Search

Best Measure for a Normal Return in an Event Study
 Posted: 10 March 2014 04:37 PM [ Ignore ]
Newbie
Total Posts:  4
Joined  2014-03-10

Hi Joost,

I am attempting to do a simple event study concerning regulators leaving office. I know the precise event dates and have the daily returns for the firms I need but I was wondering what would be the best measure for a Normal Return. In other words, what should I compare the individual firms’ stock returns to in order to get an abnormal return?

P.S. I’m an Econ PhD student at UF. You were recommended to me by a friend in Accounting.

-John

 Profile

 Posted: 11 March 2014 11:39 AM [ Ignore ]   [ # 1 ]
Total Posts:  901
Joined  2011-09-19

hi John,

Models that come to my mind to control for ‘normal’ returns are (in order of ‘usefulness’):
- market model
- capm
- 3 or 4 factor model

With the market model, the expected return is the market return for that day. Firm A goes up by 2%, market moves by 0.5%, then the abnormal return for A for that day is 1.5%. A common choice for the market model is the equal weight/value weight of firms on Nyse/Amex/Nasdaq. See: http://wrds-web.wharton.upenn.edu/wrds/tools/variable.cfm?library_id=20&file_id=67317

With size adjusted returns, you would subtract the average firms with a similar size of the firm. For this to work, you need to retrieve the size decile of the firm. For example, http://wrds-web.wharton.upenn.edu/wrds/tools/variable.cfm?library_id=20&file_id=67324

With CAPM, you would need to estimate the firm’s beta prior to the event. The beta gets multiplied with the market return to get to the expected return. For example, firm A’s beta is 1.5 and moves up by 2%; market moves by 0.5%; abnormal return is 2% - 1.5 x 0.5% = 1.25%. In this tutorial I have SAS code that computes betas. http://www.wrds.us/index.php/tutorial/view/17

Instead of a single factor like ‘beta’ or ‘size’, you can use the Fama French 3 factor model (beta, size, book-to-market), or 4-factor model (including momentum). These factors are on WRDS (and on Kenneth French’s website). (Unless you really need to, I would use the market model and size adjusted returns)

I believe UF has access to ‘Eventus’, which is a tool to do eventstudies. (Good for quick and dirty, but probably not suitable to use in a paper, because it is too much of a black box).

Hope this helps,

Joost

Signature

 Profile

 Posted: 11 March 2014 11:49 AM [ Ignore ]   [ # 2 ]
Newbie
Total Posts:  4
Joined  2014-03-10

Thank you Joost, this helps a lot. The market model with size adjusted returns does indeed sound like the best option for me. Just to make sure I completely understand how to calculate these adjusted returns, you are saying adjusted return=equal weight return-decile return for the decile of the firm in question? Also, should I used the returns that include or exclude dividends?

-John

 Profile

 Posted: 11 March 2014 12:49 PM [ Ignore ]   [ # 3 ]
Total Posts:  901
Joined  2011-09-19

hi John,

Using the market model and size adjusted returns are two distinct models, so you would get two measures for abnormal return.

market model: abnormal return = raw return - market return
size adjusted return: abnormal return = raw return - size decile return

best regards,

Joost

Signature

 Profile

 Posted: 11 March 2014 01:24 PM [ Ignore ]   [ # 4 ]
Newbie
Total Posts:  4
Joined  2014-03-10

Thank you once again, that is helpful. I have one more question. If I am looking at a subset of firms, say electric utility companies, could I calculate a market return for just the subset and compare it to the individual firms to get an abnormal return or do I need to use the entire market?

-John

 Profile

 Posted: 11 March 2014 01:38 PM [ Ignore ]   [ # 5 ]
Total Posts:  901
Joined  2011-09-19

hi John,

That is a tough call.. Maybe do both, and footnote one of them.

best regards,

Joost

Signature

 Profile

 Posted: 11 March 2014 01:44 PM [ Ignore ]   [ # 6 ]
Newbie
Total Posts:  4
Joined  2014-03-10

Do you have a on how to calculate your own market returns in sas or stata? Also, when doing the entire market return, should I use the equal weighted returns including dividends or excluding them?

-John

 Profile

 Posted: 11 March 2014 01:55 PM [ Ignore ]   [ # 7 ]
Total Posts:  901
Joined  2011-09-19

hi Jon,

I would think it shouldn’t matter (value weighting vs equal weighting); both are provided by wrds for the ‘full’ market.

When computing your own sub-market return, equal weighting is easier. You would collect the stock returns for the firms in your market, and then compute a mean return for each date. In SAS, that can be done with proc means, in Stata with ‘egen’ (by group).

best regards,

Joost

Signature