Long-Horizon Regression

[TDeval]

Sorry to bother you once again and hopefully this should be the last question I will need to ask you in order to complete all of my long horizon regressions.

I would like to use one-month returns as my dependent variable and lagged multiperiod returns as the independent variable.

As previously mentioned I am quite an EViews new user and so if you could please give me an example of the code that could be use to conduct this regression, that would be amazing. Or steps of how to go about this.

This is the same tyoe of regression as used in the Jegadeesh (1991) paper if you are familiar with that.

I have a series that contains 575 monthly stock price observations.

Thank you in advance
Tom

[startz]

Ignoring EViews for the moment, given prices how would you define the one-month return? How would you define a multiperiod return?

[TDeval]

My apologies for being so vague. By one-month return as my dependent variable I was referring to using returns of only say January (month01) and by multiperiod returns as my independent variable I was referring to using aggregated returns in all months in the aggregation interval e.g. intervals ranging from 1-10years.
Hopefully, this will allow me to examine the seasonal pattern in long-term return reversals, and specifically the extent to which temporary components in stock prices decay in the month of January. Therefore, I would also need the corresponding non-january regression for the same sample period.

I hope that this has cleared up what I mean't in my previous post.
Many thanks for your time once again
Tom

[TDeval]

oh and sorry my definition of return would be (logPt - logPt-T) T being the time period (horizon)

[startz]

oh and sorry my definition of return would be (logPt - logPt-T) T being the time period (horizon)

Something like

Code: Select all

ls (log(P) - log(P(-1)) c (log(P(-1)) - log(P(-13)) (log(P(-13)) - log(P(-25))