Nearly one in three of the highest-yielding UK shares cuts its dividend within a year

Among UK-listed shares from 2015 to 2025, those in the highest-yielding quarter of the market cut their dividend within the following 12 months 30.3% of the time. The lowest-yielding quarter cut 13.0% of the time. The gradient rises monotonically across all four quartiles: 13.0%, 17.0%, 21.8%, 30.3%.

Put plainly: roughly one in three of the highest-yielding UK large caps cut within a year, against roughly one in eight of the lowest-yielding.

The US pattern is similar at the top (23.4% for the highest-yield quartile) but the UK cuts more often at every level.

All headline figures exclude special dividends and exclude the observation dates whose forward windows span the 2020 pandemic suspensions.

The data

Cut within 12 months, by yield quartile. Specials excluded, pandemic excluded.

Pooled across both markets (n=19,312): 11.4%, 9.4%, 13.4%, 24.5%.

The US series is not monotonic. Q1 (11.4%) sits above Q2 (8.9%). The lowest yielders include companies that have only just initiated a dividend, where an early reduction is more common. The UK series is monotonic across all four.

Robustness: the specials threshold

Special-dividend detection is a heuristic (see Method), so the finding is reported across five settings. The gradient holds at every one, and the top quartile is stable:

UK Q4 moves within 30.2% to 34.3% across every setting. US Q4 within 23.4% to 25.9%. The result does not depend on the threshold chosen.

Method

• Cut definition. Rolling TTM: a cut is recorded where the forward 12-month dividend total falls more than 2% below the trailing 12-month total.

  is_cut = d_fwd < d_base * (1 - 0.02)
  

  Summing a full year each side removes payment cadence from the question. UK issuers typically pay a small interim and a larger final, so a payment-to-payment comparison fires on nearly every UK name every year for structural reasons. A genuine reduction (25p in April, 10p in November) still registers, because the annual totals differ.

• Both windows are rebuilt from raw payment records in growthInputs.dividends, not from the pre-aggregated dps fields, because specials cannot be stripped from a pre-aggregated sum.

  Each observation's array holds history only up to that observation date, so no single row can supply a forward window. Histories are therefore built as the union of a ticker's arrays across all its semiannual observations, deduplicated on ex-date. Only 32 of 127,288 (ticker, ex-date) pairs disagree on amount across observations, so the union is safe.

• Special-dividend detection. The dataset carries no is_special flag. The naive rule (payment > N x median) is wrong for the UK, where a final of roughly twice the interim is the normal pattern. A payment is flagged special only when both hold:

  1. it exceeds 2.5x the median payment of the preceding 24 months (requiring at least 3 prior payments), and
  2. it has no comparable counterpart about a year earlier: no payment within 45 days of (date minus 365) worth at least 60% of it.

  Condition 2 is what protects UK finals. A recurring large final has a counterpart in the prior year; a one-off special does not. 1,514 payments are flagged, about 1.2% of all payments.

• Quartiling is on the as-reported trailing yield, assigned within each observation date and then pooled. Pooling first would let market-wide yield drift decide which period lands in which quartile. As-reported yield is used deliberately: it is what an investor sees on a screener, and the only route to an ex-specials yield runs through dps.trail12m, whose unreliability produced 45% average yields when attempted.

• Truncation. 2,599 observations after 2024-07-15 are dropped: their forward window runs past the end of the data, where an absent total is indistinguishable from a suspension and would score as a 100% cut.

Limitations

1. Acquisitions are conflated with cuts. The manifest states: "dps targets use candidate payment streams incl. post-delisting cessation; acquisitions are conflated with cuts by this measure." An acquired company stops paying, which reads as a cut. This inflates all rates by an unknown amount. It should affect quartiles similarly, so the relationship is more robust than the levels.
2. Special detection is heuristic, not sourced from a corporate-actions feed. The sensitivity table above is the mitigation, not a substitute.
3. Survivorship is only partly corrected, bounded by delisted EOD coverage.
4. $2bn market cap floor. Nothing here applies to small caps.
5. A 2% tolerance is tight for progressive policies with small annual variation.
6. No significance testing. The sample is large and the UK gradient is monotonic across four buckets, but no confidence intervals are computed.
7. UK sample is 2,245 observations, an order of magnitude smaller than the US.

What this does not show

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

Reproduce it

node scripts/research/ttm-cut-by-yield-quartile.js --exclude 2019-07-15,2020-01-15,2020-07-15

Add --sensitivity for the threshold table, --special-mult N to change the threshold, --no-specials-filter to disable it.