Inflation Versus Price-Level Targeting - Bayesian Estimation of A Small Open DSGE Model For Switzerland-Gabler Verlag (2015)

BestMasters
Springer awards “BestMasters” to the best application-oriented master’s theses,

which were completed at renowned chairs of economic sciences in Germany, Aus-
tria, and Switzerland in 2014.
The works received highest marks and were recommended for publication by
supervisors. As a rule, they show a high degree of application orientation and deal
with current issues from different fields of natural sciences, psychology, engineer-
ing, and economics.
The series addresses practitioners as well as scientists and offers guidance for early
stage researchers.
Lukas Heim
Inflation versus
Price-Level Targeting
Bayesian Estimation of a Small
Open DSGE Model for Switzerland
With a foreword by Prof. Dr. Luca Benati
Lukas Heim
Luzern, Switzerland
BestMasters
ISBN 978-3-658-08227-7 ISBN 978-3-658-08228-4 (eBook)
DOI 10.1007/978-3-658-08228-4
Library of Congress Control Number: 2014956973
Springer Gabler
© Springer Fachmedien Wiesbaden 2015
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or
part of the material is concerned, specifically the rights of translation, reprinting, reuse of illus-
trations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and
transmission or information storage and retrieval, electronic adaptation, computer software, or by
similar or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are
exempt from the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, express or implied, with respect to the material contained
herein or for any errors or omissions that may have been made.
Printed on acid-free paper
Springer Gabler is a brand of Springer Fachmedien Wiesbaden

Springer Fachmedien Wiesbaden is part of Springer Science+Business Media
(www.springer.com)
For Mami and Paps.
Foreword
Following the recent financial crisis, academics and policymakers have
been engaged in a wide-ranging reappraisal of the pros and cons of alter-
native monetary regimes and monetary policy instruments. The crisis
highlighted two key issues which, before 2007-2008, had received scant
attention: first, the role played by asset price bubbles in macroeconomic
fluctuations; second, the fact that the probability of hitting the zero
lower bound (ZLB) is significantly greater than previously thought.
As for the former issue, the professionals’ response has been to aug-
ment the toolkit of monetary policy instruments with so-called ’macro-
prudential policies’ (e.g., limits to the size of a real estate loan compared
to the price of the house which is being bought, or the income of the
borrower). By surgically targeting disequilibria in specific asset mar-
kets’ segments, these (yet untested) instruments promise, in principle,
to take away the froth from that particular segment without the dele-
terious, economy-wide recessionary side-effects which would result from
an interest rate hike.
As for the ZLB, research has focused on several alternative ‘lines of at-
tack’. In particular, on the one hand, there has been an intensive research
effort aimed at identifying which policies (quantitative easing, ‘forward
guidance’, etc.) can jolt the economy out of the recession once the ZLB
has been reached, so that standard monetary policy has all but lost its
effectiveness. A second group of contributions has instead explored al-
ternative ways of ‘hardwiring’ into the system a set of self-stabilizing
expectations, in order to make the economy more resilient in the face of
shocks which push it to the ZLB, and to make it easier, for policymakers,
to push it out of that state once it has been reached. Under this respect,
the fundamental theoretical contributions of Krugman (1998), and of
VIII Foreword
Eggertson and Woodford (2003), have highlighted how a comparatively

simple (at least, in principle) solution to the problem posed by the ZLB
would be to create inflationary expectations in that state of the world.
The rationale behind this proposal is straightforward: given that, by def-
inition, at the ZLB the short-term interest rate is stuck at zero - so that,
different from the ‘normal state of affairs’, it does not react to macroe-
conomic developments (in particular, it does not react to fluctuations
in either inflation or inflationary expectations) - any change in expected
inflation maps, one-for-one, into a corresponding change in the opposite
direction in the real ex ante interest rate. This automatically implies
that an increase in expected inflation causes a corresponding decrease
in the real ex ante interest rate, thus stimulating demand and pushing
the economy out of the recession. Eggertson and Woodford (2003) have
shown that the optimal policy, which takes an extremely convoluted and
non-transparent form, can be well-approximated, to a quite remarkable
extent, by a simple price-level targeting rule. The reason for this is, once
again, straightforward: if the central bank commits to keeping the price
level constant, this logically implies that deflation today ought to be
followed by inflation tomorrow, simply because any price decline has to
be followed by a corresponding and identical price increase, in order to
bring prices to their original level. This implies that - if the price level
targeting policy is credible - the deflation associated with the recession-
ary shock which pushed the economy down to the ZLB automatically
creates inflationary expectations, thus introducing into the economy an
automatic stabilizer of aggregate demand. As discussed by Eggertson
and Woodford (2003), a simple price-level targeting rule presents two
fundamental advantages compared to the optimal policy. First, as pre-
viously mentioned, it is much easier in terms of formulation, and it can
therefore be easily communicated to the public. Second, because of this,
it is comparatively easy, for a central bank, to build credibility about
such a policy: quite simply, it has to stick to this policy ‘through thick
and thin’ - that is, in all states of the world, including when the ZLB is
not binding. In this way the public regularly observes the policy being
implemented, and the central bank builds up a reputation for following
such a policy.
Foreword IX
A limitation of this literature, as it currently stands, is that the over-

whelming majority of these contributions have been of a theoretical na-
ture, and have been based on calibrated simple, purely forward-looking
New Keynesian (DSGE) models. The fact that these models have been
known, for about two decades now, to provide an extremely poor fit
to real-world macroeconomic data represents a significant obstacle to
the possible widespread adoption of price-level targeting rules. Quite
obviously, before switching to an alternative monetary policy rule, poli-
cymakers want to have a reasonably precise idea of how the economy is
going to perform conditional on that rule. In particular, in the spirit of
the ‘robustness’ approach to policy, they want to be reassured that the
rule will not create ‘tail risks’, or extreme reactions to specific shocks or
circumstances. Since DSGE models’ behavior depends, to a sometimes
crucial extent, on the specific values taken by the model’s structural
parameters. It logically follows that only a comparative assessment of
the performance of a price-level targeting rule based on an estimated (as
opposed to calibrated), and relatively sophisticated DSGE model might
convince policymakers of the wisdom of adopting such a policy.
Lukas Heim’s Master’s thesis ought to be assessed against this general
background. The thesis is one of the very few existing attempts to eval-
uate the performance of a price-level targeting rule, compared to that
of a standard Taylor-type rule, based on a medium-scale DSGE model
which has been estimated based on state-of-the-art Bayesian methods.
Specifically, Lukas Heim first builds such a model for the Swiss economy
by expanding the framework proposed by Galı́ and Monacelli (2005)
and Monacelli (2005), enriching it with habit formation in consumption,
price indexation, labor market imperfections, and several additional
structural disturbances. Within this framework, the Euro area plays
the role of the large foreign economy. The model is then estimated via
standard Bayesian techniques (specifically, Random-Walk Metropolis).
Estimation is performed conditional on a standard Taylor-type rule for
both Switzerland and the Euro area. Based on the estimated model,
Lukas Heim then explores, for Switzerland, the performance of a price-
X Foreword
level targeting rule compared to that resulting from the Taylor-type rule
used in estimation. A key finding emerging from the impulse response
functions to the structural disturbances is that whereas the volatility of
the nominal interest rate is very similar conditional on the two rules; and
- exactly as expected - the volatility of inflation is quite significantly lower
under the price-level targeting rule; the volatility of the output gap is
markedly higher conditional on either productivity or preference shocks.
This is a very important result, because it suggests that in the real world
- as opposed to the simple, purely forward-looking New Keynesian model
- the adoption of a price level-targeting rule would likely produce an in-
crease in the volatility of real economic activity ‘across the board’ - that
is, conditional on both supply-side and demandside shocks. In turn, this
implies that a policymaker who assigns a non-negligible weight to the
volatility of real economic activity in her/his loss function would proba-
bly have some reservations about adopting a price-level targeting rule for
Switzerland. Under this respect, Lukas Heim’s work therefore implicitly
provides support to the monetary policy regime currently implemented
by the Swiss National Bank.
Bern, in July 2014 Prof. Dr. Luca Benati

Contents
1 Introduction 1
2 Description of the Model 5

2.1 Domestic Economy . . . . . . . . . . . . . . . . . . . . . . 6
2.1.1 Households . . . . . . . . . . . . . . . . . . . . . . 6
2.1.2 Producers . . . . . . . . . . . . . . . . . . . . . . . 12
2.1.3 Retail Firms . . . . . . . . . . . . . . . . . . . . . 15
2.1.4 Exogenous Processes . . . . . . . . . . . . . . . . . 16
2.2 Foreign Economy . . . . . . . . . . . . . . . . . . . . . . . 16
2.2.1 Households . . . . . . . . . . . . . . . . . . . . . . 17
2.2.2 Producers . . . . . . . . . . . . . . . . . . . . . . . 21
2.2.3 Monetary Authority . . . . . . . . . . . . . . . . . 24
2.2.4 Exogenous Processes . . . . . . . . . . . . . . . . . 25
2.3 General Equilibrium and Definitions . . . . . . . . . . . . 25
2.3.1 General Equilibrium . . . . . . . . . . . . . . . . . 25
2.3.2 Various Definitions: . . . . . . . . . . . . . . . . . 26
2.4 Domestic Monetary Policy Rule . . . . . . . . . . . . . . . 28
2.4.1 Inflation Targeting Rule . . . . . . . . . . . . . . . 28
2.4.2 Price-Level Targeting Rule . . . . . . . . . . . . . 28
3 Bayesian Estimation of the Parameters 31

3.1 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.2 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.3 Prior Distribution . . . . . . . . . . . . . . . . . . . . . . 37
3.4 Estimation Results . . . . . . . . . . . . . . . . . . . . . . 40
3.5 Historical Simulation . . . . . . . . . . . . . . . . . . . . . 44
XII Contents
4 Results - Inflation versus Price-Level Targeting 47

4.1 Impulse Response Functions . . . . . . . . . . . . . . . . . 47
4.2 Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
5 Conclusion 55
References 57
Appendix 61
1 Log-Linearized Equations . . . . . . . . . . . . . . . . . . 61
2 Shape of the Log-Likelihood Functions . . . . . . . . . . . 65
1 Introduction
Historically, Sweden is the only country that has ever followed an explicit
price-level regime. As discussed in Berg and Jonung (1999), the regime
dates back to the great depression in the 1930s. After this Swedish expe-
rience, the price-level targeting was never applied again. Actually, there
are many monetary authorities following an implicit or explicit inflation
targeting regime.1 The inflation targeting regime publicly announces
an inflation target that it will reach in the medium-term horizon of
one to three years (see Bernanke and Woodford (2005)). The monetary
authority usually tries to reach the announced goal by controlling the
short-term interest rate. Under the actual circumstances with very low
interest rates, the monetary policy regime is limited to the zero lower
bound on nominal interest rates. In this case, a change to a price-level
targeting regime can help to avoid the liquidity trap, as suggested by
Svensson (2001). The two policies react differently to a positive shock in
the price-level: The monetary authority under inflation targeting tries
to stabilize inflation at the new price-level (with base drift), whereas
the price-level regime tries to push the economy back to the old price-
level (no base drift). Under the assumption of credibility, this additional
push impacts the investor’s beliefs and can help to avoid or even escape
a liquidity trap.
The actual environment with low interest rates is the reason why the
discussion about advantages and disadvantages of a price-level targeting
regime became more active in the recent years. Svensson (1999) showed
on a theoretical basis, that price-level targeting lowers the short-run vari-
ability of inflation. Later on, Vestin (2006) examines the topic in the
standard New Keynesian model using loss functions for the two policies.
1
E.g., Canada, England, New Zealand, Sweden, and Switzerland.
L. Heim, Inflation versus Price-Level Targeting, BestMasters,

