CRS NOTES PITFALLS OF TAX ON SHORT-TERM TRADING BY PENSION FUNDS.

Are Pension Funds Short-Term Investors?

Donald W. Kiefer Senior Specialist in Economic Policy Office of Senior Specialists

January 31, 1992

SUMMARY

This report examines whether pension funds are short-term investors and reviews the consequences of their investment behavior. The focus of the analysis is on investments by pension funds in the stock market.

Section I of the report provides data on pension fund portfolio turnover rates and draws implications for average investment holding periods. Beginning in the mid-1960s and carrying through the late 1980s, trading activity by all types of institutional investors increased substantially. Whatever caused the increase in trading affected all types of institutional investors and the stock market as a whole, not just pension funds.

Section II considers the standard for judging the investment performance of pension funds. The analysis in this report presumes that the appropriate criterion is the effect on the rate of return earned by and the risk borne by the funds. If there is a conflict between this goal and another objective -- for example, the effect of pension fund investment patterns on corporate planning horizons -- current law and the fiduciary responsibility of pension funds require resolving the conflict in favor of the investment performance of the funds.

Section III discusses the relationship between the portfolio turnover rate and the rate of return earned on a portfolio in three different theoretical models of the financial markets.

Section IV considers eight potential explanations for the substantial increase in pension fund portfolio turnover rates since the 1960s. At the end of the section is a summary of a statistical analysis in the Appendix which attempts to measure the effects of the various factors contributing to higher turnover rates. The results suggest that the development of the futures and options markets, the spreading influence of modern portfolio theory, the decrease in transaction costs, and the general runup of stock market prices played significant roles.

Section V reviews the empirical research on the relationship between portfolio turnover rates and rates of return earned by pension funds. In general, the research finds that the portfolio turnover rate and the rate of return (net of costs) are independent. This relationship is consistent with the costly information model of the securities markets discussed in section III.

Section VI considers the criticism that pension funds evaluate their investment managers based on short-term performance. Relying on the little information available on this topic, investment managers typically seem to be given 3 to 5 years to prove their worth.

Finally, section VII discusses implications of the analysis. The principal conclusion is that -- using the effect on the rate of return earned by the funds as the standard of judgement -- there is not much evidence that pension funds trade their stocks too aggressively.

ACKNOWLEDGMENTS

The author gratefully acknowledges helpful comments and suggestions on an earlier draft from Jane Gravelle, Richard Ippolito, William Randolph, Raymond Schmitt, Vincent Treacy, John Turner, Kevin Winch, and Dennis Zimmerman.

                          TABLE OF CONTENTS

Pension funds hold nearly $1 trillion in equity investments, over one fourth of all equity capital in the U.S. economy. ¹ Together with other institutional investors they hold over 40 percent of all equity capital ² and do close to two thirds of all trading on the New York Stock Exchange. ³ Hence, their investment behavior is important, not only to the workers who depend on pension funds for retirement security, but also to the financial markets and to the companies and governments which both sponsor them and issue the securities in which they invest.

Recently, pension funds have been criticized for being short- term investors. The complaint is that the stock market, where trading is dominated by pension funds, is interested mainly in quick profits; investors are not willing to hold onto an investment patiently waiting for a long-term return. If an investment fails to provide quick results, it is alleged, the stock is dumped and another investment, which the investor believes to have more promise of short-term profit, is substituted. This investment pattern is claimed to force corporate managers to forego long-term investment projects in favor of short-term efforts to boost quarterly earnings. This orientation of corporations, in turn, is held to be one of the primary reasons for low productivity growth and our loss of international "competitiveness." Several congressional hearings have addressed this issue. ⁴

The policy implications of short-term trading by institutional investors are clearly controversial; in fact, legislation that would move in opposite directions on this issue is before Congress. A bill in the House would remove current restrictions on short-term trading by mutual funds, while a Senate bill would impose the restrictions now applied to mutual funds on pension funds.

Under current law, to maintain its tax status as a pass-through entity, a mutual fund must derive less than 30 percent of its gross income from the sale of securities, futures, and options held less than 3 months. ⁵ Mutual funds are the only type of investor subject to this restriction, which has become known as the "short-short" rule. H.R. 2735, introduced by Dan Rostenkowski, Chairman of the Ways and Means Committee, would repeal this requirement. In introducing the legislation, Mr. Rostenkowski stated,"Repeal of the rule will . . . bring the tax laws in line with the realities of present- day securities markets and investment strategies." ⁶ The legislation has been supported by the U.S. Department of Treasury (assuming the lost revenue is replaced) and the Securities and Exchange Commission. ⁷ At the same time, S. 2160, introduced in the Senate by Senators Sanford, Sasser, and Ford, would extend the short- short rule to pension funds. ⁸

Another Senate bill, S. 1654 introduced by Senators Dole and Kassebaum, would impose a tax on short-term gains received by pension plans with assets of $1 million or more. A tax rate of 10 percent would apply to gains from sale of assets held 30 days or less, and a tax rate of 5 percent would apply to gains from sale of assets held between 30 days and 180 days.

This report examines whether pension funds are short-term investors and reviews the consequences of their investment behavior. ⁹ The focus of the analysis is on investments by pension funds in the stock market. ¹⁰

I. THE FACTS: PENSION FUND TURNOVER RATES AND HOLDING PERIODS

Other than limited survey information, there are no data available on typical or average holding periods for equity investments by pension funds. For this reason, portfolio turnover rates are often cited as being indicative of short-term trading by pension funds and other institutional investors. Even this information is frequently anecdotal, however, because industry-wide statistics are no longer published. Furthermore, the relationship between portfolio turnover rates and holding periods is not as direct as is often assumed.

The first subsection of this section discusses the relationship between portfolio turnover rates and the average holding periods of the securities. The second subsection shows graphically the available pension fund turnover rates and discusses their implications for holding periods. Finally, the third subsection shows historical turnover rates on the New York Stock Exchange to provide the context for the more recent turnover rates.

A. THE RELATIONSHIP BETWEEN TURNOVER RATES AND HOLDING PERIODS

Care must be exercised in using portfolio turnover rates to draw conclusions about the holding periods of investments. In certain circumstances the turnover rate provides information about the average holding period of shares that HAVE BEEN SOLD, but it does not necessarily provide much information regarding the average holding period of shares IN THE PORTFOLIO. This can be illustrated with a simple example, details of which are given in Table 1 on the next page.

Assume a portfolio of stocks is split evenly into two components. Half of the portfolio is managed "conservatively;" the other half is managed "aggressively." The turnover rate of the conservative portion of the portfolio is 10 percent, and the turnover rate of the aggressive portion of the portfolio is 100 percent. Further assume that when shares are sold from one of the halves of the portfolio, an equal fraction of all shares in that portion of the portfolio is sold. Under these assumptions, the turnover rate for the total portfolio will be 55 percent [(0.5 x 0.1) + (0.5 x 1.0) = 0.55].

The usual approach is to assume that the average holding period of stocks in a portfolio is approximately equal to the inverse of the turnover rate. Thus, in this case, the average holding period would be estimated at 1.82 years [1 / 0.55 = 1.82]. For ease of reference, this figure will be referred to as the "turnover holding period." As indicated in Table 1, the turnover holding period is equal to the average holding period of the shares that have been sold. But it is not equal to the average holding period of the shares held in the portfolio.

The reason for this disparity is that shares in the conservative and aggressive portions of the portfolio comprise different proportions of the shares that are sold versus the shares held in the portfolio. The average holding period for the shares in the conservative portion of the portfolio is 10 years, whereas the average holding period for shares in the aggressive portion is 1

                                TABLE 1

But the two groups of stocks do not comprise equal proportions of the shares that have been sold. 91 percent [0.5 / 0.55 = 0.91] of the shares that have been sold will have come from the aggressive side of the portfolio; the other 9 percent [0.05 / 0.55 = 0.09] will have come from the conservative side. The average of these two holding periods, weighted by the proportion of sales, is 1.82 years [(1 x 0.91) + (10 x 0.09) = 1.82].

Each of these average holding periods is correct given the concept that lies behind it. Furthermore, it is not clear which of the figures is the most meaningful in trying to determine whether the investor is a short-term investor or a long-term investor. The illustration makes clear, however, that even portfolios with high turnover rates may contain a substantial amount of investments that have been held for long periods. In this sense, a high turnover rate does not necessarily demonstrate a short-term investment approach (at least not exclusively a short-term approach). In general, the turnover holding period provides the lower bound for the average period that shares in the portfolio are held. ¹¹ The actual figure may be much higher than this lower bound depending on the turnover rates within components of the portfolio and the distribution of shares across those components.

These relationships also imply that the portfolio turnover rate may erroneously rank pension funds in terms of average holding periods. That is, in comparing two pension funds, the fund with the higher turnover rate may hold shares in its portfolio for longer periods, on average, depending on the compositions of the portfolios.

B. PENSION FUND TURNOVER RATES AND IMPLICATIONS FOR HOLDING PERIODS

While data on pension fund portfolio turnover rates are available, no single source provides comprehensive information. Turnover rates have been published by several sources, but they are incomplete and inconsistent.

Figure 1 on the next page plots several portfolio turnover rates for the years for which they are available during the period 1955 through 1987. The solid lines in the graph are different turnover rates for pension funds; the dashed and dotted lines are other turnover rates included for comparison.

Until 1982 the Securities and Exchange Commission (SEC) published common stock portfolio turnover rates for private pension funds and other institutional investors. These turnover rates for pension funds, mutual funds, and life insurance companies for 1955 through 1980 (the last year for which the data were published) are graphed in Figure 1. ¹² The SEC turnover rates are based on value; the rate is equal to the average of the gross amounts of portfolio purchases and sales of stocks during the year divided by the average value of the portfolio. The averages are calculated on a value- weighted basis (as are all the turnover rates in the figure except the NYSE rate).

An article by Berkowitz and Logue (B&L) provides equity turnover rates for private pension plans for the years 1965 through 1983, mutual funds for 1971 through 1983, and for college and university endowment funds for 1974 through 1984; these series are plotted in the graph. ¹³ The definition of the turnover rate used by Berkowitz and Logue is the lesser of the value of purchases or sales divided by the average asset value of the portfolio.

[Figure 1 omitted]

McCarthy and Turner (M&T) report turnover rates for 1977 through 1987 for corporate stock portfolios of private pension plans reporting annually to the Internal Revenue Service (those plans with 100 or more participants). ¹⁴ These turnover rates, computed using the SEC definition, are also plotted in Figure 1.

The average share turnover rate for all stocks on the New York Stock Exchange (NYSE) is also graphed in the figure. ¹⁵ The NYSE turnover rates are not based on value; they equal the trading volume in numbers of shares divided by the average number of shares listed on the exchange. These data, therefore, are not strictly comparable to the other lines in the graph; there does not appear to be a reason, however, to expect a systematic difference in the data. ¹⁶

The graph suggests several conclusions. First, the reported turnover rate depends on the source of the data and how it is calculated. For 1977, for example, the pension fund portfolio turnover rate reported by the SEC is 17.4 percent; the rate reported by Berkowitz and Logue is 22.2 percent, and the rate reported by McCarthy and Turner is 28.1 percent. Given the differences in measured rates, it may not be appropriate to rely exclusively on any single data source.

Second, whatever caused the increase in trading since the early 1960s affected all types of institutional investors and the stock market as a whole, not just pension funds. Beginning in the mid-1960s and carrying through the remainder of the period shown in the graph, trading activity by all types of institutional investors increased substantially. For mutual funds and (to a somewhat lesser extent) insurance companies, after a period of relatively low and constant turnover, trading activity rose rapidly and peaked in the late 1960s and early 1970s, then, after declining, rose again in the late 1970s and early 1980s. For pension funds and the market as a whole, the earlier peak was less pronounced. But the trading increase in the late 1970s and early 1980s was substantial.

Third, the graph does not provide clear evidence that pension fund turnover rates exceed overall market turnover rates by much (if at all), or that pension fund trading has increased relative to total trading in the market. While the B&L pension fund turnover rates are higher than the NYSE rates in most years, the SEC pension fund rates are lower. The pension fund turnover rates reported by M&T are somewhat higher than the NYSE rates until 1986 and 1987, when they drop below the NYSE rates. The B&L pension fund turnover rates increase somewhat more rapidly than the NYSE rates, but the SEC and M&T pension fund turnover rates do not.

Fourth, pension funds do not appear to have portfolio turnover rates that are out of line with those for other institutional investors. Pension fund portfolio turnover was lower than that of insurance companies and substantially lower than turnover by mutual funds over this period based on the SEC data, which provide comparable figures for the three types of institutional investors. ¹⁷ Endowment funds seem to have lower turnover rates than the other institutional investors, but their turnover rates also increased substantially in the early 1980s.

These relationships raise the question of why, other than their massive size, pension funds have been singled out for criticism by those concerned about high turnover rates in the stock market.

Comparable turnover rates for institutional investors are not available for years after those shown in the graph. The NYSE rate is available, however. After peaking at 73 percent in 1987, the last year shown in the graph, the NYSE turnover rate dropped to 55 percent in 1988, 52 percent in 1989, and 46 percent in 1990. Hence, if the turnover rate of pension funds continued to move roughly parallel to turnover in the overall market, it decreased significantly in the past few years. Nonetheless, portfolio turnover rates are substantially higher today than they were 30 years ago.

The published data actually understate the growth in trading of stocks listed on the New York exchange, however, because they do not account for the growth in off-exchange trading and trading on foreign exchanges. Trades executed at the NYSE as a proportion of total trades in listed stocks dropped from 85 percent in 1980 to about 68 percent in 1990. ¹⁸ This decline could account for the more rapid growth in pension fund turnover suggested by the comparison between the B&L pension fund data in Figure 1 and the NYSE data.

It is possible that if comparable figures were available the turnover rate of pension funds may have actually declined somewhat relative to market averages during the past few years because of the movement toward index investing. Index investing involves managing a portfolio of stocks to duplicate the performance of one of the market indexes, most often the S&P 500 Index. While some periodic revision of such a portfolio is necessary to keep it balanced to track the index, index funds typically have much lower turnover rates than actively managed stock funds. Index investing was not widespread a decade ago. Now, however, 38 percent of pension funds have a portion of their portfolios invested in index funds; these funds account for about 20 percent of total pension equity assets, and the amount is growing. ¹⁹ Pension funds are moving toward indexation more rapidly than the market as a whole; only 7.3 percent of the total value of corporate stocks is held in indexed funds. ²⁰

Typically, a pension plan will use several different investment strategies for different portions of its funds. A "core" portion of the funds may be placed in an index fund or with conservative investment advisors to manage. Smaller portions of the funds may be placed with "active" or "aggressive" investment advisors in an effort to maximize the rate of return earned on the funds consistent with maintaining an acceptable level of risk. ²¹ While some of these aggressive investment advisors may have very high and attention- getting turnover rates, reaching levels as high as 200 to 300 percent in some cases, ²² the turnover rates for pension plans as a whole will be much lower. This "segmented" approach to pension fund investing underscores the point made in subsection A regarding the use of portfolio turnover rates to infer average holding periods. That is, when the securities in different components of a portfolio are held for different periods, the turnover rate may provide a misleading indication of the average holding period.

