What Is a Dividend Health Score, and How Is It Calculated?
"Is this dividend safe?" is the question every income investor wants answered, and it is a hard one. The honest answer depends on payout coverage, the balance sheet, the record, and how they interact, which is a lot to assess for every stock you own. A dividend score exists to compress all of that into a number you can read in a second.
This guide explains how the DividendAtlas Dividend Health Score is built, why it carries a confidence level, and what it does with businesses whose accounts work differently. We are describing our own methodology, so this is the explanation straight from the source.
Why a single ratio is not enough
The instinct is to reach for one number, usually the payout ratio. But no single ratio captures whether a dividend will hold.
A low payout ratio looks reassuring until you notice the company is carrying too much debt. A clean balance sheet means little if earnings are collapsing. A fifty-year record of increases is real evidence right up until a business is disrupted. Each factor is necessary and none is sufficient, which is why a score weighs several of them together and returns one comparable read.
The five pillars
The Dividend Health Score is built on five pillars, each answering a different question about the payout.
- Payout coverage. Can the company afford the dividend? We set the dividend against reported earnings and against free cash flow, each over complete financial years, and against analysts' earnings forecasts for the year ahead. Cash coverage matters because dividends are paid in cash, which is the distinction in free cash flow vs earnings payout ratio.
- Balance sheet. Can the business carry the dividend through a bad year? We look at debt against operating earnings, at how easily interest is covered, and at whether leverage is rising or falling. For banks and insurers this pillar becomes capital strength, because their balance sheets work differently.
- Earnings stability. How steady are revenue and earnings over time? A business whose profits swing widely has less room to keep a dividend steady through a weak patch.
- Track record. Has the company done this before? We read the length of the payment record, its growth, and any past cuts or suspensions.
- Forward outlook. What do analysts expect next? Consensus earnings growth, and whether forecasts are being revised up or down, add the one forward-looking input a score built on filed accounts otherwise lacks.
A company can score well on one pillar and poorly on another. The combined score reflects that balance rather than letting a single flattering number tell the whole story.
Why the score carries a confidence level
Here is the part most scores hide: a score is only as trustworthy as the data behind it.
A large company with a long history and complete financials can be scored with high confidence. A company with gaps in its reported data can still be scored, but the score rests on less evidence. The usual gaps are a short filing history, a missing cash flow figure, and companies that no analyst forecasts. Reporting a bare number in both cases would imply the same certainty for a firm read and a provisional one.
That is why every Dividend Health Score comes with a confidence level, high, medium or low, reflecting how much of the evidence the score could read. Below a minimum, we publish no score at all, and why some stocks have no Dividend Health Score explains what produces a blank.
The scale and the buckets
The score runs from 1 to 99, and each score falls into one of five buckets: very_safe, safe, borderline, risky and very_risky. The number lets you rank and compare; the bucket gives you the instant read.
Two of the best-known consumer-staples companies show the range. Procter & Gamble, the maker of Pampers and Gillette, yields 2.84%. Its Dividend Health Score is 81, in the very_safe bucket:
Procter & Gamble Company (PG.US)Consumer Defensive · Household & Personal ProductsData as of 13 Oct 2026high confidence
Kraft Heinz yields 7.13% and has paid the same 1.60 USD a share in each of the last five years. Its Dividend Health Score is 39, in the risky bucket:
Kraft Heinz Co (KHC.US)Consumer Defensive · Packaged FoodsData as of 13 Oct 2026high confidence
Both are judged on the same Defensive profile, so the distance between them is not a matter of how they are measured. The higher yield comes with a payout the score reads as much less secure. The score is not predicting that Kraft Heinz will cut. It is flagging that the payout carries more risk than the headline yield suggests, which is the signal a score exists to surface.
Different businesses, different tests
A sound score does not apply the same test to every company, because businesses are built differently. Each company is judged on a profile, and the profiles use different measures or different expectations:
- Standard, Defensive and Cyclical companies are scored on the same measures, judged against different expectations. A cyclical business is expected to swing more than a defensive one, and is not penalised for doing what its industry does.
