Key takeaways
- Morningstar's own global study of 43,698 funds found 5-star funds beat 1-star funds by approximately 0.25 percentage points a year, across horizons from one month to five years.
- Vanguard asked the same question against an index instead of a peer group: 39% of 5-star funds beat their style benchmark over the next 36 months, against 46% of 1-star funds.
- In Vanguard's data the average 36-month excess return fell steadily as the rating rose, from -0.04% for 1-star funds to -1.32% for 5-star funds.
- Funds landing just above a star threshold drew incremental net flows worth about 2.5% of assets over six months, while their returns moved between -0.01 and 0.01 percentage points.
- Sorting the same fund universe by cost discriminated far harder: 62% of the cheapest quintile of US equity funds succeeded over 2011 to 2015, against 20% of the priciest.
The short answer: a small edge over other funds, and none against the index
You're looking at two funds in the same category. One carries five stars, the other three. You want to know whether that badge tells you anything about the next few years.
The honest answer has two halves and they point different ways. Measured against other funds in the same Morningstar category, the highly rated ones have gone on to do slightly better. Measured against a market index, they haven't. Morningstar's own global study of its rating found 5-star funds beat 1-star funds by roughly 0.25 percentage points a year. Vanguard's study of the same rating found 39% of 5-star funds beat their style benchmark over the following 36 months, against 46% of 1-star funds.
Neither firm is wrong. They're scoring the rating against different things, and the distance between those two scoreboards is the subject of this piece.
What the Morningstar star rating actually measures
The rating is a grade on the past, and Morningstar has never described it as anything else. Its April 2023 methodology document sets the distribution explicitly. Funds are sorted inside their Morningstar category by risk-adjusted return, and "those funds with a rank that meets but does not exceed 10% receive a 5-star rating". The next band, out to 32.5%, takes four stars. Out to 67.5% takes three, out to 90% takes two, and the rest take one.
Three things follow from that sentence, and they matter more than most readers assume.
First, the ranking is against a peer group, not against an index. A category in which every fund trailed the market still produces its 10% of 5-star funds. Somebody has to be top decile.
Second, the rating needs history. "Investments must have at least 36 continuous months of total returns in order to receive a rating." A fund with a 30-month record carries no stars at all, whatever it did in those 30 months.
Third, the overall rating people actually see is a blend. For a fund with 120 months or more of returns, it's 50% the 10-year rating, 30% the five-year and 20% the three-year. Morningstar spells out the consequence itself: because the most recent three years sit inside all three windows, "the most recent three-year period actually has the greatest impact".
The methodology document closes modestly. The rating "gives investors the ability to quickly and easily identify funds that are worthy of further research". That is a claim about where to start looking, not about what happens next.
Morningstar's own study put the 5-star edge at about a quarter of a point a year
In 2016 Morningstar published a global test of its own rating, covering January 2003 to December 2015. The sample is large. Monthly fund counts ran from 7,669 to 44,071, and by December 2015 it spanned 43,698 funds. It keeps funds that were later merged or liquidated, which matters, because a sample that drops them flatters whatever is left.
Two methods produced two answers of very different strength. The regression test, which controls for fees and for market, size, value and momentum risk, found that among equity funds moving from 1 star to 5 stars was "correlated with 0.09-percentage-point higher returns per month on average (1.03 percentage points annualized)".
The event study found much less. That test simply sorts every fund by its rating each month and tracks what happened afterwards, with no adjustment for risk or cost. Across holding periods and market cycles, "5-star funds outperformed 1-star funds by approximately 0.25 percentage point annualized on average". Morningstar's own description of the pattern is "consistent (though weak)".
On a £10,000 holding, a quarter of a percentage point is £25 in the first year. That is the size of the prize the badge is pointing at, before you ask whether you could have identified the 5-star fund in advance.
Vanguard asked the same question against an index and got the opposite sign
In June 2010 Vanguard published Mutual fund ratings and future performance, by Christopher Philips and Francis Kinniry. They took US domestic equity funds carrying a Morningstar rating between 30 June 1992 and 31 August 2009, live and dead, and asked a different question. Over the 36 months after a rating, did the fund beat its style benchmark?
The answer ran backwards. On average 39% of 5-star funds beat their style benchmark over the next 36 months. For 1-star funds the figure was 46%.
The average excess return said the same thing with more force. Ranked from five stars down to one, the average excess return over the 36 months following a rating ran -1.32%, -1.26%, -1.00%, -0.67% and -0.04%. The chart above plots that series. Every rating bucket lost to its benchmark on average, and the best-rated bucket lost by the most.
