Estimating unobserved heterogeneity

[@helenpan5533]

Hi Ben, I don't quite agree with your point on the inconsistency of FE estimates. Actually I think strict exogeneity guarantees consistency. You said in the video that the variance of estimates goes to 0 if T goes to infinity but doesn't if N goes to infinity. I don't agree with this point. Because no matter whether it's N or T that goes to infinity, we are essentially running a pooled OLS after the data are demeaned. So regardless whether the extra data points are from the time dimension or city dimension, we get more data points for the regression. I think it certainly will reduce the variance of the estimates and by asymptotic properties of OLS estimates we will have consistency if N goes to 0.

[@lastua8562]

What would you say if you answered this question now after 6 years?

[@favprince]

Thank you for all the videos.

[@sharad5149]

Hi Ben, Why did you not drop the error term uit when you wrote down your regression equation with the dummy variables using the LSDV method?

[@lastua8562]

At 4:20, why exactly can we estimate our alpha like this? I am struggling to derive it.

[@angelgadon6132]

Hello sir Ben, Thank you for the videos. I would like to know if it is okay to include education or other demographics as an explanatory variables (like crime and unemployment) in a panel data model. Or should we not include it and let the alpha i explains it. I am a bit confused. Hoping for your respond. Once again, thank you for sharing your knowledge. 😌

[@lastua8562]

You can add whatever you like, as long as it is relevant and you have sufficient data.

[@misi7259]

Is this the last video of panel data? which one is next?

[@sissilamia3703]

plz how can we calculate the variance of alpha and the variance of erreur term.. can you make us understood by a numerical example..thank you