« U.S. Supreme Court Justices Database | Main | Insights from the Methodology of Psychology (Blumenthal #4) »

09 March 2007


William Henderson


You raise a good question. In fact, there is not a very strong correlation between compensation and interesting work or quality work. Multicollinearity is not typically a problem unless the correlations are > .7.

I also check the variance inflation factor (VIF) diagnostics. The usual rule of thumb is that VIF > 5 are problematic and may require re-specification of the model. The VIF were all lower than that.

John Coates

Interesting, and within some range, plausible. But I don't think your regression gets me there: what's the correlation between compensation and interestingness/quality of work? If it's high, as I expect, then multicollinearity makes inference about both difficult. As a thought experiment: suppose a firm scoring high on interestingness/quality of work cut salaries by 50% for incoming associates. Do we really think the associates would be as good, or that, conditional on taking the job in the first place, stick around as long?

The comments to this entry are closed.


June 2022

Sun Mon Tue Wed Thu Fri Sat
      1 2 3 4
5 6 7 8 9 10 11
12 13 14 15 16 17 18
19 20 21 22 23 24 25
26 27 28 29 30    

Site Meter

Creative Commons License

  • Creative Commons License
Blog powered by Typepad