Heterogeneity, Endogeneity, Measurement Error and Identification of the Union Wage Impact

We study union wage effects when membership is non-coercive and employers reserve the right of union recognition. Using the BHPS during 1995-2009 we find evidence of membership differentials for the lower male and middle female observed skill groups. The coverage differential at the bottom of the observed male skill distribution suggests free riding. While union members are negatively selected, there is evidence of positive selection at the bottom of the observed skill distribution. Using the unified measurement error and endogeneity bias expression, we obtain a discernible pattern between uncorrected, endogeneity corrected and fixed effects estimates of the union wage effect.