Problem: Suppose you want to estimate a model of women's earnings at age 50. You have data for a sample of employed women, provided by the alumni associations of Mills College and Smith College, on:
- A woman's salary at age 50
- Her age
- Year of graduation
- Her high school GPA
- Her college GPA
- Her college major
- Her job tenure (how many years she has been with employer)
- The fraction of her household income that she earns
Questions will be based on Classical Assumptions:
1. Suppose you include both HIGH SCHOOL GPA and COLLEGE GPA in your model. Both turn out to be statistically significant (by which I mean, each coefficient is at least twice as big as its estimated standard error). Suppose that you then drop COLLEGE GPA from your model. Would you expect the standard error of the coefficient on HIGH SCHOOL GPA to become smaller or larger as a result? Explain.
2. Why is it not a good idea to add Fraction of Household Income in the regression?