FRBSF Economic Letter
2004-05; February 13, 2004
In a speech last month to the annual meeting of the American Economic
Association, Fed Chairman Alan Greenspan said, "The Federal
Reserve's experiences over the past two decades make it clear that
uncertainty is not just a pervasive feature of the monetary policy
landscape; it is the defining characteristic of that landscape." And
he gave some examples of how the Fed made decisions about policy
in the face of such uncertainty.
Following the Russian debt default in the autumn of 1998, for
example, the FOMC [Federal Open Market Committee] eased policy
perception that the economy was expanding at a satisfactory pace
and that, even without a policy initiative, it was likely to continue
doing so. We eased policy because we were concerned about the low-probability
risk that the default might trigger events that would severely
disrupt domestic and international financial markets, with outsized
adverse feedback to the performance of the economy.
involved a more detailed description of the problem of policymaking
For example, policy A might be judged as best
advancing the policymakers' objectives, conditional on a particular
model of the economy, but
might also be seen as having relatively severe adverse consequences
if the true structure of the economy turns out to be other than
the one assumed. On the other hand, policy B might be somewhat
less effective in advancing the policy objectives under the assumed
baseline model but might be relatively benign in the event that
the structure of the economy turns out to differ from the baseline.
A year ago, these considerations inclined Federal Reserve policymakers
toward an easier stance of policy aimed at limiting the risk of
deflation even though baseline forecasts from most conventional
models at that time did not project deflation; that is, we chose
a policy that, in a world of perfect certainty, would have been
judged to be too loose.
The research literature in economics has explored the task of
decisionmaking under uncertainty and has developed theories about "precautionary" policies
and "robust" policies. This Economic Letter summarizes
some of the latest results and debates in this literature.
and asymmetric costs of policy mistakes
Most discussions of monetary
policy focus on the forecasts for
inflation and real economic activity, and these forecasts, of
course, are associated with some degree of uncertainty. In the
basic framework economists generally use to address monetary
policy issues, policymakers can ignore this uncertainty under
certain conditions and instead determine the best policy based
only on the mean, or average, forecast. For example, if the average
forecast for inflation next year is 1%, the best policy action
today is the same regardless of whether the range of forecasts
is from 0 to 2% or whether the range is wider—and, therefore,
more uncertain—say, from -2% to 4%.
One condition under which
only the forecast matters while the uncertainty surrounding the
forecast can be ignored is when the
costs of undershooting
or overshooting the policy target are symmetric. That is, if the
central bank's target for inflation is 2% and the costs of having
inflation turn out to be -2% are the same as the costs of having
it turn out to be +6%, it makes sense simply to aim for the 2%
Uncertainty becomes an issue if the costs of errors are
asymmetric. For example, suppose the inflation target is 2% and
the costs of
letting inflation fall below zero are greater than the costs of
letting it rise above 4%; in that case, a policymaker might prefer
to err on the side of higher inflation. That is, it might make
sense to adopt a policy that reduces the chances of having inflation
fall below zero, even if it does raise the chances that inflation
will end up higher than 4%. A precautionary policy would err on
the side of reducing the chance that the more costly outcome occurs.
kinds of considerations are certainly not limited to monetary policymakers.
Most of us are familiar with situations in which
focusing only on expected outcomes does not lead to the best policy.
For example, suppose you have to catch an important flight, say,
at 3:00 p.m. Do you aim to arrive right at 3:00 p.m. sharp? Probably
not. Like most people, you will try to arrive at the airport a
bit early, because the costs of running into an unexpected traffic
jam and missing the flight are typically much greater then the
costs of arriving early. The precautionary policy is to err on
the side of getting to the airport a little early, even though
this means that, on average, you waste some time reading magazines
in the boarding area.
Acting with precaution means that the policymaker
takes into account not just the expected forecasts for output and
inflation but also
the uncertainties surrounding those forecasts.
Another situation in which uncertainty, and not
just expected outcomes, can matter occurs when policymakers want
robust policies, that
is, policies that do reasonably well regardless of what surprises
may lie ahead. Finding policies that are robust is particularly
important when uncertainty makes it difficult to assign probabilities
to all the different possible future situations that could occur.
The notion of robust policies is akin to the description of "policy
A" vs. "policy B" described in Chairman Greenspan's
speech. To put it in other words, a policy that is best if one's
assumptions turn out to be correct may produce poor economic outcomes
if the assumptions turn out to be wrong; in contrast, a robust
policy may never be fully optimal for any particular future scenario,
but when policymakers face great uncertainty, a robust policy will
guard against having things turn out really badly.
