Computer Science for All? The Impact of High School Computer Science Courses on College Majors and Earnings

635,771 students across 233 regular-program Maryland high schools

Maryland public high-school students followed into college and the early labour market to ages 24-25.

Staggered rollout of high-quality computer science course offerings across schools, with "unexpected exposure" to a CS offering used as an instrument for taking one, estimated by two-stage least squares. The paper is honest about its weakness: first-stage F statistics of 13.0 to 16.6, the authors' own concession that modest first stages might bias the IV estimate upward, and a note that standard errors need inflating by roughly 1.5x for valid inference. There is NO measure of programming skill anywhere — the outcomes are administrative attainment and earnings, which is why this source establishes the vocational case for school CS rather than the instructional one.

Unexpected exposure raised the chance of taking a high-quality CS course by about 6.2 percentage points. Taking one raised the likelihood of declaring a CS major by 10.2 points — a more than sixfold increase over the base rate — and of earning a CS bachelor's degree by 5.5 points. It raised employment at 24 by 2.6 points, at 25 by 3.0 points, and annual earnings by about 8% at age 24. Female, low-SES and Black students showed larger benefits in degree attainment and earnings but had lower take-up. The reallocation matters: high-quality CS pulled students OUT of other STEM (10-13 points) and engineering (5-9 points) rather than out of non-STEM, so part of the gain is redistribution within STEM.

Low. The instrument is the timing of a school's course offering relative to a student's enrolment, which no family chose. The residual concern is instrument strength, not heredity.

The first causal analysis of its kind in the US, with no replication in another state yet.