2-period difference panel regression

[Kavorka]

(please ask for my licence codes, and I will send them to you)

This should hopefully be 4 rather standard questions for you, but I have not figured them out so I would like to have some help.

I am having some computer assignments for my students.

I have a data set based on Stock and Watson’s book Intro. to econometrics (based on Stata, but I think EViews usually is more user-friendly).

I would like to replicate the result for my students (please see the attchment fatality.wf1 which is freely accessible from the website).
First I run a cross-section for 1988
smpl 1988 1988
ls MRALL c beertax

and then for 1982
smpl 1982 1982
ls MRALL c beertax
Then I do this using a FEM between 1982-1988, in order take into account for some omitted variables which do not change over time.

However, in the book (Stock and Watson page 355, eq.( 10.8 ) they also estimate a regression for 2 period Difference panel regression (only with 1988 and 1982). Difference regression (n = 48)

I tried Ls (MRALL(1988)-MRALL(1982)) c (beertax(1988)-beertax(1982)) but it is wrong... but you can see which model I want to estimate (therefore I write this out)

Q1---Therefore, instead I would like to know how to get the following result (for a 2-period panel difference regression)
The estimated result should be as below

MRALL(1988)-MRALL(1982) = –.072 – 1.04(BeerTax(1988)–BeerTax(1982))

(all cross-sections, US states, are included)

(N.B. it fatalities per 10 000 inhabitants)

Q2--- I guess I can change the sample so that we only have 1982 together with 1988 (all cross-sections, US states), like a 2-period Difference panel regression (did not figure that out, but it would be good to know how to do this not very elegant solution anyway).

Q3--- In another part of the question they only use 3 cross sections (only 3 cross-sections, US states: CA, TX, MA, but with all years 1982-1988). How can I change so that only these 3 states are taken into account together with all years between 1982-1988)? (Panel)

Q4--- With rather awkard programming I solved how to Entity-demean the panel (WG, within-group estimator), but is there a canned routine in EViews to solve this?

Thanks!

[trubador]

For Q1 and Q2 the following trick should work:

Code: Select all

smpl @all if @year = 1988 equation eq.ls mrall-mrall(-6) c beertax-beertax(-6)

By the same token, you can confine your analysis to the states of your interest. I am not sure about the corresponding numbers of states, but you can do something along the following lines:

Code: Select all

smpl @all if state=1 or state=8 or state=13

As for your last question, I think you are looking for @meansby function:

Code: Select all

series mrall_dm = mrall - @meansby(mrall,@crossid,"1982 1988")

[EViews Glenn]

On the last question, if you want to run the regression, use the "cx=f" option to ls

[Kavorka]

Excellent help! Thanks!