New flag to detect stale compustat data

Latest Release
41

Hugh,

The confusion is that the factors use different sources for the data. Some use the actual “effective dates” on when the data is inserted in Compustat , others use the general public filing information (which has many sources types like: sec filings, 6K’s, press releases). We use these different sources so we can warn you that the data is stale, which is all we can do.

WeekIntoQ is based on when the data on Compustat database was inserted. On July 15 2019 the most recent period that is visible on Compustat is for 12/31/18. For this period the very first line items were added on 3/22/19. The difference between July 15 and March 22 is 115 days or 16.4 weeks which is truncated to 16.

LatestFilingDate looks at latest filings data regardless of what is actually in Compustat. On July 15 2019 and initial 6-K was filed on 5/7/19 for period 3/31/19. But the first data to actually make it into Compustat for 3/31/19 happened on 7/18/19 , two months later!

In short , on July 15 2019 the data is stale. If you add “StaleStmt” to your screen you will see that it’s set to 1 , basically telling you that there’s fresher data somewhere , just not on Compustat.

I suppose it’s debatable what WeeksIntoQ should use. Should it use the Compustat effective date or the period filing general source dates ? Also some of the names for these factors need to be revisited. Also, the documentation needs to be clearer.

Thanks for bringing it to our attention

PS. I do not know why Compustat has a “First 6K” filed on 5/7/19. The company press release seems to be on 5/15/19

42

Hi Marco,

Thank you very much for your explanation. I can see how difficult this is to deal with.

I have one major request if, as I hope, p123 expands how it handles this information: Please leave WeeksIntoQ as is and create additional variables as appropriate. a) I think there is something to be said for knowing how long the data has been available in CompuStat; and b) I’ve been using WeeksIntoQ, and I assume other subscribers have as well, for so long now that it would take major retooling to abandon it as a variable even if over time another iteration proves more useful.

Thanks again.

Hugh

43

Yuval, is this still the preferred method for finding calendar days since the most recent filing?

44

I’ve simplified it a bit and it works pretty well for me:

IsNA (LatestActualDays, DaysSince (Max (LatestFilingDate - 1, LatestNewsDate)))