Dividend Forecasts Explained
What an ORTEX dividend forecast is, what its probabilities mean, where the declarations come from, and how the scorecard measures accuracy.
What a forecast is
An ORTEX dividend forecast is our expectation for one future dividend of one listing: when it will be declared, its ex-date, record date and pay date, and how much it will be. We forecast every dividend-paying listing we cover and update the forecasts every day from the listing's own payment history, its reported results and the market it trades in.
Each future dividend has a stable event key: the year and its slot in the payment cycle, such as 2026-Q4 (fourth quarterly payment), 2026-H1 (interim) or 2026-FY (final). The key ties a forecast to the declaration that eventually resolves it.
Dates. Every forecast date carries a band in business days: an ex-date of 14 November with a band of 2 means we expect it between 12 and 18 November. Once the company declares, the declared dates replace the forecast ones.
Amounts. Amounts are per share in the dividend currency, adjusted for splits to today's share count. amountP50 is the central forecast. amountP10 and amountP90 bound the range we expect the declared amount to fall in four times out of five. baseAmount is the last regular payment in the same slot of the cycle, and changePct compares the central forecast with it.
Status. A forecast is forecast until it is declared, late when the expected declaration date has passed with nothing declared (daysLate counts the days), and declared once the company has declared it. A declared row shows the declared amount beside the forecast and says whether it was exact, above or below.
Reviewed. ORTEX staff review the forecasts where a miss would matter most and the ORTEX forecast is least certain. A reviewed forecast has reviewed: true, and the row shows the reviewed values: a review can change the dates, the central amount (amountP10 and amountP90 widen to include it) and the kind. The probabilities stay the ORTEX forecast's own. Every other row has reviewed: false, and so does every row of a forecast vintage, which records the forecast as the daily run made it.
What the probabilities mean
Four probabilities describe how the next payment compares with the last regular payment in the same slot:
| Field | Meaning |
|---|---|
pRaise | The payment goes up |
pHold | The payment stays the same |
pCut | The payment goes down |
pOmit | No payment at all |
The four add up to 1. pSpecial is separate: the chance of a special or extra payment around the event, which can come with any of the four. outcome names the most likely of them (special when pSpecial is 0.5 or more).
The probabilities are calibrated: across many events, a 30% figure comes true about 30% of the time. That is a claim we test in public, in the scorecard below.
Where the declarations come from
Declarations are ORTEX data. For US listings we read them from the listing venues (Nasdaq and NYSE), SEC EDGAR filings and the FINRA Daily List, usually within minutes of publication. For every other market, and for all history, they come from S&P Capital IQ.
The same dividend often arrives more than once: a company's SEC filing, then the venue's file, then S&P Capital IQ. We keep each arrival with its source and the time it could first be known (announcedAt, in UTC), so you can see which source published it first. Enterprise plans can read every arrival through the declarations feed.
The scorecard
Every forecast is frozen at fixed horizons before its expected declaration date: first (the first forecast of the event), d90, d30, d7 and d1 (90, 30, 7 and 1 calendar days before). Frozen forecasts are never revised. When the company declares, we compare each frozen forecast with what it declared:
- Ex-dates: exactly right, and within 3 business days.
- Amounts: exactly right, and within 5%.
- Probabilities: the Brier score of the raise, hold, cut and omit probabilities (0 is perfect).
- Cuts: of the events we flagged as at risk (a cut or omission probability of 25% or more), how many were cut or omitted; and of the events that were cut or omitted, how many we had flagged.
An event with nothing declared 60 days after its expected declaration date counts as an omission. Special payments are not scored. The scorecard is published by region, payment frequency and horizon.
Published and replayed forecasts
Every frozen forecast says how it was made, in its origin:
published: frozen by the daily ORTEX forecast run on the day itself, before the declaration. Published forecasts start when the ORTEX forecast service began.replayed: rebuilt afterwards from the data as it stood on that day, to extend the history back before the service began. A replayed forecast uses only the payment history, results, prices and consensus known on the day. One thing it cannot rewind: S&P Capital IQ keeps no record of its own later corrections to past dividends, so a replayed forecast sees today's version of them and can look slightly better than a published one would have been.
The scorecard counts both, so a cell's scoredSince can predate the service. The forecast history endpoint labels every row, so you can score the published forecasts on their own.
In the API
| Endpoint | What it returns |
|---|---|
GET /api/v11/stock/{namespace}/{identifier}/dividends/forecast/ | The listing's next events (num_events, 1 to 8, default 4) |
GET /api/v11/stock/{namespace}/{identifier}/dividends/forecast_history/ | Past forecasts frozen at a horizon (default d30), each beside the declared outcome and labelled with its origin |
GET /api/v11/reference/dividends/scorecard/ | The scorecard |
GET /api/v11/stock/{namespace}/{identifier}/dividends/history/ | Dividend history: past and announced dividends |
Enterprise plans also have region feeds of current forecasts (/api/v11/enterprise/bulk/dividends/forecasts/) and of declarations by announcement date (/api/v11/enterprise/bulk/dividends/declarations/), the complete forecast vintage for a past date (/api/v11/enterprise/bulk/dividends/forecast_vintage/), the scorecard, and per-stock versions of the forecast endpoints.
ORTEX publishes its forecasts country by country. Where they are not published yet, every forecast endpoint (the forecast, its history, the feeds, the vintage and the scorecard) answers an empty 200 with meta.emptyReason: notPublished, and a region feed serves the published countries only. Declarations and dividend history are always served.
Updated about 15 hours ago
