Professional Documents
Culture Documents
SW US Model
SW US Model
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
is collaborating with JSTOR to digitize, preserve and extend access to The American Economic
Review
A new generation of small-scale monetary shocks and frictions. It features sticky nominal
business cycle models with sticky prices and price and wage settings that allow for backward
wages (the New Keynesian or New Neoclassi- inflation indexation, habit formation in con-
cal Synthesis (NNS) models) has become pop- sumption and investment adjustment costs that
ular in monetary policy analysis.' Following create hump-shaped responses of aggregate de-
Smets and Wouters (2003), this paper estimates mand, and variable capital utilization and fixed
an extended version of these models, largely costs in production. The stochastic dynamics is
based on Lawrence J. Christiano, Martin driven by seven orthogonal structural shocks. In
Eichenbaum, and Charles L. Evans (CEE, addition to total factor productivity shocks, the
2005), on US data covering the period 1966:1-model includes two shocks that affect the
2004:4, and using a Bayesian estimation meth- intertemporal margin (risk premium shocks
odology. The estimated model contains manyand investment-specific technology shocks),
two shocks that affect the intratemporal margin
(wage and price mark-up shocks), and two pol-
* Smets: European Central Bank/CEPR/University of icy shocks (exogenous spending and monetary
Ghent, Kaiserstrasse 29, D-60311 Frankfurt am Main, Ger-
policy shocks). Compared to the model used in
many (e-mail: frank.smets@ecb.int); Wouters: National Bank
Smets and Wouters (2003), there are three main
of Belgium/Universit6 Catholique, Boulevard de Berliamont 5,
1000 Brussels, Belgium (e-mail: rafael.wouters@nbb.be). The
differences. First, the number of structural
views expressed are solely our own and do not necessarily shocks is reduced to the number of seven ob-
reflect those of the European Central Bank or the National servables used in estimation. For example, there
Bank of Belgium. We thank seminar participants and discus-
is no time-varying inflation target, nor a sepa-
sants at the 2003 ECB/IMOP Workshop on Dynamic Macro-
economics, the Federal Reserve Board, Princeton rate labor supply shock. Second, the model fea-
tures
University, the Federal Reserve Banks of St. Louis and a deterministic growth rate driven by
Chicago, the 2004 ASSA meetings in San Diego, Koln labor-augmenting technological progress, so
University, Humboldt University, the European Centralthat the data do not need to be detrended before
Bank, the Bank of Canada/Swiss National Bank/Federal
estimation. Third, the Dixit-Stiglitz aggregator
Reserve Bank of Cleveland Joint Workshop on Dynamic
in the intermediate goods and labor market is
Models Useful for Policy, and, in particular, Frank
Schorfheide, Fabio Canova, Chris Sims, Mark Gertler, replaced by the more general aggregator de-
and two anonymous referees for very useful and stimu- veloped in Miles S. Kimball (1995). This
lating comments.
aggregator implies that the demand elasticity
SSee Marvin Goodfriend and Robert G. King (1997),
of differentiated goods and labor depends on
Julio J. Rotemberg and Michael Woodford (1995), Richard
Clarida, Jordi Gall, and Mark Gertler (1999) and Woodford their relative price. As shown in Eichenbaum
(2003). and Jonas Fischer (forthcoming), the introduc-
586
the real
[(1 + A/y)o']. Current consumption (c,) value of the existing capital stock (q,).
depends
Modeling capital
on a weighted average of past and expected future adjustment costs as a function of
consumption, and on expected growth in hours
the change in investment rather than its level in-
troduces additional
worked (1, - El ), the ex ante real interest ratedynamics in the investment
equation,
(r, - Er+,, 1), and a disturbance term et. which
Under is useful in capturing the hump-
the assumption of no external habit formation
shaped response of investment to various shocks.
