Where a Decade of Dividend Returns Actually Came From
Take every dividend payer we cover, sort them by yield, and look at what they returned over the past ten years. The picture is brutal. On a 10,000 EUR stake, companies yielding under 2% today produced a median price gain of 22,355. Those yielding over 8% produced a median price loss of 4,375.
That finding is wrong, and the reason it is wrong is worth more than the finding.
What the wrong version looks like
Here it is in full, because it is worth seeing before it is taken apart. Every figure is a median, on a 10,000 EUR stake held for ten years with dividends reinvested, with companies grouped by the yield they carry today.
| Yield today | Companies | Median price gain | Median dividend income | Median total return |
|---|---|---|---|---|
| under 2% | 309 | 22,355 | 4,570 | 286.6% |
| 2 to 4% | 279 | 7,030 | 5,792 | 125.6% |
| 4 to 6% | 91 | 77 | 5,695 | 63.4% |
| 6 to 8% | 27 | -1,601 | 5,905 | 38.9% |
| over 8% | 17 | -4,375 | 5,853 | 23.5% |
It is a striking table. Total return falls from 286.6% to 23.5% as the yield rises. The price gain collapses from a gain of 22,355 to a loss of 4,375. It looks like a decisive verdict on chasing yield.
One column gives it away. Dividend income is almost identical in every row, from 4,570 to 5,905, with no trend at all. If a high yield genuinely meant more income, that column should rise steeply. It does not, and a table where the supposed cause has no relationship to its most direct effect is a table measuring something other than what it claims.
The problem with sorting on today's yield
A dividend yield is a dividend divided by a price. There are two ways for it to be high. The company can pay a lot, or the price can have fallen.
Sorting today's payers by yield and then examining their history therefore does something circular. The high-yield group is populated, in part, by companies that got there by losing value. You then discover that they lost value. The measurement contains its own conclusion.
Western Union shows the mechanism cleanly. It yields 13.09% today. Ten years ago it yielded 3.05%. Nobody who bought it a decade ago was buying a high-yield share. A 10,000 EUR stake in it is now worth less than it started: a price loss of 6,576 against dividends of 2,755, for a total return of -38.21%. It carries a score of 69.
Western Union Co (WU.US)Financial Services · Credit ServicesData as of 08 Sept 2026high confidence
Western Union belongs in the over-8% bucket today. It never belonged there when the decision was being made. That is true of most of that bucket.
Sorting on the yield you could actually have bought
The honest question is different: if you had bought a share ten years ago at a known yield, what happened next?
That is answerable, because we hold the closing price on the start date and the dividend per share for the last fiscal year completed before it. Dividing one by the other gives roughly the yield an investor could have seen on the day, using only information available then. It is not the exact figure a screen would have printed, since a trailing annual dividend lags a forward one, but it is built from nothing that had not yet happened, which is the property that matters.
Doing it that way, for the 711 companies where both figures resolve, produces a different picture.
| Yield 10 years ago | Companies | Median price gain | Median dividend income | Median total return |
|---|---|---|---|---|
| under 2% | 327 | 12,707 | 3,452 | 167.2% |
| 2 to 4% | 293 | 8,193 | 6,434 | 149.3% |
| 4 to 6% | 62 | 3,942 | 7,819 | 124.0% |
| 6 to 8% | 13 | 6,187 | 11,933 | 180.5% |
| over 8% | 14 | -2,585 | 10,042 | 96.6% |
All figures are on a 10,000 EUR stake with dividends reinvested, gross of tax and dealing costs.
The monotonic collapse is gone. Total returns run 167%, 149%, 124% and then 180%, which is not a pattern so much as noise around a similar answer. The band that returned most over the decade is the 6 to 8% band.
And the dividend column now behaves. It rises from 3,452 to 11,933 as the starting yield rises, which is what buying a higher yield is supposed to do and what the first table flatly denied. Two tables built from the same underlying returns disagree completely, and the only difference between them is which yield was used to sort the rows.
What actually survives the correction
Two things, and they point in opposite directions.
A higher starting yield did deliver more income. Median dividend income on the same 10,000 EUR rose from 3,452 in the lowest band to 7,819 at 4 to 6% and 11,933 at 6 to 8%. This is the least surprising result imaginable, and it is worth stating because the sorted-by-today version appears to deny it. There, dividend income was flat across every band, which was itself an artefact.
