Hi everyone.
I want to conduct f-test for the equality of coefficients of firm-level determinants across countries and I came across a paper by de Jong et al. (2008) that has used the Seemingly Unrelated Regression estimation method. I find it difficult to implement their procedure in STATA. I am sharing it here if anybody can help. The procedure followed by de Jong et al. (2008) is quoted below..
"we make use of an unrestricted
regression model (where all coefficients are allowed to vary
across countries), and seven restricted models (e.g., for tangibility
null hypothesis, we restrict that the tangibility coefficients
are the same for all countries, but other coefficients
of business risk, firm size, etc. can vary)."
"using the Seemingly Unrelated
Regression (SUR) estimation method, we get SUR by
adding all sum-squared-residuals (SSR) from all the
equations for firm-specific determinants of leverage (as
specified in Eq. (1)). For SR in each test (still using
SUR), we add the SSR from the restricted equations in
the system with respective assumptions that the relevant
coefficients are the same across countries. The values of
f-statistic provide evidence whether to reject or not the
hypotheses."
Eq (1) in this regard means OLS regressions for each country in their data set.
I will appreciate if someone would put me through on how to conduct this test in STATA.
Thank you
Usman
I want to conduct f-test for the equality of coefficients of firm-level determinants across countries and I came across a paper by de Jong et al. (2008) that has used the Seemingly Unrelated Regression estimation method. I find it difficult to implement their procedure in STATA. I am sharing it here if anybody can help. The procedure followed by de Jong et al. (2008) is quoted below..
"we make use of an unrestricted
regression model (where all coefficients are allowed to vary
across countries), and seven restricted models (e.g., for tangibility
null hypothesis, we restrict that the tangibility coefficients
are the same for all countries, but other coefficients
of business risk, firm size, etc. can vary)."
"using the Seemingly Unrelated
Regression (SUR) estimation method, we get SUR by
adding all sum-squared-residuals (SSR) from all the
equations for firm-specific determinants of leverage (as
specified in Eq. (1)). For SR in each test (still using
SUR), we add the SSR from the restricted equations in
the system with respective assumptions that the relevant
coefficients are the same across countries. The values of
f-statistic provide evidence whether to reject or not the
hypotheses."
Eq (1) in this regard means OLS regressions for each country in their data set.
I will appreciate if someone would put me through on how to conduct this test in STATA.
Thank you
Usman