DOI 10.1007/978-3-658-08228-4_1, © Springer Fachmedien Wiesbaden 2015
2 1 Introduction
His findings show that price-level targeting outperforms inflation target-

ing in an environment characterized by discretion. In the recent years,
the Bank of Canada published various writings on price-level targeting,
for example Ambler (2009).
This thesis will take a slightly different but complementary approach.
There are two main differences to be noted: The price-level targeting
regime is modeled in a small open New Keynesian economy. Therefore
the effects of a change to a price-level regime for the foreign economy can
be examined. In addition, the comparison of the two policies is based
on a stochastic simulation rather than theoretical calculations with loss
functions. Based on the simulation the variances of different variables
under the inflation targeting and the price-level targeting policy are
compared.
To derive empirically relevant parameters, the model is estimated us-
ing Bayesian methodology. The small open economy is modeled with
Swiss data whereas the large economy is based on data from the Euro
area. The data base contains the following variables for both economies:
output, inflation and interest rate. Prior assumptions about the distri-
bution of the estimated parameters are based on Robinson (2013). The
model is mainly built on the standard small open economy suggested
by Galı́ and Monacelli (2005) and Monacelli (2005) with various mod-
ifications. Price and wage stickiness are introduced using the Calvo
(1983) approach with indexation to previous inflation rates. The house-
hold’s preferences are modeled with an external habit formation term,
as it was proposed by Fuhrer (2000). Additionally, numerous stochastic
shocks are implemented: preference shock, labor supply shock, mone-
tary policy shock, among others. These modifications are important for
the empirical fitting in closed economies, as presented by Smets and
Wouters (2007). An open economy model with analogous rigidities is
discussed in Justiniano and Preston (2008).
Firstly, the estimated model is used to calculate the impulse response
functions to a domestic technology, preference and monetary policy
shock. The response of the inflation rate to all shocks are more ac-
curate under price-level targeting than under inflation targeting. Using
1 Introduction 3
a price-level targeting rule, output reacts stronger to technological and

preference changes. The monetary policy shock is followed by larger
changes in output under inflation targeting than under price-level tar-
geting. Therefore, the statement about the changes in output is am-
biguous. Secondly, the variables are simulated based on the estimated
model. The variances of the simulated inflation rates are much lower
using the price-level targeting rule. Under the inflation targeting pol-
icy the variance of the output is a little lower than under price-level
targeting. Since inflation rates and output are simultaneously targeted,
neither of the two policies can be clearly preferred.
The structure of the thesis looks as follows: Chapter 2 derives the
standard small open DSGE model. Chapter 3 describes the data and
the methodology of the Bayesian estimation. Chapter 4 discusses two
exercises of comparison between inflation and price-level targeting: cal-
culating impulse response functions and running a stochastic simulation.
Chapter 5 summarizes the findings and provides an outlook for further
studies.
2 Description of the Model
The following model mainly follows Galı́ and Monacelli (2005) and Mona-
celli (2005). It is the standard DSGE model of an open economy with
microfoundations. The model is extended by external habit formation,
labor market imperfections, incomplete asset markets, indexation of
prices and various stochastic shocks. The habit formation is based on
an argument called ’keeping up with the Joneses’, which describes social
comparisons in consumption choices. The concept was introduced to the
monetary policy literature by Fuhrer (2000).1 In the goods and in the
labor market, prices and wages are set according to a price-setting à la
Calvo (1983) where non-optimized prices and wages are adapted to past
inflation as suggested by Christiano et al. (2005). Compared to Galı́
and Monacelli (2005) and Monacelli (2005), the modified set-up contains
preference, labor supply and cost-push shocks. Shocks are directly in-
troduced in the following derivations. A small open DSGE model with
analogous modifications was given by Justiniano and Preston (2008). A
complete set of the model’s equations is provided in chapter 1 of the
appendix.
The model consists of two economies: a domestic and a foreign econ-
omy. The domestic economy is assumed to be small compared to the
foreign economy. Variables describing the foreign economy are indicated
with the superscript ’*’.
The monetary policy rule of the domestic economy is distinguished
in two cases: inflation targeting and price-level targeting. Because of
their importance to the topic, the two rules are discussed in a separate
chapter, 2.4.
1
The history of the concept of habit formation is long. For example, in the finance
literature it was applied by Constantinides (1990).

6 2 Description of the Model
2.1 Domestic Economy

The domestic economy consists of households, producers, retail firms
and a monetary authority. The following sections derive the optimal
behavior of each of these agents.
2.1.1 Households
The domestic economy has access to foreign goods through imports and
to foreign bonds through financial markets. The domestic households
undertake final consumption of domestic and foreign goods. They pro-
vide working hours as a production factor to domestic producers. The
optimal mixture of consumption and work is derived by maximizing the
household’s preferences. There exists a continuum of households in the
domestic economy. However, for the goal of deriving the optimality con-
ditions, the focus stays on a single representative household. The derived
optimality conditions hold analogously for the whole continuum.
Utility
The domestic household chooses consumption (Ct ), labor input (Nt ,
working hours), domestic bonds (Dt ) and foreign bonds (Bt ) such as
to maximize its utility function.

∞ 1

(Ct+s − Ht+s )1− σ n,t (Nt+s )1+ϕ
max E0
s
β g,t 1
− (2.1)
{Ct ,Nt ,Dt ,Bt }∞ 1− σ
1+ϕ
t=0
s=0
Ht describes the habit formation. It is not regarded as a choice vari-

able, since the habit formation is assumed to be external. Ht = hCt−1
holds in all periods with 0 < h < 1. g,t and n,t denote the preference
and the labor supply shock. The parameters σ −1 and ϕ are assumed to
be positive: σ −1 > 0 and ϕ > 0. β is the discount factor.
The maximization is limited by the following budget constraint that
has to be satisfied in each period:
2.1 Domestic Economy 7
∗
Pt Ct + Dt + Xt Bt =Rt−1 Dt−1 + Xt−1 Rt−1 φt (At )Bt−1 + Wt Nt + Tt
with φt (At ) = exp [−χ(At + rp,t )]

St−1 Bt−1
and At = (2.2)
C̄F Pt−1
The price index (Pt ) corresponds to the domestic consumer price index
(CPI). Wt is the domestic wage level. Rt and Rt∗ are the domestic and
the foreign interest rate, respectively. Tt is assumed to be a lump-sum
transfer including profits of the firms as well as taxes. Xt is the nominal
exchange rate. Following Justiniano and Preston (2008) and Kollmann
(2002), the function φ(.) is a debt elastic interest rate and At is the
real quantity of outstanding foreign debt in terms of domestic currency.
rp,t denotes the risk premium shock. Finally, C̄F is the steady state
consumption of imported goods. The debt elastic interest rate is crucial
for the stationarity of foreign debt in the log-linearized model.
The described maximization problem can be solved using the La-
grangian method with a multiplier λt .

∞ 1

s (Ct+s − Ht+s )1− σ n,t (Nt+s )1+ϕ
L = E0 β g,t [ 1
−
1− σ
1+ϕ
s=0
− λt+s (Pt+s Ct+s + Dt+s+1 + Xt+s Bt+s − Rt+s−1 Dt+s−1

∗
− Xt+s Rt+s−1 φt+s Bt+s−1 − Wt+s Nt+s − Tt+s )] (2.3)
Taking the derivatives with respect to Ct , Nt , Bt and Dt yields to the

following optimality conditions.
∂L −1
: g,t (Ct − Ht ) σ − λt Pt = 0 with Ht = hCt−1 (2.4)
∂Ct
∂L ϕ
: − n,t Nt + λt Wt = 0 (2.5)
∂Nt
∂L
: − λt + βRt Et [λt+1 ] = 0 (2.6)
∂Bt
∂L ∗
: − λt Xt + βRt Et [λt+1 Xt+1 φt+1 ] = 0 (2.7)
∂Dt
Furthermore, the budget constraint (derivative with respect to λt )

and the transversality condition have to be satisfied.
(2.4) can be used to substitute for the multiplier λt in the conditions
(2.5) to (2.7).
1/σ Wt
ϕ
n,t Nt (Ct − hCt−1 ) = (2.8)
Pt

−1/σ Pt −1/σ
g,t (Ct − hCt−1 ) = βRt Et g,t+1 (Ct+1 − hCt ) (2.9)
Pt+1

−1/σ ∗ Xt+1 Pt −1/σ

g,t (Ct − hCt−1 ) = βRt Et φt+1 g,t+1 (Ct+1 − hCt ) (2.10)
Xt Pt+1
Consumption Allocation
Maximizing its utility function, the household has chosen an optimal
consumption bundle Ct . The household’s consumption bundle is a com-
bination of a domestic consumption bundle (CH,t ) and foreign consump-
tion bundle (CF,t ).
η−1 η−1
η−1
η
1 1
Ct = (1 − α) η η
CH,t + αη η
CF,t (2.11)
The parameter η is assumed to be positive: η > 0. α indicates the

openness of the domestic economy and lies between 0 and 1: 0 < α < 1.
Furthermore, the domestic and foreign consumption bundles are them-
selves combinations of domestic produced goods, CH,t (i), and foreign
produced goods, CF,t (i), according to a Dixit-Stiglitz aggregator.
1
θ−1
θ
θ−1
CH,t = CH,t (i) θ di (2.12)
0
1
θ−1
θ
θ−1
CF,t = CF,t (i) θ di (2.13)
0
θ is the elasticity of substitution and assumed to be positive: θ > 0.

The household allocates the given expenditures for consumption (Pt Ct )
from the utility maximization regarding the aggregation index for Ct .
min Pt Ct − PH,t CH,t − PF,t CF,t

{CH,t ,CF,t }
η−1 η−1
η−1
η
1 1
η
s.t. Ct = (1 − α) η CH,t η
+ α η CF,t (2.14)
Replacing Ct by its aggregation index, it becomes an unconstrained

maximization problem.
η−1 η−1
η−1
η
1 1
L = Pt (1 − α) η η
CH,t + αη η
CF,t − PH,t CH,t − PF,t CF,t (2.15)
The corresponding optimality conditions are given by equations (2.16)

and (2.17).
η−1
η
1
−1
η−1 η−1 η
∂L 1 1 1
: Pt (1 − α) η η
CH,t + αη η
CF,t η
(1 − α) η CH,t − PH,t = 0 (2.16)
∂CH,t
η−1
η
1
−1
η−1 η−1 η
∂L 1 1 1
: Pt η
(1 − α) η CH,t η
+ α η CF,t α η CF,tη − PF,t = 0 (2.17)
∂CF,t
Solving for CH,t and CF,t delivers the following demand functions.
−η
PH,t
CH,t = (1 − α) Ct (2.18)
Pt
−η
PF,t
CF,t = α Ct (2.19)
Pt
Secondly, the same optimization procedure is followed for domestic

goods (CH,t ) and foreign goods (CF,t ). The households allocate their
expenditures for domestic and foreign goods across a continuum of do-
mestic and foreign produced goods.
min Pj,t Cj,t − Pj,t (i)Cj,t (i) di

{Cj,t (i)}
0
1
θ−1
θ
θ−1
s.t. Cj,t = Cj,t (i) θ di with j = {H, F } (2.20)
0
The Lagrangian for the unconstrained problem is given by equation

(2.21).
1
θ−1
θ 1
θ−1
L = Pj,t Cj,t (i) θ di − Pj,t (i)Cj,t (i) di with j = {H, F } (2.21)
0 0
Taking the derivative with respect to Cj,t (i) delivers the following first
order condition (FOC).
1
θ−1
θ
1
θ
∂L −1
θ−1
: Pj,t Cj,t (i) θ di Cj,t (i) θ − Pj,t (i) = 0
∂Cj,t (i)
0
with j = {H, F } (2.22)
Solving for CH,t (i) and CF,t (i) delivers the two demand functions
(2.23) and (2.24).
−θ
PH,t (i)
CH,t (i) = CH,t (2.23)
PH,t
−θ
PF,t (i)
CF,t (i) = CF,t (2.24)
PF,t
Price-Level
The following expression holds because of the price-taking and non-satiation
argument: Pt Ct = PH,t CH,t + PF,t CF,t . Using the demand functions
(2.18) and (2.19) to replace CH,t and CF,t gives the following definition
of the price-level:
1−η 1−η
1−η
1
Pt = (1 − α) PH,t + αPF,t (2.25)
Using the same methodology, the price-level for domestic goods (PH,t )
and foreign goods (PF,t ) can be calculated. In this case, the price-taker
1
and non-satiation assumption yields: PH,t CH,t = 0 PH,t (i)CH,t (i) di
1
and PF,t CF,t = 0 PF,t (i)CF,t (i) di. Replacing CH,t (i) and CF,t (i) by
its demand functions, (2.23) and (2.24), defines the price-levels PH,t and
PF,t .
1
1−θ
1
1−θ
PH,t = PH,t (i) (2.26)
0
1
1−θ
1
1−θ
PF,t = PF,t (i) (2.27)
0
Labor Supply
Finally, the households decide how many working hours they provide
to the domestic producers. There is a single domestic labor market
where domestic producers hire labor inputs at a common wage. For
the derivation of the optimal labor supply, the representative household
makes use of the producer’s labor demand function (see chapter 2.1.2).
The households are assumed to have monopolistic power in the labor
market. They are allowed to re-optimize their wage in a given period
with a probability of 1 − ξw where 0 < ξw < 1. On the contrary, the
probability of staying with the same wage as in the previous period is
simply ξw . If this is the case, the wage is adjusted according to the
following indexation rule:
log Wt (k) = log Wt−1 (k) + γw πt−1 (2.28)
Wt (k) is the wage of household k. γw denotes the degree of indexa-

tion to the previous period’s inflation and is limited by: 0 ≤ γw ≤ 1.
The inflation rate is defined as follows: πt−1 = log Pt − log Pt−1 . This
methodology was introduced by Calvo (1983) for the price-setting behav-
ior in a monopolistic good market, whereas Erceg et al. (2000) applied
the concept to the labor market.
Under this assumption, the wage-setting problem becomes dynamic.
The representative household solves the following maximization problem
by its choice of W̄t (k), that is constrained by the producer’s demand
function. A derivation is provided in chapter 2.1.2.

