Panel Data dummy variable/structural Breaks/ Shocks

[Lukas Just]

Dear Community,

I want to analyze the effect of the Brexit on the economic growth on the UK.
For that, i created a panel dataset for the different industry groups (19) for the time period 2010q1-2022q4 (48 dates) including 5 different control variables (p.E: Total Assets (Mil.). Workforce (Thsd.), Productivity(Thsd.) , Innovation( Mil.), Delator(%).)
The dv is the gross value added from the specific industry group (GVA) (Mil.).
My goal is to see if there is a discernible effect of the brexit.

To my questions:
#1 after setting my panel set:
" xtset Industry1 time, quarterly
panel variable: Industry1 (strongly balanced)
time variable: time, 2010q1 to 2022 q4
delta: 1 quarter"
can i add a dummy variable "D1" ( 0= before brexit 1= after brexit) and run:
"xtreg GVA TA WF Prod Ino Defl D1, fe"
to see if there is a valid effect ?


#2 the second approach would be a structural break test:
In my research i only found the xtbreak command for panel data, because the " estat sbsingle/ estat sbknown" is only for time-series data.
So are there other opportunities to detect structural breaks or external shocks in stata?

I would be very grateful for any feedback

best regards
Lukas

[Carlo Lazzaro]

Lukas:
welcome to this forum.
I feel qualified (or hopefully so) to reply to your first query only:
1) yes, you can add a categorical variable related to Brexit. It is a good habit to code it as -i.D1- (see fvvarlist notation). Please also note to avoid codes the invoke the difference operator (see [U] 11.4.4 Time-series varlists);
2) you are dealing with a T>N panel dataset, so you should switch to -xtregar,fe- if interested in the -fe- estimator;
3) as usual, the -fe- estimator wipes out all the the time-invariant variables.

[JanDitzen]

To the best of my knowledge xtbreak is the only command which does structural break tests in panel data in Stata. There are some possibilities in R and there is Bai and Perron's code in Gauss and Matlab, which is for time series data though.

[Lukas Just]

Dear Carlo & Jan,
thank you for your helpfull advice!
Best regards!