Some problems with welfare analysis

[jpfeifer]

See

Can we do welfare analysis after log_linearizing?

The general answer is no, you cannot do this. Due to certainty equivalence, stochastics will not affect the welfare ranking. In some cases, you can do a naive linear-quadratic approach, but generally, full second order is required. See Woodford (2002): “Inflation Stabilization and Welfare” for a thorough discussion

Generally, you need a full second order approximation. A linearized model will not do unless the steady state is undistorted. In that case, a naive linear quadratic approach will work.