€¦ · Web viewKeywords: central bank communication, open-ended announcement, non-standard...
Transcript of €¦ · Web viewKeywords: central bank communication, open-ended announcement, non-standard...
1
TITLE: Macroeconomic effects of an open-ended Asset Purchase Programme1
RUNNING TITLE: Macroeconomic effects of an open-ended APP
Authors:Lorenzo Burlon (European Central Bank, Directorate General Monetary Policy)
Alessandro Notarpietro (Banca d’Italia, Directorate General for Economics, Statistics and Research)
Massimiliano Pisani(corresponding author, Banca d’Italia, Directorate General for Economics, Statistics and Research, via Nazionale 91, 00184, Rome, Italy, [email protected])
Abstract We evaluate the effectiveness of an open-ended Asset Purchase Programme (APP) using a New Keynesian model of euro area. We assume that the central bank does not announce the ending date of the programme, and agents form their expectations about possible additional purchases beyond the horizon of the announcement according to the central bank reaction function, that links the purchases to the expected inflation gap. It is showed that the open-ended APP is more effective in stimulating macroeconomic conditions than committing ex ante to an ending date.
JEL classification: E43; E44; E52; E58.
Keywords: central bank communication, open-ended announcement, non-standard monetary policy, DSGE models, open-economy macroeconomics, euro area.
1 We thank Fabio Busetti, Martina Cecioni, Sandra Gomes, Stefano Neri, and participants at the Working Group on Econometric Modeling of the European System of Central Banks (March 2018) for useful comments. All errors are ours. The opinions expressed in this paper are those of the authors and do not necessarily reflect the views of Banca d’Italia, European Central Bank, or Eurosystem.
2
“We have proven in the past that our forward guidance is credible. This has
been the case both for our guidance on rates and on our reaction function,
notably when we laid out the contingencies that would justify launching an
asset purchase programme in response to a too-prolonged period of low
inflation.”
President of the European Central Bank, Mario Draghi2
1 IntroductionIn January 2015 the Eurosystem extended its asset purchase programme (APP) to euro-
denominated investment-grade securities issued by euro-area (EA) governments and
institutions in the secondary market. The reason for this decision was the progressive
fall of inflation that was leading to a de-anchoring of inflation expectations from levels
consistent with price stability and risks of a deflationary spiral. The implementation of
these measures contributed to improve EA macroeconomic conditions and favored a
substantial progress towards a sustained adjustment in the path of inflation.
In December 2016 the ECB started a recalibration of the non-standard monetary policy
measures, and in particular the APP.3
The design of this recalibration involved several decisions about its timing, size, and
composition. A feature of the recalibration was the choice between a programme with
an end date for the net purchases that is announced and a programme that remains open-
ended. The ECB chose to maintain the possibility of extending the flow of future net
2 Draghi (2018).3 See Draghi and Constancio (2017).
3
purchases beyond the horizon of the purchases to which it made explicit commitment. A
careful communication was used, based on the concept of “central bank reaction
function”, aimed at managing agents’ (households’ and firms’) expectations about future
purchases of sovereign bonds, linking them to expected future macroeconomic
conditions.
In this paper we evaluate the macroeconomic effectiveness of the announcement of an
open-ended APP. To this purpose, we build a large-scale New Keynesian dynamic
general equilibrium model calibrated to the EA and the rest of the world (RW). The EA
is formalized as a monetary union of two regions, Home and rest of the EA (REA),
where Home is of medium size (its GDP being around 20% of overall EA GDP).
We assume that an open-ended programme implies no explicit commitment by the
central bank to end the purchases. Agents form their expectations about future, not yet
announced, asset purchases that the central bank would implement in the quarters past
the announced horizon according to a reaction function of the central bank. The latter
links the expected purchases to the expected inflation gap, i.e., the distance of expected
inflation from the target. If the expected inflation gap is positive, i.e., if expected
inflation is below the target, then agents expect additional purchases (that have not been
announced) and no purchases otherwise. The central bank reaction function is known by
agents on the basis of the inference that they may have drawn from previous purchases
announcements.4 We provide an empirical foundation of the reaction function by
estimating the elasticity of expected purchases to the expected inflation gap using data
4 In several speeches the ECB President Mario Draghi had suggested that the communication would have been aimed at clarifying the ECB reaction function. See, for example, Draghi (2014, 2018). According to Praet (2018b), “the convergence of market expectations towards the ECB's indications about the size and horizon of net asset purchases has proceeded in tandem with the gradual rise of in inflation expectations towards our medium-term policy aim. Likewise, there is a strong co-movement between markets views of the convergence of in inflation towards the ECBs aim on the one hand and, on the other, the length of the interval between the end of net asset purchases and the first rate rise fully priced into market rates. This suggests that our monetary policy reaction function is well understood.”
4
on announced additional purchases and the inflation gap in correspondence of the
considered announcement. The estimated elasticity is positive and significant, in line
with anecdotal evidence provided by financial analysts.5 This allows us to discipline the
results of our simulations from a quantitative point of view.
We use the ECB October 2017 announcement on the pace and amount of future
purchases as a case study for our analysis. The APP consists of purchases of long-term
sovereign bonds by the EA central bank. Following Chen, Curdia, and Ferrero (2012),
we assume that in each EA country some households (labeled “restricted”) invest in
domestic long-term sovereign bonds and physical capital (they do not invest in short-
term bonds). This feature captures the presence of investors who have a demand over
specific asset classes and maturities, as formalized in the preferred-habitat framework.
The presence of these agents allows us to obtain real effects from central bank purchases
of long-term sovereign bonds, as those households’ decisions depend only on long-term
rates.6 The rest of the households (labeled “unrestricted”) have instead access to
multiple asset markets and, in particular, their decisions are directly affected by the
short-term monetary policy rates. Moreover, we assume forward guidance (FG) on the
short-term policy rate, i.e., the central bank announces the number of periods the policy
rate should be expected to remain at its current (baseline of the model) level.
The other model features are in line with those of existing large-scale dynamic general
equilibrium models of the EA, like the Eurosystem’s EAGLE (Gomes, Jacquinot, &
Pisani, 2010) and the ECB’s New Area Wide Model (Christoffel, Coenen, & Warne,
2008). In particular, we distinguish between final (non-tradable) consumption and
5 See Praet (2018b) and Skolimowski (2017).6 See Andres, Lopez-Salido, and Nelson, (2004) and Modigliani and Sutch (1966). Recently, Altavilla, Carboni, and Motto (2015), Blattner and Joyce (2016), and Ferdinandusse, Freier, and Ristiniemi (2017) have provided empirical evidence on the presence of preferred-habitat investors in the EA.
5
investment goods and between intermediate tradable and non-tradable goods, produced
according to sector-specific technologies exploiting domestic capital and labor. We also
include standard nominal (price and wage) and real (consumption habit and investment
adjustment costs) frictions.
We consider several scenarios. We compare purchases under a “full” announcement
(i.e., an announcement of a purchasing programme that specifies both the overall path of
future purchases and their end date) to the case of an “open-ended” announcement, in
which the end date is not specified and the possibility of additional future purchases is
communicated. All announcements are accompanied by the FG on the short-term policy
rate. In its announcements, the ECB connected the end of the FG to the end of the APP
by explicitly communicating that monetary policy rates should be expected to remain at
their present (low) levels for an extended period of time, and, crucially, past the horizon
of net asset purchases. This assumption on the FG is consistent with the current wording
of ECB official communication, according to which the first rate hike would occur only
“well past the horizon” of net asset purchases (Draghi & Constancio, 2017).
Moreover, we assess the effectiveness of the two alternative programmes when a
negative demand shock to EA aggregate demand reduces EA in inflation and, thus,
widens the gap from the inflation target.
