In early 1997, Morningstar Incorporated's CD-ROM service included data on 7,782 mutual funds based in the United States. The number of such funds, already large, continues to grow at a rapid rate. Given this enormous range of choice, investors crave a simple method for separating good funds from average or bad funds. Several services purport to do so. However, Morningstar's is almost certainly the most widely used in the United States, with the majority of attention given to its "star" risk-adjusted performance measures.
Evidence of the impact of the Morningstar ratings abounds. Fund manager advertisements frequently cite them. For example:
"How did one company's mutual funds receive more four- and five-star ratings than 98% of the companies with funds rated by Morningstar? We earned them."1
"... Morningstar gave an impressive overall rating of four stars to the CREF Stock Account... The CREF Bond Market Account received an overall rating of five stars ... and the CREF Social Choice Account earned four stars..."2
The influence of these ratings is thought to be considerable. According to one report:
"As billions of dollars flow into mutual funds, those ubiquitous Morningstar stars carry a lot of weight with investors. They mean big money for fund companies that can advertise top-of-the-heap "five star" results. indeed, almost 90% of the new money that flowed into stock funds last year went to funds with four-star and five-star ratings, according to Financial Research, of Chicago."3
While "the stars" are the most prominent performance measures produced by Morningstar, they are by no means the only ones. In fact, six different measures designed to take both risk and return into account can be found in Morningstar's services.
It is clearly important that investors understand the procedures used by Morningstar to compute key performance measures, the theoretical and empirical characteristics of such measures and the relationships between them and alternative measures used in academic and professional studies. Such is the goal of this paper.
Morningstar presents its measures as summaries of historic results which may or may not provide useful predictions of future results. For example:
"Many commentators insist on treating the star rating as a predictive measure or a short-term trading signal. The rating, which is clearly labeled as a historical profile, does neither. (It would be just as ill-advised to treat an alpha or a Sharpe ratio with similarly blind faith).... the rating does not reflect Morningstar's opinion of a fund's future potential; it is simply a first-stage screen that summarizes how well each fund has historically balanced risk and return."4
While such cautionary statements are well-advised, the choice of an historic performance measure is likely to be made on the presumption that the measure in question will have some predictive value for some measure evaluated over a future period. Moreover, it is usually assumed that the performance measure that can best provide information about a given future measure is the historic value of that same measure. In practice neither of these assumptions may prove warranted.
In this paper we consider several measures of performance. For each, we attempt to answer two questions:
In what investment context is the measure most suitable for summarizing forecasted performace?
How different are the historic results obtained with the measure from those obtained with other measures?
More simply put:
For what decision is the measure best?
How different is it from the others?
From a pragmatic viewpoint it is important to address both these issues. Theory may indicate that measure A is preferable to measure B for answering a particular investment question. However, if the results obtained by ranking funds on measure B are almost the same as thouse obtained by ranking them on measure A, use of the wrong measure may not prove especially harmful.
We address only in passing the third piece of the puzzle -- the extent to which a given future performance measure can be predicted using currently available information, including previous realizations of the same or other performance measures. Other studies and our very tentative results suggest that a fund's expense and turnover are likely to better predict its future net performance than any single measure of its past performance. Nonetheless, historic performance may add predictive power when used in conjunction with information about a fund's expenses and turnover. In such instances a risk-adjusted measure of historic performance is likely to prove to be more valuable than one without such adjustment, even when the measure being predicted involves little or no adjustment for risk.
1. Smith Barney advertisement in the Wall Street Journal, May 3, 1996.
2. Academe This Week, provided by TIAA CREF on the World Wide Web, June 3, 1996.
3. Karen Damato, "Morningstar Edges Toward One-Year Ratings", The Wall Street Journal, April 5, 1996, p. C1.
4. Amy C, Arnott, in Morningstar Mutual Funds, December 6, 1996
Go to Table of Contents for this paper