<p>When and how should governments use industrial policy to direct innovation to specific sectors? This paper develops a framework to analyze the costs and benefits of industrial policies for innovation. The framework is based on a model of endogenous innovation with a sectoral network of knowledge spillovers (Liu and Ma 2023), extended to capture implementation frictions and alternative policy goals. Simulations show that implementing sector-specific fiscal support is only preferable to sector-neutral support under restrictive conditions—when externalities are well measured (e.g., greenhouse gas emissions), targeted sectors generate high domestic knowledge spillovers, and administrative capacity is strong (including to avoid misallocation to politically connected sectors). If any of these conditions are not fully met, welfare impacts of industrial policy for innovation quickly become negative. For a sample of technologically advanced economies, existing industrial policies seem to be directing innovation to broadly the right sectors, but to an excessive degree in most economies, including China and the United States. Relative to existing policies, the optimal allocation of support entails slightly greater subsidies to greener sectors, but cross-sector knowledge spillovers also matter.</p>