Our results are the following. First, the announcement of an open-ended APP is more
effective in stimulating macroeconomic conditions than an announcement that commits
to an end date. Given the initial positive inflation gap, agents expect additional
purchases in the case of open-ended APP. Thus, the long-term interest rate decreases
more than in the case of announcing an end date and economic activity and inflation
increase more. Second, the effectiveness of the open-ended programme is reinforced by
6
the FG. In particular, for a given amount of announced purchases, keeping the monetary
policy rate unchanged past the purchases’ horizon stimulates economic activity and
reduces the expected inflation gap. This induces households to expect a lower amount of
additional (unannounced) purchases. Third, the open-ended APP is more effective in
stabilizing the economy if negative EA aggregate demand shocks unexpectedly
materialize and push down inflation. In this case, the widening of the inflation gap
associated with the shocks induces agents to expect a larger amount of additional
purchases by the central bank.
Among the recent studies that analyze the effectiveness of the APP, there is a growing
consensus that the adoption of the APP has had both financial and real effects, and that
the latter are mostly consistent with a portfolio rebalancing channel.7
Different from previous studies, our exercise features two EA-specific characteristics,
i.e., the monetary union and open-economy dimensions, and attempts a first description
of expectations’ formation by private agents in presence of an open-ended programme.
In Ellison and Tischbirek (2014) asset purchases are carried out according to a
systematic rule. As in that contribution, we endogenize purchases as a function of the
inflation gap, using a reaction function. Different from it, we assume that the central
bank does not systematically follow a reaction function (the reaction function is not a
systematic rule), but follows it only for a limited number of periods after the
announcement of the open ended-programme.
Burlon et al. (2017) study the macroeconomic effectiveness of an APP with announced
end-date under different assumptions regarding the duration of the purchases and of the
FG on the policy rates. One of the main conclusions of that paper is that the more distant
7 See, among others, Breckenfelder et al. (2016).
7
in the future the announced closing date of a programme, the larger the (current and
future) effectiveness of the programme itself. The model used in our analysis builds on
the one in Burlon et al. (2017) with one crucial departure, i.e., the presence of the central
bank reaction function in the case of an open-ended programme. Households know the
function and use it to form their expectations about future unannounced asset purchases
by the central bank.
Our analysis clarifies that two elements of central bank communication are crucial (from
both a theoretical and a policy perspective) for the effectiveness of non-standard
measures: (1) agents’ expectations about future unannounced purchases and (2) the
ability of the central bank to shape those expectations through (past and current)
announcements. We are the first, to the best of our knowledge, to make this type of
analysis in a fully-fledged New Keynesian model.8
Thus, from a policy perspective, our paper emphasizes the key role of central bank
communication for the effectiveness of non-standard monetary policy measures.
Communication allows the central bank to manage expectations of economic agents.
This is particularly relevant when macroeconomic conditions and the monetary policy
tools are not the standard ones and, thus, are not well known by economic agents and/or
their functioning and interaction is difficult to explain to the public.9
The paper is organized as follows. Section 2 describes the main features of the model, in
particular the non-standard monetary policy measures (sovereign bond purchases by the
central bank) and the formation of agents’ expectations on future purchases by the 8 For other analyses of non-standard monetary policy measures, that, however, do not consider the central bank reaction function, see Breckenfelder et al. (20116), Darracq Paries and Kuhl (2016), Harrison (2012), Reis (2017), and Woodford (2016). See Coenen et al. (2017) for a discussion of the role of communication of monetary policy in unconventional times.9 According to Praet (2017), “In normal times, the central bank reaction function mainly specifies the contingencies for loosening, tightening or maintaining a given policy stance. In unconventional times, it also needs to spell out which tool would provide the first margin of adjustment if a change in policy stance were to appear warranted, and how the adjustment in one policy instrument would interact with the others.”
8
central bank in the case of open-ended APP. Section 3 describes the simulated scenarios.
Section 4 reports the main results. Section 5 concludes.
2 The modelWe first provide an overview of the model. Subsequently, we describe its key features,
i.e., the reaction function and the EA central bank, (restricted and unrestricted)
households and capital goods producers, the fiscal authority, and the definition of
general equilibrium. Finally, we report the calibration.
2.1 Overview
The model is New Keynesian and represents a world economy composed of three
regions, i.e., Home, REA (Home+REA=EA), and RW. The size of the world economy
is normalized to 1. Home, REA, and RW have sizes equal to n, n*, and (1-n-n*), with 0
< n, n*< 1 and n+n* < 1.10 Home and REA share the currency and the central bank. The
latter sets the policy rate according to EA-wide variables (a standard Taylor rule holds)
and deliberately enacts non-standard monetary policy measures. The presence of the
RW outside the EA allows us to assess the role of the nominal exchange rate and extra-
EA trade for the transmission of the EA shocks.
Crucial features of the model are two: the central bank reaction function, used by
households to form expectations about future (unannounced) APP purchases in the
open-ended scenario, and financial market segmentation.
10 For each region, size refers to the overall population, to the number of firms operating in each sector and, in the case of each EA region, the number of capital producers.
9
The first key, and novel, feature is that, under an open-ended programme, when the
central bank does not commit to an end date and keeps the possibility of continuing
purchases past the announcement horizon, agents follow a rule to form their
expectations about future unannounced purchases. The rule is the central bank reaction
function, which is known by the agents, and links the expected purchases to the
expected inflation gap, i.e., the distance of expected inflation from the target.
The second crucial feature of the model is financial market segmentation as in Chen,
Curdia, and Ferrero (2012), which allows us to relax the well-known “Wallace
neutrality” and make financial assets imperfect substitute so that sovereign bonds’
purchases by the central bank have real effects in our model.11 In each EA region there
are two types of households, “restricted” and “unrestricted.” Restricted households can
invest only in the domestic long-term sovereign bond market and domestic private
physical capital. The increase in the purchases of long-term government bonds by the
monetary authority reduces long-term interest rates (their term premium component)
and therefore induces restricted households to increase consumption and investment in
physical capital. To the opposite, unrestricted households (1) have access to the
domestic short-term private bond and long-term sovereign bond markets, (2) trade a
riskless private bond with RW households, and (3) invest in physical capital. The
consumption/saving decisions of the unrestricted households depend not only on long-
term interest rates, but also on short-term ones. Thus, FG directly affects the optimal
choices of those households, in particular consumption and investment in physical
capital through the capital producers. Finally, there are capital producers that
accumulate private physical capital by demanding final investment goods subject to
11 See Wallace (1981).
10
quadratic adjustment costs on investment change. They rent out capital to the domestic
firms producing intermediate goods. They maximize profits with respect to capital and
investment taking prices as given, and evaluate returns according to a weighted average
of restricted and unrestricted households’ stochastic discount factors (where the weights
reflect the corresponding population shares). The (net) revenues are rebated in a lump-
sum way to domestic restricted and unrestricted households according to their
corresponding shares.
Remaining features of the model are standard. Households consume a final non-tradable
good which is a composite of intermediate non-tradable and tradable goods. The latter
are domestically produced or imported. All households supply differentiated labor
services to domestic firms and act as wage setters in monopolistically competitive labor
markets by charging a mark-up over their marginal rate of substitution between
consumption and leisure.
On the production side, there are perfectly competitive firms that produce two final
goods (consumption and investment goods) and monopolistic firms that produce
intermediate goods (firms are owned by domestic unrestricted households). The two
final goods are sold domestically and are produced combining all available
intermediate goods using a constant-elasticity-of-substitution (CES) production
function. The resulting bundles can have different compositions. Intermediate tradable
and non-tradable goods are produced combining domestic capital and labor. The two
factors are assumed to be mobile across sectors. Intermediate tradable goods can be
sold domestically and abroad. Since intermediate goods are differentiated, firms have
market power and, thus, are price setters and restrict output to create excess profits. We
also assume that markets for tradable goods are segmented, so that firms can set a
11
different price for each of the three markets. Adjustment costs ensure that main
variables react smoothly to a shock. Specifically, on the real side, habits and quadratic
costs prolong the adjustment of consumption and investment, respectively. On the
nominal side, quadratic costs make wages and prices sticky.12 Home and REA setups
are similar, so in what follows we report only Home key equations. The RW setup is
standard (the representative household framework holds).
