Theory: A spatial model of the policy process is used to identify conditions under which the threat of initiatives constrains legislative behavior. Hypotheses: Legislators in states that allow initiatives are expected to pass laws that more closely reflect their state's median voter's preference than legislators in states that do not allow initiatives. Methods: Logistic regression analysis of United States data relating state policies, state median voter preferences, provisions for the initiative, and several control variables in the policy area of parental consent requirements for teenage abortions. Results: Parental consent laws passed by legislatures in initiative states more closely reflect their state's median voter's preference than parental consent laws passed in states that prohibit initiatives, even after controlling for other population and institutional differences across states.