FRBSF Economic Letter
2002-11; April 19, 2002
Macroeconomic Models for Monetary Policy
This Economic Letter summarizes the papers presented at the
conference "Macroeconomic Models for Monetary Policy" held at
the Federal Reserve Bank of San Francisco on March 1-2, 2002, under the
joint sponsorship of the Federal Reserve Bank of San Francisco and the
Stanford Institute for Economic Policy Research. The papers are listed
at the end and are available at http://www.frbsf.org/economics/conferences/0203/index.html.
Models of the economy are valuable tools for monetary policymakers for
at least two reasons. First, such models can help produce forecasts of
future inflation, output, and other variables, which are crucial for a
forward-looking central banker who takes into account lags in the effects
of monetary policy. Second, macroeconomic models can help quantify the
amount of uncertainty that central bankers face in making their policy
choices—particularly through the use of alternative model simulations.
The research and discussion at this conference considered which macroeconomic
models would be most useful in guiding monetary policy. Some of the relevant
issues included the role of explicit expectations in models, the use of
multiple models, the importance of judgmental adjustments to models, identifying
model structural change, and the appropriate size and amount of detail
There are three broad categories of macroeconomic models currently being
considered for monetary policy analysis, and this conference had examples
and proponents of all three. One category contains calibrated or estimated
general equilibrium (GE) models, which are closely based on a detailed
theoretical structure that features explicitly optimizing businesses and
consumers. The paper by Smets and Wouters formulates such a model with
sticky prices and wages for the euro area to investigate business cycle
fluctuations and optimal monetary policy.
The papers by Sbordone and by Neiss and Nelson also start with a dynamic
stochastic GE model as motivation; however, their analyses focus on the
single issue of the appropriate econometric specification of the determination
of wages and prices, the subject of the paper by Rudd and Whelan as well.
Such a focus on the empirical estimates of a structural equation is the
hallmark of the second type of model used to analyze monetary policy:
the structural macroeconometric model. As was made clear in the conference
panel discussion by Adrian Pagan and David Stockton, such structural macroeconometric
models are the most common type of model used at central banks. These
models, which continue a line of research over 50 years old, have been
updated during the past decade or so with explicit expectations and better
long-run properties, but another panel discussant, Larry Christiano, suggested
that GE models could be a useful alternative.
The third category of models contains those that are almost purely statistical
in nature, particularly Vector Autoregressions (VARs). The paper by Leeper
and Zha uses a VAR to consider the plausibility of various monetary policy
actions, and the paper by Sims and Zha uses a VAR to examine changes in
the variances of shocks that buffet the economy.
Monetary policy in an estimated GE model of the euro
Smets and Wouters develop a dynamic stochastic GE model with sticky prices
and wages for the euro area. The model is theoretically quite intricate,
with features such as monopolistically competitive markets, costs to adjusting
the capital stock, habit formation by consumers, and a variable rate of
capacity utilization. Smets and Wouters attempt to estimate this model
and analyze the relative contributions of eight different kinds of shocks
to business cycle fluctuations in the euro area. They find that productivity
shocks account for only 10% of the long-run variability in output, contrary
to what so-called Real Business Cycle models would predict.
The estimated model is also used to show that historical monetary policy
in the euro area has apparently deviated from the way an optimal monetary
policy should have been set in response to various structural shocks.
An important caveat to this analysis is that the unified euro area did
not exist historically, so the analysis pertains to a synthetic history
of reconstructed euro area data. (For a complementary analysis with a
very simple macroeconometric model, see Rudebusch and Svensson 2002.)
An optimizing model of U.S. wage and price dynamics
The Phillips curve, which links inflation with an unemployment or output
gap from trend, has provided perhaps the most popular empirical description
of wage and price dynamics for the past half century; however, the theoretical
foundations of this model are considered weak and have been the subject
of almost constant debate. In her paper, Sbordone uses a GE optimizing
model to derive a theoretical "New Keynesian" Phillips curve,
which relates inflation to future expected inflation and marginal costs.
The output gap in such a model should be measured as the deviation from
potential output with a stochastic trend (i.e., incorporating actual aggregate
demand and supply shocks) rather than the usual deterministic trend.
Sbordone also argues that incorporating labor cost dynamics is crucial
to a model of price dynamics. She derives the joint dynamics of wages
and prices implied by a sticky-price model with a perfectly competitive,
flexible-wage labor market and, alternatively, with a monopolistically
competitive, sticky-wage labor market. Sbordone compares the implied dynamics
with actual postwar U.S. data, and concludes that the model performs quite
well in predicting the inflation process using the real wage. The real
output gap, when measured as deviations from a stochastic trend potential
output, also performs well in matching inflation dynamics.
Should monetary policy target labor's share of income?
In a closely related paper, Rudd and Whelan challenge the empirical results
in Woodford (2001) and Sbordone (2002), which also suggested that marginal
costs or wages worked better in predicting inflation than the output gaps
used in the traditional Phillips curve literature. Specifically, Woodford
(2001) presented evidence that using real unit labor costs (labor's share
of income) as a driving variable in the Phillips curve yields a superior
fit for inflation than a model that uses deterministically detrended real
GDP as the driving variable. For his empirical work, expectations of the
driving variables were obtained from a reduced-form VAR. However, Rudd
and Whelan find that Woodford's result is not robust and that the evidence
in favor of using the labor's income share is highly sensitive to small
changes in the specification of the VAR.
