143 ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 | 143-212
A DSGE MODEL FOR THE SLOVENIAN 
ECONOMY: MODEL ESTIMATES AND 
APPLICATION
1
 Received: June 29, 2016 
ANDREJ KUŠTRIN
2
 Accepted: February 24, 2019
ABSTRACT: The paper presents the estimation of a dynamic stochastic general equilibrium 
(DSGE) model for the Slovenian economy and its applications. The model, which is built in 
the tradition of New Keynesian models, closely follows the structure of the model developed by 
Adolfson et al. (2007) and Masten (2010). W e estimate the model using a Bayesian method on 
quarterly Slovenian macroeconomic data covering the period 1995-2014. Beyond evaluating 
the properties of the estimated model, we discuss the role of various shocks in explaining 
macroeconomic fluctuations in the Slovenian economy to illustrate the model’s potential in 
structural business cycle analysis.
Key words: DSGE models, Bayesian methods, business cycle
JEL classification: C11, E32
DOI: 10.15458/ebr.87
INTORDUCTION
New-Keynesian dynamic stochastic general equilibrium (DSGE) models have recently 
become a standard tool for macroeconomic analysis. The key feature of this class of models 
is that they are derived from the microeconomic foundations meaning that they assume 
optimizing agents which usually form rational expectations and maximize their objective 
functions subject to their respective constraints in the presence of imperfect competition 
and nominal rigidities.
3
 In recent years there have been many theoretical and empirical 
contributions developing and estimating DSGE models. The most influential papers in 
this area include Clarida et al. (1999, 2001), Beningo & Beningo (2003), Galí & Monacelli 
(2005), Christiano et al. (2005), Smets & Wouters (2003, 2007), Adolfson et al. (2007) and 
many others. 
* I am grateful for guidance and helpful comments from my supervisor Igor Masten. For helpful comments on 
earlier versions of this paper, I would also like to thank the editor and two anonymous referees. Last, I would 
like to acknowledge financial support from the Slovenian Research A gency. All remaining errors are my own 
responsibility.
1
 University of Ljubljana, Faculty of Economics, Young Researcher, Ljubljana, Slovenia, e-mail: andrej.kustrin@
ef.uni-lj.si. 
2
 For further information on the New-Keynesian models, see Galí (2008) and Woodford (2003).
*
1
2

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 144
area include Clarida et al. (1999, 2001), Beningo & Beningo (2003), Galí & Monacelli
(2005), Christiano et al. (2005), Smets & W outers (2003, 2007), Adolfson et al. (2007)
and many others.
Although the literature on the estimation of DSGE models and the subsequent use of these
models to study macroeconomic fluctuations in various countries has rapidly expanded in
recent years, no attempt has as yet been made to estimate a New-Keynesian DSGE model
for Slovenia, at least to the best of our knowledge.
2
This paper therefore seeks to fill this
gap by presenting an estimated DSGE model for the Slovenian economy .
The model that we use was inspired in the work of Adolfson et al. (2007) and Masten
(2010). Masten (2010) extended the baseline model of Adolfson et al. (2007) in two di-
rections, namely by (i) adapting the model to the small open economy case within the
euro area and (ii) enriching the fiscal block of the model. W e use a Bayesian approach
to estimate key model parameters on 15 time series for Slovenia: GDP , consumption, in-
vestment, exports, imports, government consumption, real ef fective exchange rate, real
wage, employment, GDP deflator , CPI price index, short-run interest rate, and three for -
eign variables (that is output, inflation and interest rate), which refer to the first 12 euro
area countries.
W ith this paper we want to contribute to the lar ge literature on estimated DSGE models
by applying the Bayesian method to the estimation of the DSGE model for the Slovenian
economy and therefore presenting evidence for an additional country on the estimates of
the structural parameters, and by identifying the shocks responsible for the recent recession
and the key sources of macroeconomic fluctuations in Slovenia.
After the estimation, we first present our estimates of the structural parameters. W e then
perform several checks of the model’ s empirical performance. Specifically , we evaluate
how well the model fits the data. T o do so, we compare the actual data with the one-sided
predicted values from the model. Next, we calculate statistics of the data generated by
the estimated model and compare them with those based on the actual data. Finally , we
look at the smoothed estimates of the shock innovation paths to check whether they look
stationary . In the last part of the paper , we apply the estimated DSGE model to analyse the
contribution of the structural shocks on business cycle fluctuations in the Slovenian econ-
omy . W e proceed here in three steps. First, using traditional impulse response analysis,
we look at the partial ef fects of the most important shocks included in the model on key
2
Despite their wide use, DSGE models have also certain drawbacks. The most problematic issues which are
currently much discussed in the literature are mainly concerned with: (i) unrealistic assumptions (e.g. Ricardian
equivalence, rational expectations hypothesis, infinitely-lived households, ...), (ii) unconvincing method of esti-
mation (which is a combination of calibration and Bayesian estimation), (iii) questionable assumption about the
structural parameters that are assumed to be invariant to policy changes, (iv) issue related with the use of revised
or real-time data when estimating the model, and (v) poor performance during the recent crisis. For more de-
tailed discussion of these issues, see Romer (2016), Blanchard (2016) and the other contributions (see Blanchard
(2017) for an extensive list of references). Despite these shortcomings, we decided to use a DSGE framework as
we believe that it is flexible enough to be used for our purposes, while other models are more limited in terms of
their ability to fully address the research questions under study .
2
2.1
2.1

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 145
macroeconomic variables. Second, to assess how much of the volatility of the observed
variables can be explained by the shocks included in the model, we also produce variance
decomposition analysis. Finally , we compute historical decompositions of GDP growth
and its main components in terms of various structural shocks of the model to examine
the importance of respective shocks in explaining the observed macroeconomic dynamics
over the sample period, with particular attention to the recent recessionary periods.
Previewing the results, we find that investment-specific technology shocks mostly ac-
counted for a significant portion of the drop in output growth from 2008 onwards. This
result accords with a drop in foreign and domestic orders followed by a decline in invest-
ment (mostly at the beginning of the crisis) and the lar ge amount of losses of the corporate
sector that accumulated on balance sheets of the banks in the form of non-performing bad
loans, further contributing to a contraction of lending activity , which in turn reduced invest-
ment and impeded economic activity . Furthermore, consumption preference and export
mark-up shocks were another sources that contributed negatively to GDP growth, most
likely reflecting the reduction in households’ income (in combination with the precaution-
ary saving) and the fall in exports due to the deterioration of external competitiveness, as
wages increased faster than productivity before the crisis years, respectively . The results
also suggest that fiscal shocks had a stimulating impact during the first stage of the cri-
sis. However , starting from 2010 there was a turnaround in fiscal policy due to austerity
measures adopted to consolidate public finances. The slowdown in GDP growth was also
accompanied by permanent (unit-root) technology shocks that could be considered as as-
sociated with the lack of productivity-enhancing and other structural reforms in the run-up
to the crisis. By contrast, the historical decomposition suggests that transitory (stationary)
technology shocks were stimulative for GDP growth from 2013 onwards, which may be
interpreted as resulting from a temporary greater tendency of the corporate sector to take
restructuring measures in response to the crisis to enhance its technology and production
ef ficiency . Finally , our results show that the recovery phase after 2013 is explained in
our model mainly by consumption preference shocks, which could be explained by the
increased consumer confidence, resolution of banking system problems, as well as by the
improvement in the labour market situation.
The rest of the paper is or ganized as follows. Section 2 presents the model. Section 3
presents the estimation methodology and discusses the calibration of the model, the choice
of priors and presents the data used in the estimation. Section 4 contains the estimation
results and evaluation, which are followed by an analysis of the impulse responses of
the various structural shocks and their contribution to the developments in the Slovenian
economy in Section 5. Section 6 concludes with a summary of the main findings.
2 THE MODEL
As mentioned in the introduction, to describe the Slovenian economy we use a DSGE
model presented in Adolfson et al. (2007) and Masten (2010), which is an extended ver -
3

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 146
sion of basic closed-economy new-Keynesian models, including the benchmark models
of Christiano et al. (2005), Altig et al. (201 1), and Smets & W outers (2003, 2007). The
model economy consists of households, domestic goods producing firms, importing con-
sumption and importing investment firms, exporting firms, a government which conducts
fiscal policy , and an exogenous foreign economy . As it is common in the DSGE literature,
the model incorporates several real and nominal rigidities, such as habit persistence in con-
sumption, variable capacity utilization of capital and investment adjustment costs, as well
as the price and wage stickiness. The stochastic dynamics of the model is driven by sixteen
exogenous structural shocks. The shocks considered are: permanent (unit-root) technol-
ogy , transitory (stationary) technology , investment-specific technology , markup shocks
(domestic, imported consumption, imported investment and export markup shocks), con-
sumption preference and labour supply shocks, asymmetric technology , risk premium, for -
eign V AR shocks (foreign output, inflation and interest rate shocks) and fiscal shocks (rate
of transfers to households and government spending shocks). One feature of the model
worth noting is that it includes a stochastic unit-root technology shock, which implies a
common trend in the real variables of the model. Consequently , the model can be esti-
mated with raw data without any pre-filtering. In the following we summarize the main
features of the model. T o this end we follow quite closely the mode of presentation from
Section 2 of Adolfson et al. (2014).
3
2.1. Supply side of the economy
2.1.1. Domestic firms
The domestic firms use labour together with capital to produce intermediate goods Y
i
,
which are sold to the final good producer . The production function of the final good firm
is of the Dixit-Stiglitz form:
Y
d
t
=
[∫
1
0
(
Y
d
i,t
)
1
λ
d
t
di
]λ
d
t
,λ
d
t
≥1, (1)
where λ
d,t
is a stochastic process determining the time-varying markup in the domestic
goods market. The final good producers operate in a perfectly competitive environment,
taking the prices of the intermediate goods P
d
i,t
and final goods P
d
t
as given. The pro-
duction function for each intermediate good firm i which operates under monopolistic
competition is of the Cobb-Douglas type:
Y
i,t
= z
1−α
t
ϵ
t
K
α
i,t
H
1−α
i,t
−z
t
ϕ, (2)
whereH
i,t
denotes homogeneous labour input hired by firmi, andK
i,t
is the amount of
capital services used by firmi , which can dif fer from capital stock since the model assumes
3
The detailed description of the model (including the first order conditions) is available in Adolfson et al.
(2007).
4

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 147
a variable capital utilization rate. Furthermore, z
t
is a permanent (unit-root) technology
shock, whereas ϵ
t
is a transitory (stationary) technology shock. The term z
t
ϕ indicates
fixed costs, which grow with the technology rate. Fixed costs are set in such a way that
profits are zero in steady state. Cost minimization yields the following nominal mar ginal
cost function for intermediate firmi:
MC
d
t
=
(
1
1−α
)
1−α
(
1
α
)
α
(
R
k
t
)
α
W
1−α
t
1
z
1−α
t
1
ϵ
t
, (3)
where R
k
t
is the gross nominal rental rate per unit of capital, R
t
is the gross nominal
interest rate, andW
t
is the nominal wage rate per unit of aggregate, homogeneous labour
H
i,t
. Besides solving the cost minimization problem, intermediate good firms have to
decide on price for their output. The model assumes the Calvo type staggered-price setting.
This means that at each period, each firm faces a random probability (1−ξ
d
) that it can
reoptimize its price. The reoptimized price is denotedP
d,opt
t
. W ith probabilityξ
d
a firm
is not allowed to set its prices optimally , and its price is then set according to the following
indexation rule (Smets & W outers, 2003):
P
d
t+1
=
(
π
d
t
)
κ
d
P
d
t
,
whereπ
d
t
= P
d
t
/P
d
t−1
is the (gross) inflation rate andκ
d
is an indexation parameter . The
optimization problem of a firm setting a new price in periodt is the following:
max
P
d,opt
t
E
t
∞
∑
s=0
(βξ
d
)
s
υ
t+s
[ (
π
d
t
π
d
t+1
...π
d
t+s−1
)
κ
d
P
d,opt
t
Y
d
i,t+s
−MC
d
i,t+s
(
Y
d
i,t+s
+z
t+s
ϕ )
]
, (4)
where (βξ
d
)
s
υ
t+s
denotes the stochastic discount factor , which is used to make profits
conditional upon utility . β is the discount factor , and υ
t
denotes the mar ginal utility of
households’ nominal income in periodt+s , which is exogenous to the intermediate firms.
2.1.2. Importing and exporting firms
The importing sector consists of two types of firms: firms which import consumption
goods and firms which import investment goods. There is a continuum of importing firms,
indexed byi∈ (0,1) . These firms buy a homogeneous good in the world market at price
P
∗
t
and transform it into a dif ferentiated consumption C
m
i,t
or investment good I
m
i,t
. In
addition, there is also a continuumi∈ (0,1) of exporting firms that buy a homogeneous
good on the domestic market and transform it into a dif ferentiated exported good which
is sold on the foreign market. The mar ginal cost of importing and exporting firms areP
∗
t
and P
t
, respectively . The aggregate import consumption, import investment and export
good is a composite of a continuum of i dif ferentiated imported consumption, imported
investment and exported goods, each supplied by a dif ferent firm, which follows the CES
function:
C
m
t
=
[∫
1
0
(
C
m
i,t
)
1
λ
m,c
t
di
]λ
m,c
t
,I
m
t
=
[∫
1
0
(
I
m
i,t
)
1
λ
m,i
t
di
]λ
m,i
t
,
5

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 148
X
t
=
[∫
1
0
(X
i,t
)
1
λ
x
t
di
]λ
x
t
, (5)
where1≤λ
j
t
<∞ forj ={mc,mi,x} is the time-varying flexible-price mark-up in the
import consumption (mc ), import investment (mi ) and export (x ) sector . The model as-
sumes monopolistic competition among importers and exporters and Calvo-type staggered
price setting. The price setting problems are completely analogous to that of the domes-
tic firms in Equation ( 4). From the optimization problems four specific Phillips curves,
determining inflation in the domestic, import consumption, import investment and export
sectors, can be derived.
2.2. Demand side of the economy
2.2.1. Households
In the model economy there is also a continuum of households j ∈ (0,1) , which attain
utility from consumption and leisure. The households decide on their current level of
consumption and their domestic and foreign bond holdings. They also choose the level of
capital services provided to the firms, their level of investment and their capital utilization
rate. The households can increase their capital stock by investing in additional physical
capital, taking one period to come in action, or by directly increasing the utilization rate
of the capital at hand. Thej
th
household’ s utility function is:
E
j
0
∞
∑
t=0
β
t
[
ζ
c
t
ln(C
j,t
−bC
j,t−1
)−ζ
h
t
A
L
(h
j,t
)
1+σL
1+σ
L
]
, (6)
whereC
j,t
andh
j,t
denotes levels of real consumption and labour supply of householdj ,
respectively . A
L
is a constant representing the weight that the worker attaches to disutility
of work. The model also allows for habit formation in consumption by includingbC
j,t−1
.
ζ
c
t
andζ
h
t
are preference shocks, consumption preference shock and labour supply shock,
respectively . The aggregate consumptionC
t
is a CES index of domesticC
d
t
and imported
C
m
t
consumption goods:
C
t
=
[
(1−ω
c
)
1/ηc
(
C
d
t
)
(ηc−1)/ηc
+ω
1/ηc
c
(C
m
t
)
(ηc−1)/ηc
]
ηc/(ηc−1)
, (7)
whereω
c
is the share of imported consumption goods in total consumption, andη
c
is the
elasticity of substitution between domestic and imported consumption goods. The corre-
sponding consumer price index is given by:
P
c
t
=
[
(1−ω
c
)
1/ηc
(
P
d
t
)
1−ηc
+ω
1/ηc
c
(P
m,c
t
)
1−ηc
]
1/(1−ηc)
. (8)
The model also assumes that households can purchase investment goods in order to in-
crease their capital stock. The law of motion of capital is given by:
¯
K
t+1
=(1−δ)
¯
K
t
+Υ
t
F (I
t
,I
t−1
)+∆
t
, (9)
6

