We set out to write a different article from this one. The plan was to find the base rate — of the twenty highest-yielding stocks in a given year, what proportion cut the dividend within the following twelve months? It is the obvious question, it is answerable with a screener and a decade of history, and it would settle a great many arguments.
It is not published anywhere. We went through Hartford, Ned Davis Research, Wellington, Vanguard, S&P Dow Jones Indices, WisdomTree, Morningstar and Verdad, and nobody puts that number in print. Which is itself worth knowing, because a great deal of writing about high yields is delivered with a confidence the underlying evidence does not support — including, we suspect, some of ours.
So this is what the evidence does show, what it does not, and the three questions we now use instead.
The one number that is not in dispute
Fifty-three years say the cut is the thing that costs you
Annualised total return by dividend policy, S&P 500 constituents, 31 January 1973 to 31 December 2025.
Over fifty-three years, the companies that cut produced a negative annual return and the wildest ride available. Not a lower return — a negative one. Everything else in this article is a footnote to that bar.
Ned Davis Research and Hartford Funds, The Power of Dividends: Past, Present and Future, March 2026. σ is annualised standard deviation.
Show the data
| Dividend policy | Annualised return | Beta | Std. deviation |
|---|---|---|---|
| Growers and initiators | 10.22% | 0.89 | 15.97% |
| All dividend payers | 9.20% | 0.94 | 16.71% |
| Equal-weighted S&P 500 | 7.74% | 1.00 | 17.55% |
| No change in policy | 6.87% | 1.02 | 18.45% |
| Non-payers | 4.21% | 1.18 | 21.91% |
| Cutters and eliminators | −0.96% | 1.22 | 24.80% |
Fifty-three years, every S&P 500 constituent, sorted by what the company did with its dividend. Growers and initiators compounded at 10.22% a year. Companies that cut or eliminated their dividend returned minus 0.96% a year, with the highest volatility of any group in the study.
Read that again, because it is easy to skim past. Not a lower return than the market — a negative return, across five decades, from a group of large, listed, well-covered American companies. Whatever else is arguable about dividend investing, the cost of being holding the wrong one when it cuts is not.
And note where the cut sits in the sequence. By the time a board announces a reduction, the operational problem is usually two years old and the share price has usually already fallen — which is precisely why the yield looked so attractive in the first place. The high yield and the eventual cut are not two events. They are the same event, observed at different times.
What about just buying the biggest yields?
The highest-yielding stocks do not win. The second-highest do — most of the time
Annualised return by decade for US stocks sorted into dividend-yield quintiles, %.
Be careful with this one. The popular version of the finding is that the top quintile always loses to the second — the actual data says it lost in six decades out of ten and won in four, including the current one. The reliable part is the reason: the top quintile carried a 72% average payout ratio against 49% for the second.
Wellington Management and Hartford Funds, whitepaper WP106, data to 31 December 2025.
Show the data
| Period | S&P 500 | Highest yield | Second quintile |
|---|---|---|---|
| 1930s | -1.30% | -1.50% | -1.80% |
| 1940s | 9.10% | 13.90% | 13.00% |
| 1950s | 19.40% | 18.40% | 19.70% |
| 1960s | 7.90% | 8.60% | 9.00% |
| 1970s | 5.70% | 9.50% | 10.10% |
| 1980s | 17.60% | 20.50% | 19.10% |
| 1990s | 18.20% | 12.50% | 15.60% |
| 2000s | -0.80% | 5.40% | 4.40% |
| 2010s | 13.50% | 12.90% | 13.40% |
| 2020–25 | 15.17% | 13.40% | 10.37% |
Here we have to be careful, because the popular version of this finding is stronger than the data. You will read that the highest-yielding quintile always underperforms the second-highest. On Wellington’s own numbers it underperformed in six decades out of ten — and beat the second quintile in the 1940s, the 1980s, the 2000s and the 2020s so far.
So “the top quintile loses” is a tendency, not a law, and anyone who tells you otherwise has read the summary rather than the table. What is much more solid is the reason the tendency exists: over the study period the highest-yield quintile carried an average payout ratio of 72% against 49% for the second. The top quintile is not cursed. It is simply, on average, paying out far more of what it earns — and that is a mechanical, checkable property of any individual stock you are looking at right now.
Which turns a statistical argument into a practical one. You do not have to avoid high yields. You have to avoid the ones with no cover, and you can tell which is which before you buy.
Ten cuts, and what the yield told you first
Ten cuts, and what the yield was telling you first
Each dot is a documented dividend cut since 2022. Horizontal: the yield shortly before the announcement. Vertical: how much of the dividend went.
Intel and 3M were yielding about 6% — nothing that would trip an alarm — and still cut by more than half. The yield is a symptom. It is not the diagnosis.
