Why is a squared term in panel data fixed effects considered redundant

[Julianna Grato]

I have the following equation: xtreg lwage educ black Hispanic experience expersq married union, fe
Why is experience^2 considered to be redundant. I'm confused since experience^2 changes over time?

[Joseph Coveney]

Collinearity: it might be linearly dependent with another predictor, for example, the linear term for experience.

[Julianna Grato]

Thank you very much Joseph.

[Phil Bromiley]

You'll increase our ability to provide a useful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output, and sample data using dataex (which you do).

But, in general, the squared term is not redundant to the linear in fe models. There are piles of studies using squared and linear terms in xtreg. So, if it is not estimating, I'd look more carefully. Are you using factor variable notation to do the square?

You might look at Myles Shaver's article on interactions in fixed effect models:
https://pubsonline.informs.org/doi/1...stsc.2018.0065