We examine how incentives for test prep varied between math and English language arts (ELA) on U.S. state accountability exams. We collected data on exam structure for grade 3 to 8 tests in six states that are the setting for most U.S. research in literatures where accountability matters. We show that math exams typically measured ability more precisely for students on the margin of achieving proficiency. This gave educators an incentive to spend more time preparing students for math tests than for ELA tests, consistent with the common finding of larger math effects in the literature.