Tuesday, January 25, 2011

Savelyev - Conscientiousness, Education and the Longevity of High-Ability Individuals

Working Paper from December 2010 - available on following webpage


This paper investigates whether post-high school educational investments causally affect longevity. The answer to this question is unclear in the literature, largely because not enough is known about possible confounding factors and because it is hard to find instrumental variables for post-compulsory education. I examine whether there are confounding factors among early cognitive and noncognitive traits that causally affect both education and longevity, inducing spurious correlation between them.

I represent noncognitive traits by three personality traits widely recognized in personality psychology—Conscientiousness, Openness, and Extraversion—and find that only Conscientiousness has strong effects on both education and longevity. I estimate a model which uses Conscientiousness and education as arguments of a production function for longevity. The model accounts for the endogeneity of education, measurement error in the proxies for Conscientiousness, age-dependence in the effect of education on longevity, and the interaction between education and Conscientiousness in producing longevity. I estimate the model using the 1922-1991 Terman life cycle data of children with high ability, a prospective study with unique life cycle information including detailed background characteristics, early health measures, personality ratings, IQ, and mortality observations.

My results show that Conscientiousness and education both causally increase longevity for males. However, the effect of Conscientiousness on longevity is only strong at low levels of education, and the effect of education on longevity is especially strong at low levels of Conscientiousness. In addition, I show that a failure to account for Conscientiousness leads to a generally upward bias in the estimate of the effect of education on longevity. The bias from omitting Conscientiousness is comparable to the bias from omitting all other control variables in my model. For females, the effects of education and Conscientiousness are generally not precisely determined.

1 comment:

Martin Ryan said...

Thanks for posting this Liam. Will be reading it in full.