Differences in differences with matching

[Derya Zayim]

Hi,

I'm working with a survey data which covers 5 rounds (1995, 2000, 2005 , 2010 and 2015). However, the data is not a panel but cross-section. My pre-treatment period is 1995, 2000 and 2005 and post treatment is 2010 and 2015. I'd like to apply DID and matching techniques. Is it plausible to match the sample at the first place and then apply standard DID? If so, how can I match the observations in a cross section data. As far as I know, matching is performed using the pre-treatment characteristics of the treatment and control groups. But my data structure does not allow to track the same observations in the pre and post period? One option might be to match the observations in the pre-treatment and post-treatment period, separately? Or is it completely not accurate? Any help will be appreciated.

[Clyde Schechter]

The only matching approach that makes any sense to me would be actually doing a separate match in each of the five cross-sectional data sets.

Remember that even in relatively simple situations, matching is dicey. If there are large differences in the distribution of the matching variable(s) between the treatment and control populations, then you will find yourself with many observations that have no suitable match. If there aren't, then matching only modestly improve the efficiency of your analysis unless the matching variable itself is a major determinant of the outcome. You also have to be quite certain not to match on any attribute that might lie on the causal pathway from the treatment to the outcome. And you also have to be certain not to match or condition on any variable which is caused by both the outcome and the treatment.

I'm not saying that matching is never helpful. There are some circumstances where it can be quite effective. I'm just saying that it is, in my opinion, over-rated, and you really need to think carefully before doing it. If you are going to do it, though, I cannot see any sensible approach in your data other than doing a separate match in each round.

[Derya Zayim]

Thank you very much Clyde!