Aggregate Supply in the United States: Recent
Developments and Implications for the Conduct of
Monetary Policy
Dave Reifschneider
Federal Reserve Board
William L. Wascher
Federal Reserve Board
David Wilcox
Federal Reserve Board
Paper presented at the 14th Jacques Polak Annual Research Conference
Hosted by the International Monetary Fund
Washington, DC─November 7–8, 2013
The views expressed in this paper are those of the author(s) onl
y
, and the presence
of them, or of links to them, on the IMF website does not imply that the IMF, its
Executive Board, or its mana
g
ement endorses or shares the views expressed in the
paper.
1
1
4
4
T
T
H
H
J
J
A
A
C
C
Q
Q
U
U
E
E
S
S
P
P
O
O
L
L
A
A
K
K
A
A
N
N
N
N
U
U
A
A
L
L
R
R
E
E
S
S
E
E
A
A
R
R
C
C
H
H
C
C
O
O
N
N
F
F
E
E
R
R
E
E
N
N
C
C
E
E
N
N
O
O
V
V
E
E
M
M
B
B
E
E
R
R
7
7
–
–
8
8
,
,
2
2
0
0
1
1
3
3
Page 1 of 61
Aggregate Supply in the United States: Recent Developments and
Implications for the Conduct of Monetary Policy
Dave Reifschneider, William Wascher, and David Wilcox
1
November 1, 2013
Abstract
The recent financial crisis and ensuing recession appear to have put the productive capacity of
the economy on a lower and shallower trajectory than the one that seemed to be in place prior to
2007. Using a version of an unobserved components model introduced by Fleischman and
Roberts (2011), we estimate that potential GDP is currently about 7 percent below the trajectory
it appeared to be on prior to 2007. We also examine the recent performance of the labor market.
While the available indicators are still inconclusive, some indicators suggest that hysteresis
should be a more present concern now than it has been during previous periods of economic
recovery in the United States. We go on to argue that a significant portion of the recent damage
to the supply side of the economy plausibly was endogenous to the weakness in aggregate
demand—contrary to the conventional view that policymakers must simply accommodate
themselves to aggregate supply conditions. Endogeneity of supply with respect to demand
provides a strong motivation for a vigorous policy response to a weakening in aggregate demand,
and we present optimal-control simulations showing how monetary policy might respond to such
endogeneity in the absence of other considerations. We then discuss how other considerations—
such as increased risks of financial instability or inflation instability—could cause policymakers
to exercise restraint in their response to cyclical weakness.
1
Dave Reifschneider (davidreifs[email protected]) is the deputy director (retired), William Wascher
(william.l.wa[email protected]) is the deputy director, and David Wilcox (david.wilcox@frb.gov) is the director of the
Division of Research and Statistics at the Federal Reserve Board, Washington D.C., 20551. We would like to thank
Hess Chung, Bruce Fallick, Michael Kiley, Jean-Philippe Laforte, Jesper Linde, Jeremy Rudd, and three anonymous
reviewers at the IMF for their helpful comments; we would also like to thank John Roberts for providing the code
used to estimate the FRB/US state-space model of supply-side conditions and Dennis Mawhirter for helpful research
assistance. Finally, we should note that the analysis and conclusions set forth in this paper are those of the authors
and do not indicate concurrence by other members of the research staff or the Board of Governors.
Page 2 of 61
In the United States, the collapse of a housing market bubble and the ensuing financial
crisis led to the steepest drop in real GDP and the largest increase in the unemployment rate
since the Great Depression. The fallout from these events on credit availability, balance sheets,
and confidence continues to weigh on aggregate demand, restraining the pace of recovery in the
housing market, firms’ willingness to hire and invest, and spending by consumers and state and
local governments. In addition, these demand effects have probably diminished the productive
capacity of the economy.
In this paper, we examine recent developments in potential output and discuss the
implications for monetary policy. We begin our analysis by using a standard production-
function framework and an unobserved components statistical model to estimate the extent of
supply-side damage in recent years, and to identify the components of aggregate supply where
the damage was most acute. Our results suggest that the level of potential GDP was about 6
percent below its pre-crisis trend in 2013:Q1, with a 95 percent confidence interval ranging from
3.8 to 8.1 percent; the model projects the shortfall to widen to 6¾ percent by 2013:Q4. We also
show that, in real time, this modeling apparatus would have recognized the decline in potential
output relative to its pre-crisis trend only gradually, and only after some large revisions to the
estimate of potential output since before the crisis, the downside surprise with respect to actual
output has been considerably greater; as a result, the model sees actual output as currently still
running significantly below its potential at present.
In terms of the components of aggregate supply, the model estimates the largest losses to
be in trend productivity, reflecting both a steep decline in capital accumulation and slower
growth in multifactor productivity. However, the growth in trend labor input also appears to
Page 3 of 61
have slowed in recent years, suggesting that the deep recession resulted in some structural
damage in the labor market. Motivated by the employment leg of the Federal Reserve’s dual
mandate, we examine in more detail the evidence pertaining to labor market damage. Our
analysis on this point suggests that there has been a modest rise in the natural rate of
unemployment and a steepening of the downtrend in labor force participation in recent years, but
the evidence on the likely persistence of labor market damage is less conclusive.
We then turn to a more general issue—the somewhat blurred distinction between
“supply” and “demand” shocks, and the resulting potential for monetary policy to mitigate
endogenous adverse developments in supply-side conditions. In many macroeconomic models,
aggregate supply shocks are viewed as exogenous—and specifically as outside the range of
influence of monetary policy. However, if—as we suggested earlier—some elements of
aggregate supply are significantly influenced by changes in aggregate demand, they may also be
susceptible to influence from monetary policy. Capital spending provides the clearest example,
and we include a simple simulation showing how monetary policy can mitigate the loss to the
capital stock and thus aggregate supply that results from a broad-based shock to aggregate
demand. But as discussed by Blanchard and Summers (1986), Ball (1999) and Blanchard (2003)
some time ago, and investigated more recently by Stockhammer and Sturn (2012) and Erceg and
Levin (2013), demand shocks can also have long-lasting effects on unemployment duration and
labor force attachment that, in principle, activist monetary policy might be able to check. And
finally, demand shocks and monetary policy may even be able to influence potential output over
the medium term through their effects on new business formation and research and development.
In the final section of the paper, we examine the implications of this blurring for the
“optimal” conduct of monetary policy. Taken alone, the possibility that potential output will be
Page 4 of 61
affected by adverse demand shocks through hysteresis-like effects leads optimal monetary policy
to be more activist, in order to mitigate the possible damage to the current and future supply side
of the economy. However, other considerations may militate toward restraint in the conduct of
monetary policy; these considerations include concerns about the unintended effects an
unusually aggressive monetary policy might have on financial stability or the dynamics of
inflation. Thus, in an uncertain world, a policymaker’s choice of policy will depend not only on
the extent to which he or she believes a demand shock is likely to affect potential GDP and
employment, but also on his or her view of the risks associated with actively trying to offset
these adverse supply-side developments through accommodative monetary policy.
I. Recent Supply-Side Developments: Evidence from a State-Space Model
In the wake of the financial crisis, real GDP in the United States fell 4¼ percent from its
cyclical peak in the fourth quarter of 2007 to its trough in the second quarter of 2009, and the
unemployment rate rose sharply, reaching 10 percent by late 2009. Moreover, the subsequent
recovery in economic activity has been sluggish by historical standards, with real GDP in 2013
only modestly above its pre-recession peak and the unemployment rate still nearly 3 percentage
points higher than it was through most of 2007. These features of the recession and recovery,
coupled with observations by Reinhart and Rogoff (2010) and Cerra and Saxena (2008) that past
financial crises tended to be followed by persistent shortfalls in real GDP, have led many to
speculate that the financial crisis and ensuing recession have left a permanent imprint on the
productive capacity of the U.S. economy.
2
2
See, for example, CBO (2012). Similarly, the European Central Bank (2011) estimates that the financial crisis led
to a permanent drop in the level of potential output in the Euro area, but argues that the effects on potential growth
going forward are more uncertain.
Page 5 of 61
As a first step in assessing the implications of the events of recent years for potential
output, we examine the behavior of real GDP and unemployment in the context of a simplified
version of Okun’s Law: ΔU = α (Δq* – Δq), where ΔU is the change in the unemployment rate,
Δq and Δq* are the growth rates of actual and potential GDP, and α is the Okun coefficient,
which is currently thought to be about ½ (Ball, Leigh, and Loungani, 2013). With the
unemployment rate 2.8 percentage points higher in 2013:Q2 than in 2007:Q4 and real GDP
having increased 4.4 percent during that time, this simple rule of thumb suggests that potential
output grew about 10 percent over that period, or roughly 1.8 percent per year. This compares
with an estimated annual growth rate for potential GDP of 2.7 percent using the same
methodology from 2000:Q4 to 2007:Q4.
Of course, this very simple exercise tells us little about the sources of the deceleration in
potential GDP over the past five years; nor does it allow for the possibility that the natural rate of
unemployment has changed over time; moreover, it assumes that a particularly simple version of
Okun’s Law holds without error between benchmark dates (for example, 2007:Q4 and 2013:Q2
in one of the calculations reported above). To allow for a wider array of underlying forces, we
next turn to a richer approach to estimating potential output based on an aggregate production
function. This approach allows us to decompose the estimated changes in potential output into
changes in potential labor input (including changes in the natural rate of unemployment), capital
deepening, and multifactor productivity. As we discuss in more detail later in the paper, our use
of an aggregate production function also avoids the questionable assumptions about welfare
inherent in some of the concepts of potential output derived from dynamic stochastic general
equilibrium models. Finally, the model we use embeds a relationship between economic slack
and inflation—concepts that are close correlates of the two legs of the dual mandate given by the
Page 6 of 61
Congress to the Federal Reserve, and hence critical for the conduct of monetary policy in the
United States.
3
To estimate potential output in a production-function framework, we use a version of an
unobserved components model of the supply-side of the economy developed by Fleischman and
Roberts (2011).
4
In particular, we first define (log) output in terms of the components that
comprise an aggregate production function:
y
t
≡ Σx
it
,
where the x
i
’s include the various components of labor input (e.g., population, labor force
participation, the employment rate, and the workweek), the factors influencing labor productivity
(e.g., capital deepening, labor quality, and multifactor productivity), and a variety of technical
factors that account for the different measurement systems used to construct the data series we
use in estimation. (In addition, actual output is unobserved in the model but is identified by the
comovements of real GDP, real non-farm business output, and real non-farm business income.
5
)
We then specify each element of the production function as the sum of a cyclical component, a
trend component, and an idiosyncratic residual:
x
it
= λ
i
(L) cyc
t
+ x
it
*
+ μ
it
.
Finally, as noted above, we augment the production function equations with a new-Keynesian-
style inflation equation. This equation relates current-period inflation to a survey-based measure
of long-run inflation expectations, lagged inflation, economic slack (as measured by the same
3
A production-function approach is also used by the CBO, the IMF, the ECB, and the OECD in developing their
estimates of potential output. See also Fernald (2012), Basu and Fernald (2008), Clark (1987), and Gordon (2003).
4
The state-space model that we use has been embedded in the Federal Reserve’s large-scale econometric model of
the US economy known as FRB/US. With additional developmental work, partly necessitated by the July 2013
comprehensive revision of the national income and product accounts, the specification of the state-space model
within FRB/US has evolved somewhat away from the one we use here.
5
Nalewaik (2010) shows that elements from the income side of the national income and product accounts have
substantial incremental information content relative to elements from the product side.
Page 7 of 61
cycle variable that appears in the decomposition of each element of the production function), and
changes in the relative prices of energy, food, and imports:
Δp
t
= ωΔp
t
e
+ (1-ω)Δp
t-1
+ βcyc
t
+ Z
t
Γ + ε
t
.
One element of this specification is not entirely standard—our use of a survey-based measure of
long-run inflation expectations—but as discussed by Clark (2011), Del Negro, Giannoni, and
Schorfheide (2013), and Ascari and Sbordone (2013), such measures do an excellent job of
capturing the movements in the low-frequency stochastic trend component of U.S. inflation over
the past fifty years, and thus provide a convenient way to improve the fit of price equations.
6
The decomposition of each element of the production function into an unobserved
individual trend component, an unobserved cyclical component that is common across all
variables, plus an idiosyncratic component, leads us to specify the system as a state-space model.
As in Fleischman and Roberts, we assume that each of the trend variables follows a random walk
with drift,
x
it
*
= α
i,t
+ x
i,t-1
*
+ η
it
.
For some unobserved variables, the drift parameter is constrained to equal zero (the natural rate
of unemployment) or an estimated constant, but for most trend terms the drift parameter is
assumed to follow the AR(1) process α
i,t
= .95 α
i,t-1
+ ε
i,t
. The common cyclical component
follows an AR(2) process,
cyc
t
= δ
1
cyc
t-1
+ δ
2
cyc
t-2
+ ξ
t .
Likewise, we use standard maximum likelihood methods for state-space models (specifically, the
Kalman filter) to estimate the parameters. (See the appendix for additional details of the model.)
6
From 1990 through the present, Δp
t
e
equals the median projection of inflation over the next ten years reported
quarterly in the Survey of Professional Forecasters; from 1980 through 1989, it equals the average expected rate of
inflation ten years ahead reported in the Hoey survey of financial market participants. Prior to 1980, Δp
t
e
is inferred
from the low frequency movements in actual inflation.
Page 8 of 61
Results from the state-space model based on current-vintage data are shown in Table 1.1
and Figure 1.1. According to the point estimates from the model, as indicated in the top row of
the table and the upper-right panel of the figure, potential GDP growth slowed from 2.6 percent
per year in the 2000-2007 period to 1.3 percent per year on average during the past five years, a
somewhat greater stepdown in growth than suggested by the simple Okun’s Law calculation
above. Moreover, the deceleration in potential GDP is even more pronounced during the past
three years, with the average annual change estimated at less than 1 percent. As shown in the
upper left panel of Figure 1.1, the level of potential GDP as of 2013:Q1 now is estimated to be
about 6 percent below the trajectory that appeared to be in place based on data through 2007, and
the model projects the shortfall to widen to 6¾ percent by 2013:Q4.
These figures overstate the likely hit to the supply side going forward, since the model
interprets a substantial portion of the slowdown in potential GDP growth since 2007 as reflecting
one-time adverse shocks to the level of the natural rate, labor force participation, and trend
multifactor productivity. (These level shocks are the η
it
shocks specified above.) Although such
level shocks are assumed in the model to have a permanent effect on the level of potential output,
they do not affect its expected future growth rate.
7
The importance of this distinction can be
seen in line 5 of the table, which reports the estimated growth rate of potential GDP excluding
level shocks; this adjusted rate—which represents the model’s assessment of the underlying rate
of increase in potential GDP once the level shocks have dissipated—is estimated to have slowed
somewhat less markedly in recent years. Moreover, a substantial portion of the slowdown in the
adjusted growth rate since 2007 reflects an unusually slow pace of capital deepening—a factor
7
The α
i,t
shocks are the primary source of variation in the expected growth in potential GDP, with the caveat that
persistent movements in the rate of capital deepening can also influence the expected growth rate.
Page 9 of 61
whose contribution to growth should pick up substantially over time as the recovery in business
investment and the broader economy proceeds.
Lines 2 through 4 of the table and the remaining panels of the figure provide some
evidence from the model on the sources of the reduction in the potential growth rate of the U.S.
economy. The largest contribution to the slowdown in potential output growth is from trend
labor productivity (line 3), reflecting both a sharp decline in the contribution to labor
productivity from capital deepening (capital services per trend employee hour) and smaller
increases in trend multifactor productivity since the financial crisis. The trend growth rate of
labor input (line 2) has also slowed in recent years, according to the model, owing to a modest
increase in the natural rate of unemployment and a steepening of the trend decline in the labor
force participation rate.
8
That said, even with the estimated slowdown in potential growth, the
model’s estimate of the cycle (shown in Figure 1.2) is consistent with a sharp drop in resource
utilization in 2008 and 2009 and only a gradual recovery thereafter. Similarly, the model’s
estimates of the GDP gap and the unemployment gap suggest that the economy is still some
distance away from full employment.
It took some time for these changes to the supply side of the economy to become fully
apparent in the data, and consequently many economists did not initially adjust down their
estimates of potential output in the United States following the financial crisis despite the
international evidence reported by Reinhart and Rogoff (2010) and Cerra and Saxena (2008) for
earlier financial crises. As a way of illustrating the discrepancy between the perceived effects of
the crisis as they unfolded over time versus how they appear today in hindsight, we estimate the
state-space model described above using the data for real activity and inflation that were
8
Although not explicitly accounted for in the model, some of the steepening in the trend participation rate reflects
demographic influences unrelated to the financial crisis or recession. See, for example, Aaronson et al. (2006).
