Fixed effect model vs Fama Macbeth


I have a question:
I have a set of Panel data, indexed by year and ID.
I first ran the FamaMacbeth regression:

FM<-pmg(y ~ x1+ x2 + x3+ x4 +x5 + x6, data, index=c("ID","Year"))

I first got a error message of: "Insufficient number of time periods".
Then I changed the order of the index index=c("Year","ID")), and obtained a result.

FM<-pmg(y ~ x1+ x2 + x3+ x4 +x5 + x6, data, index=c("Year","ID"))
I got a significant NEGATIVE value of x1 coefficient (which is my variable of interest).

I also want to include a fixed effect model. (Have conducted a Haussman test)
My problem is, which is the correct indexing?

When I run

fixed<-plm(y ~ x1+ x2 + x3+ x4 +x5 + x6, data=data, index=c("Year","ID"), model="within")

X1 coefficient is still NEGATIVE and significant.

BUT, if I change the order of the index:

fixed<-plm(y ~ x1+ x2 + x3+ x4 +x5 + x6, data=data, index=c("ID","Year"), model="within")

x1 coefficient is suddenly POSITIVE and significant (even at a higher significance level).

What is correct indexing, as I would like to compare my results to FM regression?

And if I have to use the index that gives me positive coefficient, how do I interpret the results as they differ from FM results?


This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.

If you have a query related to it or one of the replies, start a new topic and refer back with a link.