2.2 Expectations about future bond purchases in the open-ended APP
Consistent with the evidence on the inference by market participants reported by
Skolimowski (2017), in the “open-ended” scenarios we assume that agents (households
and firms) in both EA regions form their expectations about future central bank
purchases according to the central bank’s reaction function. The function dictates that
in correspondence of positive expected inflation gaps in the quarters past the end of
announced purchases, agents form their expectations about additional unannounced
future central bank purchases in those quarters, according to the size of the (positive)
inflation gap. Thus, we assume that in both EA regions expectations over future
unannounced purchases evolve as follows:
Et [ BAPP , t+ hL −B APP ,t+h−1
L ]=ϕAPP (π target−Et [ π t+h ] ) ,(1)
where Et is the expectation operator in quarter t, BAPP ,t+hL are purchases in quarter t+h,
π target is the constant (annualized) inflation target of the central bank, Et [ π t+h ] is the
expected (annualized) inflation rate in quarter t + h with h = 1, 2, 3, . . . , and ϕ APP>0is
12 See Rotemberg (1982).
12
a parameter regulating the amount of additional future expected net central bank
purchases corresponding to each percentage-point deviation of expected inflation from
the central bank target. The reaction function specifies the overall amount of bonds that
agents expect the EA central bank to buy in the Home and REA secondary long-term
sovereign bond markets. Following the ECB experience, country-level purchases (both
announced and expected) are determined according to each region’s capital key, which
roughly coincides with the regional share of overall EA GDP.
In the case of the ECB, even if the size, pace, and timing of the APP changed from one
announcement to the other, the official communication has always aimed these
announcements to the achievement of a sustained adjustment in the path of inflation,
“that is consistent with its aim of achieving inflation rates below, but close to, 2% over
the medium term.” Thus, it is plausible that market participants and private analysts
have strived to map the features of each announcement into some possible form of
ECB’s “reaction function,” that connects the expected adjustment in inflation to the
expected amount of sovereign bonds’ purchases.
The reaction function should not be interpreted as describing a central bank's
systematic feedback rule regarding asset purchases. 13 The reaction function describes,
in a stylized way, agents’ expectations over unannounced future purchases, and does
neither imply nor intend to model a complete control of the central bank over long-
term interest rate developments.
Sovereign bond purchases imply that the central bank purchases long-term (domestic)
sovereign bonds in the Home and REA secondary markets to affect the long-term
13 This corresponds to actual recent policy experience by the Bank of Japan and its yield-curve control policy. See Amamiya (2017) and Jones and Kulish (2013).
13
interest rates and hence the real side of the economy.
The other non-standard measure that we consider is FG. The central bank commits to
keeping the policy rate constant at its baseline (steady-state) level for a prolonged
period of time. Thereafter it follows the Taylor rule
( Rt
R )=( Rt−1
R )ρR
( Π EA ,t−3
Π EA )(1−ρR ) ρΠ
( GDPEA , t
GDPEA ,t−1 )(1−ρR ) ρGDP
. (2 )
The parameter ρR (0< ρR<1) captures inertia in interest-rate setting, while the term R
represents the steady-state gross nominal policy rate. The parameters ρΠ (ρΠ >0) and
ρGDP (ρGDP>0) are respectively the weights of EA consumer price (CPI) inflation
rate (Π EA ,t−3≡ PEA, t / PEA ,t−4) and GDP.14 The parameter Π EA is the long-run (steady-
state) constant inflation target of the central bank.
Households and firms, being forward-looking, anticipate this future change in the
monetary policy stance when making consumption and investment decisions.
2.3 Restricted households
There exists a continuum of restricted households j ', with j' ∈ (0; λR], where 0 ≤ λR≤ 1.
Their preferences are additively separable in consumption and labor effort. The generic
restricted household j' receives utility from consumption CR , t and disutility from labor
LR ,t. The household’s expected lifetime utility is
E0∑t=0
∞
βRt [ (C R,t ( j ')−hCR , t−1 )1−σ
1−σ−
LR,t ( j ' )1+φ
1+φ ](3)
14 The CPI inflation rate is a geometric average of Home and REA inflation rates, with weights equal to the correspondent (steady-state) regional GDP (as a share of the steady-state EA GDP). EA GDP is the sum of Home and REA GDPs. A similar equation describes monetary policy in the RW region.
14
where E0 denotes the expectation conditional on information set at date 0; βR is the
discount factor (0 < βR < 1), 1/σ is the elasticity of intertemporal substitution (σ>0),
and 1/φ is the labor Frisch elasticity (φ>0). The parameter h (0<h<1) represents
external habit formation in consumption.15 Restricted households have access only to
the market of long-term sovereign bonds. The budget constraint is
PtL BR, t
L ( j' )−(1+κPtL )BR, t−1
L ( j' )=Π tprof ( j ')+W R ,t ( j' ) LR ,t ( j' )−Pt CR,t ( j' )−AC R,t
W ( j ' )(4 )
where BR ,tL ( j' )is the amount of domestic long-term sovereign bonds, Π t
prof ( j' ) are profits
from ownership of the Home capital producers, and W R ,t ( j ' ) is the nominal wage. The
long-term sovereign bond BR ,tL ( j ') has price Pt
L and is formalized as a perpetuity paying
an exponentially decaying coupon κ∈ (0,1¿, following Woodford (2001). The implied
gross yield to maturity at time t on the long-term bond is defined as
RtL ≡ 1
PtL +κ .(5)
Finally, households act as wage setters in a monopolistically competitive labor market.
Each household j' supplies one particular type of labor services which is an imperfect
substitute to services supplied by other households. It sets its nominal wage W R ,t ( j ' )
taking into account the labor demand and quadratic adjustment costs ACR,tW ( j' ),
AC R ,tW ( j' ) ≡
k W
2 ( W R, t ( j' )W R ,t−1 ( j ')Π t−1
αW Π EA1−α W
−1)2
W R , t LR , t(6)
where kW>0 and 0≤ αW ≤ 1 are parameters regulating wage stickiness and inertia,
respectively, and the variable Π t−1is the previous period Home consumer price
15 Following common practice in the New Keynesian literature, the assumption of cashless economy holds in the model.
15
inflation rate. The adjustment costs are proportional to the per-capita (average) wage
bill of restricted households W R ,t LR,t .
Restricted households are crucial for the APP to have real effects in our model. Since
they cannot arbitrage between short-term and long-term bonds, their consumption
decisions directly depend only upon the long-term interest rate (rather than the short-
term rate). Therefore, the monetary policy authority can affect their consumption and
saving decisions by intervening in the secondary long-term sovereign bond market.
2.4 Unrestricted households
There exists a continuum of unrestricted households, indexed by j, with j ∈ (nλR; n].
These households have the same preferences as restricted households described in (2),
thus they consume and supply labor. The only difference is the discount factor,
0<βU<1, which can be different from that of restricted households (this allows us to
have a well-defined steady-state of the model).
Home unrestricted households have access to multiple financial assets (all denominated
in euros): the short-term (one-period) sovereign bond BtG ( j ), exchanged with the
domestic government and paying the EA (gross) monetary policy rate Rt; the short-
term private bond BtP ( j ), exchanged with REA unrestricted and RW households and
paying the (gross) interest rate RtP; the long-term sovereign bond BU ,t
L ( j ), exchanged
with the domestic restricted households, domestic government and, because of the
APP, the EA central bank. Thus, they have several opportunities to smooth
consumption when facing a shock. The budget constraint of the generic unrestricted
household j is
16
PtL BU , t
L ( j )−(1+κPtL)BU ,t−1
L ( j )+BtG ( j )−B t−1
G ( j ) R t−1+B tP ( j )−Bt −1
P ( j ) Rt −1P (1−ϕt−1 )=Π t
P ( j )+Π tprof ( j )+W U ,t ( j ) LU ,t ( j )−Pt CU ,t ( j )−TAX t ( j )−ACU , t
W ( j )−ACU , tB ( j ) .(7)
The dividends Π tP ( j )are from ownership of domestic monopolistic firms (claims to
firms’ profits are not internationally tradable). The term ϕt−1 represents an exponential
adjustment costs, needed to stabilize the position in the internationally traded bond.16
The variable profit Π tprof ( j ) denotes profits from ownership of the Home capital
producers. The term TAX t ( j ) represents lump-sum taxes. Unrestricted households
supply labor services under monopolistic competition and face quadratic adjustment
costs ACU ,tW ( j ) when setting nominal wages (the cost is similar to the one paid by
restricted households). They also pay adjustment costs ACU ,tB ( j ) on long-term
sovereign bond holdings.17
First order conditions imply no-arbitrage conditions for the unrestricted households.
