Fixed Effects at individual-area-level?

[Andreas Baltin]

Hello folks!

I have read this paper today: http://ftp.iza.org/dp9311.pdf

If you scroll to page 11, they say with regards to their methodology: We use linear regression with fixed-effects at the individual-area-level to model the direct effect of violent crime victimisation on the wellbeing of individual i living in area a in period t.

What does this mean/how does it work? They use FE in panel data, but have fixed effects on individuals, areas, and time? How does this work.
Background: I am currently running a panel data analysis and run FE with fixed effects on individuals between the time periods. How do I include area effects? I do have variables indicating the town the individual lived in at time t=0 and t=1, but I am not doing anything with these variables as of now. This variable is also numerical.
Do I include a dummy for every single city (ommitting one)?

I hope this is not a really stupid question and am greatful for any input!

Andreas

[Carlo Lazzaro]

Andreas:
you can create a categorical variables for -city- via -fvvarlist- notation.
However, since you're using a -fe- specification, please note that time-invariant predictors will be wiped out by the -fe- machinery (put differently: if all the -panelid- do not move to another city, no coefficient will be estimated).

[Phil Bromiley]

Note that it is much better to xtset your data and use xtreg or reghdfe rather than messing with all the dummies.

I'm not sure what level and area mean. You only need area and level as separate controls if individuals switch areas or levels - otherwise the individual fixed or random effect takes care of areas and levels.

The problem may be simpler than you think. As I read the paper (briefly), it says "The inclusion of an individual-area fixed-effect means that an individual living in area A is treated separately from the same individual living in area B." This sounds like it is including fe at the area level and for time . A fe at the individual level would treat every individual differently (i.e., with separate panel effect). But I could have this wrong.