Different types of fixed effects? relations?

[Shi Y. Yan]

Can anyone help with following questions? Thanks.

With panel data regression, we can control for unobserved factors with different fixed effect terms, for example

SALESit = a + bXit + ujt + mt +ni + error term

i is country; t is year; j is continent, such as Asia or Europe; SALES is the sales of cars; X is interdependent variable, such as s certain type of tax revenue (tax rate changes every year but different tax rate for different country).

Here are three fixed effect:
1. individual fixed effect, ni
2. time dummy (year), mt
3. continent-year fixed effect, ujt

Questions:

1.what is the relationship between the third fixed effect and the first two effects?

2.Any correlation problems or which part of variation is absorbed? ......

3. Is there any problems when all three fixed effects are used?

[Andrew Musau]

I would be interested in seeing how someone estimates the above model. Do you have a reference (e.g., a paper or book)?

Presumably, when you xtset your data, you have a cross-sectional identifier (in your case country), and a time variable (year). Accounting for country and time fixed effects amounts to including T-1 time dummies (assuming T periods), and N-1 country dummies (assuming N countries). Then, the only way to account for what you call continent-year fixed effect is to include continent dummies for all the years.

But countries are within continents (so continent dummies are not necessary in the presence of country dummies).

[Shi Y. Yan]

I would be interested in seeing how someone estimates the above model. Do you have a reference (e.g., a paper or book)?

Presumably, when you xtset your data, you have a cross-sectional identifier (in your case country), and a time variable (year). Accounting for country and time fixed effects amounts to including T-1 time dummies (assuming T periods), and N-1 country dummies (assuming N countries). Then, the only way to account for what you call continent-year fixed effect is to include continent dummies for all the years.

But countries are within continents (so continent dummies are not necessary in the presence of country dummies).

above is a example to make my question clear and simple.

The original regression can be found in paper "using vehicle taxes to reduce carbon dioxide emissions rates of new passanger vehicle: evidence from france, germany and sweden"

[Carlo Lazzaro]

NHH Stijn:
please note the strong preference for full real names (FAQ 6 http://www.statalist.org/forums/help) and full reference (FAQ 13 http://www.statalist.org/forums/help). Thanks.
That said, where is the example you mentioned in your last post?

[Andrew Musau]

NHH: You misrepresent the model by Klier and Linn
Using Taxes to Reduce Carbon Dioxide Emissions Rates of New Passenger Vehicles: Evidence from France, Germany, and Sweden (AEJ: Economic Policy, 2015)

I have briefly looked at an earlier working paper (because I do not have a subscription to AEA journals), and what they have as a "third" fixed effect is just an interaction of model related variables and year. Generally, this should be included in the list of regressors and it appears that they do so in later revisions of the paper (see page 17 of a presentation of the paper).

https://www.google.no/url?sa=t&rct=j...s80QFBZO3_UEaA

Their data and do files are available at AEAweb free to download. I have not looked at these but you should if you are thinking of doing something similar or challenging their work.

https://www.aeaweb.org/articles.php?...7/pol.20120256

[Shi Y. Yan]

NHH: You misrepresent the model by Klier and Linn
Using Taxes to Reduce Carbon Dioxide Emissions Rates of New Passenger Vehicles: Evidence from France, Germany, and Sweden (AEJ: Economic Policy, 2015)

I have briefly looked at an earlier working paper (because I do not have a subscription to AEA journals), and what they have as a "third" fixed effect is just an interaction of model related variables and year. Generally, this should be included in the list of regressors and it appears that they do so in later revisions of the paper (see page 17 of a presentation of the paper).

https://www.google.no/url?sa=t&rct=j...s80QFBZO3_UEaA

Their data and do files are available at AEAweb free to download. I have not looked at these but you should if you are thinking of doing something similar or challenging their work.

https://www.aeaweb.org/articles.php?...7/pol.20120256

Thanks. It helps.