Multiple Regressions per Month following George & Hwang
Dear all,
i am currently struggling to fulfill my tasks in Stata. I have to reconstruct Table V of George and Hwangs Paper 'The 52-Weeks High and Momentum Investing' from 2004, published in the Journal of Finance (https://www.bauer.uh.edu/tgeorge/papers/gh4-paper.pdf).
The goal is to explain how the return of a stock can be attributed to the stock being a winner/loser of a certain strategy or in turn how the status of being a winner/loser is related to the return.
The following regression is used in the paper:
Rit = b0jt + b1jt*Ri,t-1 + b2jt*sizei,t-1 +b3jt*JHi,t-j + b4jt*JLi,t-j + b5jt*MHi,t-j + b6jt*MLi,t-j + b7jt*FHHi,t-j + b8jt*FHLi,t-j + eit
Where Rit means the Return of Stock i in Month t,
size equals the market capitalization and all other variables are dummies which indicate wether the specific security is included in the respective strategies winner-/loser-group (Strategies are Jegadeesh/Titman, Moskwitz/Grinblatt and the papers own 52-week high strategy). j can take values from 2 to 7 because the strategies are build on a 6 month observation phase.
The next step is where my problem is:
The authors now estimate 6 cross sectional regressions for each month (to account for every month in the past that is included in the observation phase, represented by the values of j) and average the coefficient estimates.
They then proceed to take the time series average of these averages but this should be not a problem for me.
How can i implement the regression in a way that the average of the past 6 months is stored in the line of the respective stock for each month?
The command i am using now is:
Please note that retL just is the one month lagged return.
I know that this is absolutely not correct but do not know how to implement what i need.
My dataset consists of monthly data of stocks of the whole CRSP Index. Each stock/line has a stock identifier, return, one month lagged return, date(month), size and the dummy variables mentioned above.
I am using Stata 15 on Mac OS.
I hope i made clear what my issue is about and that somebody can help me with this.
Hendrik Schillings
i am currently struggling to fulfill my tasks in Stata. I have to reconstruct Table V of George and Hwangs Paper 'The 52-Weeks High and Momentum Investing' from 2004, published in the Journal of Finance (https://www.bauer.uh.edu/tgeorge/papers/gh4-paper.pdf).
The goal is to explain how the return of a stock can be attributed to the stock being a winner/loser of a certain strategy or in turn how the status of being a winner/loser is related to the return.
The following regression is used in the paper:
Rit = b0jt + b1jt*Ri,t-1 + b2jt*sizei,t-1 +b3jt*JHi,t-j + b4jt*JLi,t-j + b5jt*MHi,t-j + b6jt*MLi,t-j + b7jt*FHHi,t-j + b8jt*FHLi,t-j + eit
Where Rit means the Return of Stock i in Month t,
size equals the market capitalization and all other variables are dummies which indicate wether the specific security is included in the respective strategies winner-/loser-group (Strategies are Jegadeesh/Titman, Moskwitz/Grinblatt and the papers own 52-week high strategy). j can take values from 2 to 7 because the strategies are build on a 6 month observation phase.
The next step is where my problem is:
The authors now estimate 6 cross sectional regressions for each month (to account for every month in the past that is included in the observation phase, represented by the values of j) and average the coefficient estimates.
They then proceed to take the time series average of these averages but this should be not a problem for me.
How can i implement the regression in a way that the average of the past 6 months is stored in the line of the respective stock for each month?
The command i am using now is:
Code:
reg ret retL mcap id_topJT id_lowJT id_topMG id_lowMG id_top52 id_low52
I know that this is absolutely not correct but do not know how to implement what i need.
My dataset consists of monthly data of stocks of the whole CRSP Index. Each stock/line has a stock identifier, return, one month lagged return, date(month), size and the dummy variables mentioned above.
I am using Stata 15 on Mac OS.
I hope i made clear what my issue is about and that somebody can help me with this.
Hendrik Schillings
Attached Files
