Research indicates that teachers with high value-added (VA) improve their students’ later life outcomes, leading to VA measures featuring ever more widely in high-stakes decision-making. Yet concerns remain that VA measures might capture undesirable dimensions of learning such as rote memorization that are unlikely to contribute to a student’s lifelong success. This paper separates test score value-added measure into two components: long-run VA, which persists to the following year, and short-run VA, which does not. We find that students assigned to high long-run VA teachers fare substantially better in terms of long-run outcomes. In contrast, students assigned to high short-run VA teachers do not perform any better in terms of long-run outcomes. We also find that long-run VA is correlated with non-cognitive VA measures, whereas short-run VA is not. Policy simulations indicate that the use of long-run VA improves policy efectiveness by a factor of two over traditional VA measures.