View original document

The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.

January 2014, EB14-01

Economic Brief

Learning about Fiscal Policy Uncertainty
By Christian Matthes and Tim Sablik

In response to the financial crisis and recession of 2007–09, the federal
government enacted a number of emergency fiscal policies intended to aid
recovery. These included short-term stimulus measures, such as the American
Recovery and Reinvestment Act of 2009, and temporary tax reductions, such
as the payroll tax cut in 2010. However, the unconventional and transitory
nature of these fiscal policies may have contributed to greater economic
uncertainty. Given the slow recovery that has followed the recession,
economists are studying how such uncertainty might impact growth.
Some economists have pointed to an increase
in overall economic uncertainty as a contributing factor to the slow recovery from the 2007–09
recession. Theory suggests that uncertainty can
affect the economy in a number of ways. It might
prompt firms to delay investment or hiring decisions or make households more likely to postpone consumption and increase savings, all of
which could hamper growth. In the wake of the
recession, Congress enacted a number of emergency fiscal provisions designed to aid recovery.
These policies were often temporary measures
that were subject to last-minute modifications or
extensions. For example, Congress cut the payroll
tax rate in 2010. The measure was set to expire on
Dec. 31, 2011, but continued economic weakness
prompted Congress to pass an extension just
eight days before that date.
Policy changes such as these appear to have
contributed to an overall increase in uncertainty
in recent years. To quantify this trend, Scott Baker
and Nick Bloom of Stanford University and Steven
J. Davis of the University of Chicago developed

EB14-01 - Federal Reserve Bank of Richmond

an index to measure economic uncertainty over
several decades. They looked at newspaper coverage containing terms related to the economy,
uncertainty, and policy, as well as scheduled tax
code expirations and the level of agreement
among economic forecasters.1 They found that
the overall level of uncertainty has increased
beginning in 2008. This change seems to be
driven in large part by an increase in policy uncertainty. (See Figure 1.)
This Economic Brief explores two key questions
about the role fiscal policy uncertainty might play
in the economy. First, how do firms and households learn about changes in fiscal policy, and
how is that learning process affected by their prior
beliefs about the nature of policy changes? Second, what are the economic effects of uncertainty,
and are those effects temporary or permanent?
Modeling Learning
Many macroeconomic models of the business
cycle are built on a framework of rational expectations. This theory posits that firms and

Page 1

households (or “economic agents”) have complete
knowledge of the structure of the economy. In the
event of a change, they immediately and rationally
incorporate new knowledge into their expectations
for the future. For example, in the 2007–09 recession, the government enacted a number of fiscal
policy changes to counteract negative economic
shocks, such as the collapse of the housing market.
The rational-expectations model predicts that agents
with access to complete information would expect
the new policies to improve economic conditions,
and they would immediately adjust their expectations for future economic growth upward. This could
allow the economy to move quickly toward recovery,
or even bypass the recession altogether, since firms
and households would not be as concerned about
the negative economic shocks.
But some economists have argued that rational expectations may not be the most realistic way to model how agents react during periods of policy change.
It is not clear that agents have full knowledge of the
policies in place at any given time, especially when
those policies are subject to sudden changes. These
economists propose an alternative model framework
called adaptive learning. Under adaptive learning,
firms and households are uncertain about the current
and future structure of the economy, including fiscal

policy. They form expectations about how policies are
set based on their observations of how the government acts. Thus, unlike the full-information, rationalexpectations model, there is a period of uncertainty
surrounding each policy change. Models using the
adaptive-learning framework predict very different
responses to policy changes than rational-expectations models. Economic effects from the policies may
ultimately be larger than the impact predicted under
rational expectations, and adjustment to the new
equilibrium may take longer.2
Models with adaptive learning typically assume that
agents are uncertain about the structure of the entire
economy. This may make it difficult to isolate how uncertainty about a policy change specifically impacts
economic outcomes. In a recent working paper, one
of the co-authors of this brief (Matthes) and Josef
Hollmayr of the Deutsche Bundesbank develop a
model to explore how learning and uncertainty affect
the economic response to fiscal policy changes.3 In
order to isolate these effects, they assume agents
are fully knowledgeable about the structure of the
economy but are uncertain about how the government sets fiscal policy in response to changes in economic factors, such as past output and government
debt. In other words, agents don’t know what “rules”
the government follows when setting policy. Instead,

Figure 1: Index of Economic Policy Uncertainty
Debt-Ceiling Debate/
Eurozone Crisis




Financial Crisis

Iraq War
LTCM Collapse/
Russian Financial Crisis










Note: The index is an aggregation of four components: a scaled count of news articles that refer to the economy, uncertainty, and policy;
a discounted dollar-weighted sum of scheduled expirations of federal tax code provisions; and indexes of disagreement among professional
forecasters about future Consumer Price Index inflation and future government purchases.
Sources: Scott Baker, Nick Bloom, and Steven J. Davis at

