GMgn%Gr%PYQ Calculation - Unusual Behavior

Report a problem or error
1

Gents,

Curious if anyone has any insight on what I am observing with the GMgn%Gr%PYQ calculation for a handful of stocks. My understanding is that the calculation should be

((GMgn%Q-GMgn%PYQ)/abs(GMgn%PYQ))*100

However, as you can see in the referenced screen for a handful of stocks that calculation does not give the same results as the GMgn%Gr%PYQ Factor.

https://www.portfolio123.com/app/screen/summary/258621?mt=1

Note: Run the Screen on the Data 01/03/2003 to see the stocks I am referring to.

For all 5 of the stocks referenced in the screen the values of GMgn%Q and GMgn%PYQ are clean (no negatives, NAs, or extremely small #'s).

Not at all clear what is going on here.

Grateful for any thoughts you guys can provide.

Thanks,

Daniel

2

I’m not seeing a problem here. Perhaps it went away?

Screenshot_2021-06-01 Screener - Portfolio123.png
Screenshot_2021-06-01 Screener - Portfolio123.png1252×642 42.9 KB

3

I cannot reproduce it for that 2003 date either.

FYI - You will see some cases starting in 2020 where ((GMgn%Q-GMgn%PYQ)/abs(GMgn%PYQ))*100 != GMgn%Gr%PYQ. The rule CompleteStmt=0 will show that these are prelims. There is special handling for these in the code.

4

Yuval,

Thanks for taking a look at this so quickly. I should have pointed out that I am using the Compustat data. Are you able to take a look at it using the Compustat data?

Thanks,

Daniel

5

It does appear that all the stocks in question have CompleteStmt=False, but on that day (01/03/2003) in the Compustat data set only 580 out of 8048 stocks have CompleteStmt=True so it seems that might be a false indicator.

6

Note that the CompleteStmt Flag seems to be exhibiting some very unusual behavior in the Compustat dataset. This graph below shows the % of stocks in the ALL FUNDAMENTALS universe listed as having a Complete Stmt over time. This had to indicate some kind of problem. Does the FactSet data have the same issue?

CompleteStmtOverTime.PNG
CompleteStmtOverTime.PNG919×287 28 KB

7

No, with FactSet we don’t have that issue. Here’s the chart, which is what you’d expect. We’ll look into the CompleteStmt problem for pre-2006 numbers with Compustat and update you shortly.

Screenshot_2021-06-02 Aggregate Series - Portfolio123.png
Screenshot_2021-06-02 Aggregate Series - Portfolio123.png1119×359 23.2 KB

8

We’ve looked into the Compustat situation and found a bug that’s causing this behavior before a certain date in 2007. We’ll try and get this fixed in the next few days. This will affect some fundamentals because the bug is causing some unnecessary fallbacks.

Thank you, Daniel, for calling this to our attention. It’s a valuable fix indeed for Compustat subscribers.

9

Thanks Yuval, I really appreciate your help with this. Is there any indication as to how long this error has existed? Is it due to a recent change, or has it been around for awhile?

Also, would you mind dropping a note here when the error has been resolved.

Thanks again,

Daniel

10

It looks like it happened in November 2020. I will drop a note when it’s been resolved, but it might take a few days.

11

There are two issues going on here.

PYQ is occasionally giving you the wrong quarter when there’s a prelim fallback. Gr%PYQ is correct. Try using “exclude prelims” in your screen and you’ll see that it works. We’re going to work on a fix for that at some point soon.

The fix for CompleteStmt in Compustat will be released later today.