Racial Discrimination As a Means of Cream-Skimming? a Conjoint Experiment Among US Charter School Principals
*Names in bold indicate Presenter
Arguably, there are different mechanisms through which charter schools can select prospective students, starting at the initial stage of the enrollment process in which parents seek information of how to enroll. Here, administrative burdens in accessing relevant information can be increased (Herd and Moynihan 2019). Indeed, the availability of information is a crucial criterion for parents to learn about schools of choice, and apply to them (Hastings and Weinstein 2008). But why would charter schools select certain types of students by increasing others’ learning costs?
Charter-schools are subject to competitive pressures to sustain in the educational marketplace. Hence, we predict that they will prioritize students who are not costly and able to meet bureaucratic success criteria like high test scores. However, direct information about the future performance and/or costliness of prospective students is often not available during the initial application process of charter schools.
Drawing on the theory of statistical discrimination (Arrow 1973; Phelps 1972), we argue that charter schools instead use imperfect signals of students’ future performance and costliness. This means that in the absence of direct information, charter school principals will draw on population-based inferences about the average performance/ costliness of members of certain racial/ethnic groups, and use this population-specific statistical knowledge as a stereotype against individual applicants. As a consequence, racial/ethnic groups who are perceived to perform less well on average academically (like African-Americans and Latinx; see NCES 2017) will be discriminated against in accessing charter schools.
To test these theoretical predictions, we will send out a conjoint experiment to all charter school principals in the US in April 2019. The conjoint design will involve a choice between multiple sets of two hypothetical profiles of prospective students, and ask principals to indicate which student profile they would prioritize. Student profiles will involve randomly presented attributes, including gender, race/ethnicity, math and reading scores, English language learner status, special education needs, and parents’ occupation. By randomly presenting different levels of these attributes (including no information), we can estimate the average marginal component effect of each student attribute separately and in combination with all others. This allows us to effectively tease-out whether charter school principals engage in statistical discrimination as a means of cream-skimming.