Association for Public Policy Analysis & Management

Mounting evidence of the importance of non-tested skills in explaining students’ academic success has generated calls to incorporate measures of social-emotional learning (SEL) into evaluations of educational performance and even school accountability systems. Yet we have little evidence on the ability of existing measures of social emotional skills, most of which are based on student self-reports, to reliably capture differences in student skills when administered at scale. As a consequence, we also have a limited understanding of how specific SEL constructs vary over time and across student subgroups.

This paper exploits data from the first large-scale panel survey of students on social-emotional learning (SEL) to address two questions: (1) Do self-report surveys provide reliable and valid data on student SEL? (2) How do specific SEL constructs vary across grades, by gender, and by race/ethnicity?

Specifically, we draw on data from surveys administered to roughly 500,000 students in six large California school districts over two years designed to assess four SEL constructs: growth mindset, self-efficacy, self-management, and social awareness. The measures generally demonstrate a high degree of internal reliability overall and for most student subgroups, with the exception of growth mindset (Cronbach’s alpha = 0.7) and students below grade 5. With respect to validity, we find that the constructs are correlated in expected ways with students’ course performance (i.e., GPA), state test scores, and behavior (i.e. absences and suspensions), and these relationships are generally similar across gender and racial/ethnic subgroups. In examining reliability and validity, we also examine the extent to which student non-response and careless answering (e.g., straight-lining) compromises the data quality and the potential of alternative corrections for measurement error.

We then use these data to provide a descriptive analysis of variation in the four SEL constructs across students. We find that, unlike academic achievement, the SEL constructs do not increase monotonically across grade levels. Self-management, for example, increases in the elementary grades but falls off sharply between grades 6 and 8 before recovering in high school. We also find notable differences in trends in levels of SEL across student subgroups. For example, girls experience a sharp decline in self-efficacy between grades 5 and 9, while the self-efficacy of boys remains relatively stable. In elementary grades, black and Hispanic students report much lower levels of growth mindset and self-management than white and Asian students; these gaps narrow considerably by high school, however.

We discuss potential explanations for these and other patterns but do not explore them in detail. Rather, our focus in this paper is on using a unique dataset to establish a set of descriptive findings that, due to data limitations, have not been previously known. While our findings cannot be generalized beyond the California districts we study, the scope and scale of our data far exceeds anything in the extant literature. As a result, we expect them to play and important role in shaping the research agenda for the field of social-emotional learning.