First Difference with Fixed Effects

[Barbora Sedova]

Dear all,

I have panel data with two periods and want to apply simple first difference approach. Since I only have data for two periods, this should provide the same results as a fixed effect model. I ran both versions and it was indeed the case.
When using the first difference, the way I understand it is that in this way I get of the observation specific fixed effects.

I am a bit puzzeled though, cause I would like to replicate what Burke and Emerick (2016) are doing in their model 10. However, in their FD model they apply state fixed effects. But if it is FD shouldnt the fixed effects already be controlled for anyways?
Please, advise. Thank you.


(Burke and Emerick (2016): Adaptation to Climate Change: Evidence from US Agriculture)

[Chinmay Sharma]

It would be, if the unit of observation was the state. If the unit of observation was the individual, then using First differences on the individual observations is not equivalent to using state fixed effects.

[William Lisowski]

Having searched out a copy of the paper in question, in model 10 the observations are at the county level, rather than the state level, so as post #2 suggests, first differencing is not equivalent to the fixed effect.