I have a panel data set covering various countries over various years. I have information about the years in which these countries published an entrepreneurship policy - some countries only published one in the past 20 years, while others published a new one every five or ten years. I also have an annual estimate of the level of entrepreneurial activity for each country.

I'd like to evaluate whether the launch of these policy documents resulted in an increase of entrepreneurial activity in these countries. However, I am not sure how to approach this question methodologically, since the policies were introduced in different years and sometimes countries introduced various policies over the time that my data set covers.

I don't think I can use the typical methods for evaluating the impact of treatment (such as difference-in-difference and instrumental variables regression), since the introduction of a treatment, which in my case is the introduction of policies take place across different years and sometimes there are several "treatments" for individual countries.

I would greatly appreciate your help with this question, as I am stuck not having found guidance searching through my books and the existing questions on Cross Validated.


Just a partial answer, but maybe it'll help.

You talk about several "treatments" for a single country. One approach to deal with this phenomenon is a Difference-in-difference-in-differences estimator. There is also a good description of this estimator in Woolridge's Econometric Analysis of Cross Section and Pannel Data .

Regarding the different points of time of the changes it might be helpful to standardise the data giving t=x for the year of the impact in the measure in a specific country.

  • $\begingroup$ Dear @Ferdi - thanks a lot for your quick answer. I am sorry for formulating the question in a way that led you to misinterpret it (just fixed the description). I don't have only one country that introduced various policies over this time period, but several of them. Eg. South Korea introduced one in 2004 and then another in 2010, and Belgium in 2002, 2009, and 2015 for instance. Some of the countries only ever introduced one policy. $\endgroup$ – Sannita Sep 29 '16 at 17:09
  • $\begingroup$ Also @Ferdi - by giving t=x, what exactly do you mean? $\endgroup$ – Sannita Sep 29 '16 at 17:14
  • $\begingroup$ You do not necessarily have to look at the difference between 2004 and 2005, but you could look at the difference between the period of introducing a policy and the period before(e.g. 2003 and 2004) for South Korea or 2001 and 2002 for Belgium) $\endgroup$ – Ferdi Sep 29 '16 at 17:18
  • $\begingroup$ Thanks for the clarification @Ferdi. How would I analyze the difference between before a policy was implemented and after its implementation, given that the implementation varies between countries in my data set. Or do you mean that I would analyze each country separately from the others? $\endgroup$ – Sannita Sep 29 '16 at 17:29
  • $\begingroup$ It does not make sense to analyze each country separately from each other. You will not have a control group. If you analyze all countries together you have got a treatment and a sufficiently large control group. $\endgroup$ – Ferdi Sep 30 '16 at 6:12

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.