Panel Paper: Experimental Estimates of the Student Attendance Production Function

Saturday, November 10, 2018
8209 - Lobby Level (Marriott Wardman Park)

*Names in bold indicate Presenter

Long Tran and Seth Gershenson, American University


Student attendance is a critical input in the education production function, an intermediate outcome influenced by teachers and other school-provided inputs, and a type of non-cognitive skill valued in the labor market. Hence, education policies (e.g., Every Student Succeeds Act) are increasingly using attendance as a standard or measurable outcome with which to hold schools accountable. Holding schools and teachers accountable for student attendance, however, only makes sense if there are policy levers available that can improve attendance.

Recent research suggests that teachers do influence student attendance (Gershenson, 2016; Ladd & Sorenson, 2017; Liu & Loeb, 2017). However, a more general understanding of the extent to other classroom inputs affect student attendance is lacking. This study contributes to this gap in the literature by leveraging the experimental variation in classroom assignments created by the Tennessee’s Project STAR (Student Teacher Achievement Ratio) class size experiment to address two research questions:

(1) To what extent do classrooms vary in their aggregate impacts on student attendance?

(2) What fraction of classroom effects on student attendance is explained by observed classroom characteristics such as class size, teacher qualifications, and peer composition?

Tennessee’s Project STAR, a seminal field experiment in education, randomly assigned kindergarten students and teachers in participating schools to either small- or regular-sized classrooms in the school year of 1985-1986. The experiment continued over the next three years, following the 1985-1986 kindergarten cohort to grade 3 while also refreshing the analytic sample each year by randomly assigning new entrants to small- or regular-sized classrooms. To avoid Project STAR’s issues of noncompliance and attrition, we focus on students’ year of entry only, when compliance was essentially perfect and before students had the opportunity to drop out.

We document classroom effects on student attendance by estimating the distributions of classroom fixed and random effects. We then replace the classroom fixed effects with observed classroom inputs, including class size, teacher characteristics, and peer characteristics, to identify the specific reasons that classrooms vary in their impacts on student attendance. Preliminary results indicate that: (i) classrooms had significant aggregate impacts on chronic absence rates; (ii) smaller class size significantly decreased chronic absence rates for white students; (iii) student-teacher racial match significantly decreased chronic absence rates for black students.