That caveat notwithstanding, if the most recent pension fund turnover rates shown in Figure 1 were used to estimate the average holding period, the estimate would be about 1.5 years. If the pension fund turnover rate has decreased proportionately to the NYSE rate from 1987 to 1990, the current estimate of the pension fund average holding period would be about 2.4 years. As discussed in the previous section, this figure would represent a minimum for the average holding period of shares held by the funds; the actual average holding period could be substantially longer.

C. THE HISTORICAL CONTEXT

To provide a proper perspective, it is useful to place the data graphed in Figure 1 in a longer historical context. Current stock market turnover rates may seem high, but the rates were even higher early in the 1900s. As shown in Figure 2, the turnover rate on the New York Stock Exchange sometimes exceeded 100 percent early in the century (the figure in 1900, not shown in the graph, was 172 percent). ²³ Over the period 1910 to 1940, the turnover rate, which was also quite variable, gradually declined and became more stable. From 1940 to the mid-1970s the rate remained between 10 and 25 percent. During the late 1970s and the 1980s, the turnover rate rose substantially and then declined somewhat. The peak turnover rate of 73 percent in 1987 was still significantly lower than the levels in the early NYSE Turnover Rates 1900s. The turnover rate 1910-1990 declined to 46 percent in 1990. ²⁴

If the turnover rates in the early part of the century were used to estimate average holding periods for the overall stock market, the results in several years would be holding periods of less than one year (0.58 years in 1900, for example, and 0.79 years in 1910).

[FIGURE 2 OMITTED]

II. THE CRITICISM: IS IT BASED ON THE APPROPRIATE CRITERION?

The high portfolio turnover rates of pension funds in the 1980s led to the criticism cited in the introduction that they have become short-term investors. This criticism implies that pension funds should hold their investments longer. How long should they hold them? And how is this optimal time period identified?

The primary investment objective of pension funds -- as for any investor -- is to achieve the highest rate of return consistent with an appropriate risk exposure. In fact, this objective is mandated by Federal law for private pension funds under the Employee Retirement Income Security Act (ERISA). ²⁵ Hence, judgements about holding periods must be made with regard to their effects on risk and rates of return.

Some critics have argued that because pension funds are investing to meet financial obligations that are relatively certain and very long-term in nature, they should be long-term investors, holding their investments on average for many years. But this does not necessarily follow. ²⁶ An investor who achieves the highest rate of return during each short subperiod of a longer period will also achieve the highest rate of return over the longer period. Hence, unless there is something about the short-term strategy that prevents following a profitable long-term strategy, there is nothing inconsistent about being a long-term investor who attempts to maximize the rate of return in each short time interval. ²⁷

The criticism of pension funds for short-term investing usually is not based on a claim that investing for longer periods would increase their rate of return. ²⁸ Rather, as stated in the introduction, most of the criticism is based on the argument that pension fund investment behavior has negative effects on the companies in which they invest. The argument is that short-term holdings of corporate stock force corporate managers to focus unduly on short-term profits. But providing a stable, long-term pattern of share ownership for corporations is not the objective of pension funds. Their objective is a high rate of return consistent with the level of risk they have determined to be appropriate for their fund. Whether this objective leads to a pattern of long-term or short-term investment is a separate issue from the effects of the shareholding pattern on corporations.

Pension fund investment practices have over the years been the targets of criticisms that are tangential to the fiduciary responsibilities of the pension funds to their beneficiaries. They have been criticized, for example, for holding investments in companies that have operations in South Africa and for not making sufficient investments in localities where the pension funds operate. ²⁹ Each such criticism raises the issue of whether the critic believes pension funds should pursue the alternative goal regardless of the consequences for the rate of return earned on the funds' investments, or whether the prescription would be to take the alternative goal into account only if doing so would not sacrifice the rate of return objective.

This issue also applies to the criticism regarding short-term investing. The analysis in this report presumes that the appropriate criterion for judging the investment behavior of pension funds is the effect on the rate of return earned by and the risk borne by the funds. If there is a conflict between this goal and another objective -- for example, the effect of pension fund investment patterns on corporate planning horizons -- current law and the fiduciary responsibility of the funds require resolving the conflict in favor of the investment performance of the funds. This report concludes, however, that these two goals probably do not conflict. While there may be practical difficulties involved in implementing a policy to reduce the turnover rates of actively traded pension funds and, at the same time, avoiding unintended effects, the policy, if successful, probably would not reduce the rate of return received by pension funds.

III. THE THEORY: PORTFOLIO TURNOVER AND THE RATE OF RETURN

The expected relationship between the turnover rate of investment portfolios and the rate of return they earn depends on the conceptualization of the financial markets, specifically, on the extent and nature of their efficiency. Three alternative views of the nature of the financial markets are summarized in this section to provide a context for the remainder of the discussion.

A. EFFICIENT MARKETS; COSTLESS INFORMATION

The financial markets are characterized as highly "efficient" in the finance theory that has predominated the literature for at least the last two decades, the development of which won its three principal early contributors -- Harry Markowitz, William Sharpe, and Merton Miller -- the 1990 Nobel prize in economics. ³⁰ In an efficient financial market, the prices of securities are maintained at fundamental value. ³¹ Assuming "rational" investors, free availability of all relevant information, no restraints on trade or credit, and zero or very low transaction costs, efficient market prices result. In an efficient market, the prices of all securities always reflect all available information. ³² If new information becomes available and changes the fundamental value of a security, potential buyers and sellers of the security will adjust their prices accordingly. Hence, the next trade will occur at a price equal to the new fundamental value. If a price diverges from its fundamental value, even by a small amount, it will be noticed by investors who will profit by buying shares (if they are underpriced) or selling shares (if they are overpriced) until fundamental value is reestablished.

In an efficient market the expected return on each security properly reflects its risk, and all securities with equal risk also have equal expected returns. This implies that future changes in return on a security are random; they cannot be predicted systematically. This, in turn, implies that it is possible to earn a higher-than-market rate of return only by chance; it is not possible for an investor to "beat the market" consistently. ³³ The conclusion of this line of reasoning is that the best approach to investment management is to assemble a diversified portfolio of investments structured to earn the highest rate of return consistent with the risk being assumed; such a portfolio is referred to as an "efficient portfolio."

Efficient markets theory gave rise to the concept of index funds, which are portfolios structured to match the risk and performance of a broad market index, such as the Standard & Poor's 500 stock market index. ³⁴ Some index funds hold all the securities in the market index. Many index funds hold a subset of the securities; statistical methods are used to structure the portfolio to closely parallel the performance of the index. ³⁵ Investment in an index fund is referred to as a "passive" investment strategy, since it involves investing in the stock market as a whole. Index funds are not entirely inactive, however, since the funds must periodically be "rebalanced" as prices of the stocks in the market. index change. ³⁶ But the turnover rate of most indexed funds is relatively low.

Based on efficient markets theory, investment portfolios with higher turnover rates (higher than necessary to maintain an index fund) would, on average, be expected to have lower rates of return (net of costs). Turnover is not costless; commissions must be paid to brokers and market specialists, taxes may apply to transactions, and trades may have adverse market impacts. ³⁷ If it is not possible to earn a higher than average rate of return -- and the efficient markets theory implies that it is not -- then additional trading must, on average, yield a lower rate of return when costs are taken into account.

This does not imply that no investors who trade actively will beat the market or even that the average return for actively traded portfolios will necessarily be lower than the return on index funds every period. Since price movements in the market are random, with hundreds or even thousands of investment managers and investment strategies, the laws of probability will result in some of them beating the market each period. Some will even beat the market for extended periods, but their numbers will dwindle as time passes. During some periods, traders on average may surpass the index funds; in other periods they will fall short. But, if the efficient markets theory is the correct view of the financial markets, portfolios with higher turnover rates should, over the long term and on average, have lower rates of return.

This points out an uncomfortable inconsistency between the implications of efficient markets theory and real world behavior. If the theory is correct, then turnover rates should have remained relatively low. Investors and investment managers who try to increase their rate of return by trading more actively to take advantage of perceived opportunities for higher profit would discover that, over the longer term, they earn lower rates of return instead. For this reason, they would revert to a less aggressive trading style. Hence, turnover rates substantially in excess of those associated with index funds should not persist. As reported in section 1, however, mutual funds have had relatively high turnover rates since the mid-1960s, and other investors and the market as a whole have had high turnover rates at least since the early 1980s.

Because of the randomness of stock market price changes and rates of return, it may be difficult for investors to determine the long-term outcome of their investment strategies. Furthermore, investors exhibit an apparent eagerness to believe in the ability to beat the market. Nonetheless, stock market investment is among the most heavily studied of economic activities. Participants in the financial markets have available a greater array of data, computers, sophisticated analytical techniques, and consultants than in probably any other market. The market also attracts highly capable, well educated people. If inefficient behavior persists on a grand scale in this market, then a serious reconsideration of some of the most basic principles of economic theory may be necessary.

B. MODIFIED EFFICIENT MARKETS; COSTLY INFORMATION

Because of the dominance of efficient markets theory in the financial literature, rather little attention has been devoted to developing alternative models of the financial markets. One alternative model suggested by Grossman and Stiglitz, however, is relevant to a consideration of the relationship between portfolio turnover and return. ³⁸ The model is a modification of the standard efficient markets model to allow for costly information.

The model assumes that at least some information which is useful to investors in determining the value of investments is costly to obtain. Investors will make the outlay to obtain this information only if it is profitable to do so. Assuming it is profitable for at least some investors to obtain some information, two classes of investors will develop: "informed" investors and "uninformed" investors.

It is useful to explore how equilibrium is achieved in this model and some of its properties. Assume an initial condition in which no investors are "informed." If it is profitable to do so, some investors will make the outlay to obtain information which will enable them to make trades providing them with extra profit exceeding the cost of the information. ³⁹ As more investors become informed, the cost of information will increase (in a real world setting, the types of information needed to reap higher and higher profits will be increasingly costly to obtain) and the extra profit which can be earned from information-based trades will decline. Eventually, an equilibrium will be reached in which the expected profit from the marginal information-based trade exceeds the average profit from uninformed trades by an amount equal to the marginal cost of information.

In this equilibrium, the market exhibits a kind of economic efficiency, but the efficiency is different in nature than in the efficient markets model. In the costly information model, the prices of all securities do not fully reflect all available information. The prices only partially reflect the costly information. The discrepancy between price and fundamental value of securities is necessary to compensate informed investors for their costs of obtaining information. In this model, the uninformed investors benefit from the activities of the informed investors because market prices at least partially reflect the information known only to the informed investors.

The informed investors in this model do not hold indexed portfolios. ⁴⁰ They assemble portfolios of selected investments based on their special information in an effort to earn higher-than- market rates of return. As informed investors continuously obtain new information, they learn of new profitable trading opportunities and revise their portfolios to take advantage of them. Hence, the informed investors do not follow a passive investment strategy; they are "active" investors with portfolio turnover rates exceeding those of index funds.

In the costly information model, the relationship between portfolio turnover rates and rates of return in equilibrium would be non-negative. The relationship could be positive, or there could be essentially no relationship, depending on how costs and profits are related to the volume of trading based on costly information. ⁴¹ In either case, for investors like pension funds who rely primarily on outside investment managers, there should be little or no relationship between turnover rates and rates of return. The fees of the investment managers would be expected to absorb most of the excess returns earned on managed portfolios, although pension funds with strong negotiating positions may be able to garner some of the excess returns for themselves.

The costly information model of the securities markets has not been fully developed and scrutinized. In fact, it has largely been ignored. It is discussed here for two reasons. First, it provides a rationale for active investment management without having to discard the notion of efficiency in the securities markets. Second, the empirical research on the relationship between portfolio turnover rates and rates of return (surveyed in section V) is broadly consistent with the implications of the costly information model.

C. INEFFICIENT MARKETS

While the efficient markets paradigm currently dominates the finance literature, not all theorists accept the notion that the finance markets are efficient or even close approximations of efficiency. Over the last decade a number of studies have identified specific areas in which the financial markets are alleged not to be fully efficient. The ostensible inefficiencies range from the apparent excess rate of return earned by stocks with low price/earnings ratios and by stocks of small firms, to excess returns earned by firms neglected by market analysts, to the so-called "January effect" (the tendency for the stock market to rise during January). ⁴² Furthermore, several recently developed theories and new empirical tests suggest broader market inefficiency in which securities may commonly be mispriced for lengthy periods. Some of the theories involve myopic investors who rely on extrapolation of current trends because they are unable to fully evaluate the business prospects of companies, or "noise traders" who believe that they have, but in fact do not have, special information about the future course of stock prices.

The inefficient markets theories will not be summarized in detail here, ⁴³ but their implications for the relationship between turnover and rates of return can easily be summarized. In general, active investors can earn superior rates of return in these models. In some of the models it is even possible that the "noise traders" earn higher returns than the sophisticated investors. ⁴⁴ Hence, as in the costly information model, the equilibrium relationship between turnover and rates of return should be non-negative.

It is not known with certainty, of course, which of these theories most accurately describes the investment markets. ⁴⁵ While the efficient markets theory or a close variant of it provides the basis for most current academic research, much actual investment practice -- including investment analysis, investment advice, and the management of investment portfolios -- is based on the belief that it is possible to identify investments that will provide superior returns.

IV. THE EXPLANATIONS: WHY HAVE TURNOVER RATES INCREASED?

As reported in section I, pension fund portfolio turnover rates have increased substantially since the 1960s. In assessing the policy implications of this increase it would be useful to know why it has occurred. Unfortunately, trading volume and turnover have received far less attention in the finance literature than most other aspects of the securities markets. No systematic analysis has attempted to explain the patterns of stock market trading shown in Figure 1 (on page 6) or Figure 2 (on page 10). Presumably, however, the substantial increase in portfolio turnover since the 1960s occurred because investors thought increased trading would bring higher returns, lower risk, or some other desirable result.

This section discusses eight potential explanations for the turnover increase. ⁴⁶ Some of the first five potential explanations quite appropriately contribute to higher turnover; whether the influence of others is appropriate from a public policy perspective might be somewhat controversial. The last three potential explanations have negative public policy connotations. That is, public policy goals would be served by minimizing their influence. The last subsection provides a statistical test of the strength of some of the factors.

A. MODERN PORTFOLIO MANAGEMENT THEORY

Efficient markets investment theory forced a rethinking of many aspects of investment management. Some implications of the theory, such as the notion that it is impossible to consistently beat the market, obviously have not persuaded most investment managers. But the theory also introduced an entirely different concept of portfolio management, and this aspect of the theory has had widespread and profound effects. In general, following the prescripts of modern portfolio theory results in higher turnover rates.

Earlier investment theory, the basic principles of which are still taught in many business schools and used by many investors, emphasizes investment selection, concentrating on the criteria to use in attempting to identify "undervalued" stocks. ⁴⁷ Under this approach to investment, the intention generally is to assemble a portfolio of stocks that can be held for a long period with only occasional review and revision. Prior to the development of modern portfolio management theory, the portfolio as a whole typically was not the subject of a great deal of attention (other than assuring diversification); the focus was on individual investments within the portfolio.

