Dear researchers,

This is not a question regading Stata, but rather a more general one. I have seen the forum being the most helpful in the past regarding such questions too, so I hope it is okay that I post my question here. I am currently reading the following paper and I have two questions regarding their methodology:

The study in question is: Babenko, I., Bennett, B., Bizjak, J. M., Coles, J. L., and Sandvik, J. (2019): Clawback Provisions and Firm Risk. Charles A. Dice Center Working Paper No. 2019-13. and can be found here: http://dx.doi.org/10.2139/ssrn.3382498


In this study the researches use a modified diff-in-diff approach that allows for multiple treatment periods to asses the effect of a policy change on investment level. As a dependent variable the authors use the annual change in R&D expenditures relative to assets i.e., Array

My first question would be why do they use a relative change score instead of repeated measures. The authors do not provide a reason and I can't think of an economic reason. Other researchers in the same field haven't done this either. Is there a statistical reason that would justify doing this?

My second question is concerned with the interpretation of the coefficient, when using the above described dependent variable. The statistically significant coefficient on the treatment dummy is - 0.004. Would this implicate that the annual change in the share of R&D in assets decreases by on average 0.4 percentage points, if treated?

Thanks,

Maria