∞ γw
s Pt+s−1
max Et (ξw β) λs W̄t (k) Nt+s (k)
{W̄t (k)} Pt−1
s=0

n,t Nt+s (k)1+ϕ

−
1+ϕ
−θw
W̄t (k)
s.t. Nt+s (k) = Nt+s (j) (2.29)
Wt+s
As the following Lagrangian shows, the optimization problem become

unconstrained by plugging the demand function for Nt+s (k).

∞ γw −θw
s Pt+s−1 W̄t (k)
L = Et (ξw β) λs W̄t (k) Nt+s (j)
Pt−1 Wt+s
s=0
−θw 1+ϕ
n,t W̄t (k)
− Nt+s (j) (2.30)
1+ϕ Wt+s
This yields to the following optimality condition.

∞
∂L s Pt+s−1
: Et (ξw β) Nt+s (k) λs W̄t (k)
∂ W̄t (k) Pt−1

s=0
θw
− n,t Nt+s (k)
ϕ
=0 (2.31)
θw − 1
2.1.2 Producers
There is a continuum of producers in the domestic economy. The pro-
ducers hire labor input from the households in the labor market and sell
their outputs to domestic and foreign households. Therefore, the produc-
ers have to decide about their optimal labor demand and their optimal
price-setting behavior. For the purpose of getting the optimality con-
ditions, the focus stays on a single representative firm. The conditions
hold analogously for the whole continuum of firms.
Labor Demand
The domestic goods from firm j, Yt (j), are produced according to the
given production function.
Yt (j) = a,t f (Nt (j)) (2.32)
The function f (.) is assumed to satisfy the Inada conditions. a,t is

a domestic technology shock that is independent of j and therefore the
same for all domestic producers. Nt (j) is the aggregated labor input for
firm j that is given by the following constant elasticity of substitution
(CES) aggregator.
1
θ θw−1
θw −1 w
Nt (j) = Nt (k) θw (2.33)
0
θw is assumed to be greater than 1: θw > 1. Similar to the households,

producers allocate their expenditures for labor input optimally across all
households. Given the expenditure for labor input, producer j solves the
following problem to allocate its expenditures.
1
1
θ θw−1
θw −1 w
max Wt Nt (j) − Nt (k)Wt (k) dk s.t. Nt (j) = Nt (k) θw (2.34)
{Nt (k)}
0 0
Replace Nt (j) by the CES aggregator. The unconstrained problem is

given by the following Lagrangian.
1
θ θw−1 1
θw −1 w
L = Wt Nt (k) θw − Nt (k)Wt (k) dk (2.35)
0 0
The FOC is derived by setting the derivative of the Lagrangian with

respect to Nt (k) equal to 0.
1
θ θw−1 θ1
θw −1
− 1
∂L w w
: Wt Nt (k) θw Nt (k) θw − Wt (k) = 0 (2.36)
∂Nt (k)
0
Solving this expression for Nt (k) yields to the producer’s demand

function for labor input of type k.
−θw
Wt (k)
Nt (k) = Nt (j) (2.37)
Wt
Price-Setting
The representative producer j has monopolistic power in its good market.
The price-setting method is adopted from Calvo (1983). With probabil-
ity 1 − ξH the producer is allowed to change its price in a given period.
ξH is limited by: 0 < ξH < 1. Prices that are not re-optimized are
indexed to previous period’s inflation rate. The indexation rule is given
by the following expression.
log PH,t (j) = log PH,t−1 (j) + γH πH,t−1 (2.38)
γH defines the degree of indexation and is limited as follows: 0 ≤ γH ≤

1. πH,t is the inflation rate of the domestic goods price. It is defined as
πH,t = log PH,t − log PH,t−1 . As with every firm that has monopolistic
power, producer j wants to maximize the expected discounted value of its
profits by setting an optimal price, P̄H,t (j). The maximization problem
is constrained by the demand function for its goods.

∞ γH
s PH,t+s−1
max Et ξH Qt,t+s [P̄H,t (j) YH,t+s (j)
{P̄H,t (j)} PH,t−1
s=0
n
− PH,t+s M Ct+s YH,t+s (j)]
γH −θ
P̄H,t (j) PH,t+s−1 ∗
s.t. YH,t+s (j) = CH,t+s + CH,t+s (2.39)
PH,t PH,t−1
Qt,t+1 is the firm’s discount factor and M Ctn is the nominal marginal
cost. Again, the demand function can be replaced to derive an uncon-
strained optimization problem. The Lagrangian is given by the following
equation (2.40).

∞

s ∗
L = Et ξH Qt,t+s CH,t+s + CH,t+s

s=0
γH γH −θ
PH,t+s−1 P̄H,t (j) PH,t+s−1
P̄H,t (j)
PH,t−1 PH,t PH,t−1
γH −θ
P̄H,t (j) PH,t+s−1
−PH,t+s M Ct+s
n
(2.40)
PH,t PH,t−1
This yields to the following FOC.

∞ γH
∂L s PH,t+s−1
: Et ξH Qt,t+s P̄H,t (j)
∂ P̄H,t (j) PH,t−1

s=0
θ
− PH,t+s M Ct+s
n
=0 (2.41)
θ−1
2.1.3 Retail Firms

Domestic retail firms import foreign produced goods. Compared to pro-
ducing firms, there is no labor demand decision. The retailer has just
to set an optimal price for its goods. In the domestic reselling mar-
ket, they have monopolistic power. They are price-setters. This market
power leads to a violation of the law of one price in the short-term.
The retail firms set their prices in a Calvo-type procedure with adjust-
ment to previous period’s inflation rate. With a probability of 1 − ξF ,
the retailers are allowed to re-optimize the price for the imports. ξF lies
between 0 and 1: 0 ≤ ξF ≤ 1. The optimal price, P̄F,t (i), is set such that
the expected discounted value of its profits is maximized. The optimal
price has to satisfy the domestic demand function.

∞ γF
s PF,t+s−1
max Et ξF Qt,t+s YF,t+s (i)[P̄F,t (i)
{P̄F,t (i)} PF,t−1
s=0
∗
− Xt+s PF,t+s (i)]
γF −θ
P̄F,t (i) PF,t+s−1
s.t. YF,t+s (i) = CF,t+s (2.42)
PF,t PF,t−1
∗
PH,t is the price for foreign produced goods in the foreign economy.
Replace YF,t+s by its demand function delivers the following Lagrangian
of an unconstrained optimization problem.

∞ γF −θ
s P̄F,t (i) PF,t+s−1
L = Et ξF Qt,t+s CF,t+s
PF,t PF,t−1
γF
s=0
PF,t+s−1 ∗
P̄F,t (i) − Xt+s PF,t+s (i) (2.43)
PF,t−1
Solving the problem yields to the following optimality condition.

∞ γF
∂L s PF,t+s−1
: Et ξF Qt,t+s P̄F,t (i)
∂ P̄F,t (i) PF,t−1

s=0
θ ∗
− Xt+s PH,t+s (i) =0 (2.44)
θ−1
2.1.4 Exogenous Processes

The domestic economy is disturbed by the following shocks: technology
shock (a,t ), preference shock (g,t ), labor supply shock (n,t ), monetary
policy shock (i,t ), cost-push shocks for home goods (cpH,t ), and foreign
goods (cpF,t ) and a risk premium shock (rp,t ). The previous section
showed, how the preference and labor supply shock are introduced. All
other shocks are introduced in the log-linearized stage. For details see
chapter 1 of the appendix.
The following assumptions hold for the log-linearized stage. The mon-
etary policy shock and the cost-push shock for domestic goods are as-
sumed to be independent and identically distributed (i.i.d.) processes.
The other shocks follow an autoregressive process of order 1. These
assumptions are based on Justiniano and Preston (2008).
2.2 Foreign Economy

The foreign economy consists of the same agents as the domestic econ-
omy. However, there is no retail firm since the imports from the small
2.2 Foreign Economy 17
domestic country are negligible. The small domestic economy is engaged

in a negligible amount of financial trades. All in all, the large foreign
economy is approximately closed.
2.2.1 Households
The household’s optimization problems are similar to the problems of the
domestic households. Since the large foreign economy is approximately
closed, the foreign households only have access to foreign produced goods
and foreign debt. Therefore, the households are neither able to buy
domestic produced goods nor able to hold domestic bonds. The foreign
economy consists of a continuum of households. In the following sections,
the focus stays on a single representative household.
Utility Maximization
The representative household chooses consumption (Ct∗ ), labor input
(Nt∗ , working hours) and foreign bonds (Bt∗ ) such as to maximize its
utility function.
1− 1

∞ ∗
Ct+s ∗
− Ht+s σ∗
∗ ∗
n,t (Nt+s )
1+ϕ ∗
∗s ∗
max E0 β g,t 1
− (2.45)
{C ∗ ,N ∗ ,B ∗ }∞ 1− σ∗
1 + ϕ∗
t t t t=0
s=0
External habit formation is described by Ht∗ . Ht∗ = h∗ Ct−1

∗
holds in
∗ ∗ ∗
all periods with 0 < h < 1. g,t and n,t define the preference and the
labor supply shock. σ ∗−1 and ϕ∗ are positive parameters: σ ∗−1 > 0 and
ϕ∗ > 0. β ∗ is the foreign household’s discount factor.
The maximization is limited by a dynamic budget constraint that has
to be fulfilled in each period.
∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗
Pt Ct + Bt = Rt−1 Bt−1 + Wt Nt + Tt (2.46)
Wt∗ is the foreign wage level and Rt∗ is the foreign interest rate. Pt∗
corresponds to the foreign price-level. Tt∗ is assumed to be a lump-sum
transfer to foreign households. It includes profits of the firms and taxes.
The household’s maximization problem can be solved using the La-

grangian method with a multiplier λ∗t .
1− 1

∞ ∗
Ct+s ∗
− Ht+s σ∗
∗ ∗
n,t (Nt+s )
1+ϕ ∗
∗ ∗s ∗
L = E0 β g,t [ 1
−
1− σ∗
1 + ϕ∗
s=0
∗
∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗

− λt+s Pt+s Ct+s + Bt+s − Rt+s−1 Bt+s−1 − Wt+s Nt+s − Tt+s ] (2.47)
The optimality conditions are given by setting the derivatives with
respect to Ct∗ , Nt∗ and Bt∗ to 0.
∂L∗ ∗
∗ 1
∗ − σ∗ ∗ ∗ ∗ ∗ ∗
: g,t Ct − Ht − λt Pt = 0 with Ht = h Ct−1 (2.48)
∂Ct∗
∗
∂L ∗ ∗ϕ∗ ∗ ∗
: − n,t Nt + λt Wt = 0 (2.49)
∂Nt∗
∂L ∗
∗ ∗ ∗
∗
: − λt + β Rt Et λt+1 = 0 (2.50)
∂Bt∗
Additionally, the derivative with respect to the Lagrangian multiplier

λ∗t and the transversality condition have to be satisfied simultaneously.
Replace λ∗t in the conditions (2.49) and (2.50) using equation (2.48).
∗ ∗ϕ
∗ ∗ ∗
1/σ∗ Wt∗
n,t Nt Ct − h Ct−1 = (2.51)
Pt∗
∗ −1
∗
∗ ∗ ∗
−1/σ∗ ∗ ∗ P t+1 ∗
∗ ∗ ∗
−1/σ∗
g,t Ct − h Ct−1 = β Rt Et g,t+1 Ct+1 − h Ct (2.52)
Pt∗
Consumption Allocation
Solving the utility maximization problem, the representative household
chooses an optimal consumption bundle Ct∗ . The corresponding expen-
diture is given by Pt∗ Ct∗ . Since the foreign household has only access
to foreign produced goods, the consumption bundle is just a mixture
∗
of foreign produced goods, CF,t (i). As for domestic households, the
consumption bundle is aggregated with a Dixit-Stiglitz function.
1
θ∗
θ ∗ −1 θ ∗ −1
∗ ∗
Ct = CF,t (i) θ∗ di (2.53)
0
θ∗ describes the elasticity of substitution and is assumed to be positive:

θ∗ > 0. For an optimal allocation across the foreign produced goods, the
household minimizes the given expenditures Pt∗ Ct∗ minus the costs for
each of the foreign produced goods. The problem is constrained by the
aggregation function.
1
1
θ∗
θ ∗ −1 θ ∗ −1
∗ ∗ ∗ ∗ ∗ ∗
min P t Ct − PF,t (i)CF,t (i) di s.t. Ct = Cj,t (i) θ∗ di (2.54)
{C ∗ (i)}
F,t 0 0
The constraint is directly plugged into the optimization problem. The

Lagrangian becomes unconstrained.
1
θ∗ 1
θ ∗ −1 θ ∗ −1
∗ ∗ ∗ ∗ ∗
L = Pt CF,t (i) θ∗ di − PF,t (i)CF,t (i) di (2.55)
0 0
Taking the derivative with respect to Cj,t (i) delivers the following
optimality condition.
1
θ∗
1

θ ∗ −1 θ ∗ −1 θ∗
∂L∗ ∗ ∗ ∗ − 1∗ ∗
∗ (i)
: Pt CF,t (i) θ∗ di CF,t (i) θ − PF,t (i) = 0 (2.56)
∂CF,t
0
∗
Solving for CF,t (i) yields to the following demand for the foreign pro-
duced good i.
P∗ −θ
∗ F,t (i) ∗
CF,t (i) = Ct (2.57)
Pt∗
Price-Level
Under the assumption of price-taking and non-satiating households, the
1 ∗
following expression holds: Pt∗ Ct∗ = 0 PF,t ∗
(i)CF,t (i) di. Replace Ct∗ by
using the demand function (2.57) yields to the following definition of the
foreign price-level.
1
1
1−θ ∗
∗ ∗ 1−θ ∗
Pt = PF,t (i) (2.58)
0
Labor Supply
The foreign households provide their labor input to a continuum of for-
eign producers. The labor inputs can be hired in a single foreign labor
market. The households have to decide how many labor inputs they are
willing to provide depending on a common wage. The derivation of the
condition for an optimal labor supply makes use of the producer’s labor
demand that is discussed in chapter 2.2.2.
As in the domestic labor market, the foreign households have monop-
olistic power. The wage-setting follows a Calvo-style process where the
probability of re-optimizing the wage is 1 − ξw∗ with 0 < ξw∗ < 1. With
a probability of ξw∗ the representative household stays with its wage for
another period. In this case, the indexation rule adjusts the wage to past
inflation rates. The indexation rule is given by the following expression.
∗ ∗ ∗
log Wt (k) = log Wt−1 (k) + γw∗ πt−1 (2.59)
The degree of adjustment to the previous period’s inflation is denoted

by γw∗ . Parameter γw∗ is limited: 0 ≤ γw∗ ≤ 1. The foreign inflation
∗
rate is defined as follows: πt−1 = log Pt∗ − log Pt−1
∗
.
In order to maximize the following expression, the representative
household choses its optimal wage, W̄t∗ (k). The maximization is con-
strained by the producer’s demand function for labor input of type k.
γw∗
∞
∗
s ∗ ∗
∗
Pt+s−1 ∗
max Et ξw∗ β λs Wt (k) ∗
Nt+s (k)
{W̄ ∗ (k)} Pt−1
t

s=0
∗
∗ ∗
n,t Nt+s (k)
1+ϕ
−
1 + ϕ∗
−θw∗
∗ W̄t∗ (k) ∗
s.t. Nt+s (k) = ∗
Nt+s (j) (2.60)
Wt+s
Making use of the demand function, the maximization problem be-

comes unconstrained. The Lagrangian function of the problem is given
by equation (2.61).
γw∗ −θw∗
∗
∞
∗
s ∗ ∗
∗
Pt+s−1 W̄t∗ (k) ∗
L = Et ξw∗ β λs W̄t (k) ∗ ∗
Nt+s (j)
Pt−1 Wt+s
1+ϕ∗
s=0
−θw∗
∗
n,t W̄t∗ (k) ∗
− ∗
Nt+s (j) (2.61)
1 + ϕ∗ Wt+s
Taking the derivative with respect to W̄t∗ delivers the following opti-
mality condition.

∂L∗
∞
∗
s ∗ ∗
∗
Pt+s−1 ∗
: Et ξw∗ β Nt+s (k) λs ∗
W̄t (k)
∂ W̄t (k) Pt−1

s=0
θw∗ ∗ ∗ ϕ∗
− n,t Nt+s (k) =0 (2.62)
θw∗ − 1
2.2.2 Producers
The production side of the foreign economy exists in a continuum of pro-
ducers. The goods are produced using labor input from foreign house-
holds. The labor inputs can be hired in the foreign labor market. The
produced goods are sold to the domestic retail firm and to foreign house-
holds. There are two decisions to be considered by foreign producers:
optimal labor demand and optimal price. The derivation focuses on a
single representative producer j. All conditions hold simultaneously for
the whole continuum of firms.
Labor Demand
The technology for producing the foreign good j, Yt∗ (j), is given by the
following equation.
∗ ∗ ∗ ∗
Yt (j) = a,t f (Nt (j)) (2.63)
As for the domestic producers, the production function f ∗ (.) is as-

sumed to satisfy the Inada conditions. The foreign technology shock
∗a,t is independent of j and therefore the same for all foreign producers.
Nt∗ (j) is firm j’s aggregated labor input that is a mixture of all labor
inputs from all foreign households.
1
θ θw∗−1
θw∗ −1 w∗
∗ ∗
Nt (j) = Nt (k) θw∗ (2.64)
0
Parameter θw∗ is constrained by the following unequality: θw∗ > 1.

Firm j wants to allocate its expenditure for labor inputs optimally across
all foreign households. Therefore, the representative producer j has to
solve the following maximization problem.
1
∗ ∗ ∗ ∗
max Wt Nt (j) − Nt (k)Wt (k) dk
{N ∗ (k)}
t 0
1
θ θw∗−1
θw∗ −1 w∗
∗ ∗
s.t. Nt (j) = Nt (k) θw∗ (2.65)
0
Again, Nt∗ (j) can be replaced by the aggregator function. This yields
to an unconstrained Lagrangian.
1
θ θw∗−1 1
θw∗ −1 w∗
∗ ∗ ∗ ∗ ∗
L = Wt Nt (k) θw∗ − Nt (k)Wt (k) dk (2.66)
0 0
Taking the derivative with respect to Nt∗ (k) delivers the following
optimality condition.
1
θ θw∗−1 θ 1
∗ θw∗ −1 w∗ w∗ 1
∂L ∗ ∗ ∗ − ∗
: Wt Nt (k) θw∗ Nt (k) θw∗ − Wt (k) = 0 (2.67)
∂Nt (k)∗
0
Producer j’s demand function for labor input of type k is derived by

solving the expression above for Nt∗ (k).
−θw∗
∗ Wt∗ (k) ∗
Nt (k) = Nt (j) (2.68)
Wt∗
Price-Setting
The foreign producer j is a price-setter in its good market. This means
that producer j has monopolistic power. As in the domestic economy,
a Calvo-type price-setting process is assumed. The representative pro-
ducer is allowed to re-optimize its price with a given probability of 1−ξF ∗
with 0 < ξF ∗ < 1. The indexation rule applies to prices that are not
re-optimized. The adjustment to previous period’s inflation follows this
process.
∗ ∗ ∗
log PF,t (i) = log PF,t−1 (i) + γF ∗ πF,t−1 (2.69)
γF ∗ measures the degree of adjustment to past inflation rates. The

parameter lies between 0 and 1: 0 ≤ γF ∗ ≤ 1. The inflation rate of
∗
foreign produced goods is denoted by πF,t that is formally defined as:
∗ ∗ ∗
πF,t = logPF,t − logPF,t−1 . By setting an optimal price for its produced
∗
goods (P̄F,t ), firm j maximizes the expected present discounted value
of its profits. The optimization problem is constrained by the demand
function for its goods.

∞ ∗
γF ∗
s ∗ ∗
PF,t+s−1 ∗
max Et ξF ∗ Qt,t+s [P̄F,t (i) ∗
YF,t+s (i)
{P̄ ∗ (i)} PF,t−1
F,t
s=0
∗ n∗ ∗
− PF,t+s M Ct+s YF,t+s (i)]
γF ∗ −θ∗
∗
∗
P̄F,t (i) ∗
PF,t+s−1 ∗

s.t. YF,t+s (i) = ∗ ∗
CF,t+s + CF,t+s (2.70)
PF,t PF,t−1
The two variables Q∗t,t+1 and M Ctn∗ denote the foreign producer’s
discount factor and the nominal marginal cost, respectively. Making
use of the demand function, expression (2.70) can be transformed to an
unconstrained Lagrangian function.

∞

∗ s ∗ ∗
L = Et ξF ∗ Qt,t+s CF,t+s + CF,t+s

s=0
∗
γF ∗ ∗
∗
γF ∗ −θ∗
∗
PF,t+s−1 P̄F,t (i) PF,t+s−1
P̄F,t (i) ∗ ∗ ∗
PF,t−1 PF,t PF,t−1
∗
∗
γF ∗ −θ∗
∗
P̄F,t (i) PF,t+s−1
n∗
−PF,t+s M Ct+s ∗ ∗
(2.71)
PF,t PF,t−1
∗
The FOC is derived by taking the derivative with respect to P̄F,t (i).

∞ ∗
γF ∗
∂L s ∗ ∗
PF,t+s−1
∗ (i)
: Et ξF ∗ Qt,t+s P̄F,t (i) ∗
∂ P̄F,t PF,t−1

s=0
θ∗ ∗
− PF,t+s M Ct+s
n∗
=0 (2.72)
θ∗ −1
2.2.3 Monetary Authority

The foreign monetary authority follows a flexible inflation targeting
regime. It reacts not only to changes in the inflation rate but also
to changes in output and output growth. The policy is described with
a simple Taylor-type rule.
R∗ ρi∗ θπ∗ θy∗ θΔy∗ 1−ρi∗

Rt∗ t−1 Pt∗ Yt∗ Yt∗ ∗
= ∗ ∗
m,t (2.73)
R̄∗ R̄∗ Pt−1 Ȳ ∗ Yt−1
R̄∗ and Ȳ ∗ are the steady state values of foreign interest rate and
output. ∗m,t is the exogenous monetary policy shock. The monetary
regime tries to smooth its actions over time according to ρi∗ , what can
be seen from the lagged interest rate term. θπ∗ , θy∗ and θΔy∗ describe
the magnitude of an interest rate change due to changes in inflation,
output and output growth.
Furthermore, fiscal policy is assumed to follow a zero debt policy.
2.3 General Equilibrium and Definitions 25
2.2.4 Exogenous Processes

There are five shocks in the foreign economy: technology shock (∗a,t ),
preference shock (∗g,t ), labor supply shock (∗n,t ), monetary policy shock
(∗i,t ) and cost-push shock (∗cp,t ). The cost-push shock is introduced at
the end of the price Phillips curve in the log-linearized stage (see chapter
1 of the appendix). All other shocks were already part of the previous
derivations.
The shocks have the following specification in the log-linearized stage:
The cost-push shock and the monetary policy shock are i.i.d.. The other
shocks, namely the technology shock, the preference shock, and the labor
supply shock, follow a first order autoregressive process. Again, the
assumptions are based on Justiniano and Preston (2008).
2.3 General Equilibrium and Definitions

The following section introduces the equilibrium conditions for the goods
market. These conditions are basic assumptions in general equilibrium
models. Afterwards various definitions for an open economy model are
introduced. The definitions build linkages between the two economies.
2.3.1 General Equilibrium

The domestic economy is assumed to be small and the large foreign
economy is approximately closed. These assumptions are fundamental
for the general equilibrium in the goods market.
The domestic produced goods YH,t are sold to domestic and foreign
households. The market for domestic produced goods clears if the fol-
lowing expression holds.
∗
YH,t = CH,t + CH,t (2.74)
The demand for foreign produced goods by the domestic economy is

negligible. Therefore, foreign produced goods are sold only to foreign
households. The market clearing condition is given by the following

equation.
∗ ∗
Yt = Ct (2.75)
The foreign demand for domestically produced goods is assumed to

follow equation (2.76).
P ∗ −λ
∗ H,t ∗
CH,t = Yt (2.76)
P∗
Parameter λ is positive, λ > 0, and is allowed to be different from

the domestic elasticity of substitution across domestic goods (η). This
assumption is based on Kollmann (2002). It provides more possibilities
for the transmission mechanism of foreign shocks to impact the domestic
economy.
As previously explained, the same argumentation holds for the bond
market. Domestic households have access to domestic bonds and foreign
bonds, whereas foreign households hold only foreign bonds. Therefore,
domestic debt is in zero net supply: Dt = 0 for all t.
Additionally, the model assumes symmetries. All producers that are
allowed to re-optimize their price in period t set a common optimal
price PH,t (or Pt∗ for foreign producers). Analogously, all retailers re-
optimize their price to PF,t if they are allowed to do so in period t. The
same holds for households in their function as wage-setters. Finally,
initial wealth is common across all households. This leads to identical
optimality decisions for consumption, labor supply and savings.
2.3.2 Various Definitions

To complete the open economy model, some linkages between the do-
mestic and the foreign economy have to be introduced. The linkages are
just a matter of definition.
2.3 General Equilibrium and Definitions 27
Terms of Trade
The effective terms (st ) of trade are defined by equation (2.77).
PF,t
St ≡ (2.77)
PH,t
Law of One Price

Since the foreign economy is approximately closed, it holds that Pt∗ =
∗
PF,t . Since the law of one price fails to hold, Monacelli (2005) defined a
gap ΨF,t by the following expression.
Xt Pt∗
ΨF,t ≡ (2.78)
PF,t
Real Exchange Rate

The real exchange rate (Ωt ) is defined as the nominal exchange rate
times the ratio of the domestic over the foreign price-level.
Xt Pt
Ωt ≡ (2.79)
Pt∗
Uncovered Interest Parity

The uncovered interest parity restricts the movements of domestic and
foreign interest rates as well as changes in the nominal exchange rate.
The condition is based on the optimality conditions for domestic and
foreign bond holdings.