Company announcements and contemporaneous reporting, 2022–2026. Pre-cut yields are as reported at the time and are approximate where a source gave a range.
Show the data
| Company | Cut announced | Yield before | Old → new | Cut |
|---|---|---|---|---|
| 3M | May 2024 | over 6% | $1.51 → $0.70 quarterly | −53.6% |
| Intel | Feb 2023 | approached 6% | $0.365 → $0.125 quarterly | −65.8% |
| Walgreens Boots | Jan 2024 | over 7% | $0.4775 → $0.25 quarterly | −47.6% |
| Wendy's | Aug 2026 | about 7.6% | $0.14 → $0.07 quarterly | −50.0% |
| V.F. Corporation | Feb 2023 | about 8.5% | $0.51 → $0.30 quarterly | −41.2% |
| AT&T | Feb 2022 | about 9.1% | $2.08 → $1.11 a year | −46.6% |
| Hanesbrands | Feb 2023 | about 10% | $0.15 → nil | −100% |
| Leggett & Platt | Apr 2024 | over 10% | $0.46 → $0.05 quarterly | −89.1% |
| Dow Inc. | Jul 2025 | over 10% | $0.70 → $0.35 quarterly | −50.0% |
| Medical Properties | Aug 2023 | over 13% | $0.29 → $0.15 quarterly | −48.3% |
Every dot there is a real announcement with a date. A few things jump out that the folklore gets wrong.
The really dangerous yields were not the biggest ones. Intel yielded about 6% before its 2023 cut and then eliminated the dividend entirely in 2024. 3M was “over 6%” before halving its payout in May 2024. Six per cent trips no alarms in anyone’s screen. Meanwhile Medical Properties Trust was above 13% — visibly, obviously distressed — and cut by 48%, less than 3M did.
The cause was almost always the balance sheet, not the business. V.F. Corporation cut because gross debt hit 4.5x EBITDA against a 2.5x target. Hanesbrands eliminated its dividend at about 4.3x. Leggett & Platt’s chief executive said it plainly: reducing the dividend to free up capital to accelerate deleveraging. That ended a 53-year record of increases and cost the company its Dividend King status in a single sentence.
And it is still happening. LyondellBasell cut 49.6% in February 2026, and Wendy’s halved its dividend on 7 August 2026 at a yield of about 7.6%, with US comparable sales down 7% and full-year guidance withdrawn. This is not a story about 2020.
The S&P Dow Jones Indices data on US common stocks says the same thing more quietly: 176 dividend decreases in 2025 against 132 in 2024, a third more, while the dollar value of increases fell from $71.4 billion to $59.3 billion. Both years were far better than 2023’s 386 decreases. But the direction in 2025 was the wrong one.
Today’s big yields, and the distinction that matters most
Eight of today's big yields cost more than the company earns
Trailing dividend payout ratio, % of earnings. Anything above the dashed line is being paid out of something other than this year's profit.
Four of these are supposed to be above 100%. Yancoal, New Hope and Fortescue pay a stated share of cash flow, and Eni runs a policy-linked distribution — a high earnings-payout ratio is the design, not a warning. Pfizer, M&G and Aviva are the ones paying a fixed, progressive dividend out of earnings they did not make this year.
Trailing payout ratios via stockanalysis.com, June–August 2026; individual as-of dates vary by security. Payout ratios on an earnings basis overstate strain at companies with heavy non-cash charges.
Show the data
| Ticker | Company | Market | Yield | Payout ratio | Dividend type |
|---|---|---|---|---|---|
| PFE | Pfizer | US | 6.15% | 226% | Fixed quarterly |
| ITH | Ithaca Energy | UK | 9.67% | 206% | Policy-linked |
| NHC | New Hope | Australia | 3.72% | 186% | Variable |
| YAL | Yancoal | Australia | 4.22% | 175% | Variable |
| MNG | M&G | UK | 5.93% | 160% | Fixed progressive |
| ARCC | Ares Capital | US | 9.66% | 143% | Board-set quarterly |
| ENI | Eni | Italy | 4.74% | 135% | Policy-linked |
| AV. | Aviva | UK | 5.52% | 119% | Fixed progressive |
| ET | Energy Transfer | US | 6.39% | 92% | Growing distribution |
| MO | Altria | US | 6.18% | 89% | Fixed progressive |
| FMG | Fortescue | Australia | 6.09% | 88% | Variable |
Eight of those eleven pay out more than they earn. Before you write all eight off, split them into two groups, because they are not the same thing at all.
Group one: policy-linked payers, where a high ratio is the design
Yancoal, New Hope and Fortescue pay a stated share of cash flow or profit, and Eni runs a policy-linked distribution. When a company has told you in writing that it distributes a fixed proportion of what it generates, a payout ratio above 100% of accounting earnings tells you about depreciation and impairments, not about the dividend.