Thus, in equilibrium the interest rates paid by the different bonds are expected to be
equal to the monetary policy rate, net of the spreads induced by the longer maturity and
the adjustment costs.
2.5 Capital goods producers
There exists a continuum of mass 0<n<1 of firms, indexed by e, that produce private
16 The adjustment cost is defined as ϕt ≡ ϕb1
exp (ϕb2 (BtP−BP ))−1
exp (ϕb 2 (B tP−BP ))+1
, where ϕb 1 , ϕb 2>0 are parameters, BtP
and BPare the period t and steady-state positions of the representative Home unrestricted household, respectively. Both are taken as given in the maximization problem. A similar cost holds for the RW household.17 We assume a standard quadratic form for the adjustment cost, that is,
ACU ,tB ( j )≡
ϕbL
2 ( PtL BU , t
L ( j )−PL BUL )2(8)
where ϕb L>0 is a parameter and PL BUL is the (symmetric) steady-state value of the long-term sovereign bond. The
adjustment cost guarantees that the bond holdings follow a stationary process and that the economy converges to the steady state.
17
physical capital. Each firm e optimally chooses capital K t (e) and investment I t(e ) to
maximize profits subject to the law of capital accumulation, the adjustment costs on
investment, and taking prices as given. The law of motion of capital accumulation is
K t (e )= (1−δ ) K t−1 (e )+(1−AC I , t (e )) I t (e ) ,(9)
where 0<δ<1 is the depreciation rate. The adjustment cost on investment AC I ,t(e) is
AC I ,t (e )≡ φI2 ( I t (e )
It−1 (e )−1)
2
I t (e )(10)
where φI >0 is a parameter. Capital producers rent the existing physical capital stock
K t−1(e)at the nominal rate Rk , t to domestic firms producing intermediate tradable and
non-tradable goods. Investment is a final non-tradable good, composed of intermediate
tradable (domestic and imported) and non-tradable intermediate goods. Capital
producers buy it in the corresponding market at price P I.18 Because of the adjustment
costs on investment, a “Tobin’s Q” holds. When maximizing profits with respect to
capital and investment, capital producers discount profits using the stochastic discount
rates of restricted and unrestricted households, aggregated according to the
corresponding population shares.
2.6 Fiscal authority and supply of long-term sovereign bonds
Fiscal policy is set at the regional level. In the Home region there is a government
budget constraint. Public debt consists of two bonds issued to domestic households: the
(short-term) one-period nominal bond, that pays the monetary policy rate and is issued
18 Similarly to the consumption basket, the investment bundle is a composite of tradable and non-tradable goods. The composition of consumption and investment goods can be different.
18
to domestic unrestricted households; the (long-term) bond, issued to domestic
unrestricted and restricted households and, because of the non-standard monetary
policy measures, to the EA central bank.
The Home government follows a fiscal rule that calls for an increase (reduction) in
lump-sum taxes (as a ratio to GDP) whenever the current-period short-term public debt
(as a ratio to GDP) is above (below) the target and increasing over time.19 We assume
that the long-term sovereign bond is kept constant at its steady-state level, so that
changes in the long-term interest rate are entirely due to the non-standard monetary
policy measures.20
2.7 Equilibrium
In each country the initial asset positions, preferences, technologies, and budget
constraints are the same for households belonging to the same type and firms
belonging to the same sector.
Moreover, profits from ownership of domestic monopolistically competitive firms are
equally shared among unrestricted households. Profits from ownership of domestic
19 We choose lump-sum taxes to stabilize public finance as they are non-distortionary (i.e., they do not enter the first-order conditions of households and firms). Lump-sum taxes are paid by unrestricted households only. In this way we are able to isolate the response of restricted households to the APP from the indirect fiscal adjustments implied by the program. However, the Ricardian equivalence does not hold because restricted households hold long-term sovereign bonds but are not subject to lump-sum taxes. Thus, our assumption on the distribution of lump-sum taxes or, equivalently, on the initial distribution of public debt implies that sovereign bond holdings are net wealth.20 We include only the short-term debt in the Home fiscal rule for two reasons. First, we hold the supply of long-term government bonds fixed, so as to isolate the direct demand effects of the APP. Second, we need the fiscal rule to stabilize the short-term debt and, given that the long-term component is exogenous, the overall public debt. We take into account this distinction when we calibrate the model and more specifically the fiscal target. A similar fiscal rule holds in the REA. In the RW the rule holds for the overall public debt, as there is no distinction between short- and long-term domestic sovereign bonds.
19
capital producers are distributed to restricted and unrestricted households according to
the corresponding population shares, and are equally shared within each type. Thus, we
consider the representative household for each type (restricted and unrestricted). We
also have a representative firm for each sector (final non-tradable, intermediate
tradable, and intermediate non-tradable) and a representative capital producer in each
EA region. The implied symmetric equilibrium is a sequence of allocations and prices
such that, given initial conditions and considered shocks, households and firms satisfy
their corresponding first order conditions, the monetary rules, the fiscal rules, the EA
central bank reaction function, and the government budget constraints hold, and all
markets clear.
2.8 Taking the model to the data
The model is taken to quarterly data. The key equation is the reaction function of the
central bank (see eq. 1), which is estimated. Specifically, we estimate the elasticity of
expected purchases to the expected inflation gap using data on (a) additional
announced purchases and (b) the inflation gap in correspondence of the considered
announcement. Data are reported in Table 1
Table 1. Reaction function. Data
Announcement date
Inflation gap (pp)
Purchases (€ bn) / Month
Total additional purchases (€ bn)
January 2015 0.7 60 1080December 2015 0.3 60 350
March 2016 0.2 80 240
20
December 2016 0.5 60 540October 2017 0.4 30 270
The inflation gap is the difference between a target of 1.5% and the core inflation expected (within the Eurosystem's official projection exercises) for the quarter of the announcement in the quarter immediately before (2014Q4 for January 2015, 2015Q4 for December 2015, 2016Q1 for March 2016, 2016Q4 for December 2016, and 2017Q3 for October 2017).
Figure 1 reports the four monetary policy announcements regarding the APP,
highlighting the linear relationship between announced volumes of purchases and the
inflation gap. Table 2 contains the estimated parameters of the regression analysis.
