Pablo Egaña Del Sol

How much should we trust self-reported measures? an applied neuroscience application to social program evaluations

With Paul Sajda (Columbia U) and Nicole Moscowitz (Columbia U)

Abstract

Measures of cognitive skills —i.e. GPA, IQ, etc.— account for a small fraction of the variance in salaries and other economic outcomes (Bowles et al., 2001; Heckman et al., 2006). Therefore, there is an increasing interest in elucidating other factors that might explain that variance; in particular, the role played by non-cognitive skills. In the economics literature, empirical attempts to measure the aforementioned dimensions —cognition, personality traits, and creativity— have been insightful but are noisy due to the low reliability of the proxies used in their measurement (Calero et al., 2014; Cunha and Heckman, 2008; Cunha et al., 2010; Attanasio et al., 2015a; Almlund et al., 2011). This chapter is intrinsically related to the impact evaluation done in the Job Market Paper, and was designed as a deeper proof of concept of the relationship between emotions and self-reporting on measures of non-cognitive skills. The purpose is to show a plausible correlation between self-reported psychometric tests and transient emotional states, with the latter estimated from EEG recordings. To do so, a couple of relationships are established: namely, (i) showing the behavioral correlation between self-reported psychometric tests and transient emotional states, (ii) detecting emotional state and responsiveness from scalp EEG recordings, and (iii) showing the correlation between self-reported psychometric tests and transient emotional states using features from the EEG recordings. I argue that there is a positive correlation between emotional state and test scores that rely on self-ratings. In particular, I claim that self-reported tests used to measure both cognitive and non-cognitive skills are usually biased because of transient emotions that arise during testing.

Publication