∗ Xt
Et [λt+1 Pt+1 ] Rt − Rt rp,t =0 (2.80)
Et [Xt+1 ]
2.4 Domestic Monetary Policy Rule

Last but not least, the domestic monetary policy is needed to complete
the model. The inflation targeting rule is the standard case. It is used
for the Bayesian estimation in chapter 3. The price-level targeting is
used for different exercises in chapter 4.
2.4.1 Inflation Targeting Rule

In the inflation targeting case, the domestic monetary authority follows
a simple Taylor-type rule that accounts for inflation, output and output
growth. The monetary policy tries to smooth its actions over time, what
can be seen from the lagged interest rate term.
ρi θπ θy θΔy 1−ρi
Rt Rt−1 Pt Yt Yt
= ¯m,t (2.81)
R̄ R̄ Pt−1 Ȳ Yt−1
R̄ and Ȳ are the steady state values of the interest rate and output.
m,t denotes the exogenous monetary policy shock. The parameter ρi
measures the importance of interest rate smoothing over time. It holds
that 0 < ρi < 1. All other parameters of equation (2.81) are assumed
to be positive: θπ > 0, θy > 0 and θΔy > 0. Usually, it is assumed
that the interest rate reacts more than one-for-one to a change in the
inflation rate: θπ > 1. θy and θΔy describe the importance of the other
goals of the flexible inflation targeting regime. Its values are close to
zero: θy < 1 and θΔy < 1.
2.4.2 Price-Level Targeting Rule

Compared to the inflation targeting case, the interest rate reacts to
changes in the price-level rather than changes in the inflation rate. Out-
put, output growth and interest rate smoothing goals are assumed to be
the same.
ρi θπ θy θΔy 1−ρi
Rt Rt−1 Pt Yt Yt
= ¯m,t (2.82)
R̄ R̄ P̄ Ȳ Yt−1
2.4 Domestic Monetary Policy Rule 29
P̄ is the steady state value of the domestic price-level. As discussed in

the introduction, Sweden is the only country that has ever introduced
a price-level targeting regime for a relatively short time period (1931
to 1937). Because of this lack of data, there is no way to estimate the
parameters separately for the price-level targeting rule. Therefore, the
parameter’s values are just taken over from the inflation targeting case
that is estimated in the next chapter 3.
3 Bayesian Estimation of the
Parameters
The Bayesian analysis of the derived model is mainly based on Schorfheide
(2000) such as An and Schorfheide (2007). The parameters of the DSGE
model are estimated with a mixture of information contained in the data
and prior beliefs about the distribution of the parameters. In the follow-
ing chapter 3.1, the theoretical background is briefly introduced. More
information about the database is given in chapter 3.2. The prior distri-
butions of the estimated parameters are based on Robinson (2013). The
rest of the chapter discusses the outcome of the estimation. The outcome
includes the posterior distributions of the parameters and a historical
simulation. As already mentioned in the previous chapter (see 2.4), the
estimations are based on the standard model using an inflation targeting
rule. All calculations and simulations are conducted using Matlab.
3.1 Methodology
The Bayesian analysis simultaneously uses both a combination of infor-
mation from earlier studies (prior beliefs) and from real world data to
estimate the parameters of a DSGE model. The use of prior informa-
tion is especially appropriate, if the model is complex or if the sample
period of the data is short. In this thesis, 44 parameters are estimated
with only six data series. Without the use of prior information, the em-
pirical analysis would heavily suffer from identification problems. The
log-likelihood functions of the parameters would not have enough curva-
ture to find proper maximums. Although the information contained in
six series is rather minor, Bayesian estimation allows to adapt the param-

32 3 Bayesian Estimation of the Parameters
eters of the model to Swiss data and is preferred to a simply calibrated

model.
The starting point for the Bayesian analysis is the canonical form of
the DSGE model according to Sims (2002). It is built on log-linearized
equations provided in the appendix (see chapter 1). The canonical form
is characterized by the following equation.
Γ0 St = Γ1 St−1 + Ψvt + Πηt (3.1)
St is defined as the state vector that includes all endogenous variables

of the equation system. The vector vt contains all stochastic variables
and ηt all rational expectation forecast errors. Γ0 (42 × 42), Γ1 (42 × 42),
Ψ (42 × 12) and Π (42 × 9) denote the corresponding matrices based on
the model’s parameters to describe the equation system.
Using Sim’s algorithm, the system of equations is solved and described
by the following state-space form. The second equation is called obser-
vation equation.
St = F St−1 + vt with E[vt vt ‘] = Q (3.2)

Yt = H St + wt with E[wt wt ‘] = R (3.3)
Yt is the vector of observed macroeconomic time series. In this case,

it contains six observed variables. H (42 × 6) links the state vector and
the vector of observed series. wt denotes the measurement error. It is
assumed that vt and wt are orthogonal to each other and to their own
lagged values.
Using the Kalman filter recursion, the log-likelihood function of the
observed macroeconomic series conditional on the parameters can be
computed.1
1
For more information about the Kalman filter, see for example Hamilton (1994).
3.1 Methodology 33
T

log L(YT ) = − log |H Et−1 [Pt ]H + R
2

1

T
Yt − H Et−1 [St ] Yt − H Et−1 [St ]
−
2 (H Et−1 [Pt ]H + R)
t=1
with Et [Pt+1 ] = F Et−1 [Pt ]F + Q and YT = {yt }Tt=0 (3.4)
Pt denotes the precision matrix. The matrices H, F , R and Q depend

on the model’s parameters (Λ). Λ is a vector of all structural parameters
that are going to be estimated. Because there are only six observed
macroeconomic series, the curvature of the log-likelihood functions will
be too low to solve them properly. More curvature is implied by the
prior information about the distribution of the structural parameters.
The joint density function F (.) of the observed data and the structural
parameters is defined as follows:
F (YT , Λ) = φ(Λ)L(YT |Λ) = ψ(YT )P (Λ|YT ) (3.5)
φ(Λ) is the prior distribution of the parameters and L(YT |Λ) is the
likelihood function of the data conditional on the structural parameters.
The functions ψ(YT ) and P (Λ|YT ) denote the density of the data and
the posterior distribution of the parameters conditional on the observed
data, respectively. Expression (3.5) can be rearranged to a definition for
the posterior distribution of the parameters conditional on the data.
φ(Λ)L(YT |Λ)
P (Λ|YT ) = (3.6)
ψ(YT )
Since the density function of the observed data, ψ(YT ), is independent

of the structural parameters, it is just a known constant. Therefore,
the posterior distribution of the parameters is proportional to the prior
distribution of the parameters times the likelihood function of the data.
P (Λ|YT ) ∝ φ(Λ)L(YT |Λ) (3.7)

The parameters are estimated in order to maximize the likelihood

function. Since the maximization of sums is simpler, the likelihood func-
tion can be expressed in terms of logarithms.
log P (Λ|YT ) ∝ log φ(Λ) + log L(YT |Λ) (3.8)

Finally, the parameters are estimated in order to maximize the log-
likelihood function. The following sections provide more detailed infor-
mation about the observed series and the choice of the prior distribution
of the parameters.
3.2 Data
As shown in the previous section, the six observed series are put into the
state-space form using the vector Yt . The sample period reaches from
1979Q4 to 2005Q4. The vector is defined as follows:
∗ ∗ ∗

yt = Outputt , Interestt , Inf lationt , Outputt , Interestt , Inf lationt (3.9)
Foreign Economy
The foreign economy is modeled with data from the Euro area. The
series are denoted by a superscript ’*’. The series are published in the
Area-Wide Model (AWM) database by the European Central Bank. The
series are transformed such that they are stationary.
The foreign economy’s output (Output∗t ) is measured by the real gross
domestic product (GDP ∗ ). The logarithmic GDP ∗ is seasonally ad-
justed using a Hodrick-Prescott filter. Finally, it is multiplied by 100.
∗ ∗
Outputt = HP F ilter[ln(GDPt )] × 100 (3.10)
The variable Interest∗t is quantified by the short-term interest rate
from the AWM database. The series is not transformed.2
2
The short-term interest rate is measured in percentage points. Therefore, a mul-
tiplication by 100 is unnecessary.
3.2 Data 35
Figure 3.1: Transformed Series

The inflation rate is measured by the consumer price index (CP I ∗ ).

The base year is set to 1970Q1. The CP I ∗ is logarithmized. The loga-
rithmic CP I ∗ is first-differentiated and 1 is added. The modified series
is taken to the power 4 and multiplied by 100.
∗
∗ ∗
4
Inf lationt = { ln(CP It ) − ln(CP It−1 ) + 1 − 1} × 100 (3.11)
Obviously, the first-differentiation drops the first observation (1979Q4).

The modified sample ranges from 1980Q1 to 2005Q4.
Domestic Economy
The small domestic economy is specified with data from Switzerland.
The output data was collected and published by the State Secretariat for
Economic Affairs (SECO). The data for interest rates and for inflation
are recorded by the Swiss National Bank and the Federal Statistical
Office, respectively. As for the foreign economy, the series are made
stationary for the empirical analysis.
Outputt is quantified by the gross domestic product (GDP ). The
GDP is logarithmized and seasonally adjusted using a Hodrick-Prescott
filter. The modified series is multiplied by 100.
Outputt = HP F ilter[ln(GDPt )] × 100 (3.12)
The domestic interest rate is measured by the government bond yield.

No transformation is performed.3
The consumer price index (CP I) builds the basis for the calculation
of Inf lationt . One is added to the logarithmized and first-differentiated
CP I. The modified series is taken to the power 4, reduced by 1 and
multiplied by 100. The first-differentiation excludes the first observation
(1979Q4).
3
The government bond yield is measured in percentage points. A multiplication
by 100 is unnecessary.
3.3 Prior Distribution 37
4
Inf lationt = {[ln(CP It ) − ln(CP It−1 ) + 1] − 1} × 100 (3.13)
Figure 3.1 shows all transformed series over the whole sample horizon
(1980Q1 to 2005Q4).
3.3 Prior Distribution
Some of the parameters are calibrated and not estimated. It is the

discount factor for the domestic economy (β) and the foreign economy
(β ∗ ) that is set to 0.99 for both economies. The inverse Frisch parameters
(ϕ and ϕ∗ ) are fixed to 1. The elasticity of substitution between types
of differentiated domestic (θ) and foreign goods (θ∗ ) are both calibrated
to 8 following Woodford (2003). In the domestic economy, the interest
debt elasticity parameter (χ) is set to 0.01 according to Benigno (2009).
Additionally, the openness parameter (α) is specified to 0.185.
The prior distributions of the parameters are summarized in table
3.1. The second column denotes the type of the density function. Mean
and standard deviation are described in columns 3 and 4. The density
functions and the standard deviations are taken over from Robinson
(2013). The means are adapted from Robinson’s posterior means. All
parameters are positive.
All parameters of first-order autoregressive processes (ρ) follow a beta
distribution. Their values lie between 0 and 1. The standard deviations
(σ) characterize the stochastic variables of the model. They follow an
inverse gamma distribution with relatively large standard deviations.
Their means and standard deviations are specified to the shock. For
example, mean and standard deviation of the domestic and foreign mon-
etary policy shock are as usual assumed to be relatively small. The
prior means and standard deviations are fixed at 0.07 and 0.25 for the
domestic and foreign shock.
The parameters, considering the Calvo-type price- and wage-setting
(ξ), are beta distributed and have a range between 0 and 1. The means
for the parameters of the price-setting processes (around 0.75) are larger
Table 3.1: Prior Distribution
Density Function Mean Std. Dev.

h Beta 0.19 0.25
σ Normal 1.12 0.4
λ Gamma 0.38 0.5
η Gamma 0.57 0.75
ρa Beta 0.82 0.1
ρg Beta 0.81 0.1
ρcpF Beta 0.74 0.25
ρrp Beta 0.75 0.1
ρn Beta 0.54 0.1
ρi Beta 0.86 0.25
ξH Beta 0.71 0.1
ξF Beta 0.53 0.1
ξw Beta 0.53 0.1
θπ Gamma 1.9 0.3
θΔy Gamma 0.36 0.13
θy Gamma 0.18 0.13
γH Beta 0.47 0.2
γF Beta 0.37 0.2
γw Beta 0.53 0.2
h∗ Beta 0.53 0.1
σ∗ Normal 1.04 0.4
ρa∗ Beta 0.83 0.1
3.3 Prior Distribution 39
Density Function Mean Std. Dev.

