We use a unique measure of daily economic activity and manually audited non-pharmaceutical intervention indexes for Norway and Sweden to model the dynamics between COVID-19, policy, health, and business cycles within a SVAR framework. Our analysis documents large measurement errors in commonly used containment policy measures, significant endogeneity between the model’s variables, and a strong health-economy trade-off following both policy shocks and precautionary actions. We further document that a large share of the variation in containment policies is driven by news innovations and quantify via counterfactual simulations the output cost per life saved from following stricter versus softer policies.