Philip Morris is the ordinary version of this. It could have been bought on a 4.1% yield ten years ago. Since then the stake gained 8,137 in price and 11,568 in dividends, so more than half the return arrived as cash rather than as price. It carries a score of 73.
Philip Morris International Inc (PM.US)Consumer Defensive · TobaccoData as of 08 Sept 2026high confidence
The top band genuinely did underperform. Companies yielding more than 8% a decade ago returned a median 96.6% against 124% to 180% for every band below them, and 8 of the 14 lost capital outright. That is the yield trap, and it is real. It is also a much narrower claim than "high yield hurts returns". It is a claim about the extreme, on 14 companies. What separates a sustainable high yield from a doomed one is a question about quality rather than realised return, and we measured that separately in what a high yield costs you in safety.
Even there the answer is not simple. PennyMac Mortgage Investment Trust genuinely yielded 12.49% ten years ago, not as an artefact of a fallen price. Its stake lost 3,595 of capital and collected 13,250 of dividends, finishing up 96.56% overall. The income did what it promised while the capital eroded underneath it. We rate it risky today, at a score of 27, which is a statement about the next decade rather than the last one.
Why this matters more than the numbers
Almost every comparison of high-yield against low-yield investing you will read is built the wrong way round, including the one we nearly published. Sorting on a current yield and looking backwards is the single easiest mistake to make in this subject, because the data to do it is free and the data to do it properly is not.
The correction does not make high yield attractive. It makes the honest case narrower and more useful. The extreme is dangerous and the middle is unremarkable. A yield of 4 to 6% bought a decade ago produced roughly the same total return as a yield under 2%, in a very different shape. One arrived as price, the other as cash.
Which of those you want is a question about your circumstances rather than about which is better. Our piece on dividend growth against high yield works through that choice with a modelled example, and this article is its measured counterpart.
What this does not show
Three limits, and the first is the largest.
The high-yield bands are small. Thirteen and fourteen companies are not enough to be confident about a median, and the spread inside them is enormous: the middle half of the 6 to 8% band runs from 98% to 782%. Read those two rows as "no clear signal" rather than as a measurement.
Survivorship. Only companies still listed and still in our coverage have a ten-year record here. A company that yielded 9%, halved, and was taken private or delisted is absent, and those absences are concentrated at exactly the high-yield end. The top band's 96.6% is therefore flattering.
It is one decade. A period in which low-yield growth shares did extremely well is not a law of nature, and the under-2% band's 167% median owes a good deal to when the window happens to start and end.
None of that changes the central point, because the central point is about measurement rather than about yield. If a comparison sorts companies by what they yield today and then reports what they did before today, it is describing its own sorting rule. Check that first, in anyone's numbers, including ours.
Our screener sorts on yield alongside the Dividend Health Score, and the score is the part that is trying to say something about the decade ahead rather than the one behind.
Yield against growth is one of five decisions behind a dividend income, and not the one that moves the outcome most. The full ordering is in building a dividend income.
Frequently asked questions
- Do high-yield shares deliver worse total returns?
- Not on this evidence, once the yield is measured at the point you could have bought it. Median ten-year total returns by starting yield band were 167%, 149%, 124% and 180%, with only the band above 8% trailing at 97%.
- Why do high yielders look so bad in most comparisons?
- Because they are usually sorted by today's yield. A share whose price has halved has roughly double the yield, so sorting on it selects companies whose price fell and then reports that their price fell.
- Does a higher starting yield give you more dividend income?
- Yes, and that part holds up. On a 10,000 stake held for ten years, median dividend income rose from 3,452 in the under-2% band to 11,933 in the 6 to 8% band. What it does not do is reduce the total return.
- Is Western Union a high-yield stock?
- It yields 13.09% today and yielded 3.05% ten years ago. Nobody bought it on a 13% yield. The yield is high because the share price fell by 65.76% over the decade, which is a different thing entirely.
- What are the limits of this measurement?
- Three. The two highest yield bands hold only 13 and 14 companies. Only companies still listed and still covered appear, so the worst outcomes are missing. And it is one particular decade.
DividendAtlas provides data and research for informational purposes only. Nothing here is investment advice or a recommendation to buy or sell any security. Always do your own research.