ρg∗ Beta 0.85 0.1
ρn∗ Beta 0.56 0.1
ρi∗ Beta 0.86 0.1
ξF ∗ Beta 0.81 0.1
ξw∗ Beta 0.45 0.1
θπ∗ Gamma 1.92 0.3
θΔy∗ Gamma 0.55 0.13
θy∗ Gamma 0.1 0.13
γF ∗ Beta 0.35 0.2
γw∗ Beta 0.51 0.2
σa Inv. Gamma 0.45 1.0
σg Inv. Gamma 0.98 1.0
σcpH Inv. Gamma 0.25 1.0
σcpF Inv. Gamma 1.81 1.0
σrp Inv. Gamma 0.51 1.0
σn Inv. Gamma 5.33 1.0
σi Inv. Gamma 0.07 0.25
σa∗ Inv. Gamma 0.37 1.0
σg∗ Inv. Gamma 1.17 1.0
σcp∗ Inv. Gamma 0.15 1.0
σn∗ Inv. Gamma 5.34 1.0
σi∗ Inv. Gamma 0.07 0.25
than the means for the wage-setting processes (around 0.5). The same
holds for the indexation parameters (γ). They are beta distributed and
lie between 0 and 1. The mean for the indexation in the price-setting
case is around 0.4 and for the wage-setting case around 0.5.
The monetary policy regime is modeled by a Taylor-type rule that is
specified by the parameters θ. The parameters θ are assumed to follow
a gamma distribution. Specifically, the interest rate is assumed to react
more than one-for-one to changes in the inflation rate. θπ and θπ∗ have
to be greater than 1 with a mean at 1.9. The reactions to changes in
output (θy and θy∗ ) and output growth (θΔy and θΔy∗ ) are assumed to
be smaller.
The elasticity of intertemporal substitution (σ and σ ∗ ) follows a nor-
mal distribution with a mean around 1 and a standard deviation of 0.4.
The parameters of the household’s habit formation (h and h∗ ) are beta
distributed. According to estimations of Robinson (2013) the posterior
mean of the foreign parameter h∗ (0.53) is assumed to be larger than the
mean of the domestic parameter h (0.19). The elasticity of substitution
between foreign and domestic goods (η) follows a gamma distribution
with a relatively large standard deviation of 0.75. The mean is set to
0.57.
3.4 Estimation Results

Table 3.2 shows the estimation output. The first column delivers the
estimations of the means. The standard deviation is given in the sec-
ond column. The third column provides the median of the parame-
ter’s estimation. Finally, the last two columns show the 90%-confidence-
interval. The estimated parameter values are compared to Cuche-Curti
et al. (2009)4 as well as to other references. Convergence was reached
after 607 201 evaluations with an optimal log-likelihood value of 113.97.
4
Cuche-Curti et al. (2009) provide a benchmark DSGE model for the Swiss econ-
omy that is used at the Swiss National Bank for policy analysis. Parameters are
calibrated to previous studies about the Swiss economy and to Adolfson et al.
(2007).
3.4 Estimation Results 41
The confidence intervals are calculated using a random walk Metropolis

methodology that is one of the common Markov chain Monte Carlo al-
gorithms. The random walk Metropolis algorithm is executed with 100
000 iterations. Thereby, the acceptance rate was exactly 0.25.
The estimators for the price stickiness are around 0.65 for the Swiss
economy (ξH and ξF ) and around 0.7 for the European economy (ξF ∗ ).
Compared to Cuche-Curti et al. (2009), the domestic price stickiness is
smaller and the foreign price stickiness is perfectly in line. The parame-
ters for the wage stickiness are estimated around 0.5 for Switzerland (ξw )
and around 0.4 for the Euro area (ξw∗ ). These values are comparatively
small regarding Cuche-Curti et al. (2009) or Adolfson et al. (2007). The
estimation delivers very different results for the indexation parameters
in the Swiss economy (small values) and the European economy (large
values). Cuche-Curti et al. (2009) assumes all indexation parameters
to be 0.6. This value is comparable to the wage adjustment parameters
(γw and γw∗ ), whereas it is too high for Switzerland’s price indexation
(γH and γF ) and too low for the Euro area’s price indexation (γF ∗ ).
The habit formation parameter is relatively high in the Swiss econ-
omy (h) and relatively small in the European economy (h∗ ) compared
to the estimated value of 0.6 in Boldrin et al. (2001). The elasticities of
intertemporal substitution (σ and σ ∗ ) are close to 0. In most other refer-
ences the intertemporal elasticity of substitution is assumed to be close
to 1.5 The estimator of the elasticity of substitution between domestic
and foreign goods (η) is slightly smaller than suggested in Justiniano
and Preston (2008). The estimation of λ (0.31) is almost congruent to
its posterior mean.
The estimated values for the interest rate smoothing parameter (ρi
and ρi∗ ) are around 0.9 and in line with Adolfson et al. (2007). The re-
actions to a change in inflation, output and output growth are calibrated
with 1.5, 0.1 and 0.1 in Cuche-Curti et al. (2009). The estimations for
Switzerland show that the reaction to a change in the inflation rate (θπ )
is smaller and the reactions to changes in output (θy ) and output growth
(θΔy ) are higher compared to Cuche-Curti et al. (2009). θπ is only little
5
Parameters σ and σ ∗ are often defined as their inverted values in other references.
Table 3.2: Posterior distribution
Mean Std. Dev. Median 90%-Conf.-Interv.

h 0.8409 0.0864 0.8617 0.6716 0.9509
σ 0.1648 0.0952 0.1490 0.0405 0.3453
λ 0.3117 0.0622 0.3109 0.2148 0.4203
η 0.6804 0.2172 0.6247 0.3581 1.0885
ρa 0.7903 0.1008 0.8024 0.6089 0.9315
ρg 0.8402 0.0704 0.8561 0.7034 0.9258
ρcpF 0.7953 0.1041 0.8182 0.5930 0.9430
ρrp 0.9075 0.0280 0.9088 0.8611 0.9514
ρn 0.5376 0.1006 0.5397 0.3698 0.7014
ρi 0.9208 0.0130 0.9203 0.8998 0.9424
ξH 0.6394 0.0876 0.6398 0.4948 0.7810
ξF 0.6754 0.0481 0.6799 0.5942 0.7487
ξw 0.4841 0.0969 0.4844 0.3234 0.6445
θπ 1.0878 0.0976 1.0547 0.9913 0.13011
θΔy 0.4629 0.1432 0.4504 0.2472 0.7192
θy 0.4383 0.1596 0.4233 0.2061 0.7261
γH 0.1414 0.0814 0.1257 0.0350 0.2955
γF 0.1385 0.0966 0.1154 0.0297 0.3270
γw 0.5052 0.2033 0.5087 0.1747 0.8354
h∗ 0.3857 0.0369 0.3886 0.3213 0.4430
σ∗ 0.0298 0.0117 0.0275 0.0142 0.0511
ρa∗ 0.9665 0.0174 0.9692 0.9345 0.9903
3.4 Estimation Results 43
Mean Std. Dev. Median 90%-Conf.-Interv.

ρg∗ 0.6072 0.0577 0.6089 0.5121 0.6992
ρn∗ 0.5552 0.0998 0.5557 0.3881 0.7211
ρi∗ 0.8941 0.0157 0.8944 0.8669 0.9187
ξF ∗ 0.7274 0.0566 0.7339 0.6250 0.8093
ξw∗ 0.3913 0.0902 0.3890 0.2491 0.5455
θπ∗ 1.7617 0.1668 1.7532 1.5040 2.0496
θΔy∗ 0.7628 0.3189 0.7151 0.3266 1.3491
θy∗ 0.3333 0.2116 0.2879 0.0767 0.7370
γF ∗ 0.8282 0.0735 0.8365 0.6922 0.9322
γw∗ 0.6625 0.1793 0.6876 0.3238 0.9079
σa 0.9265 0.6919 0.7145 0.3049 2.4924
σg 1.4551 0.5761 1.3586 0.6928 2.4627
σcpH 1.6107 0.4683 1.5524 0.9085 2.4741
σcpF 4.4656 1.8373 4.0534 2.2874 8.6748
σrp 0.7287 0.3340 0.6635 0.3291 1.3670
σn 5.6177 0.9059 5.5054 4.3217 7.3237
σi 0.0918 0.0129 0.0909 0.0724 0.1144
σa∗ 3.1152 1.8851 2.7364 0.7967 6.8487
σg∗ 0.7257 0.2187 0.6884 0.4364 1.1511
σcp∗ 1.8147 0.3271 1.7993 1.3052 2.3677
σn∗ 5.8278 1.0121 5.7091 4.3008 7.7216
σi∗ 0.2615 0.0384 0.2548 0.2089 0.3376
above 1. In the Euro area, the monetary authority reacts stronger to all
types of changes in comparison with Cuche-Curti et al. (2009).
The persistences of the first-order autoregressive processes are de-
scribed by the parameters ρ. In both countries, the lowest persistence
shows up in the labor supply shock processes (ρn and ρn∗ ). This is per-
fectly in line with Justiniano and Preston (2008). The persistence of the
technological process is higher in the Euro area (ρa∗ ) than in the Swiss
economy (ρa ). The preference shocks (ρg and ρg∗ ) and the risk premium
shock (ρrp ) are a little more persistent than in Justiniano and Preston
(2008). The persistence of the cost-push shock for foreign goods (ρcpF )
is around 0.8. The cost-push shock for domestic goods is assumed to be
white noise.
The standard deviations of the monetary policy shocks are relatively
small compared to other shocks in both countries. This is in line with
the findings of Justiniano and Preston (2008). The biggest variations
show up in the labor supply shocks (σn and σn∗ ) what is likewise compa-
rable to Justiniano and Preston (2008). Relatively big differences can be
found looking at the standard deviations of the technology shock. The
standard deviation for Switzerland (σa ) is relative small compared to
the standard deviation of the Euro area (σa∗ ).
3.5 Historical Simulation

Historical simulations for all observed variables are provided in figure 3.2.
The simulation is used to examine the empirical relevance of the model.
It shows the fit to the data based on the estimated mean estimates. The
solid line illustrates the calculated variables of the model. The observed
data series are given by the dotted lines.
The domestic interest rate is fitted very closely by the Taylor-type
monetary policy rule. The two series of the data and the model almost
coincide. The same holds for the domestic output that is replicated
tightly. The domestic inflation is reproduced less closely. The estima-
ted series and the data proceed pro-cyclical but the short-run volatility
3.5 Historical Simulation 45
Figure 3.2: Historical Simulation

is lower in the model’s series. All in all, the empirical adaption for the
domestic economy is relatively close.
The data of the Euro area is replicated similarly close as the domes-
tic economy. The estimated foreign interest and inflation rate almost
coincide with the data series. The two series of the foreign output are
pro-cyclical. Again, the data series contains more short-run volatility
compared to their estimated counterpart.
4 Results - Inflation versus
Price-Level Targeting
The following sections compares inflation and price-level targeting using
the estimated model (see chapter 3). The comparison of inflation and
price-level targeting is based on impulse response functions for different
shocks and a stochastic simulation. The findings are compared to previ-
ous studies in the field of price-level targeting such as Svensson (1999)
and Vestin (2006).
4.1 Impulse Response Functions

The main results are provided in figure 4.1. The figure shows the re-
actions of the main series for monetary policy (interest rate, inflation
and output) to a domestic productivity, preference and monetary policy
shock. The shocks are normalized to one. The inflation targeting regime
is plotted by the solid line and the reactions under a price-level targeting
regime are given by the dashed line. Since the domestic economy is as-
sumed to be small, the response of foreign variables to domestic shocks
is negligible. This is the reason why no foreign variables are plotted.
Productivity Shock
Obviously, a positive technology shock increases output for a given labor
input. The direct response in period 0 is very similar in both regimes.
Afterwards, the output follows a more hump-shaped reaction under an in-
flation targeting regime. In both regimes the response is at its maximum

48 4 Results - Inflation versus Price-Level Targeting
after 5 quarters. Producers need some time to react to new technologies.