Page 2

they use their observations of the government’s
actions over time to update their expectations of
the fiscal policy rules.4
Estimating the Effects of Uncertainty
Matthes and Hollmayr compare the predictions of
their adaptive-learning model to those of a fullinformation, rational-expectations model. They run a
simulation with both models in which there is a negative economic shock followed by a general increase
in government spending.5 Under rational expectations, economic agents are immediately aware of
the fiscal policy change, and there is no period of
uncertainty or learning. The model predicts that the
policy change generates a small short-term increase
in overall output, but this comes at the cost of several
long-run effects, including higher debt and lower
Under the learning model, agents behave similarly
to the predictions under rational expectations up
until the fiscal policy change. Immediately after the
change, there is a spike in uncertainty that quickly
fades as agents learn about the new policies and
incorporate that knowledge into their decisions. This
follows the pattern Baker, Bloom, and Davis observe
in their uncertainty index. Yet despite the fact that
this period of heightened uncertainty is brief, it
generates economic outcomes that are significantly
different than those predicted under rational expectations because agents initially respond to perceived
rather than actual policy changes. In the period immediately following the policy change, average consumption is higher and average hours worked are
lower, making firms and households in the learning
model better off than those in the rational-expectations model on average. But at the same time, consumption and other variables, such as gross domestic
product (GDP), are significantly more volatile.
This short-run volatility has long-lasting effects. The
model predicts that 10 years after the policy change,
cumulative GDP is 2 percent lower. The stock of physical capital is also persistently lower due to a sudden
drop in investment immediately following the policy
change. During the period of uncertainty, agents underestimate the persistence of the increase in govern-

ment spending because they initially attribute part of
the change in policy to short-lived shocks. As a result,
they underestimate future increases in debt and
taxes, leading them to view immediate investment
as less favorable than it actually is. In contrast, agents
with rational expectations immediately realize the
new long-run levels of debt and taxes and therefore
find it more profitable to invest even during the crisis.
The extent to which the predictions of the learning
model differ from the rational-expectations model
also depends on the likelihood agents assign to policy changes. The less firms and households believe a
policy change will occur, the larger the gap in average
outcomes predicted by the two models. In essence,
it takes the agents in the learning model longer to
modify their expectations when they don’t expect a
change. This difference substantially decreases the
short-run volatility predicted by the model, but it
increases the negative long-run effects. This occurs
because the agents place a greater weight on their
prior beliefs. In the opposite scenario, when agents
think that a policy change is more likely (for example,
when policymakers announce a change ahead of
time), they react more strongly to new information as
it becomes available. The result in this case is greater
short-term volatility, as agents tend to “overreact” to
the new data.
Adaptive-learning models suggest that economists
and policymakers should exercise caution when evaluating the effects of fiscal policy changes. Assuming
that firms and households have access to full information may lead to underestimates of long-run economic effects. Matthes and Hollmayr find that even
limiting the uncertainty agents face to fiscal policy
changes yields results that are significantly different
from those predicted under rational expectations.
It is likely that firms and households face additional
uncertainties about the economy as well.
The results of the learning model also suggest a
possible role for communication about fiscal policy
changes. Increased communication by policymakers may reduce the negative long-term economic
outcomes, but to the extent that increased commu-

Page 3

nication increases the public’s belief that large policy
changes are likely, it may also substantially increase
short-term volatility.
Christian Matthes is an economist and Tim Sablik is
an economics writer in the Research Department
at the Federal Reserve Bank of Richmond.

For a full description of the methodology used to construct
the index, see Baker, Scott, Nick Bloom, and Steven J. Davis,
“Has Economic Policy Uncertainty Hampered the Recovery?”
George J. Stigler Center for the Study of the Economy and
the State, University of Chicago, Working Paper No. 242,
February 3, 2012. To learn more about the uncertainty
index, visit


See, for example, Giannitsarou, Chryssi, “Supply-Side Reforms
and Learning Dynamics,” Journal of Monetary Economics,
March 2006, vol. 53, no. 2, pp. 291–309, and Mitra, Kaushik,
George W. Evans, and Seppo Honkapohja, “Fiscal Policy and
Learning,” Centre for Dynamic Macroeconomic Analysis,
University of St. Andrews, Working Paper No. 1202, January 17,
2012, revised June 18, 2013.


Hollmayr, Josef, and Christian Matthes, “Learning about Fiscal
Policy and the Effects of Policy Uncertainty,” Federal Reserve
Bank of Richmond Working Paper No. 13-15, September 2013.


In the model, economic agents know that the government
budget constraint must hold in each period. Matthes and
Hollmayr also make certain assumptions that, while they may
not hold in reality, allow for the construction of a model that
provides insight into the real-world effects of uncertainty. For
example, the government in the model does not include a
central bank, allowing Matthes and Hollmayr to isolate the
effects of fiscal policy changes from monetary policy changes.


Matthes and Hollmayr also test their model with a variety of
other fiscal policy changes and find that overall effects are
similar to those estimated under a general spending increase.

This article may be photocopied or reprinted in its
entirety. Please credit the authors, source, and the
Federal Reserve Bank of Richmond, and include the
italicized statement below.
Views expressed in this article are those of the authors
and not necessarily those of the Federal Reserve Bank
of Richmond or the Federal Reserve System.

Richmond Baltimore Charlotte

Page 4