the independent variables. All the coefficients are positive and significant at the 1 percent
significance level. The coefficient of the normalized AFQT scores is 0.178, compared with the
normalized SES index’s coefficient of 0.083. Due to the fact that both AFQT and SES are
normalized and have the same standard deviation of 1, a higher estimated coefficient indicates a
bigger marginal effect on the independent variable. A standard deviation increase in the
normalized AFQT score increases the wage rate by 17.8 percent, which is more than twice the
effect of a one standard deviation increase of the normalized SES index. This result is consistent
with Herrnstein and Murray’s conclusion that cognitive ability, measured by AFQT, plays a
larger role in determining wage differentials than does family background. Regression 1B then
reruns this specification on 2002 NLSY79 respondent data to provide a bridge to other results
presented in this paper. Here the coefficients on both AFQT scores and SES index continue to be
positive and statistically significant, indeed growing in magnitude, though the AFQT score effect
grows by less than the SES index effect as the subject cohort ages; age becomes statistically
insignificant as a regressor for the respondent cohort.
Fisher et al. (1996) extend the analysis from the regressions of Herrnstein and Murray
(1994). They point out that the measure for social and family background used by Herrnstein and
Murray, i.e., the SES index, has “left out many important features of the social environment that
affect who is at risk of being poor” (Fisher et al. [1996], p.78). For example, Herrnstein and
Murray explain the high correlation between the AFQT test scores and future education levels (r
= 0.6 for white males) as a one directional causal effect: that schooling is in itself a signal of
cognitive ability. Smarter respondents that score higher on IQ tests will go to colleges and
graduate schools, obtain better jobs and enjoy higher living standards; therefore, schooling itself
does not directly explain wage differentials. However, Fisher et al. argue that education has an