Measurement Error in the AFQT in the NLSY79.

Many promising efforts in the social sciences aim to measure future outcomes (such as wages or health outcomes) given some base level of human capital or ability. They typically fail to recognize the proxies for human capital are all measured with error, creating bias in regression analysis. Here I show how item level data offers the opportunity to improve a broad range of economic, social and psychometric studies; an opportunity now enhanced significantly by the new release of item response level data for the Armed Forces Qualifying Test in the 1979 National Longitudinal Survey of Youth.