OLS or difference-in-differences for evaluating the impact of training on investment

[Guest]

Hi all,

I am examining the impact of training on the level of investment by microfinance clients. There are two time period - 2015 and 2016. Training was randomly given to one group in the village that meets on Tuesday while the other groups didn't receive any treatment. Since there is no significant difference in the control and treatment sample in the baseline, I am treating it as a randomised experiment. However, if I use diff-in-diff to check the robustness, its estimates are twice the estimate of OLS of the sample of 2016 data.

Just wondered what is the best approach to this? Should i use OLS or diff-in-diff for the analysis?

Or if i use diff-in-diff for robustness check, interpret the sign and not the values. For instance, training has a negative effect on the level of investment.

Thanks in advance for your time.

Kind Regards,
Guest
Researcher

[Clyde Schechter]

Your question is unclear because you are talking about two things as if they were in the same category when they are not. OLS is an estimation technique, DID is a model. You can estimate a DID model using OLS or using other approaches. And OLS can be used for lots of different models in addition to DID. So it isn't at all clear what your two analyses were. I suggest you post the actual commands you used and the actual output you got from Stata for better advice.

[Steve Samuels]

To ensure optimum readability, please post the commands and output between CODE delimiters, described in FAQ 12