Full text of Working Papers (Federal Reserve Bank of Chicago) : Monetary Policy, Output Composition and the Great Moderation, Working Paper 2007-07

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

Federal Reserve Bank of Chicago

Monetary Policy, Output Composition
and the Great Moderation
Benoît Mojon

WP 2007-07

Monetary Policy, Output Composition and the Great Moderation

Benoît Mojon

y

June 2007, (…rst draft: Novemb er 2005)

Abstract
This paper shows how US monetary policy contributed to the drop in the volatility of US output
‡uctuations and to the decoupling of household investment from the business cycle. I estimate a
model of household investment, an aggregate of non durable consumption and corporate sector
investment, in‡ation and a short-term interest rate. Subsets of the models’ parameters can vary
along independent Markov Switching processes.
A speci…c form of switches in the monetary policy regimes, i.e. changes in the size of monetary
policy shocks, a¤ect both the correlation between output components and their volatility. A regime
of high volatility in monetary policy shocks, that spanned from 1970 to 1975 and from 1979 to
1984 is characterized by large monetary policy shocks contributions to GDP components and by a
high correlation of household investment to the business cycle. This contrasts with the 1960’s, the
1976 to 1979 period and the post 1984 era where monetary policy shocks have little impact on the
‡uctuations of real output.
K eyw ords: business cycle volatility, m onetary p olicy, M arkov Switching VAR
JE L classi…cation: E3 E5

E-mail address: bmojon@frbchi.org; Federal Reserve Bank of Chicago, on leave from the European
Central Bank. Tel.: 1-312-322-2303.
y I wish to express special thanks to Daniel Waggoner and Tao Zha for sharing their code and for invaluable
suggestions in the course of this research project. I also want to thank Klaus Adam, Michel Aglietta, Filippo
Altissimo, Gadi Barlevi, Marco Bassetto, Jean Boivin, Anton Brender, Je¤ Campbell, Steve Cecchetti, Larry
Christianno, Martin Eichenbaum, Charles Evans, Jonas Fisher, Domenico Giannone, Anil Kashyap, Michele
Lenza, Simone Manganelli, Florence Pisani and Daniele Terlizzese for stimulative discussions as well as David
Miller for outstanding research assistance. The paper re‡ects the author’s view alone and neither the ones
of the Federal Reserve Bank of Chicago nor the Federal Reserve System, nor the European Central Bank.

Monetary Policy, Output Composition and the Great Moderation

1.

2

Introduction

This paper investigates the role of monetary policy in two related aspects of the great
moderation: the mid 1980’s decline of US output volatility and the decoupling of household
investment expenditure from the business cycle. This decoupling, which was particularly
obvious during the 2001 recession, sharply contrasts with the high degree of co-movement
between GDP components that prevailed before the mid-1980’s.
Why has household investment become less correlated to other GDP components and
why does it matter for assessing the role of monetary policy in the reduction of US output
volatility? Precisely, as this paper shows, because the monetary policy instability that
characterizes the …fteen years prior to the great moderation ampli…ed the ‡uctuations of
GDP components as well as their co-movement. This co-movement of two GDP components
implies, ceteris paribus, a higher variance of GDP.
These results resolve an ongoing controversy on the role of monetary policy in the great
moderation of output ‡uctuations. A widely held view is that the reaction function of the
Federal Reserve changed when Paul Volcker became Chairman of the FOMC (see for instance,
Clarida, Galì and Gertler, 2000, Cogley and Sargent, 2005a, Lubik and Schorfheide, 2004,
Boivin, 2006, and many others).1 In essence, the response of the short-term real interest rates
to in‡ation increased, became positive and therefore stabilizing for nominal as well as for
real ‡uctuations. However, changes in the coe¢ cients of the central bank reaction function
have little impact on the magnitude of output ‡uctuations. Lubik and Schorfheide (2004)
…nd that, in the indeterminacy regime that best characterizes the data of the 1970’s, the
sunspot shocks raise the variance of in‡ation and the nominal interest rate, but have hardly
any impact on the variance of output. Stock and Watson (2003), who simulate post 1984
monetary policy rules in the 1970’s, conclude that monetary policy "had the major impact of
bringing in‡ation under control but happened not to have had a large e¤ect on the cyclical
volatility of output". Moreover, several econometric tests have failed to reject stability in
the coe¢ cients of the monetary policy rule or the transmission of monetary policy.2 3
1 See also the account of changes in the policy makers views on the controlability of in‡ation and the
desirability of disin‡ation in Romer and Romer (2002), DeLong (1997, 2003) and Cogley and Sargent (2005b).
2 Bernanke and Mihov (1998) is an early example. Among the more recent contributions, Canova and
Gambeti (2005) showed that, although there is some evidence of coe¢ cient instability in the 1979-1982
period, the response of interest rates to in‡ation in the 1990’s is not clearly supperior to 1 nor to the esponse
estimated in the 1970’s. They also show that the monetary policy transmission has changed little over the
last 40 years. See also Leeper and Zha (2003) and Canova (2005).
3 Sims and Zha (2006) attempt to reconcile the con‡icting econometric tests on changes in the "Taylor"
coe¢ cient. They come to the view that one can …nd evidence of changes in the coe¢ cients of the Federal
Reserve reaction function if one is looking for such changes. Correspondinly, they stress that speci…cations
where time variation is limited to the magnitude of the shocks hitting the economy have a much better …t
of the data.

Monetary Policy, Output Composition and the Great Moderation

3

Against this background, I revisit the role of monetary policy in the mid-1980’s drop in
output volatility using a two sector Markov Switching VAR of the US business cycle. The
two sectors are household investment, i.e. the sum of durable consumption and residential
investment, and the rest of domestic private sector demand (DPD, thereafter), i.e. the
sum of non-durable consumption and corporate investment. This decomposition of output
is warranted because the drop of GDP variance is essentially the drop of the variance of
DPD, while net exports and government consumption play hardly any role, and the drop
in the variance of DPD is owed largely to the decoupling of household investment from the
business cycle. The Markov Switching VAR estimation procedure that Sims, Waggoner and
Zha (2006) developed is then used to identify when the model’s parameters changed and
which of these changes in‡uenced the covariance of the two DPD components included in
the model.
I …nd that the change in the size of monetary policy shocks is the only pattern of time
variation that has an impact on both the variance of DPD and the cross-correlation of its
sub-components. DPD variance has been very high in regimes of high volatility of monetary
policy shocks that have taken place from 1970 to 1975 and from 1979 to 1982, and lower in
the 1960s and between 1976 and 1979 and from 1985 to 2006. The regime of highly volatile
monetary policy shocks is one where these shocks have taken the business cycle in and out of
the four recessions that took place between 1969 and 1982. This contrasts sharply with the
1991 and especially with the 2001 recession, during which the monetary policy shocks did
not amplify the slowdown. Smaller monetary policy shocks reduce the relative importance
of a source of business cycle ‡uctuations that has, over the last 50 years, moved household
investment, household consumption and corporate sector investments together.
These results bring a new perspective on the role of monetary policy in the great moderation. They can be seen as an illustration of the impact of a particularly erratic monetary
policy that largely conforms to the narrative analysis of monetary policy by Romer and
Romer (2002). They stress that "in the 1970’s ... monetary policy was used aggressively to
support rapid economic growth... in‡ation became a problem and periodic severe recessions
were necessary to keep in‡ation under check." The amplitude of unsystematic monetary
policy during the early to mid-seventies and, as this paper shows, in the …rst years of Paul
Volcker’s FOMC Chairmanship, added considerable real volatility. The reduced scale of the
unsystematic monetary policy under Greenspan, i.e. not deviating from a systematic "leaning against the wind" interest rate rule, implies much smaller contributions of monetary
shocks to the volatility of GDP growth, in particular around the recessions of 1991 and 2001.
These conclusions di¤er from the view that monetary policy had little e¤ects on the scale
of output ‡uctuations for essentially three reasons. First, the model focuses on a decomposition of output that is highly relevant for the understanding of the great moderation. Most

Monetary Policy, Output Composition and the Great Moderation

4

of the literature relies on one sector models (the above references and Boivin and Giannoni,
2006, Benati, 2006, Benati and Mumtaz, 2007, Blinder and Reis, 2005, Stock and Watson,
2002, 2003, Gordon, 2005, among others). A notable exception is Justiniano and Primiceri
(2006), who analyze the great moderation using DGSE models that disentangle consumption from investment. In line with the state of the art estimated DSGE models, they use
an investment measure that aggregates corporate and household investment. This paper
focuses on another decomposition of demand because household investment have been much
less correlated to either corporate investment or non durable consumption from 1985 to 2006
than between 1965 and 1984.4
This decoupling is likely to re‡ect structural changes in the …nancing of the household
sector. The required downpayment to issue mortgages has fallen steadily since 1976 and an
increasing number of households have used mortgage re…nancing as a source of cash in the
event of a decline in [long-term] interest rates (Canner et al., 2002). To the extent that longterm interest rates are correlated to the business cycle, more households have become able to
shield their expenditure plans from business cycle induced reduction in income (Brender and
Pisani, 2005).5 Turning to automobiles, an important part of durable consumption, Ramey
and Vine (2004) found that the volatility of inventories has dropped in line with in the
persistence of car sales, which is itself driven by changes in the dynamics of interest rates.6
E¤ectively, the last two US recessions, in 1991 and 2001,7 are characterized by an increase
in the household leverage, which contrasts sharply with the four US recessions previous to
these two and may have helped dampen the business cycle. For all these reasons, I estimate
a model that decomposes output to allow an explicit description of household investment
and compare its dynamics to the dynamics of other expenditure components of GDP.
Second, the analysis presented in this paper estimates rather than imposes the dates
of changes in monetary policy regimes. The Sims, Waggoner and Zha (2006) estimation
procedure gives the possibility to test for regimes switches regarding either systematic or
unsystematic monetary policy along Markov processes that are independent from the ones
4 The correlation between the contribution of corporate investment and the contribution of durable consumption and residential investment to the growth rate of US GDP was 0.46 between 1965 Q1 and 1984 Q4.
It was -0.14 between 1985 Q1 and 2006 Q3. Household investment has also become markedly less correlated
to non-durable and services consumption while the correlation betwen corporate investment and non durable
and services consumption has not changed signi…cantly from the two decades before 1984 to the post 1984
period.
5 Campbell and Hercowitz (2006), Cecchetti et al. (2005), Dynan et al. (2006), and Fisher and Gervais
(2006) have also pointed to the broader availability of …nancial instruments as a cause of macroeconomic
stability.
6 Herrera and Pesavento (2004) con…rm a link between the drop in the volatility of inventories and a
reduction in the volatility of manufacturing sales.
7 The 2001 recession was actually the …rst time since 1947 when residitial investment kept a positive
contribution to GDP growth during a recession.

Monetary Policy, Output Composition and the Great Moderation

5

that govern potential changes in the variance of non policy shocks and the response of each
variable of the model to these shocks. The estimation lumps together some of the 1970’s
with a large period of Volcker Chairmanship of the FOMC as one regime of highly volatile
monetary policy shocks. The estimated sub-samples of monetary policy instability give
the model a much more precise account of the impact of monetary instability on output
‡uctuations than the common practise of considering all the observations before or after
1979 or 1982 as one homogenous regime. In this respect, the results con…rm the importance
of abandoning the idea of once-and-for-all non stochastic regime switches (Sims and Zha,
2006).
Third, the impact of monetary policy on output ‡uctuations is not limited to changes
in the coe¢ cients of the policy rule and in particular to what has become refered to the
"Taylor" coe¢ cient, i.e. the response of the interest rate to in‡ation. While both changes
in the coe¢ cients in the policy rule and changes in the size of monetary policy shocks are
explored, only the latter has a large e¤ect on the variance and covariance of GDP components.
The paper goes as follows. Section 2 describes the changes in the covariance structure
of the expenditure components that add up to GDP. Section 3 brie‡y presents the Markov
Switching VAR model and the form of time variation that it can test. Section 4 presents the
timing of switches in the models parameters and how they impact the covariance structure
of GDP components. Section 5 summarizes the main results and concludes.
2.
2.1.