(A = 0) and log utility in consumption
Finally,(o- = 1), a disturbance to the invest-
et represents
ment-specific
c, = C2 = 0 and the traditional purely forward- technology process and is assumed
looking consumption equation is obtained. With autoregressive process with
to follow a first-order
an marginally
steady-state growth, the growth rate y IID-Normal error term: e = Pis-+1 + -.
affects the reduced-form parameters in The
thecorresponding
linear- arbitrage equation for the
ized consumption equation. When the value of capital is
elasticity of given by
intertemporal substitution (for constant labor) is
(4) and
smaller than one ( , > 1), consumption qt = hours
qlEtq, + + (1 - ql)Etrk,
worked are complements in utility and consump-
tion depends positively on current hours worked
- (rt - *E,t t+l Sb),
and negatively on expected growth in hours
worked (see Susanto Basu and Kimball 2002).
where q, = 3y-'c(1
Finally, the disturbance term eb represents a - 6) = [(1 - 5)/(Rk +
wedge between the interest rate controlled bycurrent
(1 - 8))]. The the value of the capital stock
central bank and the return on assets (q,)
helddepends
by positively
the on its expected future
households. A positive shock to thisvaluewedge in- real rental rate on capital
and the expected
creases the required return on assets (E,r+
and reduces
1) and negatively on the ex ante real in-
current consumption. At the same time, itand
terest rate also
the risk premium disturbance.
increases the cost of capital and reducesTurning
the valueto the supply side, the aggregate
of capital and investment, as shown production
below.3functionThisis given by
shock has similar effects as so-called net-worth
shocks in Ben S. Bernanke, Gertler, and Simon (5) y, = 0 (ak' + (1 - a)l, + ea).
Gilchrist (1999) and Christiano, Roberto Motto,
and Massimo Rostagno (2003), which explicitly
model the external finance premium. The distur-
bance is assumed to follow a first-order autore- Output is produced using capital (kt) and
labor services (hours worked, I,). Total factor
gressive process with an IID-Normal error term: productivity (ea) is assumed to follow a first-
b b + b
order autoregressive process: s" = Pa't-1
t = Pbt-1
The + .lt"
dynamics of investment comes from t
tr. The parameter a captures the share of
investment Euler equation and is given by
capital in production, and the parameter OP is one
plus the share of fixed costs in production, reflect-
ing the presence of fixed costs in production.
(3) i = ilit-I + (1 -- As
il)Etit+l + i2qt
newly installed capital + et,
becomes effective
only with a one-quarter lag, current capital ser-
where i1 = 1/(1 +
vices3y(l-"c)), i2
used in production (kt) are = [1/(1
a function of
(14) r,with
where a, is the elasticity of labor supply = pr + (1 - p){r,,, T + ry(y,- yl)}
respect to the real wage and A is the habit
parameter in consumption. + rAy[(yt - yt) - (Yt-1 -- Yt-1)] + S.
