ments over time. One implication of these auxiliary state variables is that timeless perspective
policies involve commitments and are not time-consistent in the sense of Kydland and Prescott
(1977). At the same time, timeless perspective policies are not optimal in the sense of Kyd-
land and Prescott (1980), opening the door to the possibility that they may be inferior to
other suboptimal policies, such as discretion. It is important to compare the performance
of timeless perspective policies to discretion because such a comparison helps to identify and
understand situations where timeless perspective policymaking may be inferior to discretion.
More generally, such a comparison allows us to better understand when discretionary policies
perform well and when timeless perspective policies perform less well.
Rather than employ unconditional loss to compare discretion to timeless perspective poli-
cymaking (McCallum and Nelson, 2004; Sauer, 2007), I develop a measure of conditional loss
suitable for the task.4’5 In particular, I show how the auxiliary state variables that enter the
timeless perspective equilibrium can be “integrated out” to produce a measure of loss that
depends upon the state variables that describe the original optimization problem and that are
common to both equilibria. For linear-quadratic models, this integration lowers the perfor-
mance of the timeless perspective policy relative to the optimal commitment policy by terms
that quantify the conditional mean and the conditional volatility of the auxiliary states. Of
course, it is far from automatic that these adjustments will permit a timeless perspective to
be dominated by discretion. However, using standard New Keynesian DSGE models, I show
that factors that flatten the New Keynesian Phillips curve, such as nominal price rigidity,
firm-specific labor∕capital, and Kimball (1995) aggregation, can raise the conditional volatil-
ity (in particular) of the auxiliary state variables to the point where discretion becomes the
superior policy. Indeed, the intuition for this result is reasonably clear. As the Phillips curve
becomes increasingly flat, the central bank must generate greater volatility in real marginal
costs in order to stabilize inflation. To the extent that real marginal costs are correlated
4Indeed, some have interpreted the term “timeless perspective” to mean that timeless perspective policies
should be derived as the solution to an unconditional optimization problem (Blake, 2001, Jensen and McCallum,
2002, Damjanovic, Damjanovic, and Nolan, 2008). Since Woodford’s approach to timeless perspective policy
design does not do this, these studies have found that timeless perspective policies are not optimal from the
timeless perspective. The policy associated with the solution to the unconditional optimization problem is
sometimes referred to as the Blake-Jensen-McCallum alternative.
5There are several good reasons why unconditional loss should not be used to compare policies. First, the
loss function common to both the timeless perspective and discretionary optimization problems is (invariably)
conditional. Second, because the timeless perspective policy and the optimal commitment policy share the
same asymptotic equilibrium, using unconditional loss to evaluate performance amounts to comparing discretion
to the optimal commitment policy. Third, by ignoring transition dynamics, the use of unconditional loss can
generate spurious performance reversals (Kim, Kim, Schaumburg, and Sims, 2005).