Page 10 of 61
available in early June of each year between 2007 and 2013, and we use the model to generate
estimates of supply-side conditions through the first quarter of the year in question. In addition,
for each of the seven years, we use the estimated model to project the path of potential GDP and
the natural rate from the second quarter of the year forward, conditional on the assumption that
the contribution of capital deepening to growth would gradually return to its historical average.
9
The upper two panels of Figure 1.3 presents the results of this real-time exercise for the
estimated level and rate of change of potential output. Initially, the state-space model did not
interpret the available data as suggesting much of a change in the economy’s overall productive
capacity; in fact, the historical and projected path of potential GDP revised up slightly between
the 2008 and 2010 data vintages.
10
However, subsequent data vintages paint a considerably
darker picture, reflecting not only the surprisingly weak pace of the recovery but also revisions to
historical data. For example, the revisions to the National Income and Product Accounts
released in late July 2010 showed a much larger drop in output in 2008 and early 2009 than
initially reported, and as a result, estimates generated using the data available in June 2011
(green line) show potential GDP expanding only 1.3 percent per year on average in 2008, 2009
and 2010, with growth projected to remain subdued in 2011 and then to rise slowly back to 2
percent over the longer run. With another marked downward revision to historical estimates of
GDP released in July 2011, and with the unemployment rate trending down after late 2010
despite only sluggish GDP growth, the model’s real-time assessments of past and projected
9
Because the state-space model does not forecast capital deepening (in contrast to, say, trend labor force
participation), any real-time projection of potential GDP beyond the current quarter requires some assumption for
the future path of capital services. Similar considerations apply to population, which in these real-time calculations
is assumed to continue rising at the average pace observed over the previous year.
10
Comparing potential GDP estimates based on the June 2008 and June 2009 vintages of data to those based on
subsequent vintages is somewhat difficult because the measures of real output and income used to calculated
potential GDP were rebased from 2000 dollars to 2005 dollars beginning in July 2009. In figure 1.3, the 2008 and
2009 real-time estimates of the level of potential GDP are rescaled by a constant multiplicative factor that has the
effect of making these vintages’ historical estimates of real GDP from the late 1940s through the late 1990s closely
match those published at a later date.
Page 11 of 61
supply-side conditions made in mid-2012 and mid-2013 deteriorated further (orange and black
lines).
Similarly, as illustrated in the lower left panel of Figure 1.3, the model did not initially
detect much of an increase in the natural rate of unemployment. In particular, although the
model’s estimate of the natural rate did move up following the onset of the crisis, it was still no
higher than 5 percent as late as mid-2009 (blue line). However, the real-time estimates of the
natural rate of unemployment jump sharply between mid-2009 and mid-2010 (red line), as the
unemployment rate continued to rise sharply even after the recession ended. Moreover, during
that period, inflation did not decline much even though the real-time estimates of the
unemployment gap remained elevated (bottom right panel). As a result, the model’s estimate of
the sensitivity of inflation to the unemployment rate (the “slope of the Phillips Curve”)
diminished.
11
While other analysts also marked down their estimates of potential output growth, the
timing and the extent of these markdowns varied considerably (Table 1.2). The OECD and IMF
adjusted down their estimates of potential growth quickly, and by sizable amounts, on the
explicit assumption that financial crises are invariably followed by permanent supply-side
losses.
12
For example, the IMF slashed its estimate of potential output growth in 2009-10 from 2
percent to below 1 percent between late 2008 and late 2009. Although the incoming productivity
data in the United States did not look particularly dark at first—and the IMF subsequently scaled
back their estimate of the losses some—later vintages of NIPA data in the US came to validate
11
The real-time estimates also show a marked revision to the pre-crisis level of the natural rate in 2010. This shift is
a result of historical revisions to aggregate output and income released in July 2009, which altered the historical co-
movements of these series with the unemployment rate and inflation, thereby resulting in higher estimates of the
natural rate beginning back in the late 1990s.
12
Such judgmental assessments raise the question: Should the specification of a “good” state-space model allow for
discrete shifts in parameters and shocks following the onset of a financial crisis? While the answer to this question
is almost certainly “yes” in principle, the practical difficulty of doing so is quite high owing to the rarity of such
crises domestically and the uncertainties of calibrations based on international experience.
Page 12 of 61
the earlier IMF pessimism to some extent. For its part, the OECD lowered its estimate of
potential output growth for 2009-10 to about 1½ percent in early 2009, where it has remained
since. (Interestingly, both the IMF’s and OECD’s current estimates for potential growth in 2012
are 1¾ percent, noticeably above that from the state-space model.) In contrast, U.S. officials
were somewhat slower to recognize the decline in potential output growth. For example, the
Congressional Budget Office did adjust down its estimate of potential growth from 2½ percent to
2 percent in August 2009, but it subsequently made further downward adjustments to its
estimates for those years, in a pattern not unlike the vintage-based results from the state-space
model. And the Council of Economic Advisers’ estimate for those years was still at 2½ percent
in 2010. The one private-sector forecaster we surveyed, Macroeconomic Advisers, also lowered
its estimate of potential growth in 2009-10 relatively promptly to 1.2 percent by late 2009.
Of course, considerable uncertainty attends all of these estimates of potential output
growth and the natural rate of unemployment. As indicated by the blue shaded region in the
upper-right panel of Figure 1.1, the 95 percent confidence band around the state-space model’s
current estimate of the four-quarter change in potential real GDP is nearly ±1 percentage point,
while the comparable confidence band around the estimated natural rate of unemployment
(middle-left panel) ranges from about 4½ percent to 7 percent. Moreover, these ranges
undoubtedly understate the true uncertainty surrounding our model-based estimates as they do
not account for uncertainty about data revisions, the specification of the state-space production-
function model, or the possibility that other altogether-different frameworks might yield different
estimates of supply-side damage. (Indeed, one suggestive piece of evidence consistent with
these other sources of uncertainty being significant is the fact that in the real-time estimates
presented in Figure 1.3, data revisions in 2009 and 2010 accounted for a sizable portion of the
Page 13 of 61
downward adjustments to the state-space model’s estimates and projections of potential output
growth.)
The importance of uncertainty about the proper specification of the state-space model can
be illustrated by reporting the results from alternative versions that seem perfectly reasonable on
their face. To this end, we estimate two versions with alternative specifications for inflation
dynamics—one in which the coefficient on the common cycle term in the inflation signal
equation is allowed to shift discretely starting in 1995, and another in which the inflation
equation is dropped altogether. These two variations can help shed light on how the information
content of inflation for slack may have changed in recent years.
13
In addition, we explore the
sensitivity of our state-space estimates to the measures of output used in the model by
considering a version that drops the signal equation for real nonfarm income.
As shown in the upper panel of Figure 1.4, the inflation-related changes in specification
do not markedly change the estimated year-to-year movements in potential GDP growth
(including the effects of level shocks). In contrast, the model’s estimates of the natural rate are
much more sensitive to changes in the inflation equation, as these changes result in either a
modestly higher average level of U* (the result with a time-varying slope) or a much higher level
(the result when the model does not condition on inflation at all). Estimates of the natural rate
are also somewhat sensitive to the measures of aggregate output and income included in the
state-space model, particularly prior to 2000. Moreover, and unlike the situation with the
inflation alternatives, dropping nonfarm income from the model has a noticeable effect on the
13
In the version of the state-space model that allows for a shift in the Phillips curve slope, the estimated coefficient
on the cycle term drops markedly starting in the mid-1990s, falling to 0.04 from 0.16 prior to 1995.
Page 14 of 61
estimated quarter-to-quarter pattern of potential GDP growth and, among other things, results in
a somewhat higher estimate of trend growth in the past few years.
14
Not surprisingly, frameworks that differ from ours in more fundamental ways can also
yield different estimates of potential output and economic slack. As an example, Figure 1.5
reports two alternative estimates of the output gap from the Board’s DSGE model of the U.S.
economy (EDO)—one a production-function measure comparable in spirit to that generated by
the state-space model, and the other based on a Beveridge-Nelson statistical estimate of the long-
run trend in aggregate capacity.
15
Compared to the baseline version of the state-space model,
both of the EDO measures show a much narrower degree of economic slack in recent years,
especially in the case of the production-function measure; this latter result largely reflects EDO’s
assessment that the trend in aggregate hours has fallen steeply since the middle of the 2000s. Yet
another approach to estimating the output gap is that taken by Borio et al (2013), who argue that
adding information about the financial cycle in the state-space model yields more robust readings
on aggregate resource utilization because doing so allows policymakers and others to take into
account concerns about future financial imbalances; they also advocate ignoring inflation when
filtering the data on the grounds that the Phillips curve has become too flat to be useful. Borio et
al.’s measures imply much more damage to the supply side in recent years than do the estimates
generated by our baseline state-space model.
Looking forward, the trajectory of potential GDP is even more uncertain. To some
extent, this uncertainty reflects uncertainty about the future pace of technological change. As
14
Fernald (2012) provides a different perspective on the uncertainty associated with empirical specification. Using
a methodology that is similar to the CBO’s but with different assumptions about underlying technology and capital
growth, he finds a somewhat greater slowing in potential output growth following the financial crisis than does the
CBO.
15
For details on the current version of the EDO model (including the approaches used to estimate potential output),
see Chung, Kiley and Laforte (2012). Also, see Beveridge and Nelson (1981), Clark (1987), and Haltmaier (2012)
for a discussion of other applications of univariate time-series analysis to the estimation of trend output.
Page 15 of 61
Byrne, Oliner, and Sichel (2013) illustrate, views about the future pace of productivity growth—
and especially about importance of information technology—vary considerably, perhaps
bounded by Cowen (2011) and Gordon (2013) on the low end and by Baily (2013) and
Brynjolfsson and McAfee (2011) on the high end. In addition, the future path of potential output
may be sensitive to the fundamental causes of the reduction in potential output following the
financial crisis, over which there remains much debate.
In this context, and without taking any stand on the relative odds of their coming to pass,
Figure 1.6 presents three possible paths for potential output over the remainder of this decade, all
of which build off the estimated state-space model. The lower red dashed line shows a scenario
in which various trend growth rates continue at their last estimated values; this scenario also
assumes that the natural rate of unemployment remains at its recent estimated level. The
medium-growth scenario (the blue line) assumes instead that trend MFP growth will gradually
move back to its pre-crisis average rate of about 1 percent per year, that the trend rate of decline
in the labor force participation rate will moderate to a bit less than 0.2 percentage point per year,
and that the natural rate gradually returns to 5.4 percent. Thus, the medium-growth scenario
incorporates a highly persistent reduction in the level of potential GDP but not a permanent
reduction in potential output growth. In contrast, the high-growth scenario assumes a gradual
return to pre-crisis trends in the levels of capital deepening, trend MFP, and trend labor force
participation. Later in the paper, we will explore the role that monetary and fiscal policy might
play in influencing the likely path of potential GDP going forward.
II. Evidence from the Labor Market
The production-function approach provides a useful high-level perspective on the
evolution of potential output over time, and generates suggestive evidence about the sources and
Page 16 of 61
magnitudes of losses that may have been associated with the financial crisis, the deep recession,
and the persistently slow recovery. In this section, we delve more deeply into the possibility of
supply-side damage in labor markets, which carry special significance in light of the full-
employment leg of the Federal Reserve’s dual mandate. Specifically, we provide our take on the
evidence regarding three potential sources of labor market damage that have been the focus of
much recent commentary: (1) difficulties in reallocating labor across different segments of the
economy (industry, occupation, or geographic) associated with the distribution of the demand
shock caused by the financial crisis and deep recession; (2) a more general deterioration in the
efficiency of the matching process between available workers and available jobs; and (3) long-
term damage in labor markets (often referred to as hysteresis) associated with the substantial rise
in the number of long-term unemployed and a possible reduction in the employability of affected
workers.
Reallocation
Given that the financial crisis involved the bursting of a bubble in housing prices and a
steep drop in activity that was concentrated in the residential construction sector, it is not
surprising that construction employment experienced an outsized decline in late 2007 and 2008
relative to many other industry sectors (Figure 2.1). Similarly, employment in industries tied to
housing, including mortgage finance, real estate, and construction-related manufacturing also
dropped sharply at the outset of the recession.
It is worth noting, however, that recessions always affect some industries to a greater
extent than others, and in the past these imbalances have typically faded as the overall economy
recovered. As shown by the black line in Figure 2.2, the reduction in aggregate activity
associated with the financial crisis was initially distributed unevenly and led to a sharp increase
Page 17 of 61
in the variance of employment change across broad industry categories in 2008 and 2009.
16
To
shed further light on the sources of the rise in dispersion, the blue and green lines in the figure
present a decomposition of the variance in the percent changes in employment into two pieces:
the part associated with cyclical dynamics (the blue line) and the part associated with unusual—
or excess—dispersion (the green line).
17
The statistical procedure we use here interprets the
overwhelming bulk of that spike as caused by the very deep recession. In contrast, excess
dispersion increased only moderately during the recession, and by less than it had during several
other episodes in the past 50 years. A similar exercise for employment changes across states
comes to much the same conclusion (see Valletta and Kuang, 2010).
While Figure 2.2 indicates that the recession did not initially result in an unusual amount
of dispersion in employment changes across industries relative to previous recessions, Figure 2.3
suggests that the industry-specific shocks to labor demand in the recent recession were more
persistent than in the past. In particular, this figure plots the variance of the cumulative change
in industry employment shares for five past recessions from the business cycle peak up to six
years after the peak. Consistent with Figure 2.2, the first few quarters following the peak do not
look especially different in the recent recession than in the past. Beyond the first year, however,
the dispersion in the change in employment shares since late 2007 rises sharply and by two years
out is well above that following any of the earlier business cycle peaks. Thus, the change in the
16
Specifically, the black line plots the share-weighted variance of the quarterly growth rates of payroll employment
across 14 industry categories.
17
To decompose the variance into its cyclical and noncyclical components, the quarterly percent change in each
industry’s employment relative to the change in total employment is regressed on a constant term and a measure of
the business cycle. Standard variance decomposition methods are then used to decompose the overall variance of
industry employment changes into the parts associated with differences in trend growth across sectors, differences in
the normal degree of cyclicality across sectors, and differences in the residuals across sectors. The cross-terms that
include the cyclical term are allocated to the variance associated with the business cycle. Other cross-terms are
allocated to the idiosyncratic variance.
Page 18 of 61
industry composition of employment associated with the recent recession may have required a
more significant amount of labor reallocation than in earlier periods.
18
These measures of sectoral imbalances rely on fairly aggregate industry or geographic
definitions and thus may be too crude to capture the full extent of the reallocation across more
narrowly defined industries, occupations, or geographic areas that was engendered by the
financial crisis. An alternative way to assess sectoral reallocation is to focus on permanent job
loss more generally, on the grounds that permanent separations of any type are potentially
associated with substantial costs in terms of relocation and lost human capital that could slow the
pace at which workers find new jobs.
19
Figure 2.4 shows that the rate of permanent job loss—
the red line—rose sharply during the recession, briefly reaching a level more than twice as high
as it reached in the aftermath of the relatively mild recession during the early 2000s and as high
as that during the 1982 recession. Although the rate of permanent job loss has trended
downward during the past four years and is currently close to its pre-recession level, the stock of
persons still unemployed following a permanent job loss (the black line) remains noticeably
higher than prior to the recession. This suggests that many permanent job losers continue to
experience difficulties in finding a new job, consistent with the hypothesis that structural
unemployment may have increased.
20
That said, the stock of permanent job losers has now
18
In computing the variance of the cumulative changes in industry employment shares, we first removed the long-
run trends in the shares using a Hodrick-Prescott filter with λ set equal to 1,024,000 (so that we only removed the
very long-run trends). Without detrending, the variances from the 2007 peak are still generally above those in the
earlier periods two to three years out, but fall below the dispersion in employment shares following the 1981 peak
after about four years.
19
See, for example, Loungani and Rogerson (1989) and Figura and Wascher (2010). In addition, the need to
reallocate physical or organizational capital can lead to reductions in productivity and higher unemployment,
especially if the displaced capital is highly specific to the affected industry or firm (Ramey and Shapiro, 2001).
20
Two potential sources of a continued elevated stock of permanent job losers are a reduction in labor mobility as a
result the sharp drop in house prices and associated increase in homeowners who are “underwater” and the
possibility that displaced workers would resist lower wage offers (wage rigidity). Researchers who have studied
housing markets and migration have thus far found little empirical support for an effect of house lock on labor
mobility (see, for example, Molloy, Smith, and Wozniak, 2013; and Valletta, 2013). There is less evidence on the
Page 19 of 61
moved down about half way to its precession level—roughly proportionate to the improvement
in overall unemployment since its peak—and it seems likely that further improvements in
economic activity and job opportunities will lead to further reductions in the stock of permanent
job losers.
