A very common assumption of DiD is parallel pre-trend satisfaction. And when we subsample, we need to make sure the parallel trend assumption of the subsample also need to be statistically satisfied. One way to circumvent it is to use DDD (Difference-in-Differences-in-Differences).
For example, the setting is to examine the impact of anti-corruption laws on firms' asset growth in a staggered setting (laws being implemented differently among countries). Afterward, I want to examine the difference between the impact of laws on asset growth in developed and developing countries. If I subsample, I need to test the parallel trend of the developed and developing group- separately. However, if I use DDD, it would be easier that I just need to add the interaction into the regression. But I am wondering whether I need to test any further parallel trend assumption here ? I tested the parallel assumption already for my whole sample.
Related Posts with Do we need to test for any additional parallel trend when using triple difference?
query on time-series function for computing averageI am attempting to calculate average assets i.e. computed as average assets = (assets (t) + assets (…
Where is program_name stored?Dear All, I have a program as follows: program BSE_02052018 version 12 syntax, MAINdirectory(str) …
Graphic illustration of 3-way interaction between dummy, ordinal, and metric variableHi I am currently looking into how the election of Trump (main independent variable) affected satis…
Difference in Difference estimates with binary variablesDear all I am struggling in understanding and analysing my data with the diff function & I hope…
xtprobit with clustered standard errorsHi, I have panel data, with a binary dependent variable. The data is at the individual level. I want…
Subscribe to:
Post Comments (Atom)
0 Response to Do we need to test for any additional parallel trend when using triple difference?
Post a Comment