Covariances of GDP components and the great moderation
Changes in correlations of GDP components

There is already a dense literature on the great moderation of US nominal and real
business cycle volatility (see the survey by Stock and Watson, 2002 and more recent references
in Justiniano and Primiceri, 2006). Thanks to this literature, we know that the drop in real
volatility is very much shared across the di¤erent expenditure components that add up to
GDP and that it all starts in 1984.
Few papers also highlight that, beyond single components’ volatility, the correlation
and hence the covariance of some speci…c component has contributed to the fall of GDP
growth variance. These changes in correlation are important for two reasons. First, they
are suggestive of new endogenous adjustments of GDP components and thereby may lead
us to the formulation of structural interpretation of the great moderation.8 Second, they
imply that one sector models could miss an essential characteristic of the greater stability of
8 To some extent, because residuals re‡ect our "ignorance", many economists consider the Good Luck
explanation of the Great Moderation as deeply unsatisfactory.

Monetary Policy, Output Composition and the Great Moderation

6

output ‡uctuations.
Early on, Kahn et al. (2003) reported the change in the covariance of inventories, investment and sales. This evolution can be due to the impact of IT on the supply side of
inventories (McConnel and Perez-Quiros, 2000), to the impact of …nancial innovation on the
demand of durable goods or again to better monetary policy (Ramey and Vine, 2006). More
recently, Irvine and Schuh (2005), who use a supply side decomposition of output, stress
that a large share of the output volatility decline is due to a reduction of the co-movement
of sectorial business cycles. This result lead them to reject the Good Luck hypothesis. Galì
and Gambetti (2007) stress the same intuition that, had good luck been prominent in the
great moderation, we should not observe changes in the correlation of output components.
Finally, Dynan et al (2005) show that savings and income have become more correlated
than they were, an evolution which should provide some form of automatic stabilization.
This, again, could result from the improved completeness of the …nancial instruments available to US households.
In the following, I therefore propose to show whether and how the variance and the
covariance of GDP main components have changed over time. Following the analysis reported
in the …rst section of Stock and Watson (2002), I opt for the simplest form of time variation.
I compare covariances across sub-samples and before and after 1984.9
2.2.

The covariance matrix of DPD components contributions to the growth rate of GDP

My analysis of the (co)variance deviates slightly from Stock and Watson (2002) and
others in a small and but nevertheless useful respect. Given that GDP growth is a weighted
average of the growth rate of GDP components, I compute the full covariance matrix of the
GDP growth contributions of GDP components. The contributions are directly taken from
NIPA Table 1.1.2 published by the Bureau of Economic Analysis.
To start with, I focus on six components of GDP: consumption of durable goods, consumption of non-durable goods, consumption of services, …xed non-residential investment,
residential investment and changes in inventories. These components, whose sum can be
labelled DPD, are the most relevant because, as shown in Table 1, the variances and covariances involving net exports and government consumption had little impact on the post1984 drop in the volatility of GDP growth. Actually, from 1965-1984 to 1985-2007, the
variance of the DPD contribution to GDP growth has dropped more than the variance of
GDP growth.
9 Actually, McConnel and Perez-Quiros (2000) estimated 1984 as the date of the break in the volatility of
US GDP growth. Stock and Watson (2002) con…rmed that the fall in volatility is of a discrete nature and
that its most likely date is indeed 1984.

Monetary Policy, Output Composition and the Great Moderation

7

The second panel of the table reports two summary statistics of the covariance matrix
of these six components for the full post 1947 sample, and three sub-samples of about two
decades each: 1976-1965, 1965-1984 and 1985-2006. Hence the post 1984 era can be compared
to samples of similar length. The …rst summary statistic is the trace of the covariance matrix
which summarizes the change in volatility of each GDP component taken in isolation. The
second is the sum of the o¤-diagonal elements of the co-variance matrix, which allows us to
assess the importance of co-movements in GDP components in the variance of GDP growth
across samples.10
I do this decomposition both for contributions to quarter on quarter annualized GDP
growth rates (in the top panel) and for their ‡uctuations of periodicity comprised between
6 and 32 quarters as extracted by a band pass …lter (in the bottom panel).
First, Table 1 reveals that the great moderation is indeed remarkable. The variance of
the GDP growth rate is divided by …ve, from about 20 % of annualized qoq GDP growth
rate in the second sub-sample (Column c) to 4 % in the third sub-sample (column d). The
volatility of the DPD contribution drops by an even larger magnitude, which is why the rest
of this paper will abstract from net exports and government consumption.11
Second, the covariance terms in the variance of DPD fall from 11 %, i.e. nearly half of
the variance DPD growth variance in the 1965-1984 sample to zero in the post 1985 sample.
And the contribution of co-variances to the drop in variance after 1984 is even larger, in
relative terms, at business cycle frequencies.
Taking a closer look at the data indicates that not all the 15 bilateral cross-correlations
between the 6 contributions to the growth rate of GDP have dropped over time. The bulk
of the drop in the covariances after 1984 comes from the fall in the correlation between
“household investments”, i.e. durable goods consumption plus residential investment (f ht
thereafter) and the rest of DPD (f f cht , thereafter). The time series of these two subaggregate of GDP/DPD are reported in the top panels of Figure 1 and their variance and
covariance are given in Table 2. Both series have had smaller ‡uctuations after the mid1980’s. In particular, f ht has dropped much less in the two recessions of the post-1984
sample than in ones preceding it (see Figure 1). US households have increased their leverage
through out the last two decades (Report on the Survey of Consumer Finance, Federal
Reserve Monthly Bulletin 2006) including in 1991 and in 2001. This evolution may have
dampened macroeconomic ‡uctuations in the sense that household investment becomes less
dependent on their current income (Brender and Pisani, 2005, Campbell and Hercowitz,
1 0 To

…x ideas on the role of correlations and covariances in the variance
of an aggregate, we have V ar(x +
p
y) = V ar(x) + V ar(y) + 2Cov(x; y) and Cov(x; y) = Cor(x; y)
V ar(x) V ar(y):
1 1 The variance of the aggregate of net exports and government consumption also drops after 1984. However,
its covariance with DPD becomes much less negative. This is why, altogether, net exports and government
consumption explain little in the drop of GDP variance around 1984.

8

Monetary Policy, Output Composition and the Great Moderation

2004). E¤ectively, the correlation between f f cht and f ht has become slightly negative and
the covariance of household investment expenditure and the rest of DPD has stopped to
contribute positively to the variance of GDP.
To sum up, the great moderation comes from a more stable DPD and the drop in the
correlation of two speci…c sub-aggregates of DPD accounts for a sizeable fraction of the
output stabilization.12 This evolution may be related to transformations of the US …nancial
system (Dynan et al, 2005, Fisher and Gervais, 2006, Campbell and Hercowitz, 2004 and
references therein) and their e¤ects on household’s expenditure.
It is important to stress that this sectorial decomposition of output is not the one usually
chosen to estimate DSGE models. Christiano, Eichenbaum and Evans (2005), Smets and
Wouters (2007) and their numerous followers bundle together corporate and household investment as "investment" in the model. However, the correlation between the corporate and
houshold investment has dropped dramatically during the last three decades (see footnote 4).
Because this correlation is a constitutive feature of the great moderation, it could be desirable
to estimate a time varying parameters DSGE model that singles out household investment
to analyze the great moderation.13 Rather than addressing this ample task, the following
sections of the paper present the estimates of reduced-form models with time varying parameters in the spirit of Cogley and Sargent (2005a), Canova and Gambetti (2006), Primiceri
(2005), Sims and Zha (2006) among many others. These reduced-form models have the
advantage of being free of speci…c functional form assumptions necessary to estimate DSGE
models.
3.

A Markov Switching VAR evidence on the origin of the great moderation

3.1.

The baseline model

The baseline VAR used for the analyzis models a vector of four endogenous variables:
yt0 = (f f cht ; f ht ;
1 2 Stock

t ; it )

(1)

and Watson (2005) showed that the drop of the variance of GDP growth is an international
phenomenon. Unfortunately, we can’t replicate Table 1 for other countries because the quarterly national
accounts usually do not decompose investment into residential and non-residential. Moreover, quarterly
national accounts data are usually not available before 1980. Partial evidence, based on shorter time series
of consumption and investment show some decline in the 1990’s of the correlation between the two main
components of GDP in the UK, Australia, Italy, Japan, Spain and Belgium, but neither in France nor in the
Netherlands. I leave the systematic study of international data for future research. See also Cecchetti et al.
(2006) for a cross-country assessment of the e¤ects of improvements in the conduct of monetary policy.
1 3 Interestingly, two recent papers, Calza et al (2006) and Iacoviello and Neri (2007), single out residential
investment within estimated constant parameters DSGE models. See also the discussion in Monacelli (2006).

Monetary Policy, Output Composition and the Great Moderation

9

1)
f f cht is the sum of the contributions to the GDP growth rate of the consumption
of non durable goods, consumption of services, non-residential …xed investment and change
in inventories.
2)
f ht , household investment, is the sum of the contributions to the GDP growth
rate of durable consumption and residential investment,
3)
t is the in‡ation rate of the GDP de‡ator,
4)
it is the interest rate on federal funds.
While the small number of variables may limit our ability to identify exogenous monetary policy shock, this model is preferred as a baseline in order to maximize the chances to
detect regime switches in the data. I however checked that the main results hold in models that contain additional variables, including monetary aggregates, credit aggregates and
commodity prices.
The model is estimated over the 1960-2006 sample with quarterly data.14 The identi…cation of monetary policy shocks is recursive as in Christiano, Eichenbaum and Evans (1999).
In particular, the central bank information set comprises contemporaneous developments in
the other variables of the model. The three other shocks, which are not given any structural
interpretation, are also orthogonalized via a Cholesky decomposition of the VAR innovations.
One important shortcoming of analyzing the dynamics of yt0 with a standard, constant
parameters, VAR model is that it would imply that the variables have a constant covariance
matrix. It would therefore be totally at odd with the evidence of a decline in the variance
of US output. This is why we turn to the recently developed estimation procedure of Sims,
Waggoner and Zha (2006).15
3.2.

The Sims,Waggoner and Zha (SWZ) methodology

SWZ have developed a Bayesian estimation of VAR models where subsets of parameters
of the VAR evolve along independent Markov Switching processes. This section gives a brief
overview of their approach. For a comprehensive presentation, the interested reader should
report to SWZ and the seminal contributions by Hamilton (1989, 1994) and Kim and Nelson
(1999).
Following SWZ, we assume that the time series of interest can be modelled as a VAR
which parameters depend on unobserved Markov Switching processes. The model can be
1 4 The

time series where obtained from the BEA web site and, for the interest rate, monetary, credit and
the index of commodity prices, from HAVER.
1 5 Another possibility would be to …t a VAR with drifting coe¢ cients and variance, as performed by
Canova and Gambeti (2006), Cogley and Sargent (2005a) and Primiceri (2005). A systematic comparison of
the Markov Switching VARs and the VAR with drifting parameters would, however, goes beyond the scope
of this paper.

Monetary Policy, Output Composition and the Great Moderation

described as
0

yt A0 (st ) =

p
P

i=1

0

0

yt i Ai (st ) + "t

1

(st )1 for

t

T

10

(2)

where
p is the lag length,
yt is an n dimensional vector of endogenous variables (e.g. (1)),
each observation belongs to one of k states st with some probability and for each of the
1
k states of the model, A(k) is invertible and Ai (k) are n n matrices and
(k) is an n n
diagonal matrix.
Q is a Markov transition matrix of which the element qi;j gives the probability that st is
in state i and st 1 inX
state j.
qi;j = 1.
All qi;j 0 and
j

3.2.1.

Restrictions on the time variation and the transition probability matrix

The pattern of time variation as formulated in (2) is so general that it may be impossible
to estimate for lack of degrees of freedom. Without restrictions on time variation, a VAR of
n macroeconomic variables with p lags and k states would imply the estimation of k (n
(n p + 1) + (n (n 1)=2)) parameters for the VAR plus k (k 1) free elements in the
matrix of cross-states transition probabilities Q.16 This is why SWZ introduced restrictions
on the time variation patterns allowed in the model.
First, the time variation of the parameters is constrained to the long run e¤ects of variable
j on variable i, though the lag structure of the impact is similar across states. This means
that each additional state add only n n more parameters on top of the (n (n p + 1) +
(n (n 1)=2)) that are estimated in the constant parameter VAR.
Second, the transition probability matrix Q is the Kronecker product of sub-transition
probability matrices Qk that are associated to independent Markov processes. Each of these
will, eventually, govern changes in a subset of VAR parameters. We have
Q = Q1

:::

Q

where is the Kronecker product.
Such a structure for Q greatly reduces the number of transition probabilities to be estimated. In the case of our benchmark model, a 4 equations VAR, we allow the variance of
each of the four shocks of the model to take one of two values between 1960 and 2006. In
1 6 To …x ideas, a 4 variables VAR with 4 lags and 12 regimes would imply that 1020 free parameters have
to be estimated.

Monetary Policy, Output Composition and the Great Moderation

11

other words, we associate an independent two states Markov process to the variance of each
shock. This amounts to allowing for 24 = 16 di¤erent states. An unrestricted transition
probability matrix Q would require the estimation of 16 (16 1) = 240 free parameters.
Instead, the composition of four 2 2 transition probability matrices involve the estimation
of 8 free parameters.17
This formalization of the transition across regimes is very well suited for our purpose
because subsets of the parameters of a time series model can change value along independent
Markov processes. Hence changes in the coe¢ cients or the error term variance of each of
the model’s equations can evolve depending on a speci…c Markov process. This is convenient
because each subset of parameters may pertain to a speci…c economic hypothesis. The main
advantage of this methodology over the one used by Sims and Zha (2006) is the ability to focus
on particular parameters of the model and check whether the data point to regime switching
for these parameters independently from regime switching that would signal changes of other
parameters of the model.
For example, consider using a standard VAR of output growth, in‡ation and the shortterm interest rate to assess the relative importance of “Good Luck, Good Practices and
Good Policies”. A …rst Markov Process would be used to model “Good Luck” as potential
changes in the size of the shocks a¤ecting output and in‡ation. A second Markov Process
would be associated to “Good Practices”, i.e. to changes in the propagation of shocks to
the economy, i.e. the coe¢ cients of the output growth and the in‡ation equations. A third
Markov Process would pertain to “Good Policies”, changes in the coe¢ cients of the central
bank interest rate rule.
Provided the estimation spots 2 regimes for each of the three Markov processes, we now
would have characterized the sample period by 2^3 = 8 regimes. However, we can more easily
assess which one of the changes in the three subsets of parameters has indeed happened in
the mid-1980’s and whether this change of parameter translate into a sizeable impact on the
volatility of either in‡ation or GDP growth. This approach can actually be taken a step
further allowing for instance the dynamics of in‡ation to change at a di¤erent date than the
dynamics of output growth. This would imply testing for altogether 16 states characterizing
the VAR parameters for the sample under review, though identifying the dates of regime
shifts separately for each subset of parameters. This procedure is clearly more tractable
than the one consisting of allowing for 16 states where all parameters can change and then
identifying ex post which states correspond to this or that subset of parameters.
1 7 The

set up of SWZ further allows one to introduce linear restrictions on the elements of the transition
probability matrices Q :A typical application would be a zero restrictions to impose absorbing states. See
SWZ section 2.

Monetary Policy, Output Composition and the Great Moderation

3.2.2.

12

Estimation

First, the prior on the parameters , which does not vary across states, is similar to the one
introduced in Sims and Zha (1998), i.e. it is a generalization of Litterman’s prior. The prior
distribution of the VAR coe¢ cients is normal while the prior distribution of the variances of
the shocks is a gamma distribution.18 The prior for the elements of the transition probability
have a Dirilichet distribution (i.e. a convenient multidimensional Binomial distribution). The
diagonal elements of the transition probability are much larger than the o¤ diagonal elements
to re‡ect the view that the economy tends to stay in one state once it is in that state.19
The formulation of the prior uses the independence of the parameters of the model ,
the transition probability Q and the an initial guess that at time 0 every states has an equal
probability 1=h: The posterior distribution of , conditional on the data, the sequence of
states and the transition probability matrix across states is estimated by sampling sequentially from the conditional distribution of the parameters similar to a Gibbs sampler (see
SWZ section 4).
Given that the posteriors have unknown shapes the estimation is very computationally
intensive. However, SWZ have developed executables in C that can easily be implemented.20
The estimation of the mode of the posterior distribution takes up to one hour on a standard
PC and the estimation of the marginal data density up to a couple of days for most of the
models reported in this paper.
4.
4.1.

Results
Model selection

I estimate several versions of the model which di¤er in terms of the set of parameters
allowed to take di¤erent values across states of a Markov switching process and the number
of states these processes can have. Given that the estimation allows for large combinations
of time variation (actually twice the number of equations independent Markov Switching
processes) it is important to set a criterion for accepting or rejecting the relevance of each of
1 8 The models are estimated with the same hyper parameters of the priors as Sims, Waggoner and Zha
(1998), i.e., in the notation of Sims and Zha (1998), 0 = 1:0, 1 = 1:0, 2 = 1:0, 3 = 1:2, 4 = 0:1,
5 = 1:0 and 6 = 1:0.
1 9 Speci…cally, the prior distribution of the diagonal elements Qk has its mode at 0.85. In the two states
case, this implies that the prior value for the o¤ diagonal elements of Qk is 0.15. In the cases where three
states are allowed for in the estimation, we rule out transitions from state 1 to state 3 and from state 3 to
state 1 (q13 = q31 = 0). The prior mean for the transition probability from state 1 to state 2 and from state
3 to state 2 is 0.15 and from state 2 to either state 1 or state 3 is 0.075.
2 0 Programs can be downloaded from Tao Zha’s webpage:
http://home.earthlink.net/~tzha_center/

Monetary Policy, Output Composition and the Great Moderation

13

the time variation parts of the model. My criterion is to allow for time variation only when
each of the states has probability 1 for at least some time in the sample.
This criterion has e¤ectively restricted the variance of each shock to take up to three
states and the coe¢ cients up to two states. Additional states appear unlikely in the sense
that they typically never have a high or even positive probability for the sample period. The
…t of the di¤erent versions of the model is reported in Table 3.
The best …t is found for a variance only change model that allows for three di¤erent states
for the variance of the monetary policy shocks, two states for the in‡ation shocks and 3 that
govern jointly the two shocks of the …rst two equations of the model (f f cht and f ht ). The
matrix therefore takes up to 18 di¤erent values as the state of the model changes through
out the 1960-2006 sample.
Second, among models that admit changes in their coe¢ cients, i.e. in A0 and A1 :::A4 ,
the only one that improves the …t markedly with respect to the constant parameters model
allows for two states in the coe¢ cients of the interest rate equation. Moreover, allowing for
changes in coe¢ cients in more than one equation dramatically worsens the …t of the model.
Third, models that combine two states for the coe¢ cients of the interest rate equation
and, in addition, allow for changes in the variance of the model’s shocks do relatively well,
even if not as well as the best …tting model.
The most surprising aspect of these results is that the only equation for which coe¢ cients
may have changed is the monetary policy rule.21 The major transformations of the US
economy (increased openness, improved completeness of …nancial markets, the rise of IT
technologies and other forms of technical changes), as well as the great moderation itself
would naturally lead one to expect more changes in the propagation of shocks since 1960.
The results may be taken as an indication that these changes are gradual in nature and
therefore hard to capture by models with "abrupt" Markov Switching processes.22
In principle, these results are also subject to the Benati and Surico (2006) critic, i.e. the
risk that VARs estimate heteroscedasticity of shocks even in cases where the data generating
process admits a break in the coe¢ cients of the policy rule. However, we show that the
Markov Switching VAR can indeed capture changes in the coe¢ cients of the policy rule
though not the ones related to the Taylor principle.23
2 1 This is also true in versions of the VAR that have more variables including, credit, money or the price
of commodities.
2 2 Evidence of such changes were indeed obtained with VAR with drifting coe¢ cients by Cogley and Sargent
(2005) and Primiceri (2006) but not by Canova and Gambetti (2006).
2 3 See also Leeper and Zha (2001) for a critical assessment of the impact of the Taylor principle on the
stability of macroeconomic models.

Monetary Policy, Output Composition and the Great Moderation

4.2.

14

The constant parameter VAR benchmark

In the constant VAR model, which actually sets the prior for the estimates of the time
varying models (Sims, Waggoner and Zha, 2006), the impulse responses of the interest rate
and the e¤ects of the monetary policy shocks are standard. The interest rate increases following positive shocks to either of the two GDP components and following positive in‡ation
shocks. The model’s account of the transmission mechanism is also standard. A monetary policy tightening triggers a temporary decline in the two demand components and a
somewhat delayed decline in in‡ation.24
However, the time series of the shocks estimated in this constant parameter model (Figure
2) show a sharp increase in their variance for the early 1970’s and the early 1990’s. There
is also some milder evidence of heteroscedasticity of the three other shocks, which all tend
to have a lower variance after 1990. This …rst evidence of time variation in the size of the
shocks is then analyzed formally in the following sections.
4.3.

The timing of regime switches

In this section, I describe the timing of regime switches for four cases of typical speci…cation for the time variation that a¤ect subsets of the VAR parameters.
4.3.1.

Case 1: 2 states for the variance of each equation’s shock

Let’s …rst assume that the variance of each shock can take one of two values and that the
timing of this change in values is independent across equations. I therefore have potentially
four independent changes in regimes along the sample period, i.e. we can have up to sixteen
di¤erent states, each having a speci…c . Although this model does not have the best …t, it
provides a good support to illustrate the e¤ects of time variation in this class of models.25
Figure 3 reports the time series of the probabilities of the high volatility regime for each
of the equations’ orthogonalized residuals.26 Two observations are in order. First, these
estimates con…rm that the volatility of the shocks changes in the early to mid-1980s, i.e. when
Kim and Nelson (1999), McConnel and Perez-Quiros (2000) and Stock and Watson (2002)
date the great moderation. The shocks a¤ecting our …rst domestic demand component,
in‡ation and the federal funds rate drop around that time.
2 4 The impulse responses are also standard in versions of the model where M1 and the prices of commodities
are included. In particular, the shock to the interest rate, which triggers a decline in M1, can be interpreted
as a money supply shock.
2 5 Sims and Zha (2006) have showed that models that assign independent Markov processes to states of
the variance of the structural shocks …t the US data very well. A result that this paper con…rms.
2 6 The time series of the 4 shocks are very similar to the ones reported in Figure 3.

Monetary Policy, Output Composition and the Great Moderation

15

Second, I notice that the change in the volatility of the shocks to the household investment
is dropping only in the early 1990s, and that the shocks a¤ecting the in‡ation equation and
the monetary policy shocks have a higher volatility only between 1970 and 1985, with, for
the latter shock, an intermediary period of moderate volatility between 1976 Q227 and 1979
Q2. This suggests that the great moderation may partly re‡ect the end of a speci…c high
volatility episode, that took place in the 1970’s and early 1980’s and not strictly a secular
evolution (as stressed by Blanchard and Simon, 2001). In particular, the high volatility of
the monetary policy shock and the shock to the in‡ation equation covers the period of the
four (NBER dated) recessions that took place between 1969 and 1982. It also shows that it
is important that each shock can change variance independently from other shocks.
4.3.2.

Case 2: Best …t model

Among speci…cations of changes in the states of the model’s variance, the one with the
best …t allows for three states for both equation 1 and 2, two states for the variance of the
shock hitting the in‡ation equation and three states for the variance of the monetary policy
shocks. The grouping of the time variation for equation 1 and 2 is an information e¢ cient
way to economize on parameters. Turning to the Monetary Policy shock variance, allowing
for a three regimes greatly improves the …t of the model (Table 3).
The timing of the high volatility regimes is reported in Figure 4. The main change with
respect to the previous model is that I now have a regime of moderate volatility of the
monetary policy shocks, from 1969 Q2 to 1972, and a high volatility regime which spans
from 1973 Q3 to 1976 Q1, from 1979 Q4 to 1982 Q3 and for the last quarter of 1984.
4.3.3.

Case 3: Changes in coe¢ cients

Among the speci…cations allowing for changes in the propagation of shocks, the models
that admit two states for the coe¢ cients of the interest rate equation are the only ones
that imply a signi…cant improvement in the …t of the model with respect to the constant
parameter model or models with variance only changes. For all the other equations of the
system, there is little support against the benchmark of no changes in the coe¢ cients. This
…nding partially con…rm the conclusions of several previous studies that failed to reject
instability in the VAR coe¢ cients.28
The …rst version of this model is one where I allow for 2 states for the coe¢ cients in
the interest rate equation with constant variances. This speci…cation is interesting to report
2 7 Dates of regime switching are given by the …rst observation when the probability of being in a regime is
supperior to 0.5.
2 8 See in particular, Bernanke and Mihov (1998), Canova (2005), Canova and Gambetti (2006) and Sims
and Zha (2006)

Monetary Policy, Output Composition and the Great Moderation

16

even if it does not have the best …t because some, like Benati and Surico (2006) have argued
that VAR estimates tend to confuse changes in variances for what are actually changes in
coe¢ cients of the reaction function.
Interestingly, the dates of the changes in coe¢ cients of the interest rate equation coincide
with the dates estimated for a change in the variance of the monetary policy shock in the
case where these could change only for two states. The reaction function coe¢ cient changed
from 1969 Q3 to 1975 Q4 and from 1979 Q4 to1985 Q4. The main change in the coe¢ cients
is not so much related to the "Taylor principle" as to the reaction of the interest rate to the
household investment and to the size of the monetary policy shocks.
These results point to a timing of change in the central bank’s behavior that is very
di¤erent from the conclusions stressed in the literature so far (Clarida et al., 2000, Cogley
and Sargent, 2005a, Lubik and Schorfheide, 2004 and many others). First, it puts together
in one regime the period when in‡ation took o¤ and the Volker disin‡ation of the early
1980’s. The model therefore lumps together some of the Burns and most of the Volker era
as a distinctive regime that contrasts with the rest of the last half century. Perhaps more
importantly the nature of the change in the systemic part of policy seems more related to
changes in the mix of response to di¤erent components of demand and to the variance of
the monetary policy shock around the policy rule than to the reaction of the interest rate to
in‡ation.29
For the …rst part of the 1970’s and the …rst half of the 1980’s, the federal funds rate
is estimated to react to household investment four times more than it responds to the rest
of domestic demand (Table 4 upper panel). Outside this regime, the interest rate response
to household investment becomes slightly negative, the response to the rest of domestic
demand and to in‡ation increases slightly in the short run but hardly in the long run given
that interest rates are estimated to be less autocorrelated in this other regime. The second
sizable di¤erence across regimes is that the standard deviation of monetary policy shocks are
nearly …ve times as large in the former regime than in the latter. It therefore appears that,
to some extent, the change in regime of the coe¢ cients actually accounts for a change in the
amplitude of non-systematic monetary policy shocks.30 This result reinforces the case for
the importance of changes in the size of monetary policy shocks during this sample period.
2 9 This second change is actually consistent with the variance-only-change versions of the model and it
drives most of the result that brings along the change in the share of the variance of demand that is due to
interest rate shocks. This was clear in simulations (not reporte for the sake of space) where the impact of
the shocks was rescaled to analyse the sole e¤ects of the change in coe¢ cients.
3 0 In the model’s speci…cation, the strucutral shocks are normalized to a variance of one in the case where
is constant accross regimes so that the coe¢ cients measuring the impact of shocks on the variables belongs
to the set of coe¢ cients that evolve across states. Therefore, the relative size of the e¤ects of these shocks
can change change accross regimes together with all the coe¢ cients of a given equation.

Monetary Policy, Output Composition and the Great Moderation

4.3.4.

17

Case 4: The Full Monty

A natural last step is to estimate models where both the variance of the shocks and the
coe¢ cients of the model can change. One version of these models combines three states
for the variance of the …rst two shocks, two states for the variance of the in‡ation equation
shock, two states for the monetary policy shock and two states for the coe¢ cients of the
interest rate rule. The …t of this model slightly under-performs the …t of the variance only
changes model presented as Case 2.
To a large extent, the timing of regime switches in this version of the model corresponds
to a combination of case 2 and case 3 presented above. The coe¢ cients of the monetary
policy rule change consistently from 1969 Q3 to 1975 Q4 and from 1979 Q4 to1985 Q4 with
respect to their value for the rest of the sample period (see the fourth panel of Figure 5 and
the bottom panel of Table 4). Turning to the variance of the monetary policy shocks, the
estimates put a high probability of a higher variance regime from 1960Q1 to 1960 Q4 and
from 1978 Q4 to 1982 Q3 than for the rest of the sample period. The variance of the other
shocks of the model change at identical dates to the ones estimated for case 3.
Altogether, as illustrated by these four typical speci…cations, there is ample evidence of
time variation in the parameters of the VAR. More speci…cally, each shock has changed in
scale at di¤erent times and the scale of the monetary policy shocks changes has been the
largest. There is also some evidence of changes in the coe¢ cients, though, only for the
monetary policy rule. Moreover, an important aspect of this regime switch is the change in
the impact of the monetary policy shock on the interest rate. It is therefore in part equivalent
to versions of the model where only the variance of the shocks can change over time. Hence,
a consistent result across all speci…cations of time variation that I have explored is that
unsytematic monetary policy has been more erratic for the …rst part of the 1970’s and for
a few years after 1979 from what it was uniformly during the 1960’s, the post 1985 period
and from 1976 to 1979.
4.4.

Regime switches and the variance of output

I can now use the models with time varying parameters to describe the evolution od
output volatility. I compute the unconditional variance, the correlation of the variables and
variance decompositions for each of the states and analyze their evolution as given by the
probabilities of being in each state.31
3 1 The

results reported in this section and the next are based on the estimates obtained for the case 1 form
of time variation, i.e. when each shocks can be in one of two states. Results for either case 2 or case 4, which
is itself a generalization of case 3, are essentially similar. Some of these results are presented in appendix
tables A1 and A2.

Monetary Policy, Output Composition and the Great Moderation

18

Figure 6 reports the standard deviation of f f cht + f ht , and the correlation of f f cht
and f ht , as implied by the changing states of the VARs parameters. These time series are
weighted averages of these statistics across states where the weights are the probabilities of
being in each state.
The time varying VAR captures the great moderation of output ‡uctuations quite well.
The standard deviation of DPD declines from a peak of nearly 5.0 percent in the late 1970’s
to as low as 2.5 % in the early 1990’s. The model also shows a sizeable decline in the
correlation between household investment and the rest of DPD in the early 1980’s.
Figure 6 reports the evolution of domestic demand volatility and cross components correlation, holding the volatility of the monetary policy component at the value it has in the
high volatility regime. About half the drop in output volatility and all of the drop in crosscomponents correlation recovered by the model are associated to changes in the magnitude
of the monetary policy shocks.
How did monetary policy contribute to the great moderation? The model’s answer to
this question is that monetary policy shocks have spurred the real ‡uctuations of the early
1970’s and early 1980’s. And, because monetary policy shocks are a source of ‡uctations that
moves GDP components in the same direction, regimes where the variance of these shocks
is high are periods when the correlation between DPD components is higher. In facts, as
can be seen in Figure 1, the four recessions between 1969 and 1982 are characterized by the
joint decline of f f cht and f ht .
This interpretation is reinforced by the comparison of the forecast error decomposition
across states (Table 5). In the regimes of high volatility of the monetary policy shocks,
the latter explain up to 44 and 53 % of the variance of the f f cht and f ht at horizon 8
to 16 quarters.32 This is much more than what is estimated in the constant parameter
VAR model (see upper panel of Table 4, that averages all regimes) and even more than in
the pre-1970’s post 1985 regime, where monetary policy shocks have a negligible impact on
DPD components. Moreover, it is remarkable that in the low volatility regime, most of the
variance of the monetary policy instrument can be traced back to developments in demand
and in in‡ation. In other words, monetary policy shocks have not been a source of instability
except in the regimes where the volatility of monetary policy shocks is high.
There results are robust to a number of changes in the spe…cication of time variation in
the model or the set of variables included in the VAR. The Markov Switching VARs implies
a large drop in the variance of DPD during the 1980’s and that the changes in the scale
of monetary policy shocks are the only ones that led to a drop in the variance and the
3 2 These magnitude rise to 63 and 72 % (and to 50 and 59 %) for some of the regimes of high volatility in
the policy shocks in case 2 (in case 4). See Table A1 and A2.

Monetary Policy, Output Composition and the Great Moderation

19

correlation of DPD components in the …rst half of the 1980’s.33
4.5.

