event study that affects all firms - equally-weighted calender-time portfolios

[Michal Barzuza]

Hi
when the event study effect all firms, there is a concern of nonidependence of returns.
Thus recent papers have used equally-weighted calender-time portfolios approach (Eckbo et al 2018)

Is the princeton code ppropriate when it is run only on the sample of affected firms?
https://dss.princeton.edu/online_hel...ventstudy.html

if not is there an example for a code of an event study with the portfolio approach?
Thx

[David Benson]

Hi Michael, I've run event studies, but I'm not familiar with the "equally-weighted calendar-time portfolios", but hopefully I can help point you in the right direction.

1. I couldn't find the (Eckbo et al 2018) paper you mention, could you post a link?

2. The code at the Princeton site is for a "regular" event study

3. If the event in question affects all firms, or affects all firms on the same day (as when a court case is decided or a law is changed that affects all firms), this is often referred to "event date clustering" in the accounting / finance literature. This may give you some things to search for in Google Scholar.

• If event date cllustering is the cause for the non-independence, I posted some papers that might be helpful at https://www.statalist.org/forums/for...d-ols-or-xtreg

4. If you have access to CRSP and Compustat through WRDS, you probably have access to Eventus or "Event Study by WRDS" accessible here and here (obviously they are paywalled, so if you don't have access at your institution, those links won't do you any good. Sorry.)

• The Eventus 9.0 manual has a section 1.5 on Calendar-time tests, and section 1.5.2 mentions "Calendar-time portfolio regression method." Manual is http://www.eventstudy.com/Eventus-9-...d-Features.pdf
• I don't know if the full 9.0 manual is public, but you can see the 8.0 manual here http://www.eventstudy.com/Eventus-Guide-8-Public.pdf (then search within it for "calendar-time")