The effects fully disappear after more than 50 quarters for both regimes.
The inflation and price-level targeting regime react to an increase in
output with lower interest rates. The lower interest rates increase money
holdings that is used for the growing economic activity. The reaction of
the monetary authority is the strongest after 4 quarters in the inflation
targeting case and after 7 quarters in the price-level targeting case. The
effect vanishes after around 30 quarters for both regimes.
The inflation rates respond negatively to technological improvement.
The direct response in period 0 under price-level targeting is not even
half as big as the reaction under inflation-targeting. In the price-level
targeting policy, the reaction turn from negative into positive after 5
quarters, what can be interpreted as an overreaction. In the end, the
reaction to the inflation rate expires after around 15 quarters in both
regimes.
Preference Shock
The reactions to a positive preference shock are quite similar to reactions
due to a technology shock. A positive preference shock leads to a higher
demand for consumption by the households. The higher demand ends
up in higher economic activity. Therefore, output reacts positively to
a preference shock. The direct response is similar under both regimes.
The responses are hump-shaped and their maximum is reached after 6
quarters for the inflation and price-level targeting policy. Nevertheless,
the reaction in the price-level targeting case is much smaller than under
inflation targeting. Again, the effects are not persistent.
The inflation rate reacts stronger under the inflation targeting regime.
Under price-level targeting, the reaction of the inflation rate is quite
small. After 5 quarters a little overreaction becomes visible. After 15
quarters the responses disappear in both cases.
The deflationary pressure is answered with lower interest rates by the
monetary authorities under inflation and price-level targeting. In period
0, the reaction of the inflation targeting regime is stronger. The re-
actions follow a negative hump-shape with a maximum after 5 quarters
4.1 Impulse Response Functions 49
Figure 4.1: Impulse Response Functions

in the inflation targeting economy and after 10 quarters in the price-level

targeting economy. It takes more than 20 quarters for the interest rate to
fully implement the preference shock. The reactions become negligible
after 50 quarters.
Monetary Policy Shock

Since the monetary policy shock follows an first-order autoregressive
process (with ρi set to 0.92), the effect on the interest rate is constantly
decreasing. Because the assumption about the first-order autoregressive
process holds in both cases, the responses are very similar. The reactions
become negligible after around 12 quarters.
The inflation rate responds negatively to the increase in the interest
rate. However, the response under inflation targeting is much stronger
than under price-level targeting. In both cases, the reaction decreases
constantly and is close to 0 after 12 quarters. It looks similar to the
mirror-inverted interest rate process.
The reactions of output to the monetary policy shock are negatively
hump-shaped. The output impact of a change in the interest rate is
smaller under price-level targeting than under inflation targeting. Under
price-level targeting, the maximum is reached after 3 quarters, whereas
under inflation targeting the strongest response can be seen after 4 quar-
ters. Finally, the reaction is non-persistent in both cases.
To sum up, the reactions of the inflation rate are more accurate under
a price-level targeting regime. The discussed shocks lead to stronger re-
actions over the whole horizon under inflation targeting. The statement
for the changes in output is ambiguous. The responses to a productivity
and preference shock are stronger under price-level targeting. But the
reaction of output to a monetary policy shock is larger under inflation
targeting compared to price-level targeting.
The previous section discussed changes to only a selected few of the
shocks. The stochastic simulation provides a possibility to examining
the variability of variables while all shocks are randomly active.
4.2 Simulation 51
4.2 Simulation
In contrast to the historical simulation, the stochastic simulation is fully
based on random numbers. The model is filled with random shocks
using the estimated standard deviations. The simulation is calculated
for 10 000 periods. The first 100 observations are plotted in figure 4.2.
The figure shows the inflation rate, the output and the interest rate for
the domestic and the foreign economy. Again, the outcome of the model
using inflation targeting is illustrated by the solid line and the model
with price-level targeting by the dashed line.
Looking at the foreign economy, the plotted variables for the inflation
and price-level regime appear as one line. They perfectly coincide. This
is reasonable, because the domestic economy is modeled to have only a
negligible impact on the large foreign economy.
The domestic inflation has more extreme outliers under inflation tar-
geting than under price-level targeting. The short-term changes are
pro-cyclical for both regimes, but the variances are bigger for inflation
targeting. The inflation rate of the price-level targeting regime fluctu-
ates closer to its steady state value. The visible intuition is matched by
theoretical calculations. The variance over the whole simulation horizon
is much bigger for the inflation targeting economy (6.48) compared to
the price-level targeting economy (0.80).
As it can be seen from the policy rule, the monetary authority reacts
not only to changes in inflation but also to changes in output. The
differences between the two different regimes are lower for the output.
Again, the two series proceed pro-cyclical. After 45 periods and after
65 periods, it can be shown that the output is slightly farer away from
its steady state value under inflation targeting than under price-level
targeting. But over the whole simulation, the variance of the output is
slightly smaller using an inflation targeting rule (8.17) compared to a
price-level targeting rule (8.50).
The reactions of the monetary authority are described by the interest
rates. The series proceed pro-cyclical. The first 40 periods and the
last 30 periods of the domestic interest rate show that the inflation
targeting regime reacts with stronger decreases. The positive deviation

from the steady state value is similar for both policies, which can be
seen after 50 periods. This seems to be in line with Svensson (2001).
He argues that price-level targeting is more effective around the zero
lower bound. The calculated variances for the interest rates are higher
in the inflation targeting economy (3.10) than in the price-level targeting
economy (1.07). In other words, the monetary authority under inflation
targeting reacts stronger to disturbances.
4.2 Simulation 53
Figure 4.2: Simulation (first 100 observations)

5 Conclusion
The results of this thesis are based on a standard small open DSGE
model á la Monacelli (2005) and Galı́ and Monacelli (2005) with various
extensions, such as wage stickiness and habit formation. The monetary
policy rule of the domestic economy is distinguished in two cases: infla-
tion and price-level targeting. The inflation targeting regime reacts to
inflation, output and output growth. As its name implies, the price-level
targeting regime responds to changes in the price-level rather than the
inflation rate. Because of the lack of data, the model’s parameters are
only estimated for the inflation targeting regime. Finally, the parame-
ters are simply adapted to the price-level targeting case.
First results of the thesis are the estimated parameters of the model us-
ing Bayesian techniques. The parameter’s prior distributions are mainly
based on Robinson (2013) as well as Justiniano and Preston (2008). Con-
fidence intervals for the estimators are calculated using the random walk
Metropolis algorithm. Some parameters, such as the indexation param-
eters, are relatively different for Switzerland and the Euro area. The
elasticities of intertemporal substitution are close to 0. In most other
references, for example Cuche-Curti et al. (2009), the intertemporal
elasticity of substitution is assumed to be close to 1. The Taylor rule
for the domestic economy is only fulfilled by a narrow margin. The in-
flation parameter is just above 1. Cuche-Curti et al. (2009) assume the
parameter to be 1.5. The empirical relevance of the estimated model
is examined with a historical simulation. All series proceed pro-cyclical.
Main differences between the data and the model series can be found in
the short-term volatility of domestic inflation and foreign output. All in
all, the fit to the data is acceptable.

56 5 Conclusion
The estimated DSGE model is used for calculating impulse response

functions and running a stochastic simulation. The two regimes are
evaluated by looking at the pathway and the variances of key variables,
namely inflation rate, output, and interest rate. The impulse response
functions are computed for a domestic technology, preference and mon-
etary policy shock. Changes of the inflation rate to all shocks are more
accurate under price-level targeting than under inflation targeting. The
conclusion for the changes in output is ambiguous. The reactions to a
productivity and preference shock are heavier under price-level targeting,
whereas the response of the output to a monetary policy shock is larger
under inflation targeting. Using stochastic simulation, variances of all
variables can be calculated. Under price-level targeting, the variances
of the inflation and interest rate series are smaller than under inflation
targeting. On the other hand, the output varies a little more in the
price-level targeting model. This facts are in line with Svensson (1999)
and Vestin (2006) because both studies found lower variances of infla-
tion under a price-level targeting policy. Since inflation and output are
targeted simultaneously, neither of the two policies is strictly dominant.
Although, the difference in the volatilities of the output series is very
close compared to the difference in the volatilities of the inflation rates.
The thesis introduced a comparison of inflation and price-level tar-
geting using a simulation method. Previous studies were mainly based
on loss-function evaluations. The introduction of a zero lower bound on
interest rates would be a possible extension to this thesis. As Svensson
(2001) already discussed theoretically, the advantages of a price-level tar-
geting regime seem to increase at the zero lower bound. Furthermore,
this thesis only considered price stability as the monetary authority’s
main goal. Recent studies introduced other goals such as financial sta-
bility, for example White (2006).
References
Adolfson, M., S. Laséen, J. Lindé and M. Villani (2007), Bayesian estima-

tion of an open economy DSGE model with incomplete pass-through,
in: Journal of International Economics 72, p. 481-511.
Ambler, S. (2009), Price-Level Targeting and Stabilization Policy: A
Survey, in: Journal of Economic Surveys 23(5), p. 974-997.
An, S. and F. Schorfheide (2007), Bayesian Analysis of DSGE Models,
in: Econometric Reviews 26, p. 113-172.
Benigno, p. (2009), Price Stability with Imperfect Financial Integration,
in: Journal of Money, Credit and Banking 41, p. 121-149.
Berg C. and L. Jonung (1999), Pioneering Price Level Targeting: The
Swedish Experience 1931-1937, in: Journal of Monetary Economics
43(3), p. 525-551.
Bernanke, B. S. and M. Woodford (2005), The Inflation-Targeting De-
bate, The University of Chicago Press, Chicago and London.
Boldrin, M., L. J. Christiano and J. D. M. Fisher (2001), Habit Per-
sistence, Asset Returns, and the Business Cycle, in: American Eco-
nomic Review 91(1), p. 149-166.
Calvo, G. A. (1983), Staggered Prices in a Utility-Maximizing Frame-
work, in: Journal of Monetary Economics 12, p. 383-398.
Christiano, L. J., M. Eichenbaum and C. L. Evans (2005), Nominal
Rigidities and the Dynamic Effects of a Shock to Monetary Policy,
in: Journal of Political Economy 113(1), p. 1-45.

DOI 10.1007/978-3-658-08228-4, © Springer Fachmedien Wiesbaden 2015
58 References
Clarida, R., J. Galı́ and M. Gertler (1999), The Science of Monetary

Policy: A New Keynesian Perspective, in: Journal of Economic Lit-
erature 37, p. 1661-1707.
Constantinides, G. M. (1990), Habit Formation: A Revolution of the Eq-
uity Premium Puzzle, in: Journal of Political Economy 98(3), p. 519-
543.
Cuche-Curti, N. A., H. Dellas and J.-M. Natal (2009), DSGE-CH: A
dynamic stochastic general equilibrium model for Switzerland, Swiss
National Bank Economic Studies 2009(5).
Eggertson, G. B. and M. Woodford (2003), The Zero Bound on Interest
Rates and Optimal Monetary Policy, in: Brooking Papers on Eco-
nomic Activity 1, p. 139-211.
Erceg, C. J., D. W. Henderson and A. T. Levin (2000), Optimal Mon-
etary Policy with Staggered Wage and Price Contracts, in: Journal
of Monetary Economics 46(2), p. 281-313.
Fuhrer, J. C. (2000), Habit Formation in Consumption and Its Impli-
cations for Monetary-Policy Models, in: American Economic Review
90(3), p. 367-390.
Galı̀, J. and T. Monacelli (2005), Monetary Policy and Exchange Rate
Volatility in a Small Open Economy, in: Review of Economic Studies
72, p. 707-734.
Hamilton, J. D. (1994), Time series analysis, Princeton University Press,
1st edition, Princeton.
Justiniano, A. and B. Preston (2008), Can Structural Small Open Econ-
omy Models Account For The Influence of Foreign Disturbances?,
NBER Working Paper 14547.
Kollmann, R. (2002), Monetary Policy Rules in the Open Economy:
Effects on Welfare and Business Cycles, in: Journal of Monetary
Economics 49(5), p. 989-1015.
References 59
Krugman, P. R. (1998), It’s baaack: Japan’s Slump and the Return of

the Liquidity Trap, in: Brookings Papers on Economic Activity 2,
p. 137-205.
Lane, p. R. (2001), The new open economy macroeconomics: a survey,
in: Journal of International Economics 54(2), p. 235-266.
Levin, A. T. A., A. Onatski, J. C. Williams and N. Williams (2005),
Monetary Policy Under Uncertainty in Micro-Founded Macroecono-
metric Models, NBER Working Paper 11523.
Monacelli, T. (2005), Monetary Policy in a Low Pass-through Environ-
ment, in: Journal of Money, Credit and Banking 37(6), p. 1047-1066.
Robinson, T. (2013), Estimating and Identifying Empirical BVAR-
DSGE Models for Small Open Economies, Research Discussion Paper,
Reserve Bank of Australia.
Schorfheide, F. (2000), Loss-Function-Based Evaluation of DSGE Mod-
els, in: Journal of Applied Econometrics 15(6), p. 645-670.
Sims, C. A. (2002), Solving Linear Rational Expectations Models, in:
Computational Economics 20(1-2), p. 1-20.
Smets, F. and Wouters R. (2003), An Estimated Dynamic Stochastic
General Equilibrium Model of the Euro Area, in: Journal of the
European Economic Association 1(5), p. 1123-1175.
Smets, F. and R. Wouters (2007), Shocks and Frictions in US Business
Cycles: a Bayesian DSGE Approach, in: American Economic Asso-
ciation 97(3), p. 586-606.
Svensson, L. E. O. (1999), Price-Level Targeting Versus Inflation Tar-
geting: A Free Lunch?, in: Journal of Money, Credit and Banking
31(3), p. 277-295.
Svensson, L. E. O. (2001), The Zero Bound in an Open Economy: A
Foolproof Way of Escaping from a Liquidity Trap, in: Monetary and
Economic Studies 19(S1), p. 277-312.
60 References
Vestin, D. (2006), Price-Level Versus Inflation Targeting, in: Journal of

