Measuring Dividend Cuts: What Splits and Payment Timing Hide
Two artefacts impersonate a dividend cut. Correcting for them changes the measured cut rate among long-record payers by more than forty percentage points.
A dividend cut sounds like it should be simple to count: the annual total fell, so the dividend was cut. Applied to real payment records that definition is badly wrong, and wrong in one direction - it finds cuts that never happened.
A stock split halves the per-share dividend while leaving the shareholder with twice as many shares and exactly the same income. In an unadjusted series that is a 50% fall. Separately, a payment whose ex-date slips from late December to early January leaves one calendar year holding three payments instead of four, which reads as a 25% fall. Neither is a decision the company made about its dividend.
The tables below quantify both, on one fixed group of companies, by varying only how a cut is counted. The third table is the result once they are corrected for - and it is the figure this site publishes on individual stock pages.
Key Observations
- Counted naively - raw per-share amounts, any payment count - 304 of these 310 companies appear to have cut their dividend, 98.1% of them. That figure is almost entirely an artefact.
- Adjusting for stock splits removes 62 of those. A split halves the per-share payment without changing what a shareholder receives, so an unadjusted series shows a 50% fall where nothing happened.
- Requiring both years to contain the same number of payments removes 70 more. A payment that slips across a year boundary leaves one year with three instead of four - a 25% drop that is purely the calendar.
- What survives both corrections is 172 companies, 55.5%. That is the figure quoted elsewhere on this site, and the gap between it and 98.1% is the entire reason the corrections are worth making.
What each correction changes, across the same 310 long-record payers
| How cuts are counted | Payers that appear to have cut | Share |
|---|---|---|
| Raw per-share amounts, any payment count | 304 | 98.1% |
| Split-adjusted amounts only | 242 | 78.1% |
| Split-adjusted, equal payment counts (what we publish) | 172 | 55.5% |
The method against companies whose record is public
| Company | Ticker | Comparable years | Years the payout rose | Years it fell over 20% |
|---|---|---|---|---|
| The Procter & Gamble Company | PG | 55 | 55 | 0 |
| Johnson & Johnson | JNJ | 55 | 54 | 0 |
| The Coca-Cola Company | KO | 54 | 54 | 0 |
| 3M Company | MMM | 54 | 52 | 1 |
| AT&T Inc. | T | 40 | 32 | 1 |
| GE Aerospace | GE | 35 | 26 | 4 |
How often a long record still contains a cut
| Record length | Payers | Of which also cut over 20% | Share |
|---|---|---|---|
| 25 or more years of increases | 310 | 173 | 56% |
| 10 to 24 years of increases | 884 | 667 | 75% |
| Fewer than 10 | 606 | 432 | 71% |
Methodology
We sum regular dividends per calendar year for every covered security, excluding one-off specials and future-dated declarations, and excluding the current year as incomplete. A year counts as a cut when the total fell more than 20% against the prior year - the same materiality threshold used across the site. Amounts are split-adjusted, and a pair of years is compared only when both contain the same number of payments. The comparison table applies each correction in turn to the same cohort, so the only thing changing is the counting method.
Caveats
A long record that contains no cut is not a prediction. These are historical base rates for the companies we cover, and coverage reflects today's universe: companies that delisted before our data begins are absent, so historical cut rates are, if anything, understated. Nothing here is a view on any security.
Sources: Dividendly dividend history (originally sourced from FMP, cross-checked against Tiingo); split-adjusted amounts as stored. Computed Oct 19, 2026.
Related Pages
Citing this research
These figures are computed from Dividendly's own stored dividend history and are free to read. If you reference them, please cite the page and the date the figures were computed — they are recalculated as new payments are recorded.
Dividendly. “Measuring Dividend Cuts: What Splits and Payment Timing Hide.”
Computed Oct 19, 2026. https://dividendly.ai/data/measuring-dividend-cuts/
We have not set reuse terms for this data. For anything beyond quotation with attribution, ask us first — some underlying inputs come from licensed providers.
Data updated Oct 19, 2026. This research page is generated from Dividendly's structured market and dividend dataset using the approach described on our methodology and data sources pages. Data is automatically validated using defined rules and may also be periodically reviewed. Research only, not investment advice. Spotted a data error? Report it.