Professional Documents
Culture Documents
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
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.
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.
1 Introduction 1
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.
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
∗
Pt Ct + Dt + Xt Bt =Rt−1 Dt−1 + Xt−1 Rt−1 φt (At )Bt−1 + Wt Nt + Tt
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
∂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
8 2 Description of the Model
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
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)
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
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)
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)
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
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
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
2.1 Domestic Economy 11
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:
∞ γw
s Pt+s−1
max Et (ξw β) λs W̄t (k) Nt+s (k)
{W̄t (k)} Pt−1
s=0
∞ 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
∞
∂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.
2.1 Domestic Economy 13
Labor Demand
The domestic goods from firm j, Yt (j), are produced according to the
given production function.
1
θ θw−1
θw −1 w
Nt (j) = Nt (k) θw (2.33)
0
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
1
θ θw−1 1
θw −1 w
L = Wt Nt (k) θw − Nt (k)Wt (k) dk (2.35)
0 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
14 2 Description of the Model
−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.
∞ 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).
2.1 Domestic Economy 15
∞
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
∞ 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
∞ 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
16 2 Description of the Model
∗
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
∞ 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
2.2.1 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
∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗
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.
18 2 Description of the Model
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∗
∗ ∗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
2.2 Foreign Economy 19
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
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
20 2 Description of the Model
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)
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
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)
∗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
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
−θw∗
∗ Wt∗ (k) ∗
Nt (k) = Nt (j) (2.68)
Wt∗
2.2 Foreign Economy 23
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 ∗
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.
24 2 Description of the Model
∞
∗ 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
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
∗
YH,t = CH,t + CH,t (2.74)
∗ ∗
Yt = Ct (2.75)
P ∗ −λ
∗ H,t ∗
CH,t = Yt (2.76)
P∗
Terms of Trade
The effective terms (st ) of trade are defined by equation (2.77).
PF,t
St ≡ (2.77)
PH,t
Xt Pt∗
ΨF,t ≡ (2.78)
PF,t
Xt Pt
Ωt ≡ (2.79)
Pt∗
∗ Xt
Et [λt+1 Pt+1 ] Rt − Rt rp,t =0 (2.80)
Et [Xt+1 ]
28 2 Description of the Model
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.
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
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-
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
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 )
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
∗
∗ ∗
4
Inf lationt = { ln(CP It ) − ln(CP It−1 ) + 1 − 1} × 100 (3.11)
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.
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).
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.
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
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∗ ).
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).
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
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
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
52 4 Results - Inflation versus Price-Level Targeting
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)
∗ ∗ ∗ ∗
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
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
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