INterpretation of quadratic term with multicollinearity

[Tom Schmidt]

Hi all,

when I put a linear term and a quadratic termn of experience in my linear regression model, I get nice P-values, but they have a collinearity of 0.99. What is the reason? And should I still leave one of the terms in? If yes, which one?

Thanks!

[Carlo Lazzaro]

Tom:
the sky-rocketing collinearity between linear and squared terms is expected, as you can see from the following toy example:

. sysuse auto.dta
(1978 Automobile Data)

. g sq_price=price^2

. pwcorr price sq_price

             |    price sq_price
-------------+------------------
       price |   1.0000
    sq_price |   0.9846   1.0000

Keep both terms in your regression equation and determine the turning point.