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 149
where
¯
K
t
is a physical capital stock,δ is the depreciation rate of capital stock,F (I
t
,I
t−1
)
is a function that transforms investment into capital. Following Christiano et al. (2005),
F (I
t
,I
t−1
) is of the following form:
F (I
t
,I
t−1
)=
[
1−
˜
S(I
t
,I
t−1
)
]
I
t
, (10)
where
˜
S determines the investment adjustment costs through the estimated parameter
˜
S
′′
.
Υ
t
denotes the investment-specific technology shock and∆ t
represents either newly bought
capital if it is positive or sold capital if it is negative. The investment (I
t
) is a bundle be-
tween domestic and imported investment goods (I
d
t
andI
m
t
, respectively):
I
t
=
[
(1−ω
i
)
1/ηi
(
I
d
t
)
(ηi−1)/ηi
+ω
1/ηi
i
(I
m
t
)
(ηi−1)/ηi
]
ηi/(ηi−1)
, (1 1)
where ω
i
denotes the share of imported investment goods in total investment, and η
i
is
elasticity of substitution between domestic and imported investment goods. It is worth
noting that domestically produced consumption and investment goods have the same price
P
d
t
. The aggregate investment price index is therefore given by:
P
i
t
=
[
(1−ω
i
)
(
P
d
t
)
1−ηi
+ω
i
(
P
m,i
t
)
1−ηi
]
1/(1−ηi)
. (12)
Furthermore, the model assumes that each household is a monopolistic supplier of dif-
ferentiated labour service, which implies that they can determine their own wage. Each
household sells its labourh
j,t
to a firm which transforms it into a homogeneous input good
H
t
according to the following production function:
H
t
=
[∫
1
0
(h
j,t
)
1
λw
dj
]λw
,λ
w
≥1, (13)
whereλ
w
is the wage markup. The demand function for each dif ferentiated labour service
is given by:
h
j,t
=
[
W
j,t
W
t
]
λw
1−λw
H
t
. (14)
Following Erceg et al. (2000) and Christiano et al. (2005), the households are subject to
the Calvo wage rigidities, which means that in every period each household faces a random
probability 1−ξ
w
that it can change its nominal wage. If a household is allowed to re-
optimize its wage, it will set its wage toW
opt
t
taking into account the probabilityξ
w
that
the wage will not be re-optimized in the future. The households that cannot re-optimize
set their wages according to the following indexation rule:
W
j,t+1
=(π
c
t
)
κw
μ
z,t+1
W
j,t
, (15)
where κ
w
is an indexation parameter , π
c
t
is the inflation rate measured by the consumer
price index, andμ
z,t
=z
t
/z
t−1
is the growth rate of the unit-root technology shock. The
7

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 150
householdj that can re-optimize solves the following optimization problem:
max
W
opt
j,t
E
t
∞
∑
s=0
(βξ
w
)
s





−ζ
h
t+s
A
L
(hj,t+s)
1+σ
L
1+σL
+v
t+s
(
1−τ
y
+τ
tr
t+s
)(
π
c
t
...π
c
t+s−1
)
κw
×(µ z,t+1
...µ
z,t+s
)W
opt
j,t
h
j,t+s





, (16)
whereτ
y
is a labour income tax andτ
tr
t
is a time-varying rate of social transfers to house-
holds defined in more detail in Subsection 2.4.
2.3. Monetary policy
The monetary policy is modelled in a highly simplified way . It is assumed that the domestic
interest rate (R
t
) depends on the exogenously given foreign interest rate (R
∗
t
) adjusted for
the risk premium on foreign bonds (Φ
(
a
t
,
˜
ϕ t
)
):
4
R
t
= R
∗
t
Φ
(
a
t
,
˜
ϕ t
)
, (17)
where the risk premium, which is defined through the following function:
Φ
(
a
t
,
˜
ϕ t
)
= e
−
˜
ϕ a(at−¯ a)+
˜
ϕ t
, (18)
depends on the aggregate net foreign asset position of the domestic economy (a
t
) and
exogenous risk premium shocks (
˜
ϕ t
).
5
The inclusion of risk premium is necessary to
ensure a well-defined steady state in the model (Schmitt-Grohé & Uribe, 2003).
4
In Adolfson et al. (2007), the interest rate is determined according to a simple rule (expressed in log-linear
form):
ˆ
Rt = ρ
R
ˆ
R
t−1
+(1−ρ
R
)
(
ˆ
¯ π
c
t
+rπ
(
ˆ π
c
t−1
−
ˆ
¯ π
c
t
)
+ryˆ y
t−1
+rxˆ x
t−1
)
+r
∆π
∆ˆπ
c
t
+r
∆y
∆ˆyt +ε
R,t
,
where
ˆ
Rt is the short-rate interest rate,ˆ π
c
t
the CPI inflation rate,
ˆ
¯ π
c
t
a time-varying inflation tar get,ˆ yt the output
gap, ˆ xt denotes the real exchange rate, andε
R,t
is an interest rate shock.
5
Besides joining the European Monetary Union (EMU) in 2007, Slovenia went through dif ferent monetary
regimes since its independence in 1991: money based stabilization policy (1991-1995), price and real exchange
rate stability dual tar geting policy (1996-2001), and exchange rate based stabilization policy and accession to
ERM 2 exchange rate mechanism and EMU (2001-2007) (Caprirolo & Lavrač, 2003). Because these would be
very dif ficult to implement in the model, we are not modelling any break in the conduct of monetary policy when
estimating the model. Rather , we follow an uncomplicated way of monetary policy inclusion into the model
structure. That is, we keep only a modified UIP condition (17) on the monetary side without specifying any
particular form of monetary rule for the period before 2007, when Slovenia run an independent monetary policy .
Of course, for the years before 2007 (or at least 2004Q3, when Slovenia entered the ERM 2), an additional term,
−∆S
t+1
, capturing the nominal exchange rate fluctuations, must be added on the right-hand side of Equation
(17). In addition, the terms of trade channel would be af fected by the nominal exchange rate. In such a setting
the endogenously determined nominal exchange rate may reduce or amplify the impact of structural shocks,
depending on their nature. Based on the findings presented in Cúrdia et al. (2012), applying a more adequate
approach to modelling monetary policy may also improve the fit of the model. W e initially included exchange
rate as an additional variable into the model, but due to model solvability problems (i.e., more variables than
equations) we again restricted with the simple version, as described above. In other words, in order to guarantee
solvability of the model, an explicit monetary policy rule must be incorporated into the model structure. Despite
8

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 151
2.4. Fiscal policy
The government in this economy collects taxes, issues debt and uses revenues for govern-
ment consumption, transfers to households and interest on outstanding debt. The resulting
government budget constraint can be expressed as:
B
t+1
+T
t
=R
t−1
B
t
+TR
t
+P
d
t
G
t
⇔B
t+1
=B
t
+DEF
t
, (19)
whereB
t
denotes the public debt andDEF
t
is the government deficit, which is defined
as the dif ference between the government expendituresGEX
t
and total tax revenuesT
t
:
DEF
t
=GEX
t
−T
t
. (20)
The government tax revenues consist of taxes on private consumption, as well as on labour
income and capital income:
T
t
=τ
c
P
c
t
C
t
+τ
y
W
t
H
t
+τ
k
[
(R
t−1
−1)B
t
+R
k
t
K
t
+
(
R
∗
t−1
Φ
(
a
t−1
,
˜
ϕ t−1
)
−1
)
B
∗
t
+Π
t
]
,
(21)
withτ
c
,τ
y
andτ
k
being the tax rates on private consumption, labour income and capital
income, respectively , which are assumed to be fixed. In the above expression,Π
t
are total
profits, which are equal to the sum of profits earned by domestic, importing and exporting
firms,Π
d
t
,Π
m
t
andΠ
x
t
, respectively:
Π
t
=Π
d
t
+Π
m
t
+Π
x
t
, (22)
where:
Π
d
t
=P
d
t
(
C
d
t
+I
d
t
+G
t
)
+P
d
t
(C
x
t
+I
x
t
)
−MC
d
t
(
C
d
t
+I
d
t
+G
t
+C
x
t
+I
x
t
)
−MC
d
t
z
t
ϕ (23)
Π
m
t
=P
m,c
t
C
m
t
+P
m,i
t
I
m
t
−P
∗
t
(C
m
t
+I
m
t
) (24)
and:
Π
x
t
=P
x
t
(C
x
t
+I
x
t
)−P
d
t
(C
x
t
+I
x
t
). (25)
these simplifications, the model in such a structure fits the data, including the short-term nominal interest rates,
reasonably well. It is also worth noting that similar approach neglecting existence of diverse monetary policies
and flexible exchange rates prior to the EMU-start was used in the literature (see, for example, Adolfson et
al. (2007), Almeida (2009), Smets & W outers (2003), Marcellino & R ychalovska (2014) among others). The
authors estimated their models under implicit assumption that, even before the establishment of the currency area
there was a common monetary policy in the European Union. Finally , as a robustness check we re-estimated our
model using the data for the period 2004Q3 onwards, when Slovenia entered the ERM 2. Our analysis reveals
that in this case parameter estimates do not substantially vary from the estimates reported in T ables 3 and 4 in
the main text. But what is more important, we find that our main results reported in the paper (e.g., those of the
historical decompositions) persist. W e choose not to report this robustness check in the paper to save space, but
it is available upon request from the author .
9

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 152
Furthermore, the government expenditures are given by:
GEX
t
=TR
t
+P
d
t
G
t
+(R
t−1
−1)B
t
, (26)
whereTR
t
denotes transfers to households,G
t
is government consumption of goods and
services and(R
t−1
−1)B
t
stands for public debt interest payments. W e assume that trans-
fers to households are indexed to wages W
t
and hours worked H
t
with an exogenously
given rate of transfersτ
tr
t
according to the following expression (D’Auria et al., 2009):
TR
t
=τ
tr
t
W
t
H
t
. (27)
For the rate of transfers to households it is simply assumed that follow an AR(1) process
(in deviations from its steady state):
ˆ τ
tr
t
=ρ
τ
trˆ τ
tr
t−1
+ε
τ
tr
,t
. (28)
Finally , government consumption follows the log-linear rule of the following form:
6
ˆ g
t
=ρ
g
ˆ g
t−1
−ϕ π
ˆ π
c
t
−ϕ y
ˆ y
t
−ϕ b
ˆ
b
t
−ϕ d
�
def
t
+ε
g,t
. (29)
In this equation, ˆ g
t
is the percentage deviation of real government consumption (station-
arized with the unit-root technology level, z
t
) from its steady state level, ˆ π
c
t
is the CPI
inflation, ˆ y
t
reflects the output gap,
ˆ
b
t
is the public debt and
�
def
t
denotes the government
deficit which is expressed as a dif ference from its steady state, that is,
�
def
t
=def
t
−def .
ε
g,t
defines the exogenous shock aimed at capturing discretionary changes in government
consumption. ϕ π
,ϕ y
,ϕ b
andϕ d
denote the feedback coef ficients towards inflation, output
gap, public debt and government deficit deviations, respectively . ρ
g
reflects the degree of
government consumption smoothing.
2.5. Market equilibrium
In equilibrium all markets clear . The market clearing condition for the domestic goods
market is given by:
C
d
t
+I
d
t
+G
t
+C
x
t
+I
x
t
≤z
1−α
t
ϵ
t
K
α
t
H
1−α
t
−z
t
ϕ −a(u
t
)
¯
K
t
, (30)
whereC
x
t
andI
x
t
are the foreign demand for export goods which follow CES aggregates
with elasticity η
f
. Furthermore, the net foreign assets’ market clears when domestic in-
vestment in foreign bonds (denoted byB
∗
t
) equals the net position of exporting/importing
firms:
B
∗
t
=P
x
t
(C
x
t
+I
x
t
)−P
∗
t
(C
m
t
+I
m
t
)+R
∗
t−1
Φ
(
a
t
,
˜
ϕ t
)
B
∗
t
. (31)
6
Our specification for the fiscal rule is similar to those used by Erceg & Lindé (2013), with the only exception
that they do not include the inflation rate.
10

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 153
2.6. For eign economy
Since the domestic economy is a small open economy , we assume that the foreign economy
is exogenous. In particular , foreign output (ˆ y
∗
t
), foreign inflation (ˆ π
∗
t
) and foreign interest
rate (
ˆ
R
∗
t
) are exogenously modelled as an identified V AR model with two lags:
7
Φ
0
X
∗
t
=Φ
1
X
∗
t−1
+Φ
2
X
∗
t−2
+S
x
∗ε
x
∗
,t
, (32)
where X
∗
t
≡
(
ˆ π
∗
t
,ˆ y
∗
t
,
ˆ
R
∗
t
)
′
, ε
x
∗
,t
∼N(0,I
x
∗) , S
x
∗ is a diagonal matrix with standard
deviations andΦ
−1
0
S
x
∗ε
x
∗
,t
∼N(0,Σ
x
∗) .
2.7. Structural shocks
In total, the dynamics of the model is driven by 16 exogenous shock processes that are
assumed to be characterized in log-linearized form by the univariate representation:
ˆ
ξ
t
= ρ
ξ
ˆ
ξ
t−1
+ε
ξ,t
,ε
ξ,t
iid
∼ N
(
0,σ
2
ξ
)
, (33)
where ξ
t
=
{
µ z,t
,ϵ
t
,λ
j
t
,ζ
c
t
,ζ
h
t
,Υ
t
,
˜
ϕ t
,˜ z
∗
t
,τ
tr
t
,ε
g,t
}
for j = {d,mc,mi,x}. ε
g,t
is as-
sumed to be a white noise process (that is, ρ
εg
=0 ). There are also three foreign shocks
(that is, foreign output, foreign inflation and foreign interest rate shock) provided by the
exogenous (pre-estimated) foreign V AR model.
3 MODEL SOLUTION AND ESTIMA TION
In this section, we present how the DSGE model is solved and estimated.
7
The foreign V AR model is estimated for the first 12 Euro area countries over the period 1995Q1-2014Q4
and includes the following variables: output (GDP at market prices, chain linked volumes (2005), million units
of national currency); GDP deflator (GDP at market prices, price index (implicit deflator), 2005=100, national
currency); interest rate (12-month money market interest rate in percent). T o make the observed data consistent
with the model’ s concepts, we adjusted the data before entering the V AR model. Specifically , we used HP-
detrended log of GDP (we set the smoothing parameter to 1600, which is typically used with quarterly data), the
demeaned first dif ference of the log of GDP deflator and the demeaned interest rate which is divided by 400. All
data series are seasonally adjusted and adjusted by working days. The lag order of the V AR model was chosen
using the Hannan-Quinn information criterion, which suggests an optimal lag order of two periods (Lütkepohl &
Krätzig, 2004). W e also removed variables with lowest t-ratios until all remaining variables had t-ratios greater
than 2, which is often used in applied work. The estimated foreign V AR model is, therefore, given by:
ˆ π
∗
t
=0.028ˆ y
∗
t−1
+0.121ˆ π
∗
t−1
+0.279ˆ π
∗
t−2
+ε
π
∗
,t
ˆ y
∗
t
=1.667ˆ y
∗
t−1
−0.698ˆ y
∗
t−2
+ε
y
∗
,t
ˆ
R
∗
t
=1.190
ˆ
R
∗
t−1
+0.321ˆ y
∗
t−1
−0.306ˆ y
∗
t−2
−0.271
ˆ
R
∗
t−2
+ε
R
∗
,t
.
.
11

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 154
3.1. Model solution
The model presented in the previous section consists of a set of optimality conditions and
laws of motion of the shock processes. Since the model comprises the unit-root technology
shock that induces a stochastic trend in the levels of the real variables, the first step prior to
model solution is rendering the model stationary . T o this end all real variables are divided
with the trend level of technologyz
t
. The resulting stationary variables are then denoted by
lower -case letters, that is,x
t
=
Xt
zt
for a generic variablex
t
. W e then proceed with the log-
linearisation
8
to the model’ s equations of the transformed model around the deterministic
steady state
9
, where the variables are expressed as logarithmic deviations from their steady
state values, that is, ˆ x
t
=
xt−x
x
≈ lnx
t
− lnx, wherex denotes the steady state value of
a generic variablex
t
. Once the model has been stationarized and log-linearized, it can be
written in the following compact form:
E
t
{α
0
Γ
t−1
+α
1
Γ
t
+α
2
Γ
t+1
+β
2
Ψ
t+1
+β
1
Ψ
t
}=0, (34)
whereΓ
t
is a vector of endogenous variables,Ψ
t
is a vector of exogenous variables, and
α
0
,α
1
,α
2
,β
1
andβ
1
are coef ficient matrices. It is assumed thatΨ
t
evolves according to:
Ψ
t
=ρΨ
t−1
+ε
t
ε
t
∼N(0,Σ). (35)
W e use Dynare 4.4.3
10
to solve the model. The solution of the model takes the form:
11
Γ
t
=AΓ
t−1
+BΨ
t
. (36)
3.2. Data and measur ement equations
For estimation purposes the solved model can be written in the following state-space form
(Hamilton, 1994):
ξ
t+1
=Fξ
t
+υ
t+1
(37)
and:
˜
Y
t
=A
′
x
t
+H
′
ξ
t
+ω
t
. (38)
8
However , it is important to notice that dynamics in the log-linearized model is only approximation of the
true non-linear dynamics. Therefore, studying the log-linearized models is only valid for small deviations from
the model’ s steady state. For a complete list of the log-linearized equations of the model, see Appendix A.
9
W e compute the non-stochastic steady state of the model following the procedure described in Adolfson et
al. (2007). It is important to note that the steady state also depends on estimated parameters. For this reason,
when estimating the model, it is of great importance to take into account parameter dependence by using model-
local variables. For further discussion, see Pfeifer (2014a), Remark 4 (Parameter dependence and the use of
model-local variables).
10
Dynare is a software package for solving and estimating DSGE models. For more information regarding
Dynare refer to the of ficial Dynare web page http://www.dynare.org and see Mancini Grif foli (201 1), as
well as Adjemin, Bastani, Karamé, Juillard, Maih, Mihoubi, Perendia, Pfeifer , Ratto & V illemot (2014).
11
Dynare uses solution algorithms proposed by Klein (2000) and Sims (2002). For a detailed look at what
exactly is going on behind the scenes of Dynare’ s computations, the interested reader is referred to V illemot
(201 1).
12

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 155
The first equation is called the state equation, whereas the second is called the observa-
tion (measurement) equation. The symbols appearing in ( 37) and (38) have the following
meaning:
˜
Y
t
is an(n×1) vector of observed variables at timet,ξ
t
is an(r×1) vector of
unobserved variables at timet (also referred to as state vector) andx
t
is a (k×1) vector
with exogenous or predetermined variables (e.g. a constant). Furthermore,F ,A
′
andH
′
are matrices of dimension (r×r) , (n×k) and (n×r) , respectively . The (r×1) vector
υ
t
and the (n×1) vector ω
t
are uncorrelated, normally distributed white noise vectors,
therefore:
E(υ
t
υ
′
τ
)=
{
Q forτ =t
0 otherwise
E(ω
t
ω
′
τ
)=
{
R forτ =t
0 otherwise,
whereQ andR are (r×r) and (n×n) matrices, respectively . The disturbancesυ
t
and
ω
t
are assumed to be uncorrelated at all lags:
E(υ
t
,ω
′
τ
)=0 for allt andτ. (39)
In what follows, we describe how the raw data were converted to the form used in estima-
tion. In addition, we present the exact measurement equations that are employed to relate
the observed data to the model state variables. The estimates are based on quarterly Slove-
nian macroeconomic data covering the period 1995Q1-2014Q4. W e employ the following
14 variables as observables:
12
the GDP deflator (P
d
t
), the real wage (W
t
/P
d
t
), consump-
tion (C
t
), investment (I
t
), government consumption (G
t
), the real exchange rate (x
t
), the
short-run interest rate (R
t
), employment
13
(E
t
), GDP (Y
t
), exports (X
t
), imports (M
t
),
the CPI price index (P
c
t
), foreign output (for the first 12 euro area countries) (Y
∗
t
), the
foreign GDP deflator (for the first 12 euro area countries) (P
∗
t
) and the foreign interest
rate (12-month money market interest rate of the euro area) (R
∗
t
). Regarding the foreign
variables, GDP for the first 12 euro area countries is used for foreign output, and the GDP
deflator for the first 12 euro area countries is used for foreign inflation, while the foreign
interest rate refers to the 12-month money market interest rate of the euro area. Data come
from four dif ferent sources. Data on the employment and gross wages are taken from the
Statistical Of fice of the Republic of Slovenia. The sources for domestic interest rate are
the Bank of Slovenia and the Institute of Macroeconomic Analysis and Development of
the Republic of Slovenia. The rest of the data are taken from Eurostat. Since the model
comprises a stochastic unit root technology shock that induces a common stochastic trend
in the real variables of the model, we use first dif ferences to make these variables sta-
tionary . When estimating the model, the following variables are matched in growth rates
12
A detailed description of the data used in the estimation together with their sources is provided in Appendix
B. Additionally , the data are plotted in Appendix D.
13
W e assume that the employment variable (
ˆ
Et ) is related to the hours worked variable (
ˆ
Ht ) by an auxiliary
equation (expressed as a percentage deviation from the steady state):
∆ ˆ
Et =
β
1+β
Et∆ ˆ
E
t+1
+
(1−ξe)(1−βξe)
(1+β)ξe
(
ˆ
Ht−
ˆ
Et
)
.
The Calvo parameter ,ξe , representing the fraction of firms that in any period is able to adjust employment to its
desired total labour input, is estimated.
13

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 156
measured as quarter -to-quarter log-dif ferences: GDP , consumption, investment, exports,
imports, government consumption, real wage, GDP deflator , CPI price index, foreign out-
put and foreign GDP deflator . The rest of the variables are used in levels: domestic and
foreign interest rate, employment and real exchange rate. The real wage is calculated as
the nominal gross wage per employee deflated by the GDP deflator . All interest rates are
divided by 4 to express them in quarterly rates consistent with the variables in the model.
The stationary variables,x
t
andE
t
, are measured as follows: we take the logarithm of real
exchange rate and remove a linear trend, so that it is expressed in percentage deviations
from the trend, consistently with the model concepts, that is ˆ x
data
t
=
xt−x
x
, while the em-
ployment is measured as deviation around the mean, that is
ˆ
E
data
t
=
Et−E
E
. Furthermore,
in order to align the data with the model-based definitions, some additional transforma-
tions are made. First, since the model assumes that all real variables are growing at the
same rate as output, we match the sample growth rates of private consumption, invest-
ment, government consumption, exports, imports and real wage with the sample growth
rate of real GDP by removing the remaining growth rate dif ferentials. Second, the model
assumes that in steady-state, the interest rates (that is, domestic and foreign interest rate)
as well as dif ferent measures of inflation (that is, domestic, CPI and foreign inflation) are
identical, that isR = R
∗
andπ
d
= π
c
= π
∗
, respectively . This assumption is clearly
rejected by the data. T o circumvent this issue, we demean all these time series before the
model estimation and add the sample mean of domestic interest rate to the foreign interest
rate and the sample mean of domestic inflation to the CPI and foreign inflation, so that the
data match the model assumptions. All variables (except the nominal interest rates) are
seasonally adjusted and adjusted by working days. The vector of observable variables,
˜
Y
t
,
is then given by:
˜
Y
t
=
[
∆ lnP
d,data
t
∆ ln
(
W
t
/P
d
t
)
data
∆ ln
˜
C
data
t
∆ ln
˜
I
data
t
ˆ x
data
t
...
R
data
t
ˆ
E
data
t
∆ lnY
data
t
∆ ln
˜
X
data
t
∆ ln
˜
M
data
t
...
∆ lnG
data
t
∆ lnP
c,data
t
∆ lnY
∗,data
t
∆ lnP
∗,data
t
R
∗,data
t
]
′
,
(40)
14

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 157
where ∆ is the first dif ference operator . The corresponding measurement equation that
matches observed data with model’ s variables is:
˜
Y
t
=




























∆ lnP
d,data
t
∆ ln
(
W
t
/P
d
t
)
data
∆ ln
˜
C
data
t
∆ ln
˜
I
data
t
ˆ x
data
t
R
data
t
ˆ
E
data
t
∆ lnY
data
t
∆ ln
˜
X
data
t
∆ ln
˜
M
data
t
∆ lnG
data
t
∆ lnP
c,data
t
∆ lnY
∗,data
t
∆ lnP
∗,data
t
R
∗,data
t




























=


























(
π
d
−1
)
lnμ
z
lnμ
z
lnμ
z
0
4R(R−1)
0
lnμ
z
lnμ
z
lnμ
z
lnμ
z
(
π
d
−1
)
lnμ
z
(
π
d
−1
)
4R(R−1)


























+



























ˆ π
d
t
∆ ˆ
¯ w
t
+ˆ μ
z,t
∆ ˆ
˜ c
t
+ˆ μ
z,t
∆ ˆ
˜
i
t
+ˆ μ
z,t
ˆ x
t
4R
ˆ
R
t
ˆ
E
t
∆ˆy
t
+ˆ μ
z,t
∆ ˆ
˜ x
t
+ˆ μ
z,t
∆ ˆ
˜ m
t
+ˆ μ
z,t
∆ˆg
t
+ˆ μ
z,t
ˆ π
c
t
∆ˆy
∗
t
+∆
ˆ
˜ z
∗
t
+ˆ μ
z,t
ˆ π
∗
t
4R
ˆ
R
∗
t



























+



























ε
me
π
d
,t
ε
me
˜ w,t
ε
me
˜ c,t
ε
me
˜
i,t
ε
me
x,t
ε
me
R,t
ε
me
E,t
ε
me
y,t
ε
me
˜ x,t
ε
me
˜ m,t
ε
me
g,t
ε
me
π
c
,t
ε
me
y
∗
,t
ε
me
π
∗
,t
ε
me
R
∗
,t



























, (41)
whereε
me
i,t
denotes the measurement error for the respective variable. The standard devi-
ation of specific measurement error is calibrated at 10% of the standard deviation of the
corresponding observed domestic variables, while the measurement errors for the foreign
variables are set to 0, as in Adolfson et al. (2007).
3.3. Estimation methodology
Structural parameters of the model are either calibrated or estimated. The values for the
parameters that are calibrated (and thus kept fixed throughout the estimation) are chosen
in accordance with the practice in the literature calibrating small open-economy models.
Their values are discussed in Subsection 3.4. All remaining parameters are estimated with
a Bayesian estimation method, which has become a standard econometric technique for
estimating DSGE models.
14
In the following, we briefly describe the main features of
the method. The key idea of the Bayesian estimation method
15
is that it combines the
prior belief of the parameters with empirical data to form the posterior distributions of the
parameters. The posterior distributions are obtained by using the Bayes theorem:
p
(
θ|
˜
Y
t
)
=
p
(
˜
Y
t
|θ
)
p(θ)
p
(
˜
Y
t
) , (42)
14
All estimates are performed using Dynare version 4.4.3 in Matlab R2012b.
15
A detailed explanation of the estimation method can be found in An & Schorfheide (2007), Adolfson et al.
(2007), Canova (2007), Fernández-V illaverde (2010) and Smets & W outers (2003, 2007) among many others.
The reader is also referred to Dynare Manual for additional explanation of the estimation method.
15

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 158
whereθ is a vector of the parameters to estimate,p
(
θ|
˜
Y
t
)
is the density of the parameters
conditional on data (the posterior), p
(
˜
Y
t
|θ
)
is the density of the data conditional on the
parameters (the likelihood),p(θ) is the unconditional density of the parameters (the prior)
and p
(
˜
Y
t
)
is the mar ginal density of the data.
16
Given that the mar ginal density of the
data is a constant term or equal for any parameter , equation ( 42) can be rewritten as:
p
(
θ|
˜
Y
t
)
∝p
(
˜
Y
t
|θ
)
p(θ)≡K
(
θ|
˜
Y
t
)
, (43)
whereK
(
θ|
˜
Y
t
)
is the posterior kernel. T aking logs of ( 43), we get:
lnK
(
θ|
˜
Y
t
)
= lnp
(
˜
Y
t
|θ
)
+ lnp(θ)= lnL
(
˜
Y
t
|θ
)
+ lnp(θ). (44)
Before the estimation can begin, we need to specify the priors for the parameters to be es-
timated and evaluate the likelihood function of the observed data. The choice of priors is
discussed in Subsection 3.5. The likelihood function of the observed data is evaluated by
generating forecasts from the state-space system, ( 37) and (38), with the use of the Kalman
filter . Conceptually , the Kalman filter consists of calculating the sequence
{
ξ
t+1|t
}
T
t=1
and
{
Σ
ξ
t+1|t
}
T
t=1
, whereξ
t+1|t
denotes the optimal forecast ofξ
t+1
based on observation
of y
t
≡
(
˜
Y
′
t
,
˜
Y
′
t−1
,
˜
Y
′
t−2
,...,
˜
Y
′
1
,x
′
t
,x
′
t−1
,x
′
t−2
,...,x
′
1
)
′
and Σ
ξ
t+1|t
denotes the mean
squared error of this forecast. The algorithm works forward in time and is conducted as
follows:
17
Fort =1 , the algorithm needs to be provided with initial values for a one-step
ahead forecast of time t states, ξ
t|t−1
, and respective forecast error variance-covariance
matrix,Σ
ξ
t|t−1
. Based on this a one-step ahead forecast of timet data,
˜
Y
t|t−1
and respec-
tive variance-covariance matrix, Σ
˜
Y
t|t−1
are computed. The algorithm then updates the
forecasts of timet states,ξ
t|t
, and a respective variance-covariance matrix,Σ
ξ
t|t
. The final
step is to compute a one-step ahead forecast of time t +1 states, ξ
t+1|t
, and respective
variance-covariance matrix, Σ
ξ
t+1|t
. These steps are iterated for t =2 ,3,4,...,T . The
log-likelihood function (based on the data up to timet) can be written as follows (Hamilton,
1994):
T
∑
t=1
lnL
(
˜
Y
t
|x
t
, y
t−1
,F,A
′
,H
′
,Q,R
)
=−
T
∑
t=1



n
2
log2π+
1
2
log|Σ
˜
Y
t|t−1
|
+
1
2
∑
T
t=1
(
˜
Y
t
−
˜
Y
t|t−1
)
′
(
Σ
˜
Y
t|t−1
)
−1
(
˜
Y
t
−
˜
Y
t|t−1
)


.
(45)
16
It is defined as:
p
(
˜
Yt
)
=
∫
Θ
p
(
θ,
˜
Yt
)
dθ,
where p
(
θ,
˜
Yt
)
denotes the joint density of the parameters and the data.
17
The presentation here follows Hamilton (1994).
16

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 159
Finally , the posterior distribution is obtained in two steps: first, by maximizing the log
posterior density with respect to θ , the posterior mode θ
m
and an approximate covari-
ance matrix, based on the inverse Hessian matrix evaluated at the posterior mode,Σ
θ
m =
H
(
θ
m
|
˜
Y
)
−1
, is obtained and second, the posterior distribution is simulated by using
the Monte-Carlo Markov-Chain (MCMC) sampling method, specifically the Metropolis-
Hastings (MH) algorithm. The idea behind this algorithm is the following (Mancini Grif-
foli, 2013): first, the algorithm chooses a starting point (posterior mode), then it draws a
candidate valueθ
∗
from an arbitrary candidate (or jumping) distribution:
J(θ
∗
|θ
i−1
)∼ N (θ
i−1
,cΣ
θ
m), (46)
where θ
i−1
is the last accepted draw , Σ
θ
m denotes the inverse of the Hessian computed
at the posterior mode, and c is the scale factor , which is chosen to ensure an appropriate
acceptance rate. In the next step, the algorithm computes the acceptance ratio:
α = min


K
(
θ
∗
|
˜
Y
t
)
K
(
θ
i−1
|
˜
Y
t
),1


. (47)
The algorithm then accepts or discards the proposalθ
∗
according to the following rule:
θ
i
=
{
θ
∗
with probabilityα
θ
i−1
otherwise
.
If the parameter value is accepted, the mean of the distribution is updated with the new draw
θ
i
. These algorithm steps are repeated many times to simulate the posterior distribution.
3.4. Calibrated parameters
In this section, we present the calibrated parameters of the model.
18
Their values are taken
mainly from Adolfson et al. (2007) unless otherwise stated. The discount factor , β , is
fixed to 0.993, implying a steady-state interest rate of 1 1%,
19
which matches the average
interest rate in the sample period. The share of capital in production,α , is calibrated to 0.30.
The depreciation rate of capital,δ , is set to 0.013. W e calibrate the capital utilization cost
parameter ,σ
a
, to10
6
. The elasticity of substitution between domestic and foreign goods,
η
c
, is calibrated to 5. Labour disutility constant,A
L
, is calibrated to 7.5. As in Christiano,
Eichenbaum and Evans (2005), we set the labour supply elasticity ,σ
L
, to 1, and the wage
mark-up, λ
w
, to 1.05. The steady state mark-ups are calibrated at: 1.222 in the domestic
goods market (λ
d
), 1.633 in the imported consumption goods market (λ
m,c
) and 1.275 in
the imported investment goods market (λ
m,i
). The steady state foreign terms of trade,γ
f
,
is calibrated to 1. The rest of the parameters, as well as the steady state relationships, are
18
It is important to note that a time period is taken to be a quarter .
19
This follows from the first order condition of the households’ bond holdings,R =
πμz−τ
k
β
(1−τ
k
)β
.
17

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 160
calibrated using the averages of Slovenian data for the period 1995Q1-2014Q4. The shares
of imports in consumption and investment,ω
c
andω
i
, are set to 0.67 and 0.40, respectively .
The steady state rate of transfers to households, τ
tr
, is calibrated to 0.50. The ratios of
government expenditures (
gex
y
), taxes (
t
y
), government consumption (
g
y
), and debt services
(
r
y
) in GDP are 0.37, 0.36, 0.19, and 0.02, respectively . Further , the share of government
consumption, social transfers and debt services in total government expenditures,
g
gex
,
tr
gex
and
r
gex
, are set to 0.51, 0.44 and 0.05, respectively . The tar get value of debt-to-GDP ratio,
b
∗
y
, is assumed to be 240% in the steady state, which is consistent with the reference value of
public debt established by the Maastricht T reaty , which equals 60% of yearly output. The
steady state quarterly gross inflation rate,π
d
, is equal to 1.01. Finally , the average ef fective
tax rates on consumption, labour income and capital income,τ
c
,τ
y
andτ
k
, amount to 0.17,
0.48 and 0.22, respectively . An overview of the calibrated parameters is found in T able 1,
while T able 2 provides an overview of the steady state relationships.
T able 1: Calibrated parameters of the model
Parameter Description Calibrated value
β Households’ discount factor 0.993
α Capital share in production 0.30
η
c
Substitution elasticity betweenC
d
t
andC
m
t
5
σ
a
Capital utilization cost parameter 10
6
A
L
Labour disutility constant 0.3776
σ
L
Labour supply elasticity 1
δ Depreciation rate of physical capital 0.013
λ
w
W age mark-up 1.05
λ
d
Mark-up in the domestic goods market 1.168
λ
m,c
Mark-up in the imported consumption goods market 1.619
λ
m,i
Mark-up in the imported investment goods market 1.226
ω
i
Share of imports in investment 0.40
ω
c
Share of imports in consumption 0.67
τ
c
Consumption tax rate 0.1 14
τ
y
Labour income tax rate 0.48
τ
k
Capital tax rate 0.22
τ
tr
Rate of transfers to households 0.50
3.5. Prior distributions of the estimated parameters
Before the Bayesian estimation method starts, the prior distributions of the estimated pa-
rameters need to be specified. As the name suggests, prior distribution describes the avail-
able information about the parameters prior to observing the data used in the estimation.
The observed data is then used to update the prior , through the Bayes theorem, to the poste-
rior distribution of the model’ s parameters. In specifying the prior distributions we mainly
18

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 161
T able 2: Steady state r elationships
Parameter Description V alue
π
d
Steady state quarterly gross inflation rate 1.01
gex
y
Share of government expenditures in GDP 0.37
t
y
Share of taxes in GDP 0.36
g
y
Share of government consumption in GDP 0.19
g
gex
Share of government consumption in government expenditures 0.51
r
y
Share of debt services in GDP 0.02
r
gex
Share of debt services in government expenditures 0.05
tr
gex
Share of social transfers in government expenditures 0.44
b
∗
y
T ar get value of debt-to-GDP ratio 2.4
rely on choices from Adolfson et al. (2007).
Throughout the analysis we use four main distributions: beta distribution, inverse gamma
distribution, normal distribution and gamma distribution. For the parameters bounded be-
tween 0 and 1 we choose beta distribution. Parameters belonging to this group are nominal
stickiness parametersξ , indexation parametersκ, the habit persistenceb and the persistence
parameters of the shock processes ρ. W e set the mean of prior distributions for the price
stickiness parameters to 0.5 with standard deviation 0.2, while the mean for the indexation
parameters is set to 0.4 with standard deviation 0.1. However , there are three exceptions.
For the Calvo parameter for domestic firms we set the prior mean to 0.85 with a standard
deviation of 0.1, while for the Calvo parameter for exporting firms we choose a prior mean
equal to 0.75 with a standard deviation of 0.1. For the wage indexation parameter we im-
pose a prior mean of 0.5 with a standard deviation of 0.2. The prior on habit persistence
has a mean of 0.65 and a standard deviation of 0.2. W ith the exception of the shocks to
the unit-root technology , stationary technology and government consumption, we set the
prior means of the persistence parameters for the structural shocks equal to 0.5 with a stan-
dard deviation of 0.2. For the unit-root technology , stationary technology and government
consumption shocks we choose a mean of 0.6 and a standard deviation of 0.2.
W e use inverse gamma distribution to describe our priors about the parameters that are
assumed to be positive. These parameters are the standard deviations of shocks and the
substitution elasticities between goods, η . W e set the prior mean of the substitution elas-
ticity between domestic and foreign investment goods, η
i
, equal to 0.8, while the prior
mean of the substitution elasticity among goods in the foreign economy , η
f
, is set to 1.5.
Continuing with the standard deviations of shocks
20
, we set the standard deviation of the
stationary technology shock,σ
ϵ
, to 0.007, and the standard deviation of the unit-root tech-
nology shock, σ
µ z
, is assumed to be 0.002, which is the value used by Altig et al. (201 1).
20
In order to decrease the degree of non-linearity when estimating the model, the mark-up shocks in the Phillips
curves, as well as the investment-specific technology shock, the labour supply shock and the consumption pref-
erence shock enter into the equations in an additive way .
19

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 162
The size of the risk premium shock, σ
˜
ϕ , and the prior on the risk premium parameter
related to net foreign assets,
˜
ϕ , are set to 0.0005 and 0.045, respectively . Based on the
residuals from a first-order autoregression of the series obtained when substracting the
HP-trend in domestic output from the HP-trend in foreign output, we set the size of the
asymmetric technology shock,σ
˜ z
∗ , to 0.003. The consumption preference, labour supply
and investment-specific technology shocks, σ
ζc
, σ
ζ
h
and σ
Υ
, respectively , are assumed
to have the prior mean of 0.002, which is similar to Adolfson et al. (2007). Since we
have little information about the properties of these shocks, we choose very loose priors
with infinite variances. Regarding the foreign shocks, there are three standard deviations
of shocks which need to be specified, namely the standard deviation of the foreign output
shock, foreign inflation shock and foreign interest rate shock. W e fix their values at the
standard deviations of residuals obtained from a pre-estimated foreign V AR model. The
standard deviation for the foreign output shock,σ
y
∗ , is, therefore, set to 0.004, the foreign
inflation shock,σ
π
∗ , is assumed to have a standard deviation of 0.002, while the standard
deviation for the foreign interest rate shock,σ
R
∗ , is set to 0.003.
Finally , turning to the parameters of the fiscal rule, the prior on the persistence parameter
(ρ
g
) follows a beta distribution with a mean of 0.6 and a standard deviation of 0.2. The
priors on the feedback coef ficients are assumed to be gamma distributed. W e set their
values as follows: the prior mean of the feedback coef ficient on inflation (ϕ π
) and output
gap (ϕ y
) is set to 0.25 with a standard deviation of 0.15, while the prior on the feedback
coef ficient on public debt (ϕ b
) and government deficit (ϕ d
) is somewhat lower and has a
mean equal to 0.05 and a standard deviation of 0.01. For the steady state quarterly gross
growth rate, µ z
, we choose normal distribution with prior mean centred around 1.006,
implying an annual growth rate of about 2.4%.
4 ESTIMA TION RESUL TS AND EV ALUA TION
In this section, we present and evaluate the estimation results.
4.1. Posterior distributions of the estimated parameters
In total we estimated 50 parameters: 17 friction parameters, 5 policy parameters and 28
shock processes parameters. The posterior mode and an approximate covariance matrix,
based on the inverse Hessian matrix evaluated at the mode, have been computed by using
a standard numerical optimisation routine, namely Christopher Sims’ optimizer csminwel,
on the log posterior density . After having optimized the log posterior density , the draws
from the posterior distribution have been obtained by simulating two parallel Markov
chains of 300,000 draws of the Metropolis-Hastings algorithm, ignoring the first 50% of
draws as burn-in. The average acceptance rate is roughly 32% across the two Metropolis-
Hastings blocks used.
20

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 163
After the estimation, we performed several diagnostic tests to assess the quality of the
estimated model. More precisely , we (i) looked at the quality of the posterior kernel opti-
mization, (ii) assessed the conver gence of the Markov chains by using both the univariate
conver gence diagnostics proposed by Brooks & Gelman (1998) as well as the multivariate
conver gence diagnostics, and (iii) compared the plots of the prior and posterior distribu-
tions. This latter diagnostic can be found in Appendix C. T o have sensible estimates, the
patterns of the prior and posterior distributions should be reasonably distinct. If the pos-
terior looks like the prior , either the prior is a very accurate reflection of the information
in the data or , more usually , the parameter under consideration is only weakly identified
and t he data does not provide much information to update the prior (Canova, 2007). On
the other hand, if the prior and posterior distribution are far away from each other , this
typically indicates that there is a disagreement between the information provided by the
data and the prior knowledge about the true parameter value. In addition, the posterior
distribution should be approximately normal in shape, which is in line with the asymptotic
properties of Bayesian estimation, and the mode should be in the center of the posterior
distribution.
As seen in these plots, the most of the estimated parameters are well identified as their
posterior distribution is reasonably dif ferent from the prior distribution. Moreover , for the
majority of the parameters, the variance of the posterior is lower compared to the prior
distribution, indicating that data are quite informative. The mode check plots (not pre-
sented here) indicate that the optimization procedure was able to precisely find a robust
maximum for the posterior kernel. Finally , both univariate and multivariate conver gence
graphs (also not presented here) confirm that the parameters are generally characterized
by good conver gence.
21
The estimation results are summarized in T able 3, which provides prior distributions, pos-
terior estimations and 90% confidence intervals of the estimated parameters. Let us now
briefly discuss the estimation results. Beginning with the Calvo price stickiness parame-
ters, we find that the domestic price stickiness parameter ξ
d
is estimated at 0.90, which
implies the average duration of prices of about 10 quarters.
22
The values for the other
sectors (ξ
m,c
, ξ
m,i
and ξ
x
) are estimated as follows: the estimated price stickiness param-
eter for the imported consumption,ξ
m,c
, is equal to 0.71, suggesting that prices remain on
average unchanged for 3 quarters. Furthermore, the posterior mean of the price stickiness
parameter for the imported investment,ξ
m,i
, is estimated at 0.52. The average duration of
prices in this sector is therefore 2 quarters. The export price stickiness parameter , ξ
x
, in
turn, is estimated at 0.87, corresponding to an average price duration of 7 quarters. The pos-
terior mean of the Calvo wage stickiness parameter ,ξ
w
, is nearly 0.56. This value implies
that wages are reset as frequent as twice a year . Considering next the indexation param-
eters, we find that the posterior mean of the degree of wage indexation, κ
w
, is estimated
at 0.60, while the remaining indexation parameters (κ
d
, κ
m,c
, κ
m,i
, κ
x
) are estimated at a
lower value. The posterior mean of the habit persistence parameter in consumption, b, is
21
Due to space limitations these graphs are not presented here, but they are available from the author upon
request.
22
A verage duration of prices comes from
1
1−ξ
d
.
21

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 164
estimated at 0.94. Furthermore, our estimates suggest the substitution elasticity between
domestic and foreign investment goods,η
i
, of around 0.29, while the posterior mean for the
substitution elasticity among goods in the foreign economy ,η
f
, is estimated at 1.37. The
investment adjustment cost parameter ,
˜
S
′′
, is estimated to be equal to 8.65. The posterior
mean of the risk premium parameter related to net foreign assets,
˜
ϕ a
, is 0.03.
Regarding the parameters in the fiscal policy rule, we find that the feedback coef ficient
of government consumption to inflation,ϕ π
, is estimated at 0.22, the estimated feedback
coef ficient of output gap,ϕ y
, is 0.08, while the estimated feedback coef ficients of public
debt and government deficit,ϕ b
andϕ d
, are equal to 0.06 and 0.05, respectively . It is worth
noting that the latter two parameters are driven by a prior . This can be explained by the
fact that we do not use the data on public debt and government deficit in the estimation.
The persistence parameter in the fiscal rule,ρ
g
, is estimated to be 0.50, which indicates a
moderate degree of persistence in government consumption.
Finally , we consider the parameters associated with the persistence and volatility of shock
processes (see T able 4). W e find that the autoregressive parameters are estimated to lie
between 0.22 for the consumption preference shock and 0.96 for the unit-root technology
shock. In general, the level of persistence of stochastic processes is not very high, indicat-
ing that the model contains a suf ficiently persistent endogenous propagation mechanism.
T urning to the estimated standard deviations of shocks, we find that the most volatile are
the imported investment mark-up shocks and the investment-specific technology shock,
with standard deviations of 0.3345 and 0.0309, respectively , while the least volatile is the
unit-root technology shock with a standard deviation equal to 0.0013.
22

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 165
T able 3: Prior and posterior distribution of structural parameters
Prior distribution Posterior distribution
Description Parameter T ype Mean Std. Dev ./Df Mode Std. Dev . Mean 5 % 95 %
Friction parameters
Calvo wages ξw Beta 0.500 0.200 0.5775 0 .0881 0.5568 0.4173 0.6918
Calvo domestic prices ξ
d
Beta 0.850 0.100 0.9044 0 .0206 0.9018 0.8639 0.9365
Calvo import consumption prices ξm,c Beta 0.500 0.200 0.7569 0.0957 0.7051 0.5352 0.8673
Calvo import investment prices ξm,i Beta 0.500 0.200 0.6293 0.1099 0.5195 0.2910 0.7509
Calvo export prices ξx Beta 0.750 0.100 0.8954 0.0439 0.8655 0.7689 0.9702
Calvo employment ξe Beta 0.500 0.200 0.81 12 0.0319 0.8232 0.7762 0.8757
Indexation wages κw Beta 0.500 0.200 0.5927 0.1770 0.6016 0.3491 0.8291
Indexation domestic prices κ
d
Beta 0.400 0.100 0.2013 0.0643 0.2181 0.1064 0.3172
Indexation import consumption prices κm,c Beta 0.400 0.100 0.3379 0.1 1 10 0.3510 0.1951 0.5056
Indexation import investment prices κm,i Beta 0.400 0.100 0.3250 0.0957 0.3349 0.1814 0.4848
Indexation export prices κx Beta 0.400 0.100 0.3293 0.0996 0.3309 0.1786 0.4764
Investment adjustment cost
˜
S
∗
Normal 7.694 1.500 8.6319 1.2547 8.6526 6.5778 10.6581
Habit formation b Beta 0.650 0.200 0.9442 0.0214 0.9413 0.9081 0.9761
Substitution elasticity investment ηi Inv . Gamma 0.800 inf 0.2860 0.0552 0.2925 0.2001 0.3847
Substitution elasticity foreign η
f
Inv . Gamma 1.500 2 1.1934 0.4019 1.3696 0.6160 2.1375
T echnology growth µ z Beta 1.006 0.005 1.0061 0.0005 1.0061 1.0054 1.0068
Risk premium
˜
ϕ a Beta 0.045 0.02 0.0234 0.0103 0.0296 0.01 15 0.0460
Policy parameters
Policy rule: lagged gov . consumption ρg Beta 0.600 0.200 0.5149 0.1585 0.4987 0.3314 0.6619
Policy rule: inflation ϕ π Gamma 0.25 0.15 0.1900 0.0990 0.2234 0.0623 0.3765
Policy rule: output gap ϕ y Gamma 0.25 0.15 0.0649 0.0491 0.0834 0.0148 0.1451
Policy rule: public debt ϕ b
Gamma 0.05 0.01 0.0508 0.0160 0.0553 0.0381 0.0716
Policy rule: gov . deficit ϕ def
Gamma 0.05 0.01 0.0495 0.0101 0.0514 0.0339 0.0674
23

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 166
T able 4: Prior and posterior distribution of shock pr ocesses
Prior distribution Posterior distribution
Description Parameter T ype Mean Std. Dev ./Df Mode Std. Dev . Mean 5 % 95 %
Persistence parameters
Unit-root technology shock ρµ z
Beta 0.600 0.200 0.9707 0.0167 0.9577 0.9272 0.9899
Stationary technology shock ρϵ Beta 0.600 0.200 0.8697 0.1096 0.7887 0.6071 0.9613
Investment-specific technology shock ρΥ Beta 0.500 0.200 0.2872 0.1381 0.2775 0.0823 0.4677
Asymmetric technology shock ρ˜ z
∗ Beta 0.500 0.200 0.9872 0.0091 0.9822 0.9678 0.9971
Consumption preference shock ρ
ζc
Beta 0.500 0.200 0.1938 0.0979 0.2228 0.0773 0.3700
Labour supply shock ρ
ζ
h
Beta 0.500 0.200 0.5307 0.3261 0.5435 0.2473 0.8126
Risk premium shock ρ˜
ϕ Beta 0.500 0.200 0.9329 0.0153 0.9317 0.9076 0.9577
Domestic mark-up shock ρ
λ
d
Beta 0.500 0.200 0.4599 0.0988 0.4742 0.3051 0.6394
Import consumption mark-up shock ρ
λm,c
Beta 0.500 0.200 0.4727 0.2155 0.5425 0.2412 0.8520
Import investment mark-up shock ρ
λ
m,i
Beta 0.500 0.200 0.1670 0.1219 0.3121 0.0500 0.5801
Export mark-up shock ρ
λx
Beta 0.500 0.200 0.2893 0.1458 0.3566 0.0859 0.6103
Rate of transfers shock ρ
τ
tr Beta 0.500 0.200 0.6550 0.1 105 0.5941 0.3635 0.8286
Standard deviations
Unit-root technology shock σµ z
Inv . Gamma 0.002 inf 0.001 1 0.0002 0.0013 0.0008 0.0017
Stationary technology shock σϵ Inv . Gamma 0.007 inf 0.0126 0.0056 0.0210 0.0071 0.0380
Investment-specific technology shock σΥ Inv . Gamma 0.002 inf 0.0304 0.0040 0.0309 0.0246 0.0367
Asymmetric technology shock σ˜ z
∗ Inv . Gamma 0.003 inf 0.0037 0.0005 0.0038 0.0030 0.0047
Consumption preference shock σ
ζc
Inv . Gamma 0.002 inf 0.0044 0.0006 0.0044 0.0035 0.0054
Labour supply shock σ
ζ
h
Inv . Gamma 0.002 inf 0.0010 0.0005 0.0015 0.0006 0.0023
Risk premium shock σ˜
ϕ Inv . Gamma 0.0005 inf 0.0030 0.0003 0.0032 0.0026 0.0036
Domestic mark-up shock σ
λ
d
Inv . Gamma 0.003 inf 0.0024 0.0003 0.0024 0.0018 0.0030
Import consumption mark-up shock σ
λm,c
Inv . Gamma 0.003 inf 0.0026 0.0004 0.0028 0.0019 0.0037
Import investment mark-up shock σ
λ
m,i
Inv . Gamma 0.003 inf 0.2533 0.0756 0.3345 0.1470 0.5169
T able continued on next page
24

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 167
Prior distribution Posterior distribution
Description Parameter T ype Mean Std. Dev ./Df Mode Std. Dev . Mean 5 % 95 %
Export mark-up shock σ
λx
Inv . Gamma 0.003 inf 0.01 12 0.0043 0.01 13 0.0044 0.0184
Government consumption shock σg Inv . Gamma 0.002 inf 0.0049 0.0006 0.0050 0.0041 0.0058
Rate of transfers shock σ
τ
tr Inv . Gamma 0.002 inf 0.0018 0.0005 0.0016 0.0007 0.0023
25

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 168
4.2. Assessing the empirical performance of the model
After having presented and evaluated the estimation results, we now proceed with the
assessment of the empirical performance of the estimated model. This is done in three di-
rections. First, we evaluate the absolute (in-sample) fit of the model. Second, we compare
the unconditional second moments in the estimated DSGE model with those based on the
actual data. Finally , we look at the smoothed estimates of the innovation component of
structural shocks.
In Figure D.1 in Appendix D we first plot the actual series used in the estimation along with
filtered variables obtained by the one-sided Kalman filter for each of the fifteen observable
variables. The thin red line depicts the mean estimate of the one step ahead forecast of
the endogenous variables (best guess for the endogenous variables at time t+1 given
information up to the current observations t), derived from the Kalman filter , whereas the
thick black line represents the actual data (Pfeifer , 2014b). As it can be seen from the sub-
plots, the in-sample fit of the model is satisfactory in most of the cases since the model
predictions closely follow the path of the observed historical data. However , the model is
not good at capturing government consumption.
The common practice in the DSGE literature is to analyse how well the model’ s moments
match those from the actual data. As a next step we therefore compare the second mo-
ments in the data (for the period 1995Q2-2014Q4) with those in the model (calculated
at the posterior mean). The results are presented in T able 5. The first column presents
the standard deviations of the selected observed variables and their counterparts implied
by the estimated model. The second column reports the first order autocorrelation coef-
ficients. The last two columns show correlations with GDP growth rates and domestic
inflation, respectively . Several results are worth highlighting here. First, our model is
able to replicate quite well the volatilities of some observables, in particular those of the
growth rate of GDP and government consumption, but generates much high volatile con-
sumption growth rates as compared to the data. Consumption growth in the model is three
times more volatile than in the data. Furthermore, we can observe that the model replicates
quite closely the positive correlation of investment and government consumption growth
rates with GDP growth rates. The correlation between investment and GDP growth rates
is 0.67 in the model, while it is 0.77 in the data. These numbers are respectively 0.09 and
0.14 for the correlation of government consumption. On the other hand, the correlation of
imports and exports seems to be underestimated by the model (0.77 in the data and 0.40 in
the model for imports, 0.70 in the data and 0.41 in the model for exports), while the corre-
lation of consumption is slightly overestimated by the model (0.42 in the data and 0.64 in
the model). Furthermore, the model is quite successful in predicting the persistence of the
observables, except the persistence of consumption, investment and imports that is over -
predicted compared to their empirical counterparts. From the table we can also see that
the model is less successful in replicating the observed correlations between the respective
variables and inflation.
26

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 169
Finally , Figures (E.1a)-(E.1b) in Appendix E plot the estimated structural shocks of the
model. The values plotted are obtained using the two-sided Kalman filter and represent the
most likely values for the respective shock in a particular period, whereas the green areas
provide the highest posterior density intervals (HPDI) (Pfeifer , 2014b). For the estimates
to be sensible, they should be stationary around zero. As can clearly be seen from the
figures, the estimates tend to fluctuate around zero over time and look clearly stationary ,
which gives some positive indication on the statistical validity of the estimated model.
27

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 170
T able 5: Unconditional second moments in the data (1995Q2-2014Q4) and in the model
Standard dev . Autocorrelation Corr with∆ lnY
t
Corr with∆ lnP
d
t
Data Model Data Model Data Model Data Model
∆ lnP
d
t
0.95 0.75 0.62 0.74 0.03 -0.46 1 1
∆ ln
(
W
t
/P
d
t
)
1.09 1.17 0.49 0.59 0.74 0.46 -0.01 -0.38
∆ ln
˜
C
t
1.51 4.91 -0.09 0.67 0.42 0.64 0.1 1 -0.75
∆ ln
˜
I
t
7.06 8.28 -0.15 0.35 0.77 0.67 -0.13 -0.22
ˆ
E
t
3.17 5.13 0.99 0.99 -0.36 0.02 -0.42 0.03
∆ lnY
t
1.18 1.43 0.49 0.47 1 1 0.03 -0.46
∆ ln
˜
X
t
2.80 3.13 0.45 0.50 0.70 0.41 -0.07 -0.05
∆ ln
˜
M
t
3.97 4.34 -0.02 0.30 0.77 0.40 -0.13 -0.05
∆ lnG
t
0.76 0.78 0.22 0.29 0.14 0.09 0.47 0.14
Notes: Standard deviations are in percent. V alues implied by the model are calculated at the posterior mean estimates of
the model’ s parameters.
28

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 171
5 APPLICA TION: WHA T STRUCTURAL SHOCKS DRIVE THE SLOVENIAN
ECONOMY?
After having verified the empirical performance of the model, we use the estimated DSGE
model to analyse historical contributions of structural shocks to the business cycle develop-
ments in the Slovenian economy . In particular , we focus our attention to analyse the main
driving forces behind the real GDP growth and its components during the sample period
with special focus on the recent recessions. Before proceeding to such analyses, it is useful
to discuss the impulse response functions and variance decompositions to understand the
reaction and properties of the shocks.
5.1. Impulse r esponse analysis
This section briefly discusses the impulse response functions of some selected variables
23
from shocks that appear to be, based on a historical decomposition of the data (discussed
in more detail in Subsection 5.3), the most important in driving macroeconomic fluctua-
tions in Slovenia. The results are reported in Figures F. 1 -F. 9 in Appendix F, displaying
impulse responses up to 20 quarters. These figures portray a Bayesian version of the im-
pulse responses which are presented in terms of mean responses of endogenous variables
(solid line) together with the 5% and 95% posterior intervals (dashed lines). Notice that all
quantities are reported as log deviations from the steady state (i.e. percentage deviations).
W e first focus on the impulse responses to a permanent (unit-root) technology shock that
captures permanent shifts in total factor productivity (see Figure F. 1 ). As can be seen in the
figure, this shock induces all variables (except the real exchange rate and private consump-
tion) to rise. After the initial drop, it has also a positive impact on private consumption.
Figure F. 2 plots the impulse responses to a transitory (stationary) technology shock. As
expected, this shock has expansionary ef fects on the economy . When such a shock hits
the economy , mar ginal cost of domestic firms decreases, which reduces domestic inflation
and increases domestic output. Since the increase of output outperforms the decrease in
inflation, government consumption is decreased by the government. One noticeable ob-
servation is that the model predicts negative response for employment. The reason for this
is that the model includes various rigidities that restrict the increase in aggregate demand,
which further induces a fall in employment as firms have become more productive.
Next, we present the impulse response functions to an investment-specific technology
shock (also referred to as a shock to the mar ginal ef ficiency of investment), which af fects
the transformation of investment into physical capital (see Figure F. 3 ). A positive realiza-
23
Although the model includes 81 endogenous variables, we restrict our attention to key variables only . These
variables include the GDP and its main components, domestic inflation, CPI inflation, real wages, employment
and real exchange rate.
29

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 172
tion of this shock is associated with an increase in investment. This induces an increase
in aggregate demand and output in the economy . Interestingly , domestic inflation slightly
decreases after this shock, while private consumption increases. Further , if we look at the
impulse responses of government consumption, we can see that the government reduces its
consumption to dampen demand. The expansion in the economy drives up imports, while
exports suf fer from higher prices caused by increasing domestic mar ginal costs.
Figure F. 4 refers to the case when the economy is hit by a consumption preference shock.
This shock causes an increase in consumption, investment and output. T o meet the higher
demand, firms increase capital utilisation and employment. Firms therefore face rising
mar ginal costs, and they respond by increasing prices. Higher domestic prices drive up
CPI inflation. This, in turn, induces the government to decrease its consumption to counter
the expansion in the economy .
Next, we present the impulse responses to a negative labour supply shock (i.e., an increase
in the disutility of working, ζ
h
t
). The impulse responses are presented in Figure F. 5 . This
shock leads to a decline in hours worked and to an increase in the real wage. This increase
in the real wage leads to an increase in mar ginal cost and inflation. Through the usual
aggregate demand ef fects, the result is a recession in the economy .
In the following, we discuss the impulse responses to four mark-up shocks. Figure F. 6 de-
picts the impulse responses to a domestic mark-up shock. As a consequence of this shock,
domestic inflation increases. Higher domestic prices lead to a decrease in demand of do-
mestic consumption and investment goods. Consumption demand is also shifted towards
imported goods that are cheaper than domestic production. V olumes of imports therefore
increase. As a consequence, lower domestic production has a negative impact on both
hours worked and wages. Higher domestic prices also negatively af fect the competitive-
ness of exports. All these factors cause a decrease in the GDP growth rate. Government
consumption, which follows a fiscal rule, decreases on impact in response to the increase
in inflation.
Figures F. 7 and F. 8 contains the impulse response functions to an imported consumption
and investment shock, respectively . After the imported consumption shock, the prices of
imported consumption goods increase. This leads households to buy fewer of these goods.
The increase in imported consumption inflation also drives up CPI inflation. Because im-
ported consumption goods are now more expensive relative to the domestic ones, expendi-
ture switching towards domestic goods works to expand the economy . As domestic firms
see mar ginal costs go up, they increase their prices and domestic inflation increases. This
has a negative ef fect on exports. Nevertheless, output increases because of increased do-
mestic demand. The government therefore reduces demand in the economy by decreasing
its consumption.
Figure F. 8 shows impulse response functions to an imported investment mark-up shock.
Following this shock, the prices of imported investment goods rise up. As a consequence,
the resulting relative price ef fects induce investment to fall whereas consumption increases.
30

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 173
Decrease in demand for imported investment goods causes a reduction in imports. Domes-
tic inflation rises up. Due to higher domestic prices, export decreases. Because of reduced
exports, production in the economy falls and thus output decreases. T o stimulate the econ-
omy , the government increases its consumption.
Finally , in Figure F. 9 we present the dynamics of the economy following an export mark-
up shock. After this shock prices of exported goods rise up. This leads to a fall in exports
and consequently domestic firms produce less output. Lower production forces firms to
reduce demand for labour and capital services, pushing down wages and rental rate of
capital. This reduces mar ginal costs, allowing domestic firms to reduce prices on domestic
goods. The fall in domestic inflation also works to reduce CPI inflation. Consequently ,
this has a positive ef fect on domestic demand. Since the increase in domestic demand is
not suf ficient to of f-set the fall in exports, output falls. Fiscal policy therefore responds by
raising government consumption.
5.2. V ariance decompositions
In this section, we use the estimated model to decompose the unconditional variances
of the observable variables into the contributions of the structural shocks. Although the
primary interest of this paper is to investigate the background of the GDP (and its main
components) fluctuations, we also present results for some other macroeconomic aggre-
gates. The results are presented in T able 6, where we report the unconditional variance
decomposition analysis (i.e., evaluated at the infinite horizon)
24
computed at the posterior
mean for selected observable variables.
T o facilitate the presentation, we divide the shocks into five categories. The first contains
technology shocks: the stationary (ε
ϵ,t
), unit-root (ε
µ z,t
), investment-specific (ε
Υ,t
), and
asymmetric technology (ε
˜ z
∗
,t
) shocks. The second category includes supply shocks: the
labour supply shock (ε
ζ
h
,t
) and shocks to the mark-ups of the domestic (ε
λ
d
,t
), imported
consumption (ε
λ
mc
,t
), imported investment (ε
λ
mi
,t
), and export (ε
λ
x
,t
) goods. The third
category contains the domestic demand shock: the consumption preference shock (ε
ζ
c
,t
).
The fourth category includes foreign shocks: the uncovered interest rate parity (ε
˜
ϕ,t
), for -
eign output (ε
y
∗
,t
), foreign inflation (ε
π
∗
,t
) and foreign interest rate (ε
R
∗
,t
) shocks. Finally ,
we have the fiscal policy shocks: the government spending (ε
g,t
) and rate of transfers to
households (ε
τ
tr
,t
) shocks.
It is evident from the table that technology shocks play the most important role in fluc-
tuation of the GDP and investment growth. Our results show that roughly 50% and 80%
24
In general, Dynare allows for two types of variance decompositions. The first one is the variance decom-
position that is obtained under stoch_simul command and is calculated at the calibrated parameter combination,
e.g., at the posterior mean (at the average over the parameters), while the second one is the so-called Bayesian
variance decomposition, which is the average of the variance decomposition over the parameter draws. Further -
more, Dynare distinguishes between the conditional and unconditional variance decomposition. The first one is
at a particular time horizon, while the second one is at horizon infinity .
31

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 174
of fluctuations in GDP and investment growth rates, respectively , are due to technology
shocks. In this context, the most important technology shocks are the investment-specific
technology shocks. Among the 50% (80%) of fluctuations of GDP (investment) growth
rates explained by technology shocks, investment-specific technology shocks account for
around 38% (78%). Furthermore, we can observe that supply shocks are the main drivers
of fluctuations in the domestic inflation, real wages, consumption, exports and imports.
More specifically , domestic inflation is mainly driven by domestic mark-up shocks. In
our case they account for 41% of the total variation. Moreover , domestic mark-up and
imported consumption mark-up shocks appear to have a leading role in explaining con-
sumption growth fluctuations. They explain about 28% of total volatility . The shocks
most responsible for the variability of real wages are labour supply shocks (34%). The
export mark-up shocks turn out to be the key drivers for the exports, contributing to ap-
proximately 91% of total volatility , while the imported investment mark-up shocks play
the most important role in accounting for the variation in imports, explaining about 54%
of total volatility in that variable. Next, our estimates suggest that demand shocks have
some importance in our framework in the sense that they explain about 15% of the vari-
ance in GDP growth rates, but their contribution to the remaining variables is negligible.
Finally , we can observe that foreign and fiscal shocks explain a small fraction of variability
in all variables and thus do not play an important role in explaining the Slovenian business
cycle.
25
25
The small impact of foreign shocks may be due to the simplified representation of the foreign block, which
is modelled as a V AR model.
32

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 175
T able 6: V ariance decompositions (in %) evaluated at the infinite horizon
V ariable T echnology Supply Demand Foreign Fiscal
lnP
d
t
33.70 57.01 0.27 7.34 0.10
∆ (
lnW
t
/P
d
t
)
36.95 60.59 0.19 1.03 0.38
∆ ln
˜
C
t
27.52 66.27 3.45 2.59 0.07
∆ ln
˜
I
t
79.82 16.03 0.24 2.90 0.01
ˆ
E
t
60.30 32.25 4.56 2.45 0.06
∆ lnY
t
49.60 33.26 14.84 0.98 0.62
∆ ln
˜
X
t
4.13 92.52 0 2.54 0
∆ ln
˜
M
t
41.70 55.1 1 0.17 2.19 0
∆ lnG
t
35.88 5.44 0.27 1.62 55.85
Notes: The unconditional variance decomposition is performed at the posterior mean esti-
mates of the model’ s parameters. Shocks are aggregated as explained in the main text.
5.3. Historical decompositions
The economic developments in Slovenia in recent years have been characterized by one of
the biggest decline in economic growth in the European Union. Since 2008, Slovenia has
experienced a double-dip recession. After a significant decline in GDP over the 2008-2009
period, the period of short-lived recovery began, but in the last quarter of 201 1 Slovenia
again dropped into recession. The question that arises is, what were the main driving forces
behind the decline in GDP during the recent recession? T o answer this question, we calcu-
late historical decompositions that allow us to investigate the role of shocks in explaining
the movement of observable variables over the sample period. In discussing the results,
we focus on four variables: GDP , private consumption, investment, import and export. All
variables are in real terms. The historical decomposition of real GDP growth is provided
in Figure 1, while the remaining graphs are presented in Appendix G. In all graphs the bold
black line represents the estimate of the smoothed observed variables
26
(best guess for the
observed variables given all observations) derived from the Kalman smoother , while the
coloured bars correspond to the contribution of the respective smoothed shock to the devia-
tion of the smoothed observable variable from its steady state (Pfeifer , 2014b). Bars above
the horizontal axis represent positive shock contributions, while bars below the horizontal
axis show negative contributions.
Figure 1 decomposes the growth rate of real GDP dynamics over the underlying period.
Prior to the crisis, Slovenia was characterized by a very high growth rate of the GDP .
During the period 1996-2005, the annual growth rate of real GDP averaged 4%. The high-
est growth rate of GDP was achieved in the years 2006-2008, reaching its peak in 2007,
when it was 6.7%. The historical decomposition results show that while domestic mark-up
26
The smoothed series results from the Kalman smoother . They are the best guess of the variables given the
information for the whole sample. Given that they are observed, their best guess is the actual value. Hence, there
should be no dif ference unless one assumes they are observed only with measurement error (Pfeifer , 2014b).
33

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 176
shocks had a positive ef fect on economic growth in this period, stationary technology and
labour supply shocks had a negative one. The intuition behind this result can be explained
as follows: During this period domestic, mark-up shocks appear to have contributed signif-
icantly to lower inflation, which in turn stimulated the economy . This could be contributed
to the entrance of Slovenia to the EMU, since this process was characterized by the ef forts
aimed to achieve s uf ficiently low inflation to satisfy the euro adoption, as well as by the
conver gence of previously high Slovenian interest rates towards lower interest rates in the
Euro Area. This enabled the Slovenian banks to get access to low interest rate credits from
abroad, which were mainly intended for the corporate sector . The competition among
banks has further induced a decrease in ef fective interest rates and thus reduced borrow-
ing costs for a business. At the same time, negative transitory (stationary) technology and
labour supply shocks that resulted in lower ef ficiencies of production and higher wages,
caused an upward inflationary pressures, which af fected GDP growth negatively .
Since the first quarter of 2008, the GDP growth has experienced persistent declines un-
til reaching the bottom in the first quarter of 2009. In that quarter , the GDP declined by
about 5% relative to the previous quarter . As the model’ s estimates suggest, this negative
dynamics was mainly driven by investment-specific technology , consumption preference
and export mark-up shocks. Investment-specific technology shocks to a lar ge extent re-
flect a drop in foreign and domestic orders, followed by a decline in investment. This
ef fect was further compounded by tougher access to financial resources. Consumption
preference and export mark-up shocks also contributed to the slowdown in economic ac-
tivity in the 2008-2009 period. It is likely that consumption preference shocks reflect the
reduction in households’ income (in combination with the precautionary saving) while ex-
port mark-up shocks could capture the loss of external competitiveness from an increase
in wages, reflected in a series of negative labour supply shocks identified right before the
crisis. The analysis also points out that fiscal and foreign shocks played a smaller but nev-
ertheless noticeable role in driving the Slovenian business cycle. According to the model,
fiscal shocks have contributed positively over the whole pre-crisis period. The positive
ef fect was still visible in the early stages of the crisis, when loose fiscal policy mitigated
the economic slowdown, although during the ongoing recession fiscal tightening (as a re-
sult of austerity measures adopted to consolidate public finances) was suppressing GDP .
However , it should be noted that the impact of fiscal policy shocks was small compared
to other shocks, which suggests a relatively minor significance of changes in fiscal pol-
icy for cyclical fluctuations in GDP growth. The results regarding the ef fects of foreign
shocks show that the direction of foreign shocks has reversed its course in 2010, from
having a negative ef fect on GDP growth during the period 2008-2010 to having a positive
influence by the middle of 2010, where the latter can be attributed to improvements in
the economic situation in Slovenia’ s main trading partners. Moreover , the historical de-
composition also suggests that investment-specific technology shocks continued to be the
main sources of blocking recovery in more recent years, especially in the years 2012-2013,
when Slovenia fell into its second recession, in which GDP declined by about -2% (quarter -
on-quarter) in real terms. This result is obviously connected to banking system problems.
Namely , the Slovenian banking system has accumulated a lar ge amount of non-performing
bad loans in the last years, resulting in a credit crunch which in turn caused a cutback in
34

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 177
corporate investment and impeded economic activity . The model also identifies the im-
portant role of permanent (unit-root) technology shocks in explaining the movements of
real GDP growth, from having a small but positive impact on GDP growth dynamics in
the pre-crisis period to having negative one in recessionary periods, in particular between
2012-2013, and whose negative ef fects also lasted during the recovery phase. This re-
sult could be considered as associated with the lack of productivity-enhancing and other
structural reforms in the run-up to the crisis. On the other hand, transitory (stationary)
technology shocks have had a positive impact on economic growth, especially from 2013
onwards. This result may be interpreted as resulting from a temporary greater tendency
of corporate sector to take restructuring measures in response to the crisis to enhance its
production ef ficiency . If we compare the two recessionary periods, we can observe that in
contrast to the first period (2008-2009), when investment-specific technology shocks were
accompanied by consumption preference and export mark-up shocks, in the second period
(2012-2013), export mark-up shocks made virtually no contribution to the downturn, point-
ing to a recovery in exports thanks to improving foreign demand, and more importantly ,
to wage moderation and productivity gains, which translated into considerable competi-
tiveness gains and strong export performance. According to the results obtained, it can
also be observed that at the end of the sample period, when the recovery of ficially began,
consumption preference shocks were the main contributors to the pace of economic recov-
ery , presumably due to the increased consumer confidence, resolution of banking system
problems, and the improvement in the labour market situation.
T urning now to the main components of GDP , Figure G.1 in Appendix G shows the histor -
ical decomposition of consumption growth. As can be seen from the figure, the movement
of consumption growth was af fected by a variety of structural shocks. Consistent with
the variance decomposition results, the shocks most important for explaining dynamics
of consumption growth rates over the sample period were stationary technology shocks,
investment-specific technology shocks, domestic mark-up shocks and imported consump-
tion mark-up shocks. Figures G.2-G.3 (see Appendix G) plot the historical decomposition
results for investment and imports, respectively . As can be seen from the figures, the
investment-specific technology shocks and the imported investment mark-up shocks ex-
plained most of the variation in these two variables. Finally , Figure G.4 (see Appendix
G) portrays historical decomposition of the growth rates of exports. As illustrated in the
figure, almost all historical variation in exports was due to export mark-up shocks.
35

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 178
Figure 1 : Historical decomposition of GDP gr owth in terms of structural shocks
Notes: The smoothed observed time series is plotted excluding its mean.
6 CONCLUSION
The main objective of this paper is to present and estimate an open-economy DSGE model
for the Slovenian economy . The model we use closely follows that of Adolfson et al.
(2007) and Masten (2010). Using a data set that extends from 1995Q1 to 2014Q4 for
Slovenia, we estimate the model using Bayesian estimation techniques and compute the
contribution of structural shocks to the cyclical variation of key macroeconomic variables.
After the estimation, we first describe the estimation results and perform several tests on
the quality of the estimation process. Further , we evaluate the model’ s empirical perfor -
mance. Overall, the estimation results are satisfactory . The diagnostic tests indicate that
the estimation is robust in what concerns the quality of the numerical posterior kernel opti-
mization and the conver gence of the MCMC procedure. Furthermore, the majority of the
parameters appear to be well identified by the data and the data fit of the model is good.
The obtained estimates for the structural parameters of interest are generally in line with
the literature and, in most cases, seem to make sense from an economic point of view .
In the last part of the paper , the empirical importance of various types of structural shocks
in explaining macroeconomic fluctuations in the Slovenian economy is studied using im-
pulse responses, variance and shock decompositions. Our main findings can be summa-
rized as follows. The variance decomposition results show that the investment-specific
technology shock is the major driving force of the growth rates of GDP and investment.
36

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 179
Moreover , domestic mark-up shocks are estimated to have a leading role in explaining con-
sumption growth and inflation fluctuations. The labour supply shocks explain the majority
of the variance of real wages. The variance of imports growth rates is explained mainly
by imported investment mark-up shocks, while the exported mark-up shocks account for
most of the variation in exports growth rates. The ef fect of consumption preference shocks
on the economy is estimated to be rather limited, with the lar gest influence on the GDP
and consumption growth rates. Finally , fiscal and foreign shocks are estimated to have a
negligible ef fect in our framework.
Last, using historical decompositions, we estimate the individual contributions of each
structural shock to the movements in GDP growth rates (and its main components) over
the sample period, focusing mainly on the two recessionary periods: 2008-2009 and 2012-
2013. Our results suggest that investment-specific technology shocks accounted for a sig-
nificant portion of the drop in output from 2008 onwards. This result accords with a drop
in foreign and domestic orders followed by a decline in investment (mostly at the begin-
ning of the crisis), as well as with a significant tightening of credit availability , thereby
reducing expenditures on investment, which produced a decrease in the aggregate demand
and output. Consumption preference and export mark-up shocks were another important
sources that contributed to the slowdown in economic activity , especially in the first reces-
sion (2008-2009), most likely reflecting the reduction in households’ income (in combina-
tion with the precautionary saving) and the fall in exports, mainly due to the deterioration
of external competitiveness as wages increased faster than productivity before the crisis
years, respectively . A noticeable but smaller impact was also exerted by foreign and fiscal
shocks. Furthermore, the results show that permanent (unit-root) technology shocks also
contributed to the developments of GDP growth rates during the analysed period. While
in the pre-crisis period these shocks had a small but positive impact on GDP growth rates,
in periods of the crisis, they contributed importantly to the GDP decline. This result could
be considered as associated with the lack of productivity-enhancing and other structural
reforms in the run-up to the crisis. On the contrary , transitory (stationary) technology
shocks had a stimulating impact, especially from 2013 onwards. This finding may capture
the ef fect of measures adopted to improve production ef ficiency . The comparison between
the two recessions also shows that the role of export mark-up shocks decreased in 2010,
from having a significantly negative ef fect on GDP growth during the period 2008-2009
to making virtually no contribution to the economic downturn between 2012-2013, point-
ing to a recovery in exports thanks to improving foreign demand, but more importantly ,
to wage moderation and productivity gains, which translated into considerable competi-
tiveness gains and strong export performance. In addition, consumption preference also
importantly contributed to the sur ge in GDP growth in the most recent years, which could
be interpreted as a consequence of the increased consumer confidence, the resolution of
banking system problems and the recovery in the labour market.
37

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 180
REFERENCES
Adjemin, S., Bastani, H., Karamé, F ., Juillard, M., Maih, J., Mihoubi, F ., Perendia, G.,
Pfeifer , J., Ratto, M., & V illemot, S. (2014). Dynare: Reference Manual V ersion 4
(W orking Paper No. 1). Retrieved May 10, 2015, from
http://www.dynare.org/wp- repo/dynarewp001.pdf
Adolfson, M., Laseén, S., Lindé, J., & V illani, M. (2007). Bayesian Estimation of an
Open Economy DSGE Model with Incomplete Pass-Through. Journal of International
Economics, 72(2), 481-51 1.
Adolfson, M., Laseén, S., Lindé, J., & Svensson, L. E. O. (2014). Monetary Policy T rade-
Of fs in an Estimated Open-Economy DSGE Model. Journal of Economic Dynamics &
Contr ol, 42 (C), 33-49.
Almeida, V . (2009). Bayesian Estimation of a DSGE Model for the Portuguese Economy .
Bank of Portugal W orking Papers Series No. 14. Retrieved April 12, 2018, from https:
//papers.ssrn.com/sol3/papers.cfm?abstract_id=1535285
Altig, D., Christiano L., Eichenbaum, M., & Linde, J. (201 1). Firm-Specific Capital, Nom-
inal Rigidities and the Business Cycle. Review of Economic Dynamics, 14(2), 225-247.
An, S., & Schorfheide, F . (2007). Bayesian Analysis of DSGE Models. Econometric
Review , 26 (2-4), 1 13-172.
Benigno, G., & Benigno, P . (2003). Price Stability in Open Economies. Review of Eco-
nomic Studies, 70(4), 743-764.
Blanchard, O. (2016). Do DSGE Models Have a Futur e?. Retrieved February 8, 2017,
from https://piie.com/system/files/documents/pb16- 11.pdf
Blanchard, O. (2017). Further Thoughts on DSGE Models. Retrieved February 8, 2017,
from https://piie.com/blogs/realtime- economic- issues- watch/further-
thoughts- dsge- models
Brooks, S., & Gelman, A. (1998). General Methods for Monitoring Conver gence of Itera-
tive Simulations. Journal of Computational and Graphical Statistics, 7 (4), 434-455.
Calvo, G. (1983). Staggered Prices in a Utility-Maximizing Framework. Journal of Mon-
etary Economics, 12 (3), 383-398.
Canova, F . (2007). Bayesian Analysis of DSGE Models by S. An and F . Schorfheide.
Econometric Reviews, 26 (2-4), 187-192.
38

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 181
Caprirolo, G., & Lavrač, V . (2003). Monetary and Exchange Rate Policy in Slovenia.
Ezoneplus W orking Paper No. 17G. Retrieved April 12, 2018, from
https://www.econstor.eu/bitstream/10419/31944/1/480746710.pdf
Christiano, L., Eichenbaum, M., & Evans, C. (2005). Nominal Rigidities and the Dynamic
Ef fects of a Shock to Monetary Policy . Journal of Political Economy , 1 13 (1), 1-45.
Clarida, R., Galí, J., & Gertler , M. (1999). The Science of Monetary Policy: A New
Keynesian Perspective. Journal of Economic Literatur e, 37 (2), 1661-1707.
Clarida, R., Galí, J., & Gertler , M. (2001). Optimal Monetary Policy in Open vs. Closed
Economies: An Integrated Approach. American Economic Review , 91 (2), 248-252.
Cúrdia, V ., Ferrero, A., Cee Ng, G., & T ambalotti, A. (2012). Inter est Rate Rules in
DSGE Models: T racking the Efficient Real Inter est Rate. Retrieved April 12, 2018, from
https://www.newyorkfed.org/medialibrary/media/research/economists/
tambalotti/Draft_20130322_Web.pdf
D’Auria, F ., Pagano, A., Ratto, M., & V ar ga, J. (2009). A Comparison of Structural
Reform Scenarios Across the EU Member States: Simulation-Based Analysis Using the
QUEST Model with Endogenous Growth (Economic Papers 392). Brussels: European
Commission.
Erceg, C., Henderson, D., & Levin, A. (2000). Optimal Monetary Policy with Staggered
W age and Price Contracts. Journal of Monetary Economics, 46 (2), 281-313.
Erceg, C., & Lindé, J. (2013). Fiscal Consolidation in a Currency Union: Spending Cuts
vs. T ax Hikes. Journal of Economic Dynamics & Contr ol, 37 (2), 422-445.
Fernández-V illaverde, J. (2010). The Econometrics of DSGE Models. SERIEs, 1(1-2),
3-49.
Galí, J., & Monacelli, T . (2005). Monetary Policy and Exchange Rate V olatility in a Small
Open Economy . Review of Economic Studies, 72 (3), 707-734.
Galí, J. (2008). Monetary Policy , Inflation and the Business Cycle: An Intr oduction to the
New Keynesian Framework. Princeton: Princeton University Press.
Hamilton, J. (1994). T ime Series Analysis. Princeton University Press.
Klein, P . (2000). Using the Generalized Schur Form to Solve a Multivariate Linear Ratio-
nal Expectations Model. Journal of Economic Dynamics & Contr ol, 24(10), 1405-1423.
Lütkepohl, H., & Krätzig, M. (2004). Applied T ime Series Econometrics. New Y ork, NY :
Cambridge University Press.
39

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 182
Mancini Grif foli, T . (2013). DYNARE User Guide: An Intr oduction to the Solution &
Estimation of DSGE Models. Retrieved May 10, 2015, from
http://www.dynare.org/documentation- and- support/user- guide/Dynare-
UserGuide- WebBeta.pdf/view
Marcellino, M., & R ychalovska, Y . (2014). Forecasting with a DSGE Model of a Small
Open Economy within the Monetary Union. Journal of For ecasting, 33(5), 315-338.
Masten, I. (2008). Optimal Policy in Presence of Balassa-Samuelson-T ype Productivity
Shocks. Journal of Comparative Economics, 36 (1), 120-141.
Masten, I. (2010). Dinamični stohastični model splošnega ravnovesja Slovenije SLODSGE
1.0. Retrieved May 15, 2014, from https://www.dlib.si/stream/URN:NBN:SI:DOC-
FVPZR20C/f89623bf- 3439- 4bba- 8c89- a9f92a58dc0e/PDF
Pfeifer , J. (2014a). A Guide to Specifying Observation Equations for the Estimation of
DSGE Models. Retrieved May 10, 2015, from https://sites.google.com/site/
pfeiferecon/dynare
Pfeifer , J. (2014b). An Intr oduction to Graphs in Dynar e. Retrieved May 10, 2015, from
https://sites.google.com/site/pfeiferecon/dynare
Romer , P . (2016). The T r ouble with Macr oeconomics. Retrieved June 10, 2017, from
https://ccl.yale.edu/sites/default/files/files/The%20Trouble%20with%
20Macroeconomics.pdf
Schmitt-Grohé, S. & Uribe, M. (2003). Closing Small Open Economy Models. Journal
of International Economics, 61 (1), 163-185.
Sims, C. A. (2002). Solving Linear Rational Expectations Models. Computational Eco-
nomics, 20 (1-2), 1-20.
Smets, F ., & W outers, R. (2003). An Estimated Dynamic Stochastic General Equilibrium
Model of the Euro Area. Journal of the Eur opean Economic Association, 1 (5), 1 123-1 175.
Smets, F ., & W outers, R. (2007). Shocks and Frictions in US Business Cycles: A Bayesian
DSGE Approach. American Economic Review , 97 (3), 586-606.
V illemot, S. (201 1). Solving Rational Expectations Models at First Order: What Dynare
Does (W orking Paper No. 2). Retrieved May 10, 2015, from http://www.dynare.org/
wp- repo/dynarewp002.pdf
W oodford, M. (2003). Inter est and Prices: Foundations of a Theory of Monetary Policy .
Princeton: Princeton University Press.
40

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 183
APPENDICES
A COMPLETE MODEL IN LOG-LINEARIZED FORM
This appendix presents the log-linearized equations of the model. In what follows, a vari-
able with a hat denotes the log deviation from steady-state values (ˆ x
t
=
xt−x
x
≈ lnx
t
− lnx
for any variable x
t
, where x is the steady-state level), while the overhead tilde indicates
that a variable is measured as dif ference from its steady-state value, i.e. ˜ x
t
≡ x
t
− x.
Because the model comprises the unit-root technology shock, all real variables have to
be scaled with the trend level of technology z
t
in order to render them stationary . The
resulting stationary variables are denoted by lower -case letters, that is,x
t
=
Xt
zt
.
Domestic Phillips curve:
ˆ π
d
t
=
β
1+βκ
d
E
t
ˆ π
d
t+1
+
κ
d
1+βκ
d
ˆ π
d
t−1
+
(1−ξ
d
)(1−βξ
d
)
ξ
d
(1+βκ
d
)
(
� mc
t
+
ˆ
λ
d
t
)
. (48)
Phillips curve for the imported consumption goods:
ˆ π
m,c
t
=
β
1+βκ
m,c
E
t
ˆ π
m,c
t+1
+
κ
m,c
1+βκ
m,c
ˆ π
m,c
t−1
+
(1−ξ
m,c
)(1−βξ
m,c
)
ξ
m,c
(1+βκ
m,c
)
(
� mc
m,c
t
+
ˆ
λ
m,c
t
)
.
(49)
Real mar ginal cost for domestic firms:
� mc
t
= αˆ r
k
t
+(1−α)
ˆ
¯ w
t
−ˆ ϵ
t
(50)
Rental rate of capital:
ˆ r
k
t
=ˆ µ z,t
+
ˆ
¯ w
t
+
ˆ
H
t
−
ˆ
k
t
(51)
Phillips curve for the imported investment goods:
ˆ π
m,i
t
=
β
1+βκ
m,i
E
t
ˆ π
m,i
t+1
+
κ
m,i
1+βκ
m,i
ˆ π
m,i
t−1
+
(1−ξ
m,i
)(1−βξ
m,i
)
ξ
m,i
(1+βκ
m,i
)
(
� mc
m,i
t
+
ˆ
λ
m,i
t
)
.
(52)
Real mar ginal cost for the importing firms (consumption goods):
� mc
m,c
t
=−� mc
x
t
−ˆ γ
x,∗
t
−ˆ γ
mc,d
t
. (53)
41

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 184
Real mar ginal cost for the importing firms (investment goods):
� mc
m,i
t
=−� mc
x
t
− ˆ γ
x,∗
t
− ˆ γ
mi,d
t
. (54)
Phillips curve for the exporting firms:
ˆ π
x
t
=
β
1+βκ
x
E
t
ˆ π
x
t+1
+
κ
x
1+βκ
x
ˆ π
x
t−1
+
(1−ξ
x
)(1−βξ
x
)
ξ
x
(1+βκ
x
)
(
� mc
x
t
+
ˆ
λ
x
t
)
. (55)
Real mar ginal cost for the exporting firms:
� mc
x
t
= � mc
x
t−1
+ˆ π
d
t
− ˆ π
x
t
. (56)
Real wage equation:
E
t


α
0
ˆ
¯ w
t−1
+α
1
ˆ
¯ w
t
+α
2
ˆ
¯ w
t+1
+α
3
(
ˆ π
d
t
−
ˆ
¯ π
c
t
)
+α
4
(
ˆ π
d
t+1
−ρ
¯ π
cˆ
¯ π
c
t
)
+α
5
(
ˆ π
c
t−1
−
ˆ
¯ π
c
t
)
+α
6
(
ˆ π
c
t
−ρ
¯ π
cˆ
¯ π
c
t
)
+α
7
ˆ
ψ
z,t
+α
8
ˆ
H
t
+α
9
ˆ
ζ
h
t
+α
10
ˆ τ
b
t


=0 , (57)
where:


















