# Subsampling result explanation?

I want to examine whether the impact of laws on asset growth difference between developed and developing countries.

One way to do so is by adding the interaction between the variable of interest and a dummy variable representing for developed countries

Dependent_variables= pt + developed_dummy*pt + Independent_variables + fixed effects + error term

where developed_dummy equalling to 1 if this observation is in developed country. The purpose is to see whether the pt effect is stronger in the developed countries, is not it? And with the result below, how could I conclude? Can I compare the law effect between the developed and developing countries or Can I compare the law effect between the developed and the whole sample?

where developedpt=developed_dummy*pt

And what should I conclude if the coefficient of developedpt is significant?

And with the result below, how could I conclude?

From the output above you can conclude that those laws do not affect the dependent variable in developed countries differently (of course, assuming there are no other issues or methodological problems not mentioned above).

Can I compare the law effect between the developed and developing countries or Can I compare the law effect between the developed and the whole sample?

If you want to make some alternative estimations on subsamples you should compare subsample of developing countries with subsample of developed country not just whole sample, since whole sample will include the developed countries.

However, subsampling and running separate regressions will cause you to loose power, not only sample will be smaller but on each subsample you need to use the same number of parameters (-1 for the interaction dummy).

Unless you have some specific reason to subsample (e.g. you have long panel an you believe the error structure is different between developed countries so you need special specification for either subsample) you should not do subsampling.

And what should I conclude if the coefficient of developedpt is significant?

If we would pretend the coefficient would be significant, you could conclude that the effect of the laws in developed countries is higher by $$0.00914$$ relative to non-developed countries. So the total effect of this laws on the dependent variable in developed countries will be: $$-0.0115+0.00914= -0.00236$$.

• (1)Thanks @1muflon1, "Unless you have some specific reason to subsample (e.g. you have long panel and you believe the error structure is different between developed countries so you need special specification for either subsample) you should not do subsampling.". Do you mean that if I do not have a special reason, I should not run the equation "Dependent_variables= pt +Independent_variables + fixed effects + error term" for the developing sample and developed sample separately and then compare them together? Jul 22, 2021 at 5:07
• (2)"those laws do not affect the dependent variable in developed countries differently". So differently compared to "developing countries" not the "whole sample" right? Jul 22, 2021 at 5:08
• @BeautifulMindset 1. yes you should not do it separately unless there is a reason for it. There is no advantage of doing that separately and there are disadvantage to doing that unless there are reasons other than you want to see how effects differ. 2. Yes differently compared to developing countries, if you care about different effects between groups you don’t compare that effect to whole sample but to other groups whole sample already includes everyone
– 1muflon1
Jul 22, 2021 at 12:26
• Hi @1muflon1, -0.115 as above is the impact of laws on the whole sample rather than non-developed sample right? Aug 8, 2021 at 23:00