Monetary policy and output ‡uctuations

The contribution of monetary policy shocks to output ‡uctuations can best be visualized
by the historical contribution of these shocks to the two components of DPD included in the
model. This is done in Figure 7 which reports the time series of monetary policy shocks and
their historical contributions to the federal funds rate, to f f ch and to cf , 34 together with
the dating of regime shifts as estimated with the SWZ methodology and NBER recessions.
The top row of graph reports the timing of the high volatility when two regimes are allowed
for. The bottom two graphs show how, between 1969 and 1982, the monetary shocks explain
the evolution of both GDP components in and out of the recessions. This contrast sharply
with 1991 and 2001, when monetary policy shocks revamp the economy out of the recession
but did not contribute much to getting it there in the …rst place.35
I also use the historical decomposition of the VAR to simulate a counterfactual experiment
whereby, for the full sample, the Monetary Policy shocks are re-scaled to a "Greenspanian"
amplitude. As can be seen in Figure 8, this reduction of monetary policy shocks volatility
greatly reduces the volatility of DPD between 1970 and the mid-1980’s. Monetary policy
shocks had been essentially stimulative between 1970 and 1975. As the result, output appear
to have been stimulated twice in 1972 and in 1976 by nearly 2 percent in excess of what it
would have been if the policy shocks had the same scale as in the 1960’s or under the Chairmanship of Greenspan. The reverse situation occurs from 1978 to 1984. An important aspect
of this experiment though, is that the …rst subsample of monetary policy volatility comprises
some observations of "excessive tightening" and the second one, under the Chairmanship of
Volker, contains episodes of "excessive loosening" of the monetary policy instrument. These
occurences explain why the model outlines these two periods as one regime of excess volatility of unsystematic monetary policy rather than two di¤erent regimes with di¤erent sets of
coe¢ cients.
3 3 In the case of the best …t model (case 2), the regime that spans from 1973 to 1975 and from 1979 to
1982, where the variance of monetary policy shocks is the highest, implies a correlation of 0.54 between the
two DPD components, while this correlation is only 0.28 in the low volatility regime.
3 4 These are actually based on standard …xed parameter VARs, estimated by OLS. The times series of the
two GDP components are 4 quarters averages of the contributions in order to improve the readibility of the
graph. This transformation of the data does not change the timing of the regimes shifts estimated in the
SWZ Markov Switching VARs.
3 5 Extending the sample period to include the second half of the 1950s indicates that the monetary policy
shock was stimulative during the 1957 recession and did contribute little to the 1960 recession. This reinforces
the view that monetary policy had a special role in either triggering or reinforcing the recessions in the 1970’s
and early 1980’s. See also the discussion in Romer and Romer (2002).

Monetary Policy, Output Composition and the Great Moderation

4.6.

20

Robustness

The …nding that monetary policy shocks have had such a large impact on the business
cycle volatility could in principle re‡ect the in‡uence of other variables omitted in the model.
First, in‡ation of commodity prices shocks are the number one suspect for the turbulences
the US economy went through in the 1970’s. Second, Sims and Zha (2006) insist on the role
of money in the decisions of the Federal Reserve in the 1970’s. The omission of money from
the model could lead one to confuse monetary policy shocks for what have actually been
changes in the federal funds rate driven by money demand shocks.
The model (1) was therefore augmented in turn with the yoy in‡ation rate of commodities,
the yoy growth rate of M1 and with both of these variables. In these estimates, not reported
for the sake of space, I allow for changes in the volatility of the shocks a¤ecting the equation
of either variables and …nd indeed two regimes for each of these volatilities.36 However, this
does not a¤ect the timing of the changes in the variance of monetary policy shocks, nor the
impact of these changes in the covariance matrix of domestic demand components.
5.

Summary and conclusions

An important characteristic of the sharp decline in the US business cycle volatility is the
drop in the correlation between household investments, i.e. the sum of residential investment and durable consumption, and the aggregate of corporate investment and non-durable
goods and services consumption. I therefore investigate the joint dynamics of these two
sub-aggregates of GDP in a time varying Markov Switching VAR model that also includes
in‡ation and the short-term interest rate.
I …nd strong evidence of higher volatility of all shocks during a period starting with the
1969 recession and ending in the aftermath of the 1982 recession. In particular, the change
in variance has been the largest for monetary policy shocks between 1970 and 1975, and
from 1979 Q4 to 1984. During these episodes, the variance and the correlation of the two
GDP components are much higher than for the rest of the last …ve decades.
These results point to an important role of monetary policy instability in raising the
volatility of GDP in the 1970’s and early 1980’s. Interestingly these e¤ects of monetary
policy are not associated to changes in the coe¢ cients of the monetary policy rule, nor do
they square with the usual timing of monetary policy regimes into the terms of the Presidents
of the FOMC. In particular, the regime of high monetary policy includes periods of Arthur
Burns’Chairmanship and several years of the one of Paul Volcker.
3 6 Models with changes in coe¢ cients for these equations were also estimated, but the speci…cations with
changes in the volatility of shocks had a much better …t.

Monetary Policy, Output Composition and the Great Moderation

21

One the one hand, the results are consistent with the view that Alan Greenspan has
created little GDP volatility. On the other hand, they also also stress that the 1984 sudden
drop in the US business cycle volatility largely re‡ects the end of a very particular regime of
monetary policy. One where monetary policy had such a large in‡uence on the ‡uctuations
of demand components that they were highly correlated to one another, thereby further
spurring the variance of GDP. After 1984, the Federal Reserve could be praised for having
managed to avoid the return to this type of interest rate instability as the short-term interest
rate remained much closer to a "neutral" monetary policy stance. This may be seen as an
important achievement of Alan Greenspan.
Three other aspects of the results should be highlighted. First, although the great moderation comes largely from the variance of the VAR shocks, it does not imply that "Good
Luck" has been the driving factor of output stabilization. This paper has stressed the role
of unsystematic monetary policy which re‡ects the distance of the interest rate from the
estimated monetary policy rule. The paper showed how the drop in scale of non-systematic
monetary policy contributed to the mid-1980s drop of the variance and correlation of output
components. Second, the results are consistent with the view that it is easier for the central
bank to avoid recessions or not having to trigger recessions when in‡ation expectations are
anchored than when they are not. They provide a quantitative assessment of the changes
in the conduct of monetary policy described in the narrative analyzis of Romer and Romer
(2002).
Finally, I do not …nd changes in the propagation of shocks. The e¤ects of monetary policy
on the two GDP components that I analyze in the model have not changed. This is somewhat
surprising given the evolution of …nancial markets and, for instance, the increasing recourse
of households to mortgage re…nancing at times of declining interest rates. The Markov
Switching VAR simply did not capture changes in the coe¢ cients of demand components
equations that would improve the …t of the model to the data. One potential explanation is
that Markov Switching regimes are not able to spot gradual changes in the coe¢ cients of the
model. It would be useful that future research …nd out whether time variation models with
parameters’drift capture an evolution of GDP components’responses to monetary policy.
References
Ahmed S., An. Levin, and B.-A. Wilson (2004), “Recent U.S. Macroeconomic Stability:
Good Policies, Good Practices, or Good Luck?” Review of Economics and Statistics
86 (3): 824-832.

Monetary Policy, Output Composition and the Great Moderation

22

Baxter, M. and R. King (1999), “Measuring Business Cycles: Approximate Band-Pass Filters
For Economic Time Series”, Review of Economics and Statistics 81(4), 575-593.
Benati L. and P. Surico (2006): “VAR Analyses of the Great Moderation”, mimeo, Bank of
England.
Benati L. and H. Mumtaz (2007), “US Evolving Macroeconomic Dynamics: A Structural
Investigation”, ECB Working Paper 746.
Bernanke, B., and I. Mihov (1998): “Measuring Monetary Policy”, Quarterly Journal of
Economics, 113 (3), 869–902.
Blanchard, O., and J. Simon (2001): “The Long and Large Decline in U.S. Output Volatility”,
Brookings Papers on Economic Activity, 135-164.
Blinder A. And R. Reis (2005), “Understanding the Greenspan Standard”, Princeton University, CEPS Working Paper No. 114.
Boivin, J. (2006): “Has U.S. Monetary Policy Changed? Evidence from Drifting Coe¢ cients
and Real-Time Data,” Journal of Money, Credit and Banking, 38 (5), 1149-1174.
Boivin J. and M.P. Giannoni (2006), “Has Monetary Policy Become More E¤ective?”, The
Review of Economics and Statistics, 88 (3): 445-462.
Brender A. and F. Pisani (2005): The New American Economy, Economica, Paris.
Calza A., T. Monacelli and L. Stracca (2007), “Mortgage Markets, Collateral Constraints
and Monetary Authority: Do Institutions Matter?”, CEPR Discussion Paper 6231.
Campbell J. and Z. Hercowitz (2004), “The Role of Households’ Collateralized Debts in
Macroeconomic Stabilization”, Federal Reserve Bank of Chicago WP 2004-24.
Canner G., K. Dynan and W. Passmore (2002): “Mortgage Re…nancing in 2001 and Early
2002”, Federal Reserve Bulletin, December, 469-481.
Canova F. (2005), “Monetary Policy and the Evolution of the US Economy: 1948-2002”,
Pompeu Fabra Universtity, mimeo.
Canova F. and L. Gambetti (2005): “Structural Changes in the US Economy: Bad Luck of
Bad Policy?”, mimeo, University Pompeu Fabra.
Cecchetti SG, A Flores-Lagunes and S Krause (2005), “Assessing the Sources of Changes
in the Volatility of Real Growth”, Proceedings of the Reserve Bank of Australia Annual
Conference, 2005.

Monetary Policy, Output Composition and the Great Moderation

23

Cecchetti SG, A Flores-Lagunes and S Krause (2006),“Has Monetary Policy Become More
E¢ cient? A Cross-Country Analysis”, The Economic Journal 116 (2), 408-433.
Christiano L., M. Eichenbaum and C. Evans (2005), “Nominal Rigidities and the Dynamic
E¤ects of a Shock to Monetary Policy”, Journal of Political Economy 113 (1), 1-45.
Christiano, L. J., M. Eichenbaum and C. Evans, (1999), "Monetary policy shocks: what
have we learned and to what end?”, in J. B. Taylor and M. Woodford (eds.), Handbook
of Macroeconomics, Volume 1A, 65-148.
Clarida R., J. Galì and M. Gertler (2000), “Monetary policy rules and macroeconomic stability: evidence and some theory”, Quarterly Journal of Economics 115 (1), 147-180.
Cogley, T., and T. J. Sargent (2005a): “Drifts and Volatilities: Monetary Policies and
Outcomes in Post WWII U.S.”, Review of Economics Dynamics 8 (2), 262-302.
Cogley, T., and T. J. Sargent (2005b): “The Conquest of US In‡ation: Learning and Robustness to Model Uncertainty,” Review of Economics Dynamics 8 (2), 528-563.
DeLong B. (1997), “America’s Only Peacetime In‡ation: the 1970’s”, in C. Romer and D.
Romer (Editors), Reducing in‡ation, NBER Studies in Business Cycles 30. University of
Chicago Press.
DeLong B. (2003), “Comment on James Stock and Mark Watson, ’Has the Business Cycle
Changed?’", in Monetary Policy and Uncertainty, Federal Reserve Bank of Kansas City,
9-56.
Dynan K., D. Elmendorf and D. Sichel (2006). “Can Financial Innovation Explain the Reduced Volatility of Economic Activity?”, Journal of Monetary Economics 53 (1), 123-150.
Fisher, J. and M. Gervais (2006), “First-Time Home Buyers”, 2006 Meeting Papers 432,
Society for Economic Dynamics.
Galì J; and L; Gambetti (2007), “On the Sources of the Great Moderation”, Pompeu Fabra
Universtity, mimeo.
Giannone D., L. Reichlin and L. Sala (2002), “Tracking Greenspan: Systematic and Unsystematic Monetary Policy Revisited”, IGiER mimeo.
Gordon R. (2005), “What Has Caused the Decline in US business Cycle Volatility?”, NBER
working paper 11777.

