Questions re BHAR and eventstudy2

[Marcus Silkisberg]

Hi all,

My thesis partner and myself are in the midst of writing our master's thesis on long- and short-term SPAC performance using an event study methodology

We are using Thomas Kaspereit's fantastic eventstudy2 command to calculate abnormal returns

We have calculated short term CAR using the market model (i.e. mod(FM)) and are now trying to calculate long-term abnormal returns using BHAR

However, we have encountered some issues ...

The "help eventstudy2" specify that the stata user may employ any event window that they wish. However, when we try to use event windows exceeding >20 days, our "carfile" caps the number of days at 20. This is an issue as we want to compute abnormal returns for >20 days (for example 90 days, or 120 days). Below, I am attaching the code that we have used (short but sweet):


eventstudy2 Company_ID Date using Company_3, ret(Return) car1LB(-1) car1UB(1) car2LB(-1) car2UB(10) car3LB(-1) car3UB(30) car4LB(-1) car4UB(90) mod(BHAR) marketfile (Market_Russell) mar(Market_returns) idmar(Market_ID)

Note that we are using three separate dta files as input:
(1) Market returns
(2) Security returns
(3) Event days with Security ID

As our code worked perfectly in the short term - we do not believe there is an issue with our labelling or code.

I am attaching our results files if that's helpful

We have one related questions as well:

(1) Suppose we would like to base our analysis on weekly returns, instead of monthly. Is there an easy way in stata to transform market return and security return sheet? Considering our data is collected from CRSP and all our company returns are stacked

If anyone is able to contribute with some stata-expert knowledge, that would be very appreciated

Best,
Marcus

[Thomas Kaspereit]

Dear Marcus,

eventstudy2 uses a default event window of [-20,+20]. If the user specifies CAR/BHAR windows exceeding theses limits, it cuts back those windows and issues a warning.

Please use the evwlb and evwub options to extend the event window in order to make it encompass your CAR/BHAR windows.

I cannot comment on your second question but would like to emphasize that eventstudy2 was designed for studies based on daily returns. Although I think that it could also work with lower frequency data, if the date variable is properly defined as an integer variable with distance "1" between each unit of time, it has not been extensively tested on lower frequency data.

Best Thomas