You can find the spreadsheets for this video and some additional materials here: drive.google.com/drive/folders/1sP40IW0p0w5IETCgo464uhDFfdyR6rh7 Please consider supporting NEDL on Patreon: www.patreon.com/NEDLeducation
@NEDL Hello, this is a great video which is helping me a lot . However, can you please recommend, how to write the regression equation using CAPM, and how to name the results once you apply the linest function (in 5 row and 4 row table)? Also, would you recommend to use the linear regression with dummy variables, or the AR and CAR method? Please, a response would save me right now. Best, Aida
@NEDL, Many thanks for all your educative videos. Could you please share your insight on using dummy variable on a corporate bond index using Eviews. A kind and prompt response will be appreciated.
To be clear, that was the day of the announcement of the game? If so would not be expected to wait a few weeks in order to see that the community did not liked the game and then sells went down?
Great videos from you guys Saba (hope I spelled your name correctly). If time permits could you show a similar event analysis announcement but looking at multiple companies, dealing with more than one company. Thank you
Hi Muhammad, and glad you are enjoying the videos! As for your suggestion, this is indeed possible, and I will most likely release a video on that sometime over the summer. Long story short, if there are many assets exposed to the same type of event at different points in time, event studies can be applied to a "quasi-portfolio" of such assets at day t with regards to the event date.
@@NEDLeducation Hi Savva, thanks a lot for your kind considerations. And thanks for the further explanation on the quasi-portfolio concept. I was actually thinking about how to used the event study to assess whether a stock market is actually weak-form or semi-strong form efficiency using a number of earnings or dividend announcements of companies listed from the market.
Also if I want to claculate my results for different event windows like +2, -2 (5days) and +5,-5(11 days window) instead of anticipated and adjustment window separately , can I assign 1 as dummy variable from +2 to -2 (5days) and similarly I can assign 1 as dummy variable from +5 to -5 (11 days ) in a separate column rather that taking anticipation and adjustment window
Hello sir! First of all, I have to say that this is my very first comment on youtube (and I have been using it for a bunch of years) and I just can congratule you for all the videos you have upladed. In addition, I wanted to ask you two questions: - The results for the adjustment/anticipation window are daily results, right? I mean, we would expect that the impact of the event for the 10 days after it happened would have an impact on the stock price of -2.41% (CAPM) per day, right? Or it would be on a cumulative basis for the whole adjustment period? - When conducting this study, should we check for heterokedasticity problems? If so, could you share with us some tips on how to do it? Thank you in advance!
Hi Jaime, and thanks so much for the feedback! Glad you are enjoying the channel. As for your question, you can either calculate CAR (cumulative abnormal return) for the full period or AAR (average abnormal return) per day. Dummy variable regressions naturally give you AAR, so here it is abnormal return per day. As for heteroskedasticity, you can naturally implement robustness procedures like the heteroskedasticity-consistent standard errors (th-cam.com/video/EOZvh5r6U4c/w-d-xo.html) in a dummy variable regression framework.
Hello Sir, Thank you so much for this video .....indeed very helpful .....I am doing my research related to impact of monetary policy announcement on banking index..I mean I have taken multiple monetary policy announcement on one index. so can I use this regression with dummy variable...
Hi, and happy the video helped! Yes, indeed, you can use the dummy variable regression approach if you have multiple events, simply put 1 whenever you have an event and 0 otherwise (analogously with anticipation and adjustment).
Hi!, I would like to know please how would the regression equation for the market-adjusted model method look like and how would you write the hypothesis on a paper or thesis. It's urgent please!! Thanks a lot for your work, it's an amazing channel.
Hi Lev, and glad you are enjoying the channel! As for your question, you would write and equation like this: Rit - Rmt = b0 + b1*ANTit+b2*EVENTit+b3*ADJit + eit, where Rit is the return of the ith stock on day t; Rmt is the market return on day t; b0 is the constant; ANTit, EVENTit, and ADJit are the anticipation, event, and adjustment dummies respectively; b1, b2, and b3 are the average abnormal return estimators for the anticipation, event, and adjustment periods; and eit is the error term. Hope this helps!
@@NEDLeducation Thanks a lot for your quick response!, in the case of the t-test and P-Values how will the notations be, which is the level of significance for t statistics and the degrees of freedom for the P value?