Figure 1. Empirical relation between APP purchases and inflation gaps
21
Jan 2015
Dec 2015Mar 2016
Dec 2016
Oct 2017
0
200
400
600
800
1000
1200
0 0.2 0.4 0.6 0.8 1
Table 2. Reaction function. Estimates
Dependent variable: Purchases (€bn) / Month
Inflation gap (pp) 118.615***(23.234)
Observations 5R-squared 0.785Robust standard errors in parentheses*** p<0.01, ** p<0.05, * p<0.1
The coefficient is estimated to be positive and significant. Consistent with the above
estimates, the parameters of the central bank reaction function are set so as to
correspond to a quarterly flow of additional net purchases equal to € 360 billion, or
10% of quarterly EA GDP, for each pp deviation of EA annualized inflation from the
target. This amount is in line with the estimates reported in Skolimowski (2017) and
based on data on Eurosystem APP purchases and EA inflation gaps.21
21 Skolimowski (2017) also includes data on the targeted longer-term refinancing operations (TLTRO) in the
22
We also provide evidence that, by linking APP and FG announcements,
communication shaped expectations about the subsequent monetary policy rates lift-
off. In October 2017 the ECB extended the net purchases past the end of the year,
signaling that, starting in January 2018, purchases would “continue at a monthly pace
of €30 billion until the end of September 2018, or beyond, if necessary, and in any case
until the Governing Council sees a sustained adjustment in the path of inflation
consistent with its inflation aim.”22 Moreover, the expected path of the policy rate was
linked to the intended horizon of the net asset purchases. In fact, concerning the key
ECB interest rates, the (so-called “well past” FG) statement said that the ECB would
“expect them to remain at their present levels for an extended period of time, and well
past the horizon of our net asset purchases.”
Figure 2 reports evidence of the effects of the “well past” FG announcement on
expectations about the future monetary policy rate path. The day of the announcement
(October 26, 2017), the path of 3-month Euribor rates derived from futures contracts (a
widely used market-based measure of expected future monetary policy rate) shifted
outward. The markets forecasted the lift-off of monetary policy rates (10 basis points
above the spot level, dashed black line) to begin in the first quarter of 2019 (red line),
while the day before the announcement markets had forecasted the first lift-off in the
last quarter of 2018 (blue line).
Figure 2. October 2017 announcement and expectations about the monetary policy rate
estimation. When we simulate the model, the amounts of purchases of Home and REA long-term sovereign bonds are proportional to the corresponding EA GDP share.22 See Praet (2018a).
23
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
Spot Dec-17 Mar-18 Jun-18 Sep-18 Dec-18 Mar-19 Jun-19 Sep-19 Dec-19 Mar-20 Jun-20
25/10/2017
26/10/2017
In our analysis the response of the interest rate on long-term sovereign bonds to the
central bank purchases plays a key role. We use the existing evidence for the EA
reported in Altavilla, Carboni, & Motto (2015) to calibrate the parameter regulating the
adjustment costs paid by the unrestricted household on deviations of long-term
sovereign bond positions from steady-state levels. We set them to 0.0005 in both Home
and REA blocs ( the parameters regulating the adjustment cost on the internationally
traded private bond position paid by Home unrestricted households and RW
households are set to 0.0055).
Turning to the rest of the model parameters, Home represents a generic EA Member
State, relatively small (its GDP is 17% of EA GDP) and open (the import-to-GDP ratio
is 26%).23 The sizes of Home and REA GDPs as shares of world GDP are set to 3%
and to 15%. We deliberately keep the two regions relatively symmetric to get a
transparent transmission mechanism of the simulated shocks. We set some parameter
23 Details and tables reporting the calibration are in the working paper version (Burlon, Notarpietro, & Pisani, 2018).
24
values so that steady-state ratios (private consumption-, public consumption-, and
investment-to-GDP ratios) are consistent with EA national account data. For the
remaining parameters we resort to previous studies and estimates available in the
literature.
The short-term interest rate is equal to 0.24% on an annual basis. The spread between
short- and long-term bond is equal to 1.7 percentage points (pp). In each EA region the
share of restricted households is set to 0.10 and, thus, the share of unrestricted
households to 0.90. The (resulting) share of capital producers owned by restricted
households is set to 0.10, which is equal to the share of restricted households in the
population.
The chosen shares allow, in correspondence of a standard monetary policy shock and
taking the calibration of other parameters as given, to (i) obtain impulse responses in
line with estimates prevailing in the literature, especially with respect to the relative
response of investment and total consumption and (ii) align the responses of
consumption of restricted households with those of unrestricted households.24
Short-term public debt (ratio to yearly GDP) is set to 8% for both Home and REA.
Long-term public debt is set to 93.3% of GDP for both Home and REA. Thus, total
public debt as a share of GDP is 101% in both Home and REA. Such figure is slightly
larger than the average ratio for the EA, which has settled around 90% since 2012. The
duration of the long-term sovereign bond is equal to the average duration of EA long-
term sovereign bonds (8 years).
We assume that in each country long-term sovereign bond holdings are equally shared
24 The resulting share of consumption by unrestricted households is around 0.93 of total consumption and the share by restricted households is 0.07, in both EA countries.
25
between unrestricted and restricted households.
The central bank of the EA uses the policy rate to target the contemporaneous EA-wide
consumer price inflation (the corresponding parameter is set to 1.7) and the output
growth (the parameter is set to 0.1). The interest rate is set in an inertial way and hence
its previous-period value enters the rule with a weight equal to 0.87. The values are
identical for the corresponding parameters of the Taylor rule in the RW.
The chosen calibration yields impulse response functions to a standard monetary policy
shock (+0.25 basis points, bp) for GDP and inflation in each EA region that are in line
with those of the workhorse estimated models of the EA in the literature.25
3 Simulated scenariosWe consider four scenarios. They are all simulated under perfect foresight, i.e., there is
no uncertainty around the expected future values of the model variables.
In constructing the economic environment at the beginning of our simulations we
assume an initial gap of 0.1 pp between the inflation target and actual annualized
inflation. The value of 0.1 aims at representing a macroeconomic outlook characterized
by an inflation rate that is still not aligned with the target of the central bank in the
foreseeable future, at least not in absence of the monetary policy stimulus. In one
scenario (the fourth), we assume that an unexpected negative EA aggregate demand
shock affects the EA economy after the APP announcement. In all scenarios the central
25 See, for example, the models of the EA economy NAWM (Christoffel, Coenen, & Warne, 2008) and EAGLE (Gomes, Jacquinot, & Pisani, 2010). See also the model GEM (Pesenti 2008).
26
bank announcements are fully credible. In the first scenario, labeled “Full
announcement,” we assume that the central bank makes an announcement of future
purchases over a fully specified period of time and, in particular, it announces the end
date of the programme.26
We calibrate the size of the announced future purchases to mimic the October 2017
announcement by the ECB of future additional purchases between January and
September 2018. This consists of € 30 billion per month worth of additional purchases
for three quarters (from 2018Q1 to 2018Q3), with no additional purchases in the first
period after the announcement (2017Q4). Moreover, we assume that the FG keeps the
policy rate at the level prevailing before the announcement until the end of the
purchases, that is, for 4 quarters (until 2018Q3). Thereafter, the monetary policy rate
follows the Taylor rule (see Eq. 2).
In the second scenario, labeled “Open-ended announcement,” the ECB does not
announce an end date of the APP and, moreover, announces that it keeps the possibility
of further prolonging the purchases beyond 2018Q3. In this case, agents, as in the
previous scenario, fully believe the announced purchases (from 2018Q1 to 2018Q3).
Moreover, they know the central bank reaction function (see Eq. 1) and use it to form
expectations about additional unannounced future purchases on the basis of the
expected inflation gap. We assume, in a deliberately conservative way, that the reaction
function is active for two quarters (2018Q4 and 2019Q1). Expectations on the FG on
the policy rate are the same as in the first scenario. In the third scenario, labeled
“Open-ended announcement with prolonged FG,” we assume that the same conditions
26 For simplicity, in all simulations the decay of the outstanding amounts is expected to start immediately after the end of the purchases. Thus, we abstract from the effects of the announcement on the reinvestment policy which is an additional possible instrument in the APP toolbox.
27
of the second scenario hold but for the FG on the policy rate, which keeps the policy
rate unchanged also in two additional periods: 2018Q4 and 2019Q1. This assumption
on the FG is consistent with the current wording of ECB official communication,
according to which which the first rate hike would occur only “well past the horizon”
of net asset purchases (Draghi & Constancio, 2017).
In the fourth scenario, labeled “Negative demand shock,” we simulate a negative EA-
wide aggregate demand shock, which induces a reduction in EA consumption and
investment, calibrated to induce a drop in EA annualized in inflation of around 0.1pp in
the average of the first year of simulation, so to double the size of the inflation gap.
