School improvement research has insufficiently considered the importance of intervening in schools with declining academic performance. Fields such as engineering and medicine have prioritized predicting decline to save structures or patients before they are in peril. Unfortunately, in education, school improvement policies and interventions are only enacted once schools reach low levels of academic performance. In this study, we apply sophisticated statistical models to analyze more than 10 years of longitudinal student achievement data in English/language arts and mathematics in the US state of Texas. We find that a considerable number of schools consistently decline over time. Some significant predictors of decline included shifting student demographics and changes in the percent of economically disadvantaged students. Higher starting percentages of students labeled as English language learners also increased the likelihood of decline, but increasing percentages of English language learners over timereduced the rate of decline. Leadership stability also appears to be important to impeding decline. We close by discussing implications for research, policy, and practice.