Socio demographic controls

[Finn Sykes]

Hello,

I am regressing conservative party contact on the likelihood to vote conservative and using individual level panel data. I'm starting with pooled OLS and building up to individual fixed effects and time fixed effects. I am controlling for a bunch of socio demographic characteristics in each regression: my question is are some of these too similar to each other and therefore do I need to pick one, for example I have:

working status
income
household income
social grade
education level

I also control for:
country
constituency

does it matter that some of these controls are obviously very similar and likely correlated with each other? I understand some of them may be kind of useless my question is - is it a problem having them all in there?

[Carlo Lazzaro]

Finn:
welcome to this forum.
Please follow the FAQ to post more effectively in the future. Thanks.
That said:
1) you should rely on the literature in your research field to spot the controls that are more frequently used;
2) technically speaking, if two variables tells the very same things, omission due to colinearity will take control;
3) why starting with a pooled OLS if (I guess) you're dealing with a panel dataset?

[Finn Sykes]

thank you Carlo,

sorry will do that in the future,

I'll have look at the literature. Point 2 is helpful thank you. Starting with pooled OLS was just to build up my model its not for any real inference.

kind regards,

Finn