Similarly, due to nominal wage stickiness
and partial indexation of wages to inflation, real
The monetary
wages adjust only gradually to the desired wage authorities follow a generalized
mark-up: Taylor rule by gradually adjusting the policy-
controlled interest rate (r,) in response to in-
flation and the output gap, defined as the
difference between actual and potential output
(13) w, = wlw,_-
(John B. Taylor 1993). Consistently with the
+ (1 - Wl)(E,wt+ 1 + Etrt+1) DSGE model, potential output is defined as the
level of output that would prevail under flexible
prices and wages in the absence of the two
- W2Tt- + W337Tt1 - W4i.Lt + et,
"mark-up" shocks.5 The parameter p captures
the degree of interest rate smoothing. In addi-
tion, there is a short-run feedback from the
change in the output gap. Finally, we assume
with w1 = 1/(1 + 3,yl - ), w2 = (1 +
that the monetary policy shocks (er) follow a
3yl - w)/(1 + p -ly1), w3 = w/I(1 +
py'-c), and w4 = 1/(1 + 3y'-c)[(l - first-order autoregressive process with an IID-
3 '-cfw)(1 - w)/(w((w- 1)ew + 1))]. Normal error term: E=PR+N
The real wage w, is a function of expected Equations (1) to (14) determine 14 endoge-
and past real wages, expected, current, and past nous variables: y, ct, i,, q,, kt, k,, z r t, -tr,T
inflation, the wage mark-up, and a wage- pt, wt, I,, and r,. The stochastic behavior of the
markup disturbance (e"). If wages are perfectly system of linear rational expectations equations
flexible ( = 0), the real wage is a constant is driven by seven exogenous disturbances: total
mark-up over the marginal rate of substitution factor productivity (es), investment-specific
between consumption and leisure. In general, technology (el), risk premium (es), exogenous
the speed of adjustment to the desired wage spending (ef), price mark-up (s'), wage mark-up
mark-up depends on the degree of wage sticki- (e'), and monetary policy (sr) shocks. Next we
ness ( and the demand elasticity for labor, turn to the estimation of the model.
which itself is a function of the steady-state
labor market mark-up (,w - 1) and the curva-
ture of the Kimball labor market aggregator II. Parameter Estimates
shows that the estimated DSGE model can com- prisingly, the BVAR(4) model performs worse
pete with standard BVAR models in terms of than the simple VAR(1) model at longer hori-
empirical one-step-ahead prediction perfor- zons. Moreover, the improvement appears to be
mance. These results are confirmed by a more quite uniform across the seven macro variables.
traditional out-of-sample forecasting exercise
reported in Table 3. Table 3 reports out-of- IV. Model Sensitivity: Which Frictions Are
sample RMSEs for different forecast horizons Empirically Important0
over the period 1990:1 to 2004:4. For this ex-
ercise, the VAR(1), BVAR(4), and DSGE The introduction of a large number of fric-
model were initially estimated over the sample tions raises the question of which of those are
1966:1-1989:4. The models were then used to really necessary to capture the dynamics of the
forecast the seven data series contained in data.Y, In this section, we examine the contribu-
from 1990:1 to 2004:4, whereby the VAR(1) tion of each of the frictions to the marginal
and BVAR(4) models were reestimated every likelihood of the DSGE model.
quarter, and the DSGE model was reestimated Table 4 presents the estimates of the mode
every year. The measure of overall performance of the parameters and the marginal likelihood
reported in the last column of Table 3 is the logwhen each friction (price and wage stickiness,
determinant of the uncentered forecast error co- price and wage indexation, investment ad-
variance matrix. justment costs and habit formation, capital
The out-of-sample forecast statistics confirm utilization, and fixed costs in production) is
the good forecast performance of the DSGE drastically reduced one at a time. This table
model relative to the VAR and BVAR models. also gives an idea of the robustness of the
At the one-quarter-ahead horizon, the BVAR(4) parameters and the model performance with
and the DSGE model improve with about the respect to the various frictions included in the
same magnitude over the VAR(1) model, con-model. For comparison, the first column re-
firming the results from Table 2. Over longer produces the baseline estimates (mode of the
horizons up to three years, however, the DSGE posterior) and the marginal likelihood based
model does considerably better than both theon the Laplace approximation for the model
VAR(1) and BVAR(4) model. Somewhat sur- without training sample.
