I'm quite new to Stata and econometrics in general. I have taken two undergraduate courses in econometrics and I'll now do my first real analysis and have some questions. All comments are highly appreciated!
I'll use a Diff-in-Diff to estimate an excise tax on certain goods, the tax is based on the weight of the underlying good (in an attempt to lower overall consumption). Goods with lower weight will get taxed less, but this also means that two goods with the same pre-tax price will get taxed differently.
I want to estimate the pass through of the tax, how much is passed on to consumers, how much more will consumers pay in taxes on certain goods, etc. I want to estimate both the introduction + the repeal of the tax, and possible compare any symmetric differences. I've already visually verified the parallel trend assumption.
I have data over 10 years with monthly average prices and weights of a couple of different types of products/ goods. The tax is introduced in year 3, raised slightly in year 4, and repealed in year 9. I'm thinking about using the standard command:
reg log(price) after treat diff
where log(price) is monthly average log price, after is a dummy = 1 if the observation is after the introduction of the tax, treat is a dummy = 1 for goods subject to the tax, and diff is the interaction variable (after*treat).
Since I have different types of goods with different types of weights/volumes, I'm wondering if I should regress one treatment product at a time, or every product subject to the tax together? Pros/cons?
I'm also wondering if it matters if I use all observations from every time period when I study the introduction/repeal of the tax? For example, the tax is introduced in year 3, and repealed in year 9, should I use all 10 years or maybe year 1-5 when I estimate the introduction and year 6-10 on the repeal? Does this matter?
Thirdly, I have thought about time fixed effects as well, would the estimation be significantly better with a monthly dummy variable etc? There will be a lot of dummies over a 10 year period then, pros/cons ?
Those are some practical questions I can't seem to find in the schoolbooks etc, I guess you learn by doing over time. All comments are highly appreciated! Thanks in advance.
Sincerely,
Joakim
I'll use a Diff-in-Diff to estimate an excise tax on certain goods, the tax is based on the weight of the underlying good (in an attempt to lower overall consumption). Goods with lower weight will get taxed less, but this also means that two goods with the same pre-tax price will get taxed differently.
I want to estimate the pass through of the tax, how much is passed on to consumers, how much more will consumers pay in taxes on certain goods, etc. I want to estimate both the introduction + the repeal of the tax, and possible compare any symmetric differences. I've already visually verified the parallel trend assumption.
I have data over 10 years with monthly average prices and weights of a couple of different types of products/ goods. The tax is introduced in year 3, raised slightly in year 4, and repealed in year 9. I'm thinking about using the standard command:
reg log(price) after treat diff
where log(price) is monthly average log price, after is a dummy = 1 if the observation is after the introduction of the tax, treat is a dummy = 1 for goods subject to the tax, and diff is the interaction variable (after*treat).
Since I have different types of goods with different types of weights/volumes, I'm wondering if I should regress one treatment product at a time, or every product subject to the tax together? Pros/cons?
I'm also wondering if it matters if I use all observations from every time period when I study the introduction/repeal of the tax? For example, the tax is introduced in year 3, and repealed in year 9, should I use all 10 years or maybe year 1-5 when I estimate the introduction and year 6-10 on the repeal? Does this matter?
Thirdly, I have thought about time fixed effects as well, would the estimation be significantly better with a monthly dummy variable etc? There will be a lot of dummies over a 10 year period then, pros/cons ?
Those are some practical questions I can't seem to find in the schoolbooks etc, I guess you learn by doing over time. All comments are highly appreciated! Thanks in advance.
Sincerely,
Joakim
Comment