4) One challenge of measuring the returns to schooling is that not all students have the same major or require the same skill sets. Some majors (like STEM fields) require very technical skills and are generally highly rewarded in the marketplace. Other majors (like well, lets not name any, but you know who THEY are ;) ) dont. Suppose I ran a regression of ln(wage) = a + b1*STEM + b2*OTHER + b3*experience + b4*male + b5*South + e. That is, I think your wage depends on where you live (people in the South earn less), your gender (men earn more), years of experience (more is better), and what your major was. I classify everyone into one of three groups: 1 Graduated from a STEM field; 2 Graduated with any other degree; 3 Did not graduate. My estimates show that b1>b2>0 at the 5% level. That is, STEM people earn (lets say) 25% more than non-graduates and other BA/BS degrees earn 10% more, holding all else constant.

A. Explain why my estimate of the returns to STEM fields is likely biased.

B. Come up with an instrumental variable that you could use to separate the STEM-majors from the rest.

C. Do you expect your new IV estimate of the returns to STEM majors to be larger or smaller? WHY?