On the informational loss inherent in approximation procedures : welfare implications and impulse responses

This paper shows the inappropriatedness of approximation procedures for welfare rankings across suboptimal policy strategies. On the grounds of a simple general equilibrium model, we find that even commonly accepted techniques are not suitable to achieve accurate welfare orderings. This result points to a non-universality of these methods since we unveil welfare reversals when we compare them with the implications of the corresponding Ramsey problem. We infer that the spurious outcomes originate from restricting the approximations to only first and second-order moments. The order of approximation needed to obtain accuracy obviously depends on the underlying type of general equilibrium model and on its parameter values. But this creates uncertainty about the correct degree of approximation adopted by researchers in order to obtain clear welfare insights. Therefore, we strongly recommend that normative analyses should rely exclusively on the exact optimality conditions delivered by the Ramsey problem. Nonetheless, we are able to propose approximation methods in order to characterize macroeconomic fluctuations triggered by small disturbances. -- Monetary Policy ; Macroeconomic Policy Consistency ; First-order approximation ; Second-order approximation ; Purely quadratic approach