Questions / Quant research, Backtest integrity

Quarter end is not publication time

Foundation, about 6 minutes. Question R022.

An earnings value is labeled March 31 but was first published April 20. A historical model trades on April 5. Which use is valid?

  1. AUse the value on April 5 with its March 31 period label.
  2. BUse the March 31 period label as the value’s first availability date.
  3. CWait until the April 20 publication before making the value available.
  4. DInterpolate the value between the previous report and the March 31 label.

Answer it in the app It's free. You get hints, a second try if you miss, and a follow-up question.

Show the answer and worked solution

Answer: C. Wait until the April 20 publication before making the value available.

  1. The relevant clock is publication availability, not the date the accounting period ended.
  2. On April 5 the stated figure has not yet been published.
  3. Exclude it from that decision; any prior available release must be represented under its own vintage and timing.

Interview shortcut. Use publication time, not period end.

Takeaway. Backtest joins should use information availability, not just economic period labels.

Why the other choices fail

  • Use the value on April 5 with its March 31 period label. A fixed lag can still precede the April 20 release.
  • Use the March 31 period label as the value’s first availability date. A reporting-period end identifies the economic period, not when the number became known.
  • Interpolate the value between the previous report and the March 31 label. Interpolating toward a not-yet-published figure still allows future information into earlier features.

In the app this has a follow-up where one thing about the setup changes. Try the follow-up.

More backtest integrity questions