fixed effects and dummy variable

[Rebecca Nnamani]

hi all,
i am trying to use 'xtreg avginflation cbi exrate i.dum, fe' but the dummy variable keeps getting omitted because of collinearity. is there anything i can do to fix this? please i a am not a pro at stata and i am just looking for advice/help. thanks!

[George Ford]

Mundlak regression might work.

ssc install xthybrid

See Jeff Wooldridge's work on this.

[Rebecca Nnamani]

Mundlak regression might work.

ssc install xthybrid

See Jeff Wooldridge's work on this.

hi thank you for the help but i'm afriad it is still not working. i have no idea what's wrong

[Carlo Lazzaro]

Rebecca:
one of the most valuable FAQ recommendations is to post what you typed and what Stata gave you back. Thanks.

[George Ford]

What does i.dum measure and what is your FE?

[Rebecca Nnamani]

Rebecca:
one of the most valuable FAQ recommendations is to post what you typed and what Stata gave you back. Thanks.

thank you for this. i will make sure to do this when asking for help!

[Rebecca Nnamani]

What does i.dum measure and what is your FE?

i.dum is the level of development for the 8 countries (0 for developed and 1 for developing) and for xtset i used the countries and years

[George Ford]

which variable is omitted in estimation and what does it measure?

[Andrew Musau]

i.dum is the level of development for the 8 countries (0 for developed and 1 for developing) and for xtset i used the countries and years

Unless a country transitions from developing to developed during the sample period or vice versa, the coefficient on the variable will be collinear with the country effects, rendering it unidentifiable. If your intention is to account for the developmental differences between countries, the country effects already do this, making it unnecessary to include this variable in the model. However, if you aim to examine the impact of development on your outcome, consider incorporating income per capita. This variable varies across countries and over years, largely signifying a country's developed or developing status. Income per capita data is readily available from sources like the World Bank and the Penn World Tables.

[Rebecca Nnamani]

which variable is omitted in estimation and what does it measure?

the dummy variable for developing countries: i.dum

[Rebecca Nnamani]

Unless a country transitions from developing to developed during the sample period or vice versa, the coefficient on the variable will be collinear with the country effects, rendering it unidentifiable. If your intention is to account for the developmental differences between countries, the country effects already do this, making it unnecessary to include this variable in the model. However, if you aim to examine the impact of development on your outcome, consider incorporating income per capita. This variable varies across countries and over years, largely signifying a country's developed or developing status. Income per capita data is readily available from sources like the World Bank and the Penn World Tables.

omg thank you so so much it worked after i used xtset year! and just to check it's alright for me to use year as the xtset?

[George Ford]

I don't think that's correct. xtset year is going to give you a year fixed effect, not a country one.

I hink you need

xtset country

but do not include i.dum as a regressor.

[Rebecca Nnamani]

I don't think that's correct. xtset year is going to give you a year fixed effect, not a country one.

I hink you need

xtset country

but do not include i.dum as a regressor.

ah ok thank you. do you have any solutions to this as i am back to square one?

[Carlo Lazzaro]

Rebecca:
if your -i.dum- is collinear with your -panelid-, the -fe- estimator will omit it, as it is time-invariant.
Please note that:
a) the -fe- estimator wipes out all the time-invariant variables;
b) the -fe- estimator uses the within-panel demeaning approach. As the mean of a constant is the constants itself, the related coefficient cannot be estimated.

[George Ford]

xtset country year

** simple
xtreg avginflation cbi exrate , fe

** maybe
xtreg avginflation cbi exrate i.year, fe

** alternate
reghdfe avginflation cbi exrate, absorb(dum) 
reghdfe avginflation cbi exrate, absorb(dum year)