Monetary Policy, Output Composition and the Great Moderation

24

Hamilton J. (1989), “A New Approach to the Economic Analysis of Nonstationary Time
Series and the Business Cycle”, Econometrica 57(2), 357-384.
Hamilton J. (1994), “Time Series Analysis”, Princeton University Press.
Herrera A.-M. and E. Pesavento (2004), “The Decline in U.S. Output Volatility: Structural
Changes and Inventory Investment”, mimeo, Michigan University.
Iacoviello M; and S. Neri (2007), “The Role of Housing Collateral in an Estimated Two-Sector
Model of the US Economy”, Boston College mimeo.
Irvine F. and S. Schuh (2005), “The Roles of Co-Movement and Inventory Investment in
the Reduction of Output Volatility”, Federal Reserve Bank of Boston Working Paper No
05-9.
Justiniano A. and G. Primiceri (2006), “The Time Varying Volatility of Macroeconomic
Fluctuations”, mimeo, Northwestern University.
Kahn, J., M. McConnell, and G. Perez-Quiros (2002), “On the Causes of the Increased
Stability of the U.S. Economy”, Federal Reserve Bank of New York, Economic Policy
Review 8 (1): 183-202.
Kim, C. and C. Nelson (1999): “State Space Models with Regime Switching”, MIT Press,
London and Cambridge, Mass.
Leeper E; and T. Zha (2003), “Modest Policy Interventions”, Journal of Monetary Economics
50 (8), 1673-1700.
Monacelli T. (2006), “New Keynesian Models , Durable Goods and Borrowing Constraints”,
CEPR Discussion Paper 5916.
McConnell, M., and G. Perez-Quiros (2000), “Output Fluctuations in the United States:
What Has Changed Since the Early 1980s?" American Economic Review 90, 1464-1476.
McConnell M., R. Peach and A. Al-Haschim (2003), “After the Re…nancing Boom: Will
Consumers Scale Back Their Spending?”, Current Issues in Economics and Finance 9,
No. 12.
Leeper E. and T. Zha (2001), “Assessing Simple Policy Rules: A View from a Complete
Macroeconomic Model”, Federal Reserve Bank of Saint-Louis Review, 83-110.
Lubik T. and F. Schorfheide (2004), “Testing for Indeterminacy: An application to U.S.
Monetary Policy”, American Economic Review 94 (1), 190:217.

Monetary Policy, Output Composition and the Great Moderation

25

Primiceri G. (2005): “Time Varying Structural Vector Autoregressions and Monetary Policy”, Review of Economic Studies 72, pp821-852.
Primiceri G. (forthcoming): “Why In‡ation Rose and Fell: Policymakers’ Beliefs and US
Postwar Stabilization Policy,” Quarterly Journal of Economics.
Ramey, V. and D. Vine (2004) “Tracking the Source of the Decline in GDP Volatility: An
Analysis of the Automobile Industry”,” NBER working paper 10384.
Ramey V. and D. Vine (2005), “Declining Volatility in the US Automobile Industry”, NBER
Working Paper 11596.
Romer C. and D. Romer, “The Evolution of Econmic Understanding and Postwar Stabilization Policy”, The Federal Reserve Bank of Kansas City - Economic Review.
Sellon G., (2002), “Changing U.S. …nancial system: Some Implications for the Monetary
Transmission”, The Federal Reserve Bank of Kansas City - Economic Review, First
Quarter 2002.
Sims C., D. Waggoner and T. Zha (2006), “Methods for Inference in Large Multiple-Equation
Markov-Switching Models”, Federal Reserve Bank of Atlanta, mimeo.
Sims C. and T. Zha (1998), “Bayesian Methods for Dynamic Multivariate Models”, International Economic Review 39(4), 949-968.
Sims C. and T. Zha (2006), “Were There Regime Switches in U.S. Monetary Policy?”,
Forthcoming, American Economic Review.
Smets, F. and R. Wouters, (2007), “Shocks and Frictions in US Business Cycles: A Bayesian
DSGE Approach,” ECB Working Paper 722, forthcoming in the American Economic
Review.
Stock J., and M. Watson (2002), “Has the Business Cycle Changed and Why?" NBER
Macroeconomics Annual 2002 : 159-218.
Stock J., and M. Watson (2003), “Has the Business Cycle Changed? Evidence and Explanations,” in Monetary Policy and Uncertainty, Federal Reserve Bank of Kansas City:
9-56.
Stock J., and M. Watson (2005), “Understanding Changes in International Business Cycle
Dynamics”, Journal of the European Economic Association 3 (5), 968-1006.

Table 1: Decomposition of the variance of the US Private Sector Domestic Demand into the variances and co-variances of its main
components
a
1947-2006

b
1947-1965

c
1965-1984

d
1985-2006

c-b

d-c

Quarterly growth rates and contributions
GDP growth rate variance
Domestic private demand contribution variance
Net exports+Government consumption variance
2 Covariance of the above 2 sub-aggregates of GDP

17
20
4
-7

28
33
8
-14

21
25
3
-6

4
5
1
-2

-6
-9
-5
8

-17
-20
-2
5

Decomposition of DPD contribution variance
Trace of the covariance matrix of its 6 components
Sum of off diagonal terms among the 6 components

13
7

23
10

13
11

5
0.0

-10
1

-8

Band pass filtered (6,32,8) quarterly growth rates and contributions
GDP growth rate variance
8
Domestic private demand growth rate variance
12
Trace of the covariance matrix of its 6 components
4
Sum of off diagonal terms among the 6 components
8

18
27
8
19

8
10
3
7

2
2
1
1

-10
-17
-5
-12

-11

-7
-8
-2

-6

Notes: Domestic private demand (DPD) contribution to GDP growth is the sum of 6 sub-aggregates: residential investment, durable consumption, consumption of nondurable goods, consumption of services, fixed non-residential investment and inventory investment. The table decomposes the covariance matrix of these 6 subaggregates. All data come from "Table 1.1.2. Contributions to Percent Change in Real Gross Domestic Product". The Band Pass Filter methodology is presented in
Baxter and King (1999).

Table 2: Decomposition of the variance of the US Private Sector Domestic Demand into the variances and co-variances of
Households Investments and the aggregate of other components
a
b
c
d
c-b
d-c
1947-2006 1947-1965 1965-1984 1985-2006
Quarterly growth rates and contributions
Domestic private demand growth rate variance
20
33
25
5
-9
-20
Trace of the covariance matrix of its 2 components
3
27
17.6
5.5
-10
-12
Sum of off diagonal terms among the 2 components
17
6
7.1
-0.5
1
-8
correlation of the 2 components

0.27

Band pass filtered (6,32,8) quarterly growth rates and contributions
Domestic private demand growth rate variance
12
Trace of the covariance matrix of its 2 components
8
Sum of off diagonal terms among the 2 components
4
correlation of the 2 components

0.65

0.26

0.46

-0.14

27
17
11

10
6.6
3.7

2
1.8
0.2

0.71

0.63

0.20

-17
-10
-7

-8.3
-4.8

-3.5

Notes: Domestic private demand (DPD) contribution to GDP growth is the sum of 2 sub-aggregates: household investments defined as the sum of residential
investment and durable consumption and the aggregate of non-durable goods consumption, services consumption, fixed non-residential investment and
inventory investment. The table decomposes the covariance matrix of these 2 sub-aggregates. All data come from "Table 1.1.2. Contributions to Percent
Change in Real Gross Domestic Product".

Table 3: Marginal likelihood of 4 variables model depending on its form of time variation
Marginal likelihood
Marginal likelihood
Constant parameters VAR
-1139.9
Specifications with changes only in the variances
Specifications with changes only in the coefficients
ffch
fh
dpgdp
ffr
ffch
fh
dpgdp
ffr
constant constant constant
2S
-1085.5
Case 3 constant constant constant 2 S
-1069.6
2S
constant constant constant
-1131.1
2S
constant constant constant
-1132.4
constant 2 S
constant constant
-1136.4
constant 2 S
constant constant
-1135.2
constant constant 2 S
constant
-1126.5
constant constant 2 S
constant
-1131.7
constant constant constant
2S
-1085.5
constant constant constant
3S
-1070.7
constant constant constant
3S
-1053.4
constant constant
2S
2S
-1353.4
constant constant
2S
2S
-1072.1
constant constant
2S
3S
-1609.8
constant constant
2S
3S
-1040.1
constant constant
3S
3S
-1085.5
constant constant
3S
2S
-1068.7
constant
2S
2S
2S
-1068.6
constant
2S
2S
3S
-1036.5
Case 1
2S
2S
2S
2S
-1059.9
2S
2S
2S
3S
-1027.8
3S
2S
2S
-1054.5
3S
Case 2
2S
3S
-1022.5
3S
3S
3S
-1019.2
Specification with changes both in the variance and in the coefficients
Variances
Coefficients
2S
2S
2S
2S
constant constant constant
2S
-1032.2
2S
2S
2S
3S
constant constant constant
2S
-1029.1
3S
Case 4
2S
2S
constant constant constant
2S
-1027.7
3S
2S
3S
constant constant constant
2S
-1024.6
Note: The variance of the shocks and the coefficients can be either constant; take one of 2 values (2S) or one of three values (3S).
indicates models for which the results are discussed in the text.

Table 4: Coefficients of the interest rate equation accross regimes
VAR model (Case 3)
ff+ch
cf
dpgdp
ffr
Regime 1: 1969 Q3 to 1975 Q4 and 1979 Q4 to1985 Q4
0.04
0.18
0.02
0.98
Regime 2: 1960's, 1976:Q1 to 1979:Q3 and post 1986Q1
0.05
0.00
0.05
0.95
VAR model (Case 4)
ff+ch
cf
dpgdp
Regime 1: 1969 Q3 to 1975 Q4 and 1979 Q4 to1985 Q4
0.05
0.20
0.03
Regime 2: 1960's, 1976:Q1 to 1979:Q3 and post 1986Q1
0.06
-0.01
0.04

ffr
0.97
0.95

Note: the table reports the sum of A0(j,4) and sum/lags of A(j,4)lags. Confidence
intervals based on draws from the posterior distribution of the models parameters are
avalaible upon request.

Table 5: Forecast error variance decomposition across states of different volatility of Monetary Policy Shocks
VAR model with constant parameters
contrib of MP shock to
ff+ch
cf
4 quarters
14
21
8 quarters
16
21
12 quarters
16
21
16 quarters
16
21

dpgdp
3
2
2
2

ffr
64
47
38
33

contrib to FFR of shocks to the equation of
ff+ch
cf
dpgdp
ffr
19
4
14
64
25
8
20
47
28
11
23
38
28
12
27
33

VAR with time variation specification sv12 (ML=-1059.9)
Regime 1: High MP shock volatility, 1969Q2 to 1975 Q4 and 1979Q4 to 1984Q4
contrib of MP shock to
contrib to FFR of shocks to the equation of
ff+ch
cf
dpgdp
ffr
ff+ch
cf
dpgdp
ffr
4 quarters
30
48
9
90
5
1
4
90
8 quarters
43
53
8
80
8
5
7
80
12 quarters
44
53
7
73
11
8
9
73
16 quarters
44
53
10
68
12
9
12
68
Regime 2: Low MP shock volatility of the 1960's, the mid-1970's and the "Greenspan era"
contrib of MP shock to
contrib to FFR of shocks to the equation of
ff+ch
cf
dpgdp
ffr
ff+ch
cf
dpgdp
ffr
4 quarters
3
6
1
40
31
7
22
40
8 quarters
5
7
1
22
32
19
27
22
12 quarters
5
7
1
16
33
24
27
16
16 quarters
5
7
1
13
32
25
30
13
Note: report the percent of the variance of the variables explained by the monetary policy shocks (left panel) and the contribution of the shocks of each
equation to the variance of the federal funds rate (right panel), all numbers are percentages. Confidence intervals based on draws from the posterior
distribution of the models parameters are avalaible upon request.