Matching efficiency and the Beveridge Curve
Proponents of the view that the rate of structural unemployment has risen also point to
the Beveridge curve—the relationship between the unemployment rate and job vacancies—as
evidence for this view.
21
Through long stretches of time, the vacancy rate and the
unemployment rate have traced out a seemingly stable schedule that is often interpreted as
reflecting changes in aggregate demand playing out in the context of a labor market exhibiting
roughly constant structural unemployment. As shown in Figure 2.5, which measures job
vacancies using data from the “Job Openings and Labor Turnover Survey” or JOLTS, the data
points from late 2000 through 2008 represented one such period in which the Beveridge curve
appeared to be stable. Beginning in mid-2009, however, it became apparent that the Beveridge
curve had shifted to the right. During the past three years, the vacancy rate has been rising and
the unemployment rate has been falling, consistent with the usual negative relationship between
these two series, but the locus of points traced out has been distinctly to the right of the one that
prevailed during the 2000s.
It would be tempting to conclude from the rightward shift in the Beveridge curve that
structural unemployment had increased around the time of the financial crisis and the onset of
the ensuing recession—and that may ultimately prove to be the right conclusion. But before
effect of wage rigidity on the speed of labor reallocation in recent years, although Daly et al. (2013) find that
downward nominal wage rigidity increased during the recession.
21
See, for example, Hassett (2013). The use of the Beveridge Curve to help distinguish between structural and
cyclical increases in unemployment was also prominent in the debate between Lilien (1982) and Abraham and Katz
(1986).
Page 20 of 61
drawing that conclusion, it is important to note that the Beveridge curve can shift for a variety of
reasons, some of which are cyclical rather than structural in nature.
22
Indeed, as the labor market
has improved following past recessions, the vacancy-unemployment locus has typically
exhibited a counter-clockwise loop (that is, unemployment has eventually declined more than
would be apparently consistent with the normal stable downward-sloping schedule). One factor
that could generate such looping behavior is extended unemployment insurance benefits. All
else equal, when extended UI benefits become available, some unemployed individuals may
experience a reduction in the incentive to maintain their intensity of job search, while other
individuals who would otherwise have dropped out of the labor force may be induced to report
themselves as unemployed (and undertake sufficient search to qualify as such) in order to receive
benefits. For both reasons, the measured unemployment rate associated with any given job
vacancy rate may increase.
23
During the recovery phase, as the availability of extended UI
benefits is curtailed and eventually eliminated, the process unwinds, and the unemployment rate
comes down by more than one would predict based only on the job vacancy rate. In the most
recent episode, extended UI benefits have been available since 2008. During the past two years,
however, availability of extended UI benefits has been greatly curtailed, with the number of
recipients now down to about one quarter of its peak. Despite that fact, and as can be seen in the
figure, there has been no apparent shift back to the left in the Beveridge curve, lending further
credence to the view that structural unemployment may have increased.
24
22
See Diamond (2013) for an extensive treatment of the Beveridge Curve and a discussion of the relevance of the
recent evidence for assessing the extent to which the currently high level of unemployment is structural or cyclical
in nature.
23
Some recent evidence on this point is provided by Farber and Valletta (2013), who find that extended UI benefits
reduced the exit rate from unemployment and increased the duration of unemployment spells. They also find that
the effect on unemployment exit and duration stemmed primarily from a reduction in exits from the labor force
rather than from a decrease in the job finding rate.
24
In addition, Davis, Faberman, and Haltiwanger (2012) argue that persistently weak demand has caused employers
to be more selective in choosing whom to hire, resulting in a decrease in the ratio of hires to vacancies and thus an
Page 21 of 61
Other analysts (for example, Lazear and Spletzer, 2012; Sahin et al., 2012) have
constructed measures of mismatch from disaggregated unemployment and vacancy data. The
Lazear-Spletzer and Sahin et al. measures of industrial mismatch, shown in Figure 2.6, rose
sharply during the recession but began to fall back in 2010 and by early this year were at or close
to their levels prior to the recession—a development that Lazear and Spletzer interpreted as
suggesting “that changes in industrial mismatch are cyclical, rather than structural.” Their
occupational mismatch indexes show a similar pattern, although the Sahin et al. measure is still a
little on the high side. In contrast, Sahin et al. find essentially no evidence that geographic
mismatch (not shown in the figure) increased during the recession or that it is currently above
normal levels.
25
Of course, there may also be movements in the Beveridge curve that represent changes in
the efficiency of the job matching process more broadly (that is, not specifically associated with
occupational, industry, or geographic mismatch), and which could be viewed as stemming from a
change in structural unemployment. In this regard, Barnichon and Figura (2013) propose a
model that attempts to decompose movements in the aggregate Beveridge Curve into various
components and, in particular, to isolate the outward shift associated with a decline in matching
efficiency.
26
As indicated in Figure 2.7, the estimates from this model suggest that structural
unemployment increased by nearly 1½ percentage points between the onset of the recession and
the end of 2011. However, the model also suggests that matching efficiency has begun to
improve more recently, although it remains well below where it was prior to the recession.
outward shift in the Beveridge Curve without an increase in structural unemployment (assuming that the effect fades
as real activity recovers).
25
We thank Aysegul Sahin for providing updated estimates of the mismatch indices presented in Sahin et al.
26
To calculate their measure of matching efficiency, Barnichon and Figura regress the job-finding rate of the
unemployed on the ratio of vacancies to unemployment. The residuals from the regression represent shifts in the
Beveridge curve associated with changes in the efficiency of job matching.
Page 22 of 61
Using a somewhat different framework, Daly et al. (2012) combine the Beveridge curve with a
job creation curve and estimate that the natural rate of unemployment was about 6 percent at the
end of 2011, about 1 percentage point above its level prior to the recent recession.
Hysteresis
An important and unusual aspect of the recent recession and the subsequent slow
recovery—and one that heightens the risk that structural labor market damage may have been
sustained already or may emerge—is the sharp increase in long-term unemployment since the
onset of the financial crisis. As shown in Figure 2.8, the number of individuals unemployed for
more than 26 weeks as a share of the labor force rose to 4.3 percent in April 2010 and has since
fallen only to 2.7 percent, as compared with ¾ percent in 2007; likewise the share of the
unemployed who have been out of work more than 26 weeks peaked at about 45 percent in early
2011 and remains above one-third today, well above the levels experienced during any previous
post-World War II recession. Long-term unemployment is of particular concern because
individuals out of work for extended periods of time may find that their skills, reputations, and
networks deteriorate, resulting in a persistently higher level of structural unemployment or a
steeper downtrend in the labor force participation rate. Although such effects do not appear to
have been important in the United States in the past, they have been evident in other advanced
economies and the unprecedented durations of unemployment during the present episode in the
United States may reduce the relevance of historical experience in this country.
It is well known that individuals with longer spells of unemployment find it more
difficult to become reemployed. For example, as indicated in Figure 2.9, job finding rates for the
long-term unemployed are nearly always well below those for individuals with shorter
unemployment spells. In the past, however, researchers have found it difficult to separate the
Page 23 of 61
effects of unobserved heterogeneity in the individuals experiencing long spells of unemployment
from duration dependence. To address this issue, Kroft, Lange, and Notowidigdo (2013)
recently conducted an experiment and found that, all else equal, potential employers were much
less likely to call back job applicants with longer spells of unemployment than applicants with
shorter spells, evidence that is consistent with duration dependence in unemployment. Although
the aggregate implications of this finding are unclear, under some interpretations employers’
aversion to long unemployment spells could result in hysteresis.
If hysteresis-type effects were taking hold, one might expect to see an improvement in
the job-finding rates for the unemployed with shorter durations but not for those with longer
spells of unemployment. Thus far, the evidence on this point seems mixed. Job-finding rates
have edged up, on balance, in recent years for those unemployed less than 27 weeks and by more
than the job finding rates for the longer-term unemployed. However, the differences are not
large and the data on these flows are fairly noisy. At the same time, as indicated in Figure 2.10,
the rate of exit from the labor force among the long-term unemployed has risen since late 2009,
but broadly speaking, no more than has the rate for those with shorter unemployment spells;
moreover, this upward drift reverses an earlier move in the other direction--a pattern that seems
consistent with the earlier increases and more recent declines in the length of eligibility for
extended unemployment insurance benefits.
There have also been some concerns that the availability of disability insurance (DI)
would induce a larger number of those with poor job prospects to permanently leave the labor
force—concerns that were based in part on evidence that those receiving disability payments
tend to remain out of the labor force until retirement (Autor and Duggan, 2006). However, as
indicated in Figure 2.11, the proportion of DI recipients has deviated only slightly in recent years
Page 24 of 61
from its longer-term upward trend. In addition, Mueller, Rothstein, and von Wachter (2013) find
little evidence that the expiration of UI benefits causes individuals to move onto DI rolls. That
said, applications for disability insurance have risen noticeably in recent years and, if approved,
could cause cyclically-induced exits from the labor force to become permanent.
Inflation
Finally, some observers point to the absence of an ongoing steep deceleration in wage
and price inflation as evidence that there has been a noticeable increase in structural
unemployment.
27
However, the results from the state-space model provide another
interpretation, namely that changes in inflation are less informative about labor market slack (and
resource utilization more generally) than in the 1970s and 1980s because of the substantial
flattening of the Phillips curve over the past two decades. Moreover, other factors, such as
downward nominal wage rigidity and well-anchored inflation expectations, appear to be more
likely explanations for the recent behavior of wages and prices. For example, Daly et al. (2013)
present a model in which downward nominal wage rigidity reduces the responsiveness of wage
inflation to the unemployment gap (thus flattening the Phillips curve).
28
They then go on to
show that the behavior of wages since 2007 is well explained by the model given the sizable
degree of downward nominal wage rigidity observed during this period. Similarly, Del Negro,
Giannoni, and Schorfheide (2013) show that in a standard DSGE model with a significant degree
of price rigidity, inflation expectations remain fairly stable causing inflation to depend more on
expected future marginal costs than on economic slack. Of course, the stability of long-run
inflation expectations is not a fundamental property of the economy that monetary authorities
27
See, for example, Gordon (2013), who argues that the relationship between short-run unemployment and inflation
has been stable in recent years, and thus that the sharp rise in long-term unemployment represents an increase in
structural unemployment.
28
As these authors note, this research builds on previous work by Tobin (1972) and Akerlof, Dickens, and Perry
(1996).
Page 25 of 61
can exploit forever, but rather something that can be taken advantage of only cautiously and only
for a time (although that time could be quite long).
Summing up
In the end, we see the evidence of recent years as suggesting that the natural rate of
unemployment may have moved up between ½ and 1½ percentage points since the onset of the
recent recession, roughly in line with the estimates from the state-space model. However, the
evidence also suggests that the factors leading to this increase have begun to reverse and that
further increases in aggregate demand might therefore bring about further healing in the labor
market. Such ultimately transitory damage likely has only modest (at most) implications for
inflation, given the apparent stability of long-run inflation expectations and the estimated flatness
of the Phillips curve. For this reason, policymakers may wish to consider whether it would better
to focus only on those shifts in supply-side conditions that are expected to persist over the long
run when assessing the amount of economic slack that policy seeks to close.
As shown in Figure 2.12, other measures of labor market performance—including the
National Federation of Independent Business’ measure of jobs hard to fill and the Conference
Board’s measure of job availability—have moved broadly in line with the state-space model’s
estimate of the unemployment gap, suggesting that our measure of labor market slack—and thus
our measure of the natural rate of unemployment—is not out of line with the views of
households and businesses.
In addition, while the labor force participation is clearly on a
longer-run downtrend caused by the aging of the U.S. population, it seems likely that at least
some of the currently low level of the participation rate is associated with weak aggregate
demand.
29
In our estimation, however, there is a big wildcard attending the labor market
29
There is a wide range of views as to how much of the decline in participation in recent years reflects cyclical
influences. In particular, Erceg and Levin (2013) estimate that cyclical factors account for about 2 percentage points
Page 26 of 61
outlook: the possibility of hysteresis effects associated with the continued high level of long-term
unemployment. In particular, it seems quite plausible that continued lengthy spells of
unemployment could lead to permanent damage in the productivity or employability of those
who remain willing to work, and could lead others to throw in the towel and permanently exit the
labor force.
III. Some Policy Implications of Recent Supply-Side Developments
Although considerable uncertainty attends any estimate of potential output and
employment, the preceding analysis strongly suggests that the U.S. economy has experienced
significant supply-side damage since 2007; broadly speaking, these results are consistent with
the now-conventional claim that major financial crises tend to reduce a nation’s productive
potential. However, we argue in this section that the implications for monetary policy may differ
sharply from what is commonly presumed because much of the supply-side damage could be an
endogenous response to weak aggregate demand. If so, then an activist monetary policy may be
able to limit the amount of supply-side damage that occurs initially, and potentially may also
help to reverse at a later stage such damage as does occur. By themselves, such considerations
militate toward a more aggressive stance of policy and help to buttress the case for a highly
aggressive policy response to a financial crisis and associated recession. In section 4, we discuss
other considerations that may incline policymakers toward a less aggressive policy response.
Contrasting views
In setting monetary policy, central banks have traditionally tried to distinguish between
trend and cyclical movements by disentangling the effects of exogenous “supply” shocks (which
are assumed to influence the economy’s long-run equilibrium) from the effects of “demand”
of the decline in labor force participation since 2007, while Aaronson, Davis, and Hu (2012) put the cyclical decline
at roughly 1 percentage point, and Hornstein (2013) finds only a small participation rate gap at present.
Page 27 of 61
shocks (which are assumed to drive the economy away from its steady state). The rationale for
this distinction is presumably rooted at least partly in the long-run neutrality of monetary policy:
However important it may be in influencing the paths of real variables in the short run, monetary
policy cannot affect output, employment, or unemployment once prices have fully adjusted and
the effects of other nominal rigidities have faded away. For this reason, monetary policymakers
have to accept the real long-run equilibrium of the economy as something that is determined
outside the sphere of monetary influence, and they need to recognize that it would be fruitless or
even outright damaging to seek a different set of real outcomes. (See Barro and Gordon, 1983.)
The standard textbook presentation of a vertical Phillips Curve has this flavor: In the long
run, output must return to a level that is determined by the location of the vertical aggregate
supply curve, and is invariant with respect to the conduct of monetary policy. The aggregate
demand curve may be buffeted by factors over time, including the stance of monetary policy, but
in the long run, the location of the aggregate demand curve matters only for the value of the
equilibrium real interest rate consistent with stable inflation.
30
While the sharp separation between supply shocks and demand shocks—and the
identification of the first with circumstances that monetary policymakers must accept as given
and the second as factors that they may be able to usefully counteract—is characteristic of
particularly simple models, it greatly oversimplifies the real world. As Blanchard and Summers
(1986) noted many years ago in the European context, weak real activity may give rise to long-
lived hysteresis effects in labor markets, thereby providing a strong motivation for governments
30
Dornbusch and Fischer (1978), the first edition of their macro text, describes the expectations-augmented Phillips
curve in this manner (pp. 404-405) and specifically references a long-run vertical Phillips curve (page 410). For the
first known presentation of the standard textbook vertical supply curve to the Federal Open Market Committee, see
page 25 of the document provided at:
http://www.federalreserve.gov/monetarypolicy/files/FOMC19831115material.pdf
. To be sure, ideas along these
lines had been presented and discussed among the staff at the Board for more than a decade. An early formulation
came in the paper that Robert Lucas presented at a 1971 conference hosted at the Board (Lucas, 1972).
Page 28 of 61
to implement policies (fiscal or otherwise) to both check the magnitude of economic downturns
and so limit the supply-side damage that occurs, and to later boost the pace of activity as the
economy recovers to repair that damage that has occurred. Ball (1999) subsequently expanded
on this idea by examining cross-country evidence on the role of monetary policy in influencing
the magnitude of unemployment hysteresis effects, and concluded that policy-related supply-side
effects were substantial for many European economies—a conclusion that has been reaffirmed in
empirical work by Stockhammer and Sturn (2012).
We would go beyond this literature, however, and argue that the potential endogeneity of
supply-side developments extends well beyond the labor market, and includes such factors as
multifactor productivity and capital deepening. To this end we review below several
mechanisms that all have the characteristic of blurring the distinction between supply and
demand, and therefore prompt a careful consideration of the factors that monetary policy must
accommodate versus those it can counteract.
Before describing these mechanisms, we should note that the statistical methods
commonly used to distinguish “cycle” from “trend” may exacerbate the blurring problem in
severe recessions and slow recoveries. Most if not all of these statistical methods identify the
“trend” with low-frequency movements in the variables of interest; the remaining movements are
assumed to be cycle or noise. In a typical model of this type, the dividing line between
“cyclical” frequencies and “trend” frequencies is generally something like five or six years. That
distinction may be appropriate for the dynamics of most recessions, but the adjustment of labor
force participation, the unemployment rate, and productivity to the events of the last few years
arguably will play out over an even longer span of time. For example, the recession and slow
recovery may impair job matching and other aspects of labor market functioning for quite a few
Page 29 of 61
years. Moreover, these same conditions may lead a significant number of older workers to drop
out of the labor force permanently at an earlier age than otherwise would have occurred, thereby
depressing the participation rate for possibly a decade or more. In both cases, the distinction
between cyclical and “trend” movements in the participation rate and related variables is not as
clear as would be suggested by standard filtering methods.
Three mechanisms for blurring
Among the mechanisms blurring the delineation between the factors that monetary
policymakers must accept versus those they can influence are the potential effects of weak
aggregate activity on potential labor supply. As many policymakers and outside analysts have
noted and as we discussed in Section 2, the unusual length and severity of the Great Recession,
together with the fact that unemployment has been atypically concentrated among the long-term
unemployed, seem likely erode the skills and workforce attachment of some unemployed
persons. Historically, there has been much less evidence of hysteresis in US labor markets than
in European ones, but, as we noted earlier, the severity and unprecedented characteristics of the
recent recession suggest the possibility that the United States will not remain free of hysteresis-
type effects this time. In principle, hysteresis in labor markets could cause a period of slack
demand to have long-lasting adverse implications for the productive capacity of the economy.
Accordingly, the ultimate effects of a financial crisis on the potential supply of labor could
depend critically on the degree to which monetary policy can limit the initial contraction in real
activity, and the speed with which it is able to restore aggregate demand to its normal and
sustainable level.
A second channel through which persistent weak aggregate demand could affect
aggregate supply involves some aspects of multifactor productivity. Evidence suggests that new-
Page 30 of 61
business formation suffers disproportionately during business-cycle downturns, and it is certainly
the case that the annual number of start-ups has fallen noticeably since 2007 (upper panel of
Figure 3.1). Moreover, employment growth at young firms has also been extremely weak by
historical standards over the last few years (lower panel). Haltiwanger et al. (2012) show that
young and small businesses were especially hard-hit during the recession and weak recovery,
reflecting credit constraints and the steep drop in house prices, which reduced the ability of
entrepreneurs to finance startups or expansions with home equity. If start-ups play a
disproportionate role in promoting innovation because they embody the latest technologies, then
the “demand” factors that have restrained new business formation since the onset of the financial
crisis are also probably working to damp growth in multifactor productivity.
Additionally, cyclical changes in research and development (R&D) can have long-lasting
effects on multifactor productivity. Simple models generate the prediction that R&D investment
will be countercyclical as businesses would be expected to shift resources toward investments
with longer-term payoffs when the opportunity costs of allocating resources away from current
production is lower. Empirically, however, R&D investment appears to move in a procyclical
manner.
31
If so, then recessions should have a persistent adverse effect on the growth of
multifactor productivity. Relatedly, Shleifer (1986) finds that the diffusion of new technologies
is slower in recessions than in expansions. In light of this research, and given that real R&D
investment has grown only 1.6 percent per year since late 2007, as compared to 3.6 percent on
average from 1990 through 2007, it seems reasonable to assume that at least some of the
31
See, for example, Diego and Gertler (2006). Barlevy (2007) argues that the procyclicality of R&D reflects
externalities that cause firms to undertake more R&D in economic booms than would be optimal. In contrast,
Aghion (2012) shows that credit constraints can limit the capacity for firms to invest in R&D during recessions if
profits—and thus internal funds—are too low to finance such investments directly.
Page 31 of 61
cumulative reduction in trend MFP over the last few years is an endogenous response to weak
aggregate demand.
Finally, aggregate demand, and hence monetary policy, can potentially influence the
economy’s productive potential through its effects on capital deepening. Under the production-
function approach to supply-side estimation discussed earlier and employed by the
Congressional Budget Office (2001), the International Monetary Fund (2010), the ECB (2010),
and many other official institutions, the current level of the capital stock is a key determinant of
potential output. Thus, in this accounting framework, the substantial cutback in business outlays
on equipment and structures that typically occurs in response to the diminished sales prospects,
heightened uncertainty, and tight credit conditions of deep recessions
32
acts not only to reduce
current aggregate demand but also to lower the estimated productive capacity of the economy in
the future. Although such demand-induced capital deepening effects are presumably not literally
permanent, they are likely to persist for many years given the substantial adjustment costs that
characterize business investment.
33
Some alternative approaches to measuring resource utilization attempt to side-step this
issue by estimating potential output using an “equilibrium” concept of the capital stock in place
of the actual level. For example, it is common practice in DSGE modeling to define economic
slack using a flex-price concept of potential output, in which the latter is computed by simulating
32
In theory, the reduced pace of business capital deepening in the United States seen since 2007 could be the result
of technology shocks that have reduced the marginal return on capital. Arguing against this interpretation, however,
is the elevated level of profitability. Alternatively, one might argue that the decline in business investment has been
driven at least in part by reduced access to capital associated with permanently tighter underwriting standards and
other structural changes in credit markets. Whether the latter phenomenon is best thought of as a technological
rather than a demand development is open to debate, however; in any event, the restrictions on credit availability
that have emerged since the financial crisis have been more important for households than for businesses (especially
large ones). For these reasons, we believe that most of the observed slowdown in business investment is primarily a
response to a weak demand environment and heightened uncertainty about the future pace of recovery.
33
Such drawn-out capital accumulation dynamics are a standard feature of estimated macro models, including the
Federal Reserve Board’s workhorse FRB/US model and its two DSGE models, EDO and SIGMA.
Page 32 of 61
how the economy would have evolved over history in the absence of both wage-price frictions
and markup shocks. (See Neiss and Nelson, 2003.) This approach yields measures of the
equilibrium capital stock and potential output that, at least in theory, are exogenous to the
transitory fluctuations in aggregate demand and accompanying changes in monetary policy that
occur in the wake of a financial crisis, while allowing the “efficient” effects of changes in tastes
and technology on productivity, the composition of output, and other real factors to show
through. Thus, policymakers who employ the flex-price concept of potential output arguably
have the advantage of seeing through the transitory (albeit drawn out) swings in capital
deepening when crafting policy.
34
While we think it important to distinguish permanent movements in capital from
transitory fluctuations, we nonetheless believe that standard flex-price calculations of potential
output are problematic. As Woodford (2003) has pointed out, an important rationale for
allowing the actual (rather than equilibrium) level of capital services to affect the estimated level
of potential output is that firms’ marginal costs and productive capacity, and thus aggregate
inflation, depend on the actual capital stock, which evolves slowly over a time horizon relevant
for monetary policy. This line of argument suggests to us that central banks should design their
strategies with an eye to both the predicted future path of capital and the effects of their policy
actions on that path (and hence the evolution of potential output, actual employment, and
inflation).
35
Moreover, the standard flex-price calculation ignores the potential for movements in
34
On the surface, purely statistical methods for extracting trend output, such as the Beveridge-Nelson decomposition
or the Hodrick-Prescott filter, might also seem to avoid this issue because they do not condition on any measure of
the capital stock. For the reasons discussed earlier, however, such methods have the problem of ascribing to the
“trend” any movements in output associated with drawn-out fluctuations in capital services and other inputs,
whether or not they are endogenous.
35
Even if an estimate of potential output generated by a DSGE model is based on the actual business capital stock,
comparing that estimate to one based on the production-function approach may be problematic because the model’s
measure of capital may differ noticeably from the official government measure. In part, such differences can arise
because DSGE models often define business capital to include residential capital and the stock of consumer durable
Page 33 of 61
aggregate demand to influence potential labor input and trend multifactor productivity—effects
that in turn will alter any calculation of the equilibrium capital stock—because these channels are
not accounted for in the standard models used by central banks, DSGE or otherwise. Finally, we
would note that completely delinking the estimated level of potential output from the actual
capital stock, and instead basing it entirely on a theoretical calculation of what the stock would
be in the absence of all nominal frictions and mark-up shocks, suffers from the problem that the
identification of frictions and shocks, and hence the estimated level of potential output, can be
quite sensitive to model specification and assumptions about the nature of shocks—a point
discussed by Kiley (2012).
Quantitative assessment
The foregoing discussion leads to the obvious question: How much of the reduction in
aggregate supply during the past several years has represented an endogenous response to weak
aggregate demand that monetary policy should strive to mitigate, versus an exogenous
development to which monetary policy probably had to acquiesce? Of course it is difficult to
pinpoint the composition of what happened in the past several years, but the state-space model
we described earlier suggests that a reduction in capital deepening—which we view as mostly an
endogenous response to weak demand—caused almost half of the cumulative shortfall in
potential output from its pre-crisis trend. For the other possible channels, we interpret the
available evidence as indicating a modest adverse shift in the basic parameters of the labor
market expected to prevail over the longer run; at the same, we stress that it is far too early to
rule out the possibility that evidence of more substantial effects may emerge before the economy
goods, unlike the non-farm business sector measure used in the state-space analysis discussed earlier. In addition,
DSGE models may use implicitly use a different methodology for translating the business capital stock into an
aggregate flow of capital services. Finally, DSGE models often treat the capital stock as an unobserved variable, an
assumption that can result in yet more differences from the official series.
Page 34 of 61
has fully recovered. And the underlying causes—and likely persistence—of the apparent recent
deceleration in trend multifactor productivity are even murkier, although it seems likely that the
depressive effects of the recession and reduced credit availability on the rate of new business
formation and on R&D expenditures has played some role.
As we noted at the start of this section, our assessment that much of the recent supply-
side damage is endogenous has potentially important implications for the conduct of monetary
policy. In particular, such damage provides an additional rationale for policymakers to take
highly accommodative actions in response to sharp contractions in real activity. In the next
section, we illustrate this effect using simulations of a financial crisis under “optimal” policy
analysis. In carrying out this analysis, however, we also consider some additional factors that
may act to push monetary policy in a less accommodative direction.
IV. Optimal Policy, Endogenous Supply-Side Effects, and Other Considerations
The relevance of the endogeneity discussed in the previous section for monetary (or
fiscal) policy relates to the possibility that policymakers may be able craft strategies with an eye
to influencing both the supply and the demand sides of the economy. In particular, such policies
might differ appreciably from standard strategies that treat the natural rate of unemployment,
trend labor force participation, and other components of potential output as if they were
exogenous. Along these lines, Adolfson et al (2011) have used simulations of the Riksbank’s
macro model to show that “optimal” strategies which define potential output using the actual
(endogenous) capital stock differ noticeably from ones that define potential GDP using the flex-
price equilibrium (and thus policy-invariant) capital stock.
In a similar vein, we conduct simulations using the Federal Reserve workhorse FRB/US
macro model that allow for the possibility that a financial crisis and the resulting shortfall in
Page 35 of 61
aggregate demand endogenously cause a reduction of potential labor input and capital deepening,
along the lines of what seems to have happened during the Great Recession in the United States.
Leaving aside some potentially important countervailing considerations, we find that when
policymakers recognize the endogeneity of supply-side conditions and optimize accordingly,
they adopt a more aggressive approach to the conduct of policy in response to a recession.
However, we also emphasize that these other considerations—including concerns that an
aggressive policy stance may lead to an increased risk of financial instability or unacceptably
high inflation —may appropriately cause policymakers to exercise greater caution.
The FRB/US model
FRB/US is a large-scale model of the economy that has been used extensively by the staff
of the Federal Reserve Board since the mid-1990s to study a wide range of monetary and fiscal
policy issues. Although FRB/US does not have the tight micro-foundations of a DSGE model,
its equations are grounded in the assumption that households and firms are forward-looking and
engage in optimization subject to adjustment costs and habit persistence. Roughly 25 percent of
consumer spending is estimated to be carried out by rule-of-thumb consumers, while the rest is
attributable to life-cycle households who discount the future at a high rate owing to idiosyncratic
income risk. The model is very detailed and includes equations for eleven different components
of private consumption, investment, exports, and imports; standard asset pricing equations for a
variety of long-term interest rates, the stock market, and the exchange rate; a comprehensive
accounting of government spending and taxation at both the federal and the state and local
levels; and a small-scale foreign sector. Wage and price dynamics are characterized by a new-
Keynesian Phillips curve in which marginal costs move with the unemployment gap, defined as
Page 36 of 61
the actual unemployment rate less the current value of the natural rate (which, as discussed
below, will be augmented to include hysteresis effects).
36
In the version of FRB/US used in this paper, all financial market participants and agents
involved in wage-price setting are assumed to be rational and monetary policymakers enjoy
complete credibility; furthermore, these particular private-sector expectations are assumed to
incorporate perfect foresight about the future path of the economy once shocks hit.
37
These
expectational assumptions are important for our analysis because the fallout from the illustrative
financial crisis, like the one actually experienced in 2008, is sufficiently severe and protracted to
cause short-term interest rates to be constrained by the zero lower bound for several years. As a
result, and because for simplicity we leave aside the possibility of large-scale asset purchases, the
main tool available to monetary policymakers for stimulating the economy in the near term is to
promise to keep the federal funds low in the future, thereby putting downward pressure on long-
term interest rates. In the model, this pressure in turn reduces the borrowing costs of households
and firms, boosts corporate equity prices and other types of household wealth, and promotes net
exports through a lower real exchange rate.
Calibration of hysteresis effects
To facilitate the study of endogenous supply-side effects, we modify the standard version
of FRB/US (which already includes capital accumulation equations) to also incorporate
illustrative hysteresis-like responses of both unemployment and labor force participation to
36
For more information on the FRB/US model, see Brayton and Tinsley (1996), Brayton, Tinsley and Williams
(1997), Brayton, Mauskopf, Reifschneider, Tinsley and Williams (1997), and Reifschneider, Tetlow, and Williams
(1999).
37
Other private-sector expectations—most importantly, households’ assessments of future income—are generated
using a small-scale VAR model; thus, households are forward looking but have only an approximate sense of the
dynamics of the economy. Making all private-sector expectations model-consistent (that is, fully rational) would
have no qualitative effect on the results reported in this paper but would considerably slow the convergence speed of
the optimal-control simulations discussed below, owing to the highly nonlinear nature of the model with the zero
lower bound imposed.
Page 37 of 61
changes in aggregate activity. In particular, we assume that the natural rate of unemployment
and the trend labor-force participation rate evolve as follows:
(
)
(
)
* * ** * *
1 1111
* * ** * *
1 1111
.96 .04 .02
.96 .04 .04
t t tttt t
t t tttt t
U U U fUUUU
LFPR LFPR LFPR f U U U U
ε
ξ
− −−−−
− −−−−
= + + − −+
= + − − −+
In these expressions, U* is the natural rate of unemployment and LFPR* is the trend rate
of labor force participation. Both variables move persistently in response to the unemployment
gap (U-U*) and direct shocks (ε and ξ). Nonetheless, both variables also return (very slowly) to
their fixed long-run values U** and LFPR**.
38
These dynamics are consistent with the idea
that, although financial crises and deep recessions can have persistent effects on labor supply by
disrupting labor market functioning, impairing unemployed workers’ skills, and causing pre-
mature permanent departures from the labor force, such events do not alter demographic
conditions, the social safety net, or other fundamental determinants of long-run conditions in the
labor market.
39
Because a change in interest rates affects aggregate demand and thus the level of overall
employment, monetary policy in the adjusted version of FRB/US can influence potential output
not only through the capital accumulation channel but also through the potential supply of
labor.
40
Its ability to do so in the model simulations, however, depends importantly on f(U-U*),
a function that indexes the relative strength of the hysteresis effect, and so plays a key role in
determining the magnitude of endogenous supply-side effects in the model simulations. As
38
In the standard version of FRB/US, which incorporates the state-space model discussed in the first section of the
paper, the equivalents to U** and LFPR** are subject to permanent shocks; these shocks are idiosyncratic and
unrelated to shortfalls in aggregate demand. Such shocks are not relevant for the analysis considered in this section
of the paper, however, and so are suppressed here to simplify the analysis.
39
The Scandinavian labor markets do appear to have changed permanently after their financial crisis, but these long-
run changes plausibly reflected legislative changes to labor laws and other aspects of the social safety net.
40
In contrast, we assume that the FRB/US simulations do not provide any mechanism for activist monetary policy to
offset the adverse supply-side effects of direct shocks to U* or LFPR*.
Page 38 of 61
illustrated in Figure 4.1, f(.) is assumed to depend on the unemployment gap in a highly
nonlinear manner. Specifically, we assume that the level of resource utilization has no effect on
potential labor supply when the unemployment rate is below U*, but that the marginal effect of
labor market slack on the natural rate and trend labor force participation mounts rapidly as U
rises above U*.
