The way monetary policy is conducted is a key element in New Keynesian models, and crucially determines allocations’ properties. We show that assuming monetary authorities follow a Taylor rule may bias estimation of New Keynesian type models for two reasons. The first one is theoretically trivial, and is a standard misspecification bias that occurs if the actual conduct of policy does not follow the model specified Taylor rule. The second one is more subtle, and we refer to it as a determinacy bias. It occurs when wrongly assuming a Taylor rule restricts the set of admissible model deep parameters when one requires the equilibrium to be determinate, as is almost always the case in the applied literature. Using US data, we show that the determinacy bias is a serious problem in small scale New Keynesian models, as the slope of Phillips curve is biased upwards. The misspecification bias is a serious problem when estimating a medium-scale model, as it affects the contribution of the various shocks to macroeconomic fluctuations. We propose an alternative agnostic specification of the policy rule that is immune to both misspecification and determinacy biases.