As noted above, deciduous trees with yearly life cycles experience a period if intensified growth in the spring.
By contrast, in four of the six specifications in which one or two of the life-cycle stage experiences are excluded from the specification of relation (1v, 1n), the HJ tests indicate that there is a specification problem, consistent with concerns that the excluded instruments are correlated with the omitted life-cycle experience which belongs in the model.
Estimates that do not account for the endogenous determination of these three life-cycle experiences are misleading.
For verbal skills, however, this estimated effect is halved if IV estimates are used for school attainment, whether or not the other life-cycle experiences are included.
For the final specification with all three life-cycle experiences, we also present IV limited information maximum likelihood (LIML) results using the same set of excluded instruments.
For the full specifications with all three life-cycle experiences (Set 3), the first- and second-stage diagnostics are satisfactory for both types of skills; the instruments we utilize have reasonable power and we fail to reject the overidentification tests.
Without controls for the endogenous determination of the life-cycle experiences, the effects of HAZ at age 6 years and of skilled job tenure on adult nonverbal skills appear to be substantially underestimated.22.22
The endowments and the actual or expected values of the family, community, and stochastic factors for the three life-cycle stages are all on the right side of each of the reduced-form demand relations for each individual life-cycle stage (relations 4, 5, and 7) and, therefore, the three life-cycle experiences are likely correlated.
Therefore, our preferred estimates with all three life-cycle stages treated as endogenous indicate a substantial gain from higher HAZ at age 6 years in terms of both verbal and nonverbal skills more than 20 years later, controlling for both school- and post-school experiences and for the endogenous determination of all three of these life-cycle experiences.
For nonverbal skills, the combination of including post-school experience and treating all three life-cycle experiences as behaviorally determined results in a decline in the estimated coefficient on school attainment to 0.04 SD for every additional grade, and it is no longer statistically significant.