The shock hits the EA economy after the (full or open-ended) announcement.
4 Results
4.1 Full announcement
Figure 3 reports the quarterly response of key EA-wide variables in the case of full
announcement, i.e., when the central bank announces the amounts and pace of future
purchases and, crucially for the purposes of our analysis, the end date of the
programme.27 The announcement consists of € 30 billion per month worth of additional
Figure 3: Full announcement. EA variables.
27 In what follows we do not report the details by country because the impacts are rather symmetric. They are available upon request.
28
0 5 10 15 20 25 30 35 400
2
4CB holdings
0 5 10 15 20 25 30 35 400.22
0.24
0.26Policy rate
0 5 10 15 20 25 30 35 40-0.1
0
0.1Long term int. rate
0 5 10 15 20 25 30 35 40-0.05
0
0.05Exchange rate (+=appr.)
0 5 10 15 20 25 30 35 400
0.05
0.1GDP
0 5 10 15 20 25 30 35 40-0.05
0
0.05CPI inflation
Notes: Quarters on the horizontal axis. On the vertical axis, % deviations from the baseline for GDP and exchange
rate (positive values correspond to an appreciation of the euro); annualized pp deviations from the baseline for
inflation and long-term interest rate; levels for the policy rate; pp deviations from the baseline for the central bank
(CB) holdings measured as % of EA annual GDP.
purchases for three quarters (from 2018Q1 to 2018Q3), with no additional purchases in
the first period after the announcement (2017Q4). The overall announced purchases
amount to roughly 2.5% of annual EA GDP. Moreover, we assume FG, i.e., the central
bank announces to keep the policy rate at the level prevailing before the announcement
(the baseline, steady-state level in our model) until the end of the purchases (i.e., for 4
quarters, until 2018Q3). Thereafter, the monetary policy rate follows the Taylor rule.
As a result of the announced purchases, the interest rate on long-term sovereign bonds
decreases by 8bp on impact and, thereafter, gradually returns to its baseline level (the
size of the initial decrease is in line with the range of estimates in the existing literature
once we rescale the size of the shock). The policy rate is constant at its baseline
29
(steady-state) level during the first year, and increases (by a very small amount)
starting from the fifth quarter. The nominal exchange rate of the euro slightly
depreciates on impact, consistent with the uncovered interest parity condition, which
holds in the model, and the implemented expansionary monetary policy measures.
The lower long-term interest rate induces restricted households to increase
consumption and investment in physical capital (not reported to save on space). The
FG on short-term interest rates favors unrestricted households’ aggregate demand.
Following the higher domestic aggregate demand, EA GDP persistently increases
above its baseline level (by 0.06% at the peak). Similarly, the CPI inflation rate
increases (by 3bp at the peak).28
Overall, the announced measures are expansionary, as agents make consumption and
investment decisions on the basis of their expectations about the expansion of the
central bank's balance sheet and the duration of the FG. In other terms, agents’
expectations about the future path of interest rates fully reflects the (credible)
announcement of the central bank about the intended accommodative monetary policy
stance.
4.2 Open-ended announcement
In this scenario the central bank does not announce an end date of the APP. Instead, it
announces that it keeps the possibility of further prolonging the purchases beyond
2018Q3 conditionally on inflation developments. Agents, as in the scenario reported in
28 The dynamics of international trade (not reported to save on space) is characterized by an increase in imports, driven mostly by the relatively large increase in investment in physical capital, which has a large imported component (this effect on imports dominates the negative one associated with the depreciation of the exchange rate). Exports slightly increase in the medium run.
30
the previous section, fully believe the announced purchases (implemented from
2018Q1 to 2018Q3). Expectations on the FG are the same as in the first scenario (i.e.,
the monetary policy rate is kept constant during the initial 4 quarters, until 2018Q3,
and, thereafter, it follows the Taylor rule).
Crucially, and different from the “Full announcement” scenario, it is now assumed that
agents form their expectations about additional unannounced future purchases
according to the reaction function (Eq. 1), that links expected purchases to the inflation
gap. We assume, in a deliberately conservative way, that the function is active for two
quarters, 2018Q4 and 2019Q1 (we investigate the effects of the function being active
for more periods in the next sections).
Figure 4 reports the results and compare them to those of the “Full announcement”
scenario.
Households and firms now expect more purchases than in the “Full announcement”
scenario. The reason for the larger expected purchases relies on the reaction function.
Purchases announced by the central bank, that feed the model simulation, are not large
enough to close the initial inflation gap. Thus, agents expect additional future
purchases, in 2018Q4 and 2019Q1, to further reduce the inflation gap, even if they
were not announced. The endogenous profile of expected unannounced future
purchases proceeds beyond September 2018 at roughly the same pace as the one fully
announced up to that date (around € 30 billion per month).
Figure 4: Open-ended announcement. EA variables.
31
0 5 10 15 20 25 30 35 400
2
4CB holdings
0 5 10 15 20 25 30 35 400.2
0.25
0.3Policy rate
0 5 10 15 20 25 30 35 40-0.2
0
0.2Long term int. rate
0 5 10 15 20 25 30 35 40-0.1
0
0.1Exchange rate (+=appr.)
0 5 10 15 20 25 30 35 400
0.1
0.2GDP
0 5 10 15 20 25 30 35 40-0.1
0
0.1CPI inflation
Full announcementOpen-ended
Notes: Quarters on the horizontal axis. On the vertical axis, % deviations from the baseline for GDP and exchange
rate (positive values correspond to an appreciation of the euro); annualized pp deviations from the baseline for
inflation and long-term interest rate; levels for the policy rate; pp deviations from the baseline for the central bank
(CB) holdings measured as % of EA annual GDP.
Relative to the “Full announcement” scenario, the macroeconomic impact is larger,
because of the additional future (expected) purchases. Following the larger increase in
the expected central bank’s demand of sovereign bonds, long-term rates decrease by an
additional 6bp. Consistent with the more expansionary stance, the exchange rate
depreciates to a larger extent. The additional decrease in the long-term interest rates
favors a larger increase in aggregate demand for consumption and investment. EA
GDP reaches 0.1% above the baseline level by the end of 2018, and the inflation rate
closes half the initial gap, at 5bp above the baseline level by the end of 2018. The FG
effectiveness is also indirectly enhanced by the additional expected purchases, because
the short-term real interest rates decrease to a larger extent, consistent with the larger
32
increase in inflation and the expectations of constant nominal rates in the initial
quarters. As a result, the policy rate increases to a larger extent in the subsequent
quarters, consistent with the Taylor rule (Eq. 2) that has to stabilize a larger increase in
inflation and GDP. Overall, the additional expansionary effects of the open-ended
announcement are non-negligible, even if the additional purchases are relatively small
(around € 25 billion per month) and the reaction function is active for two quarters
only (six months).
The simulations lead to a conclusion that is relevant from a policy perspective: agents’
expectations can be relevant for the relative macroeconomic effectiveness of closed- vs
open-ended announcements. The latter are more expansionary than the former if agents
expect that the central bank will implement further measures if the ones announced are
not sufficient to achieve the target. In this respect, not only the central bank’s past
actions are important, but a crucial element is the central bank’s confirmation and
communication that the duration of the programme and inflation dynamics are
intertwined.
Moreover, as we show in the next section, the effects are larger if the reaction function
is active for more periods, implying not only more additional expected purchases but
also a longer duration of the FG on the monetary policy rate (so called “signaling”
effect of the purchase programme).
4.3 Open-ended announcement with prolonged FG
In this Section we evaluate the interaction between the open-ended and FG
announcements. In particular, we focus on announced references by the central bank to
33
continue to expect key policy rates to remain at their present levels “for an extended
period of time, and well past the horizon” of net asset purchases.29 In this case, agents
expect a longer FG to accompany the additional purchases commanded by the reaction
function.
For expected purchases, the same conditions of the open-ended scenario reported in the
previous section hold.
