The measurement-error-corrected univariate regression is shown in column (3) of Panel A. We estimate that a 1-standard-deviation increase in the additive SNP factor is associated with 1.318 (95% confidence interval = [1.238, 1.398]) additional years of schooling. Relative to the PGI coefficient in column (1), this coefficient is larger by a factor of 1.318 / 0.844 = 1.56. In the regression with controls for parental education, shown in column (4), we estimate a corrected coefficient of 1.104 (95% confidence interval = [1.022, 1.186]) additional years. Relative to column (2), this is an increase by a factor of 1.104 / 0.619 = 1.78. Since for EA in the HRS, h^SNP2≈0.25 and R^2≈0.10, according to the rule of thumb mentioned above, both coefficients should be expected to have increased by a factor of 1.58 (≈0.25/0.10). The increase is larger than that from column (2) to (4) due to the positive correlations between the PGI, the controls, and the dependent variable.
