Educational accountability has grown substantially over the last decade, due in large part to the No Child Left Behind Act of 2001. Accordingly, educational researchers and policymakers are interested in the statistical properties of accountability models used for NCLB, such as status, improvement, and growth models; as well as others that are not currently used for NCLB, such as value-added models. This study examines the statistical properties of accountability measures that are based on ACT's Educational Planning and Assessment System (EPAS). Utilizing data on 1,019 high school cohorts and over 70,000 students with test scores from three time points (8th, 10th, and 11th/12th grades), different types of accountability measures are contrasted and key statistical properties are discussed--including reliability, associations with prior mean academic achievement and school contextual factors, and associations with college enrollment and retention rates. Our findings highlight how status, improvement, growth, and value-added models can lead to different conclusions about a school's effectiveness. Unlike status, improvement, and growth models, value-added models attempt to isolate and measure the school's effect on student's learning. Thus, value-added measures have smaller associations with prior mean academic achievement and, by extension, school contextual factors such as poverty level and proportion of racial/ethnic minority students. This study also highlights the need for reporting the statistical uncertainty about estimates of schools' effects so that results can be properly interpreted. Appended are: (1) States and Locales of High School Cohorts Studied; and (2) Projection Parameters from WSR Method for Projecting ACT Scores. (Contains 34 tables and 2 figures.)