A simple example
of a robust policy is to carry an umbrella all the time, regardless
of the weather forecast. Most of the time,
you won't need it, but always having the umbrella with you protects
you against the worst-case outcome—getting drenched in a downpour.
and Sargent (2004) have investigated a way of thinking about the
uncertainty policymakers face by imagining the situation of
a policymaker who knows that any model of the economy on which
policy is based is likely to be mis-specified in unknown ways.
Now make it even worse: imagine that the policymaker fears that
any model she uses will turn out to be wrong in ways that produce
particularly bad outcomes. It is as if the policymaker feared that
events would conspire to make her look as bad as possible. A robust
policy would be a policy that does well in this worst-case scenario.
critics have argued that basing policy choices on the worst-case
outcome gives too much importance to what may be very unlikely
events (Svensson 2000). Leaving for the airport so early that even
in the worst traffic jam possible you still arrive in time for
your flight probably means that you end up wasting too much time
waiting at the airport and, as a consequence, fail to accomplish
other important tasks you could have worked on at the office or
at home. Or building a boat to survive in "the perfect storm" may
make it too heavy and difficult to sail 99% of the time. In terms
of the earlier description of "policy A" vs. "policy
B," policy B may be benign in the event the worst-case outcome
occurs, but it might be significantly less effective than policy
A in all but this worst case. If the worst-case outcome is very
unlikely, adopting policy B might lead to poor outcomes almost
all the time.
Basing policy on a distorted model
One interpretation of robust
policies is that these policies are optimal for a distorted model
of the economy rather than for the
model the policymaker actually believes characterizes the economy.
The distortions are designed to capture the worst-case outcomes
that might face the policymaker. For example, shocks to the inflation
rate pose central banks with a particularly difficult policy problem—attempting
to limit fluctuations in inflation will lead to increased fluctuations
in real economic activity. If such shocks turn out to be very transitory,
the problem is not serious; but if the shocks end up lasting longer,
the problem is worse. Because persistent shocks are more serious,
a policymaker who desires a robust policy will respond to all inflation
shocks as if they were going to be persistent, more persistent
that he actually expects they will turn out to be (Walsh 2003).
idea that a central bank would deliberately use a distorted model
of the economy raises some troubling issues. The trend in
recent years among many central banks has been towards more transparency—providing
clearer statements about policy goals and forecasts. It might be
difficult to explain policies to the public if they were based
on a model of the economy that the central bank knew to be wrong,
even if the distortions were designed to yield more robust policies.
Another difficulty centers on the role of staff economists and
policymakers. Staff economists would need to know the policymaker's
preferences over different macroeconomic outcomes in order to prepare "distorted
forecasts" the policymaker would find useful. The staff economists
would not be able to present a set of alternatives based on their
best estimate of the true model of the economy, letting the policymaker
choose from among these alternatives.
While basing policy on an
explicitly distorted model of the economy may be undesirable,
analyzing worst-case scenarios can be useful
as a means of assessing the risks policymakers face. Consider
the situation presented by the possibility of a deflation that
central bank views as costly. A central bank not concerned with
robustness would assess the costs of deflation and adjust them
according to the likelihood that a deflation will occur. If this
probability were small, it would have little impact on actual
policy choices. In contrast, a central bank concerned with designing
robust policy would assume that, should a negative inflation
shock occur, it might turn out to be more persistent than expected,
the economy into a serious deflation. It would choose a policy
that protects against the possibility that a persistent, negative
inflation shock leads to deflation, behaving, in essence, as
if the chances of deflation were higher than they actually are.
other words, at the first sign of a negative inflation shock,
the central bank would respond as if it expected the shock to persist,
in order to ensure that a deflation does not occur—even
probability of this worst-case scenario is remote.
Because policymakers face great uncertainty about the
future course of the economy, the impact policy actions will
have, and what sorts
of shocks might hit the economy, they need to weight both
the expected outcomes under the chosen policy and the consequences
events take unexpected turns. If the costs of upside and
risks to the economy are asymmetric, prudence calls for precautionary
policies designed to reduce the likelihood that the most
costly situation develops. Evaluating outcomes in worst-case scenarios
can be useful in assessing whether a policy is robust, ensuring
that, come what may, things don't turn out too badly.
Professor, University of California, Santa Cruz,
and Visiting Scholar, FRBSF
[URL accessed February 2004.]
Hansen, L.P., and T.J. Sargent.
2004. Robust Control and Economic Model Uncertainty. Princeton,
NJ: Princeton University Press.
Svensson, L.E.O. 2000. "Robust Control
Made Simple." Princeton
C.E. 2003. "Implications of a Changing Economic Structure
for the Strategy of Monetary Policy." Prepared for the Economic
Policy Symposium, Federal Reserve Bank of Kansas City (August).