@@levaddler2477 Hi, and thanks for the follow-up question! You can easily apply the T.DIST.2T function and apply it to the absolute value of the t-stat. The degrees of freedom would be the number of observations minus two.
Hello NEDL! I'm uncertain if you already explained this but what are the numbers we get out when we use the LINREGR formula. The top row under let us say the constant return model is the average abnormal return. Then u say the next row is the std. error but it ends there. I'm unsure what the rest of the numbers are referring to. The 4th row under anticipation has degrees of freedom. I want to use the data I have as a picture, but unfortunately, I cannot present numbers with no text associated with them. I learned a lot! Thank you!
Hi Nikolaj, and glad you liked the video! I have explained the LINEST (or LINREGR in other languages) output template in many videos, but long story short: in the first and second row, as you correctly stated, it reports the coefficients and respective standard errors. The third row reports the model r-squared on the left and the regression standard error (prediction standard error, standard deviation of the error term) on the right. The fourth row reports the F-stat (used to test for the significance of the overall model) on the left and the degrees of freedom on the right. The fifth row reports the explained and the residual squared sum (explained and unexplained variance) on the left and the right, respectively (so if you divide ESS by ESS+RSS, you get r-squared). Hope it helps!
Hello sir, first of all accept my kind regards for clearly presenting the event study. I request you to kindly conduct a event study in stata with ‘eventstudy2’ command. Thanks in Advance
Hi Mahesh, and thanks so much for the feedback! There are many packages that can be used for event studies, I do not use stata much in my teaching/research, but potentially, I will investigate your suggestion sometime in the distant future.
Hi Caio, and great question! The dummy variable regression is an excellent framework for this, if you believe the events are similar in nature and should have similar impact on the abnormal return, you can code the event, anticipation, and adjustment as shown here, simply inputting a 1 in the dummy variable column for each event date. If you have got multiple time series, they can be merged together in a panel and dummy variables coded in the same way. Hope it helps!
Hello NDL, I love your videos. I am currently performing an event study (CAPM) on the effect the announcement of trump that the US will leave the pairs climate accord has on the stock return for environmentally friendly and unfriendly companies. I have 441 companies for which degree of freedom do I have to incorporate to test the significance of the results?
Hi, and glad you like the videos! First of all, a very good research idea! As for your question, you can form a portfolio of such companies (equal- or value-weighted, for example), and test the significance of the event using the same approach. Alternatively, you can form a panel from your companies and code the event dummy variable for all of them. Hope it helps!
Could adjustment and event factors have different signs? And what would be interpreted in that case in relation to over or under-reaction in that case?
Hi Akansh, and thanks for the question! Yes, they can! If there are significant adjustment and event factors of different signs, you could interpret is as the market incorrectly anticipating the nature/the impact of the event. Hope it helps!
Hello sir, thank you for your session sir but I'm stuck with AAR, CAAR and T-statistics (market model(... Can you tell me sir how can I proceed to get above values
Hi mate, thanks so much for the helpful video, been a lot of help. A bit confused how you would lay this out in terms of a methodology and equations on a paper? as im confused what to include etc, any ideas?
Hi, and glad the video helped! As for the outline, you can represent event study regressions as: R(t) = b0 + b1*Event(t) + b2*Anticipation(t) + b3*Adjustment(t) + e(t); R(t) - M(t) = b0 + b1*Event(t) + b2*Anticipation(t) + b3*Adjustment(t) + e(t); R(t) = b0 + b1*M(t) + b2*Event(t) + b3*Anticipation(t) + b4*Adjustment(t) + e(t) for constant return, market-adjusted, and CAPM models, respectively, where R(t) is the stock return, Event(t), Anticipation(t), and Adjustment(t) are respective dummy variables, M(t) is the market return, and e(t) is the i.i.d. disturbance term. In terms of classical references on event studies, see MacKinlay (1997) and Brown and Warner (1985). Hope it helps!
Hi Amber, and glad you liked the video! To implement event studies for multiple companies, you can use a quasi-portfolio approach, calculating average returns across a set of companies at a particular date respective to company-specific ex-dividend dates. Might do a video on quasi-portfolio approach at some point! Hope it helps!
Hi! Thanks a lot for this great video! super useful! I just don't have clear if you are estimating CAR or BHAR with this method. Could you please clarify this? Thanks a lot again!
Hi Luis, and thanks so much for the comment! Dummy variable regressions naturally estimate average abnormal returns (AARs), which are basically CARs per day. Hope it helps!
