a profit when the index fund manager makes the trade. Such procedures make some sense if an index is used for many purposes; but there is no reason any index fund must be based on a widely used index - if using that index for the fund harms the fund's investors by raising the fund's transaction costs, and causes the index and the fund to underperform a similar fund based on less congested index transition rules.

      The higher reconstitution transaction costs for funds based on the benchmark indexes in the Exhibit happen largely because other indexers and other investors enter orders to make similar portfolio changes at about the same time as the fund trades. Financial industry procedures and, to a limited extent, regulations on index change disclosure have unnecessarily and inappropriately required identical treatment and pre-announcement of index information, whether the index is a major benchmark index or a custom index developed for a single fund. Disclosure before the index change is effective is certainly necessary and appropriate for major benchmark indexes - but there are very few major benchmark indexes among the tens of thousands of equity indexes calculated daily. Disclosure is not appropriate, necessary or desirable for a fund-specific index. If indexing had started with customized fund indexing rather than benchmark indexes, publication of changes in the index used by a fund before the fund trades would never have been an issue - and properly measured reconstitution trading costs for most index funds would be far lower than they are today.

      Even if changes, annual reconstitutions and/or re-balancings in major benchmark indexes have to cause trading frenzies, there is no reason less prominent indexes have to announce and implement their changes in the same manner. The shareholders of a fund based on a custom index do not need to know the precise index rules for handling a corporate action anymore than shareholders in an actively-managed fund need to know (or get to know) exactly how their manager will react to an earnings report or a management change. If an index is designed to be fund friendly - in this case, to reduce the fund's transaction costs - it could be very advantageous to the fund's shareholders and not adversely affect other parties with legitimate interests, to announce index changes only after the change has been implemented by the fund.

      This fundamental change in the revelation of index changes has surprisingly few implications for how indexes are used for purposes other than as templates for index funds. If the implementation of a fund's index is reflected in the fund portfolio, the changes in the fund's net asset value are the best measure of changes in the index. The fund might be required to publish information on the conformity of its implementation with the rules it adopts for its index, but derivatives on the fund shares and on the NAV-based index could trade without advance notice of index/portfolio changes.10

      The Self-Indexing Fund - Daily Portfolio Disclosure

      A form of exchange-traded fund structure which we call the Self-Indexing Fund (U.S. patent applied for) was originally devised as an actively-managed ETF structure. It turns out to be the best mechanism for reducing reconstitution/transaction costs in an index fund as well. The Self-Indexing Fund manager establishes the initial composition of the Index by purchasing an initial portfolio or by accepting an initial deposit meeting certain specifications. The basis for the initial deposit and for subsequent changes in the fund may be a formally calculated but unpublished index or it can be a portfolio basket developed using any of the traditional techniques of active portfolio managers or new techniques devised by the manager or by others. The template index for an index fund will be of interest to anyone who would measure how the fund's manager tracks it and how the template index performs relative to standard benchmark indexes covering the same capitalization range, style or sector. While there is room to add value with sound trading judgment, there will not be scope for an index-based fund manager to beat the template index by trading more astutely than other fund managers using the same index because the manager will be the only fund manager using this specific index. By avoiding the periodic trading frenzies of conventional indexing, the new fund indexes and the new funds should beat the benchmarks by a wider average margin than the most effective conventional index manager can beat the benchmark index. This change in procedure reduces the fun and challenge of being an index fund manager and suggests a shift in intellectual resources to index rule design. Ironically, under these circumstances the better index fund managers may be more reluctant than their less effective competitors to adopt the new process; but the greater transaction cost savings from the new indexing implementation should assure long-term outperformance by the new funds.

      The template index would be used and its performance published to reflect the index tracking record of the manager. The performance of the fund will also be evaluated against standard benchmark indexes covering similar market space. For most purposes, however, it is the implementation in the Fund that matters going forward. For this reason, and because the Self-Indexing Fund is also a suitable vehicle for an actively-managed fund, subsequent references to "an index" are to the Self-Indexing Fund Index based on the fund's NAV.

      The portfolio is offered to investors in the form of ETF shares, and the initial and subsequent values of the ETF shares are published as a Self-Indexing Fund Index which can be used for the same purposes as any other equity index. The value of the Self-Indexing Fund Index is determined by changes in the NAV of the fund shares. The fund manager changes the composition of the fund index by selling some of the components of the portfolio and/or buying new components for the portfolio. The fund manager discloses the Self-Indexing Fund portfolio daily, and posts a creation basket and a redemption basket for public dissemination shortly after the market closes each business day. The fund creates additional fund shares by issuing them in exchange for deposits of portfolio components in multiples of the daily creation basket and redeems fund shares by accepting multiples of a preset number of fund shares in exchange for portfolio components in the form of like multiples of the daily redemption basket. The creation and redemption process is quite similar to the process for current equity index ETFs. The principal difference is that the "Authorized Participant" (the market maker or investor who creates or redeems fund shares) must notify the fund's distributor of plans to create or redeem fund shares earlier in the trading day than notice is required for most equity index ETFs. The earlier notice permits the manager of the Self-Indexing Fund to defer selling stock that is needed to be delivered as part of the posted redemption basket and to sell incoming shares from the same creation baskets as shares in the same stocks are sold to change the fund portfolio. Changes in the portfolio today are not reflected in net asset value until the next business day, so the creation basket will be an excellent hedging portfolio for market makers right up to the market close.