For the FG on the policy rate, the central bank announces that it will keep the policy
rate unchanged also in the two additional periods of expected purchases induced by the
open-ended announcement, that is, in 2018Q4 and 2019Q1.
Figure 5 shows the results. The expected extension of the FG induces lower real short-
term interest rates, which favor aggregate demand for consumption and investment in
physical capital by unrestricted households. There is a larger increase in GDP and its
components. As the announcement with the prolonged FG proves to be more
expansionary, the impact on inflation is larger (around 7bp). Thus, the inflation gap at
the onset of the open-ended programme is smaller and, consistently, the endogenous
reaction function-based determination of the additional future purchases by the central
bank leads to an expected profile of future holdings that is actually lower than the
profile that is expected in absence of a prolonged FG. The expected peak holdings
reach at most 3.5% of annual EA GDP instead of the 4%, which is reached when
agents do not associate an extension of the FG to the extension of the APP. Consistent
with the lower expected stock of central bank’s purchases, the long-term interest rates
decrease by a slightly lower amount.
29 See Draghi and Constancio (2017).
34
Figure 5: Open-ended announcement with prolonged FG. EA variables.
0 5 10 15 20 25 30 35 400
2
4CB holdings
0 5 10 15 20 25 30 35 400.2
0.25
0.3Policy rate
0 5 10 15 20 25 30 35 40-0.2
0
0.2Long term int. rate
0 5 10 15 20 25 30 35 40-0.2
0
0.2Exchange rate (+=appr.)
0 5 10 15 20 25 30 35 400
0.1
0.2GDP
0 5 10 15 20 25 30 35 40-0.1
0
0.1CPI inflation
Full announcementOpen-endedOpen-ended with FG (6 q)
Notes: Quarters on the horizontal axis. On the vertical axis, % deviations from the baseline for GDP and exchange
rate (positive values correspond to an appreciation of the euro); annualized pp deviations from the baseline for
inflation and long-term interest rate; levels for the policy rate; pp deviations from the baseline for the central bank
(CB) holdings measured as % of EA annual GDP.
We further investigate the quantitative effects of the open-ended programme by
assuming that the agents expect not only a longer FG, but also that unannounced
purchases will last for more periods than those considered in the previous simulations.
Specifically, it is assumed that agents expect three additional quarters of unannounced
purchases (instead of two as in previous simulations) and three additional quarters of
FG (seven quarters, from 2017Q4 to 2019Q2). We report the results in Figure 6.
Figure 6: Open-ended: additional periods of purchases and FG. EA variables.
35
0 5 10 15 20 25 30 35 400
2
4CB holdings
0 5 10 15 20 25 30 35 400.2
0.25
0.3Policy rate
0 5 10 15 20 25 30 35 40-0.2
0
0.2Long term int. rate
0 5 10 15 20 25 30 35 40-0.2
0
0.2Exchange rate (+=appr.)
0 5 10 15 20 25 30 35 400
0.1
0.2GDP
0 5 10 15 20 25 30 35 40-0.2
0
0.2CPI inflation
Open-ended with FG (6 q)Open-ended with FG (7 q)
Notes: Quarters on the horizontal axis. On the vertical axis, % deviations from the baseline for GDP and exchange
rate (positive values correspond to an appreciation of the euro); annualized pp deviations from the baseline for
inflation and long-term interest rate; levels for the policy rate; pp deviations from the baseline for the central bank
(CB) holdings measured as % of EA annual GDP.
Once we extend the open-ended periods from two to three, the front-loading of the
macroeconomic impact leads to an inflation rate that reaches 0.1pp above the baseline,
thus closing the initial gap in the first year. The higher resulting inflation rate implies
that the holdings in 2019Q1 are lower than in the case with only two open-ended
periods. GDP increases more than in the previous cases, as the larger expected
purchases and the longer expected duration of FG induce a further expansion in private
expenditure. Similarly, the macroeconomic effectiveness of FG is enhanced by the
additional expected purchases, because the implied higher (expected) inflation reduces
the short-term real interest rates, given that their (EA-wide) nominal counterpart is kept
constant.
36
Overall, the reported results suggest that the macroeconomic effectiveness of the open-
ended announcements can be enhanced if those announcements are able to
systematically and credibly link agents’ expectations on future purchases to
expectations on the duration of FG. Moreover, the effectiveness of FG in stimulating
the economy can also be enhanced by the expected future purchases, as long as the
latter are large enough to stimulate expected inflation in a non-negligible way.30
4.4 Negative demand shock
A negative demand shock can move inflation further away from the monetary policy
target. In what follows we compare the effectiveness of the full- and open-ended
announcement in counteracting the recessionary effects of a negative demand shock
that materializes immediately after the (alternative) announcements.31
We present the results of these simulations in Figure 7.
In the first simulation (labelled as “Shock,” dash-dotted black line in the charts), a
negative demand shock reduces EA GDP by almost 0.5% and the EA inflation rate by
0.1 annualized pp at the trough. We consider the following alternative scenarios on top
of the previous one: in the second (“Shock+full announcement,” solid black line),
Figure 7: Negative EA aggregate demand shock. EA variables.30 The size of additional expansionary effects in the case of prolonged FG is consistent with recent contributions on the macroeconomic effectiveness of FG. See Del Negro, Giannoni, & Patterson, (2012). In our model the decisions of restricted households are not directly affected by the short-term interest rates. Thus, it is likely that the effectiveness of FG is, in our model, lower than in the standard representative agent New Keynesian model.31 The negative demand shock is a combination of negative consumption preference and investment specific shocks in the two EA regions and in the RW that feed the model in the initial three quarters.
37
0 5 10 15 20 25 30 35 40-5
0
5CB holdings
0 5 10 15 20 25 30 35 400.1
0.2
0.3Policy rate
0 5 10 15 20 25 30 35 40-0.2
0
0.2Long term int. rate
0 5 10 15 20 25 30 35 40-0.02
0
0.02Exchange rate (+=appr.)
0 5 10 15 20 25 30 35 40-0.5
0
0.5GDP
0 5 10 15 20 25 30 35 40-0.2
0
0.2CPI inflation
ShockShock + Full announcementShock + Open-ended (2 q)Shock + Open-ended (3 q)
Notes: Quarters on the horizontal axis. On the vertical axis, % deviations from the baseline for GDP and exchange
rate (positive values correspond to an appreciation of the euro); annualized pp deviations from the baseline for
inflation and long-term interest rate; levels for the policy rate; pp deviations from the baseline for the central bank
(CB) holdings measured as % of EA annual GDP.
full announcement of the additional APP purchases until September 2018; in the third,
open-ended announcement with prolonged FG until 2019Q1 (“Shock+open-ended
(2q),” dashed red line) and, finally, in the fourth, open-ended announcement with
prolonged FG until 2019Q3 (“Shock+open-ended (4q),” green line).
All announcements help contain the worsening of macroeconomic conditions and, at
the same time, limit the fall in the short-term policy rate.
The open-ended announcement with FG until 2019Q1 is more accommodative than the
full announcement, as agents expect higher purchases also in the two quarters of the
open-ended leg of the programme. Specifically, the negative effects on inflation of the
negative demand shock widen the initial inflation gap. Given the reaction function,
38
agents expect the central bank to further increase the bond purchases beyond the
announced ones. Thus, long-term interest rates decrease to a larger extent, further
stimulating economic conditions. The improvements in inflation and economic activity
(i.e., the faster economic recovery) imply a steeper subsequent rise in the short-term
monetary policy rate.32
The macroeconomic effectiveness of the open-ended announcement with FG in
counterbalancing the recessionary shock is further enhanced if agents expect the
additional purchases and FG to last for longer.
Overall, our results clearly show that an open-ended programme is more effective than
a full-announcement in stabilizing macroeconomic conditions in the face of unexpected
shocks that drive inflation further below the target. This suggests that an open-ended
announcement is particularly desirable if economic conditions are uncertain, because,
as long as that type of communication induces agents to believe in a reaction of the
central bank, it helps stabilize expectations about future macroeconomic outcomes.