Monetary Economics 53(7). p. 1361-1376.
White, W. R. (2006), Is Price Stability Enough?, Bank for International
Settlements Working Paper 205, Bank for International Settlements,
Basel.
Woodford, M. (2003), Interest and Prices, Foundations of a Theory of
Monetary Policy, Princeton University Press, 1st edition, Princeton
and Oxford.
Appendix
1 Log-Linearized Equations
This is the complete set of the small open DSGE model’s equations.
They are transformed to the canonical form and used for running the
Bayesian estimation.
Foreign Economy
∗ h∗ ∗ 1 ∗ ∗1−h
∗ ∗ ∗

yt − ∗
yt−1 = ∗
Et [yt+1 ] − σ ∗
it − Et [πt+1 ]
1+h 1+h 1+h
∗
∗1−h ∗
−σ (1 − ρg )g,t (1)
1 + h∗
∗ 1 ∗ ∗ ∗
πt = (γH ∗ πt−1 + β Et [πt+1 ]
1 + β ∗ γH ∗

(1 − ξH ∗ )(1 − ξH ∗ β ∗ ) ∗ ∗ ∗ ∗
+ ωH ∗ yt + wt − (1 − ωH ∗ )a,t + cp,t ) (2)
ξH ∗ (1 + ωH ∗ θ ∗ )

w∗ w∗ ∗ (1 − ξw∗ )(1 − ξw∗ β ∗ ) ∗ ∗ ∗
πt =βEt [πt+1 − γw πt ] + (vt − wt ) + γw πt−1
ξw∗ (1 + ϕ∗ θ ∗ )

∗ ∗ σ ∗−1 ∗ ∗ ∗ 1 ∗ h∗ σ ∗−1 ∗ ∗
with vt ≡ ϕ + yt − ϕ a,t − g,t − y + n,t (3)
1 − h∗ σ ∗ 1 − h∗ t−1
∗ ∗ ∗
w t = πt
w∗
− πt − wt−1 (4)
∗ ∗ ∗
nt = yt − a,t (5)
∗ ∗
∗ ∗ ∗
∗
it = ρi∗ it−1 + (1 − ρi∗ ) θπ∗ πt + (θy∗ + θdy∗ )yt − θdy∗ yt−1 + i,t (6)
∗ ∗
∗ ∗ ∗
∗
it = ρi∗ it−1 + (1 − ρi∗ ) θπ∗ πt + (θy∗ + θdy∗ )yt − θdy∗ yt−1 + i,t (7)
∗ ∗ ∗ ∗
g,t = ρg g,t−1 + vg,t (8)

DOI 10.1007/978-3-658-08228-4, © Springer Fachmedien Wiesbaden 2015
62 Appendix
∗ ∗ ∗ ∗
a,t = ρa a,t−1 + va,t (9)
∗ ∗ ∗ ∗
n,t = ρn n,t−1 + vn,t (10)
Domestic Economy
h 1 1−h
ct = ct−1 + Et [ct+1 ] − σ (it − Et [πt+1 ] + σ(1 − h)(1 − ρg )g,t (11)
1+h 1+h 1+h
w
w
(1 − ξw )(1 − ξw β)
πt =β Et [πt+1 ] − γw πt + γw πt−1 + (vt − wt )
ξw (1 + ϕθ)
σ −1 −1
with vt ≡ ϕ(yt − a,t ) + (ct − hc−1 ) − σ g,t + n,t (12)
1−h

1 (1 − ξH )(1 − ξH β)
πH,t = (βEt [πH,t+1 ] + (wt
1 + βγH ξH (1 + ϕωp )
γ η−1
(1 − α) γ F
H
+ 1− γ η−1
st + ωp yt − (1 − ωp )a,t ) + γH πH,t−1 + ch,t )
α + (1 − α) γ F
H
1−η 1−η
1
1−α θ
θ−1 θ
with γH ≡ and γF ≡ (13)
1−α θ−1

θ (1 − ξF )(1 − ξF β) θ
πF,t = πF,t−1 + ψF,t + β Et [πF,t+1 ] − πF,t + cf,t (14)
θ−1 ξF θ−1
1 ∗
zt = (zt−1 + qt−1 ) − qt − ct − (1 − λ )ψF,t
β
γF η−1
(1 − α) γH
−
∗
1 − λ − (1 + η) γF η−1 st + y t
∗
(15)
α + (1 − α) γH
1 ∗ ∗

qt = Et [qt+1 ] − (it − Et [πt+1 ]) + it − Et [πt+1 ] − χzt + rp,t (16)
1+χ
w
wt = πt − πt − wt−1 (17)
nt = yt − a,t (18)
it = ρi it−1 + (1 − ρi ) (θπ πt + (θy + θdy )yt − θdy yt−1 ) − i,t (19)
πt = πH,t + αΔst (20)
g,t = ρg g,t−1 + vg,t (21)
Log-Linearized Equations 63
a,t = ρa a,t−1 + va,t (22)

n,t = ρn n,t−1 + vn,t (23)
rp,t = ρrp rp,t−1 + vrp,t (24)
cf,t = ρcf cf,t−1 + vcf,t (25)
Market Clearing
−η
−η

(1 − α)γH ∗ (1 − α)γH ∗
yt = 1− −η −1 −η
yt + 1− −η −1 −η
λ
(1 − α)γH + αγH γF (1 − α)γH + αγH γF
γF η−1
−η
(1 − α)γH (1 − α) γH
− −η −1 −η
η γ η−1 −1 st
(1 − α)γH + αγH γF α + (1 − α) F
γH
−η
−η

(1 − α)γH ∗ (1 − α)γH
+ 1− −η −1 −η
λ ψF,t + −η −1 −η
ct (26)
(1 − α)γH + αγH γF (1 − α)γH + αγH γF
Definitions
st − st−1 = πF,t − πH,t (27)
Δst ≡ st − st−1 (28)
γ η−1
(1 − α) F
γH
qt = ψF,t + γF η−1 st (29)
α + (1 − α) γH
Δqt = qt − qt−1 (30)
Rational Expectations
∗ ∗
yt = Et−1 [yt ] + ηt
y∗
(31)
∗ ∗
πt = Et−1 [πt ] +
π∗
ηt (32)
w∗
πt =
w∗
Et−1 [πt ] +
πw∗
ηt (33)
ct = Et−1 [ct ] +
c
ηt (34)
πt = Et−1 [πt ] + ηt
π
(35)
πH,t = Et−1 [πH,t ] + ηt
πH
(36)
64 Appendix
πF,t = Et−1 [πF,t ] + ηt

πF
(37)
w w
πt = Et−1 [πt ] + ηt
πw
(38)
qt = Et−1 [qt ] +
q
ηt (39)
Shape of the Log-Likelihood Functions 65
2 Shape of the Log-Likelihood Functions
Figure 1: Log-Likelihood Functions 1

66 Appendix


68 Appendix


Inflation Versus Price-Level Targeting - Bayesian Estimation of A Small Open DSGE Model For Switzerland-Gabler Verlag (2015)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Inflation Versus Price-Level Targeting - Bayesian Estimation of A Small Open DSGE Model For Switzerland-Gabler Verlag (2015)

Uploaded by

Copyright:

Available Formats

BestMasters

Springer awards “BestMasters” to the best application-oriented master’s theses,

Library of Congress Control Number: 2014956973

Printed on acid-free paper

Springer Gabler is a brand of Springer Fachmedien Wiesbaden

Eggertson and Woodford (2003), have highlighted how a comparatively

A limitation of this literature, as it currently stands, is that the over-

Bern, in July 2014 Prof. Dr. Luca Benati

2 Description of the Model 5

3 Bayesian Estimation of the Parameters 31

4 Results - Inﬂation versus Price-Level Targeting 47

L. Heim, Inflation versus Price-Level Targeting, BestMasters,

His ﬁndings show that price-level targeting outperforms inﬂation target-

a price-level targeting rule, output reacts stronger to technological and

L. Heim, Inflation versus Price-Level Targeting, BestMasters,

2.1 Domestic Economy

Ht describes the habit formation. It is not regarded as a choice vari-

with φt (At ) = exp [−χ(At + rp,t )]

− λt+s (Pt+s Ct+s + Dt+s+1 + Xt+s Bt+s − Rt+s−1 Dt+s−1

Taking the derivatives with respect to Ct , Nt , Bt and Dt yields to the

Furthermore, the budget constraint (derivative with respect to λt )

−1/σ ∗ Xt+1 Pt −1/σ

The parameter η is assumed to be positive: η > 0. α indicates the

θ is the elasticity of substitution and assumed to be positive: θ > 0.

min Pt Ct − PH,t CH,t − PF,t CF,t

Replacing Ct by its aggregation index, it becomes an unconstrained

The corresponding optimality conditions are given by equations (2.16)

Secondly, the same optimization procedure is followed for domestic

min Pj,t Cj,t − Pj,t (i)Cj,t (i) di

The Lagrangian for the unconstrained problem is given by equation

with j = {H, F } (2.22)

log Wt (k) = log Wt−1 (k) + γw πt−1 (2.28)

Wt (k) is the wage of household k. γw denotes the degree of indexa-

n,t Nt+s (k)1+ϕ

As the following Lagrangian shows, the optimization problem become

This yields to the following optimality condition.

Yt (j) = a,t f (Nt (j)) (2.32)

The function f (.) is assumed to satisfy the Inada conditions. a,t is

θw is assumed to be greater than 1: θw > 1. Similar to the households,

Replace Nt (j) by the CES aggregator. The unconstrained problem is

The FOC is derived by setting the derivative of the Lagrangian with

Solving this expression for Nt (k) yields to the producer’s demand

log PH,t (j) = log PH,t−1 (j) + γH πH,t−1 (2.38)

γH deﬁnes the degree of indexation and is limited as follows: 0 ≤ γH ≤

This yields to the following FOC.

2.1.3 Retail Firms

Solving the problem yields to the following optimality condition.

2.1.4 Exogenous Processes

2.2 Foreign Economy

domestic country are negligible. The small domestic economy is engaged

External habit formation is described by Ht∗ . Ht∗ = h∗ Ct−1

The household’s maximization problem can be solved using the La-

Additionally, the derivative with respect to the Lagrangian multiplier

θ∗ describes the elasticity of substitution and is assumed to be positive:

The constraint is directly plugged into the optimization problem. The

The degree of adjustment to the previous period’s inﬂation is denoted

Making use of the demand function, the maximization problem be-

As for the domestic producers, the production function f ∗ (.) is as-

Parameter θw∗ is constrained by the following unequality: θw∗ > 1.

Producer j’s demand function for labor input of type k is derived by

γF ∗ measures the degree of adjustment to past inﬂation rates. The

2.2.3 Monetary Authority

 R∗ ρi∗  θπ∗  θy∗  θΔy∗ 1−ρi∗

R∗ ρi∗ θπ∗ θy∗ θΔy∗ 1−ρi∗