TABLE 4--TESTING THE EMPIRICAL IMPORTANCE OF THE NOMINAL AND REAL FRICTIONS IN
We focus first on the nominal frictions. Re- ation plays a very important role in the model
ducing the degree of nominal price and wage dynamics. On the contrary, restricting the
stickiness to a Calvo probability of 0.10 is about price indexation parameter to a very low
equally costly in terms of a deterioration of thevalue of 0.01 leads to an improvement of the
marginal likelihood. In both cases the marginal marginal likelihood, suggesting that empiri-
likelihood falls very significantly by about 50. cally it would be better to leave this friction
A lower degree of price stickiness leads to a out. Moreover, leaving out either friction does
strong increase in the estimated degree of price not have any noticeable impact on the other
indexation from 0.22 to 0.84. In addition, the parameters.
variance and the persistence of the price Turning to the real frictions, the most impor-
mark-up shocks increase as a result. The other tant in terms of the marginal likelihood are the
parameters are less affected. The main impact investment adjustment costs. Reducing the elas-
of reducing the degree of wage stickiness on the ticity of adjustment costs to a very low level
other parameters concerns the elasticity of leads to a deterioration of the marginal likeli-
wages with respect to employment: the labor hood by 160. Also, reducing habit formation in
supply elasticity becomes much smaller and consumption is quite costly, although much less
falls from a value of 1.92 to 0.25. In terms of so than reducing investment adjustment costs.
short-run dynamics, these changes more or lessThe reduced hump-shaped endogenous dynam-
cancel out, leaving the impact of labor effort onics of the model due to these restrictions is
wage dynamics unaffected. In this case, thecompensated mainly by higher and more persis-
variance and the persistence of the wagetent exogenous shocks to productivity, invest-
mark-up shock increases. ment, consumption, and government spending.
While both Calvo frictions are empiricallyThe other real frictions fall, while the nominal
quite important, neither price nor wage index- rigidities increase. The presence of variable capital
Forecast horizon
fixed costs in production does
Inflation
mechanically in
crease the standard deviation of the productivit wage mark-up
shock.
pice mark-up
01 02 Q4r0
productivty
investment adjustment costs are the most im- 9096" wage mao-up
so%-
Investment
40%r
V. Applications 30W-
exogenous spend.
20%"
nsk prrmum
10:-
Q1 02 Q4 o0 Q40 1
productivity
use it to investigate a number of key macroeco- FIGURE 1. FORECAST ERROR VARIANCE DECOMPOSITION
nomic issues. In this section, we address the (At the mode of the posterior distribution)
following questions. First, what are the main
driving forces of output0 Second, can the model
replicate the cross correlation between output than 50 percent of the forecast error variance of
and inflation0 Third, what is the effect of a output up to one year. Each of those shocks can
productivity shock on hours worked0 And be categorized as "demand" shocks in the sense
fourth, why have output and inflation become that they have a positive effect on output, hours
less volatile0 We study these issues in each worked, inflation, and the nominal interest rate
subsection in turn. under the estimated policy rule. This is illus-
trated in Figure 2, which shows the estimated
A. What Are the Main Driving Forces of mean impulse response functions to each of
Output0 those three shocks. Not surprisingly, the risk
premium shock explains a big part of the
Figure 1 gives the forecast error variance short-run variations in consumption, while the
decomposition of output, inflation, and the fed- investment shock explains the largest part of
eral funds rate at various horizons based on the investment in the short run (not shown).14
mode of the model's posterior distribution re- In line with the results of Matthew D. Shapiro
ported in Section III. In the short run (within a and Mark Watson (1989), however, it is primar-
year) movements in real GDP are primarily ily two "supply" shocks, the productivity and
driven by the exogenous spending shock and the the wage mark-up shock, that account for most
two shocks that affect the intertemporal Euler of the output variations in the medium to long
equations, i.e., the risk premium shock which run. Indeed, even at the two-year horizon, to-
affects both the consumption and investment
Euler equation and the investment-specific tech-
nology shock which affects the investment Eu- 14 The full set of impulse response functions, as well as
ler equation. Together, they account for more the associated confidence sets, are available upon request.
04 0
00
0.
.0.41
o. o
sl .0k
4.5v
81, 0.
-0.9
04 025
OLT 0,1
oj
O.Z
0,12
40..
