This paper provides rigorous evidence (for 12,130 participants in a series of naturally occurring randomized lotteries) that a large-scale high school reform initiative (New York City's creation of 100+ small high schools of choice between 2002 and 2008) can markedly and consistently increase high school graduation rates (by 9.5 percentage points overall and for many different student subgroups) for a large population of educationally and economically disadvantaged students of color without increasing annual school operating costs. These findings are directly relevant to current debates by policymakers and practitioners about how to improve the educational prospects of disadvantaged students in the United States.