Vanguard's summary is blunt: "higher-rated funds are no more likely to outperform a given benchmark than lower-rated funds".
Both results can be true, because they are scored against different things
Here's the reconciliation, and it isn't a fudge.
Morningstar's scoreboard is the category. Its question is whether a 5-star fund beats a 1-star fund sitting in the same peer group. Vanguard's scoreboard is a Russell style index. Its question is whether a 5-star fund beats the market it invests in.
Those two come apart whenever a category as a whole trails its index, which is the ordinary case rather than the exception. The SPIVA scorecard exists to measure exactly that gap. If large-blend funds collectively lose to their benchmark after costs, the top decile of that category can be genuinely better than the bottom decile and still lose to the index. Being best of a losing group is a real distinction. It just isn't the distinction most people think they're buying when they sort a fund list by stars.
The obvious objection is that the two studies cover different eras. Morningstar starts its sample in January 2003 deliberately, because "In 2002, Morningstar enhanced the star rating with new peer groups and a new measure of risk-adjusted return". Vanguard describes the old system in its own footnote: before 30 June 2002, "Morningstar rated funds' risk-adjusted excess returns versus broad benchmarks such as the U.S. stock market, the U.S. bond market, or the international stock market", a method that "does not account for additional broad risk factors such as growth/value or large/small". Most of Vanguard's window sits under that system.
That objection is a fair one, and Vanguard tested it. Comparing results either side of June 2002, they report that they "find no significant differences". Vanguard is also not a neutral referee here. It sells the index funds that a relative rating scores as average, and its opening pages are about exactly that irritation: Morningstar's target-date research ranked Vanguard's Target Retirement Funds first out of 20 competing products while the star rating gave each of them 3 out of 5 stars. Both caveats stand. What survives them is the scoreboard point, which doesn't depend on the era at all.
The gap by holding period, and where it stops widening
Morningstar's event study reports equity results at six horizons. Average cumulative return, 1-star cohort first, then 5-star:
- 1 month: 0.66% against 0.78%
- 12 months: 8.78% against 9.72%
- 36 months: 25.78% against 27.74%
- 60 months: 33.86% against 35.17%
The gap widens out to three years and then stops. At 60 months the 4-star cohort returned 35.43%, ahead of the 5-star cohort's 35.17%. The ordering isn't clean at the top over five years, and it isn't clean in the middle either. At 12 months the 2-star cohort returned 8.82% against the 1-star cohort's 8.78%, a difference of four basis points that is indistinguishable from noise.
Morningstar reads its own short-horizon result as momentum. "Like traditional momentum, the star rating is a backward-looking performance indicator (albeit risk-adjusted), and as such, it makes sense that the relevance of this information should decay as time goes on." For the longer horizons it offers a quite different mechanism. The persistence, it writes, "is likely caused by the star rating's high correlation to fees".
That second sentence is the one worth sitting with. If the five-year edge is mostly a fee effect wearing a badge, the rating is a noisy proxy for a number printed on the same page of the factsheet.
What a top rating predicts reliably is money, not returns
Jonathan Reuter and Eric Zitzewitz turned the rating's own arithmetic into an experiment. Because stars are assigned on percentile cutoffs, two funds whose returns differ trivially can land on opposite sides of a boundary. One gets the extra star for reasons that have nothing to do with skill. Their NBER working paper uses monthly Morningstar data covering "virtually every mutual fund in operation between December 1996 and August 2009".
Funds just above a threshold took in incremental net flows over the next six months "equal to approximately 2.5 percent of assets under management". The pull was strongest at the top of the scale: 0.946 percentage points at the boundary between 4 and 5 stars, against 0.337 at the boundary between 1 and 2 stars. It also lasted. Incremental flows ran 0.59 percentage points in the first month, 1.75 through month six, 2.34 through month twelve and 2.88 through month twenty-four.
The returns of those same funds barely moved. Reuter and Zitzewitz report that the estimated coefficients on the threshold in the return regression are "economically small, ranging between -0.01 and 0.01 percentage points, and are not statistically significantly different from zero at conventional levels".
So an extra star moved real money and changed nothing measurable about the fund. That is close to a clean measurement of what a rating does, and it is the mechanism behind performance chasing rather than an argument against it.