As noted above, efficient markets investment theory focuses on investment PORTFOLIOS rather than individual investments. The emphasis is on assembling and maintaining properly balanced portfolios of stocks to attain the maximum rate of return possible given the risk being assumed. This approach requires frequent adjustment or "rebalancing" of the portfolio to maintain the appropriate risk/return tradeoff. ⁴⁸ The management of an index fund provides an example. Even though an index fund is, by today's standards, regarded as a passive (long-term) investment approach, it is not a "buy and hold" strategy. Trading to rebalance the portfolio and maintain its relationship to the market index frequently involves simultaneously buying or selling dozens or even hundreds of individual stocks. ⁴⁹ Many index funds are rebalanced monthly. The amount of trading required for the rebalancing varies with the volatility of the market, but annual turnover rates of 15 percent to 25 percent are not unusual for index funds. ⁵⁰

The development of modern portfolio theory had a substantial influence on the way virtually all investment managers assemble and adjust their portfolios to diversify their investments and minimize risk. As efficient markets theory began influencing investment management in the mid-1960s, the common perception of what should be considered normal portfolio turnover rates probably increased. A turnover rate of 15 to 25 percent, which is the range of total market turnover during the period from 1940 to the mid-1970s, would now be regarded as a minimum rate associated with passive or "inactive" investment management. An actively managed portfolio would be expected to have a higher, perhaps much higher, turnover rate. In this environment, even some portfolio managers with turnover rates in excess of 200 percent or more may view themselves as long-term investors. ⁵¹

B. ACTIVE INVESTMENT MANAGEMENT

Substantial portions of pension fund equity portfolios are actively managed in efforts to "beat the market" and earn superior returns. This is the behavior many critics have in mind when they criticize pension funds for speculative investing, or for rapidly shifting funds from one investment to another in search of high rates of return.

But pension funds did not start out as aggressive investors. As shown in Figure 1 (on page 6), the portfolio turnover rates of pension funds remained significantly below those of mutual funds until the early 1980s. The graph suggests that over this period pension funds may have become more aggressive in seeking higher rates of return. Indeed, there is some evidence of such a shift, but it is largely impressionistic.

Until the late 1950s, most pension fund assets were invested conservatively in bonds. During the late 1950s and early 1960s, corporate pension fund managers allocated increasing portions of their funds to the equity market. The value of equities first exceeded the value of bonds in private pension portfolios in 1963. The equity portion continued to expand; during the last two decades equities have accounted for between 60 percent and 70 percent of private pension assets. ⁵²

During the 1960s and 1970s, the perspective of corporate managers shifted away from viewing their defined-benefit pension funds ⁵³ as separate entities from their corporations. Such funds are now seen as integral parts of the financial operations of the overall corporation. ⁵⁴ The 1974 ERISA legislation added to this shift in perspective by decreeing that defined-benefit pensions are legal obligations of the sponsoring companies. For these and other reasons, during the early part of this period the principal objective of stock market investment of private pension funds shifted from earning high dividend yields to seeking capital appreciation. ⁵⁵

A higher rate of return on pension fund assets will permit lower contributions to the fund in the future or possibly even termination of the fund and reversion of excess fund assets. ⁵⁶ During the early 1980s, pension fund reversion became a substantial source of corporate capital. ⁵⁷ Congress imposed a 10 percent excise tax on pension plan reversions in the Tax Reform Act of 1986 and raised the rate of the tax to 15 percent in 1988 and either 20 percent or 50 percent (depending on the circumstances of the termination of the pension plan) in 1990. ⁵⁸ The prospect of pension fund reversion may, for this reason, have diminished; nonetheless, a higher rate of return still permits lower future pension contributions. ⁵⁹

Following the lead of corporate pension plans, public pension funds shifted from primary reliance on bonds to heavy commitments to stocks during the 1970s and especially during the 1980s. ⁶⁰ A higher rate of return on pension fund assets carries with it the same benefit of reduced future pension contributions for public entities as for private companies. And while governmental bodies have not terminated pension plans and reclaimed excess funds, they have sometimes dipped into pension assets to help fund government. programs. ⁶¹

C. DECREASED TRANSACTION COSTS

Stock market transaction costs for institutional traders are much lower today than they were in the early 1960s. This decrease is attributable to the deregulation of brokerage fees in the late 1960s and early 1970s, the removal of the Federal and New York securities transfer taxes, and perhaps to other reduced costs as well. Reduced trading costs quite naturally lead to more trading under all models of financial market behavior.

Until 1968, brokerage commissions on listed securities were set by the stock exchanges, and there were no discounts for large transactions. With the rise of large institutional investors in the early 1960s, however, the regulated commission system came under substantial pressure because there are significant economies of scale in securities trading (that is, a trade involving 10,000 shares does not cost 100 times more to transact than a trade involving 100 shares). To compete for the trading business of the institutional investors, the brokerage firms developed a system of "give-ups," in which a specified portion of the commission on a trade would be used by the brokerage firm to provide other services, such as market research, to the institutional investor, sometimes paying third parties to provide the services. So while the stated commission was not negotiable, the effective commission (stated commission minus give-ups) was negotiated.

The so-called "third" and "fourth" markets also developed during the mid-1960s in response to demand for cheaper execution of trades by institutional investors. ⁶² The third market refers to over-the- counter trading of listed securities by brokers who are not members of the exchange and, therefore, were not bound by the fixed commission schedule. The fourth market refers to trading directly between two investors (usually institutions), perhaps through a computerized trading system, but without the participation of a traditional broker (hence, without payment of a brokerage commission). Use of the third and fourth markets by institutional traders, as well as the system of commission "give-ups," led to the eventual demise of fixed-rate commissions.

The regulated commission system was abolished in a series of steps from 1968 to 1975. In December 1968, a volume discount was permitted on trades of 1000 shares or more costing less than $90 per share. In April 1971, commissions on trades exceeding $500,000 became negotiable; in April 1972, the size of the trade was lowered to $300,000. Finally, on May 1, 1975, sometimes referred to as "Mayday" in the brokerage industry regulated commissions were eliminated. ⁶³ Since that date, commissions have declined substantially for large traders.

Part of the decrease in commissions during this period is attributable to the development of more efficient trading technology, such as block trading. Prior to the development of block trading, an institution wanting to buy or sell a large number of shares generally did so in a series of small trades over a period of several days or weeks. In the mid-1960s, brokerage firms developed the ability to trade large blocks of stock in a single transaction or a collection of simultaneous transactions. ⁶⁴ Such "block trading" is much cheaper to execute than a series of small trades. The use of block trades grew substantially during the period of rapid turnover growth. In 1965, only 3.1 percent of the trading volume on the New York Stock Exchange was in block trades of 10,000 shares or more; by 1975 this percentage had increased to 16.6 percent. During the next decade, block trading increased substantially so that since 1984, half of all shares traded on the New York Stock Exchange have been in block trades. ⁶⁵

In addition to the decrease in brokerage commissions, trading costs also declined because of the repeal of transactions taxes at both the Federal and State levels. Until 1966 the Federal Government levied a tax on stock transactions equal to 4 cents per $100 of stock value. The New York State transactions tax, which had an effective rate somewhat less than 0.1 percent of transaction value, was phased out, disappearing in 1981. ⁶⁶

The estimated transaction costs of institutional investors over the period from 1955 through 1990 are graphed in Figure 3; the costs are shown as a percentage of the value of shares traded. ⁶⁷ The costs rose from 1955 to 1964 as a result of increases in the regulated brokerage commissions. Commission give-ups and repeal of the Federal transactions tax reduced trading costs by more than 40 percent from 1964 to 1968, but declining share prices raised costs back to nearly their 1964 level by 1974. ⁶⁸ The continuing decrease in commissions after deregulation in 1975 (and the elimination of the New York State transactions tax) have steadily reduced trading costs for institutional investors. Trading costs in 1990 are estimated at 0.25 percent of the value of shares traded, less than half of the cost in the mid-1960s and mid-1970s. For the largest trades by the largest institutions, the costs are even lower; for these trades commissions are estimated to be between 0.05 percent and 0.18 percent of the value of shares traded. ⁶⁹

ESTIMATED TRANSACTION COSTS FOR INSTITUTIONAL INVESTORS 1955-1990

[Figure omitted.]

A substantial decline in transaction costs would be expected to result in increased trading activity. ⁷⁰ In the efficient markets model of the financial markets, the efficient portfolio would be rebalanced more frequently. In the costly information and inefficient markets models, active investors are searching for new investments which will provide superior returns. The expected return must be high enough to cover the cost of trading and still be attractive. With lower trading costs, more potential investments will meet this test.

Indeed, the graph of transaction costs in Figure 3 looks somewhat like a mirror image of the turnover rates graphed in Figure 1 on page 6. Hence, at the intuitive level, transaction costs seem to be a likely candidate to play a role in explaining the pattern of turnover rates. The statistical analysis summarized in subsection I below confirms the significance of transaction costs as an explanatory factor. Contrary to some assertions, however, transaction costs are not found to be the most important variable explaining higher turnover rates.

D. USE OF OPTIONS AND FUTURES

In the last 15 years the use of options and futures in the U.S. financial markets has increased dramatically. These derivative financial instruments are used primarily for three purposes by institutional investors: to increase the income on a portfolio, to adjust market exposure without trading the underlying securities, and for hedging.

Options and futures may increase the portfolio turnover rates of pension funds in two ways. First, in some data sources, options and futures transactions are counted as purchases and sales in calculating turnover; this directly increases measured turnover even if there are no transactions involving the stocks in the portfolio. This problem affects the turnover data reported by Berkowitz and Logue (cited in footnote 13) graphed in Figure 1 on page 6. Ostensibly, the problem does not affect the data reported by McCarthy and Turner (cited in footnote 14) also graphed in Figure 1. Form 5500 (used by pension plans to report data to the I.R.S.), from which the data reported by McCarthy and Turner are derived, specifically requests information on equity transactions. There is no place on the form, however, to report data on transactions involving options and futures, so it seems likely that some futures and options transactions may be reported erroneously as equity transactions.

The second way that options and futures may increase portfolio turnover is that many investment strategies employing these instruments result in increased trading of stocks. Some of these strategies are illustrated here, but this discussion is not intended to be a comprehensive or detailed treatment of options and futures investment strategies.

One common income-generating technique is selling covered calls on stocks in a portfolio. ⁷¹ In some data sources, the sale of the calls would be counted as a sale of securities. In a stable or declining market, the calls will expire worthless so that no stocks in the portfolio will be "called," or sold. In this case, the calls merely generate extra income on the portfolio. Alternatively, if the market rises and the calls are exercised, the sale of the called securities adds to portfolio turnover. Reinvestment of the sales proceeds also adds to turnover even if the same stocks are reacquired. In this case, the underlying investment position of the portfolio would remain intact, but several transactions would have been recorded.

Another trading technique using options and futures to generate extra income is arbitrage. Arbitrage takes advantage of price discrepancies between options or futures and their underlying securities (which could be the market itself in the case of options and futures on the market indexes) to earn a virtually riskless return (in the case of pure arbitrage). If the S&P 500 future is temporarily overpriced (by a sufficient amount), for example, a large institutional investor can earn a guaranteed profit by selling the future and simultaneously buying the S&P 500 stocks (program trading, in which hundreds of stocks can be bought or sold simultaneously, makes this possible). When the future expires, it will have the same value as the S&P 500 stocks. At that point the position is "unwound"; that is, the stocks are sold and the proceeds are used to liquidate the future. This kind of arbitrage is part of the "invisible hand" that keeps prices properly aligned in the various financial markets.

Arbitrage is inherently short term in nature, but it is not speculative; ⁷² the return is assured and does not depend on the direction or magnitude of price changes. Hence, it is not the stereotypical short-term investment, gambling on a high return (the returns to arbitrage are small) and selling after a short period to switch to another investment if the return does not materialize. It does, however, raise portfolio turnover rates.

The derivative financial instruments can also be used to adjust exposure to the markets. If an investment manager wants to temporarily increase or decrease the sensitivity of a portfolio to changes in the stock market, for example, this can be accomplished using options and/or futures rather than buying or selling shares. Such a "synthetic" investment adjustment is cheaper than buying or selling stocks to make the adjustment (especially when market impact costs are considered), and avoids disturbing the long-term equity holdings in the portfolio. ⁷³

Options and futures are also used to hedge or to manage investment risk while seeking a somewhat higher rate of return. A portfolio manager fearing a market decline, for example, can buy puts covering some or all of the stocks in his portfolio. ⁷⁴ If the market does decline, the puts will limit the loss the portfolio sustains. If the market rises, the portfolio will participate in the share price increase and thus earn a higher rate of return than if the shares in the fund had been sold. In the case of a rising market, the cost of the puts (which expire worthless) amounts to a premium for "insurance" against the risk of a market decline. ⁷⁵

Hedges can also be devised to cover the purchase or sale of securities. If a fund manager knows that he has to sell a block of securities in the near future, for example, buying puts on the securities can lock in a minimum selling price. ⁷⁶

The use of options and futures for hedging transactions raises the possibility that the recent higher portfolio turnover rates of pension funds ironically could be, at least in part, the manifestation of a longer-term investment strategy. Zvi Bodie of Boston University, who has written extensively on pension fund investment strategies and behavior, has made essentially this point. He argues that the ERISA legislation in 1974 altered pension fund investment behavior in important ways. ⁷⁷ As noted in subsection B above, ERISA decreed that defined benefit pensions are legal obligations of the sponsoring companies; it also established minimum funding requirements. Bodie maintains that these changes forced pension funds to be long-term investors and to protect their portfolios as much as possible against loss of value. ⁷⁸ He claims these needs of pension funds were a major part of the impetus behind much of the innovation in the financial markets over the past decade and a half, including not only options and futures, but also bond immunization techniques, portfolio insurance, and a wide variety of other modern financial tools. ⁷⁹ A feature of many of these approaches is that they enable the investment manager to maintain an invested position but, at the same time, protect the investment against instability. To the extent that these approaches are responsible, the higher portfolio turnover is actually the result of a strategy that facilitates holding investments for longer periods. ⁸⁰

E. THE BULL MARKET

Conventional wisdom on Wall Street holds that trading volume is higher in bull markets and lower in bear markets. Much of the empirical research is consistent with this opinion, but the explanation for the pattern is not well understood. The explanations that have been suggested rely on the nature of the demand for and supply of securities in the market by bullish and bearish investors, ⁸¹ the way in which new information affects trading volume and share prices, ⁸² and technical aspects of market structure and functioning, such as asymmetries in the ability of investors to take long and short positions in stocks. ⁸³ None of the proffered explanations fully accounts for the observed pattern.

The period since the mid-1970s has been one of generally rising stock prices, hence, the bull market may contribute to the higher turnover in pension fund portfolios and in the market as a whole. At the intuitive level, this is not a very convincing explanation because equally strong bull market conditions prevailed in the 1950s and early 1960s, but turnover remained low. Nonetheless, the statistical analysis presented in subsection I below does find that stock price increases play a role, albeit a small one, in the increasing pension fund turnover.

F. THE EMPLOYEE RETIREMENT INCOME SECURITY ACT (ERISA)

One possibility that has been suggested is that ERISA, which provides the basic law governing the investment of pension fund assets, may be partially responsible for the increase in pension fund portfolio turnover rates. ⁸⁴ Because passage of the law in 1974 coincides with the beginning of the substantial increase in turnover rates (see Figure 1 on page 6), the claim merits some consideration.

The relevant portion of ERISA is the "prudence rule" which states that pension plan fiduciaries must discharge their duties ". . . with the care, skill, prudence, and diligence under the circumstances then prevailing that a prudent man acting in a like capacity and familiar with such matters would use in the conduct of an enterprise of a like character and with like aims . . ." ⁸⁵ Fiduciaries are personally liable for any losses sustained by a pension fund as a result of their failure to meet this requirement.

