Reading is a foundational skill, yet fewer than one in three US fourth graders read at a proficient level. How best to teach reading and improve educational attainment has been the subject of an active policy debate for decades, with some arguing for unrestricted funding increases and others supporting targeted changes to specific pedagogy. Recently, phonics-based approaches have surged in popularity, even though the largest attempt to put similar methods into classrooms was previously abandoned out of concerns that the program was ineffective. This paper studies Reading First—a nearly $1 billion per year federal program established under No Child Left Behind that promoted phonics-based instruction in K–3 classrooms. Exploiting administrative data from Texas and program eligibility formulas, we estimate effects via intent-to-treat difference-in-differences and instrumental variables specifications. We find that eligibility increased third-grade reading test scores by 0.1 standard deviations, and effects persisted through at least the eighth grade. An analysis of digitized exam questions shows that Reading First students outperformed peers on questions that required basic decoding skills and those with more words, a benefit that also drove improvements in math scores. District expenditure data show that Reading First dollars were spent on reading materials and teacher salaries. Finally, we follow exposed cohorts into early adulthood finding that the phonics-based intervention increased on-time high school graduation, college enrollment, four-year college completion, and adult earnings at age 25 by around $500.