One thing the rating does predict, on Morningstar's own account, is survival. "5-star funds were far likelier to survive the full event-study horizon, especially the 60-month horizon, than lower-rated funds." A fund that gets merged away costs its holders a decision, and in a taxable account it can cost them a realised gain they didn't choose to take.
Cost sorted funds 62 to 20 where the star rating moved a quarter of a point
Russel Kinnel's May 2016 study for Morningstar sorted funds into quintiles by expense ratio as at the end of 2010, then measured the next five years. A fund counted as a success only if it both survived and beat its category group, so a fund that was merged away counts as a failure rather than vanishing from the sample.
In US equity, the cheapest quintile had a success ratio of 62%. Then 48%, 39%, 30% and 20% for the priciest. Kinnel's phrase for it is that cheapest-quintile funds were "3 times as likely to succeed as the priciest quintile". The pattern repeated everywhere he looked: 51% against 21% in international equity, 59% against 17% in taxable bonds.
Set that next to 0.25 percentage points. One signal separates outcomes 62 to 20. The other separates them by a quarter of a point a year. Both findings come from the same firm, in the same year.
The appendix carries a quieter detail. The average subsequent three-year star rating, over 2011 to 2013, ran 3.43 for the cheapest US equity quintile and 2.51 for the priciest. Fees predicted stars. That direction of causation is easier to establish than the one most fund lists imply, and it is why fee drag keeps turning up as the variable that moves.
The FCA looked at fund ratings in the UK and split the same way
The UK regulator ran this question in its Asset Management Market Study, published as MS15/2.3 in June 2017. Its finding on rating providers has two halves, and they mirror the split above almost exactly.
Ratings and best-buy lists "can on average help investors identify funds that outperform funds not on best buy lists". And then: "our analysis shows that over our assessment period ratings and best buy lists did not on average identify funds that outperformed their Morningstar category benchmarks".
Better than the funds nobody recommended. Not better than the benchmark.
The same study found active funds sold in the UK beat their benchmarks before charges and underperformed "by around 60 basis points (bps)" after them. That 60 basis points is the arithmetic any rating has to overcome before it can point at a winner. It also explains why the FCA's rules on financial promotions, restated in COBS 4.5A.10R as amended on 23 October 2025, require five years of complete 12-month performance periods and "a prominent warning that the figures refer to the past and that past performance is not a reliable indicator of future results".
What this evidence cannot tell you
Start with vintage. Morningstar's global study covers 2003 to 2015 and was published in 2016. Vanguard's covers 1992 to 2009 and was published in 2010. Reuter and Zitzewitz use data ending in August 2009. None of them reaches the last decade, and Morningstar has revised the methodology since, removing the load adjustment from US and European star ratings on 31 October 2016. A study of the 2016 to 2026 window might find something different, and nobody in this piece has run one.
Then sponsorship. Two of these studies grade Morningstar's own product. The third was written by a firm that sells the index funds a relative rating tends to score as average. The FCA is the only disinterested party in the list, and its analysis is now nine years old.
Then the arithmetic of averages. Every figure here describes a cohort. Morningstar's 0.25 percentage points is the middle of a distribution wide enough to hold excellent and terrible funds at every rating level. It says nothing about the odds on one fund.
Geography cuts both ways. Vanguard's sample is US domestic equity and the Reuter and Zitzewitz sample is US. Morningstar's study is genuinely global, with European domiciles at 43.7% of the universe by the end of the period and US funds at 18%, which is unusual for research of this kind and worth crediting.
And none of it addresses the fund in front of you. A star rating summarises a category-relative record. It doesn't know whether the manager who produced that record is still there, whether the strategy has drifted, or whether the fund has grown past the size at which its process worked.
What would change the conclusion
If categories stopped trailing their benchmarks after costs, the two scoreboards would converge and the disagreement in this piece would dissolve. Best of the peer group would then mean beating the index. That is an arithmetic condition, not a forecast, and cost is the thing standing in its way.
If the fee spread inside a category collapsed, Morningstar's own explanation for the longer-horizon gap would go with it. The firm attributes the persistence to the rating's correlation with fees. Compress the fees and you compress the signal.
If someone published a rating study covering 2016 to 2026, all of this would need rechecking. Whether the relationship survived that stretch is simply not something these papers can answer, and neither firm has published the test.
The number to watch isn't the star count. It's what sits beside it on the same page: the ongoing charge, the size of the fund, and how long the current manager has been running it. Those are the variables the research above found actually moved outcomes. Unlike the rating, none of them is a summary of the past dressed up as a signal about the future.