Corporate investment + non-durable consumption (ff+ch)

Household investment (durables+ residential, fh)

10

10

5

5

0

0

-5

-5

-10
1960

1970

1980

1990

2000

-10
1960

Inflation of GDP deflator (dp)
20

15

15

10

10

5

5

1970

1980

1990

1980

1990

2000

Federal funds interest rate (ffr)

20

0
1960

1970

2000

0
1960

1970

1980

1990

2000

Figure 1: Time series of contributions to GDP growth, GDP deflator inflation and the federal funds rate
Grey areas indicates the periods designated by the NBER as recessions

epsilon(ff+ch)

epsilon(fh)

10

4

8

3
2

6

1

4

0
2
-1
0

-2

-2

-3

-4
-6
1960

-4
1970

1980

1990

2000

-5
1960

1970

epsilon(dp)

1980

1990

2000

Monetary Policy Shocks

1.5

6
5

1

4
3

0.5

2
1

0

0
-1

-0.5

-2
-1
1960

1970

1980

1990

2000

-3
1960

1970

1980

1990

2000

Figure 2: Identified shocks as estimated in the constant parameters VAR model

Probability of being in a regime of high volatility for the orthogonalized residual of the ff+ch equation
1
0.5
0
1960

1965

1970

1975

1980

1985

1990

1995

2000

2005

1995

2000

2005

1995

2000

2005

2000

2005

same with respect to the fh equation
1
0.5
0
1960

1965

1970

1975

1980

1985

1990

same with respect to the dp equation
1
0.5
0
1960

1965

1970

1975

1980

1985

1990

Probability of being in a regime of high volatility for the monetary policy shocks
1
0.5
0
1960

1965

1970

1975

1980

1985

1990

1995

Figure 3: Case 1 model, Probabilities of being in high volatility states for the identified shocks
Grey areas indicates periods designated by the NBER as recessions

Probability of being in a regime of high volatility for the orthoganalized residual of the ff+ch and fh equation
1

0.5

0
1960

1965

1970

1975

1980

1985

1990

1995

2000

2005

1995

2000

2005

2000

2005

same with respect to the dp equation
1

0.5

0
1960

1965

1970

1975

1980

1985

1990

Probability of being in a regime of high volatility for the monetary policy shocks
1

0.5

0
1960

1965

1970

1975

1980

1985

1990

1995

Figure 4: Case 2 model, Probabilities of being in high volatility states for the identified shocks
(green-dotted for high blue full line for either high or intermediate volatility regimes)
Grey areas indicates periods designated by the NBER as recessions

Probability of being in a regime of high volatility for the orthoganalized residual of the ff+ch and fh equation
1
0.5
0
1960

1965

1970

1975

1980

1985

1990

1995

2000

2005

1995

2000

2005

1995

2000

2005

1995

2000

2005

same with respect to the dp equation
1
0.5
0
1960

1965

1970

1975

1980

1985

1990

Probability of being in a regime of high volatility for the monetary policy shocks
1
0.5
0
1960

1965

1970

1975

1980

1985

1990

same with respect to the coefficient change equation
1
0.5
0
1960

1965

1970

1975

1980

1985

1990

Figure 5: Case 4 model, Probabilities of being in high volatility states and regime 2 coefficients of the monetary policy rule
(green-dotted for high blue full line for either high or intermediate volatility regimes)
Grey areas indicates the periods designated by the NBER as recessions

Markov Switching VAR estimates of the standard deviation of Domestic Private Demand (full line)
and its value when the volatility of monetary shocks stays high (dotted, green line)
6

5

4

3

2

1960

1965

1970

1975

1980

1985

1990

1995

2000

2005

idem for the correlation of household investment (fh) and the rest of DPD (ff+ch)

0.5
0.4
0.3
0.2
0.1
0

1960

1965

1970

1975

1980

1985

1990

1995

2000

2005

Figure 6: Time variation in the standard deviation of Domestic Private Demand (upper panel) and the correlation of its two
components (lower panel) and holding the variance of monetary policy shock at its highest level (green-dotted)
Grey areas indicates the periods designated by the NBER as recessions

Monetary Policy Shocks and high volatility MP shock regime
6

Contribution of MP Shocks to the FFR (Blue), demeaned FFR (Green)
10

5
4
3

5

2
1
0

0

-1
-2
-3
1960

1970

1980

1990

2000

Contribution of MP Shocks to FFCH (Blue), and demeaned FFCH (Green)
6

-5
1960

4

2

2

0

0

-2

-2

-4

-4

1970

1980

1990

2000

1980

1990

2000

and to FH (Blue), demeaned FH (Green) and NBER recessions
6

4

-6
1960

1970

-6
1960

1970

1980

1990

2000

Figure 7: Monetary policy shocks, and their historical contribution to the federal funds rate and GDP components
Top panel: dark areas designate the period when the model estimates a higher volatility of monetary policy shocks.
Bottom panel: grey areas indicates the periods designated by the NBER as recessions

Original and counterfactual scaled-down MP Shocks
4

Contribution of original and scaled-down MP shocks (thick red line) to the FFR
10

3
5

2
1

0
0
-1

-5

-2
-3
1965

1970

1975

1980

1985

1990

-10
1965

to DPD, demeanded DPD (Green)
3

4

2

2

1

0

0

-2

-1

-4

-2

1970

1975

1980

1985

1975

1980

1985

1990

Additional Output Fluctuations due to Monetary Policy Volatility

6

-6
1965

1970

1990

-3
1965

1970

1975

1980

1985

1990

Figure 8: Effects of scaling down monetary policy shocks on the federal funds rate and
the historical contribution of monetary policy shocks to Domestic Private Demand
Top panel: dark areas designate the period when the model estimates a higher volatility of monetary policy shocks.
Bottom panel: grey areas indicates the periods designated by the NBER as recessions

Appendix: Additional Tables
Table A1: Forecast error variance decomposition across states of different volatility of Monetary Policy Shocks
Best fit VAR model, 3 regimes for the shocks f equation 1 and 2, 2 regimes for equation 3, 3 regimes for the MP Shocks (ML=-1022.9)
Regime 1: High MP shock volatility, 1973 Q3 to 1976 Q1, from 1979 Q4 to 1982 Q3 and 1984 Q4
contrib of MP shock to
contrib to FFR of shocks to the equation of
ff+ch
cf
dpgdp
ffr
ff+ch
cf
dpgdp
ffr
4 quarters
46
68
25
97
2
1
1
97
8 quarters
62
72
21
92
3
3
2
92
12 quarters
63
72
20
89
5
4
2
89
16 quarters
63
73
26
86
5
5
3
86
Regime 2: Moderately high MP shock volatility, 1969 Q1 to 1971 Q4
contrib of MP shock to
ff+ch
cf
dpgdp
4 quarters
9
19
3
8 quarters
17
23
3
12 quarters
18
23
2
16 quarters
18
23
4

contrib to FFR of shocks to the equation of
ff+ch
cf
dpgdp
14
5
7
18
15
11
21
20
12
22
22
14

ffr
74
56
47
41

Regime 3: Low MP shock volatility of the 1960's, the mid-1970's and the "Greenspan era"
contrib of MP shock to
contrib to FFR of shocks to the equation of
ff+ch
cf
dpgdp
ffr
ff+ch
cf
dpgdp
4 quarters
2
5
1
39
33
11
18
8 quarters
4
6
1
22
31
27
20
12 quarters
4
6
0
16
32
32
20
16 quarters
4
6
1
13
32
33
22

ffr
39
22
16
13

ffr
74
56
47
41

Note: report the percent of the variance of the variables explained by the monetary policy shocks (left panel) and the contribution of the shocks of each equation to
the variance of the federal funds rate (right panel), all numbers are percentages.Confidence intervals based on draws from the posterior distribution of the models
parameters are avalaible upon request.

Table A2: Forecast error variance decomposition across states of different volatility of Monetary Policy Shocks
Regime 1: High Variance of Monetary Policy shocks and Coeffients combination 1 : Period 1979Q4 to 1982Q4
contrib of MP shock to
contrib to FFR of shocks to the equation of
ff+ch
cf
dpgdp
ffr
ff+ch
cf
dpgdp
ffr
4 quarters
41
56
18
85
5
7
2
85
8 quarters
49
59
13
78
9
11
2
78
12 quarters
50
59
12
74
11
13
3
74
16 quarters
49
59
14
70
12
15
3
70
Regime 2: Low Variance of Monetary Policy shocks and Coeffients combination 1 : 1969 Q2 to 1975 Q4 and 1982 Q4 to1985 Q4
contrib of MP shock to
contrib to FFR of shocks to the equation of
ff+ch
cf
dpgdp
ffr
ff+ch
cf
dpgdp
ffr
4 quarters
18
28
6
63
13
18
5
63
8 quarters
24
30
4
52
19
24
5
52
12 quarters
24
30
4
46
21
27
6
46
16 quarters
24
30
5
42
22
28
7
42
Regime 3: High Variance of Monetary Policy shocks and Coeffients combination 2 : Period 1960 Q1 to 1960 Q4 and 1978 Q3 to 1979 Q3
contrib of MP shock to
contrib to FFR of shocks to the equation of
ff+ch
cf
dpgdp
ffr
ff+ch
cf
dpgdp
ffr
4 quarters
6
13
2
62
25
2
11
62
8 quarters
13
18
2
48
27
10
15
48
12 quarters
14
18
2
40
28
15
17
40
16 quarters
14
19
3
36
28
18
18
36
Regime 4: Low Variance of Monetary Policy shocks and Coeffients combination 1 : Period 1960's, 1976-1979 and post 1986
contrib of MP shock to
contrib to FFR of shocks to the equation of
ff+ch
cf
dpgdp
ffr
ff+ch
cf
dpgdp
ffr
4 quarters
2
4
1
33
42
4
20
33
8 quarters
4
6
1
22
40
15
23
22
12 quarters
5
6
1
17
38
21
24
17
16 quarters
5
6
1
14
37
24
25
14
Note: report the percent of the variance of the variables explained by the monetary policy shocks (left panel) and the contribution of the shocks of each equation to the
variance of the federal funds rate (right panel), all numbers are percentages. Confidence intervals based on draws from the posterior distribution of the models
parameters are avalaible upon request.