Hello sir, iam Alia from Indonesia. First of all thank you so much for your explanation video about event study. I wanna ask you about my research. I would like to do event study about lockdown announcement in Indonesia using firm characteristics as independent variable and CAR as dependent variable. In your opinion what is the best regression model for this event?
Hi Aliakito, and glad you liked the video! I feel the regression approach would be quite fitting for your research design. You can calculate cumulative abnormal returns for your sample firms using some model (constant return, market-adjusted, or CAPM) and then regress these onto firm-level characteristics. A simple multiple linear regression would be a logical first step. Hope it helps!
Good Evening Sir, Hope you are doing well. Sir can you please clear my query regarding event studies. Can we run regression in event study without taking dummy variable into consideration? ARo = a+a1TC + a2 Dy + a3 Beta + a4 Av. Vol + a5Av.ab Vol + a6 Size Can I run regression on this model in event study.
Hi Amber, and thanks for the question! The dummy variable approach is useful to calculate the average abnormal returns themselves. If you have already calculated these using another approach and want to determine what impacts their magnitude, the approach you suggested is valid. Hope it helps!
Hi Alvaro, and thanks for the question. The regression equation would be: 1) stock return = a + b1*dummy(anticipation) + b2*dummy(event) + b3*dummy(adjustment) + error 2) stock return - market return = a + b1*dummy(anticipation) + b2*dummy(event) + b3*dummy(adjustment) + error 3) stock return = a + b*market return + b1*dummy(anticipation) + b2*dummy(event) + b3*dummy(adjustment) + error For constant return, market-adjusted, and CAPM models, respectively, with b1, b2, and b3 estimating average abnormal returns for anticipation, event, and adjustment periods. Hope it helps!
You can find the spreadsheets for this video and some additional materials here: drive.google.com/drive/folders/1sP40IW0p0w5IETCgo464uhDFfdyR6rh7
Please consider supporting NEDL on Patreon: www.patreon.com/NEDLeducation
New design looks great!
@NEDL Hello, this is a great video which is helping me a lot . However, can you please recommend, how to write the regression equation using CAPM, and how to name the results once you apply the linest function (in 5 row and 4 row table)? Also, would you recommend to use the linear regression with dummy variables, or the AR and CAR method? Please, a response would save me right now. Best, Aida
@NEDL, Many thanks for all your educative videos. Could you please share your insight on using dummy variable on a corporate bond index using Eviews. A kind and prompt response will be appreciated.
Great learning!! Need to practice and implement. BTW new haircut is nice!!
Thank you so much!
Hi Anil, and thanks for the compliment :)
To be clear, that was the day of the announcement of the game? If so would not be expected to wait a few weeks in order to see that the community did not liked the game and then sells went down?
Great videos from you guys Saba (hope I spelled your name correctly). If time permits could you show a similar event analysis announcement but looking at multiple companies, dealing with more than one company. Thank you
Hi Muhammad, and glad you are enjoying the videos! As for your suggestion, this is indeed possible, and I will most likely release a video on that sometime over the summer. Long story short, if there are many assets exposed to the same type of event at different points in time, event studies can be applied to a "quasi-portfolio" of such assets at day t with regards to the event date.
@@NEDLeducation Hi Savva, thanks a lot for your kind considerations. And thanks for the further explanation on the quasi-portfolio concept. I was actually thinking about how to used the event study to assess whether a stock market is actually weak-form or semi-strong form efficiency using a number of earnings or dividend announcements of companies listed from the market.
Also if I want to claculate my results for different event windows like +2, -2 (5days) and +5,-5(11 days window) instead of anticipated and adjustment window separately , can I assign 1 as dummy variable from +2 to -2 (5days) and similarly I can assign 1 as dummy variable from +5 to -5 (11 days ) in a separate column rather that taking anticipation and adjustment window
Yes, this would work!
Hello sir!
First of all, I have to say that this is my very first comment on youtube (and I have been using it for a bunch of years) and I just can congratule you for all the videos you have upladed.
In addition, I wanted to ask you two questions:
- The results for the adjustment/anticipation window are daily results, right? I mean, we would expect that the impact of the event for the 10 days after it happened would have an impact on the stock price of -2.41% (CAPM) per day, right? Or it would be on a cumulative basis for the whole adjustment period?
- When conducting this study, should we check for heterokedasticity problems? If so, could you share with us some tips on how to do it?
