If you’ve been following Morningstar’s research for long, you know how important we think the cost of investing is. The expense ratio is the most proven predictor of future fund returns – and our data agrees. That’s also what academics, fund companies, and, of course, Jack Bogle, find when they run the data. But it’s been a couple of years since I provided the proof, so we have updated my data to show just how strong and dependable fees are as a predictor of future success. That’s not to say investors should only consider cost when selecting a fund. There are many other variables, but investors should make expense ratios their first or second screen.
How We Ran the Test
To begin any test of predictive power, we use historical data so that we are using data investors would have had access to at the time. That includes funds that no longer exist. In fact, that’s a key part of the story because higher-cost funds are much more likely to fail and be merged away. If you do not factor them in, you will see better performance from higher-cost funds than was the reality, as those that survived naturally are more likely to have produced better performance – while so many failures have been culled.
We looked at a few different measures to test how expense ratios worked: total return over the ensuing period, load-adjusted returns, standard deviation, investor returns, and subsequent Morningstar Rating.
In addition, we calculated a success ratio for all the above measures. The success ratio is our way of factoring in mutual funds that were merged away or liquidated over the ensuing time period. The other figures only include data on funds that survived the whole time period. But the success ratio asks, “What percentage of funds survived and outperformed their category group?” Only funds that managed both were included in the success ratio, as it is hard to argue that funds that no longer exist or underperformed were successful.
For our tests, we began by grouping funds into quintiles – five sets based on cost – within their peer group and then rolled that up into an asset class. That means we ordered each Morningstar Category into quintiles. Then we grouped all the cheapest-quintile funds in an asset class, then the second-cheapest-quintile funds, and so on. We also ran all of the above tests against a universe in which only one share class per fund was included.
Some readers of past studies wondered whether fees were as strong for selecting between funds rather than among share classes of the same fund. So, to eliminate comparisons of share classes of the same fund, we limited this test to the oldest share class of a fund. The first set of tables included all share classes and the second set show just one share class per fund. We looked at the five years to the end of December 2015, the four years to the end of 2015, and so on.
The Answer: Cost Really is a Good Predictor of Success
We’ve done this over many years and many fund types, and expense ratios consistently show predictive power. Using expense ratios to choose funds helped in every asset class and in every quintile from 2010 to 2015. For example, in U.S. equity funds, the cheapest quintile had a total-return success rate of 62% compared with 48% for the second-cheapest quintile, then 39% for the middle quintile, 30% for the second-priciest quintile, and 20% for the priciest quintile. So, the cheaper the quintile, the better your chances.
All told, cheapest-quintile funds were three times as likely to succeed as the priciest quintile. If you’re wondering why only one quintile had a success ratio above 50%, it’s because many funds did not survive the time period. If no funds were merged away, then the overall success rates would average something close to 50%.
As it was, about 20% of the funds were merged away, making 40% the average success ratio point. The pattern was pretty similar in other asset classes. For example, international-equity funds had a 51% success ratio for the cheapest quintile compared with 21% for priciest. Balanced funds had a 54% success rate for the cheapest quintile compared with 24% for the priciest.
.png)
The predictive power also holds up in the other areas we tested. It points investors to a better outcome for investor returns and for load-adjusted returns. That makes some sense, as both are fairly closely tied to total returns. It was a much weaker predictor of standard deviation, though that’s not a big surprise, as fees and volatility are not very closely linked. For U.S. equity funds and sector funds, standard deviation was a hair lower for lower-cost funds. There wasn’t much pattern for the other asset classes.
Funds with high costs, especially in bonds, do tend to take greater risk in order to produce a competitive yield. However, that generally means taking on more credit risk, and credit risk damps standard deviation except when it blows up. So, what if we limit our fee test to just one share class per fund? It actually shows stronger predictive power. For example, the success rate of returns in U.S. equity funds rose to 64% with just one share class versus 62% with all of them, and the priciest quintile falls to 15% versus 20% for all share classes.
This was true in most asset classes except for international equity, where the success rates became more compressed. More important than the slight improvement in results is the larger point that this clearly helps you choose among funds and that the share-class criticism of fee studies does not hold up.
