-1
$\begingroup$

Consider the following model:

$Y_{it} = \alpha + \alpha_1 T_{it} + \alpha_2 Post_{it} + \alpha_3 Post_{it} \times T_{it} + u_{it}$

where $i$ refers to individuals and $t=1,2$ to the first (pre treatment) or second (post treatment) period. Furthermore, $Post$ is an indicator variable that is 1 for a single post-treatment period, and $T$ is an indicator variable that is 1 if a specific observation is part of the control group.

Further assume that the measurement of my outcome variable changes between the two periods (for both, the treatment and control group). For example, because a reform also changes how the outcome variable is measured. I am wondering under which conditions I still can estimate a difference-in-difference model.

$\endgroup$
0
$\begingroup$

If the method of measurement changes, it cannot be considered to be the same series. You would have to go back in time and recalculate the pre-treatment series with the new formula and then carry out your analysis.

For an overview of event studies, given below is a useful paper

https://scholar.harvard.edu/borusyak/publications/revisiting-event-study-designs

$\endgroup$

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.