There are two ways that have been suggested in which the prudence rule might result in higher pension fund turnover. First, critics have alleged that in some circumstances pension fund managers are virtually required to sell shares in their portfolios. This criticism is most frequently voiced with regard to tender offers in hostile takeovers. Such tender offers usually contain substantial takeover premiums; that is, the offered price for the shares is substantially above the market price prior to the offer. The complaint is that a pension fund investment manager does not realistically have the option of turning down such an offer even if he thinks long-term investment objectives would be better served by doing so. Such a decision might not appear "prudent," or at least it might be difficult to demonstrate its prudence if the need should arise. ⁸⁶

Concerned about this possible bias in the effects of the prudence rule, the Labor and Treasury Departments issued a joint statement in 1989 indicating that pension fund managers do not have to sell shares in tender offers to meet the prudence requirement if they have reason to believe that long-term investment interests are better served by holding the shares. ⁸⁷ The statement has apparently had little effect on pension fund behavior. ⁸⁸

The second way in which the prudence rule may increase pension fund turnover is more indirect. The prudence rule has been interpreted legally to require expertise in investment management; hence, the fiduciary of a pension plan must either be a professional investment manager or must hire one to handle the investment of fund assets. ⁸⁹ Hiring outside investment managers may also be perceived as a way of partially shifting the responsibility for investment decisions. ⁹⁰ If the prudence rule results in greater reliance on outside investment managers, it could possibly contribute to higher pension fund turnover because many of these managers have a highly active approach to portfolio management. On the other hand, this is not a necessary outcome; nothing in ERISA prevents use of index funds or active investment managers with low turnover rates as the outside investment managers.

While these arguments have some plausibility, the statistical analysis reported in subsection I below does not find evidence that ERISA contributed to higher pension fund portfolio turnover rates. Hence, either the statistical analysis is not capable of detecting the effects or, while the effects may occur to some extent, they do not contribute significantly to the higher turnover rates.

G. FADS, SPECULATION, OR "SHORT-TERMISM"

Other explanations offered by some analysts for high portfolio turnover rates are that some stocks become "faddish" or that investment managers engage in speculation. ⁹¹ Some critics also point to "short-termism," the alleged tendency of investment managers to focus unduly on short-term earnings or growth prospects and insufficiently on longer-term considerations.

A survey of investment managers by Shiller and Pound found evidence they interpreted as being consistent with these hypotheses. ⁹² The authors selected a random "control" group of stocks and an "experimental" group of stocks. The experimental group consisted of stocks having the highest price increases during the previous year and high price/earnings ratios. They sent questionnaires to institutional stockholders identified as holding one of the stocks in reports filed with the Securities and Exchange Commission. They received 30 useable responses from institutional investors holding the control stocks and 41 useable responses from those holding the experimental stocks.

Shiller and Pound were primarily interested in investigating excess price volatility in the stock market, but their evidence also relates to the turnover issue. They found that the investors holding the experimental shares were more likely to be influenced in their investment decisions by information from other investors or investment newsletters and less likely to have selected the investments based on careful research. The control group investors most commonly selected their stocks based on considerations of fundamental value. The experimental group selected stocks primarily based on short-term growth expectations. The majority (55 percent) of the experimental group said they would sell their stocks in response to negative news related to the stock, but only 21 percent of the control group said they would. Perhaps not surprisingly, the turnover rate of the experimental stocks was 70 percent higher than the rate for the control stocks.

While the results found by Shiller and Pound are interesting and suggestive, there is also contrary evidence. If fads develop among investment managers, then they should exhibit "herding" behavior; that is, many managers should simultaneously be buying or selling the same stocks. Recent research, however, indicates that herding is not prevalent. Lakonishok, Shleifer, and Vishny studied the trading behavior of 769 pension fund equity investment managers over calendar quarters from 1985 through 1989. ⁹³ For the most part, they found no evidence of herding. While some managers were selling stocks of a particular company, others were buying, so, on net, the actions of the various managers essentially canceled each other. The only exception was in the stocks of small companies, where weak evidence of herding was found. These stocks are not traded extensively by institutional investors, however.

The price patterns of stocks can also provide indirect evidence regarding the trading behavior of investment managers. If fads, speculation, and inappropriate emphasis on short-term considerations were common and widespread, the prices of stocks would frequently be pushed upward or downward to levels inconsistent with their fundamental values. There is a large body of research devoted to studying whether this occurs, but interpretation of the evidence is complex and controversial. Some authors believe they have found evidence of systematic mispricing. Others have different interpretations. ⁹⁴ At the moment, therefore, understanding the extent to which these factors contribute to the high pension fund turnover rates requires further research.

H. "WINDOW DRESSING"

Some authors have ascribed some portion of pension fund turnover to what has been called "window dressing." This term has been used most often in connection with one type of active investment management behavior, but it will be used here to refer to any effort to create a false impression of more astute active portfolio management.

The activity typically identified as window dressing is selling stocks that have performed badly in an active portfolio just prior to reporting time. ⁹⁵ When investment managers issue their quarterly or annual reports, the stocks in their portfolios are listed. If the portfolio has several stocks that have underperformed the market, the portfolio manager may expect that clients will be critical of these investments and question the skill with which the portfolio is being managed. Selling some of the underperforming stocks prior to reporting time does not improve investment performance at that point, but may avoid the criticism. Of course, some of these stocks may be sold for other reasons in the normal course of managing the portfolio, but the allegation is that some stocks which would otherwise be retained in the portfolio are sold to create an appearance of greater skill in investment selection.

At least one analysis has found some evidence of window dressing. Lakonishok, et. al., examined the equity portfolios of 769 pension funds and found evidence that the funds tended to sell losing stocks disproportionately, and this tendency was most pronounced in the fourth quarter, which precedes the annual financial report. ⁹⁶ This finding is consistent with the expectation that more client attention is focused on the annual report than on quarterly reports. The authors also found that the funds tend to BUY underperforming stocks disproportionately. Buying a stock after it has declined may appear shrewd; holding a stock while it is declining probably does not. They also found that these tendencies are weaker for large pension funds, which is consistent with the expectation that large funds would be more sophisticated in monitoring their investments so window dressing by their investment managers would be less effective.

Another kind of behavior that may be regarded as a form of window dressing was discussed in a theoretical paper by Trueman. ⁹⁷ He observed that it is difficult for investors to judge the skill of their active investment managers. Even the most skillful active investment managers will buy some stocks that do not perform very well and will have periods when their portfolios underperform the market. Under such circumstances, investors probably rely on other indicators as well as the rate of return on the portfolio to draw conclusions about the investment manager's skill. One such indicator may be the portfolio turnover rate. If an active investment manager is skillful at obtaining information that provides insights regarding which stocks would be good investments, this presumably will lead to trading based on this information. Hence, the amount of trading that active investment managers engage in may be an indirect indicator of their ability to obtain special investment information or insights. If so, there is an incentive for investment managers who do not have special information to increase the amount of their trading activity to create the impression that they do. ⁹⁸

Nonetheless, it is unlikely that trading merely to create the impression of greater skill in investment selection accounts for much of the increased turnover rates of pension funds. Pension plan sponsors are increasingly monitoring the trading practices and turnover rates of their investment managers in an effort to minimize costs. ⁹⁹ In an environment in which pension sponsors strive to purchase the best investment management (however defined) at the least cost, an investment manager who increases turnover unnecessarily will become uncompetitive. Furthermore, as shown in Figure 1 (on page 6), mutual funds have the highest turnover rates among institutional investors. But most mutual fund shareholders probably do not know the turnover rates of the funds in which they hold shares or which stocks are held by the funds. This information is buried in the prospectus which most mutual fund investors do not read carefully. Hence, it is unlikely that mutual fund managers are tempted to increase turnover to appear more skillful. Other factors must be more important in explaining the high turnover rates.

Both window dressing and the factors discussed in the previous subsection (fads, speculation, and "short-termism") would raise turnover without enhancing return. If these factors do, in fact, significantly affect turnover, therefore, they could erode any potential positive relationship between turnover and the rate of return even if the financial markets are not perfectly efficient.

I. STATISTICAL ANALYSIS OF PENSION FUND TURNOVER

This subsection presents a summary of a statistical analysis of pension fund turnover in an effort to explain the substantial increase from the 1960s to the 1980s. A fuller discussion of the analysis is provided in the Appendix.

Ideally one would like to examine the effect of each factor identified as a potential contributor to higher pension fund turnover rates in the discussion in subsections A through H of this section. Unfortunately, this cannot be done very precisely for some of the factors because they are not easily quantified.

The approach here uses six variables to try to explain the pattern of pension fund turnover rates: a proxy variable for the influence of modern portfolio theory, transaction costs, futures transactions, a bull market indicator, ERISA, and a variable for the range of market prices each year. The analytical procedure is multiple regression analysis. The turnover rates used in the analysis discussed here are a composite data series created from the three series of pension fund turnover rates graphed in Figure 1 (on page 6). ¹⁰⁰ For each year in which two or more turnover rate estimates are available, the figure for the composite series was derived by averaging the estimates. The resulting series spans the period from 1955 through 1987.

The first explanatory variable used in the analysis is intended to represent the spread of the influence of modern portfolio theory on actual portfolio management. No available statistic comes to mind as a reasonable measure or estimate of this factor, so a somewhat ad hoc proxy is used. The proxy takes a value of O from 1955 through 1964 (prior to the time that modern portfolio management theory is thought to have had any practical influence); it rises linearly from O to 1 over the period 1965 to 1980 (the period during which the influence of the theory is thought to have grown and become virtually universal), and it takes a value of 1 thereafter. A positive relationship is expected between this variable and portfolio turnover.

The second variable used in the analysis is the cost of transactions to institutional investors expressed as a fraction of average share price (the variable reflects the brokerage commission and the securities transactions taxes; it does not reflect the bid- ask spread or the market impact of trades). This variable is graphed in Figure 3 (on page 24). A negative relationship is expected between transaction costs and turnover; that is, as transaction costs decline, turnover is expected to rise.

The third variable is market-index futures trading volume. It serves one or both of two purposes. It serves as a proxy for the extent to which pension fund equity turnover may be overstated due to trading in futures and options. And/or it serves as an indicator of the extent to which trading strategies using futures also generate turnover in stocks. Futures and options on the equity market indexes, which are the primary instruments used by the institutional traders, did not begin trading until 1982. Hence, this variable has non-zero values only in the last six years of the period covered by the analysis.

The fourth variable is the real S&P 500 Index at the end of the year (the S&P 500 Index divided by the GNP price deflator). This variable serves as the bull market indicator; more trading is expected to occur in bull markets.

The fifth variable used in the analysis is a dummy variable for ERISA. This variable tests whether ERISA resulted in a shift in turnover rates. It is not clear whether such a shift would have been immediate or spread over several years, so several variations of the variable were tried, yielding similar results.

The sixth variable is the range of the S&P 500 Index expressed as a fraction of the prior year's closing index. The expectation is that this variable will be positively related to changes in turnover. This relationship can be derived from one or more of several lines of reasoning. One is that the range of stock prices serves as a proxy for the amount of new information affecting the market, which drives the trading volume of active investment managers. A closely related reason is that greater price variability requires more trading to rebalance portfolios. Greater price variability also would result in higher turnover by investment managers subject to fads and those engaging in speculation. For these managers, greater price variability presumably creates more trading "opportunities." "Window dressing" could also result in higher turnover in a more variable market because more stocks may need to be traded to rid portfolios of those that had underperformed the market.

The results of six regressions are discussed in the Appendix. One regression equation is shown and discussed here to illustrate the results and their implications. The equation is as follows:

      T = 0.3831 M - 18.6916C + 0.0049F + 0.0358S&P + 0.1484R

The implications of the regression results can be revealed by using them to attempt to explain the change in pension fund turnover from 1974 to 1987, a period that brackets most of the turnover increase shown in Figure 1 (on page 6). In 1974, the composite turnover rate was 16.6 percent; in 1987 it was 65.8 percent. The changes in the independent variables in the equation explain 71 percent of the increase in turnover during this period; the rest of the increase remains unexplained by this model. The biggest contributor to the increase is the change in futures trading from O in 1974 to 25.1 million contracts in 1987. According to the coefficient in the equation, this change increased measured turnover by 12.4 percentage points. The second largest contributor to the turnover increase according to the regression equation is the spread of the influence of modern portfolio management theory, which increased turnover by 11.5 percentage points during this period. The decrease in transaction costs was the third most significant factor, increasing turnover by 6.7 percentage points. The real increase in the S&P 500 stock index over the period accounts for 3.0 percentage points of the rise in trading. Finally, the wider trading range of the S&P 500 in 1987 (46.6 percent) compared to 1974 (38.5 percent) explains 1.2 percentage points of the increase in turnover according to the estimates.

These results suggest that the factors discussed in this section which may have perceived negative public policy implications have not exerted significant influence in raising pension fund portfolio turnover rates. The ERISA legislation was not intended to increase pension fund turnover rates. If it had this effect, that would be regarded by some observers as a negative (though perhaps unavoidable) consequence of the legislation. The regression results, however, suggest that ERISA was not an important factor. ¹⁰³ Since other types of institutional investors also increased their turnover during this period (see Figure 1, page 6), it is not surprising to find that ERISA did not play a significant role.

The effects of fads, speculation, short-termism, and window dressing on portfolio turnover rates also could be viewed negatively from a public policy perspective. The variable in the regressions which relates most directly to these factors is the range of the S&P 500. ¹⁰⁴ This variable also relates to active investment management (which would also be viewed negatively by adherents of efficient markets theory), so it is not clear how much of its influence is attributable to which factors. While this variable is statistically significant in the regressions, it does not account for much of the increase in pension fund portfolio turnover rates. The results, therefore, suggest that the influence of these factors has not been substantial. Of course, it also must be acknowledged that the variables in the regressions are, at best, imperfect proxies for these factors, and the regressions do not fully explain pension fund turnover. Hence, these or other factors could influence pension fund turnover in ways not captured by the regressions.

A further implication of the results is that the decrease in transaction costs is responsible for only a limited amount of the increase in pension fund turnover. This result contrasts with claims by some authors that the increase in turnover can be explained entirely by the decreased transaction costs since the mid-1960s and especially since the deregulation of brokerage fees in 1975. ¹⁰⁵ The analysis here suggests that the development of the futures and options markets and the influence of modern portfolio management theory have played larger roles.

V. THE EFFECTS: TURNOVER RATES AND RATES OF RETURN

The discussion in section II indicated that the appropriate criterion for judging the investment behavior of pension funds is the effect on the rate of return earned by and the risk borne by the funds. Section III delineated the relationship between turnover rates and rates of return under three different theories of the financial markets. This section reviews the rather limited factual evidence regarding this relationship. ¹⁰⁶ Generally, the most recent results seem to be most consistent with the implications of the costly information model of the financial markets (see the discussion in section III B).

More research has addressed the overall rate of return than the role of turnover in determining the rate of return. Much of the work on turnover focuses on market timing -- switching assets among stocks, bonds, and cash equivalents -- rather than turnover within the equity component of a portfolio. And more research has focused on mutual funds than on pension funds because mutual fund data are more readily available. Mutual funds and pension funds are, of course, not identical entities. Nonetheless, the research on mutual funds is relevant here because mutual funds use similar investment methods, and in some cases the same investment managers, as pension funds.

