To illustrate our proposed measurement-error correction, we apply it to several analyses reported in a recent paper relating educational attainment (and labour market outcomes) to a PGI for educational attainment20. The paper uses data from the HRS, one of our validation cohorts (N = 8,537; 58.3% female; median age = 83). As a preliminary analysis, the paper reports some straightforward tests of the relationship between educational attainment (EA) and the EA PGI. In Panel A of Table 3, we reproduce their univariate regression of EA on the PGI and their multivariate regression that additionally includes controls for mother’s and father’s EA. In the univariate regression, shown in column (1), a 1-standard-deviation increase in the PGI is associated with 0.844 (95% confidence interval = [0.793, 0.895]) additional years of schooling. This association is reduced to 0.619 (95% confidence interval = [0.572, 0.666]) years in column (2), once the controls are included.
