Period fixed effects and pooled cross sectional data

[Chiara De Siena]

Hi all,

I have a pooled cross sectional dataset from 2004 to 2018, and two categorial dependent variables: economic status (employed, unemployed, inactive) and contract type (temporary or permanent job).
I wish to see:

1) to what extent the relationship between age groups and economic status is different over time...
2) to what extent the relationship between age groups and contract type is different over time...

(3)...considering gender differences (not a simple control variable).

I created dummy variables for all the years, but including only T-1 time dummies. So my stata code is:

 bys gender: mlogit ecostat i.y2005##i.agec i.y2006##i.agec i.y2007##i.agec i.y2008##i.agec i.y2009##i.agec i.y2010##i.agec i.y2011##i.ageg i.y2012##i.agec i.y2013##i.agec i.y2014##i.agec i.y2015##i.agec i.y2016##i.agec i.y2017##i.agec i.y2018##i.ageg i.edu3 i.mregion i.hpart i.depch c.isei08


bys gender: logit tempj i.y2005##i.ageg i.y2006##i.ageg i.y2007##i.ageg i.y2008##i.ageg i.y2009##i.ageg i.y2010##i.ageg i.y2011##i.ageg i.y2012##i.ageg i.y2013##i.ageg i.y2014##i.ageg i.y2015##i.ageg i.y2016##i.ageg i.y2017##i.ageg i.y2018##i.ageg i.edu3 i.mregion i.hpart i.depch c.isei08 i.lastempj i.firms

My supervisor asked me "why are you using year instead of period fixed effects?"

So I am asking you: what is wrong with my syntax? Can you help me to understand what he's meaning? Because I understand his question as fixed effects (xtreg..., fe), but aren't fixed effect models only for pane data?

Thank you to anyone trying to help me out!

Bests,

Chiara

[Carlo Lazzaro]

Chiara.
is your dataset in -long- or -wide- format? If the llater were true, you should -reshape- to the -long- one, so that you can have one variable only for -i.years-.
It would also be advisable to share an example/excerpt of your data via -dataex-. Thanks.

[Chiara De Siena]

Carlo:
It is not a panel dataset, so does the wide/long format consideration apply also in the case of repeated cross sectional data?
Anyway, since data are confidential, I can only share a screenshoot of data browse (since I don't know how to provide a fake example of my data)

year - after 2018 there's 2017, 2016 and so on
hid - household id
pid - personal id

Can this be helpful?

[Carlo Lazzaro]

Chiara:
thanks for clarifying.
What I fail to get is why you're creating categorical variabes for -year- instead of using -i.year- notation, so that your interaction becomes:

i.year##i.age_c

[Chiara De Siena]

Hi Carlo,
I am creating a categorical variable with year because of my hypotheses.
Anyway, I finally understood what was asked me to do: aggregating some years together to create periods, and then interacting them with age groups as you just wrote. I just got confused because I associate "fixed effect" to panel data.

Thank you for you help, as always.

Kind regards,

Chiara