The earliest studies of mutual funds found that they underperformed the market averages or, if they outperformed the averages on a gross basis (before management fees and expenses), the margin was not sufficient to cover their management expenses. ¹⁰⁷ These studies concluded that investors in mutual funds were apparently purchasing diversification of their investments, not superior returns. This general conclusion regarding the results of active investment management was part of the impetus behind the movement of pension funds and other institutional investors toward index funds.

More recent studies have found results indicating that mutual funds now match the performance of the market indexes on a net (after expenses) basis. ¹⁰⁸ A recent exhaustive review of this literature concluded, "The results of statistical studies of mutual funds' returns are generally consistent with the hypothesis that funds' risk-adjusted performance, net of expenses, is statistically indistinguishable from index funds." ¹⁰⁹

Some of these studies specifically considered the effects of portfolio turnover. A recent example is an article by Ippolito that evaluated the performance of 143 mutual funds over the period 1965 through 1984 and found that they beat the market by margins sufficient to cover their costs. On a net of costs basis, the returns earned by the mutual funds were independent of the turnover rate, fees, or expenses. That is, funds that followed a more active investment approach earned an excess return sufficient to cover the additional costs of the higher turnover. The author interpreted these results as being consistent with the costly information model. ¹¹⁰

Research on pension fund performance has followed a pattern similar to that on mutual funds. The earlier work found that pension funds, on average, underperformed the stock market. More recent research concludes that pension fund performance matches the stock market.

An extensive study of pension fund performance was done by Berkowitz, Logue, & Associates, Inc. for the U.S. Department of Labor in 1986. ¹¹¹ The study analyzed the investment performance of 120 private pension plans covered by ERISA and 36 public pension plans over the period 1968 through 1988. With the exception of the bear market years of 1973 to 1975, the study concluded that the equity component of the pension funds did not underperform the S&P 500 stock market index. The study also found no consistent relationship between the portfolio turnover rate and the rate of return earned by the pension funds, either for the total portfolios or for just the equity components. This result is also consistent with the costly information model of the stock market.

Probably the most cited study of the effects of portfolio turnover on pension fund investment performance was done by Ippolito and Turner. ¹¹² Their research was based on data for approximately 1,500 pension plans over the period 1977 through 1983. They compared results for the pension plans with results for mutual funds derived from the data used for Ippolito's study (cited in footnote 110). Like earlier studies, they found that pension funds earned lower rates of return (on a risk-adjusted basis) than mutual funds. They attributed at least part of the difference, however, to a deficiency in the capital asset pricing model, which was used to compute the returns, and to overstatement of the returns of the mutual funds because load charges were not deducted.

Ippolito and Turner found that aggregate portfolio turnover was not related to the rate of return earned by pension funds; this was true both for turnover associated with market timing and turnover associated with security selection. They also tested the relationship involving turnover in just the stock portion of the portfolio and found it to be negative. The authors summarized the implication of their finding as follows: "the average pension plan in 1983 had stock turnover equal to roughly 30 percent of its overall portfolio value; the results imply that this cost the typical plan roughly 60 basis points." ¹¹³ For mutual funds, the authors found that none of the turnover measures, including stock turnover, were related to the rate of return.

A recent article by Brinson, Singer, and Beebower also derived results generally consistent with the costly information model of the financial markets. ¹¹⁴ These authors studied investment results for 82 large pension plans over the period 1977 through 1987. Unlike most of the other articles on this topic, this paper did not use regression analysis to estimate the effects of the various factors contributing to the rate of return. Instead, the contribution of each factor was estimated with reference to a "normal portfolio" for each pension plan. The normal portfolio for a pension plan was defined as a portfolio with constant asset allocations to the components of the portfolio -- stocks, bonds, Treasury bills, etc. -- equal to the average allocations for the pension plan over the 10-year period, with the assets in each component indexed. The contribution of market timing -- switching assets from one market to another -- was estimated by calculating the incremental return of a portfolio using the actual allocations of the pension plan each year (with each component indexed) compared to the normal portfolio. The contribution of security selection -- turnover within each component -- was estimated by calculating the incremental return of a portfolio with the actual return within each portfolio component compared to the normal portfolio.

The result of the analysis was the conclusion that, on average, nearly all of the return of the pension funds was determined by the basic decision regarding the allocation of the assets to the various investment markets. Active investment management, involving both market timing and security selection, had essentially no effect on the rate of return.

Hence, using the effect on the rate of return as the criterion for judgment, the review of research in this section provides no clear basis for concern regarding pension fund portfolio turnover rates. With the exception of the finding by Ippolito and Turner that turnover in the equity component of pension portfolios detracts from the overall rate of return, the results are generally consistent with the costly information model of the securities markets. That is, in general, the portfolio turnover rate and the rate of return (net of costs) are independent. Stated differently, higher turnover neither subtracts from nor adds to the rate of return.

VI. THE INTERMEDIARIES: PENSION FUNDS AND INVESTMENT MANAGERS

An additional criticism of pension funds regarding short-term investing is that they evaluate investment managers based on performance over very short periods; calendar quarters are usually mentioned. The statement of Ira M. Mil1stein, Chairman of the New York State Pension Investment Task Force, . . . before the Joint Economic Committee is typical. Referring to a discussion among money managers at a conference on short-term investment, he stated:

"The participants cited as a contributing factor to a short-term focus the fact that money managers are often evaluated, hired, compensated, and fired based on short-term performance. All too often, that performance is measured quarterly, perhaps only over a year . . . " ¹¹⁵

This issue is not related directly to the question of how long pension funds hold onto their common stock investments; a pension fund could have a very high turnover rate in a portfolio managed by an investment manager with whom it has a long-term relationship. The issue is generally raised as supporting evidence for the criticism that pension funds have a short-term investment perspective.

In considering the way pension funds evaluate their investment managers, it is useful to distinguish between passive and active managers. Passive managers are hired to maintain a portfolio with a specified structure, usually tracking a market index such as the S&P 500. Their management fees are low because they do not have to do extensive research on individual companies or industries, and their low turnover results in low brokerage commissions (as a fraction of assets under management). They should be evaluated based on how well they track the designated market index.

Active managers, on the other hand, are hired to beat a market index. That is the service they sell, and the service is substantially more expensive than investing in an index fund. The management fees of active managers are higher, in part, because of higher research costs. Their trading costs are also higher because of higher turnover rates. In hiring active managers, pension funds are paying extra for extra performance, so it is appropriate to evaluate the performance on an ongoing basis. Indeed, doing otherwise would amount to a failure to fulfill fiduciary responsibilities. If an active manager cannot provide performance that is superior to an index fund, then the assets should be placed with another active manager who can provide superior performance, or they should be indexed.

Two recent developments have affected the evaluation of active investment managers. The first is the development of custom benchmarks to judge their performance; the second is performance fees. No longer are active managers all judged on the basis of their performance compared to the S&P 500 market index. Since active managers have many different styles, and in fact are frequently hired by pension fund sponsors to pursue specific investment approaches, the S&P 500 is in many cases an inappropriate benchmark for judging their performance. One manager could beat the S&P 500 and another could lag behind, even though they both may perform equally well in pursuing their designated investment approach.

To address this problem, some pension funds and their investment advisors have devised custom benchmarks for the investment approaches followed by their active portfolio managers. These custom benchmarks are market indexes similar in concept to the S&P 500, except they are structured to track model portfolios representing the designated investment approaches of the active managers. These custom benchmarks have several implications. First, investment managers are protected from the misleading appearance of poor performance during a period when their investment approach underperforms the broad market, so long as they perform well in comparison to their benchmark. Second, it is more difficult for managers to achieve superior performance; even if they outperform the broad market, they may not exceed the performance of their custom benchmarks. On the other hand, in evaluating their active managers, pension sponsors have to understand that the amount by which fund managers can be expected to outperform custom benchmarks is limited. Taken to its limit, judging active managers against custom benchmarks, in effect, converts all managers into index funds, each with its own index.

Performance fees make the management fee paid by the pension fund to the investment manager a function of performance. Generally, the fee is comprised of a base fee and a variable amount. The base fee is usually lower than the management fee would be under a fixed fee arrangement. The variable amount of the fee is usually a percentage multiplied by the return achieved by the manager minus the return on a designated index, either the S&P 500 or a custom benchmark. In some arrangements, the variable portion of the fee can be negative as well as positive; in others, the variable amount can add to the fee but not reduce it. Most performance fee formulas have upper and lower limits on the amount of the fee. ¹¹⁶ Performance fees are now used by 15 percent of pension funds to compensate their investment managers; 26 percent of public pension plans use them (the U.S. Department of Labor did not approve use of performance fees for plans governed by ERISA until 1986). ¹¹⁷

Performance fees also have several implications. They provide a greater incentive for investment managers to achieve superior performance, since their compensation is at least partly determined by the performance. This means managers will receive reduced fees when their performance is poor, but when they achieve their objectives of providing superior performance they will receive bonuses. There is some evidence from a study of mutual funds that performance fee arrangements do improve investment performance. ¹¹⁸

Performance fee arrangements may also make pension plans more tolerant of periodic weak performance by investment managers. Since the management fee declines when performance does not exceed the benchmark, the loss to the . . . pension fund (compared to investment in an index fund) is smaller. Hence, a performance fee system makes an investment manager's compensation more dependent on performance, but it may simultaneously reduce the likelihood of being terminated for a period of weak performance.

Of course, a performance fee arrangement still leaves the question of the time period over which performance should be evaluated. In determining the management review period, pension funds have to balance costs on opposite sides of the issue. On the one hand, it is expensive to retain an active manager whose performance in terms of extra return fails to cover his cost. On the other hand, it is expensive to switch investment managers; the cost amounts to about 1.5 percent of the assets under management. ¹¹⁹

There is not a great deal of information about what constitutes standard practice on this issue. A recent survey reported that in 44.9 percent of performance fee systems, the fee is based on performance over a 1-year period; 4.1 percent of the systems use a 2- year period, and 40.8 percent use a 3 to 5-year period. ¹²⁰ Another indicator of investment manager review periods is discussions in guidebooks on pension asset management; they tend to suggest that 3 to 5-year periods are typically used to evaluate investment manager performance. ¹²¹

There is also some survey information regarding how long investment managers who are terminated by pension funds had served before being dismissed. ¹²² One survey by Sentinel Pension Institute reported that over the 5-year period ending in 1987 no pension fund had terminated an investment manager after less than one year of service; 7.5 percent of managers who had been terminated lost the account after a period of one to two years, 52.5 percent were dismissed after 2 to 5 years, and 40 percent were retained longer than 5 years. ¹²³ Another survey by the Financial Executives Institute asked pension funds over what time period investment managers were evaluated and reported that the "nearly unanimous response" was 3 to 5 years or over "a complete market cycle." The survey also gathered data on investment managers that had been terminated. About 8 percent of investment managers retained by the pension funds were terminated annually. The average tenure of those terminated was 8.7 years in 1987, 8.0 years in 1988 and 7.6 years in 1989 ¹²⁴

Summarizing the information on this topic: quarterly data are used in evaluating investment managers, but over time periods of at least one year. Very few investment managers are terminated after only one year. More typically, managers are given from 3 to 5 years to prove their worth. Some observers claim this is insufficient time. On the other hand, pension funds pay much more for active management of their assets than it would cost to index them. If the active management does not yield superior returns, arguably the cost should not be paid indefinitely.

VII. THE IMPLICATIONS

This report has examined whether pension funds are short-term investors, that is, whether they place undue emphasis on short-term investment results. The criticisms of pension fund investment practices generally focus on the implications of short-term investment for the companies in which the funds invest. This report has taken the perspective, however, that the investment practices should be judged by whether they are consistent with their objective: that of maximizing the risk-adjusted rate of return earned on pension fund assets. This objective is the natural goal of pension plan sponsors and beneficiaries; it is also imposed by Federal law and by the fiduciary responsibilities of pension fund managers.

Based on this criterion, there is not much evidence that pension funds trade their stocks too aggressively. Portfolio turnover rates of pension funds and other investors increased substantially (that is, average holding periods shortened) after the mid-1960s, but there are several reasonable explanations for this increase. Higher portfolio turnover may have been a consequence of the development of modern portfolio management theory. The options and futures markets developed, providing low-cost means of more precisely controlling risk and expected return. Transactions costs decreased substantially. In addition, bull market conditions prevailed for much of the 1980s.

Some factors suggested as explanations for high pension fund portfolio turnover may have negative public policy implications. The Employee Retirement Income Security Act imposed restrictions and responsibilities on pension plan fiduciaries that may have unintended (though perhaps unavoidable) effects on portfolio management. Some investment managers may be influenced by fads in securities selection, may engage in speculation, or may be short-sighted in their investment analysis. Some may engage in "window dressing." While each of these influences may occur to some extent, there does not seem to be much evidence of them in overall market behavior, and there are reasons to believe they are limited. If these factors were important, for example, high portfolio turnover would be associated with a reduced rate of return.

Most of the evidence, however, suggests that this is not the case. Most recent research has found that turnover rates and rates of return are unrelated; that is, higher turnover does not lead to higher or lower rate of return.

This relationship has sometimes been stated as a criticism: "active management of pension assets has not resulted in increased values for the participants and beneficiaries." ¹²⁵ If higher portfolio turnover resulted in higher rates of return, however, this relationship would imply a disequilibrium condition according to the costly information theory of the securities markets. It would signal the existence of opportunities to earn superior profits that were not being fully exploited. As investment managers traded more actively to capture these profits, the market would move toward the equilibrium condition in which all investment strategies should provide roughly equivalent risk-adjusted rates of return. The casual observer may view this equilibrium and conclude that if higher turnover does not yield higher returns, then portfolio managers should simply reduce their turnover rates. But this would once again create a disequilibrium in which profitable strategies were not being fully exploited, and the adjustment would repeat itself.

Given the dynamic nature of the financial markets, the market may always be moving toward equilibrium but never quite achieve it. The equilibrium condition, therefore, may never be observed precisely (measurement error can also be significant), but the movement should tend to be in this direction.

In general terms, the recent experience of pension funds in the stock market is consistent with this expectation. Until a few years ago, pension funds were criticized for investment performance which, on average, lagged behind the market. It is possible that they engaged in active trading beyond the point of profit maximization. This experience led a number of pension funds to index a significant portion of their assets. The quality of investment management probably also has improved over the last decade with the extensive computerization of the industry, the development of the options and futures markets, and the creation of new highly sophisticated investment strategies and techniques (the products of the so-called "rocket scientists"). The more recent research tends to find results more consistent with the expected equilibrium condition.

Given the nature of this equilibrium, it is interesting to consider the potential effects of a public policy designed to slow down active security trading by pension funds. The proponents of applying a tax to short-term investment profits of pension funds have said their goal is not to raise tax revenue, but to reduce short-term stock market investment and its presumed negative effects on corporate incentives. ¹²⁶ If a policy could be designed to achieve these objectives, it probably would not adversely affect the rate of return received by the pension funds. The principal "losers" from such a policy would be investment managers and stock brokers. The financial markets would be somewhat less efficient because less resources would be devoted to obtaining new information, liquidity would decrease, prices might not track fundamental value as closely, and financial capital would be allocated less efficiently. Some analysts, however, have argued that these effects would be small and inconsequential. ¹²⁷ Other investors, including less active pension funds, may or may not be disadvantaged by these changes depending on the effect on the overall rate of return in the market.