- Real estate investment trusts are judged on funds from operations rather than earnings, because depreciation crushes their reported profit without consuming cash. For property companies whose reported profit includes the yearly revaluation of their buildings, the earnings payout ratio is not measured at all.
- Banks and insurers are judged on capital strength rather than ordinary debt ratios, with life and non-life insurers measured against their own peers.
- Business development companies, the US lending vehicles, are judged on net investment income, because their reported earnings include unrealised changes in the value of their loans.
Applying one formula to all of them would misjudge whole sectors. The profile is shown on every stock's score, so you can see which test was applied.
What happens when the method changes?
The method is versioned, and every weight and threshold is fixed within a version. When we change any of them, we recompute every company's score history on the new method as well. A score chart therefore never mixes two methods, and a move in the line is a change in the company, not in how we measure it.
We change the method when the evidence says a test is wrong for a class of business, and the profiles above are mostly the record of those changes.
Does it work?
A score is only worth reading if it predicts something. We tested ours on 32,167 point-in-time readings between 2021 and 2025, each computed only from data available at the time, and counted how often a cut followed within eighteen months.
| Rating at the time | Cut within 18 months |
|---|---|
very_risky | 20.0% |
risky | 16.5% |
borderline | 13.0% |
safe | 7.4% |
very_safe | 1.2% |
The cut rate falls at every step. The full method, the misses included, is in does a dividend health score actually predict cuts.
What a score cannot see
A score reads the past and the present: reported financials, the payment record, the current balance sheet, and today's analyst forecasts. It cannot see the future, and some things that end dividends arrive without warning in the numbers. A regulatory shock, a lost lawsuit, an accounting fraud, or a board choosing to redirect cash into an acquisition can all cut a payout that scored well the day before.
This is why the score is a risk signal, not a verdict. The backtest above makes the point: even the top bucket cut 1.2% of the time. A high score lowers the odds of a cut. It does not lower them to zero. Use it as a fast first read that tells you where to relax and where to look harder, not as a substitute for paying attention.
How to use it
- Screen quickly. Sort a universe by score to separate well-covered payers from poorly covered ones before you dig deeper.
- Read the yield in context. A high yield with a high score is the rare good kind. A high yield with a low score is the classic yield trap.
- Monitor over time. A score that drifts down is flagging the same deterioration described in our seven warning signs of a dividend cut, ideally before the cut rather than after.
You can see the full Dividend Health Score breakdown for any stock, or screen European and US payers by score in the screener. The score does the heavy reading for you, so your own judgement starts from a well-informed place.
What the score automates can also be done by hand on one company. Our five-step check on whether a dividend will hold is that method written out.
Frequently asked questions
- What is a dividend health score?
- A single figure that estimates how likely a company is to keep paying its dividend. It combines several factors, such as payout coverage, the balance sheet and the dividend record, into one number, so you can judge a payout at a glance rather than reading the financial statements yourself.
- How is the Dividend Health Score calculated?
- It scores a company on five pillars, payout coverage, balance sheet, earnings stability, track record and forward outlook, and combines them into a score from 1 to 99. Each company is judged on the profile that fits its business, and every score carries a bucket and a confidence level.
- What do the Dividend Health Score buckets mean?
- The score runs from 1 to 99 and maps to one of five buckets, `very_safe`, `safe`, `borderline`, `risky` and `very_risky`. The number lets you compare and rank; the bucket gives you the plain-language read.
- Why does a dividend score need a confidence level?
- Because a score is only as good as the data behind it. The confidence level, high, medium or low, tells you how much of the evidence the score could actually read. Below a minimum, we publish no score at all rather than a number we cannot stand behind.
- Does a high score mean the dividend will not be cut?
- No. In our backtest, companies rated `very_safe` still cut within eighteen months 1.2% of the time, against 20.0% for those rated `very_risky`. A high score makes a cut much less likely. It does not rule one out.
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.