Hi everyone,
I have a question about SUR estimation with unbalanced panels and cross-equation linear constraints.
I need to run a SUR regression with 40 equations corresponding to 40 different countries, including 9 regressors and covering a period of 180 months (1995-2009). My panel is quite unbalanced with some consecutive observations missing at the start of the sample for some countries and single observations missing in between. The data is currently arranged in wide format. In addition, I need to constrain the coefficients of the regressors to be the same across the 40 countries, yielding 39 x 9 constraints.
Y1 = b0 + b1*X11 + … + b9*X91 + e1
Y2 = b0 + b1*X12 + … + b9*X92 + e2
.
.
.
Y40 = b0 + b1*X140 + … + b9*X940 + e40
where the coefficients b1 to b9 are restricted to be the same across equations.
Sureg allows me to estimate the SUR model with a set of linear constraints but excludes months with missing values. Suregub by Kit Baum works for unbalanced panels but does not support the constraint() option. Xtsur includes random effects which do not apply in my case.
I also tried the approach recommended by Allen McDowell (2004) “From the help desk: Seemingly unrelated regression with unbalanced equations“. However, this did not work either since I cannot impose the constraints with xtgee.
Is there any other way to use all available observations in a SUR regression while imposing linear constraints? I appreciate any advice.
Best,
Tatjana
I have a question about SUR estimation with unbalanced panels and cross-equation linear constraints.
I need to run a SUR regression with 40 equations corresponding to 40 different countries, including 9 regressors and covering a period of 180 months (1995-2009). My panel is quite unbalanced with some consecutive observations missing at the start of the sample for some countries and single observations missing in between. The data is currently arranged in wide format. In addition, I need to constrain the coefficients of the regressors to be the same across the 40 countries, yielding 39 x 9 constraints.
Y1 = b0 + b1*X11 + … + b9*X91 + e1
Y2 = b0 + b1*X12 + … + b9*X92 + e2
.
.
.
Y40 = b0 + b1*X140 + … + b9*X940 + e40
where the coefficients b1 to b9 are restricted to be the same across equations.
Sureg allows me to estimate the SUR model with a set of linear constraints but excludes months with missing values. Suregub by Kit Baum works for unbalanced panels but does not support the constraint() option. Xtsur includes random effects which do not apply in my case.
I also tried the approach recommended by Allen McDowell (2004) “From the help desk: Seemingly unrelated regression with unbalanced equations“. However, this did not work either since I cannot impose the constraints with xtgee.
Is there any other way to use all available observations in a SUR regression while imposing linear constraints? I appreciate any advice.
Best,
Tatjana
Comment