This non-linear specification has two important policy implications. First,
accommodative monetary policy can limit the amount of endogenous damage to labor supply if it
can limit the amount of time the unemployment gap is above about 1¼ percent (the value of the
unemployment rate gap above which f() becomes substantial). Although this specification is ad
hoc, and in particular the threshold of 1¼ percent is somewhat arbitrarily chosen, such
“threshold” behavior in general seems consistent with the observation that warning signals that
figure prominently in today’s landscape, such as a marked increase in long-duration
unemployment and a persistent fall in labor force participation, were largely missing in the
milder recessions seen earlier in the post-World-War-II period in the United States. A second
important implication of this specification is that policymakers cannot undo labor market damage
once it has occurred, but must instead wait for it to fade away on its own accord; in other words,
there is no special advantage, given this specification, to running a high-pressure economy along
the lines suggested by Okun (1973).
41
Such quasi-irreversibility seems consistent with both the
tendency for older workers who leave the labor force prematurely on account of unemployment
to never return and the persistent stigma experienced by the long-term unemployed.
41
The specification of f(U-U*) as well as the coefficients of the two equations have been calibrated to yield
endogenous movements in U* and LFPR* that, in the context of the financial crisis scenario discussed below,
appear roughly consistent with the experience of the last few years. Arguably, it would have been better to estimate
these equations (and the shape of the scaling function) rather than calibrate them. However, given the lack of
historical evidence for hysteresis effects in the United States prior to the current episode, and given that our
simulations are intended to explore the possible implications of recent events (as opposed to the most likely ones),
we doubt that results from any time-series exercise would be particularly illuminative.
Page 39 of 61
Simulated effects of an illustrative financial crisis
Using the modified FRB/US model, we now develop an illustrative scenario involving a
major financial crisis that has persistent effects on both aggregate demand and aggregate supply;
by design, the macroeconomic effects of this shock are broadly similar to those seen to date since
2007. In this scenario, the economy is hit with a disruption to financial markets that causes a
sharp contraction in household spending, business investment, and employment in response to
higher risk premiums on a range of financial assets, falling house prices, and direct shocks to
spending and hiring similar to those experienced during the financial crisis and subsequent deep
recession.
42
In addition, the economy experiences exogenous disruptions to productivity and
labor market functioning in addition to those that arise endogenously in response to weak
aggregate demand. Finally, the effects of all these adverse events are exacerbated by zero-lower-
bound (ZLB) restrictions on the ability of monetary policymakers to counteract the weakness in
aggregate demand, and by the failure of the fiscal authorities to initiate any discretionary
countercyclical policy response.
Results (expressed as deviations from a steady-state baseline) for this scenario under an
inertial policy rule
43
are summarized in Figure 4.2.
44
As can be seen in the upper left panel, the
42
These direct shocks are presumed to reflect those effects of a financial crisis that operate through channels not
formally accounted for in the model’s structure, such as a reduced access to credit as a result of tighter lending
standards and persistent balance-sheet problems, increased uncertainty about future household income and corporate
earnings, and a general deterioration in consumer and business confidence. In the context of many DSGE models
(including the Fed’s EDO model), the effects of such disruptions are typically captured through an economy-wide
risk premium shock intended to provide a theoretical explanation for the correlated downturn in consumption and
investment. Nevertheless, like FRB/US, current DSGE models do not really provide a satisfactory accounting of the
various linkages between financial markets and the real economy that come into play during a financial crisis.
43
Specifically, the rule is R(t) = .85 R(t-1) + .15 {2 + PI(t) + 0.5 [PI(t) – 2] + 1.0 Y(t)}, where R is the nominal
funds rate, PI is the four-quarter rate of core PCE inflation, and Y is the output gap. A non-inertial version of this
rule is discussed in Taylor (1999).
44
In the baseline, the unemployment rate, inflation, and the nominal federal funds rate are constant at 5.5 percent, 2
percent, and 4.5 percent, respectively. The results reported in this section are largely insensitive to these baseline
assumptions, with the critical exception of nominal interest rates. Because the simulations incorporate the zero
lower bound constraint, the baseline setting of the federal funds rate has an important bearing on the ability of
monetary policy to offset the financial crisis.
Page 40 of 61
illustrative financial crisis and its restraining effect on aggregate demand cause the output gap to
widen more than 6 percentage points after two years and inflation to fall more than 1½
percentage points relative to baseline. In response, the inertial policy rule causes the federal
funds rate to drop 450 basis points over the first two years, after which no further reduction is
possible because of the ZLB. Nevertheless, by adhering to an inertial rather than non-inertial
rule, policymakers are able to provide greater stimulus to near-term activity because the inertial
rule takes a gradualist approach to returning the funds rate to a normal level after the ZLB no
longer binds, thereby reducing bond yields and improving financial conditions more generally.
45
Economic conditions gradually begin to improve starting in the third year, although the pace of
recovery is painfully slow—a profile similar in many respects to the actual experience of the
U.S. economy since the recession ended in mid-2009.
As shown in the upper-right panel, the scenario—like the actual economy in recent years,
according to our state-space results—features a noticeable deterioration in the economy’s
productive capacity, with potential GDP more than 4 percent below its baseline level by the fifth
year of the simulation. Most of this decline represents an endogenous response to the
persistently weak state of aggregate demand: Just over 40 percent of it is attributable to less
capital deepening as a result of a lower level of business investment, while a slightly smaller
portion is attributable to hysteresis effects that add almost ½ percentage point to the natural rate
and reduce the trend labor force participation rate by a full percentage point (shown in the
bottom two panels). However, not all the damage is endogenous: About 20 percent of the
reduction in potential GDP reflects the combined influence of an exogenous drop in trend
multifactor productivity and a ¼ percentage point rise in the natural rate caused by direct shocks
45
For simplicity, in this simulation and the others that follow we ignore the possibility that policymakers could use
large-scale asset purchases to mitigate the constraint imposed by the zero lower bound.
Page 41 of 61
to the U* equation. As can be seen, this supply-side damage—both endogenous and
exogenous—takes years to fade away, and is still noticeably depressing actual output and
employment more than a decade after the initial crisis.
“Optimal” policy responses
While the inertial policy rule prescribes a fairly aggressive response to the financial
crisis, it nevertheless does not prevent unemployment from rising sharply and remaining elevated
for years; nor does it prevent inflation from remaining persistently well below target.
Policymakers who recognize the likely magnitude and persistence of the crisis at the onset would
obviously be interested in policies that would deliver better outcomes. As discussed by
Svensson (2003 and 2005), one standard approach to this problem is to use optimal control
techniques. Under this approach, policymakers first specify a loss function that reflects their
preferences regarding outcomes for employment, inflation, and other conditions. They then
solve for the path of the funds rate that minimizes the loss function, conditional on the dynamics
of the economy (as approximated by some model) and the expected evolution of the underlying
shocks to the economy.
46
Optimal-control solutions are, not surprisingly, sensitive to the specification of
policymakers’ preferences as reflected in the loss function, as well as the specification of the way
that they perceive the economy as operating. Of relevance for the issues addressed in this paper
is that the “optimal” policy response to a crisis will depend on whether policymakers recognize
the effects of their actions on the supply side of the economy. In addition, such responses may
be sensitive to the desired objectives of policy—for example, is the central bank explicitly
aiming only to close the conventionally-defined unemployment gap in addition to stabilizing
46
See Svensson and Tetlow (2005) for an illustration of this technique using the FRB/US model and a discussion of
its use in its FOMC briefing documents. Also see Yellen (2012) for an illustration of its application to the current
economic situation.
Page 42 of 61
inflation, or is it instead trying to bring employment and/or output back to the levels that would
prevail in the absence of hysteresis-like effects? If an economic slump has resulted in
persistently lower labor force participation and less capital accumulation, then a policymaker
who aimed to close the unemployment gap would acquiesce to a greater loss in employment and
output than could be achieved with a more aggressive response (even taking account of changes
due to inflation in policymaker utility); as we will illustrate below, this non-equivalence holds
even if the unemployment gap is defined using a measure of the long-run equilibrium natural rate
U** instead of the more conventional U*.
We illustrate these sensitivities by displaying outcomes that are derived under alternative
assumptions regarding the policymaker’s perceptions of the dynamics of the economy and the
likely implications of the financial crisis for the supply side of the economy, holding the
specification of the loss function constant. In our baseline specification of the loss function—
which conforms in spirit with the FOMC’s dual mandate—policymakers at time t0 (the onset of
the financial crisis) wish to find the path for the federal funds rate R over the next M quarters that
would be expected to minimize a quadratic loss function L that penalizes (a) squared deviations
of unemployment from the conventionally-measured natural rate; (b) squared deviations of
inflation from the policymaker’s 2 percent goal; and (c) squared changes in the policy rate, as
follows:
47
( )
( ) ( )
{ }
2
22
*
0 0 10 0 20 3 0
0
2
N
j
t t tj tj tj tj
j
LE U U R
βα απ α
++ + +
=
= − + −+∆
∑
.
47
In addition to aiming to keep unemployment near its natural rate and inflation near the FOMC’s 2 percent target,
the loss function penalizes quarter-to-quarter movements in the federal funds rate. In reality, such movements
would be destabilizing and thus would have adverse effects on financial markets and the broader economy, implying
that such movements would be avoided in optimal-control simulations because of their effects on the unemployment
gap. However, the FRB/US model does not incorporate any mechanism for such volatility to affect financial
conditions and real activity through risk premiums or some other channel, so the third term is added to the loss
function to prevent unrealistically large quarterly movements in short-term interest rates in the optimal-control
simulations.
Page 43 of 61
In our optimal-control analysis, M (the number of quarters in the optimized path of R) is always
set to 100 quarters while N (the number of quarters over which the loss function is evaluated) is
set to 160 quarters; in addition, the discount factor β is set to .99 and the three α loss weights are
all set to unity.
48
Beyond quarter t0+M, when the optimized path ends, the federal funds rate is
assumed to follow the prescriptions of the inertial policy rule.
49
Using this baseline loss function, we optimize the path of the funds rate subject to
different policymaker beliefs about the nature of the economy and the effects of the financial
crisis. These contrasting beliefs are bookended on one side by a poorly-informed view that
ignores changes in supply-side conditions altogether, and on the other side by one based on a full
understanding of supply-side dynamics—a progression that helps to illuminate the marginal
effect of different supply-side considerations on optimal policy and associated macroeconomic
outcomes. Specifically, we consider three cases:
• In the first case, policymakers fail to recognize the damage to potential labor input and trend
multifactor productivity, both endogenous and exogenous, that will occur in the wake of the
crisis. (The one aspect of the damage to the supply side that they correctly anticipate is the
reduction in business capital and hence capital deepening.) Moreover, policymakers
mistakenly view the future evolution of the natural rate and trend labor force participation as
invariant to changes in monetary policy and aggregate real activity more generally.
Accordingly, they view the outcomes reported in Figure 4.2 as too pessimistic because they
48
Increasing the value of either M or N has essentially no effect on our simulation results, as does modestly
changing the discount factor or altering the relative loss weights (say, by increasing one of them to 5).
49
Optimal-control strategies of this sort raise issues of time consistency and how policy should be reoptimized in
light of previous commitments and incoming data surprises. These questions are beyond the scope of this paper,
however, and in the simulations discussed below we assume that policymakers do not re-optimize the trajectory for
the path of the funds rate beyond t0.
Page 44 of 61
incorrectly forecast the natural rate, trend labor force participation, and trend multifactor
productivity to follow the paths projected before the crisis.
50
• In the second case, policymakers do understand that financial crises adversely affect the
in Figure 4.2 if monetary policy follows the prescriptions of the inertial policy rule.
Policymakers err, however, in failing to recognize that some of this projected supply-side
damage could be reversed under a more aggressively countercyclical monetary policy; that
is, in optimizing the path of the federal funds rate they mistakenly treat the projections of the
natural rate and trend labor force participation shown in Figure 4.2 as exogenous.
• In the third case, policymakers correctly understand both the underlying outlook for the
economy as illustrated in Figure 4.2 and the “true” dynamics of the economy as captured by
the hysteresis-modified FRB/US model. Only in this case, then, is optimal policy computed
using full information.
In the first two cases, policymakers compute an “optimal” path of the federal funds that is based
on incorrect information—that is, the wrong model and/or underlying forecast. Forcing the
funds rate to follow this path in the context of the true economy would, however, not deliver the
outcomes expected by policymakers. To simulate the effects of misinformed “optimal” policy in
the context of the true economy, we assume that the central bank responds to the unexpected
movements in output and inflation by deviating from the funds rate path originally planned by
the amount prescribed by the inertial rule. Alternatively put, policymakers implement the two
50
Although movements in trend labor force participation and trend multifactor productivity do not figure directly in
the loss function, their recognition (or lack thereof) by policymakers matters to any computation of optimal
monetary policy because changes in these supply-side factors alter policymaker forecasts of potential output,
permanent household income, and expected future profits, and hence aggregate demand, employment, and the
unemployment rate.
Page 45 of 61
misinformed optimal strategies by first computing what quarter-by-quarter adjustments (or add-
factors) to the inertial rule would be necessary to replicate the optimal funds rate path conditional
on their initial expectations for the future evolution of the output gap and inflation; then, as
events actually unfold, they follow the prescriptions of the inertial rule plus the add factors that
were computed at t0.
51
(Implementing the well-informed optimal strategy in the same manner
would yield the originally-planned path for the funds rate and associated predicted outcomes, as
output and inflation evolve as originally predicted, implying no need to adjust the funds rate over
time using the adjusted inertial policy rule.)
The results from this exercise for resource utilization, inflation, and other indicators of
demand-side conditions are plotted in Figure 4.3. As indicated by the blue dashed lines in the
upper-left panel, among the three cases the most accommodative planned response is chosen by
policymakers who neither fully anticipate the supply-side fallout from the financial crisis nor
recognize the effects of their actions on potential labor supply, while the least accommodative
planned response is undertaken by policymakers who anticipate the supply-side damage but fail
to recognize their ability to mitigate it (the red dashed lines).
Actual outcomes for both output and the unemployment rate under the misinformed
optimal strategies (the sold blue and red lines) turn out to be fairly close to those achieved under
the fully-informed plan (the green solid line) for the first eight years or so, reflecting adjustments
to the originally-planned funds rate paths that are undertaken in response to what policymakers
see as unexpected movements in output and inflation. A different result obtains with respect to
inflation. Specifically, when policymakers ignore supply-side developments altogether in
crafting an optimal response to the crisis, actual inflation runs persistently above that achieved
51
For simplicity, after time t0 policymakers’ estimates of the output gap used in the adjusted inertial policy rule are
assumed to reflect the true level of potential output, including hysteresis effects.
Page 46 of 61
under the full-information strategy, and thus this strategy can be regarded as inappropriately
loose in hindsight. Conversely, when policymakers are somewhat better informed but still fail to
recognize their ability to mitigate hysteresis effects, actual inflation runs persistently below the
full-information strategy, and so policy turns out to be inappropriately tight.
Figure 4.4 compares the (actual) supply-side effects of the financial crisis under these
various optimal policies relative to what occurs under the inertial policy rule. As can be seen,
because all three optimal strategies provide more stimulus to aggregate demand, all result in
significantly less damage to the labor market and capital deepening. As a result, the peak decline
in potential GDP relative to baseline is roughly cut in half, with the largest improvement
occurring under full-information optimal policy, and the smallest under the strategy that
anticipates the supply-side damage but fails to take account of policy’s ability to mitigate it.
That said, the differences in supply-side outcomes across the three “optimal” strategies are
relatively minor.
To this point, the policymaker in our simulations has not cared directly about the
behavior of the natural rate or trend labor force participation, only about the conventionally-
measured unemployment gap and the deviation of inflation from the 2 percent target. Intuitively,
optimal policy should become even more accommodative if the central bank did not target the
unemployment gap but instead aimed at keeping the employment-to-population ratio near the
trend level that would prevail in the absence of hysteresis effects and exogenous (but ultimately
transitory) shocks to the natural rate.
52
This intuition is supported by FRB/US simulations, the
results from which are plotted as the magenta lines in Figures 4.3 and 4.4. As can be seen, this
strategy holds the nominal federal funds rate at zero appreciably longer than what occurs under
52
Alternatively, policymakers could aim to target a trend employment-to-population ratio that incorporated the
effects of hysteresis. However, this strategy yields results that are quite similar to that obtained under the baseline
specification of the loss function.
Page 47 of 61
U-U* targeting. (The differences are more noticeable for the real federal funds rate because
inflation is higher.) This strategy results in persistently lower unemployment and higher real
GDP, which in part reflects the effectiveness of the strategy in mitigating hysteresis effects in the
labor market, increasing capital deepening, and boosting potential output. And although the
strategy also results in inflation noticeably above the 2 percent target for several years, that
additional inflation is worthwhile from the perspective of policymakers, both because it mitigates
the effects of the ZLB and so helps to boost real activity through lower real interest rates, and
because it keeps inflation close to the 2 percent target during the first five years of the
simulation.
Offsetting considerations and other caveats
By themselves, the simulation results presented in Figures 4.3 and 4.4 would seem to
suggest that monetary policymakers should consider adopting more aggressive responses to deep
recessions than would be suggested by standard policy rules in order to mitigate endogenous
supply-side effects. This conclusion, however, overlooks the fact that policymakers may have
countervailing concerns that are not accounted for in the optimal-control exercises. In particular,
policymakers may be worried that pursuing a highly accommodative monetary policy for a long
time could inadvertently sow the seeds for a future financial crisis. Such a development might
occur if persistently low short-term interest rates were to prompt firms to take on increasing
amounts of leverage—thereby decreasing the stability of the financial system—or prompt
investors to take on an inappropriate amount of risk in a reach for yield. In light of these risks,
policymakers might appropriately opt for a more conservative response to a major economic
downturn, even if they recognized the potential adverse effects on the supply side of the
economy.
Page 48 of 61
To illustrate this possibility, we compute optimal policy responses to a major financial
crisis in a scenario in which persistently low short-term interest rates would eventually result in a
second financial crisis. In this exercise, the magnitude of the original financial shock is the same
as before and is accompanied by the same endogenous supply-side effects. However,
policymakers now confront an unpleasant trade-off: The more they attempt to stimulate
aggregate demand by promising to keep current and future short-term rates at a low level, the
greater is the magnitude of a second financial crisis, which is assumed to occur, for sure, in the
tenth year of the scenario. Specifically, the longer and the lower they hold the nominal federal
funds rate below 1½ percent over the nine years following the onset of the initial financial crisis,
the larger are the shocks that hit the economy in the tenth year. For simplicity, these second-
round shocks are re-scaled versions of those that occurred during the first crisis, where the
scaling factor is
( )
36
1
1.5
tt
t
dR
µ
=
Ω= −
∑
,
with
1
t
d =
if
1.5
t
R
<
, 0 otherwise. (An implication of this specification is that the policymakers
cannot offset the destabilizing financial effects of very low interest rates by pushing them to
unusually high levels later.) The parameter μ is calibrated so that the deterioration in real
activity following the second crisis is about the same as in the first if policymakers follow the
prescriptions of the inertial policy rule throughout.
Results from this exercise are reported in Figure 4.5 for two different types of optimal-
control policies. In both cases, policymakers strive to keep the employment-to-population ratio
near its time-invariant long-run equilibrium level (i.e., E**/P) and inflation near 2 percent while
trying to avoid large quarter-to-quarter movements in the federal funds rate. In the first case (the
green lines), policymakers elect to avoid a second crisis altogether by optimizing subject to the
Page 49 of 61
constraint that the federal funds rate is never allowed to fall below 1½ percent. By contrast, in
the second case (the red lines) they allow the optimal funds rate path to fall to zero, thereby
creating a second “optimal” financial crisis with outcomes for real activity and inflation that are
roughly 60 percent as bad as those associated with the first crisis. In either case, optimal
monetary policy turns out to be noticeably more restrictive on average over the first nine years
than in the situation where very accommodative monetary policy does not have adverse effects
on financial stability (magenta lines). Nevertheless, the threat of a second financial crisis does
not mean that policymakers necessarily eschew driving short-term interest rates to zero for a
time. Even though policymakers can avoid the second crisis altogether by always keeping the
funds rate above 1½ percent, that strategy exacerbates the macroeconomic impact of the first
round of shocks by more than enough to make the overall loss appreciably worse than the
alternative optimal policy that allows the funds rate to fall temporarily to zero. In fact,
constraining the optimal path to never fall below 1½ percent increases the value of the loss
function by more than half.
53
In principle, the willingness of policymakers to pursue an aggressive monetary response
to a recession should depend on their views about the magnitude of the risks of supply-side
damage and to financial stability, as well as the expected efficacy of actions taken to mitigate
these. That is, policymakers are engaged in a cost-benefit calculation that balances, on the one
hand, the expected macroeconomic benefits from stronger aggregate demand and less adverse
supply-side effects, and on the other hand, the expected losses from sparking a future crisis. If
the risk of undermining the stability of the financial system was nil, then policymakers would
53
Given that the cumulative difference between the paths of the unemployment rate under the two strategies is close
to zero, it may seem surprising that the minimum 1½ percent strategy results in such a higher loss. The explanation
is the quadratic nature of the loss function, which causes the cost of an incremental increase in the unemployment
rate to climb sharply as the starting level of the unemployment gap widens.
Page 50 of 61
presumably wish to become more aggressive as the effectiveness of monetary policy declines
because the additional reduction in interest rates would be costless. But if low interest rates are
viewed as posing financial stability risks, increased activism should decline as the expected costs
of such action increase.
Of course, threats to financial stability are probably not the only offsetting concern that
might limit policymakers’ willingness to fight endogenous supply-side damage; for example,
they may also be reluctant to implement a highly accommodative strategy because of concerns
about its potential adverse effects on inflation expectations and inflation dynamics more
generally. In the optimal-control analysis presented in this paper, wage and price expectations
are rational, policymakers enjoy complete credibility, and the parameters of the new Keynesian
inflation process are stable and invariant to changes in monetary policy—assumptions that
almost certainly do not hold in reality. And even though Kiley (2007), Laforte (2007), and
others have found that empirical new Keynesian inflation models of the sort used in FRB/US and
in DSGE models do provide a reasonable approximation to the observed behavior of inflation
over the past twenty years or so, policymakers might well worry that inflation dynamics could
evolve in a highly undesirable and costly direction if monetary policy were to depart markedly
from recent historical norms, perhaps even returning to the instability seen during the 1970s.
Finally, we should stress that the preceding analysis ignores uncertainty, which is
ubiquitous in the real world. In the wake of a financial crisis, policymakers cannot be sure about
the extent of supply-side damage that has occurred even well after the fact, let alone the
proportion that reflects an endogenous response to weak aggregate demand. In addition, they
cannot be sure about the ability of a more accommodative policy stance to check the initial
damage that occurs or to subsequently repair it, particularly in an environment in which the
Page 51 of 61
ability of monetary policy to influence aggregate demand may be impaired. Finally, the effects
of persistently accommodative monetary policy on financial stability and the stability of inflation
expectations are also highly uncertain. How policymakers should respond to such pervasive
uncertainty is not obvious, especially if they (or the private agents on whose behalf they act) are
not risk-neutral. On the one hand, Brainard-type considerations might argue for taking a more
cautious approach to trying to head off supply-side damage than suggested by the optimal-
control simulations, given uncertainty about the effectiveness of monetary policy in mitigating
supply-side damage. On the other hand, a robust control approach might call for a more
aggressive response if the adverse tail event of primary concern involved endogenous supply-
side damage.
54
In any event, uncertainty about both the extent and nature of supply-side
damage, as well as about the possible side effects of a persistently accommodative stance of
policy, greatly complicates the decision-making process because it forces policymakers to weigh
the costs and probabilities associated with a range of risks and possible outcomes.
Conclusions
This paper has reviewed the evidence for supply-side damage in the wake of the financial
crisis and considered some of its implications for monetary policy. In the labor market,
matching efficiency seems to have been somewhat impaired, the natural rate of unemployment
appears to have risen somewhat, and trend labor force participation appears to have moved
noticeably lower relative to what would have been expected based on pre-crisis trends. In
addition, the capital stock and trend multifactor productivity are appreciably lower than what
would have been predicted in 2007. Our point estimates suggest that, in combination, these
developments—whose eventual magnitude was arguably apparent only in hindsight—shaved
54
Of course, if policymakers were instead concerned about minimizing the risk of a future financial crisis, then
robust control might argue for a less activist strategy.
Page 52 of 61
almost 7 percent off the level of potential output relative to its pre-crisis trend. That said, the
uncertainty about this estimate is extremely high and the implications for future growth are quite
uncertain.
Despite this supply-side damage, our point estimates also suggest that the level of
economic slack has been and remains quite high. As has been noted by a number of observers,
this factor by itself would argue for a highly accommodative monetary policy, particularly in an
environment of what appears to be quite well-anchored inflation expectations. We have argued
that the case for aggressive policy is strengthened further by the likelihood that much of the
supply-side damage is an endogenous response to weak aggregate demand. As our simulation
analysis illustrates, optimal monetary policy becomes noticeably more accommodative in the
wake of a major financial crisis if the natural rate of unemployment and trend labor force
participation are subject to hysteresis-like effects that policy can potentially mitigate. However,
we have also argued that policymakers may appropriately be restrained from pursuing a highly
aggressive response to a deep recession if they fear the attendant risks to financial stability, or
are concerned that inflation expectations may become unanchored. More generally, the
pervasive uncertainty in which policymakers operate may encourage them to proceed with
caution.
Page 53 of 61
References
Aaronson, Daniel, Jonathan Davis, and Luojia Hu (2012). “Explaining the Decline in the U.S.
Labor Force Participation Rate,” Chicago Fed Letter, No. 13.
Aaronson, Stephanie, Bruce Fallick, Andrew Figura, Jonathan Pingle, and William Wascher
(2006). “The Recent Decline in the Labor Force Participation Rate and Its Implications
for Potential Labor Supply,” Brookings Papers on Economic Activity, no. 1, pp.69-154.
Abraham, Katherine G. and Lawrence F. Katz (1986). “Cyclical Unemployment: Sectoral
Shifts or Aggregate Disturbances?” Journal of Political Economy, 94(3): 507-522.
Adolfson, Malin, Stefan Laseen, Jerper Linde, and Lars E. O. Svensson (2011). “Optimal
Monetary Policy in an Operational Medium-Sized DSGE Model,” Journal of Money,
Credit, and Banking, 43(7), 1287-1331.
Aghion, Philippe, Philippe Askenazy, Nicolas Berman, Gilbert Cette, and Laurent Eymard
(2012). “Credit Constraints and the Cyclicality of R&D Investment: Evidence from
France,” Journal of the European Economic Association, 10(5): 1001-1024.
Akerlof, George A., William T. Dickens, and George L. Perry (1996). “The Macroeconomics of
Low Inflation,” Brookings Papers on Economic Activity, no. 1, pp.1-75.
Ascari, Guido and Argia M. Sbordone (2013). “The Macroeconomics of Trend Inflation,”
Federal Reserve Bank of New York Staff Reports No. 628 (August).
Autor, David and Mark Duggan (2006). “The Growth in the Social Security Disability Rolls: A
Fiscal Crisis Unfolding,” Journal of Economic Perspectives, 20(3): 71-96.
Baily, Martin, James Manyika, and Shalabh Gupta (2013). “U.S. Productivity Growth: An
Optimistic Perspective,” International Productivity Monitor, No. 25(1):3-12.
Ball, Laurence (1999). “Aggregate Demand and Long-Run Unemployment,” Brookings Papers
on Economic Activity, 2, 189-251.
Ball, Laurence, Daniel Leigh and Prakash Loungani (2013). “Okun’s Law: Fit at 50?” NBER
Working Paper 18668.
Barlevy, Gadi (2007). “On the Cyclicality of Research and Development,” American Economic
Review, 97(4): 1131-1164.
Barnichon, Regis and Andrew Figura (2013). “Labor Market Heterogeneities and the Aggregate
Matching Function,” unpublished manuscript, September 2013.
Barro, Robert J. and David B. Gordon (1983). “A Positive Theory of Monetary Policy in a
Natural-Rate Model,” Journal of Political Economy, 91(4): 589-610.
Basu, Susanto and John G. Fernald (2009). “What Do We Know (and Not Know) About
Potential Output?” Federal Reserve Bank of St. Louis Review (July), 187-214.
Page 54 of 61
Beveridge, S. and C. R. Nelson (1981). “A New Approach to the Decomposition of Economic
Time Series Into Permanent and Transitory Components with Particular Attention to
Measurement of the Business Cycle,” Journal of Monetary Economics 7(x): 151-174.
Blanchard, Olivier J. and Lawrence H. Summers (1986). “Hysteresis and the European
Unemployment Problem,” NBER Macroeconomics Annual, 1, 15-90.
Blanchard, Olivier J. (2003). “Monetary Policy and Unemployment,” remarks at Monetary
Policy and the Labor Market: A Conference in Honor of James Tobin, New School, New
York, November 2002.
Borio, Claudio, Piti Disyatat and Mikael Juselius (2013). “Rethinking Potential Output:
Embedding Information about the Financial Cycle,” Bank of International Settlements
Working Papers No. 404.
Brayton, Flint and Peter Tinsley, eds. (1996). “A Guide to FRB/US—A Macroeconomic Model
of the United States,” Federal Reserve Board Finance and Economics Discussion Series
Paper No. 1996-42.
Brayton, Flint, Andrew T. Levin, Ralph Tryon, and John C. Williams (1997). “The Evolution of
Macro Models at the Federal Reserve Board,” Carnegie-Rochester Conference Series on
Public Policy 47, 227-245.
Brayton, Flint, Eileen Mauskopf, David Reifschneider, Peter Tinsley, and John C. Williams
(1997). “The Role of Expectations in the FRB/US Macroeconomic Model,” Federal
Reserve Bulletin 83 (April), 43-81.
Brynjolfsson, Erik and Andrew McAfee (2011). Race Against the Machine: How the Digital
Revolution of Accelerating Innovation, Driving Productivity, and Irreversibly
Transforming Employment and the Economy. Digital Frontier Press.
Byrne, David M., Steven D. Oliner, and Daniel E. Sichel (2013). “Is the Information
Technology Revolution Over?” International Productivity Monitor, 25(1): 20-36.
Cerra, Valerie and Sweta Chaman Saxena (2008). “Growth Dynamics: The Myth of Economic
Recovery,” American Economic Review, 98(1): 439-57.
Chung, Hess, Michael Kiley, and Jean-Philippe Laforte (2012). “Unemployment During the
Great Recession in the EDO Model of the U.S. Economy: The 2012 EDO Model.”
Federal Reserve Board Finance and Economics Discussion Paper No.
Clark, Peter K. (1987). “The Cyclical Component of U.S. Economic Activity,” Quarterly
Journal of Economics 102(4): 797-814.
Clark, Todd (2011). “Real-Time Density Forecasts from Bayesian Vector Autoregressions with
Stochastic Volatility,” Journal of Business and Economic Statistics, 29, 327-341.
Comin, Diego and Mark Gertler (2006). “Medium-Term Business Cycles,” American Economic
Review, 96(3): 523-551.
Page 55 of 61
Congressional Budget Office (2001). “CBO’s Method for Estimating Potential Output: An
Update,” Congress of the United States, Washington, D.C. (August).
Congressional Budget Office (2012). “What Accounts for the Slow Growth of the Economy
After the Recession?” Congress of the United States, Washington, D.C. (November).
Daly, Mary C., Bart Hobijn, Aysegul Sahin, and Rob Valletta (2012). “A Search and Matching
Approach to Labor Markets: Did the Natural Rate of Unemployment Rise?” Journal of
Economic Perspectives, 26(3): 3-26.
Daly, Mary C., and Bart Hobijn (2013). “Downward Nominal Wage Rigidities Bend the Phillips
Curve,” Federal Reserve Bank of San Francisco Working Paper No. 2013-08.
Davis, Steven J., Jason Faberman and John C. Haltiwanger (2012). “Recruiting Intensity During
and After the Great Recession: National and Industry Evidence,” American Economic
Review, 102(3):, 584-88.
Del Negro, Marco, Marc P. Giannoni, and Frank Schorfheide (2013). “Inflation in the Great
Recession and New Keynesian Models,” Federal Reserve Bank of New York Staff
Reports No. 618 (May).
Delong, J. Bradford, and Lawrence H. Summers (2012). “Fiscal Policy in a Depressed
Economy,” Brookings Papers on Economic Activity, Spring, 233-297.
De Masi, P. (1997). “IMF Estimates of Potential Output: Theory and Practice,” Staff Studies for
the World Economic Outlook, December.
Diamond, Peter (2013). “Cyclical Unemployment, Structural Unemployment,” Federal Reserve
Bank of Boston Working Paper No. 13-5.
Dornbush, Rudiger, and Stanley Fischer (1978). Macroeconomics, McGraw-Hill.
Erceg, Christopher J. and Andrew T. Levin (2013). “Labor Force Participation and Monetary
Policy in the Wake of the Great Recession,” working paper (IMF).
European Central Bank (2000). “Potential Output and Output Gaps,” ECB Monthly Bulletin,
October 2000: 37-48.
European Central Bank (2011). “Trends in Potential Output,” ECB Monthly Bulletin, January
2011, 73-85.
Farber, Henry S. and Robert G. Valletta (2013). “Do Extended Unemployment Benefits
Lengthen Unemployment Spells? Evidence from Recent Cycles in the U.S. Labor
Market,” Federal Reserve Bank of San Francisco Working Paper Series No. 2013-09.
Fernald, John G. (2012). “Productivity and Potential Output Before, During and After the Great
Recession,” Federal Reserve Bank of San Francisco Working Paper 2012-18.
Figura, Andrew, and William Wascher (2010). “The Causes and Consequences of Sectoral
Reallocation: Evidence from the Early 21
st
Century,” Business Economics, 45(1): pp.49-
68.
Page 56 of 61
Fleischman, Charles A. and John M. Roberts (2011). “From Many Series, One Cycle: Improved
Estimates of the Business Cycle from a Multivariate Unobserved Components Model,”
Federal Reserve Board Finance and Economics Discussion Series No. 2011-46.
Fort, Teresa, John Haltiwanger, Ron S. Jarmin, and Javier Miranda (2013). “How Firms Respond
to Business Cycles: The Role of the Firm Age and Firm Size,” NBER Working Paper
No. 19134.
Gordon, Robert J. (2003). “Exploding Productivity Growth: Context, Causes, and
Implications,” Brookings Papers on Economic Activity, 2003, 2, 207-79.
Gordon, Robert J. (2012). “Is U.S. Economic Growth Over? Faltering Innovation Confronts the
Six Headwinds,” NBER Working Paper No. 18315.
Gordon, Robert J. (2013). “U.S. Productivity Growth: The Slowdown Has Returned After a
Temporary Revival,” International Productivity Monitor, No. 25(1):13-19.
Haltmaier, Jane. (2012). “Do Recessions Affect Potential Output?” Federal Reserve Board
International Finance Discussion Paper No. 1066 (December).
Hassett, Kevin A. (2013). “A Long-Term Problem for the Economy,” National Review (May
10).
Hornstein, Andreas (2013). “The Cyclicality of the Labor Force Participation Rate,” Working
Paper, Federal Reserve Bank of Richmond.
Kiley, Michael T. (2007). “A Quantitative Comparison of Sticky-Price and Sticky-Information
Models of Price Setting,” Journal of Money, Credit, and Banking, 39(S1): 101-125.
Kiley, Michael T. (2012). “Output Gaps,” Federal Reserve Board Finance and Economic
Discussion Series 2010-27.
Kroft, Kory, Fabian Lange, and Matthew J. Notowidigdo (2013). “Duration Dependence and
Labor Market Conditions: Evidence from a Field Experiment,” Quarterly Journal of
Economics, 128(3): 1123-1167.
Laforte, Jean-Philippe (2007). “Pricing Models: A Bayesian DSGE Approach for the U.S.
Economy,” Journal of Money, Credit, and Banking, 39(s1), 127-154.
Lazear, Edward P. and James R. Spletzer (2012). “The United States Labor Market: Status Quo
or a New Normal?”in The Changing Policy Landscape, Federal Reserve Bank of Kansas
City Economic Policy Symposium, 405-451.
Lilien, David M. (1982), “Sectoral Shifts and Cyclical Unemployment,” Journal of Political
Economy 90 (August): 777-93.
Loungani, Prakesh, and Richard Rogerson (1989). “Cyclical Fluctuations and the Sectoral
Reallocation of Labor: Evidence from the PSID,” Journal of Monetary Economics, 23(2):
259-273.
Page 57 of 61
Lucas, Robert E. Jr. (1971). “Econometric Testing of the Natural Rate Hypothesis,” The
Econometrics of Price Determination, Board of Governors of the Federal Reserve
System.
Molloy, Raven Saks, Christopher L. Smith and Abigail Wozniak (2013). “Internal Migration in
the United States,” Journal of Economic Perspectives 25(3): 173-196.
Mueller, Andreas I., Jesse Rothstein, and Till M. von Wachter (2013). “Unemployment
Insurance and Disability Insurance in the Great Recession,” unpublished working paper,
University of California at Berkeley.
Nalewaik, Jeremy J. (2010) “The Income- and Expenditure-Side Estimates of U.S. Output
Growth,” Brookings Papers on Economic Activity Spring: 71-106.
Neiss, Katherine S. and Edward Nelson (2003). “The Real Interest Rate Gap as an Inflation
Indicator,” Macroeconomic Dynamics, 7, 239-62.
Okun, Arthur M. (1973). “Upward Mobility in a High-Pressure Economy,” Brookings Papers on
Economic Activity, 1973:1, pp. 207-252.
Orphanides, Athanasios (2003). “The Quest for Prosperity Without Inflation,” Journal of
Monetary Economics, 50(3): 633-663.
Orphanides, Athanasios and Simon van Norden (2002). “The Reliability of Output Gap
Estimates in Real Time,” Review of Economics and Statistics, 84(4): 569-583.
Orphanides, Athanasios and John C. Williams (2006). “Inflation Targeting Under Imperfect
Knowledge,” in Federal Reserve Bank of San Francisco Working Paper Series 2006-14.
Orphanides, Athanasios, Richard D. Porter, David Reifschneider, Robert Tetlow and Frederico
Finan (2000). “Errors in the Measurement of the Output Gap and the Design of
Monetary Policy,” Journal of Economics and Business, 52(1/2): 117-141.
Ramey, Valerie A., and Matthew D. Shapiro (2001). “Displaced Capital: A Study of Aerospace
Plant Closings,” Journal of Political Economy, 109(5): 958-992.
Reifschneider, David, Robert Tetlow, and John C. Williams (1999). “Aggregate Disturbances,
Monetary Policy, and the Macroeconomy: The FRB/US Perspective,” Federal Reserve
Bulletin 85 (January), 1-19.
Reinhart, Carmin M. and Kenneth S. Rogoff (2010). This Time is Different: Eight Centuries of
Financial Folly. Princeton University Press (Princeton).
Sahin, Aysegul, Joseph Song, Giorgio Topa, and Giovanni L. Violante (2012). “Mismatch
Unemployment,” NBER Working Paper No. 18265.
Shleifer, Andrei (1986). “Implementation Cycles,” Journal of Political Economy, 94(6): 1163-
1190.
Stockhammer, Englebert and Simon Sturn (2012). “The Impact of Monetary Policy on
Unemployment Hysteresis,” Applied Economics, 44, 2743-2756.
Page 58 of 61
Svensson, Lars E. O. (2003). “What is Wrong with Taylor Rules? Using Judgment in Monetary
Policy through Targeting Rules,” Journal of Economic Literature 41(2): 426-77.
Svensson, Lars E. O. (2005). “Monetary Policy with Judgment: Forecast Targeting,”
International Journal of Central Banking 1(1): 1-54.
Svensson, Lars E. O. and Robert Tetlow (2005). “Optimal Policy Projections,” International
Journal of Central Banking, 1(3): 177-207.
Taylor, John B. (1999). “A Historical Analysis of Monetary Policy Rules,” in John B. Taylor,
ed., Monetary Policy Rules. University of Chicago Press, 319-341.
Tobin, James (1972). “Inflation and Unemployment,” American Economic Review, 62, 1-18.
Valletta, Robert G. and Katherine Kuang (2010). “Is Structural Unemployment on the Rise?”
Federal Reserve Bank of San Francisco Economic Letter 2010-34 (November).
Valletta, Robert G. (2013). “House Lock and Structural Unemployment,” Federal Reserve Bank
of San Francisco Working Paper Series 2012-25 (April).
Yellen, Janet (2012). “The Economic Outlook and Monetary Policy,” speech at the Money
Marketeers of New York, New York (April 11),
http://www.federalreserve.gov/newsevents/speech/yellen20120411a.htm
Page 59 of 61
Appendix—the State Space Model
1. Real GDP (per capita, logged)
gdp = wedge1 + tmfp/.965 + (.035/.965)*lveoa + .725*(terate + tlfpr + tww + wedge2)
+ .275*lks + .725*lqualt + cycle + β11*β6 + β11*e_nfbp + [white-noise error, var=β100
2
]
2. Real non-farm business output (per capita, logged)
nfbp = tmfp/.965 + (.035/.965)*lveoa + .725*(terate + tlfpr + tww + wedge2) + .275*lks
+ .725*lqualt + β10*cycle + β6 + e_nfbp
3. Real non-farm business income (per capita, logged)
nfbi = tmfp/.965 + (.035/.965)*lveoa + .725*(terate + tlfpr + tww + wedge2) + 0.275*lks
+ .725*lqualt + β10*cycle – β6 + e_nfbi
4. Workweek, nonfarm business sector (logged)
wwnfb = tww + 0.72*[wwnfb(-1)-tww(-1)] + φ20*[cycle-cycle(-1)] + φ(22)*cycle
+ [white-noise error, var=β104
2
]
5. Employment, nonfarm business sector (per capita, logged)
enfb = terate + tlfpr + wedge2 + φ30*cycle + φ31*[enfb(-1)-terate(-1)-tlfpr(-1)-wedge2(-1)]
+ [white-noise error, var=β105
2
]
6. Employment-to-population ratio (logged)
erate = terate + φ50*cycle + φ51*[erate(-1)-terate(-1)] + [white-noise error, var=β106
2
]
7. Labor force participation rate (logged)
lfpr = tlfpr + φ40*cycle + φ41*[lfpr(-1)-tlfpr(-1)] + [white-noise error, var=β107
2
]
8. Core PCE inflation
pcex = β401*pcex(-1) + (1-β401)*epi(-1) + β404*[.50*cycle + .33*cycle(-1) + .17*cycle(-2 )]
+ β405*MA(rpe(-1),6) + β406*MA(d84*rpe(-1),6) + β408*rpm + β409*rpm(-1)
+ β407*wpc + [white noise error, var=β109
2
]
Note: MA(X,n) denotes the n-quarter moving average of X
9. Business cycle (state variable)
cycle = β1*cycle(-1) + β2*cycle(-2) + [white noise error, var=β111
2
]
10. Nonfarm business output error (state variable)
e_nfbp = β602*e_nfbp(-1) + [white noise error, var=β125
2
]
11. Nonfarm business income error (state variable)
e_nfbi = β602*e_nfbi(-1) + [white noise error, var=β126
2
]
12. Trend level of the GDP-NFB output wedge (state variable)
wedge1 = wedge1(-1) + .25*gwedge1 + [white-noise error, var=β112
2
]
13. Trend growth rate of the GDP-NFB output wedge (state variable)
gwedge1 = .95*gwedge1(-1) + .05*β213 + [white-noise error, var=(4*.03326*β112)
2
]
Page 60 of 61
14. Trend level of multi-factor productivity (state variable)
tmfp = tmfp(-1) + .25*gtmfp + [white-noise error, var= β114
2
]
15. Trend growth rate of multi-factor productivity (state variable)
gtmfp = 0.95*gtmfp(-1) + 0.05*β214 + [white-noise error, var= β115
2
]
16. Trend NFB workweek (state variable)
tww = tww(-1) + .25*gtww + [white-noise error, var=.01]
17. Trend growth rate of the NFB workweek (state variable)
gtww = .95*gtww(-1) + .05*β216 + [white-noise error, var=β117
2
]
18. Trend level of the wedge between household and NFB payroll employment (state variable)
wedge2 = wedge2(-1) + 0.25*gwedge2 + [white-noise error, var=(.01*β118)
2
]
19. Trend growth rate of the wedge between household and NFB payroll employment (state variable)
gwedge2 = .95*gwedge2(-1) + [white-noise error, var=β119
2
]
20. Trend level of the labor force participation rate (state variable)
tlfpr = tlfpr(-1) + 0.25*gtlfpr + [white-noise error, var=.0025]
21. Trend growth rate of the labor force participation rate (state variable)
gtlfpr = 0.95*gtlfpr(-1) + [white-noise error, var=β123
2
]
22. Natural rate of employment (state variable)
terate = terate(-1) + [white-noise error, var=β124
2
]
Exogenous variables
lveoa trend energy-output ratio (logged)
lks capital services (per capita, logged)
lqualt labor quality (logged)
rpe PCE energy prices relative to core PCE prices, weighted by energy share of consumer spending
rpm non-oil import prices relative to core PCE prices, weighted by import share of domestic spending
wpc wage-price controls (1971q3 to 1974q1 =1, 1974q2 to 1974q4 = -3.67, =0 otherwise)
d84 dummy variable (= 1 from 1985q1 on, = 0 otherwise)
epi expected long-run inflation (as reported in the Survey of Professional Forecasters from 1990 to
the present and in the Hoey survery from 1981 to 1990; prior to 1981 expectations are inferred
by a trend extraction procedure using actual inflation)
61
Page 61 of 61
Table A.1 Estimation Results for the State-Space Model
(Sample period 1963:Q2 to 2013:Q1)
Coefficient
Standard
Error
z-
Statistic
Probability Coefficient
Standard
Error
z-
Statistic
Probability
β1 1.5165 0.0583 26.00 0.00 β126 0.4004 0.0369 10.84 0.00
β2
-0.5529
0.0593
-9.33
0.00
β213
-0.3378
0.0639
-5.29
0.00
β6
0.3115
0.3119
1.00
0.32
β214
0.8627
0.2705
3.19
0.00
β10 1.3896 0.0226 61.55 0.00 β216 -0.2079 0.1285 -1.62 0.11
β11
0.7193
0.0314
22.90
0.00
β401
0.5762
0.0643
8.96
0.00
β100 0.0586 0.0125 4.70 0.00 β404 0.0996 0.0279 3.57 0.00
β104
0.2147
0.0141
15.21
0.00
β405
0.5174
0.1859
2.78
0.01
β105
0.1619
0.0143
11.33
0.00
β406
-0.3467
0.3235
-1.07
0.28
β106 0.0819 0.0152 5.38 0.00 β407 -0.4646 0.1024 -4.54 0.00
β107
0.2126
0.0150
14.14
0.00
β408
0.3316
0.1452
2.28
0.02
β109 0.7605 0.0436 17.44 0.00 β409 0.2738 0.1778 1.54 0.12
β111
0.5702
0.0382
14.93
0.00
β602
0.9124
0.0328
27.78
0.00
β112
0.1254
0.0184
6.83
0.00
φ20
0.2511
0.0392
6.41
0.00
β114 0.2299 0.0539 4.27 0.00 φ22 0.0472 0.0132 3.57 0.00
β115
0.1361
0.0653
2.09
0.04
φ30
0.4518
0.0260
17.36
0.00
β117 0.0662 0.0276 2.40 0.02 φ31 0.6599 0.0249 26.53 0.00
β119
0.1142
0.0364
3.14
0.00
φ40
0.0427
0.0165
2.58
0.01
β123
0.1169
0.0306
3.82
0.00
φ41
0.7573
0.0883
8.57
0.00
β124 0.1359 0.0191 7.10 0.00 φ50 0.2933 0.0197 14.86 0.00
β125
0.5172
0.0405
12.76
0.00
φ51
0.5334
0.0336
15.87
0.00
Log likelihood
-610.625
Akaike info criterion
6.506253
Parameters
40
Schwarz criterion
7.165917
Diffuse priors 0 Hannan-Quinn criterion 6.773209
Table 1.1
State-Space Model Estimates of Recent Changes in U.S. Supply-Side Conditions
1990-99
2000-07
2008
2009
2010
2011
2012
1. Potential GDP (Q4/Q4 percent change in level)
3.1
2.6
2.3
1.8
0.6
1.1
0.9
Contribution in percentage points of movements in:
2. Trend labor input
1.4
0.5
0.9
0.3
0.2
0.5
0.3
2a. Population
1.1
1.2
1.1
1.1
1.0
1.0
1.0
2b. Labor force participation rate
0.0
-0.2
-0.2
-0.3
-0.6
-0.7
-0.6
2c. Natural rate
0.0
0.0
0.0
-0.4
-0.3
0.2
0.1
2d. Workweek
-0.1
-0.2
0.1
0.0
0.2
0.0
-0.1
2e. Wedge (Priv. payroll vs. HH employment)
0.2
-0.3
-0.1
0.0
-0.1
-0.1
-0.1
3. Trend labor productivity
2.1
2.5
1.7
1.7
0.9
0.9
1.1
3A. Labor quality
0.4
0.3
0.4
0.4
0.3
0.3
0.3
3B. Capital deepening
0.8
0.8
0.4
0.0
0.0
0.2
0.4
3C. Multifactor productivity*
0.9
1.4
0.9
1.3
0.6
0.4
0.4
4. GDP-NFB output wedge
-0.4
-0.3
-0.1
-0.2
-0.6
-0.2
-0.5
5. Potential GDP growth ex. level shocks (annual average)
3.1
2.5
2.1
1.5
1.2
1.2
1.3
6. Natural unemployment rate (annual average)
5.39
5.46
5.36
5.72
5.82
5.87
5.75
7. Trend labor force participation rate (annual average)
66.77
66.04
65.49
65.32
65
64.57
64.16
*Includes the effects of trend movements in energy intensity of domestic production.
Table 1.2
Other Estimates of Potential Output Growth
(Percent change)
Average Growth in 2009-2010
2012
As of 2008
As of 2009
Latest
Latest
State-space model 2.1 2.0 1.3 1.3
IMF
1
2.1 0.9 1.3 1.8
OECD
2
2.3 1.5 1.6 1.8
CBO
3
2.4 1.9 1.7 1.7
CEA
4
2.9 2.5 2.1 2.1
Macro. Advisers
5
2.6 1.2 ? 1.2
1. IMF estimates are from October 2008, October 2009, and April 2013.
2. OECD estimates are from December 2008, December 2009, and June 2013.
3. CBO estimates are from September 2008, August 2009, and February 2013.
4. CEA estimates are from January 2009, February 2010, and March 2013.
5. Macroeconomic Advisers estimates are from October 2008, December 2009, and
April 2013.
16,000
15,000
14,000
13,000
12,000
11,000
10,000
9,000
8,000
7,000
90 92 94 96 98 00 02 04 06 08 10 12
2000-2007 trend
actual GDP
potential GDP
billions of chained 2005 dollars
Potential GDP (Level)
-5
-4
-3
-2
-1
0
1
2
3
4
5
6
90 92 94 96 98 00 02 04 06 08 10 12
4-quarter change in actual GDP
4-quarter change in potential GDP
percent
Potential GDP (Growth)
3.5
4.0
4.5
5.0
5.5
6.0
6.5
7.0
7.5
8.0
8.5
9.0
9.5
10.0
90 92 94 96 98 00 02 04 06 08 10 12
actual unemployment rate
natural rate
percent
Natural Rate of Unemployment
63.2
63.6
64.0
64.4
64.8
65.2
65.6
66.0
66.4
66.8
67.2
67.6
90 92 94 96 98 00 02 04 06 08 10 12
actual LFPR
trend LFPR
percent
Trend Labor Force Participation Rate
-0.2
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8
90 92 94 96 98 00 02 04 06 08 10 12
percentage points
Contribution of Actual Capital Deepening
80
82
84
86
88
90
92
94
96
98
100
102
104
106
108
90 92 94 96 98 00 02 04 06 08 10 12
2000-2007 trend
trend MFP
index (2007Q4=100)
Trend MultiFactor Productivity
Figure 1.1 State-Space Model Estimates of Potential GDP and its Components
(shaded region denotes 95% confidence interval)
-10
-8
-6
-4
-2
0
2
4
6
-10
-8
-6
-4
-2
0
2
4
6
1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012
Common cycle (2-sided estimate)
GDP gap (2-sided estimate)
percent
Output Gap
-3
-2
-1
0
1
2
3
4
5
6
-3
-2
-1
0
1
2
3
4
5
6
1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012
Common cycle (2-sided estimate, rescaled)
Unemployment gap (2-sided estimate)
percent
Unemployment Gap
Figure 1.2. Estimates of Resource Utilization
(shaded region denotes 95% confidence interval)
12,000
12,500
13,000
13,500
14,000
14,500
15,000
15,500
16,000
2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017
2008 data*
2009 data*
2010 data
2011 data
2012 data
2013 data
Potential Output
0.4
0.8
1.2
1.6
2.0
2.4
2.8
3.2
2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017
2008 data
2009 data
2010 data
2011 data
2012 data
2013 data
percent
Potential Output Growth (4-quarter)
billions, 2005 dollars
4.4
4.6
4.8
5.0
5.2
5.4
5.6
5.8
6.0
6.2
2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017
2008 data
2009 data
2010 data
2011 data
2012 data
2013 data
Natural Rate of Unemployment
*Estimated using real output and income data reported in 2000 dollars and then adjusted to
2005 dollars using a constant multipliative adjustment factor.
-2
-1
0
1
2
3
4
5
2005 2006 2007 2008 2009 2010 2011 2012 2013
2008 data
2009 data
2010 data
2011 data
2012 data
2013 data
percent
Unemployment Gap
percent
Figure 1.3. Real-Time Estimates and Projections of Potential Output and the Natural Rate
0.4
0.8
1.2
1.6
2.0
2.4
2.8
3.2
3.6
4.0
4.4
0.4
0.8
1.2
1.6
2.0
2.4
2.8
3.2
3.6
4.0
4.4
1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012
baseline model
model with post-1994 shift in Phillips curve slope
model without inflation
model without nonfarm income
percent
Potential Output Growth (4-quarter)
5.0
5.2
5.4
5.6
5.8
6.0
6.2
6.4
6.6
6.8
7.0
5.0
5.2
5.4
5.6
5.8
6.0
6.2
6.4
6.6
6.8
7.0
1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012
baseline model
model with post-1994 shift in Phillips curve slope
model without inflation
model without nonfarm income
percent
Figure 1.4. Sensitivity of Supply-Side Estimates to Changes in Model Specification
Natural Rate of Unemployment
-12
-10
-8
-6
-4
-2
0
2
4
6
-12
-10
-8
-6
-4
-2
0
2
4
6
86 88 90 92 94 96 98 00 02 04 06 08 10 12
EDO production function measure
EDO Beveridge-Nelson measure
state-space model
percent
Figure 1.5
Output Gap Estimates -- State-Space Model Versus EDO
7,000
8,000
9,000
10,000
11,000
12,000
13,000
14,000
15,000
16,000
17,000
18,000
7,000
8,000
9,000
10,000
11,000
12,000
13,000
14,000
15,000
16,000
17,000
18,000
1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 2020
low growth scenario
medium growth scenario
high growth scenario
billions
Level
0.4
0.8
1.2
1.6
2.0
2.4
2.8
3.2
3.6
4.0
4.4
0.4
0.8
1.2
1.6
2.0
2.4
2.8
3.2
3.6
4.0
4.4
1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 2020
low growth scenario
medium growth scenario
high growth scenario
percent
4-Quarter Percent Change
Figure 1.6. Alternative Scenarios for the Future Evolution of Potential GDP
Figure 2.1 - Employment Losses in Housing-Related Industries
2004 2005 2006 2007 2008 2009 2010 2011 2012 2013
40
50
60
70
80
90
100
110
40
50
60
70
80
90
100
110
Index, January 2007 = 100
Private Employment
Construction Employment
Real Estate Employment
Mortgage Finance Employment
Note: Shaded areas are NBER dated recessions.
Figure 2.2 - Dispersion in Employment Change Across Industries
1962 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012
-5
0
5
10
15
20
25
30
35
40
-5
0
5
10
15
20
25
30
35
40
Q2
Idiosyncratic dispersion
Total dispersion
Dispersion attributed to usual cyclicality
Note: Shaded areas are NBER dated recessions.
Figure 2.3 - Variance in Cumulative Change in Industry Employment Shares over the Business Cycle
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
0
1
2
3
4
5
0
1
2
3
4
5
Quarters since peak
Business cycle peak 2007:Q4
Business cycle peak 2001:Q1
Business cycle peak 1990:Q3
Business cycle peak 1981:Q3
Business cycle peak 1973:Q4
Long-term trends in industry employment shares were removed with a Hodrick-Prescott Filter.
Figure 2.4 - Permanent Job Loss
1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8
Percent of employment
0.5
1.5
2.5
3.5
4.5
5.5
6.5
Percent of labor force
Sept.
Stock (right axis)
Rate (left axis)
Note: Shaded areas are NBER dated recessions.
Figure 2.5 - Beveridge Curve
3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 8.5 9.0 9.5 10.0
1.4
1.6
1.8
2.0
2.2
2.4
2.6
2.8
3.0
3.2
3.4
3.6
1.4
1.6
1.8
2.0
2.2
2.4
2.6
2.8
3.0
3.2
3.4
3.6
Job openings rate (percent)
2013:Q3
2013:Q2
2001:Q1 to 2007:Q4
2008:Q1 to 2013:Q1
Unemployment rate (percent)
Note: Observation for 2013:Q3 is average of July and August. Source: Job Openings and Labor Turnover Survey & Current Population Survey.
Figure 2.6 - Industrial Mismatch
2005 2006 2007 2008 2009 2010 2011 2012 2013
0.10
0.15
0.20
0.25
0.30
0.35
0.40
Index
0.0
0.3
0.6
0.9
1.2
1.5
1.8
2.1
2.4
Percentage points (contribution to the unemployment rate)
Mass.: National Bureau of Economic Research, September); Aysegul Sahin, Joseph Song, Giorgio Topa, and Giovanni L. Violante (2010). "Mismatch Unemployment." NBER
Working Paper Series 18265 (Cambridge, Mass.: National Bureau of Economic Research, August 2012), authors provided updated estimates.
June
July
Dec.
Dec.
Industry - Sahin et al. (right axis)
Occupation - Sahin et al. (right axis)
Industry - Lazear & Spletzer (left axis)
Occupation - Lazear & Spletzer (left axis)
Note: Shaded areas are NBER dated recessions.
Sources: Edward P. Lazear and James R. Spletzer (2012). "The United States Labor Market: Status Quo or a New Normal?" NBER Working Paper Series 18386 (Cambridge,
Figure 2.7 - Barnichon-Figura Estimate of Matching Efficiency
1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
Percentage points
Q3
Figure 2.8 - Long-Term Unemployment
1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012
0
1
2
3
4
5
0
1
2
3
4
5
Percent of labor force
Q3
Note: Shaded areas are NBER dated recessions.
Figure 2.9 - Job Finding Rates by Unemployment Duration
2007 2008 2009 2010 2011 2012 2013
0
5
10
15
20
25
30
35
0
5
10
15
20
25
30
35
Percent
Sept.
1-26 weeks
27-52 weeks
53+ weeks
Notes: 3-month moving averages of seasonally adjusted monthly data. Shaded areas indicate periods of business recession as defined by the NBER.
Figure 2.10 - Labor Force Exit by Unemployment Duration
2007 2008 2009 2010 2011 2012 2013
15
20
25
30
35
40
15
20
25
30
35
40
Percent
53+ weeks
27-52 weeks
1-26 weeks
Sept.
Notes: 3-month moving averages of seasonally adjusted monthly data. Shaded areas indicate periods of business recession as defined by the NBER
Figure 2.11 - Disability Insurance (SSDI) Recipiency
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013
2.0
2.2
2.4
2.6
2.8
3.0
3.2
3.4
3.6
3.8
2.0
2.2
2.4
2.6
2.8
3.0
3.2
3.4
3.6
3.8
Percent of population 16 years and over
Note: Shaded areas are NBER dated recessions.
Figure 2.12 - Alternative Measures of Slack
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013
-3
-2
-1
0
1
2
3
4
5
-3
-2
-1
0
1
2
3
4
5
Percentage points
Jobs hard-to-fill
Job availability
Unemployment rate gap
Note: Shaded areas are NBER dated recessions.
Figure 3.1 - Startups and Young Business Employment
1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010
350
400
450
500
550
600
650
Thousands
4
6
8
10
12
14
16
Percent of all Establishments
No. of estabs. (left axis)
Share of all estabs. (right axis)
New Establishments
1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010
0
2
4
6
8
10
12
14
0
2
4
6
8
10
12
14
Percent
Age 1-2Age 0
Note: Shaded areas are NBER dated recessions.
Source: U.S. Census Bureau, Business Dynamics Statistics.
Employment at Startups and Young Businesses
-7
-6
-5
-4
-3
-2
-1
0
-7
-6
-5
-4
-3
-2
-1
0
01 02 03 04 05 06 07 08 09 10 11 12 13 14 15
output gap
inflation (4-qtr)
federal funds rate
percent
Output Gap, Inflation, and Interest Rates
year
-9
-8
-7
-6
-5
-4
-3
-2
-1
0
-9
-8
-7
-6
-5
-4
-3
-2
-1
0
01 02 03 04 05 06 07 08 09 10 11 12 13 14 15
real GDP
potential GDP
capital services
percent
GDP, Potential Output, and Capital Services
year
0.0
0.4
0.8
1.2
1.6
2.0
2.4
2.8
3.2
3.6
0.0
0.4
0.8
1.2
1.6
2.0
2.4
2.8
3.2
3.6
01 02 03 04 05 06 07 08 09 10 11 12 13 14 15
unemployment rate
natural rate
percent
Unemployment and the Natural Rate
year
-1.6
-1.4
-1.2
-1.0
-0.8
-0.6
-0.4
-0.2
0.0
-1.6
-1.4
-1.2
-1.0
-0.8
-0.6
-0.4
-0.2
0.0
01 02 03 04 05 06 07 08 09 10 11 12 13 14 15
actual
trend
percent
Labor Force Participation
year
Figure 4.2. Financial Crisis Scenario Under the Inertial Policy Rule
(deviations from baseline)
-5.0
-4.5
-4.0
-3.5
-3.0
-2.5
-2.0
-1.5
-1.0
-0.5
0.0
02 04 06 08 10 12 14
OC computed assuming no S-S damage (expected outcome) OC computed assuming no S-S damage (actual outcome) OC computed assuming no S-S feedback (expected outcome)
OC computed assuming no S-S feedback (actual outcome) OC computed assuming S-S feedback OC computed with adjusted E/P targeting assuming S-S feedback
inertial policy rule
percent
Federal Funds Rate
year
-5.5
-5.0
-4.5
-4.0
-3.5
-3.0
-2.5
-2.0
-1.5
-1.0
-0.5
0.0
0.5
02 04 06 08 10 12 14
percent
Real Federal Funds Rate
year
-9
-8
-7
-6
-5
-4
-3
-2
-1
0
1
2
3
02 04 06 08 10 12 14
percent
Real GDP (level)
year
-1.6
-1.2
-0.8
-0.4
0.0
0.4
0.8
1.2
1.6
2.0
2.4
2.8
3.2
3.6
02 04 06 08 10 12 14
percent
Unemployment Rate
year
-1.6
-1.4
-1.2
-1.0
-0.8
-0.6
-0.4
-0.2
0.0
0.2
0.4
02 04 06 08 10 12 14
percent
Labor Force Participation Rate
year
-1.8
-1.6
-1.4
-1.2
-1.0
-0.8
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
02 04 06 08 10 12 14
percent
Inflation (4-quarter)
year
Figure 4.3. Macroeconomic Effects of the Illustrative Financial Crisis Under Optimal Control (OC) Policy
Without and With Recognition of Supply-Side Damage and Policy Feedback Effects on Supply-Side Conditions
(FRB/US simulation results expressed as deviations from baseline)
.0
.1
.2
.3
.4
.5
.6
.7
.8
01 02 03 04 05 06 07 08 09 10 11 12 13 14 15
OC computed assuming no S-S damage (actual outcome) OC computed assuming no S-S feedback (actual outcome)
OC computed assuming S-S feedback OC computed with adjusted E/P targeting assuming S-S feedback
inertial policy rule
percent
Natural Rate of Unemployment
year
-1.1
-1.0
-0.9
-0.8
-0.7
-0.6
-0.5
-0.4
-0.3
-0.2
-0.1
0.0
01 02 03 04 05 06 07 08 09 10 11 12 13 14 15
percent
Trend Labor Force Participation Rate
year
-8
-6
-4
-2
0
2
4
6
8
01 02 03 04 05 06 07 08 09 10 11 12 13 14 15
percent
Capital Services
year
-5
-4
-3
-2
-1
0
1
2
01 02 03 04 05 06 07 08 09 10 11 12 13 14 15
percent
Potential GDP
year
Figure 4.4. Supply-Side Effects of the Illustrative Financial Crisis Under Optimal Control (OC) Policy
Without and With Recognition of Supply-Side Damage and Policy Feedback Effects on Supply-Side Conditions
(FRB/US simulation results expressed as deviations from baseline)
-5.0
-4.5
-4.0
-3.5
-3.0
-2.5
-2.0
-1.5
-1.0
-0.5
0.0
02 04 06 08 10 12 14 16 18 20
Outcomes under OC policy in the absence of leverage effects Outcomes under OC policy in the presence of leverage effects
Outcomes under OC policy that constrains the funds rate to avoid leverage effects Outcomes under inertial rule in the presence of leverage effects
percent
Federal Funds Rate
year
-5.2
-4.8
-4.4
-4.0
-3.6
-3.2
-2.8
-2.4
-2.0
-1.6
-1.2
-0.8
-0.4
0.0
0.4
02 04 06 08 10 12 14 16 18 20
percent
Real Federal Funds Rate
year
-8
-7
-6
-5
-4
-3
-2
-1
0
1
2
3
4
02 04 06 08 10 12 14 16 18 20
percent
Real GDP
year
-2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
02 04 06 08 10 12 14 16 18 20
percent
Unemployment Rate
year
-1.6
-1.4
-1.2
-1.0
-0.8
-0.6
-0.4
-0.2
0.0
0.2
0.4
02 04 06 08 10 12 14 16 18 20
percent
Labor Force Participation Rate
year
-2.4
-2.0
-1.6
-1.2
-0.8
-0.4
0.0
0.4
0.8
1.2
1.6
02 04 06 08 10 12 14 16 18 20
percent
Inflation (4-Quarter Rate)
year
Figure 4.5. Illustrative Financial Crisis Under Optimal-Control Policies that Target the Adjusted Employment-Population Ratio
When Persistently Low Interest Rates Generate Adverse Leverage and So Cause a Second Recession
(FRB/US simulation results expressed as deviations from baseline)
