Testing students and using test information to hold schools and, in some cases, teachers accountable for student achievement has arguably been the primary national strategy for school improvement over the past decade and a half. Tests are also intended to be used as a diagnostic tool to identify individual student needs, so that students can be set on a trajectory for long-term academic success. We use panel data from three states -- North Carolina, Massachusetts and Washington State -- to investigate how accurate early measures of achievement are in predicting later high school outcomes. We contribute to the literature in three distinct ways. First, the long panels we employ allow us to quantify the accuracy of models predicting how early (3rd and 4th grade) measures of student background and achievement predict several later schooling outcomes: 8th grade test achievement, high school course-taking, and high school graduation. Second, we test the extent to which predictions based on distinct segments of student data (e.g. grades 3 to 8, then 8 to 12) sacrifice forecast accuracy; this is of particular policy relevance for states or localities that do not yet have long administrative data panels. Finally, we test the degree to which the use of parameter estimates from models predicting schooling outcomes derived from one state diminish the accuracy of predicting outcomes in other states.