Common Trends Question

[Juan Tomas Rago]

Fellows!

I'm sorry for what is a probably basic question but I'm really needing help.

I'm currently doing my Bachelors thesis and I'm taking a diff-in-diff approach, It's all adding up, but i need to asses the Common Trends Assumption in a way other than graphically.

My data is structured by county, quantity of congregations (which is what i need to prove are parallel in time), population and then dummies for treatment areas and treatment moment. I have data for 1952, 1971, 1980, 1990, 2000 and 2010, treatment occurs between 2000 and 2010.

When I graph the avg quantity of congregations between the treated and control it's clear that they're parallel, but I need to be able to confirm otherwise.

Any help?

Thanks in advance!!

[Clyde Schechter]

Something like this:

regress outcome_variable i.treatment##c.year if year < 2000 // POSSIBLY SOME OPTIONS

The coefficient of 1.treatment#year in the output estimates the difference in slopes between the treatment and control groups' outcome trajectories (assuming the trends are more or less linear over time), and if the parallel trends is true, this coefficient will be close to zero.

[Juan Tomas Rago]

Hello Clyde!

Thanks for the fast response!

I got the following coef: .0936971

Would you say that's about enough to asses parallelism? also the P-value is 0.086

[Clyde Schechter]

Please show the complete output. It isn't really possible to try to decide what "close to zero" means without the full context. And the p-value isn't helpful at all.

[Juan Tomas Rago]

This is the full thing: