Skip to main content
By Fruit Stand Team · November 15, 2026 · Guides If your price data has close and adjusted_close, you already have dividend history. On each ex-dividend date the factor adjusted_close / close steps up, and the dividend equals the prior close times 1 - 1/step. On our Snowflake data this reproduces Coca-Cola’s $0.51 and $0.53, Johnson & Johnson’s $1.30 and $1.34, and NVIDIA’s $0.25 to the cent. Below is the method, the SQL for one ticker and for every US stock at once, how to keep splits out, dividend growth and yield, and where the derived numbers stop being reliable.
Fruit Stand publishes the Historical Stock Prices listing used here. Every query below was run on November 15, 2026, against prices through November 13, 2026. Declared dividends were checked against company press releases and SEC filings, and other providers’ details on their own pages, the same day.

Where bulk dividend history comes from

Per-ticker dividend history is easy to find for free. Every US stock at once, in a form you can query, usually isn’t. If you need declaration, record and payment dates, or forecasts, a paid dividend feed is the right tool. If you need ex-dates and amounts for many stocks, and you already have adjusted prices, from our listing or any other source, you can compute them yourself.

How the adjustment factor encodes each dividend

Adjusted prices are built by multiplying every price before an ex-date by 1 - dividend / prior_close. Adjusted Close vs. Close walks through that adjustment; here we run it backward. Let factor = adjusted_close / close. It’s 1.0 on the latest date and smaller further back. On a normal day it doesn’t change. On an ex-date it jumps from prev_factor to factor, and the ratio between them is
The ex-date is the row where the step lands, and prev_close is the raw close the trading day before. Because it uses raw closes, the result is in as-traded dollars: the amount the company declared, not a split-adjusted one.

The SQL for one ticker

This returns Coca-Cola’s dividends since January 2025. The date filter sits in the outer query so lag() still sees the last trading day of 2024.
The two filters do the work:
  • prev_close * (1 - prev_factor / factor) >= 0.005 keeps steps worth at least half a cent a share. adjusted_close is stored to four decimal places, so the factor wobbles slightly from day to day on cheap stocks. On stocks above $1 that wobble is normally well under half a cent.
  • factor / prev_factor < 1.2 drops splits, covered next.
The 0.5101 values are that same rounding. Round to the cent, or compare with declared amounts within $0.001.

Derived vs. declared amounts

The same query for Johnson & Johnson, Apple and NVIDIA, January 2025 to October 2026:
Every amount matched to the cent, and every ex-date we checked matched the company’s record date. That’s expected: since US stocks moved to T+1 settlement in May 2024, the ex-date and the record date are the same day.

Keeping splits out

A split moves the factor too. NVIDIA’s 10-for-1 split took effect on June 10, 2024 (8-K). This query lists every day NVIDIA’s factor changed in the rest of 2024, before any filter:
The split day has a step of 10.0 and an “implied dividend” of $1,087.99, which is nonsense. The < 1.2 cap removes it. Splits always land on simple ratios: 2.0 for 2-for-1, 1.5 for 3-for-2, 10.0 for 10-for-1. Reverse splits give a step below 1, which the half-cent filter already drops. The trap is the lower bound. A common shortcut is to treat only steps above 1.0005 as dividends. NVIDIA’s $0.01 on a $120 stock is a step of 1.00008, so that shortcut silently deletes NVIDIA’s dividends. Across 2025 it would have dropped 94 real dividends at 32 listed stocks, mostly low-yield payers with high share prices. A threshold on the dollar amount (half a cent) instead of the ratio keeps them.

Dividend growth by year

Sum the derived amounts by calendar year. Coca-Cola, 2015 to 2025:
Each year is four times that year’s quarterly rate, from $0.33 in 2015 to $0.51 in 2025. Two cautions. Grouping by ex-date year can give a fifth payment in one year and three in the next when an ex-date crosses January
  1. And amounts are as traded, so across a split you must divide earlier amounts by the split ratio before comparing.

Every stock that paid in a year

The same logic across every common stock listed on NYSE, Nasdaq and the other national exchanges, for 2025, grouped by how often each one paid:
That’s 8,823 derived dividends at 2,096 stocks. To get the list itself, select * from divs instead of the final grouping. The close >= 1 filter leaves out sub-dollar stocks, where four-decimal rounding makes the factor too noisy to trust. OTC stocks are left out because their adjusted prices are noisier too. Then check the rows the < 1.2 cap removed that aren’t clean split ratios:
It returns 21 rows for 2025. They’re a mix of large special dividends, spin-offs and odd-ratio share exchanges. Western Digital’s 1.323 on February 24, 2025, for example, is the Sandisk spin-off, which shareholders received as stock (8-K/A). Decide on these by hand.

Trailing 12-month dividend yield

Trailing yield is the last 12 months of dividends over the latest close:
Trailing yield lags raises. NVIDIA’s window still holds two $0.01 payments, so its yield at the current $0.25 rate would be about 0.61% ($1.00 / $162.85). For the yield on a full investment including price moves, see How to Calculate Stock Returns From Daily Prices in SQL.

Limitations

  • The ex-date is the vendor’s. The derived date is the day the vendor applied the adjustment. It normally matches the exchange’s ex-date, but if the vendor’s date is wrong, so is yours. Spot-check a few against the company’s investor relations page.
  • No declaration, record or payment dates. Prices only reveal the ex-date. For US stocks since May 2024 the record date equals the ex-date; declaration and payment dates aren’t recoverable.
  • Special dividends, return of capital and spin-offs look like regular dividends. A one-time payment under the 1.2 cap is counted like any other. So is a small spin-off, and a 5% or 10% stock dividend shows up as a step of 1.05 or 1.10.
  • Rounding. Expect errors of up to about a tenth of a cent, more on cheap stocks and in early history, where the factor is small. Fractional amounts such as $0.485 survive, but round to the nearest tenth of a cent before comparing.
  • Our data has no dividend table and no delisted stocks. The listing covers securities that still trade, so a count across past years misses companies that have since been acquired or delisted.

F.A.Q.

Yes. On the ex-date, divide that day’s factor (adjusted_close / close) by the previous day’s to get the step, then compute prev_close * (1 - 1/step). On Coca-Cola, Johnson & Johnson, Apple and NVIDIA in 2025–2026 it matched every declared amount to the cent.
Because the adjustment is a cumulative multiplier over every dividend and split since that date. Only the day-over-day change in the factor isolates a single dividend.
Cap the step below 1.2 and require the implied amount to be at least half a cent. Splits sit at simple ratios (1.5, 2, 3, 10), and reverse splits fall below 1. Don’t use a 1.0005 floor on the ratio: it drops small dividends on high-priced stocks, such as NVIDIA’s $0.01.
Not from the usual APIs. EODHD’s free plan gives dividends one ticker at a time, and its bulk downloads need a paid plan. ORATS and Wall Street Horizon are paid feeds. If you already have adjusted daily prices, deriving dividends costs nothing extra.
No. It has daily raw and adjusted prices for US stocks, ETFs and funds, and the queries above derive ex-dates and amounts from them. It doesn’t have declaration or payment dates.
To derive dividend history for every US stock in SQL, start a free 30-day trial of the Historical Stock Prices listing on the Snowflake Marketplace. The schema is in the dataset docs. Questions? Reach out at contact@fruitstand.dev.