      Intra-day index calculations are identical to capitalization-weighted index calculations and the weights are the portfolio weights used in that day's 4:00 p.m. NAV calculations. If the fund holds a cash balance or is leveraged, the index will include a cash component (which presumably earns interest) or will be leveraged. In short, the index fully reflects the fund's investment characteristics and investment process.

      Conclusion

      Unlike today's exchange-traded equity index funds, which attempt to create a portfolio to track the performance of a benchmark index, the fund-created index has the essential characteristics of the fund-friendly, non-disclosed index, used as a template or pattern to construct the exchange-traded index fund. The Self-Indexing Fund operates as index funds should always have operated - in the exclusive interest of the fund's shareholders.

      The Self-Indexing Fund Index can be used as a basis of comparison with other funds and their managers and with other indexes. It can serve as the basis of a family of derivatives which can be based either directly on the fund (physically settled) or on the index created by the fund (cash settled). Through the creation of its index and disclosure of changes in the Self-Indexing Fund and/or Index, the Self-Indexing Fund will report its portfolio and/or index changes within a few hours after the market closes on each business day when a portfolio change occurs.11

      
      

      ENDNOTES:
      1 The trial court decision, Hellerstein (2001), in the McGraw-Hill vs. Vanguard litigation (U.S. District Court, Southern District of New York, 00 Civ. 4247 (AKH)) on introduction of exchange-traded share classes, indicates that the license fee Vanguard pays S&P is capped at $5,000 per year. This figure is more correctly stated as $50,000 per year (still a relatively low number) in Lucchetti and Lauricella (2001). In contrast, the more recently negotiated index license fee for the 500 SPDRs is generally estimated to be approaching $10,000,000 per year.
      2 See Jacques (1988) and hundreds of articles and brokerage firm reports published since.
      3 See Gastineau (2002), Chapter 6, including the endnotes.
      4 See, for example, Bogle (1999) p. 134.
      5 See, Sinquefield (1991) for an excellent discussion of the advantages of patient trading.
      6 See, for example, Garnick, et al. (2001).
      7 See, for example, Plexus Group (2001) which measures average price changes relative to pre-announcement prices or effective membership prices.
      8 See Part B of the SPDR Prospectus, the Section headed "The Portfolio - Adjustments to the Portfolio."
      9Some passive managers are not slavish adherents to an index. They are willing to delay reconstitution trades indefinitely and implement them with cash flow from dividends and new investments in the fund. This is a different process than we describe here, however, because it offers the manager very wide discretion. The process we will describe retains the manager's flexibility within a narrow time interval but reduces the transaction costs associated with tracking the index closely. It lowers costs by avoiding concentration of multiple index fund orders within a short time.
      10 There may be regulatory reluctance to embrace cash settled derivatives on an index with changes that are not disclosed until after a fund which both creates the index and is based on it has transacted in its portfolio. This is not a major issue. There can be no such objection to derivatives that settle in the fund shares and they are likely to be the most popular derivatives anyway.
      11 For more information on the Self-Indexing Fund, see Gastineau (2002) pp. 189 - 200.

      
      

      References:

      1. Bernstein, Peter L., Capital Ideas, Free Press, 1992, pp. 233 - 252.

      2. Bogle, John C., Common Sense on Mutual Funds, Wiley, 1999.

      3. Ellis, Charles D., "The Loser's Game," Financial Analysts Journal, July - August 1975, pp. 19 - 26.

      4. Garnick, Diane M., John Davi, Steve Kim and Silvio Lotufo, "Index Dynamics: S&P Turnover Accelerated in Y2K," Merrill Lynch, January 11, 2001.

      5. Gastineau, Gary L., The Exchange-Traded Funds Manual, John Wiley & Sons, 2002.

      6. Hellerstein, Judge Alvin K., Memorandum and Order of Judgement in "The McGraw-Hill Companies vs. Vanguard Index Trust, et al.," 00 Civ. 4247 (AKH), United States District Court, Southern District of New York April 25, 2001.

      7. Jacques William E., "The S&P 500 Membership Anomaly, or Would You Join this Club?", Financial Analysts Journal, November/December 1988, pp. 73 - 75.

      8. Julian, Bill, "Catching the Winners," Salomon Smith Barney Portfolio Strategist, August 23, 2001, pp.8 - 11.

      9. Lucchetti, Aaron and Tom Lauricella, "Vanguard and S&P Seem Such a Nice Pair; Why the Nasty Spat?" Wall Street Journal, August 23, 2001, pp. A-1 and A-18.

      10. Malkiel, Burton, G., A Random Walk Down Wall Street, Norton, First Edition, 1973.

      11. Morgenson, Gretchen, "What the Sales Brochure Didn't Tell You," Forbes, April 7, 1997, pp. 90 - 96.

      12. Plexus Group, "Is Eliminating Tracking Error Hazardous to Your Client's Wealth?", Commentary 64, January 2001.

      13. Samuelson, Paul, A., "Challenge to Judgment," Financial Analysts Journal, Fall 1974, pp. 17 - 19.

      14. Sinquefield, Rex A. "Are Small Stock Returns Achievable?", Financial Analysts Journal, January/February 1991, pp. 45 - 50.

      15. SPDR Trust, Series 1 Prospectus, "Adjustments to the Portfolio Deposit," January 31, 2001, pp. B-53 - B-56.