Rudd and Whelan also show that the principal reason for the good fit
obtained by Sbordone (2002) is not the use of the labor income share as
a driving variable, but rather an additional—and unrealistic—assumption
that the nominal marginal cost evolves independently of the price level.
Indeed, after imposing a similar assumption that nominal output evolves
independently of the price level, Rudd and Whelan find that the New Keynesian
Phillips curve with a traditional output gap (defined using a deterministically
trended potential output) performs just as well as the labor income share
version. Furthermore, the reason that both models obtain fairly good results
is that both estimation equations include lagged inflation as one of the
explanatory variables. Because inflation is highly persistent, the lagged
inflation term helps explain a lot of the variation in inflation. Accordingly,
the use of the labor income share does not improve the inflation prediction
Inflation dynamics, marginal cost, and the output
The Neiss and Nelson paper also focuses on the structural modeling of
inflation dynamics and argues that the output gap obtained using a smooth
deterministic trend for potential output is not appropriate, because potential
output should be stochastic and correspond to the output level that would
prevail if there were no nominal rigidities in the economy (i.e., if prices
and wages are flexible). In other words, potential output should be affected
by real shocks over the business cycle and should not follow a smooth
path, as typically assumed. However, rather than replacing the output
gap with a marginal cost measure based on labor costs (as in the Sbordone
paper), Neiss and Nelson advocate a new output gap that is constructed
to be consistent in theory with a dynamic stochastic GE model.
Neiss and Nelson start by formulating a GE model characterized by habit
formation and capital investment adjustment costs. They calibrate this
model and discuss procedures for constructing an empirical potential output
gap series that is consistent with the model. Using post-war data for
the U.S., the U.K., and Australia, they find that output gaps defined
in a manner consistent with their model perform as well as unit labor
costs in predicting inflation. Therefore, they conclude that there is
little evidence to support the recent emphasis on the role of labor market
rigidities for modeling inflation.
Empirical analysis of policy interventions
Leeper and Zha attempt to provide a methodology for analyzing the response
of the economy to changes in monetary policy. Such analyses are hindered
by the Lucas critique, which states that changes in policy also affect
the behavior of rational agents, and such behavioral changes can invalidate
the model relationships estimated under the previous policy regime. As
also described by Rudebusch (2002), the Lucas critique complicates the
assessment of proposed policy actions.
Leeper and Zha first formulate a six-variable monthly structural VAR
model and show that the monetary policy shocks identified have a fairly
stable impact on the economy from 1959 to 1998, which suggests that there
have been no substantive changes in agents' beliefs about the policy regime.
Therefore, this estimated structural VAR model can be used to analyze
the effects of hypothetical changes in policy as long as those changes
are not too different from historical actions (and thus avoid the Lucas
critique). Leeper and Zha propose a statistical metric for judging whether
the hypothetical policy interventions are large enough to be considered
changes in the policy regime. They apply this metric to judge whether
hypothetical policies represent a recognizable break from past policy
in two different contexts: first, in assessing the usual central bank
forecasting assumption of constant nominal interest rates and, second,
in examining various policy actions that the Federal Reserve could have
taken in the 1990s.
Sims and Zha formulate a six-variable structural VAR in which they allow
for certain types of parameter variation over time. In estimating the
VAR, they first allow both the policy rule and the variances of the structural
shocks to change over time, and then they compare the fit of this model
to one in which only the shock variances are allowed to change. From their
estimates, Sims and Zha conclude that allowing for changing shock variances
over time is more important for improving model fit than allowing for
a changing policy rule. Furthermore, counterfactural exercises suggest
that, even without the presence of a "Volcker regime," inflation
in the U.S. after 1979 would have declined and the recession in the early
1980s would have been smaller, although inflation would have fallen less
rapidly in this alternative and output growth would have been much slower
after 1984. In contrast, the conventional wisdom on this subject (described
in Rudebusch 2002) is that the systematic component of Federal Reserve
monetary policy has changed dramatically over time and has at least partly
accounted for the rise and fall of the historical U.S. inflation rate.
|Glenn D. Rudebusch
Leeper, Eric, and Tao Zha. "Empirical
Analysis of Policy Interventions." Indiana University and Federal
Reserve Bank of Atlanta.
Neiss, Katharine, and Edward Nelson. "Inflation
Dynamics, Marginal Cost, and the Output Gap: Evidence from Three Countries."
Bank of England.
Rudd, Jeremy, and Karl Whelan. "Should
Monetary Policy Target Labor's Share of Income?" Federal Reserve
Sbordone, Argia. "An
Optimizing Model of U.S. Wage and Price Dynamics." Rutgers University.
Sims, Christopher, and Tao Zha.
"Macroeconomic Switching." Princeton University and Federal
Reserve Bank of Atlanta.
Smets, Frank, and Raf Wouters. "Monetary
Policy in an Estimated Stochastic Dynamic General Equilibrium Model of
the Euro Area." European Central Bank and National Bank of Belgium.
Rudebusch, Glenn. 2002. "Assessing the Lucas Critique in Monetary
Policy Models," FRBSF Working Paper 2002-02.
Rudebusch, Glenn, and Lars Svensson. 2002. "Eurosystem
Monetary Targeting: Lessons from U.S. Data." European Economic Review
46, pp. 417-442.
Sbordone, Argia. 2002. "Prices and Unit Labor
Costs: A New Test of Price Stickiness." Journal of Monetary Economics
49(2), pp. 265-292.
Woodford, Michael. 2001. "The Taylor Rule and
Optimal Monetary Policy." American Economic Review 91(2), pp. 232-237.