The shortage of secondary STEM teachers has led to periodic calls over the past four decades for federal intervention. For more than 15 years, the National Science Foundation Robert Noyce Teacher Scholarship Program has constituted the most important response at the national level to these calls. The main activity is direct support to undergraduate students and degree holders to become K-12 STEM teachers. Here we examine the Noyce program by looking beyond the qualities of particular university programs and measuring long-term outcomes on a statewide level. We accomplished this through a collaboration of eight Texas universities that pooled data on Noyce scholars supported as far back as 2003. Making use of a state longitudinal dataset, we examined whether recipients of Noyce Scholarships are more likely than other math and science teachers to remain in teaching, determined where Noyce scholars taught, and estimated student learning in classes taught by Noyce Scholars. Noyce Scholars are more likely than other STEM teachers from the same universities to teach marginalized students, and students of Noyce Scholars obtained higher value-added scores in mathematics than students of other teachers at the same schools. These are some of the most important intended outcomes. However, Noyce Scholars did not stay in teaching as long as other STEM teachers from their universities, and they were more likely to switch away from schools with low-income students. The Noyce program currently operates far below the scale needed to start reducing national STEM teacher shortages. We conclude with prospects for expansion and modification of the Noyce program.