These policy effects may be of more theoretical than practical interest, however, because it may be very difficult to design a tax (or other policy instrument) to have only the desired effect and no (or minimal) unintended consequences. It would be difficult to design the tax, for example, so it did not apply to the limited but inevitable short-term trading that occurs in rebalancing index funds and other balanced portfolios. It would be difficult to avoid discouraging hedging and other uses of options and futures, many of which facilitate long-term holding of stocks. It also would be difficult to limit the effect of the tax to short-term equity trades of pension portfolios without disrupting the relationships among stocks, bonds, and derivative instruments and without creating opportunities for arbitrage or avoidance. Any revenue collected by the tax would reduce the rate of return received by pension funds and other affected investors. These side effects of potential policy changes, therefore, need to be studied carefully.

APPENDIX: STATISTICAL ANALYSIS OF PENSION FUND TURNOVER

This Appendix presents a statistical analysis which attempts to explain the substantial increase in pension fund turnover from the 1960s to the 1980s.

THE EXPLANATORY VARIABLES

Six independent variables are used in the analysis; they are discussed here in the same order as in section IV of the report.

The first variable is intended to represent the spread of the influence of modern portfolio theory on actual portfolio management. No available statistic comes to mind as a reasonable measure or estimate of this factor, so a somewhat ad hoc proxy is used. The proxy takes a value of O prior to the time that modern portfolio management theory is thought to have had any practical influence; it rises linearly from O to 1 over the period the influence of the theory is thought to have grown and become virtually universal, and it takes a value of 1 thereafter.

Modern investment theory originated in an article by Markowitz on portfolio selection which was published in 1952. ¹²⁸ While the article had a substantial impact on finance theory, it had no direct or immediate impact on actual portfolio management because the computations that would have been required to implement the theory were beyond the capabilities of the time. Application of the theory became practical with the publication of an article by Sharpe in 1963 which provided a simplification of the theory that, along with further refinements over the years, has become known as the Capital Asset Pricing Model (CAPM). ¹²⁹ Allowing for some lag in application, the proxy for the influence of modern portfolio management theory used in the analysis begins taking non-zero values in 1965.

It is more difficult to identify a year as the time when the influence of modern portfolio theory on actual portfolio management became universal. The decade of the 1970s is sometimes referred to as the period during which institutional investors replaced "old style" portfolio management with modern quantitative, computer-based techniques. Hence, 1980 is chosen as the end of this period. The proxy for the influence of modern portfolio management theory, therefore, assumes a value of 0 from 1955 through 1964, increases by 0.0625 per year from 1965 through 1980, and equals 1 from 1980 through the end of the period.

The second variable used in the analysis is the cost of transactions to institutional investors expressed as a fraction of average share price. The relationship between this variable and turnover is expected to be negative; that is, as transaction costs decline, turnover is expected to rise.

Transaction costs have four elements: the brokerage commission, the securities transactions tax, the bid-ask spread, and market impact. It is possible to put together time series over the period covered by the regressions only for the first two of these elements. ¹³¹

Brokerage commissions paid by institutional investors are not reported by any single source for the entire period under study; they must be pieced together from a variety of sources. Per-share brokerage costs for 1977 and later are provided by Greenwich Associates based on their surveys of institutional investors. ¹³¹ The costs for 1975 and 1976 are derived from the Greenwich Associates data on commission discounts, comparing those years with the discount and per share trading cost in 1977 (Greenwich Associates does not report brokerage costs for 1975 and 1976). Brokerage commissions for the years 1964 through 1968 are estimated based on data published by the Securities and Exchange Commission on brokerage fees prior to deregulation ¹³³ and on commission "give-ups" during this period. ¹³³ Trading commissions for 1969 through 1974, for which no reported data are known, were estimated by linear interpolation between the estimated values for 1968 and 1971. ¹³⁴ Commissions prior to 1964 were estimated based on the commission schedules reported by Smidt. ¹³⁵ The commission per share data were divided by the average price of shares traded each year as reported in various issues of the Fact Book published by the New York Stock Exchange (NYSE).

The Federal securities transaction tax was repealed effective January 1, 1966. It had a rate of 4 cents per $100 of stock sold. In estimating transaction costs the tax is assumed to fall one half on the seller and one half on the buyer of the securities. Estimates of the burden of the New York State transactions tax, which ended in 1981, were derived based on collections data for the tax and the total value of shares traded as reported in the NYSE Fact Book. The resulting series of estimates of transactions costs is shown in figure 3 on p. 24.

The third variable used in the analysis is market-index futures trading volume. As discussed in section IV D, this variable serves one or both of two purposes. It serves as a proxy for the extent to which pension fund equity turnover may be overstated due to trading in futures and options. And/or it may indicate the extent to which trading strategies using futures also generate turnover in stocks. Futures and options on the equity market indexes, which are the instruments primarily used by the institutional traders, did not begin trading until 1982. Hence, this variable has a value of zero until that date. Once initiated, however, trading grew rapidly; over 25 million market-index futures were traded in 1987. ¹³⁶

The fourth variable used in the analysis is the real S&P 500 Index at the end of the year (the S&P 500 Index divided by the GNP price deflator). This variable serves as the bull market indicator. The presumption is that more trading will occur in bull markets.

The fifth variable is the ERISA variable. The purpose of this variable is to test whether ERISA resulted in a shift in turnover rates. It is not clear whether such a shift would have been immediate or spread over several years. For this reason, several variations of the variable were tried; the results were qualitatively similar. The variable used in the regressions shown below assumes that a shift would have occurred over a two-year period. It has values of 0 in years prior to 1974, 0.5 in 1974 (the year ERISA was adopted), and 1 in 1975 and later years.

The sixth variable is the range of the S&P 500 index expressed as a fraction of the prior year's closing index. ¹³⁷ The expectation is that this variable will be positively related to changes in turnover. This relationship can be derived from one or more of several lines of reasoning.

The theory of stock market trading posits that trading by active investors occurs in response to new information that is interpreted differently by different traders. ¹³⁸ New information and its interpretation by traders cannot be quantified directly. But new information also leads to price changes in the market, which can be measured. The presumption is that if prices vary over a wider range during a year, the flow of new information affecting traders' valuations of stocks was greater and therefore transactions volume will be greater. ¹³⁹

A closely related reason to expect a positive relationship between the range of prices and trading activity is that greater price variability requires more trading to rebalance portfolios. For portfolios that are managed to maintain roughly fixed proportions of total assets in certain companies, certain sectors, or in stocks with certain characteristics (index funds are managed this way, for example), higher price variability will require more trading to maintain the desired asset allocation.

Greater price variability also would result in higher turnover for investment managers subject to fads and those engaging in speculation. For these managers, greater price variability presumably creates more trading "opportunities." "Window dressing" could also result in higher turnover in a more variable market because more stocks may need to be traded to rid portfolios of those that had underperformed the market.

THE RESULTS

Results of the regressions are reported in table 2 on the next page. The dependent variables in the regressions are the pension fund turnover rates graphed in figure 1 (on page 6). For each dependent variable, the first regression equation includes all of the independent variables. The second equation omits the variables that did not have significant coefficients.

In equation 1, which is based on the SEC turnover data, the coefficients on all of the variables have the expected signs. The constant term and the coefficient on the ERISA variable are not significantly different from 0. The futures variable does not appear in this equation because the index futures did not begin trading until after the period covered by the SEC turnover data. The constant term and the ERISA variable are dropped in equation 2.

The Cochrane-Orcutt procedure was used to correct for serial correlation in each of the regressions. ¹⁴⁰ The serial correlation coefficient (RHO) is quite high, indicating that the model does not provide a complete explanation of changes in the pension fund turnover rate (a condition that is not unexpected).

 Table 2

Regressions were also performed using the other two series of pension fund portfolio turnover rates shown in figure 1 on page 6. The results were less satisfactory than equations 1 and 2 because serial correlation between the independent variables is a problem over the shorter time periods spanned by these data. The results based on the McCarthy-Turner data are not shown; results based on the Berkowitz-Logue data are shown in equations 3 and 4.

In regression 3 the transaction cost coefficient and the real S&P 500 coefficient are not significant. The ERISA coefficient is significant (barely) and has the wrong sign. The ERISA coefficient is also highly correlated with some of the other independent variables. When it and the insignificant constant term are dropped from the regression, the other variables all have significant coefficients with the expected signs, as shown in equation 4.

To make use of all the available pension fund turnover rate data, another set of regressions was performed using a composite data series created from the three series shown in figure 1. For each year in which two or more turnover rate estimates are available, the figure for the composite series was derived by averaging the estimates. The resulting series spans the period from 1955 through 1987. This approach has obvious imperfections: at the endpoints of the actual series the composite series will move in a way not reflected in any of the actual series. The tradeoff is the ability to derive estimates based on the longer time period. The regression results, shown in equations 5 and 6, are quite consistent with those in the other equations (especially equations 1 and 2), so the distortions do not appear to be large.

The implications of the estimated equations can be explored by using them to attempt to explain the changes in turnover rates over some periods. This is done in table 3 on the next page. The first two columns in the table use equation 4, based on the Berkowitz-Logue data, and equation 6, based on the composite data, to explain the increase in the turnover rate from 1974 to 1983, a period spanned by both data sets. Over this particular period, the equation based on the Berkowitz-Logue data explains more of the change in the turnover rate. The third column in the table uses equation 6 to explain the change in the turnover rate from 1974 to 1987, a period that extends beyond the Berkowitz-Logue data.

While the equations and time periods differ in terms of the total explanatory power of the estimated equation and the contribution of each independent variable, certain themes emerge. First, in the three sets of estimates shown in table 3, the two variables with the greatest explanatory power are the proxy variable for modern portfolio management and the variable for futures trading. Second, while the transactions cost variable is important, it accounts for no more than 30 percent of the shift in turnover explained by the estimated equations. Finally, while the two market- movement variables -- the real S&P 500 index and the range of the S&P 500 Index -- have significant coefficients in the regressions, they explain relatively little of the change in turnover rates over these periods.

 Table 3

² New York Stock Exchange, Institutional Investor Fact Book 1990, January 1990, p. 4.

³ Trades of 5,000 shares or more, a rough indication of institutional trading, accounted for 64.8 percent of volume on the New York Stock Exchange in 1990. See New York Stock Exchange, Fact Book 1991, 1991, p. 14.

⁴ The Impact of Institutional Investors on Corporate Governance, Takeovers, and the Capital Markers, Hearing Before the Subcommittee on Securities, Committee on Banking, Housing and Urban Affairs, U.S. Senate, 101st Congress, lst Session, October 3, 1989; Corporate Time Horizons, Hearings Before the Joint Economic Committee, U.S. Congress, 10lst Congress, lst Session, November 8 and 14, 1989; and Effects of Short-Term Trading on Long-Term Investments, Hearing Before the Committee on Finance, U.S. Senate, 10lst Congress, 2nd Session, March 21, 1990.

⁵ Mutual funds are referred to as "pass-through" entities because, so long as they meet certain requirements, for purposes of taxation they are not regarded as receiving the income on their investments The income "passes through" to their shareholders, who include it in their taxable income

⁶ Dan Rostenkowski, Introduction of a Bill Relating to Tax Treatment of Mutual Funds, Congressional Record, June 24, 1991, p. E 2323.

⁷ Kevin G. Salwen and Jonathan Clements, Short-Term Funds Trading Wins Support, The Wall Street Journal, September 18, 1991, p. Cl, C9.

⁸ The bill would extend the short-short rule to apply to pension funds, but not as an amendment to the tax code. Rather, the restriction would be added to the list of prohibited transactions under the Employee Retirement Income Security Act (ERISA).

⁹ Even if pension funds are short-term investors, this behavior may not have significant effects on stock market prices or the behavior of corporations. These issues are not considered in this report. They were the subject of an earlier analysis; see U.S. Library of Congress, Congressional Research Service, Stock Market "Short-Termism:" Implications for Corporate Planning Horizons, Report No 91-448 RCO by Donald W. Kiefer, May 29, 1991, 31 p.

¹⁰ The report does not address some topics that are related to short-term investment by pension funds. For example, the report does not discuss the role of pension funds in corporate takeovers and leveraged buyouts (LBOs), the subject of some of the criticisms of pension investment behavior. The report also focuses only on investment by pension funds in equity securities because this is where the criticism has been concentrated.

¹¹ This statement assumes that the turnover rate is representative of the investment behavior of the portfolio; that is, the turnover rate is not taken from a period when turnover was uncharacteristically low or high

¹² These data for 1964 through 1980 are taken from U.S. Securities and Exchange Commission, SEC Monthly Statistical Review, June 1979, May 1981, and February 1982. The data for 1955 through 1963 are from Institutional Investor Study Report of the Securities and Exchange Commission, Committee Print, Supplementary Volume I, Committee on Interstate and Foreign Commerce, U.S. House of Representatives, 92nd Congress, lst Session, March 10, 1971, table 3- 32, p. 146.

¹³ Stephen A. Berkowitz and Dennis E. Logue, The Portfolio Turnover Explosion Explored, Journal of Portfolio Management, Spring 1987, p. 38-45. The authors provide both unweighted and dollar- weighted turnover rates for pension funds; the estimates are very similar. The unweighted rates are shown in Figure 1 because they are reported for more years.

¹⁴ David D. McCarthy and John A. Turner, Stock Turnover in Private Pension Portfolios, in John A. Turner and Daniel J. Beller, Eds., Trends in Pensions 1992, U.S. Department of Labor, U.S. Government Printing Office, forthcoming, p. 531-562. 1987 was the last year for which Form 5500 (used by pension plans to report data to the I.R.S.) required reporting portfolio turnover rates.

¹⁵ The data are taken from Fact Book 1991 (cited in footnote 3).

¹⁶ The two methods of calculating the turnover rate would yield the same result if the average price of shares that are traded equalled the average price of shares that are listed (or are held in the portfolios of the institutions whose turnover rates are being measured).

¹⁷ It might be expected that mutual funds would have lower portfolio turnover rates than pension funds and other institutional investors because many mutual fund shareholders are subject to the capital gains tax, which reduces turnover (see Donald W. Kiefer, Lock-In Effect Within a Simple Model of Corporate Stock Trading, National Tax Journal, March 1990, p. 75-94). It is not clear, however, that mutual funds take the tax consequences for their shareholders into account in determining investment strategy

¹⁸ Roger Lowenstein and Craig Torres, Big Board Stocks May be Rising, But Exchange Seat Prices Languish, The Wall Street Journal, June 22, 1990, p. Cl-C2. Adjusted for this change, figures comparable to the 1980 NYSE turnover rate shown in Figure 1 would be 91 percent in 1987, 69 percent in 1988, 65 percent in 1989, and 58 percent in 1990.

¹⁹ New Ways to Play the Indexing Game, Institutional Investor, November 1988, p. 93-98; and James A. White, Index Funds Start Raising Their Sights, The Wall Street Journal, October 20, 1989, p. Cl, C22.

²⁰ Jayne Levin Basket Funds: the Pros and Cons of Index Funds, Investment Dealers' Digest, January 29, 1990, p. 16-19.

²¹ "Active" or "aggressive" portfolio management and a high turnover rate do not necessarily indicate high risk. Typically, the average beta for pension plan equity portfolios is less than 1; this means the portfolios have a lower risk level than the overall stock market.

