The Promise and Pitfalls of Using Imprecise School Accountability Measures

O ver the last decade, states have constructed elaborate incentive systemsfor schools using school-level test scores. By the spring of 2002, nearlyevery state and the District of Columbia had implemented some form of accountability for public schools using test scores. For instance, the state of Cali-fornia spent nearly $700 million on financial incentives in 2001, with bonuses of up to $25,000 for teachers in schools with the largest test score improvements. The federal No Child Left Behind Act of 2001 mandates even broader accountability, requiring all states to test children in grades three through eight within three years and to intervene in schools failing to achieve specific performance targets. Economists have paid scant attention to the properties of school accountability systems.1 The nature of the incentives presented to schools ultimately depends upon the strengths and weaknesses of the school-level mean test score measures upon which most accountability systems are based. Accordingly, in this article, we describe the statistical properties of school test score measures, which are less reliable than is commonly recognized, and explore the implications for school incentives. Many accountability systems that appear reasonable at first glance perform in perverse ways when test score measures are imprecise