Working Paper Series
A series of research studies on regional economic issues relating to the Seventh Federal
Reserve District, and on financial and economic topics.
Standing Facilities and Interbank Borrowing: Evidence from the Federal Reserve’s
New Discount Window
Craig Furfine

WP-04-01

Netting, Financial Contracts, and Banks: The Economic Implications
William J. Bergman, Robert R. Bliss, Christian A. Johnson and George G. Kaufman

WP-04-02

Real Effects of Bank Competition
Nicola Cetorelli

WP-04-03

Finance as a Barrier To Entry: Bank Competition and Industry Structure in
Local U.S. Markets?
Nicola Cetorelli and Philip E. Strahan

WP-04-04

The Dynamics of Work and Debt
Jeffrey R. Campbell and Zvi Hercowitz

WP-04-05

Fiscal Policy in the Aftermath of 9/11
Jonas Fisher and Martin Eichenbaum

WP-04-06

Merger Momentum and Investor Sentiment: The Stock Market Reaction
To Merger Announcements
Richard J. Rosen

WP-04-07

Earnings Inequality and the Business Cycle
Gadi Barlevy and Daniel Tsiddon

WP-04-08

Platform Competition in Two-Sided Markets: The Case of Payment Networks
Sujit Chakravorti and Roberto Roson

WP-04-09

Nominal Debt as a Burden on Monetary Policy
Javier Díaz-Giménez, Giorgia Giovannetti, Ramon Marimon, and Pedro Teles

WP-04-10

On the Timing of Innovation in Stochastic Schumpeterian Growth Models
Gadi Barlevy

WP-04-11

Policy Externalities: How US Antidumping Affects Japanese Exports to the EU
Chad P. Bown and Meredith A. Crowley

WP-04-12

Sibling Similarities, Differences and Economic Inequality
Bhashkar Mazumder

WP-04-13

Determinants of Business Cycle Comovement: A Robust Analysis
Marianne Baxter and Michael A. Kouparitsas

WP-04-14

The Occupational Assimilation of Hispanics in the U.S.: Evidence from Panel Data
Maude Toussaint-Comeau

WP-04-15

1

Working Paper Series (continued)
Reading, Writing, and Raisinets1: Are School Finances Contributing to Children’s Obesity?
Patricia M. Anderson and Kristin F. Butcher

WP-04-16

Learning by Observing: Information Spillovers in the Execution and Valuation
of Commercial Bank M&As
Gayle DeLong and Robert DeYoung

WP-04-17

Prospects for Immigrant-Native Wealth Assimilation:
Evidence from Financial Market Participation
Una Okonkwo Osili and Anna Paulson

WP-04-18

Individuals and Institutions: Evidence from International Migrants in the U.S.
Una Okonkwo Osili and Anna Paulson

WP-04-19

Are Technology Improvements Contractionary?
Susanto Basu, John Fernald and Miles Kimball

WP-04-20

The Minimum Wage, Restaurant Prices and Labor Market Structure
Daniel Aaronson, Eric French and James MacDonald

WP-04-21

Betcha can’t acquire just one: merger programs and compensation
Richard J. Rosen

WP-04-22

Not Working: Demographic Changes, Policy Changes,
and the Distribution of Weeks (Not) Worked
Lisa Barrow and Kristin F. Butcher

WP-04-23

The Role of Collateralized Household Debt in Macroeconomic Stabilization
Jeffrey R. Campbell and Zvi Hercowitz

WP-04-24

Advertising and Pricing at Multiple-Output Firms: Evidence from U.S. Thrift Institutions
Robert DeYoung and Evren Örs

WP-04-25

Monetary Policy with State Contingent Interest Rates
Bernardino Adão, Isabel Correia and Pedro Teles

WP-04-26

Comparing location decisions of domestic and foreign auto supplier plants
Thomas Klier, Paul Ma and Daniel P. McMillen

WP-04-27

China’s export growth and US trade policy
Chad P. Bown and Meredith A. Crowley

WP-04-28

Where do manufacturing firms locate their Headquarters?
J. Vernon Henderson and Yukako Ono

WP-04-29

Monetary Policy with Single Instrument Feedback Rules
Bernardino Adão, Isabel Correia and Pedro Teles

WP-04-30

2

Working Paper Series (continued)
Firm-Specific Capital, Nominal Rigidities and the Business Cycle
David Altig, Lawrence J. Christiano, Martin Eichenbaum and Jesper Linde

WP-05-01

Do Returns to Schooling Differ by Race and Ethnicity?
Lisa Barrow and Cecilia Elena Rouse

WP-05-02

Derivatives and Systemic Risk: Netting, Collateral, and Closeout
Robert R. Bliss and George G. Kaufman

WP-05-03

Risk Overhang and Loan Portfolio Decisions
Robert DeYoung, Anne Gron and Andrew Winton

WP-05-04

Characterizations in a random record model with a non-identically distributed initial record
Gadi Barlevy and H. N. Nagaraja

WP-05-05

Price discovery in a market under stress: the U.S. Treasury market in fall 1998
Craig H. Furfine and Eli M. Remolona

WP-05-06

Politics and Efficiency of Separating Capital and Ordinary Government Budgets
Marco Bassetto with Thomas J. Sargent

WP-05-07

Rigid Prices: Evidence from U.S. Scanner Data
Jeffrey R. Campbell and Benjamin Eden

WP-05-08

Entrepreneurship, Frictions, and Wealth
Marco Cagetti and Mariacristina De Nardi

WP-05-09

Wealth inequality: data and models
Marco Cagetti and Mariacristina De Nardi

WP-05-10

What Determines Bilateral Trade Flows?
Marianne Baxter and Michael A. Kouparitsas

WP-05-11

Intergenerational Economic Mobility in the U.S., 1940 to 2000
Daniel Aaronson and Bhashkar Mazumder

WP-05-12

Differential Mortality, Uncertain Medical Expenses, and the Saving of Elderly Singles
Mariacristina De Nardi, Eric French, and John Bailey Jones

WP-05-13

Fixed Term Employment Contracts in an Equilibrium Search Model
Fernando Alvarez and Marcelo Veracierto

WP-05-14

Causality, Causality, Causality: The View of Education Inputs and Outputs from Economics
Lisa Barrow and Cecilia Elena Rouse

WP-05-15

3

Working Paper Series (continued)
Competition in Large Markets
Jeffrey R. Campbell

WP-05-16

Why Do Firms Go Public? Evidence from the Banking Industry
Richard J. Rosen, Scott B. Smart and Chad J. Zutter

WP-05-17

Clustering of Auto Supplier Plants in the U.S.: GMM Spatial Logit for Large Samples
Thomas Klier and Daniel P. McMillen

WP-05-18

Why are Immigrants’ Incarceration Rates So Low?
Evidence on Selective Immigration, Deterrence, and Deportation
Kristin F. Butcher and Anne Morrison Piehl

WP-05-19

Constructing the Chicago Fed Income Based Economic Index – Consumer Price Index:
Inflation Experiences by Demographic Group: 1983-2005
Leslie McGranahan and Anna Paulson

WP-05-20

Universal Access, Cost Recovery, and Payment Services
Sujit Chakravorti, Jeffery W. Gunther, and Robert R. Moore

WP-05-21

Supplier Switching and Outsourcing
Yukako Ono and Victor Stango

WP-05-22

Do Enclaves Matter in Immigrants’ Self-Employment Decision?
Maude Toussaint-Comeau

WP-05-23

The Changing Pattern of Wage Growth for Low Skilled Workers
Eric French, Bhashkar Mazumder and Christopher Taber

WP-05-24

U.S. Corporate and Bank Insolvency Regimes: An Economic Comparison and Evaluation
Robert R. Bliss and George G. Kaufman

WP-06-01

Redistribution, Taxes, and the Median Voter
Marco Bassetto and Jess Benhabib

WP-06-02

Identification of Search Models with Initial Condition Problems
Gadi Barlevy and H. N. Nagaraja

WP-06-03

Tax Riots
Marco Bassetto and Christopher Phelan

WP-06-04

The Tradeoff between Mortgage Prepayments and Tax-Deferred Retirement Savings
Gene Amromin, Jennifer Huang,and Clemens Sialm

WP-06-05

Why are safeguards needed in a trade agreement?
Meredith A. Crowley

WP-06-06

4

Working Paper Series (continued)
Taxation, Entrepreneurship, and Wealth
Marco Cagetti and Mariacristina De Nardi

WP-06-07

A New Social Compact: How University Engagement Can Fuel Innovation
Laura Melle, Larry Isaak, and Richard Mattoon

WP-06-08

Mergers and Risk
Craig H. Furfine and Richard J. Rosen

WP-06-09

Two Flaws in Business Cycle Accounting
Lawrence J. Christiano and Joshua M. Davis

WP-06-10

Do Consumers Choose the Right Credit Contracts?
Sumit Agarwal, Souphala Chomsisengphet, Chunlin Liu, and Nicholas S. Souleles

WP-06-11

Chronicles of a Deflation Unforetold
François R. Velde

WP-06-12

Female Offenders Use of Social Welfare Programs Before and After Jail and Prison:
Does Prison Cause Welfare Dependency?
Kristin F. Butcher and Robert J. LaLonde
Eat or Be Eaten: A Theory of Mergers and Firm Size
Gary Gorton, Matthias Kahl, and Richard Rosen
Do Bonds Span Volatility Risk in the U.S. Treasury Market?
A Specification Test for Affine Term Structure Models
Torben G. Andersen and Luca Benzoni

WP-06-13

WP-06-14

WP-06-15

Transforming Payment Choices by Doubling Fees on the Illinois Tollway
Gene Amromin, Carrie Jankowski, and Richard D. Porter

WP-06-16

How Did the 2003 Dividend Tax Cut Affect Stock Prices?
Gene Amromin, Paul Harrison, and Steven Sharpe

WP-06-17

Will Writing and Bequest Motives: Early 20th Century Irish Evidence
Leslie McGranahan

WP-06-18

How Professional Forecasters View Shocks to GDP
Spencer D. Krane

WP-06-19

Evolving Agglomeration in the U.S. auto supplier industry
Thomas Klier and Daniel P. McMillen

WP-06-20

Mortality, Mass-Layoffs, and Career Outcomes: An Analysis using Administrative Data
Daniel Sullivan and Till von Wachter

WP-06-21

5

Working Paper Series (continued)
The Agreement on Subsidies and Countervailing Measures:
Tying One’s Hand through the WTO.
Meredith A. Crowley

WP-06-22

How Did Schooling Laws Improve Long-Term Health and Lower Mortality?

WP-06-23

Bhashkar Mazumder
Manufacturing Plants’ Use of Temporary Workers: An Analysis Using Census Micro Data
Yukako Ono and Daniel Sullivan

WP-06-24

What Can We Learn about Financial Access from U.S. Immigrants?
Una Okonkwo Osili and Anna Paulson

WP-06-25

Bank Imputed Interest Rates: Unbiased Estimates of Offered Rates?
Evren Ors and Tara Rice

WP-06-26

Welfare Implications of the Transition to High Household Debt
Jeffrey R. Campbell and Zvi Hercowitz

WP-06-27

Last-In First-Out Oligopoly Dynamics
Jaap H. Abbring and Jeffrey R. Campbell

WP-06-28

Oligopoly Dynamics with Barriers to Entry
Jaap H. Abbring and Jeffrey R. Campbell

WP-06-29

Risk Taking and the Quality of Informal Insurance: Gambling and Remittances in Thailand
Douglas L. Miller and Anna L. Paulson

WP-07-01

Fast Micro and Slow Macro: Can Aggregation Explain the Persistence of Inflation?
Filippo Altissimo, Benoît Mojon, and Paolo Zaffaroni

WP-07-02

Assessing a Decade of Interstate Bank Branching
Christian Johnson and Tara Rice

WP-07-03

Debit Card and Cash Usage: A Cross-Country Analysis
Gene Amromin and Sujit Chakravorti

WP-07-04

The Age of Reason: Financial Decisions Over the Lifecycle
Sumit Agarwal, John C. Driscoll, Xavier Gabaix, and David Laibson

WP-07-05

Information Acquisition in Financial Markets: a Correction
Gadi Barlevy and Pietro Veronesi

WP-07-06

Monetary Policy, Output Composition and the Great Moderation
Benoît Mojon

WP-07-07

6