Non-cognitive skills: Their formation, measurement, and interaction effects on earnings

Abstract

I investigate the impact of university education on students' non-cognitive skills using Australian longitudinal data. To isolate the skill-building effects of tertiary education, the education decisions and non-cognitive skills of 575 adolescents are followed over eight years. A standard skill production function is estimated, and it shows a robust positive relationship between university education and extraversion, and agreeableness for students from disadvantaged backgrounds. The effects are likely to operate through exposure to university life rather than through degree-specific curricula or university-specific teaching quality. As extraversion and agreeableness are associated with socially beneficial behaviours, university education may have important non-market returns.

I explore the reasons why recent studies have found evidence for personality types in several large personality data sets, despite the abundance of research that has replicated the Big Five personality traits. I show that many key statistics that are routinely used to guide factor extraction, show internal consistency, and demonstrate factor replicability do not imply the absence of clusters. Using two personality data sets, I compare the relative model fit of various factor and mixture models and find that a mixture model provides a better relative fit. Importantly, even the best mixture model fails to capture all the nuances in the personality data, suggesting the need for alternative latent variable models.

I examine the interaction effects between non-cognitive and cognitive skills on labour market earnings. One challenge facing researchers is that in a typical wage regression the number of two-way interactions that can be tested for is practically limited. I combine the use of multivariate adaptive regression splines with interaction test statistics to identify interaction effects using Australian data.