Monetary policy loss functions: two cheers for the quadratic

Following Blinder’s (1997) suggestion, we examine the implications for the optimal interest rate rule which follow from relaxing the assumption that the policymaker’s loss function is quadratic. We investigate deviations from quadratics for both symmetric and asymmetric preferences for a single target and find that (i) other characterizations of risk aversion than implied by the quadratic only affect dead-weight losses, unless there is multiplicative uncertainty; (ii) asymmetries affect the optimal rule under both additive and multiplicative uncertainty but result in interest rate paths observationally similar, and in some cases equivalent, to those implied by a shifted quadratic; (iii) the use of asymmetric loss functions leads to important insights on the issue of goal independence and monetary policy delegation; (iv) non-quadratic preferences, per se, are neither sufficient nor necessary to generate the ‘Brainard conservatism principle’ and thus do not offer much added value when analyzing policy issues of caution and gradualism. Our results suggest that in the context of monetary policymaking the convenient assumption of quadratic losses may not be that drastic after all.