Using Unemployment Rates as Instruments to Estimate Returns to Schooling

I use state unemployment rates during a person's teenage years to estimate the returns to schooling. A higher unemployment rate reduces the opportunity costs of attending school. Using the same 1980 Census data set that Angrist and Krueger (1991) use, I also estimate returns to schooling with a modified version of their quarter-of-birth instrument. The estimates from the two-stage least squares (2SLS) model using the unemployment rate and the model using the quarter-of-birth instruments are almost identical. In addition, these 2SLS estimates are larger than the ordinary least squares (OLS) estimates, supporting this counterintuitive, yet prevalent, result in the literature.