Avoiding Look-Ahead Bias with Financial Statement Release Timing
Summary
The document considers when net income and operating cash flow can legitimately enter a backtest that trades around quarterly reports. Its key principle is to use information only after it became public. An earnings date may identify a press release, but releases vary: some include income and cash-flow figures, while others provide only selected earnings measures. The release date alone therefore does not guarantee that every input to a strategy was available.
The answer recommends checking the disclosures for the companies in the sample and notes that disclosure differences can create selection bias if the backtest includes only firms whose releases contain the desired data. It presents the SEC filing date as a safer, simpler trigger when the researcher cannot verify earlier availability. The document does not establish that every XBRL filing contains usable figures at a uniform time, nor does it resolve vendor-specific timestamps; realistic testing still depends on matching each signal to the actual public availability of its inputs.
Key ideas
- A backtest should use financial statement figures only after they became public.
- Earnings press releases do not consistently provide operating cash flow and income figures.
- Check company disclosures before treating an earnings date as the trigger for every signal.
- Selective availability of figures can bias a backtest toward companies with fuller releases.
- An SEC filing date is a safer trigger when earlier data availability has not been verified.
Tags
Full text
# When is cashflow data available exactly? # When is cashflow data available exactly? I want to perform a backtest that should be as realistic as possible. My strategy uses, among other things, the Net Income and Cash Flow from operating activities from the Income and Cash Flow statements. The strategy usually goes long or short at the release of the quarterly statement. Therefore, the timing is extremely important, as prices move dramatically around that time. When do I have the aforementioned figures in reality (I'm using Refinitiv data)? I do have the historical dates for when the 10-Q XBRL was filed at the SEC and also the date and time (BMO or AMC) of the press release (commonly referred to as the "Earnings Date") there. - Do these press releases always contain operating cash flow and income (I'm fine with unaudited and, if necessary, only non-GAAP taxonomies)? - Is it then appropriate to use the "Earnings Date" as the date/time which triggers the signals in the backtest? Or do I only have these figures once the XBRL was filed? I checked a sample of press releases in the investor relations sections of some companies. Some include both cash flow and income figures. In others, only non-GAAP earnings per share were mentioned, but no cash flow statement numbers. ## Answer by phdstudent (score 2) https://quant.stackexchange.com/a/81621 I think you answered your own question. To properly backtest a strategy you can only use information once it became public available. So directly to your questions. - Do these press releases always contain operating cash flow and income (I'm fine with unaudited and, if necessary, only non-GAAP taxonomies)? Nothing like checking it yourself. If they do great, and you can use this to backtest, if not, then you need to think about adverse selection. What companies are likely to disclose this on press releases? So your backtesting may be really biased. - Once XBRL was filed seems that right, safe, and easy way to go about this backtesting.
Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.