This dissertation studies the education-to-work pipeline in low-income countries, examining how policies shape entry into higher education, how educational expansion affects local economies, and how labor markets absorb university graduates. Each chapter uses a distinct identification strategy: a randomized experiment, a staggered difference-in-differences design, and a dosage-style difference-in-differences, to provide causal evidence on a different stage of this pipeline. The first chapter examines demand-side frictions in graduate labor markets. Young university graduates in sub-Saharan Africa face persistently high unemployment despite rising employer demand for skilled workers. While substantial research has addressed barriers facing job seekers, much less is known about whether employers themselves contribute to poor matching outcomes. I conduct a firm-level randomized experiment with 376 firms posting 510 job openings in Kampala, Uganda, providing an intervention that helps firms identify and communicate their hiring needs when writing job descriptions. I vary whether firms use the resulting job descriptions as public advertisements or only for internal candidate screening, separately identifying effects through better signaling versus improved evaluation criteria. Treated firms post significantly more detailed job descriptions that attract nearly 40 percent larger applicant pools with better skill alignment, and the intervention reduces personal network-based hiring from 8 percent to essentially zero. However, these recruitment improvements do not translate into better short-term employment matches: wages, employer-reported skill fit, and performance ratings show no significant improvement on average. Heterogeneity analysis suggests that organizational capacity shapes outcomes. Among firms with existing HR departments, structured evaluation tools improve skill match quality, while firms without HR show no such gains. The results establish that firms’ inability to systematically articulate hiring requirements is a meaningful friction in developing economy labor markets, but that addressing this constraint alone is insufficient without complementary hiring infrastructure. The second chapter, joint with Brian Dillon, examines how the rapid proliferation of higher education institutions across East Africa affects surrounding communities beyond direct enrollment impacts. The opening of a university or college in a small city constitutes a form of place-based public investment, yet little is known about whether such expansions generate broader local spillovers beyond direct enrollment effects. Leveraging the staggered timing of university and college openings across Uganda over 2000–2020 and employing recent advances in difference-in-differences estimation, we find that the first opening of a higher education institution in a city reduces agricultural employment, increases self-employment in sales and services (especially for women), and substantially improves secondary school completion and literacy rates among local youth. The ten-year treatment effect on literacy represents a 65 percent increase relative to the control mean. These results suggest that university openings contribute to shifts in occupational patterns and educational aspirations that benefit local populations well beyond the students who enroll, and provide evidence that the indirect returns to educational expansion may be large in low-income country settings. The third chapter, co-authored with Isaac Ahimbisibwe and Henry Joe Opio, provides the first causal evaluation of Uganda’s 2006 compulsory science policy, which mandated chemistry, physics, and history at the secondary level and simultaneously reserved 70% of national merit university scholar- ships for STEM majors. Despite the policy’s stated goal of expanding STEM participation, declining aggregate performance has prompted stakeholders to call for repeal. We exploit pre-policy variation across schools in science participation to construct a school-level treatment intensity measure and implement a dosage-style difference-in-differences design using administrative examination records for 1,820 schools from 2000 to 2020, linked to university scholarship application records covering 407,213 applicants. The mandate reduced aggregate O-level performance, with effects concentrated in schools that had enough infrastructure to comply but insufficient capacity to expand science instruction without crowding out quality in other subjects. Examining the composition of students’ best-six aggregates suggests these declines reflect adverse effects on learning rather than a mechanical composition effect. Performance costs are largest among high-ability students, consistent with an opportunity cost mechanism. Despite these costs, the mandate significantly increased STEM applications at the university level, by 5.8 percentage points relative to pre-reform rates. These gains are concentrated among students from low-capacity schools and lower-ability students, for whom the science credential constraint was most binding pre-reform, and the mandate reduced the gender gap in applications to competitive STEM programs. The costs and benefits of the policy therefore fell on different students, with implications for how such mandates should be evaluated.