We add to an emerging body of literature on input subsidies in Africa south of the Sahara. Our analysis focuses on demand for seed, characterising smallholders with a high predicted demand for hybrid seed who were not reached by the subsidy programme. We use cross‐sectional data from the 2010 agricultural season and an instrumented control function approach to test the hypothesis that the subsidy on hybrid maize seed in Zambia is selectively biased. Consistent with other literature, we find that the subsidy is a recursive determinant of seed demand, but in 2010, its recipients had more land, more assets, and lower poverty rates. Findings illustrate the social costs of the programme as currently designed and highlight the need to build alternative supply channels if poorer maize growers are to grow hybrid seed.