UK companies with a dividend cut in the past five years are 1.7x as likely to cut again within a year

Corrected 9 October 2026. The first version of this note said UK companies with a prior cut were 2.5x as likely to cut again (28.9% against 11.5%). That overstated it. Its prior-cut flag could see part of the future: a cut spotted six months before the observation date was measured over a year that ran six months past it, into the period the note was predicting. It also summed raw dividend amounts, which read share splits and consolidations as cuts. Corrected, the UK figure is 1.7x (22.6% against 13.7%). Every number below is the corrected one. How we found it.

Among UK-listed shares observed every six months from 2015 to 2024, those that had cut their dividend at some point in the previous five years cut again within the following 12 months 22.6% of the time. UK companies with a clean five-year record cut 13.7% of the time. The gap is 9.0 percentage points, a 1.7x lift.

In the US the effect is much stronger: 23.9% against 6.6%, a 3.6x lift.

The prior-cut group cuts more often in every yield quartile. Within the top yield quartile on the LSE, companies with a prior cut cut again 31.6% of the time, against 17.8% for clean records (n=342 against n=225, a gap of 13.8 points).

The data

Cut within 12 months, by five-year prior-cut record. Specials excluded, pandemic excluded.

Allowing for chance, we are 90% confident the UK gap lies between 6.3 and 11.6 points (US: 16.3 to 18.3; all markets: 15.7 to 17.4). Those intervals treat each observation as independent. The same company appears in many periods, so the true uncertainty is somewhat wider.

Within yield quartile. The gap survives controlling for yield.

The prior-cut flag is not just a yield proxy. Within every yield quartile, in both markets, the prior-cut group cuts more often.

Robustness: specials threshold

All markets, with both the prior-cut flag and the outcome recomputed at each setting:

The gap moves by 0.2 points across every setting. The finding does not depend on the specials threshold.

Method

• Cut definition. The same TTM rule as the yield note: is_cut = d_fwd < d_base * (1 - 0.02), summed over each company's full payment record, dated by payment date, in split-adjusted amounts, with one-off specials excluded by the two-condition detector (2.5x the trailing 24-month median AND no comparable payment about a year earlier). The forward outcome is the dataset's own cut label (cutWithin.m12, frozen v4); the script recomputes it and matches the frozen label on all 19,356 rows.

• Prior-cut flag, with no lookahead. An observation at t0 is flagged when the company had a TTM cut in any 12-month window that ended within the five years before t0. Windows are checked every six months, starting 12 to 60 months before t0, so every one of them closes on or before t0 and none overlaps the period being predicted. The flag reads the company's full payment record, including the years before it entered the dataset.

• The dataset's precomputed flag (dividendCutInLast5Years) agrees with this one on 86.4% of rows. It is not used in any reported number.

• Quartiling is on the as-reported trailing yield, within each market and observation date, with at least 20 companies a date. The same convention as the yield note.

• Truncation. Observations after 2024-07-15 are dropped, because their forward window runs past the end of the data.

What would have falsified it. A gap that vanished within any single yield quartile would have meant the prior-cut flag was only a yield proxy. Every quartile keeps a positive gap, though on the LSE it is small in Q1 (4.8 points). A gap that collapsed when specials detection moved would have made the result an artefact of the heuristic. It moves by 0.2 points.

Limitations

1. Acquisitions are conflated with cuts. An acquired company stops paying, which reads as a cut. This inflates all rates. It should affect both groups similarly.
2. A prior cut may not mean the same thing in every sector. For REITs, structural distribution changes can produce a false-positive prior-cut flag. The sector split in the first version of this note has not been re-run on the corrected basis.
3. Split-adjusted amounts absorb some genuine resets. A dividend rebased after a rights issue or demerger is a cut in the plain sense, but the adjusted record may smooth it, on both sides of the comparison.
4. Specials detection is heuristic, not sourced from a corporate-actions feed. The sensitivity table is the mitigation.
5. Survivorship is only partly corrected, bounded by delisted price coverage.
6. $2bn market cap floor. Nothing here applies to small caps.
7. The five-year lookback is a choice. A three-year or seven-year lookback would shift the levels.

What this does not show

It does not show that a prior cut causes the next cut, and it does not support a claim about any individual share. It is a base rate across a decade. It says nothing about total return: a company that cuts may still outperform.

The all-markets gap of 16.6 points is the headline effect of the flag on its own. Within yield quartiles the gaps run from 4.8 to 17.0 points, so a pitch line that controls for yield should quote the within-quartile figure for its market.

Reproduce it

cd dividendmapper && npx tsx scripts/research/ttm-cut-by-prior-cut-label.ts

It prints the attribution below, the corrected tables and the threshold table, and fails if its recomputed outcome disagrees with the frozen label on any row.

Correction

The first version of this note (13 August 2026) reported 28.9% against 11.5% on the LSE, a 2.5x lift. Its analysis script was never committed, so on 9 October 2026 we rebuilt its method from the note's own description. That rebuild gives 28.6% against 11.4%, which is close enough to confirm the reading.

• Lookahead. The old flag counted a cut at any earlier observation in the past five years. An observation six months before t0 measures a cut over the following 12 months, a window that ends six months after t0. Part of the cut it detects falls in the period the note was trying to predict, which inflates the lift. Removing that alone, on the old data, cuts the UK gap from 17.2 to 6.8 points.
• Raw amounts and incomplete histories, the same defect found in the yield note. Moving to the full, split-adjusted payment record, with no lookahead, gives the corrected 9.0 points.

A prior cut still predicts another one. In the UK the gap is about half the 17.4 points first reported, and in the US it remains large.