Table 3.

Application of measurement-error correction

Panel A. Association Between EA and the PGI, Without and With Controls for Parental EA
	Original		Corrected
	(1)	(2)	(3)	(4)
EA PGI	0.844 (0.026)	0.619 (0.024)	1.318 (0.041)	1.104 (0.042)
Father’s EA	-	0.154 (0.010)	-	0.112 (0.010)
Mother’s EA	-	0.176 (0.011)	-	0.141 (0.012)
# Obs.	8,537	8,537	8,537	8,537
Panel B. Interaction Between PGI and Family SES Predicting High School and College Completion
	(1)	(2)	(3)	(4)
	High school	College	High school	College
EA PGI	0.095 (0.008)	0.055 (0.008)	0.166 (0.014)	0.103 (0.014)
Family SES	0.069 (0.009)	0.031 (0.010)	0.063 (0.009)	0.034 (0.010)
EA PGI X Family SES	−0.047 (0.009)	0.068 (0.010)	−0.084 (0.015)	0.101 (0.016)
# Obs.	8,387	8,387	8,387	8,387

Notes: Each column reports estimated regression coefficients, with standard errors in parentheses. Panel A: Columns (1) and (2) replicate results from Papageorge and Thom’s Table 2 columns 1 and 2. Panel B: Columns (1) and (2) replicate results from Papageorge and Thom’s Table B.2 panel B columns 2 and 4. Panels A and B: Columns (3) and (4) apply our measurement-error-corrected estimator to the feasible-regression results in Columns (1) and (2). A value of ρ = 1.52 was used in the correction. All regressions include indicators for birth year, sex, interactions of birth year and sex, and 10 principal components of the genetic data (coefficients not reported). The regressions in Panel B also control for mother and father’s educational attainment and an indicator for whether these values are missing (these data are missing for 2000 individuals). Our panel B regressions differ from Papageorge and Thom as we do not include a cubic of the PGI as control variables. We omitted the cubic terms because our measurement-error-corrected estimator does not account for non-linear transformations of the PGI.