EViews.com

Hi!

We are currently working with an event study and experience some problems with setting estimation period. We are running the following regression as Cross-Section SUR (PCSE):

LOG_Returns c LOG_Returns(-1) WMP(-1) WMP WMP(+1) Monday Tuesday Wednesday Thursday ,

where LOG_Returns represent index returns for seven separate countries, WMP is market return and the remaining are dummies for weekdays.

The problem ouccurs when we want to estimate the coefficients with an estimation period. Ex: We have 100 events, and when each event ouccurs we want to measure the regression coefficients using data from the previous 200 days, estimation period = [-200,0] where 0 is the day of the event. We want to save the residual from all these 100 equations as abnormal return, such that we get a series with abnormal returns for our 100 events.

How can we resolve this issue? (little experience in EViews)

How/where are the events stored?

EViews Gareth wrote:How/where are the events stored?

We have organized the data in excel as:

A B C D E F G-K

Country 1 Date LOG_Returns WMP Event1 Event2 Mon-Thur
.
.
Country 2 Date LOG_Returns WMP Event1 Event2 Mon-Thur
.
.
Country 7 Date LOG_Returns WMP Event1 Event2 Mon-Thur
.
.


For Event1, Event2 and Mon-Thur we have dummies 1 if it occurs and 0 if not.

So the events are common across cross-sections? (I think they have to be, but want to be sure)

EViews Gareth wrote:So the events are common across cross-sections? (I think they have to be, but want to be sure)

Yes

EViews Gareth wrote:So the events are common across cross-sections? (I think they have to be, but want to be sure)

I must add: We have dates for LOG_returns in column B and we have added a new column C, with the dates of the market return. We have used a VLOOKUP formula in excel to have these dates mach eachother. How can we do the same thing in EViews?

I still can't quiet visualise how the data are structured. Could you provide the Excel file?

EViews Gareth wrote:I still can't quiet visualise how the data are structured. Could you provide the Excel file?

Yes. In this excel file we have already matched the dates between log returns and market return.

mikaldj wrote:

EViews Gareth wrote:I still can't quiet visualise how the data are structured. Could you provide the Excel file?

Yes. In this excel file we have already matched the dates between log returns and market return.

EDIT "Column C should be removed"

Something like this:

would work, but you have some data issues I think which prevent it from running fully.

EViews Gareth wrote:Something like this:

Code: Select all

smpl @all
series temp = win
for !i=1 to @obsrange
   if temp(!i)=1 then
      temp(!i)=0
      %enddate = @otod(!i)
      %startdate = @otod(!i-99)
      smpl {%startdate} {%enddate}
      equation eq_!i.ls LOG_Returns c LOG_Returns(-1) WMP(-1) WMP WMP(+1)
      !i=0
   endif
next


would work, but you have some data issues I think which prevent it from running fully.

Thanks! Will try it out tomorrow and let you know if it worked

mikaldj wrote:

EViews Gareth wrote:Something like this:

Code: Select all

smpl @all
series temp = win
for !i=1 to @obsrange
   if temp(!i)=1 then
      temp(!i)=0
      %enddate = @otod(!i)
      %startdate = @otod(!i-99)
      smpl {%startdate} {%enddate}
      equation eq_!i.ls LOG_Returns c LOG_Returns(-1) WMP(-1) WMP WMP(+1)
      !i=0
   endif
next


would work, but you have some data issues I think which prevent it from running fully.

We ran the code, and we got some results. However, the code yields several residuals in a row i.e that we have residuals on a monday tuesday and wednesday, this should not be possible. We only need the residual on the day of the event. Another issue, we need to estimate the regressions as Cross-section SUR (PCSE).