Thank you in advance!
Hi Jaime, and thanks so much for the feedback! Glad you are enjoying the channel. As for your question, you can either calculate CAR (cumulative abnormal return) for the full period or AAR (average abnormal return) per day. Dummy variable regressions naturally give you AAR, so here it is abnormal return per day. As for heteroskedasticity, you can naturally implement robustness procedures like the heteroskedasticity-consistent standard errors (th-cam.com/video/EOZvh5r6U4c/w-d-xo.html) in a dummy variable regression framework.
@@NEDLeducation Thank you very much! Gonna check that video right now!
Hello Sir,
Thank you so much for this video .....indeed very helpful .....I am doing my research related to impact of monetary policy announcement on banking index..I mean I have taken multiple monetary policy announcement on one index. so can I use this regression with dummy variable...
Hi, and happy the video helped! Yes, indeed, you can use the dummy variable regression approach if you have multiple events, simply put 1 whenever you have an event and 0 otherwise (analogously with anticipation and adjustment).
Hi!, I would like to know please how would the regression equation for the market-adjusted model method look like and how would you write the hypothesis on a paper or thesis. It's urgent please!! Thanks a lot for your work, it's an amazing channel.
Hi Lev, and glad you are enjoying the channel! As for your question, you would write and equation like this:
Rit - Rmt = b0 + b1*ANTit+b2*EVENTit+b3*ADJit + eit,
where Rit is the return of the ith stock on day t; Rmt is the market return on day t; b0 is the constant; ANTit, EVENTit, and ADJit are the anticipation, event, and adjustment dummies respectively; b1, b2, and b3 are the average abnormal return estimators for the anticipation, event, and adjustment periods; and eit is the error term.
Hope this helps!
@@NEDLeducation Thanks a lot for your quick response!, in the case of the t-test and P-Values how will the notations be, which is the level of significance for t statistics and the degrees of freedom for the P value?
@@levaddler2477 Hi, and thanks for the follow-up question! You can easily apply the T.DIST.2T function and apply it to the absolute value of the t-stat. The degrees of freedom would be the number of observations minus two.
hey
how are you doing this with more than one company - for example with 50 companies?
Hello NEDL! I'm uncertain if you already explained this but what are the numbers we get out when we use the LINREGR formula. The top row under let us say the constant return model is the average abnormal return. Then u say the next row is the std. error but it ends there. I'm unsure what the rest of the numbers are referring to. The 4th row under anticipation has degrees of freedom. I want to use the data I have as a picture, but unfortunately, I cannot present numbers with no text associated with them. I learned a lot! Thank you!
Hi Nikolaj, and glad you liked the video! I have explained the LINEST (or LINREGR in other languages) output template in many videos, but long story short: in the first and second row, as you correctly stated, it reports the coefficients and respective standard errors. The third row reports the model r-squared on the left and the regression standard error (prediction standard error, standard deviation of the error term) on the right. The fourth row reports the F-stat (used to test for the significance of the overall model) on the left and the degrees of freedom on the right. The fifth row reports the explained and the residual squared sum (explained and unexplained variance) on the left and the right, respectively (so if you divide ESS by ESS+RSS, you get r-squared). Hope it helps!
@@NEDLeducation Thank you so much! Just what i needed :D
Hello sir, first of all accept my kind regards for clearly presenting the event study. I request you to kindly conduct a event study in stata with ‘eventstudy2’ command. Thanks in Advance
Hi Mahesh, and thanks so much for the feedback! There are many packages that can be used for event studies, I do not use stata much in my teaching/research, but potentially, I will investigate your suggestion sometime in the distant future.
Do you have any tips on how this could be applied for more than one event? I have 80+ events for my study
Hi Caio, and great question! The dummy variable regression is an excellent framework for this, if you believe the events are similar in nature and should have similar impact on the abnormal return, you can code the event, anticipation, and adjustment as shown here, simply inputting a 1 in the dummy variable column for each event date. If you have got multiple time series, they can be merged together in a panel and dummy variables coded in the same way. Hope it helps!
Hello NDL, I love your videos. I am currently performing an event study (CAPM) on the effect the announcement of trump that the US will leave the pairs climate accord has on the stock return for environmentally friendly and unfriendly companies. I have 441 companies for which degree of freedom do I have to incorporate to test the significance of the results?
