Learn
Japanese dividend data pitfalls
eventExample values are taken from API responses as of October 23, 2026. Financial and dividend data are refreshed daily, keyed to 16:30 (TDnet) and 19:00 (EDINET), Japan time.
What you will learn
- check_circleDividends come as split-adjusted
amountand unadjustedrawAmount. Useamountfor long-term comparisons. - check_circle
forecastOrActualseparates forecasts from actuals. Adding annual rows to per-period rows double counts. - check_circleDividend growth and no-cut streaks are counted from split-adjusted annual dividends of completed fiscal years. A year with no dividend resets them.
Why dividend data is easy to get wrong
The per-share dividend changes meaning with every stock split. After a 1-for-5 split, the same total payout is one fifth per share.
Dividend rows also mix company forecasts with confirmed actuals, and per-period rows (interim, year-end) with annual totals. Aggregating without separating them distorts increase and cut detection.
Split-adjusted vs unadjusted
In dividends, amount is split-adjusted and rawAmount is unadjusted (per share as announced). When no adjusted value exists, amount holds the unadjusted value.
Use amount for trends and growth rates. Use rawAmount to check what was actually paid per share at the time.
Split history is available from splits. Toyota split each share into five with an ex-date of September 29, 2021 (effective October 1, 2021).
Forecast vs actual, annual vs per-period rows
Dividend rows vary along two axes: forecast or actual, and which period the dividend covers.
| Field | Meaning |
|---|---|
forecastOrActual | forecast = company forecast, actual = actual |
dividendType | annual = annual total row. Other values are per-period rows such as interim and year-end |
periodEndDate | Record date (period end) |
exDividendDate | Ex-dividend date |
announcementDate | Announcement time |
Watch for double counting
The same period end can have both an annual row (dividendType of annual) and per-period rows. ZENITH Code's annual dividend uses the annual row when it is above zero; otherwise it sums per-period actuals and forecasts. It never adds the annual row to the per-period rows.
Detecting increases and cuts, with sources
dividend-changes returns increases and cuts detected from timely disclosures such as earnings summaries. Each row has the annual dividend before and after, the detection date and the source document ID (sourceDocId).
Omit the symbol to search the whole market over a date range. Use changeFlag=decrease for cuts only.
curl -sG "https://api.zenithcode.net/stable/dividend-changes" \ --data-urlencode "from=2026-09-01" \ --data-urlencode "to=2026-09-30" \ --data-urlencode "changeFlag=decrease" \ -H "x-api-key: YOUR_API_KEY"
As in every other endpoint, changeRate and the yields here are 0–1 decimals (0.246 = 24.6%). Multiply by 100 to show them as percentages.
How ZENITH Code counts dividend streaks
dividend-streaks returns the number of consecutive years of increases and of no cuts. The rules are as follows.
- Streaks are counted from the series of annual dividends (per share, split-adjusted) for fiscal years whose period end has passed. Years still in progress are only forecasts and are excluded.
- Increase streak: consecutive years in which the dividend was strictly higher than the year before.
- No-cut streak: consecutive years of an increase or an unchanged dividend. A flat year breaks the increase streak but not the no-cut streak.
- Years with a zero dividend stay in the series, so suspending the dividend resets the count. If the latest year is zero, the no-cut streak is 0.
- A missing calendar year is marked
gap, and counting stops there. - The result is checked against the progressive-dividend years stated in earnings summaries (
reportedNonDecreaseYears), andmatchesReportedshows whether they agree.
Example: Toyota Motor (7203)
Through the fiscal year ended March 2026, Toyota has a 6-year increase streak and a 13-year no-cut streak. The 13 years match the figure in its earnings summary (matchesReported is true).
The 13 years carry capped: true. The count stopped at the oldest year in the dividend data (2013), so the real streak may be longer.
For the fiscal year ending March 2027, an annual dividend forecast of ¥100 (forecast) was announced on May 8, 2026.
| Fiscal year | Annual dividend (¥) | Change (¥) | Change | Method (`basis`) |
|---|---|---|---|---|
| FY ended Mar 2026 | 95 | +5 | +5.56% | annual-row |
| FY ended Mar 2025 | 90 | +15 | +20.00% | annual-row |
| FY ended Mar 2024 | 75 | +15 | +25.00% | annual-row |
| FY ended Mar 2023 | 60 | +8 | +15.38% | annual-row |
| FY ended Mar 2022 | 52 | +4 | +8.33% | sum-of-periods |
Split-adjusted dividends per share. annual-row uses the annual row as is; sum-of-periods adds interim and year-end. Source: /stable/dividend-streaks?symbol=7203 as of October 23, 2026.
curl -sG "https://api.zenithcode.net/stable/dividend-streaks" \ --data-urlencode "symbol=7203" \ -H "x-api-key: YOUR_API_KEY"
Ex-dividend dates
From the ex-dividend date on, buying the stock no longer entitles you to that period's dividend. It is normally one business day before the record date (period end).
dividends-calendar returns market-wide ex-dividend dates for a range of up to 93 days, covering both past and upcoming dates.
Caveats
- Commemorative and special dividends: responses currently have no field separating them from ordinary dividends. A year with a one-off dividend can make the next year look like a cut.
- ETFs and REITs: the streak rules are designed for ordinary shares with fiscal-year dividends. The source data counts ETFs and REITs by calendar year, so
matchesReportedmay disagree. - Comparing across splits: lining up
rawAmountmakes a split year look like a cut. Useamountfor trends. - History length: how far back data goes varies by stock. For example, Mitsubishi Corporation (8058) has 41 dividend rows from the fiscal year ended March 2013 and a 10-year increase streak.
FAQ
For investors and developers
- Why do pre-split dividends look small?
amountis split-adjusted to today's share basis, so increases and cuts compare correctly. The per-share amount at the time is inrawAmount.- What is the difference between the increase streak and the no-cut streak?
- The increase streak counts years of raising the dividend every year. The no-cut streak counts years with an increase or a hold. A flat year breaks the first but not the second.
- Can I count streaks excluding commemorative dividends?
- Responses currently have no field separating commemorative or special dividends. ZENITH Code's streaks are counted from annual dividends that include them.
For AI builders and researchers
- What should an AI watch for when totaling dividends?
- Tell it not to add annual rows (
dividendTypeofannual) to per-period rows. If you only need annual dividends,years[].annualDividendin dividend-streaks is the safe choice. - Can I verify the source of an increase?
- Yes. Every row in dividend-changes carries the document ID of the disclosure it was detected from (
sourceDocId) and the announcement time. - Can a streak be shorter than the truth?
- Yes. A count with
capped: trueinreportedHistorywas cut off at the oldest year in the dividend data.seriesOldestYeargives that year.
Related articles
Disclaimer
This guide explains how the data works and how to use the API. It does not recommend buying or selling any security and is not investment advice. Example values reflect responses at the time of writing and may change with later filings or corrections. Investment decisions are your own responsibility.