0.09.
km
0.11
0,02. O.OF
-0.0
em* -0,04
FIGURE 2. THE ESTIMATED MEAN IMPULSE RESPONSES TO FIGURE 3. THE ESTIMATED IMPULSE RESPONSE TO A WAGE
"DEMAND" SHOCKS MARK-UP SHOCK
Note: Bold solid line: risk premium shock; thin solid line: Note: The solid line is the mean impulse response; the
exogenous spending shock; dashed line: investment shock. dotted lines are the 10 percent and 90 percent posterior
intervals.
gether the two shocks account for more than 50 B. Determinants of Inflation and the Output-
percent of the variations in output. In the longer Inflation Cross Correlation
run, the wage mark-up shock dominates the
productivity shock. Those shocks also become Figure 1 also contains the variance decompo-
dominant forces in the long-run developments sition of inflation. It is quite clear that, at all
of consumption and, to a lesser extent, invest- horizons, price and wage mark-ups are the most
ment. Not surprisingly, the wage-markup shock important drivers of inflation. In the short run,
is also the dominant factor behind long-run price mark-ups dominate, whereas in the me-
movements in hours worked. As shown in Fig- dium to long run, wage mark-ups become rela-
ure 3, a typical positive wage mark-up shock tively more important. Even at the medium- to
gradually reduces output and hours worked by long-run horizons, the other shocks explain only
0.8 and 0.6 percent, respectively. Confirming a minor fraction of the total variation in infla-
the large identified VAR literature on the role of tion. Similarly, monetary policy shocks account
monetary policy shocks (e.g. Christiano, for only a small fraction of inflation volatility.
Eichenbaum, and Evans 2000), monetary policy This is also clear from Figure 4, which depicts
shocks contribute only a small fraction of the the historical contribution of the different types
forecast variance of output at all horizons. of shocks to inflation over the sample period.
Figure 4 shows the historical contribution of The dominant source of secular shifts in infla-
each of four types of shocks (productivity, de- tion is driven by price and wage mark-up
mand, monetary policy, and mark-up shocks) to shocks. Monetary policy did, however, play a
annual output growth over the sample period. It role in the rise of inflation in the 1970s and the
is interesting to compare the main sources of the subsequent disinflation during the Volker pe-
various recessions over this period. While the riod. Moreover, negative demand shocks con-
recessions of the early 1990s and the beginning tributed to low inflation in the early 1990s and
of the new millennium are driven mainly by the start of the new millennium.
demand shocks, the recession of 1974 is due There are at least two reasons why the vari-
primarily to positive mark-up shocks (associ- ous demand and productivity shocks have only
ated with the oil crisis). Monetary policy shocks limited effects on inflation. First, the estimated
play a dominant role only in the recession of the slope of the New Keynesian Phillips curve is
early 1980s when the Federal Reserve, under very small, so that only large and persistent
the chairmanship of Paul Volker, started the changes in the marginal cost will have an im-
disinflation process. pact on inflation. Second, and more importantly,
-2
-4
-6
196 1967 1968 196 1970 197 1972 1973 1974 1975 1976 197 1978 197 1980 198 1982 1983 1984 1985 1986 1987 198 198 190 19 192 193 194 195 196 197 198 19 20 201 20 203 204
Iproducivtyemanlk-
0-
-2
-4
-6
196 1967 1968 196 1970 197 1972 1973 1974 1975 1976 197 1978 197 1980 198 1982 1983 1984 1985 1986 1987 198 198 190 19 192 193 194 195 196 197 198 19 20 201 20 203 204
SproducivtyOemanEl0k-I
Notes: The demand shocks include the risk premium, investment-specific technology, and exogenous spending shocks; the
mark-up shocks include the price and wage mark-up shocks. Trend per-capita growth is estimated at 1.73 percent, whereas
mean inflation is estimated at 3.17 percent.