The written policies are quite specific once you go and read them. BHP guarantees a minimum 50% of underlying attributable profit every reporting period, with anything above that at the board’s discretion. Rio Tinto targets 40–60% of underlying earnings through the cycle. Woodside commits to a minimum 50% payout ratio while stating that the board retains full discretion over whether a dividend is payable at all. Petrobras distributes 45% of free cash flow, conditional on gross debt staying under a stated ceiling.
None of that is a trap. It is a different instrument. The mistake is not owning them — it is annualising one good quarter and calling the result a yield. Our analysis of the shipping dividend cycle walks through what that looks like in practice, where a 23% headline yield contained roughly 4% of genuinely repeatable cash. Star Bulk’s policy states it plainly: the board may distribute up to 100% of cash flow after debt amortisation, maintenance capex and a $2.1 million-per-vessel cash buffer, with a minimum of five cents a quarter and every payment subject to quarterly board approval. Read that sentence and you know exactly what you own.
Group two: fixed, progressive payers above 100%
Pfizer at 226%, M&G at 160% and Aviva at 119% are a different proposition. These companies pay a fixed, progressive dividend — the kind you are supposed to be able to plan around — out of earnings that did not cover it. There may be perfectly good reasons: non-cash charges, a patent-cliff year, an insurance accounting quirk. That is exactly the work to do before buying, rather than after.
The rule we use is simple. A high payout ratio on a variable policy is information. A high payout ratio on a fixed promise is a question that needs answering. The mechanics of how the ratio is calculated, and where it flatters or slanders a company, are in our guide to dividend terminology.
The closest thing to a base rate that exists
Since nobody publishes the number we wanted, here is the best proxy we found. Vanguard studied a universe of roughly 700 high-yielding US stocks — the top 60% of yielders among the 2,000 largest Russell 3000 names — across April 2012 to April 2023, sorted by a quality score.
The highest-quality quintile produced 89 dividend cuts over that period. The lowest-quality quintile produced 220 — roughly two and a half times as many. Vanguard’s own summary of the universe as a whole is that companies in it “rarely lower dividends”.
That is the honest shape of it. High yield alone is not the risk factor people think. High yield plus weak fundamentals is, and the spread between the best and worst quality within the same high-yield universe is wider than the spread between high-yield and the market. Separately, Simply Safe Dividends reports 938 cuts tracked since 2015, of which 97% were rated below 60 on its safety scale before the announcement. These things are, mostly, visible in advance.
The three questions we actually ask
1. Is the dividend a promise or a formula? Find the policy in the company’s own words — investor-relations pages carry it, usually under “dividend policy”, in one paragraph. A formula (“50% of underlying profit”) means the payment will fall in a bad year and that this is not a cut. A promise (a fixed, progressive dividend) means a fall is a cut, and boards defend those far past the point of prudence, which is why the eventual reduction is so large when it comes.
2. What is the leverage, not the payout ratio? Look at V.F. at 4.5x, Hanesbrands at 4.3x, Leggett & Platt deleveraging. In every one of those cases the debt covenant, not the income statement, decided the dividend. Net debt to EBITDA above about 4x in a business with any cyclicality is a better predictor than any yield screen we know of.
3. Does free cash flow cover the cash cost of the dividend? Not earnings — cash. Multiply the dividend per share by the share count and compare it with free cash flow. It takes two minutes with a filing and it is the single most informative thing you can do. Our screen for safe high-yield dividend stocks runs through the full method, and our work on return of capital and NAV erosion covers the fund version of the same problem, where the distribution can be partly your own money coming back.
When a big yield is completely fine
We are not arguing for avoiding yield. Altria at 6.18% on an 89% payout, Energy Transfer at 6.39% on 92% with a growing distribution, Enterprise Products at 5.74% on 76% — these are mature businesses distributing most of what they make, which is what mature businesses are supposed to do. The 2020s have been kind to the top yield quintile precisely because a lot of those companies were priced for a decline that did not arrive.
The same logic applies one level up, to whole sectors. The utilities in our look at who pays for the AI grid mostly have comfortable payout ratios and are still delivering 2% dividend growth, because the cash is going into the ground; the defence primes in our US-versus-Europe comparison have low payout ratios and low yields for the opposite reason. Neither shows up in a yield screen.
What we are arguing is that the yield number itself carries almost no information. It is a ratio of a payment that already happened to a price that already fell. Everything useful — cover, leverage, policy, cyclicality — sits behind it, and takes an hour to check. Given that cutters returned less than nothing over fifty-three years, an hour is cheap.
If you want the constructive version of this argument rather than the defensive one, the dividend income strategy we use starts from cover rather than yield, and our 2026 global payout outlook sets out where the growth is actually coming from this year.
