Gallery:Step 8|Step 8: Riskese

The Slippery Slope of Fund Expenses

Gallery:Step 8|Step 8: Riskese

Investors not infrequently founder on the shoals of incorrect assumptions. Among the most prominent are the existence of money manager skill and a simple faith that past asset class returns are precisely predictive of future returns.

More subtle, but no less corrosive, is neglect of fund expenses. You say that the long term return your stock funds is about 11%? Dream on. 11% is the market return, and your odds of ever seeing it are not good. The market return is what you will get investing in a "frictionless" environment without fund expenses, commissions, spreads, and impact costs. As a first approximation, then,

Your Return = Market Return - Expenses.

Of course, a superior manager may earn a return in excess of the market return and offset her high expenses. There's only one problem-in the long run, there seem to be no superior managers. Even in the short term, superior performance seems to persist only weakly, and not nearly enough to negate expenses.

I'm fond of testable hypotheses, and the above formula presents a good one. If one plots fund performance against expense, one ought to see a correlation.

Below is a graph of return versus expense for the 5-year period ending 3/99 for all diversified domestic stock funds:

 If you look closely, you'll see that the fund cloud has a tendency to slope down and to the right-the higher the expense, the lower the return. The wonders of modern microprocessors and software allow us to rapidly calculate a "regression slope" (the straight line in the graph) which tells us how many dollars of return we lose for each dollar of expense. And here's the ghastly surprise; the slope is -2.22. In other words, every added dollar of expense actually deducts more than 2 dollars of return. This is not a fluke. The 95% confidence limits of this analysis are between -2.54 and -1.91, making it extremely unlikely that this result is a statistical aberration.

I performed the same analysis for the same period for each of the 9 Morningstar style boxes, and came up with the following results:

Fund Style Box

Return/Expense Slope

Large Blend


Large Growth


Large Value


Midcap Blend


Midcap Growth


Midcap Value


Small Blend


Small Growth


Small Value


Note that the slope is >1.0 in 8 of 9 cases, and >2.0 in 4 of 9 cases. It is statistically different from 1.0 in 5 of the 9 cases with 95% confidence, in spite of the relatively small number of funds in many of the categories.

So things are even worse than they seem. For every dollar of added expense, we lose two dollars of return. How can this be? Jack Bogle, in "Common Sense on Mutual Funds," notices this same phenomenon. He found that the return/expense slope for large blend funds was -1.80. He noted:

Our intuition might tell us that each point of cost should cost exactly one point of return, but something much more onerous is taking place. Although the causative factors are not exactly clear, one explanation seems to hold some extra merit: High-cost funds tend to have high turnover, and portfolio transactions carry a substantial cost of their own.

My thoughts exactly. But unfortunately, this explanation does not carry the day. If one adds in turnover as a second variable in the regression, the return/turnover slope is negative, but with only 0.47% of return lost for each 100% of turnover. This is not nearly as impressive as one would expect. Further, superimposing turnover adds only a minimal amount of statistical power to the analysis, with an adjusted R-squared of 0.131, versus 0.129 for expense alone. Finally, doing two-factor expense/turnover regressions for the 9 Morningstar boxes shows no particular effect of turnover with small caps, where one would expect to see it most clearly.

One possibility is that turnover is indeed important, but that the specific measure used-a "spot" figure in the Morningstar database, is too unstable. Turnover from year to year is a good deal less steady than expense ratio, and it is quite possible that analysis using turnover averaged over several years would show a more impressive relationship.

Alternate explanations are possible. One is moral turpitude. A fund organization which sees nothing particularly wrong charging its shareholders 200 basis points for a large cap fund is also likely quite comfortable with a wide range of other questionable activities. Such as front-running, a less than arms-length relationship with the organization's investment banking and bond-trading division, or perhaps simply a lax eye in general towards quality of execution. Readers even more evil-minded than this author will surely think of others.

Another possibility, suggested by Steve Dunn, is that the high expense funds, knowing that they are behind the eight ball, undertake high-risk strategies in a futile attempt to bridge the gap.

If in the overall sample turnover does not incur significant additional expenses, one then has to ask why. In fact, the negative effect of turnover is most clearly seen with the 3 value style boxes, and in 2 of the 3 blend boxes. All 3 growth boxes have a positive slope on turnover, indicating that turnover improves return. We'll talk about the precise reasons why in the September EF, and in the process solve the small growth active management anomaly-i.e., why small growth indexing is a persistently losing strategy.

The other interesting piece of data to fall out of the analysis is the increasing importance of expense over time. When 1-year returns are examined, the R-squared for fund expenses is only 0.005. In other words, only 0.5% of 1-year returns can be explained by costs. However, at 5 years this rises to12.9%, and at 15 years fully 36% of fund return is explained by expenses.

What is happening here is that over time the variation of manager performance, which is random, "washes out," leaving expense as the single most important factor determining return.

Jack Bogle, Chairman Emeritus at Vanguard, finds it difficult to get through an essay or speech without repeating "costs matter." They most certainly do.

copyright (c) 1999, William J. Bernstein
Reprinted by permission. All rights reserved.

Opinions expressed in this article are those of the author and do not necessarily reflect the views of or persons affiliated with