5 ConclusionWe have used a large-scale New Keynesian DSGE model for the EA to evaluate the
benefits of an open-ended APP when the inflation rate is not yet aligned with the
monetary policy target over the medium term. We have showed that an open-ended
APP helps maintain an accommodative monetary policy stance. Eschewing a full
announcement of the future path of purchases increases the effectiveness of the 32 We do not consider the zero lower bound (ZLB) constraint on the monetary policy rate, but the steeper rise in the policy rate that we obtain under the open-ended programme suggests that the latter should favor an earlier exit from the ZLB. This further corroborates the greater effectiveness of the open-ended programme in stimulating macroeconomic conditions. We leave the analysis of the ZLB for future research.
39
committed future purchases by managing expectations over unannounced future
purchases through appropriate and systematic communication. Moreover, the
effectiveness increases even further if the central bank communication also induces a
correlation in agents’ expectations between the continuation of net future purchases
and that of the FG on policy rates. Finally, an open-ended APP may prove all the more
adequate if additional, unforeseen negative shocks stave off even further the
achievement of the inflation target over the medium term.
From a policy perspective, our paper emphasizes the role of communication and its
impact on agents’ expectations. Communication, by allowing the central bank to
manage expectations of economic agents, becomes particularly relevant when
macroeconomic conditions and the monetary policy tools are not the standard ones and,
thus, are not well known by economic agents and/or their functioning is very difficult
to explain to the public. An appropriate communication of the non-standard monetary
policy measures should ensure consistency over time, clarify the state contingency of
the policy stance, and provide clarity on the use of specific tools in a multi-instrument
context. In this way, the central bank can reduce uncertainty and, thus, shape
expectations in such a way to maximize the effectiveness of the policy measures
themselves.
From a more theoretical perspective, in this paper we have not assessed the role of
announcing a reinvestment policy, which in principle could enhance the effectiveness
of the open-ended APP, as long as agents anticipate a larger size of the central bank
balance sheet. We leave this issue for future research.
40
ReferencesAltavilla, C., Carboni, G., & Motto, R. (2015). Asset purchase programmes and
financial markets: lessons from the euro area, Working Paper Series 1864,
European Central Bank November.
Amamiya, M. (2017). History and Theories of Yield Curve Control. Keynote Speech at
the Financial Markets Panel Conference to Commemorate the 40th Meeting of the
Bank of Japan, Bank of Japan, January.
Andres, J., Lopez-Salido, J. D., & Nelson, E. (2004). Tobin's imperfect asset
substitution in optimizing general equilibrium. Journal of Money, Credit and
Banking, 36 (4), 665-690.
Blattner, T., & Joyce, M. A. S., (2016). Net debt supply shocks in the euro area and the
implications for QE, Working Paper Series 1957, European Central Bank
September.
Breckenfelder, J., De Fiore, F., Andrade, P., Karadi, P., & Tristani, O. (2016). The
ECB's asset purchase programme: an early assessment. Working Paper Series
1956, European Central Bank, September.
Burlon, L., Gerali, A., A. Notarpietro, & Pisani, M. (2017). Macroeconomic
effectiveness of non-standard monetary policy and early exit. A model-based
evaluation. International Finance, 20 (2), 155‒173.
Burlon, L., Notarpietro, A., & Pisani, M. (2018). Macroeconomic effects of an open-
ended Asset Purchase Programme. Temi di discussione (Economic working papers)
1185, Bank of Italy, Economic Research and International Relations Area.
41
Chen, H., Curdia, V. & Ferrero, A. (2012). The macroeconomic effects of large-scale
asset purchase programmes. Economic Journal, November 2012, 122 (564), F289‒
F315.
Christoffel, K., Coenen, G., & Warne, A. (2008). The New Area-Wide Model of the
euro area: a micro-founded open-economy model for forecasting and policy
analysis. Working Paper 0944, European Central Bank, October.
Coenen, G., Ehrmann, M., Gaballo, G., Hoffmann, P., Nakov, A., Nardelli, S., Persson,
E., & Strasser G. (2017). Communication of monetary policy in unconventional
times. Working Paper 2080, European Central Bank, June.
Darracq Paries, M., & Kuhl, M. (2016). The optimal conduct of central bank asset
purchases. Working Paper 1973, European Central Bank, November.
Del Negro, M., Giannoni, M., & Patterson, C. (2012), The Forward Guidance Puzzle,
Staff Reports, Federal Reserve Bank of New York.
Draghi, M. (2014). Monetary policy communication in turbulent times. Speech at the
Conference “De Nederlandsche Bank 200 years: Central banking in the next two
decades,” Amsterdam, 24 April 2014.
Draghi, M. (2018). Monetary Policy in the Euro Area. Speech at the Conference “The
ECB and Its Watchers” XIX Conference, organized by the Institute for Monetary
and Financial Stability, Frankfurt, 14 March 2018.
Draghi, M., & Constancio, V. (2017). Introductory statement. Press conference, ECB,
Frankfurt am Main, 26 October 2017.
Ellison, M. & Tischbirek, A. (2014). Unconventional government debt purchases as a
supplement to conventional monetary policy. Journal of Economic Dynamics and
42
Control, 43 (C), 199‒217.
Ferdinandusse, M, Freier, M., & Ristiniemi, A. (2017). Quantitative easing and the
price-liquidity trade-off, Working Paper 335, Sveriges Riksbank.
Gomes, S., Jacquinot, P., & Pisani, M. (2010) The EAGLE. A model for policy analysis
of macroeconomic interdependence in the euro area. Working Paper 1195,
European Central Bank May.
Harrison, R. (2012), Asset purchase policy at the effective lower bound for interest
rates. Bank of England working paper 444, Bank of England January.
Jones, C. & Kulish, M. (2013) Long-term interest rates, risk premia and
unconventional monetary policy. Journal of Economic Dynamics and Control, 37
(12), 2547-2561.
Modigliani, F. & Sutch, R. (1966). Innovations in interest-rate policy. American
Economic Review, March, 56 (1/2), 178‒197.
Pesenti, P. (2008), The Global Economy Model: Theoretical Framework. IMF Staff
Papers, June, 55 (2), 243‒284.
Praet, P. (2017). Communicating the complexity of unconventional monetary policy in
EMU. Speech at “The 2017 ECB Central Bank Communications Conference”,
Frankfurt am Main, 15 November 2017.
Praet, P. (2018a). Maintaining price stability with unconventional monetary policy.
Speech at the Council of the European Union, Brussels, 29 January 2018.
Praet, P. (2018b). Assessment of quantitative easing and challenges of policy
normalization. Speech at “The ECB and Its Watchers XIX Conference”, Frankfurt
am Main, 14 March 2018.
43
Reis, R. (2017). QE in the Future: The central bank's balance sheet in a fiscal crisis.
IMF Economic Review, April 2017, 65 (1), 71‒112.
Rotemberg, J. J. (1982). Monopolistic price adjustment and aggregate output. Review
of Economic Studies, October, 49 (4), 517‒531.
Skolimowski, P. (2017). This Magic Formula Reveals How the ECB Might Taper.
Online article on Bloomberg.com, 17 July 2017 July 2017.
Wallace, N. (1981). A Modigliani-Miller theorem for open-market operations.
American Economic Review, June, 71 (3), 267‒274.
Woodford, M. (2001). Fiscal requirements for price stability. Journal of Money, Credit
and Banking, August 2001, 33 (3), 669‒728.
Woodford, M. (2016). Quantitative easing and financial stability. In Albagli, E.,
Saravia, D., & Woodford, M. (eds.), Monetary Policy through Asset Markets:
Lessons from Unconventional Measures and Implications for an Integrated World,
Vol. 24 of Central Banking, Analysis, and Economic Policies Book Series, Central
Bank of Chile, 2016, chapter 6, 151‒233.