Comparing financial results of 4 countries.

[Abid Jahangir]

I a doing a comparative research study which involves a panel data set of 632 firms listed in four different countries, for a period of 12 years. One of my objectives is to see if the relationship between dependent variable and the independent variables in the sample varies from one country to another. Although I can run a regression in stata, either individually or with countryid dummies to know the differences in the relationship. However, I think I must also test the statistical significance of the country specific differences. what option do I have?

[George Ford]

If you have a fully intereactive model (i.country#c.X), then the coefficients measure the difference between the i.country and the base country. The t-statistic is a test of the difference.

Can also you margins to get individual country effects. Look also at contrast.

[Abid Jahangir]

Thank You Mr. George Ford for your valuable suggestions. I believe you meant that if I just run a regression with countryID dummies the coefficients themselves will provide a measure of differences in the relationship between dependent and independent variables and performing a separate test of statistical significance may not be necessary.

[George Ford]

Think of rep78 as being a country id.

There are many possible variants (could absorb the country fixed effect is not interested in the values). But this gives you the coef and t-stat on the difference in the constant and slope coefficients across "countries".

sysuse auto, clear
reg mpg price weight i.rep78 b1.rep78#(c.price c.weight)