1
$\begingroup$

I am running a regression where I regress return of a stock on the market return. There is a specific date in my sample on which some event occurred. I believe the effect of this event to last only on that date. I want my model coefficients to be free of this effect because I want to predict what the return would be, had this event not occurred. I am thinking of including a dummy for that date (i.e. it is one for that date and zero for the rest of the sample). Is this an appropriate way to handle this issue?

Thank you.

$\endgroup$
1
  • 4
    $\begingroup$ It might be better to delete the data for that date, if you can justify doing so. $\endgroup$ – Peter Flom Jun 10 '12 at 22:10
0
$\begingroup$

Yes I think that can be appropriate. What Peter suggests would work too and possibly better at obtaining your objective of no influence on the other coefficients but you have no really good reason to remove it and you may want the effect of the event as part of the model.

$\endgroup$
2
  • $\begingroup$ A dummy that has a unique value for a single case (as seems to be the situation here) is tantamount to deleting the case from the regression and only creates useless additional output. In a more general setting, where multiple cases might correspond to the dummy indicator, this recommendation would make more sense. I still wonder, though, whether the dummy ought not to indicate all dates after the event. How can an event affect only the returns for that date without (perhaps in some complex way) also affecting subsequent returns? $\endgroup$ – whuber Jun 11 '12 at 12:14
  • $\begingroup$ It could be a single point blip like an object whizzing by on a radar screen. I do think it is different because it models the special point rather than ignore it. $\endgroup$ – Michael R. Chernick Jun 11 '12 at 12:26

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.