0.00
Co"eton:on9f9 In do9t'k)
P
00
of
Azz B.Z
4Ar om
02
-0.3
tS 4 5 f 1 8 1 10 I1 6 a a I z
output hous
OfA
qr
0
a41j
403
-0.02
4x8~
0.09
.006 0(
under the estimated monetary policy reaction and current output, and the positive correlation
function, the Fed responds quite aggressively to between current output and inflation one year
emerging output gaps and their impact on infla- ahead. Moreover, the correlations generated by
tion. This is reflected in the fact that at the short- the DSGE model are significantly different
and medium-term horizon more than 60 percent from zero. Decomposing the cross-covariance
of variations in the nominal interest rate are duefunction in contributions by the different types
to the various demand and productivity shocks, of shocks, we find that the negative correlation
in particular the risk premium shock (third panel between current inflation and future output is
of Figure 1). Only in the long run does the wage driven primarily by the price and wage mark-up
mark-up shock become a dominant source of shocks. In contrast, the positive correlation be-
movements in nominal interest rates. tween the current output gap and future inflation
In the light of these results, it is interesting to is the result of both demand shocks and mark-up
see to what extent our model can replicate the shocks. Monetary policy shocks do not play a
empirical correlation function between output role for two reasons. First, they account for only
and inflation as, for example, highlighted in a small fraction of inflation and output devel-
Gall and Gertler (1999). Figure 5 plots the em- opments. Second, as shown in Figure 6, accord-
pirical correlation function of output (detrended ing to the estimated DSGE model, the peak
using the Hodrick-Prescott filter) and inflation effect of a policy shock on inflation occurs
(estimated over the period 1966:1-2004:4), as before its peak effect on output.
well as the median and the 5 percent and 95
percent equivalent generated by the model's
posterior distribution. In order to generate this C. The Effect of a Productivity Shock on
distribution, 1,000 draws from the posterior dis- Hours Worked
tribution of the model parameters are used to
generate artificial samples of output and infla- Following Gall (1999), there has been a
tion of the same sample size as the actual data- lively debate about the effects of productivity
set. For each of those 1,000 artificial samples, shocks on hours worked and about the implica-
the autocorrelation function is calculated and tions of this finding for the role of those shocks
the median and 5 and 95 percentiles are derived. in US business cycles. Gall (1999), Francis and
Figure 5 clearly shows that the DSGE model isRamey (2005), and Galf and Pau Rabanal
able to replicate both the negative correlation(2004) have argued that due to the presence of
between inflation one to two years in the past nominal price rigidities, habit formation, and
E~ I
919
ii nificantly positive only after two years.'5 Under
Ila
il
ai
05
tion, nominal and real interest rates fall, but not
enough to prevent the opening up of an output
ol~
Uc a~
hotw
mation results show that it is mainly the esti-
output
94i
G
rai
importance of capital adjustment costs that ex-
0.2 ;
c,,vj
plain the negative impact of productivity on
o-l i
.P,~O
hours worked, thereby confirming the analysis
-0.04
4.07n a.W
of Francis and Ramey (2004). Indeed, also un-
ab
der flexible prices, hours worked would fall
imerest rate
ination
significantly as indicated in the upper-right-
FIGURE 7. THE ESTIMATED IMPULSE RESPONSES TO A hand panel of Figure 7. Given these estimates, it
PRODUCTIVITY SHOCK is unlikely that a more accommodative mone-
tary policy would lead to positive employment
Note: The solid lines represent the estimated actual mean
responses and the 10 percent and 90 percent posterior in- effects. The relatively low medium-run positive
terval; the dashed lines represent the counterfactual flexible- effects on hours worked are due to two factors.
wage-and-price responses. First, although persistent, the productivity
shock is temporary. As a result, output already
starts returning to baseline when the effects on
hours worked start materializing. A different
adjustment costs to investment, positive produc- stochastic process for the productivity shock,
tivity shocks lead to an immediate fall in hours which implies a gradual introduction of higher
worked. Given the strongly positive correlation total factor productivity, could increase the ef-
between output and hours worked over the busi- fect on hours worked.16 Second, a positive pro-
ness cycle, this implies that productivity shocks ductivity shock reduces the fixed cost per unit of
cannot play an important role in the business cy- production, and therefore less labor is required
cle. In contrast, using alternative VAR specifica- for a given output.
