Published: January 10, 2022

Cobb-Clark, D., Kong, N. & Schildberg-Hörisch, H. (2021). ‘The Stability of Self-Control in a Population Representative Study’, Life Course Centre Working Paper Series, 2021-25. Institute for Social Science Research, The University of Queensland.


Deborah Cobb-Clark, Nancy Kong, and Hannah Schildberg-Hörisch.

Download: Life Course Centre Working Paper 2021-25

Non-technical summary:

Self-control is one of the most frequently studied concepts in social science. Despite this, we know surprisingly little about the extent of stability in self-control. Reaching consensus on this issue has no doubt been made difficult by the plethora of strategies used to operationalize self-control. Moreover, research has focused largely on children, adolescents, and selected samples of young adults (e.g., university students, those incarcerated), implying that we know a great deal about self-control at certain ages, and virtually nothing at others. This is perhaps understandable given that human development is characterized by a growing capacity for self-regulation into early childhood. Still, the existing evidence leaves us in the dark about the stability of self-control over much of the remaining life cycle.

We use data from a population representative panel survey to assess the stability in a well-established measure of self-control—the Brief Self-Control Scale (BSCS). Our data come from the German Socio-Economic Panel Innovation Sample (SOEP-IS), which, to our knowledge, is the only comprehensive, population representative panel survey to provide repeated self-control measures. The BSCS is a domain-general measure of trait self-control with high internal consistency, test-retest reliability, and predictive validity. Importantly, people’s BSCS scores are highly predictive of their physical, mental, and financial well-being.

Our research makes an important contribution by demonstrating that self-control exhibits mean-level, rank-order, and individual-level stability at the population level over the medium term. Moreover, changes in people’s measured self-control are unrelated to key life events including employment shocks, relationship breakdown, family death, childbirth, or length of exposure to the COVID-19 pandemic. Together, these results provide strong support for the perspective that self-control is a stable personality trait much like the Big-Five personality traits and locus of control. Any change in people’s self-control after reaching adulthood appears to be small, economically unimportant, and likely exogenous to their life experiences.

This is good news for economists who are increasingly modeling the behavioral consequences of personality traits, including self-control, on people’s life outcomes. Empirically, it is often convenient to assume that self-control is fixed over the relevant study period because this implies that self-control is exogenous to the outcome of interest, allowing an important threat to causal identification to be avoided. This stability assumption is also particularly useful when self-control is measured contemporaneously, ex post or perhaps years prior to the observed outcome since it allows the use of lead or lagged self-control measures. Estimation strategies relying on this stability assumption are, of course, likely to produce biased estimates if self-control is, in fact, not stable over the relevant time frame. Our results provide an empirical underpinning to the core identification assumption underlying many empirical analyses. Conceptually, our results are also important in highlighting that self-control, as captured in the BSCS, can be regarded as a stable personality trait. This provides an avenue for theoretical modeling and empirical evidence to be more closely integrated, and testable hypotheses about the role of self-control in economic decision making to be derived.