Home › How it works

How we predict dividend announcements

What a dividend announcement is

Before a company can pay a dividend, its board of directors has to meet and formally declare one — the announcement. Only after that can an ex-dividend date (or "ex-date") exist: the date by which you must already own the shares to be entitled to the payment. Buy on the ex-date or later and the dividend goes to the previous owner; on the ex-date the share price typically drops by roughly the dividend amount, because new buyers are no longer entitled to it. The ex-date is what matters for timing a purchase — it's why a dividend-capture strategy cares about knowing it as early as possible, and why long-term investors track it to plan when a purchase will and won't earn the next payout.

Why predicting the announcement is useful

Most dividend calendars only show a date once it exists — once a company has already announced it, or gone ex-dividend. That's accurate but late: nothing is owed to shareholders, and no ex-date can even be set, until the board has actually declared. We go earlier and forecast the announcement itself — the first link in the chain, weeks or months before the board meeting that produces it. Most dividend payers are remarkably consistent about when they declare, not just how much, which is what makes that estimate possible.

The method: three independent readings that have to agree

For each stock we compute three separate estimates of its next declaration date, from its own history:

• Cadence — the last declaration, plus the ticker's own median gap between past declarations.
• Seasonal — the declaration from around a year earlier, carried forward exactly 52 weeks (364 days, not 365 — that preserves the weekday, since boards meet on weekdays).
• Board's calendar slot — the anchor declaration's position in the month (say, "the third Thursday of February"), projected onto the same month of the target year.

We publish a date only when independently-computed readings agree with each other within a few days — all three for High quality, or the two gap-based ones for MediumMed when the calendar slot disagrees or has no candidate. Agreement between them turns out to be a stronger signal than simply asking whether a ticker has been predictable in the past — it asks whether this particular quarter is callable, not just whether the ticker generally behaves. That is why the quality badges count how many independent routes agreed, rather than reporting a single model's own confidence score — a number a model can be sure of and wrong about at the same time.

Quality ratings, and knowing when not to guess

When all three readings agree, the estimate carries the High quality badge. When the calendar-slot reading disagrees or produces no candidate at all, but cadence and seasonal still agree with each other, it carries MediumMed quality instead — a real but less precise estimate. When even cadence and seasonal disagree, we publish nothing at all. Irregular declaration timing is the commonest reason a ticker carries no announcement estimate, but it isn't the only one — a stock also needs a long enough quarterly history, and its own recent estimates need to have actually held up, before we'll publish. We would rather show no estimate than a misleading one.

How accurate this is

Every figure below is a backtest: for each past declaration, the estimate is reconstructed using only data available before it, then checked against what the board actually did (2023-onwards holdout). Both quality badges are known a median of about 91 days ahead of the actual announcement:

We name the right week 84% of the time at High quality and 78% at MediumMed quality — which is why every estimate leads with a week rather than a single day. The full breakdown is on the accuracy page, declaration by declaration.

Limitations

These are statistical estimates from past declaration behaviour, not confirmed dates. In particular:

• A company's board can cut, suspend, raise, or reschedule a dividend at any time — a change no history-based method can anticipate before it happens.
• Coverage depends on a long, consistent quarterly declaration history — roughly three years' worth — so most stocks we cover don't get an announcement estimate at all. Their pages show the company's own declared date once it exists, or the payment record alone.
• This is not, and must not be treated as, financial or investment advice. Always confirm against the company's official announcement before acting.

See the current estimates on the predicted announcement dates page, or browse all covered stocks.