Revisiting The Bell Curve Debate Regarding the Effects of Cognitive Ability on Wages

wage rate. It also reveals nothing about the possible differences in wage returns to intelligence
across demographic groups. However, we can examine the difference in the marginal wage
return to education over the two time periods. A closer look at regressions 2B and 2D indicates
that the estimated coefficient on education declines from 0.091 in the first period to 0.048 in the
second period. The drastic drop can probably be explained by the proposition that the marginal
return to education peaks during college years as the marginal pecuniary return of completing a
master’s degree or a PhD is actually lower than that of completing college (see Jacobsen and
Skillman, 2004). Most of the respondents in the NLSY79 had finished college, if they did go to
college, before 1994. From 1994 to 2002, changes in education level are due mostly to
respondents returning to school for a master’s degree or a PhD. Thus, the marginal return of
education after 1994 is substantially lower than its value before 1994.

IV The Hausman-Taylor Estimator

Both cross-sectional and panel regressions have advantages and disadvantages. Cross-
sectional analysis allows for coefficient estimations on time-invariant variables such as g, race,
and gender, yet it omits unobservable individual characteristics like the quality of education,
potentially causing bias in the estimated coefficients. Panel analysis accounts for unobservable
individual characteristics and yields consistent estimates. However, because of the violation of
the strict exogeneity assumption in the random effects model, only the fixed effects specification
can be used for our purpose, pre-empting any possibility to estimate coefficients on time-
invariant variables. Thus, neither the cross-sectional nor the fixed effects model is appropriate
for testing these hypotheses.

13

<<   13   14   15   16   17   18   19   >>

More intriguing information