tions and identification strategies, Christiano,
Eichenbaum, and Robert Vigfusson (2004),
Luca Dedola and Stefano Neri (2004), and Gert D. The "Great Inflation" and the "Great
Peersman and Roland Straub (2005) have ar- Moderation": Subsample Estimates
gued that the empirical evidence on the effect of
a productivity shock on hours worked is not In this section we first compare the estimates
very robust and could be consistent with a for two subsamples in order to investigate the
positive impact on hours worked. stability of the full-sample estimates, and then
In Section VA we have already discussed that examine, using those estimates, why output and
productivity shocks play an important, but not inflation volatility have fallen in the most recent
dominant, role in driving output developments period. The first subsample, corresponding to
beyond the one-year horizon in our estimated the period 1966:2-1979:2, captures the period
model. At business cycle frequencies, they ac- of the "Great Inflation" and ends with the ap-
count for about 25-30 percent of the forecast pointment of Paul Volcker as chairman of the
error variance. Figure 7 presents the response of Federal Reserve Board of Governors. The sec-
the actual and the flexible-price level of output, ond subsample, 1984:1-2004:4, captures the
hours worked, and nominal interest rate to a
productivity shock in the estimated model.
Overall, the estimates confirm the analysis of 15 This picture does not change very much when we do
not allow for a positive effect of productivity on exogenous
Gall (1999) and Francis and Ramey (2004). A
spending.
positive productivity shock leads to an expan- 16 See, for instance, Rotemberg (2003) for arguments
sion of aggregate demand, output, and real favoring a slow appearance of major productivity advances
wages, but an immediate and significant reduc- in output growth.
more recent period of the "Great Moderation," same. These results are consistent with the find-
in which not only was inflation relatively low ings of Athanasios Orphanides (2003), who
and stable, but also output and inflation vol- shows, using real-time data estimates, that what
atility fell considerably (e.g., Margaret M. has changed in US monetary policy behavior
McConnell and Gabriel Perez-Quiros 2000). since the early 1980s is the relative response to
Table 5 compares the mode of the posterior output. They are, however, at odds with the
distribution of the DSGE model parameters results of Jean Boivin and Marc Giannoni
over both periods. (2006), which find that a stronger central bank
The most significant differences between the response to inflation in the second subperiod
two subperiods concern the variances of the can account for a smaller output response to
stochastic processes. In particular, the standard monetary policy shocks estimated in identified
errors of the productivity, monetary policy, and VARs. In our case, the lower response to the
price mark-up shocks (and to a lesser extent the output gap actually increases the output re-
investment shock) seem to have fallen. The sponse of a monetary policy shock in the second
persistence of those processes has changed period.
much less. One exception is the risk premium Interestingly, it turns out that the degree of
shock, which has become even less persistent in price and wage stickiness has increased in the
the second subperiod. second period, while the degree of indexation
Somewhat surprisingly, the steady-state in- has fallen. The latter is consistent with single-
flation rate is only marginally lower in the sec- equation subsample estimates of a hybrid New
ond subperiod (2.6) versus the first period (2.9). Keynesian Phillips curve by Gall and Gertler
What is different is the central bank's reaction (1999). This finding is also consistent with the
coefficient to the output gap, which is halved story that low and stable inflation may reduce
and is no longer significant in the second period. the cost of not adjusting prices and therefore
In contrast, the response to inflation is only lengthen the average price duration leading to a
marginally higher in the second period, and the flatter Phillips curve. At the same time, it may also
response to the change in the output gap is the reduce rule-of-thumb behavior and indexation