Hi, and glad you like the videos! First of all, a very good research idea! As for your question, you can form a portfolio of such companies (equal- or value-weighted, for example), and test the significance of the event using the same approach. Alternatively, you can form a panel from your companies and code the event dummy variable for all of them. Hope it helps!
Could adjustment and event factors have different signs? And what would be interpreted in that case in relation to over or under-reaction in that case?
Hi Akansh, and thanks for the question! Yes, they can! If there are significant adjustment and event factors of different signs, you could interpret is as the market incorrectly anticipating the nature/the impact of the event. Hope it helps!
Hello sir, thank you for your session sir but I'm stuck with AAR, CAAR and T-statistics (market model(... Can you tell me sir how can I proceed to get above values
Hi mate, thanks so much for the helpful video, been a lot of help. A bit confused how you would lay this out in terms of a methodology and equations on a paper? as im confused what to include etc, any ideas?
Hi, and glad the video helped! As for the outline, you can represent event study regressions as:
R(t) = b0 + b1*Event(t) + b2*Anticipation(t) + b3*Adjustment(t) + e(t);
R(t) - M(t) = b0 + b1*Event(t) + b2*Anticipation(t) + b3*Adjustment(t) + e(t);
R(t) = b0 + b1*M(t) + b2*Event(t) + b3*Anticipation(t) + b4*Adjustment(t) + e(t)
for constant return, market-adjusted, and CAPM models, respectively, where R(t) is the stock return, Event(t), Anticipation(t), and Adjustment(t) are respective dummy variables, M(t) is the market return, and e(t) is the i.i.d. disturbance term.
In terms of classical references on event studies, see MacKinlay (1997) and Brown and Warner (1985). Hope it helps!
@@NEDLeducation perfect thank you so much.
Can you please do the regression for ex dividend dates using more than one company and years.
Hi Amber, and glad you liked the video! To implement event studies for multiple companies, you can use a quasi-portfolio approach, calculating average returns across a set of companies at a particular date respective to company-specific ex-dividend dates. Might do a video on quasi-portfolio approach at some point! Hope it helps!
@@NEDLeducation thank you. you are doing an amazing job..👍
Hi! Thanks a lot for this great video! super useful! I just don't have clear if you are estimating CAR or BHAR with this method. Could you please clarify this? Thanks a lot again!
Hi Luis, and thanks so much for the comment! Dummy variable regressions naturally estimate average abnormal returns (AARs), which are basically CARs per day. Hope it helps!
Hello sir, iam Alia from Indonesia. First of all thank you so much for your explanation video about event study. I wanna ask you about my research. I would like to do event study about lockdown announcement in Indonesia using firm characteristics as independent variable and CAR as dependent variable. In your opinion what is the best regression model for this event?
Hi Aliakito, and glad you liked the video! I feel the regression approach would be quite fitting for your research design. You can calculate cumulative abnormal returns for your sample firms using some model (constant return, market-adjusted, or CAPM) and then regress these onto firm-level characteristics. A simple multiple linear regression would be a logical first step. Hope it helps!
@@NEDLeducation Thank you so much for the answer sir 🙏🏻
Thx
Good Evening Sir,
Hope you are doing well. Sir can you please clear my query regarding event studies. Can we run regression in event study without taking dummy variable into consideration?
ARo = a+a1TC + a2 Dy + a3 Beta + a4 Av. Vol + a5Av.ab Vol + a6 Size
Can I run regression on this model in event study.
Hi Amber, and thanks for the question! The dummy variable approach is useful to calculate the average abnormal returns themselves. If you have already calculated these using another approach and want to determine what impacts their magnitude, the approach you suggested is valid. Hope it helps!
@@NEDLeducation Thanks alot.
Very informative. Thanks alot.
What would the multiple regressions look like?
Thanks.
Hi Alvaro, and thanks for the question. The regression equation would be:
1) stock return = a + b1*dummy(anticipation) + b2*dummy(event) + b3*dummy(adjustment) + error
2) stock return - market return = a + b1*dummy(anticipation) + b2*dummy(event) + b3*dummy(adjustment) + error
3) stock return = a + b*market return + b1*dummy(anticipation) + b2*dummy(event) + b3*dummy(adjustment) + error
For constant return, market-adjusted, and CAPM models, respectively, with b1, b2, and b3 estimating average abnormal returns for anticipation, event, and adjustment periods.
Hope it helps!
NEDL thanks NEDL, it really helped!! You just gained a new sub ;)