²² A listing of 875 investment advisors filing forms with the Securities and Exchange Commission for activities in the four quarters ending in June, 1989, reported 119 investment advisors with a turnover rate of 100 percent or greater; of those, 22 had rates exceeding 200 percent, and 5 had rates in excess of 300 percent. See: Carolyn Kay Brancato, Institutional Investors and Corporate America: Conflicts and Resolutions, reprinted in The Impact of Institutional Investors on Corporate Governance, Takeovers, and the Capital Markets (cited in footnote 3), Exhibit 17, p. 70-85. Some of the firms with the highest turnover rates, however, may be specialists not engaged in the management of pension fund assets; see the statement of Robert Shultz in Effects of Short-Term Trading on Long-Term Investments (cited in footnote 3), p. 38.

²³ The turnover rates in figure 2 do not apply just to pension funds or institutional investors; they are aggregate turnover rates for the New York Stock Exchange (comparable to the NYSE line in Figure 1).

²⁴ The turnover rate data are from the New York Stock Exchange Fact Book, various years.

²⁵ 29 U.S.C. sec 1104(a)(1) states that a pension plan "fiduciary shall discharge his duties with respect to a plan solely in the interest of the participants and beneficiaries and (A) for the exclusive purpose of: (i) providing benefits to participants and their beneficiaries; and (ii) defraying reasonable expenses of administering the plan; . . . (C) by diversifying the investments of the plan so as to minimize the risk of large losses, unless under the circumstances it is clearly prudent not to do so. . . ."

²⁶ The record of the Fidelity Magellan Fund demonstrates clearly that being a successful long-term investor and having a high portfolio turnover rate are not incompatible. The fund had the best 10-year performance record among all mutual funds during the period from 1983 to 1991; see Jonathan Clements, Fidelity's Magellan Fund: Top Dog No More, The Wall Street Journal, July 11, 1991, p. C1, C16. During the 1980s, the fund had a portfolio turnover rate that ranged from 77 percent to 277 percent and averaged 127 percent; source; Growth Fund Research, Inc., Growth Fund Guide, various issues.

²⁷ To emphasize the importance to the long-run investor of achieving the highest rate of return each period, Harry M. Markowitz uses the following example:

"Investment for the long run'. . . is concerned with a hypothetical investor who neither consumes nor deposits new cash into his portfolio, but reinvests his portfolio each period to achieve maximum growth of wealth over the indefinitely long run. . . . In the long run, thus defined, a penny invested at 6.01% is better -- eventually becomes and stays greater -- than a million dollars invested at 6%."

Harry M. Markowitz, Investment for the Long Run: New Evidence for an Old Rule, Journal of Finance, December 1976, p. 1273.

²⁸ Some critics point to evidence that, in some cases, investment managers with high turnover rates have lower rates of return, but this usually is not the primary focus of their criticisms. Section V of the report reviews the evidence regarding the relationship between turnover and rate of return.

²⁹ See the discussion in Hillel Gray, New Directions in the Investment and Control of Pension Funds, Washington, D.C., Investor Responsibility Research Center, 1983, 120 p.

³⁰ Mark P. Kritzman, What Practitioners Need to Know: The Nobel Prize, Financial Analysts Journal, January-February 1991, p. 10-12, 15.

³¹ The fundamental value of a security is the discounted present value of the expected cash flows to be received from owning the security.

³² This discussion characterizes the strong form of the efficient markets theory, which is the most commonly employed version of the theory. A weaker version of the theory would have implications closer to those discussed in the next subsection on the costly information model. For a discussion of the different versions of the efficient markets theory, see Burton G. Malkiel, Is the Stock Market Efficient?, Science, March 10, 1989, p. 1313-1318.

³³ Stating the position very succinctly, one of the originators of the efficient market theory wrote that the conclusion that active investment management should be able to beat the market "can only be justified by assuming that the laws of arithmetic have been suspended for the convenience of those who choose to pursue careers as active managers." William F. Sharpe, The Arithmetic of Active Management, Financial Analysts Journal, January-February 1991, p. 7.

³⁴ Actually, an index fund will fall slightly behind the index it is structured to follow because of administrative expenses and transactions costs; market indexes such as the S&P 500 Index make no allowance for investment costs.

³⁵ Index funds can also be designed to track narrower indexes for particular industries or sectors within the market or foreign markets; see James A. White, The Index Boom: It's No Longer Just the S&P 500-Stock Index, The Wall Street Journal, May 29, 1991, p. C1, C8.

³⁶ [sic]

³⁷ [sic]

³⁸ See Sanford J. Grossman, On the Efficiency of Competitive Stock Markets Where Trades Have Diverse Information, Journal of Finance, May 1976, p. 573-585; and Sanford J. Grossman and Joseph E. Stiglitz, On the Impossibility of Informationally Efficient Markets, The American Economic Review, June 1980, p. 393-408.

³⁹ The effect of the activities of informed investors on the return received by uninformed investors is unclear. The more efficient allocation of investment capital may raise the overall rate of return on securities in the market so the higher profits of informed investors are not derived at the expense of uninformed investors. If the overall rate of return on securities in the market does not rise, or does not rise sufficiently, then the extra profit received by informed investors will reduce the return received by other investors.

⁴⁰ To keep the mathematics tractable, in the model developed by Grossman and Stiglitz (cited in footnote 38) there are only two securities, a risk-free security and a risky security. In this context, an index fund is not a meaningful concept. Some of the statements in this subsection are generalizations from the Grossman/Stiglitz model to apply to a more realistic setting.

⁴¹ Trading based on costly information should expand until the marginal trade produces the risk-adjusted market rate of return. The issue is whether in this equilibrium ALL information-based trades yield the market return or, alternatively, infra-marginal trades still yield higher-than-market returns. The answer depends on the nature of the cost function for the process of obtaining information and on the nature of information-based trades in equilibrium. In the Grossman/Stiglitz model, all information based trades yield the same return because there is only one type of information, it has a single price, and all trades of the risky asset occur at the same price.

⁴² For a listing and analysis of a large number of the suspected market inefficiencies, see Bruce I. Jacobs and Kenneth N. Levy, Disentangling Equity Return Regularities: New Insights and Investment Opportunities, Financial Analysts Journal, May June 1988, p. 18-43. Some of these "inefficiencies" might be explained by the costly information model discussed in subsection B.

⁴³ See the discussion in Stock Market "Short-Termism." Implications for Corporate Planning Horizons (cited in footnote 9), especially p. 9-12 and 21-31.

⁴⁴ See the discussion in J. Bradford De Long, Andrei Shleifer, Lawrence H. Summers, and Robert J. Waldmann, The Survival of Noise Traders in Financial Markets, The Journal of Business, January 1991, p. 1-19.

⁴⁵ For different conclusions regarding the implications of current knowledge, see Malkiel, Is the Stock Market Efficient? (cited in footnote 32); and Peter Fortune, Stock Market Efficiency: An Autopsy?, New England Economic Review, March/April 1991, p. 17-40.

⁴⁶ It might occur to some readers that the increase in portfolio turnover rates could be partially attributable to the rapid growth of pension funds during this period. This is not the primary explanation, however. The turnover measure used by Berkowitz and Logue (cited in footnote 13), which is based on the lesser of the value of portfolio purchases or sales, is not affected by growth of the portfolio. And McCarthy and Turner (cited in footnote 14) found that only 20 to 30 percent of turnover in pension plans is attributable to plan growth, shrinkage, or change in portfolio mix. The rest of turnover is attributable to stock selection.

⁴⁷ The classic text for this approach is Benjamin Graham, David L. Dodd, and Sidney Cottle, Security Analysis, 4th Edition, New York, McGraw-Hill, 1962.

⁴⁸ See the discussion in Statement of the American Academy of Actuaries, in Effects of Short-Term Trading on Long-Term Investments (cited in footnote 3), p. 116-117.

⁴⁹ See Gary L. Gastineau, A Short History of Program Trading, Financial Analysts Journal, September-October 1991, p. 4-7.

⁵⁰ See Financial Executives Institute Committee on Investment of Employee Benefit Assets, Survey of Pension Fund Investment Practices, in Effects of Short-Term Trading on Long-Term Investments (cited in footnote 3), p. 105-106.

⁵¹ A recent article about Jeff Vinik, manager of the Fidelity Growth & Income Fund (which has a 215 percent turnover rate), stated, "Vinik says he believes in being a long-term investor as long as a company shows growth and progress, and he won't sell a stock based on a one-time event or temporary bad news." Stan Hinden, A Whiz Kid's Stock Answers, The Washington Post, October 20, 1991, p. H1, H7-8.

⁵² See Arnold J. Hoffman, Pension Funds and the Economy, 1950- 87, in John A. Turner and Daniel J. Beller, Eds., Trends in Pensions 1989, Washington, D.C., U.S. Department of Labor, 1989, p. 95-117.

⁵³ Defined-benefit pension plans are those in which the future benefit to the employee is specified. This contrasts with defined contribution plans, in which the current contribution is specified. In defined contribution plans, the future benefit is whatever the contributions plus accrued investment return will provide. See the discussion in U.S. Library of Congress, Congressional Research Service, Explanations for the Trend Away From Defined Benefit Pension Plans, Report No. 91-647 EPW by Angela Chang, August 25, 1991, 25 p.

⁵⁴ For a recent treatment that emphasizes this approach, see Walter R. Good and Douglas A. Love, Managing Pension Assets, McGraw- Hill, New York, 1990, 216 p. A review of empirical findings consistent with this perspective is provided in Berkowitz, Logue & Associates, Inc., Study of the Investment Performance of ERISA Plans, Prepared for the Office of Pension and Welfare Benefits, U.S. Department of Labor, July 21, 1986, p. VII-4, VII-5.

⁵⁵ See the discussion in Joseph A. Komma, What's Behind Investment Performance: The Story the Figures Don't Tell, Trusts & Estates, October 1975, p. 718-720, 766.

⁵⁶ This identifies an apparent contradiction in ERISA. Fund assets are to be managed "for the exclusive purpose of providing benefits to participants and their beneficiaries," but the participants and beneficiaries do not bear the investment risk; the plan sponsor does. See the discussion of this and related issues in Daniel Fischel and John H. Langbein, ERISA's Fundamental Contradiction: The Exclusive Benefit Rule, The University of Chicago Law Review, v. 55, 1988, p. 1105-1160.

⁵⁷ See Fran Hawthorne, Raiding the Corporate Pension Fund, Institutional Investor, December 1983, p. 101-116, 113, and Jack L. VanDerhei and Scott E. Harrington, Pension Asset Reversions, in Trends in Pensions 1989 (cited in footnote 52), p. 187-210.

⁵⁸ The tax is analyzed in U.S. General Accounting Office, Pension Plan Terminations: Recapturing Tax Benefits Contained in Asset Reversions, November 1989, 22 p.

⁵⁹ Participants in defined-benefit pension plans may also benefit from a higher rate of return on pension assets because the plan sponsor may be more willing to adjust the benefit formulas to compensate for inflation. The discussion in the text is not intended to imply that defined-contribution pension funds are not aggressively invested; many are. In the case of defined-contribution pension plans, the benefit to the plan sponsor, who usually oversees the investment, is less direct. Earning a high rate of return on defined contribution pension funds, however, would be expected to result in a more satisfied work force and fewer demands for higher pension contributions in the future.

⁶⁰ See James A. White, Giant Pension Funds' Explosive Growth Concentrates Economic Assets and Power, The Wall Street Journal, June 28, 1990; and Hoffman (cited in footnote 52).

⁶¹ See Eric Schine, It's the States' Turn to Play 'Raid the Pension Fund,' Business Week, July 8, 1991, p. 30-31; and Maggie Mahar, The Great Pension Raid, Barron's, December 2, 1991, p. 8-9, 13-19.

⁶² See Donald M. Feuerstein, Toward a National System of Securities Exchanges: The Third and Fourth Markets, Financial Analysts Journal, July-August 1972, p. 57-59, 82-86.

⁶³ The information in this discussion was derived from John C. Schreiner and Keith V. Smith, The Impact of Mayday on Diversification Costs, Paper No. 710, Krannert Graduate School of Management, Purdue University, October 1979, p. 1-5; U.S. Securities and Exchange Commission, The Securities Industry in 1980, September 1981, p. 81- 84; and Seymour Smidt, The Growth of Equity Trading in World Markets, Discussion Paper, Cornell University, December 23, 1988, p. 14-16.

⁶⁴ See the discussion in The Growth of Equity Trading in World Markets (cited in footnote 63).

⁶⁵ Fact Book 1991 (cited in footnote 3).

⁶⁶ There are two additional elements of transaction costs that have probably decreased somewhat during the past 25 years, but estimates of these costs over this period are not available. The costs are the bid-ask spread and market impact costs, both of which are functions of liquidity in the market. As trading increased substantially, the average bid-ask spread and market impact costs should have decreased.

⁶⁷ Estimation of the transaction costs is explained in more detail in the Appendix.

⁶⁸ The transaction costs are expressed as a percentage of average share price, so even though the PER SHARE trading costs continued to decline during this period, the costs as a fraction of share value rose.

⁶⁹ Large institutional investors pay an average brokerage commission of about 6.5 cents per share traded, see James A. White, Big Investors' Commission Rates May Have Hit a Low, The Wall Street Journal, April 24, 1990, p. C1, C9. Some institutional trades are executed for as little as 2 to 4 cents per share; see Jeffrey M. Laderman, The Business That Brokers Would Love to Ditch, Business Week, March 27, 1989, p. 106. In calculating the figures in the text, the average share price is assumed to be $37, based on data in Fact Book 1991 (cited in footnote 3).

⁷⁰ See the discussion in Andrew H. Chen, Frank C. Jen, and Stanley Zoints, The Optimal Portfolio Revision Policy, The Journal of Business, January 1971, p. 51-61; George M. Constantinides, Capital Market Equilibrium with Transaction Costs, Journal of Political Economy, August 1986, p. 842-862, and M.H.A. Davis and A.R. Norman, Portfolio Selection With Transaction Costs, Mathematics of Operations Research, November 1990, p. 676-713.

⁷¹ A call gives the holder the right to buy the underlying security at a specified price within a specified period. A "covered" call is a call sold with regard to a stock held in the portfolio; the opposite is a "naked" call, which is a call sold on a stock which the seller does not own.

⁷² Futures and options can, of course, be used as speculative investments. There are, however, no data that permit determining what portion of futures and options activity by pension funds is speculative as opposed to hedging or arbitrage activity.

⁷³ See the discussion in Saul Hansell, Is the World Ready for Synthetic Equity?, Institutional Investor, August 1990, p. 54-61.

⁷⁴ A put gives the holder the right to sell the underlying security at a specified price within a specified time period.

⁷⁵ A variant of this strategy called a zero-cost-collar, which involves the simultaneous purchase of puts and sale of calls, apparently resulted in a lot of stock trading at the end of 1991; see Randall Smith and Barbara Donnelly, Year-End Rally Tightened "Zero- Cost-Collars," The Wall Street Journal, January 9, 1992, p. C1, C15.

⁷⁶ For survey results on use of futures and options by pension funds, see: Pension forum: Dipping Into Derivatives, Institutional Investor, December 1990, p. 173.

⁷⁷ An alternative view, that ERISA had little effect on the operation of pension plans, is provided in Richard A. Ippolito, A Study of the Regulatory Effect of the Employee Retirement Income Security Act, Journal of Law and Economics, April 1988, p. 85-125.

