Association for Public Policy Analysis & Management

Practitioners, scholars, and policymakers increasingly recognize that success in school and life depends on a combination of cognitive and noncognitive or social-emotional factors (Duckworth & Seligman, 2005, 2006; Heckman, Stixrud, & Urzua, 2006; Jencks, 1979). Noncognitive factors are crucial to school performance and persistence (Farrington et al., 2012), and are important outcomes in their own right. While recent research (Durlak et al., 2011; Yeager & Walton, 2011) gives much reason for optimism, we are a long way from applying this knowledge to systematically improve instructional practice. Practitioners need guidance in selecting social-emotional/noncognitive frameworks, identifying appropriate measures, and using assessment data to improve student outcomes. The Becoming Effective LearnersStudent survey (BEL-S), developed from our model of noncognitive factors (Farrington et al., 2012), simultaneously measures academic behaviors, academic perseverance, academic mindsets, and learning strategies in the context of specific classrooms to illuminate the mechanisms in complex motivational processes.

In 2015-2016, we conducted the BEL Partner Project with 11 schools, districts, and charter networks (the BEL Partners). Project goals were to: (1) understand local problems of practice; (2) undertake analyses to illuminate potential levers for improvement; (3) develop and field test data visualizations, data use protocols, and other data supports embedded in practice; (4) build capacity of partner schools/districts to use survey data for continuous improvement; and (5) contribute to the knowledge base about educators’ use of noncognitive data to improve students’ classroom experience.

We administered the BEL-S survey to students in grades 6-12 in fall 2015 and spring 2016 (fall n=7509, spring n=2587) from a sample of 25 urban, suburban, and rural schools in 7 states. Each student was pre-assigned to report on their experiences in two specific “Target Classes.” For each target class, they were asked about their academic mindsets, learning strategies, perseverance, and academic behaviors within the context of that class, as well as a set of 16 items about the class itself (teacher instructional practices and supports). We were able to aggregate student responses to the classroom/teacher level. We also received administrative data with students’ prior and post course grades, test scores, and demographic data. The unique survey design gave us two data points on each noncognitive measure for each student. Using multi-level models, we found that measured differences in classroom environments predicted statistically significant within-student differences in self-reported noncognitive factors, which in turn predicted significant within-student differences in academic performance (course grades) across the two classes.

We ran five half-day meetings over the school year with designated teams at each partner site where we brought further analysis of survey data and engaged with practitioners in interpretation and next-steps planning. The project provided a direct service to our practice partners and ensured that we understand the kinds of data displays and accompanying training, tools, and supports necessary for BEL data to actually lead to improvements in practice. Deep but urgent work “on the ground” in multiple school sites helped us understand how data on student noncognitive factors and classroom experiences can help school practitioners in their daily work.