Data study
The Restatement Index
93,324 annual reports of 13,282 companies, fiscal years 2010 to 2024, and every later filing that changed one of their numbers.
- How often a number in an annual report is changed by a later filing, and by how much
- Which lines move and which hold, and why changed revenue almost always goes one way
- When the change arrives, and what that means for a backtest without lookahead bias
PDF, 19 pages.

Foreword
Why we measured this
Every financial model and backtest relies on one fundamental assumption: the historical numbers in a financial database reflect what companies actually reported at the time.
But that's not always the case. Financial statements don't necessarily stay the way they were originally filed. When companies publish subsequent reports, they often include figures from previous years, and sometimes those numbers have changed.
And that's an important distinction, especially for investment research. If a model uses today's version of a historical number, it might unknowingly be using information that wasn't available at the time. That's lookahead bias, and it's particularly difficult to detect because the numbers themselves aren't necessarily wrong.
So we decided to measure just how widespread this actually is.
We analyzed 93,324 annual reports, tracked their financial figures across subsequent SEC filings, and measured how frequently those numbers changed, by how much, and when those changes occurred.
Andy ReimannFounder, StockFit API
Inside the study
6 questions, answered with data
Each part of the study is built around one question. The answers, the charts and every number behind them are in the PDF.



01
How often
How often does a later filing change a number that was already reported?
02
Which numbers
Which line items change most, and which ones hold?
03
Revenue
When revenue changes, which way does it go, and what does net income do?
04
When it arrives
How long after the first report does a change arrive, and which filing brings it?
05
Who is affected
Is it a handful of companies, or almost all of them?
06
Is it changing
Do reported numbers change more or less often than they used to?
The data
What it is built on
- annual reports, each followed through the filings after it
- 93,324
- companies, delisted ones included
- 13,282
- reported numbers from three financial statements
- 16.5M
- fiscal years, 2010 to 2024, annual and quarterly
- 15
Definitions
How the study counts
- What counts as a change?
- A later filing that reports another amount for the same line item and the same period, at least 1% away from the amount before. A number that was only rounded again, or shown with the opposite sign, is not a change.
- What is the Restatement Index?
- The share of a fiscal year's annual reports that had at least one number changed within 400 days of the report. The fixed window gives every fiscal year the same time, so the years can be compared.
- What is lookahead bias?
- Using information in a backtest or a model that was not known at the time. With financial data it happens when a later, changed number stands in for the number that was first reported.
- Is every change a restatement?
- The study counts every change of a reported number, whatever its reason. A formal restatement that corrects an error is one reason among several, and the study looks at how they differ.
- Where does the data come from?
- From the filings companies submit to the SEC, as the StockFit API serves them: income statement, balance sheet and cash flow statement, as first filed and as changed later. Every change was also compared with the SEC's own record of each filing.
- Is this investment advice?
- No. The study describes what happened in past filings. It makes no forecast and recommends no trade.
Read the answers
Six parts, 19 pages, every chart built from point-in-time data.
PDF, 19 pages.