⁷⁸ An additional reason for plan sponsors to want to guard against a decrease in the value of pension fund assets is accounting rules adopted in 1987 which require underfunded pension liabilities to appear on corporate balance sheets, a prospect corporate financial officers strive to avoid. See Financial Accounting Standards Board, Statement of Financial Accounting Standards No. 87, Employers' Accounting for Pensions, Original Pronouncements, Accounting Standards as of June 1, 1991.

⁷⁹ Zvi Bodie, Managing Pensions and Retirement Assets: An International Perspective, Journal of Financial Services Research, December 1990, p. 187-188.

⁸⁰ Modern financial management techniques result in high turnover rates not only for equities but also for fixed income investments. In fact, fixed income portfolios frequently have higher turnover rates than equity portfolios. Some bond immunization strategies, which provide assurance of the value of a bond portfolio at a future date of a financial obligation, result in high turnover. A variety of other approaches, such as securities lending, repurchase agreements, or interest rate arbitrage are used to increase the rate of return on fixed income portfolios with minimal risk. High turnover rates for fixed income securities, however, presumably do not have significant implications for corporate planning horizons and are not currently a public policy concern. In terms of the effect on investors, it would be ironic if public policy restricted or discouraged hedging techniques with regard to equity investments but imposed no restraints on hedging in the debt market, since equity investments are generally riskier than debt.

⁸¹ See Thomas Epps, Security Price Changes and Transaction Volumes: Theory and Evidence, The American Economic Review, September 1975, p. 586-597.

⁸² See Thomas E. Copeland, A Model of Asset Trading Under the Assumption of Sequential Information Arrival, The Journal of Finance, September 1976, p. 1149-1168.

⁸³ See Jonathon M. Karpoff, The Relation Between Price Changes and Trading Volume: A Survey, Journal of Financial and Quantitative Analysis, March 1987, p. 109-126.

⁸⁴ Two other channels through which ERISA might be partially responsible for higher pension fund turnover rates were mentioned above in subsection B in the discussion on active investment management and in subsection D in the discussion on hedging.

⁸⁵ 29 U.S.C. sec. 1104(a)(1)(B).

⁸⁶ In a well-known case involving the Grumman Corporation pension plan, the Secretary of Labor sued the trustees of the plan based on their failure to tender shares in a takeover bid (the plan also purchased shares at the inflated post-bid price). The courts determined that the pension plan trustees had acted imprudently. The case may not be broadly applicable, however, since it also involved a conflict-of-interest issue and Grumman Corporation itself was the takeover target. See the discussion in the article by Fischel and Langbein (cited in footnote 56).

⁸⁷ M. Peter McPherson, Deputy Secretary of the Treasury, Relationships Between Pension Funds and Corporate Management, Remarks before the Council of Institutional Investors, April 4, 1989, 11 p.

⁸⁸ See the statement of Honorable Nicholas F. Brady, Secretary, U.S. Department of the Treasury, in Effects of Short-Term Trading on Long-Term Investments (cited in footnote 3), p. 11.

⁸⁹ Krk [sic] F. Maldonado, Fiduciary Responsibilities Under ERISA, The Labor Lawyer, Fall 1986, p. 819-837.

⁹⁰ Andrew Sigler, Chairman and Chief Executive Officer of Champion International, has stated, "There has been a very strong feeling among the trustees of a fund . . . that you should not exercise judgment on fund decisions; rather, you should pick outside managers and let them exercise judgment in order to avoid liability. This has been the recommendation of lawyers . . . ." Source: Effects of Short-Term Trading on Long-Term Investments (cited in footnote 3), p. 18-19.

⁹¹ Speculation takes different forms, some of which are beneficial and contribute to economic efficiency. As used here, speculation refers to investment based on following current trends or on "hot tips" rather than careful analysis.

⁹² The survey and results are presented in two NBER working papers: Robert J. Shiller and John Pound, Survey Evidence of Diffusion of Interest Among Institutional Investors, NBER Working Paper No. 1851, March 1986, 25 p.; and John Pound and Robert J. Shiller, Speculative Behavior of Institutional Investors, NBER Working Paper No. 1964, June 1986, 27 p.

⁹³ Josef Lakonishok, Andrei Shleifer, and Robert W. Vishny, Do Institutional Investors Destabilize Stock Prices? Evidence on Herding and Feedback Trading, NBER Working Paper No. 3846, September 1991, 30 p.

⁹⁴ For a review of this research, see Stock Market "Short- Termism:" Implications for Corporate Planning Horizons (cited in footnote 9), 31 p.

⁹⁵ Several other techniques have also been employed to "dress up" a portfolio at reporting time; see Solveig Jansson, The Fine Art of Window Dressing, Institutional Investor, December 1983, p. 139- 140.

⁹⁶ Josef Lakonishok, Andrei Shleifer, Richard Thaler, and Robert Vishny, Window Dressing by Pension Fund Managers, The American Economic Review, May 1991, p. 227-231.

⁹⁷ Brett Trueman, A Theory of Noise Trading in Security Markets, Journal of Finance, March 1988, p. 83-95.

⁹⁸ One guidebook for pension plan sponsors provides an example that is consistent with the perspective that more trading may be indicative of greater skill. The example involves evaluating the performance of two hypothetical investment managers. Both manage portfolios of 30 stocks over a 5-year period. The first manager rebalances the portfolio each year. Each year 18 of the 30 stocks in the portfolio outperform the market, and over the 5-year period the portfolio outperforms the market by an average of 3 percent per year. The second manager invests in an index fund of 30 stocks of small growth companies at the beginning of the 5-year period and maintains the investment without revision throughout the period. This portfolio also outperforms the market by an average of 3 percent per year. In discussing how to evaluate the performance of the two managers, the book states:

"Yet the favorable results in the second example hardly represent useful evidence of skill in active management. The manager did not make individual decisions about each of the 30 issues in the portfolio but, rather, one decision about a style- based index fund at the beginning of the period. Clearly, one correct decision in only one attempt in the second example is much less impressive than 90 correct decisions in 150 attempts in the first example."

Source: Managing Pension Assets (cited in footnote 54), p. 136.

⁹⁹ Hilary Rosenberg, What Price Turnover?, Institutional Investor, October 1988, p. 211-223.

¹⁰⁰ Two other sets of regressions using the individual turnover rate data series are reported in the Appendix.

¹⁰¹ RHO is the serial correlation coefficient; D.W. is the Durbin-Watson statistic. See the Appendix for details.

¹⁰² The first and last coefficients are significant at the 99 percent level of confidence; the others are significant at the 95 percent level.

¹⁰³ ERISA also apparently had no significant effect on private pension fund rates of return; see Study of the Investment Performance of ERISA Plans (cited in footnote 54).

¹⁰⁴ Speculation could also be related to the bull market variable.

¹⁰⁵ See, for example, The Growth of Equity Trading in World Markets (cited in footnote 63); and J. Harold Mulherin and Mason S. Gerety, Daily Trading Volume on the New York Stock Exchange, Mimeo, December 1989, 20 p. The claims in these papers were not based on statistical analyses but merely on the observation that transaction costs declined substantially during the period in which turnover rates rose.

¹⁰⁶ Some studies provide evidence of a positive relationship between turnover and rate of return based on calculations of rate of return on hypothetical portfolios over specified test periods; see, for example, James P. D'Mello, Karen E. Lahey, and Inayat U. Mangla, An Empirical Test of the Relative Valuation Model of Portfolio Selection, Financial Analysts Journal, March-April 1991, p. 82-86. Some of these studies take into account the effects of estimated brokerage commissions, but they usually ignore market impact costs and fund management fees. Hence, their results may not be attainable in practice. As one article noted sardonically:

". . . to maximize rate of return, you should exit the money management business and sell your advice in the form of a newsletter. That way, your performance will be judged only on paper, unencumbered by the drag imposed by real-world implementation."

Source: Andre F. Perold and Robert S. Salomon, Jr., The Right Amount of Assets Under Management, Financial Analysts Journal, May-June 1991, p. 31. Studies of actual rates of return earned in practice generally do not report a positive relationship between turnover and rate of return (net of costs). The review presented here does not include any studies based on hypothetical portfolios.

¹⁰⁷ See William F. Sharpe, Mutual Fund Performance, The Journal of Business, January 1966, p. 119-138; Michael C. Jensen, The Performance of Mutual Funds in the Period 1945-1964, Journal of Finance, May 1968, p. 389-416.

¹⁰⁸ All of the research referred to here adjusts the returns for differences in risk between the funds and the market index. For examples of this research see Stanley J. Kon and Frank C. Jen, The Investment Performance of Mutual Funds: An Empirical Investigation of Timing, Selectivity, and Market Efficiency, The Journal of Business, April 1979, p. 263-289; Hany A. Shawky, An Update on Mutual Funds: Better Grades, The Journal of Portfolio Management, Winter 1982, p. 29-34; and Mark Grinblatt and Sheridan Titman, Mutual Fund Performance: An Analysis of Quarterly Portfolio Holdings, The Journal of Business, July 1989, p. 415.

¹⁰⁹ Richard A. Ippolito, On Studies of Mutual Fund Performance: 1962-1991, working paper, September 1991. ¹¹⁰ Richard A. Ippolito. Efficiency with Costly Information: A Study of Mutual Fund Performance, 1965-1984. The Quarterly Journal of Economics, February 1989, p. 1-23.

¹¹¹ Study of the Investment Performance of ERISA Plans (cited in footnote 54).

¹¹² Richard A. Ippolito and John A. Turner, Turnover, Fees and Pension Plan Performance, Financial Analysts Journal, November- December 1987, p. 16-26

¹¹³ 60 basis points equals 0.6 percent. One paper maintains that if higher turnover is associated with a lower rate of return, this reflects the cost of a strategy of hedging against losses. The pension funds knowingly accept a somewhat lower rate of return because of the value of protection against a sharp loss. See Zvi Bodie and Leslie E. Papke, Pension Fund Finance, Mimieo, June 1990, p 34.

¹¹⁴ Gary P. Brinson, Brian D. Singer, and Gilbert L. Beebower, Determination of Portfolio Performance II: An Update, Financial Analysts Journal, May-June 1991, p. 40-48.

¹¹⁵ Statement of Ira M. Millstein, in Corporate Time Horizons (cited in footnote 3), p. 90.

¹¹⁶ The structure of performance fee systems is discussed in Jeffrey Bailey, Some Thoughts on Performance-Based Fees Financial Analysts Journal, July-August 1990, p. 31-40.

¹¹⁷ Pension forum: Performance Fees Make Headway, Institutional Investor, June 1989, p. 129-130.

¹¹⁸ Joseph H. Golec, Do Mutual Fund Managers Who Use Incentive Compensation Outperform These Who Don't?, Financial Analysts Journal, November-December 1988, p. 75-78

¹¹⁹ Dennis E. Logue, Pension Plans at Risk: A Potential Hazard of Defect Reduction and Tax Reform, NCPA Policy Report #119, National Center for Policy Analysis, October 1985, p. 15.

¹²⁰ Performance Fees Make Headway (cited in footnote 117).

¹²¹ See Managing Pension Assets (cited in footnote 54), p. 153-154; and Sidney Cottle, Pension Asset Management: The Corporate Decisions, Financial Executives Research Foundation, 1980, p. 254- 255.

¹²² Termination of investment managers does not always imply disapproval of their performance. Some switches among investment managers may result from a change in investment policy by the pension fund which requires moving funds to managers with different investments approaches.

¹²³ See Statement of the American Association of Private Pension and Welfare Plans, in Effects of Short-Term Trading on Long- Term Investments (cited in footnote 3), p. 134.

¹²⁴ Survey of Pension Fund Investment Practices (cited in footnote 50).

¹²⁵ Robert A.G. Monks, Answers to Questions, in The Impact of Institutional Investors on Corporate Governance, Takeovers, and the Capital Markers (cited in footnote 3), p. 263.

¹²⁶ As stated earlier, it is not certain whether these presumed negative effects on corporate incentives actually occur, see the discussion in, Stock Marker "Short Termism" Implications for Corporate Planning Horizons (cited in footnote), 31 p.

¹²⁷ See Joseph E. Stiglitz, Using Tax Policy to Curb Speculative Short-Term Trading, Journal of Financial Services Research, Vol 3, 1989, p. 101-115, and Lawrence H. Summers and Victoria P. Summers, When Financial Markets Work Too Well: a Cautious Case for a Securities Transactions Tax, Journal of Financial Service, Research, Vol. 3, 1989, p. 261-286. Both articles are reprinted in Franklin R. Edwards, Editor, Regulatory Reform of Stock and Futures Markets, Boston, Kluwer Academic Publishers, 1989.

¹²⁸ Harry M. Markowitz, Portfolio Selection: The Journal of Finance, March 1952, p. 77-91.

¹²⁹ William F. Sharpe, A Simplified Model for Portfolio Analysis, Management Science, No. 1, 1963, p. 277-293.

¹³⁰ Estimated of the bid-ask spread for 1963 to 1982 in Richard Roll, A Simple Implicit Measure of the the Effective-bid-Ask Spread in an Efficient Market, Journal of Finance, September 1984, p. 1127-1139.

¹³¹ Memorandum from Greenwich Associates, September 26, 1991, 3 p.

¹³² The Securities Industry in 1980 (cited in footnote 63), p. 92.

133/ Institutional Investor Study Report of the Securities and Exchange Commission, Committee Print, Volume 4, Committee on Interstate and Foreign Commerce, U.S. House of Representatives, 92nd Congress, 1st Session, March 10, 1971, p. 2192-2193.

¹³⁴ The period from the mid-1960s to the mid-1970s is also when the third and fourth markets were particularly active. There are no known estimates of trading costs in these markets. Since trading, in these markets diminished once commissions were deregulated, however, their trading costs may not have been such lower than the net costs (the commission less gives-ups) of trading through the traditional brokerages; see Frank K Reilly, Secondary Markets, in Frank J. Fabozzi and Frank G. Zarb, Eds., Handbook of Financial Markets: Securities, Options, Futures, Homewood, Illinois, Dow Jones- Irwin, 1981, p. 150-151.

¹³⁵ Smidt, The Growth of Equity Trading in World Markets (cited in footnote 63), p. 31.

¹³⁶ The data are taken from Futures Industry Association, Volume of Futures Trading: 1960 Through 1987, 8 p. The figures are total futures contracts traded for the Major Market Index, the Major Market Maxi Index, the NASDAQ 100 Index, the S&P 500 Index, the S&P 100 Index, the S&P OTC 250 Index, the Value Line Index, the Mini Value Line Index, the NYSE Composite Index, and the NYSE Financial Index. Some of these futures are no longer traded.

¹³⁷ Specifically, the variable equals the high of the S&P 500 Index for the year minus the low of the index, divided by the S&P 500 Index at the end of the prior year.

¹³⁸ See The Relation Between Price Changes and Trading Volume: A Survey (cited in footnote 83); and Thomas W. Epps, The Demand for Broker's Services: the Relation Between Security Trading Volume and Transaction Cost, The Bell Journal of Economics, Spring 1976, p. 163- 194.

¹³⁹ Of course, this relationship is not precise. A large amount of new information could develop in a year, but it could cause stock prices to vary upward and downward within a narrow price range. The presumption is that, on average, more new information results in a greater price range.

¹⁴⁰ The procedure was only partially successful, as indicated by the Durbin-Watson statistic, which is in the inconclusive range for each of the regressions.