Time series means of coefficient estimates from cross-sectional regressions.

[Hakan Gunduz]

Hi all,

I have 29 annual cross-sectional regressions that I want to summarise the findings in a single table. I have been following a paper doing the same but I don't understand how exactly they calculate the significance levels.

The paper says that "the table presents the time series mean of the coefficient estimates from the 30 yearly regressions. The significance levels are for a test of the hypothesis that the time series mean is equal to zero , using the time series standard error of the mean estimate for each coefficient."

If I understand it correctly to calculate the coefficient they simply take the average of the coefficients from yearly regressions. But how exactly they calculate the significance levels based on standard errors? Is it basically a ttest of standard errors from yearly regressions?

Many thanks!

[Phil Bromiley]

If you use sureg or suest, then you may be able to test for the mean of variables across the regressions which will give you a standard error on the mean. The alternative is to interact i.year with all the rhs variables so you get separate parameters by year but all the parameters appear in the same regression. then again, you can test for the sum or mean of the variables.