α
0
α
1
α
2
α
3
α
4
α
5
α
6
α
7
α
8
α
9
α
10


















=




















b
w
ξ
w
σ
L
λ
w
−b
w
(
1+βξ
2
w
)
b
w
βξ
w
−b
w
ξ
w
b
w
βξ
w
b
w
ξ
w
κ
w
−b
w
βξ
w
κ
w
(1−λ
w
)
−(1−λ
w
)σ
L
−(1−λ
w
)
(1−λ
w
)
τ
b
(1−τ
y
+τ
b
)




















and:
b
w
=
[λ
w
σ
L
−(1−λ
w
)]
[(1−βξ
w
)(1−ξ
w
)]
.
Euler equation for consumption:
E
t



−bβˆ c
t+1
+
(
μ
2
t
+b
2
β
)
ˆ c
t
−bμ
z
ˆ c
t−1
+bμ
z
(ˆ μ
z,t
−βˆ μ
z,t+1
)
+(μ
z
−bβ)(μ
z
−b)
ˆ
ψ
z,t
+(μ
z
−bβ)(μ
z
−b)ˆ γ
c,d
t
−(μ
z
−b)
(
μ
z
ˆ
ζ
c
t
−bβ
ˆ
ζ
c
t+1
)


 =0 . (58)
First order condition w .r .t. i
t
:
E
t
{
ˆ
P
k
′
,t
+
ˆ
Υ
t
−μ
2
z
˜
S
′′
[(
ˆ
i
t
−
ˆ
i
t−1
)
−β
(
ˆ
i
t+1
−
ˆ
i
t
)
+ˆ μ
z,t
−βˆ μ
z,t+1
]}
=0 . (59)
42

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 185
First order condition w .r .t. b
t+1
:
E
t




−
ˆ
ψ
z,t
+
πµ
z
−τ
k
β
µ z
π
(
ˆ
ψ
z,t+1
− ˆ µ z,t+1
−ˆ π
d
t+1
+
ˆ
R
t
)
+β
τ
k
µ z
π
(
ˆ
ψ
z,t+1
− ˆ µ z,t+1
−ˆ π
d
t+1
)




=0 . (60)
First order condition w .r .t.
¯
k
t+1
:
E
t



ˆ
ψ
z,t
+ˆ µ z,t+1
−
ˆ
ψ
z,t+1
−
β(1−δ)
µ z
ˆ
P
k
′
,t+1
+
ˆ
P
k
′
,t
−
µ z
−β(1−δ)
µ z
ˆ r
k
t+1


 =0 . (61)
Law of motion for capital:
ˆ
¯
k
t+1
=(1−δ)
1
µ z
ˆ
¯
k
t
−(1−δ)
1
µ z
ˆ µ z,t
+
[
1−(1−δ)
1
µ z
]
ˆ
Υ
t
+
[
1−(1−δ)
1
µ z
]
ˆ
i
t
.
(62)
Capacity utilization rate:
ˆ u
t
=
ˆ
k
t
−
ˆ
¯
k
t
=
1
σ
a
ˆ r
k
t
. (63)
Aggregate resource constraint:
(1−ω
c
)
(
γ
c,d
)
ηc
c
y
(
ˆ c
t
+η
c
ˆ γ
c,d
t
)
+(1−ω
i
)
(
γ
i,d
)
ηi
i
y
(
ˆ
i
t
+η
i
ˆ γ
i,d
t
)
+
g
y
ˆ g
t
+
y
∗
y
(
ˆ y
∗
t
−η
f
ˆ γ
x,∗
t
+
ˆ
˜ z
∗
t
)
= λ
d
[
ˆ ϵ
t
+α
(
ˆ
k
t
− ˆ µ z,t
)
+(1−α)
ˆ
H
t
]
−
(
1−τ
k
)
r
k
¯
k
y
1
µ z
(
ˆ
k
t
−
ˆ
¯
k
t
)
.
(64)
Equilibrium law of motion for net foreign assets:
ˆ a
t
=−y
∗
� mc
x
t
−η
f
y
∗
ˆ γ
x,∗
t
+y
∗
ˆ y
∗
t
+y
∗
ˆ
˜ z
∗
t
+(c
m
+i
m
)ˆ γ
f
t
−c
m
[
−η
c
(1−ω
c
)
(
γ
c,d
)
−(1−ηc)
ˆ γ
mc,d
t
+ˆ c
t
]
+i
m
[
−η
i
(1−ω
i
)
(
γ
i,d
)
−(1−ηi)
ˆ γ
mi,d
t
+
ˆ
i
t
]
+
R
πµ
z
ˆ a
t−1
.
(65)
CPI inflation:
43

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 186
ˆ π
c
t
=
[
(1−ω
c
)
(
γ
d,c
)
1−ηc
]
ˆ π
d
t
+
[
(ω
c
)(γ
mc,c
)
1−ηc
]
ˆ π
m,c
t
. (66)
Investment price inflation:
ˆ π
i
t
=
[
(1−ω
i
)
(
γ
d,i
)
1−ηi
]
ˆ π
d
t
+
[
(ω
i
)
(
γ
mi,i
)
1−ηi
]
ˆ π
m,i
t
. (67)
Gross domestic product:
ˆ y
t
= λ
d
[
ˆ ϵ
t
+α
(
ˆ
k
t
− ˆ µ z,t
)
+(1−α)
ˆ
H
t
]
. (68)
Real ef fective exchange rate:
ˆ x
t
=−ω
c
(γ
c,mc
)
−(1−ηc)
ˆ γ
mc,d
t
−ˆ γ
x,∗
t
− � mc
x
t
. (69)
Employment equation:
∆ ˆ
E
t
=
β
1+β
E
t
∆ ˆ
E
t+1
+
(1−ξ
e
)(1−βξ
e
)
(1+β)ξ
e
(
ˆ
H
t
−
ˆ
E
t
)
. (70)
Domestic interest rate:
ˆ
R
t
=
ˆ
R
∗
t
+
ˆ
˜
ϕ t
−
˜
ϕ a
ˆ a
t
. (71)
Government budget constraint:
b
ˆ
b
t
+t
ˆ
t
t
=
Rb
π
d
µ z
(
ˆ
R
t−1
+
ˆ
b
t−1
−ˆ π
d
t
− ˆ µ z,t
)
+tr
�
tr
t
+gˆ g
t
. (72)
Government expenditures:
� gex
t
=
tr
gex
�
tr
t
+
g
gex
ˆ g
t
+
r
gex
R
πµ
z
(R−1)
ˆ
R
t−1
+
r
gex
ˆ
b
t−1
πµ
z
−
r
gex
ˆ π
t
πµ
z
−
r
gex
ˆ µ z,t
πµ
z
.
(73)
T ransfers to households:
�
tr
t
=� τ
tr
t
+
ˆ
¯ w
t
+
ˆ
H
t
. (74)
Fiscal policy rule for government consumption:
ˆ g
t
= ρ
g
ˆ g
t−1
−ϕ π
ˆ π
c
t
−ϕ y
ˆ y
t
−ϕ b
ˆ
b
t
−ϕ d
�
def
t
+ε
g,t
. (75)
44

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 187
T ax on consumption:
ˆ
t
a
t
= ω
c
(γ
c,mc
)
−(1−ηc)
ˆ γ
mc,d
t
+ˆ c
t
. (76)
T axes and contributions on wages:
ˆ
t
b
t
=
ˆ
¯ w
t
+
ˆ
H
t
. (77)
Public debt interest payments:
˜
t
c
t
=
Rb
πµ
z
ˆ
R
t−1
+
(R−1)b
πµ
z
ˆ
b
t−1
−
(R−1)b
πµ
z
ˆ π
t
−
(R−1)b
πµ
z
ˆ µ z,t
. (78)
Interest on the amount of the capital services:
˜
t
d
t
=
r
k
k
µ z
(
ˆ r
k
t
+
ˆ
k
t
− ˆ µ z,t
)
. (79)
Interest on the amount of foreign bond holdings:
˜
t
e
t
=
R−1
πµ
z
˜ a
t−1
. (80)
Profit of domestic firms:
�
Π
d
t
= y
(
λ
d
−1
λ
d
)
ˆ y
t
−
1
λ
d
(y+ϕ )
[
α
(
ˆ µ z,t
+
ˆ
H
t
−
ˆ
k
t
)
+
ˆ
¯ w
t
−ˆ ϵ
t
]
. (81)
Profit of importing firms:
�
Π
m
t
=
{
c
m
(
γ
mc,d
−
1
λ
d
)
η
c
[
−(1−ω
c
)
(
1
γ
c,mc
γ
mc,d
)
1−ηc
]
+γ
mc,d
c
m
}
ˆ γ
mc,d
t
+c
m
(
γ
mc,d
−
1
λ
d
)
ˆ c
t
+
{
i
m
(
γ
mi,d
−
1
λ
d
)
η
i
[
−(1−ω
i
)
(
1
γ
i,mi
γ
mi,d
)
1−ηi
]
+γ
mi,d
i
m
}
ˆ γ
mi,d
t
+i
m
(
γ
mi,d
−
1
λ
d
)
ˆ
i
t
+
1
γ
f
[(
η
m,c
η
m,c
−1
)
c
m
+
(
η
m,i
η
m,i
−1
)
i
m
]
ˆ γ
f
t
.
(82)
Profit of exporting firms:
45

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 188
�
Π
x
t
=−y
∗
� mc
x
t
. (83)
T otal tax revenue:
ˆ
t
t
=
τ
c
t
a
t
ˆ
t
a
t
+
τ
y
t
b
t
ˆ
t
b
t
+
τ
k
t
(
˜
t
c
t
+
˜
t
d
t
+
˜
t
e
t
+
˜
t
f
t
)
. (84)
Deficit:
�
def
t
= gex� gex
t
−t
ˆ
t
t
. (85)
Debt-to-GDP ratio:
ˆ
b
y,t
=
ˆ
b
t
− ˆ y
t
. (86)
Deficit-to-GDP ratio:
�
def
y,t
=
�
def
t
y
−
def
y
ˆ y
t
. (87)
Relative prices:
ˆ γ
mc,d
t
=ˆ γ
mc,d
t−1
+ˆ π
m,c
t
−ˆ π
d
t
(88)
ˆ γ
mi,d
t
=ˆ γ
mi,d
t−1
+ˆ π
m,i
t
−ˆ π
d
t
(89)
ˆ γ
x,∗
t
=ˆ γ
x,∗
t−1
+ˆ π
x
t
−ˆ π
∗
t
(90)
ˆ γ
f
t
= � mc
x
t
+ˆ γ
x,∗
t
(91)
ˆ γ
c,d
t
= ω
c
(γ
mc,c
)
(1−ηc)
ˆ γ
mc,d
t
(92)
ˆ γ
i,d
t
= ω
i
(
γ
mi,i
)
(1−ηi)
ˆ γ
mi,d
t
(93)
Exogenous shock processes:
ˆ
ξ
t
= ρ
ξ
ˆ
ξ
t−1
+ε
ξ,t
,ε
ξ,t
iid
∼ N
(
0,σ
2
ξ
)
, (94)
46

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 189
where ξ
t
=
{
µ z,t
,ϵ
t
,λ
j
t
,ζ
c
t
,ζ
h
t
,Υ
t
,
˜
ϕ t
,˜ z
∗
t
,τ
tr
t
,g
t
}
for j= {d,mc,mi,x}.
47

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 190
B DA T A SOURCES AND DESCRIPTION
T able B.1: List of variables used in the estimation and their sour ces
Symbol Description Country Source
Yt GDP . Gross domestic product in millions of euro, chain-linked vol-
umes, reference year 2005, SA
Sl Eurostat
Ct Private consumption. Household and NPISH final consumption ex-
penditure in millions of euro, chain-linked volumes, reference year
2005, SA
Sl Eurostat
It Investment. Gross fixed capital formation in millions of euro, chain-
linked volumes, reference year 2005, SA
Sl Eurostat
Gt Government consumption. Final consumption expenditure of gen-
eral government in millions of euro, chain-linked volumes, refer -
ence year 2005, SA
Sl Eurostat
Xt Exports. Exports of goods and services in millions of euro, chain-
linked volumes, reference year 2005, SA
Sl Eurostat
Mt Imports. Imports of goods and services in millions of euro, chain-
linked volumes, reference year 2005, SA
Sl Eurostat
Wt Gr oss wages and salaries. Gross wages and salaries (income struc-
ture of GDP), current prices, millions of euro, SA
Sl SORS
Et Employment. Employment (domestic concept), persons (in 1000),
SA
Sl SORS
P
d
t
GDP deflator . Price index, reference year 2005, SA Sl Eurostat
P
c
t
CPI index. Consumer price index, current month/average of the
year 2005, not SA
Sl Eurostat/ECB
xt Real exchange rate. Real ef fective exchange rate, consumer price
index deflator , reference year 2005, 28 trading partners
Sl Eurostat
Rt Domestic inter est rate. Monetary interest rate on new loans to non-
financial corporations in domestic currency in percent
Sl BS/IMAD
Y
∗
t
For eign GDP . Gross domestic product in millions of euro, chain-
linked volumes, reference year 2005, SA
EA12 Eurostat
P
∗
t
For eign GDP deflator . Price index, reference year 2005, SA EA12 Eurostat
R
∗
t
For eign inter est rate. 12-month money market interest rate in per -
cent
EA12 Eurostat
Notes: SA: seasonally adjusted; SORS: Statistical Of fice of the Republic of Slovenia; IMAD: Institute of
Macroeconomic Analysis and Development of the Republic of Slovenia; BS: Bank of Slovenia
48

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 191
C PRIOR AND POSTERIOR DISTRIBUTIONS
Figure C.1a : Prior and posterior distributions of the structural parameters, friction pa-
rameters
0.2 0.4 0.6 0.8 1
0
1
2
3
4
ξw
0.5 0.6 0.7 0.8 0.9 1
0
5
10
15
ξd
0.2 0.4 0.6 0.8 1
0
1
2
3
ξmc
0 0.2 0.4 0.6 0.8 1
0
0.5
1
1.5
2
2.5
ξmi
0.4 0.6 0.8 1
0
2
4
6
ξx
0.2 0.4 0.6 0.8
0
2
4
6
8
10
12
ξe
0.2 0.4 0.6 0.8 1 1.2
0
0.5
1
1.5
2
κw
0 0.2 0.4 0.6
0
1
2
3
4
5
6
κd
0 0.2 0.4 0.6 0.8
0
1
2
3
4
κm,c
0 0.2 0.4 0.6 0.8
0
1
2
3
4
κm,i
0 0.2 0.4 0.6 0.8
0
1
2
3
4
κx
0.2 0.4 0.6 0.8 1
0
5
10
15
b
Notes: Prior (black) vs. posterior (red) distributions for the estimated structural parameters. The gray dashed
vertical line is the posterior mode obtained from the posterior kernel maximization. Estimates obtained from
Bayesian estimation of the DSGE model using Slovenian macroeconomic data from 1995Q1-2014Q4.
49

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 192
Figure C.1b : Prior and posterior distributions of the structural parameters, friction pa-
rameters (cont.)
0 1 2 3 4
0
2
4
6
ηi
0 2 4 6
0
0.2
0.4
0.6
0.8
ηf
5 10 15
0
0.05
0.1
0.15
0.2
0.25
0.3
˜ S
′′
1.004 1.005 1.006 1.007 1.008
0
200
400
600
800
μz
0 0.05 0.1
0
10
20
30
˜ φa
Notes: Prior (black) vs. posterior (red) distributions for the estimated structural parameters. The gray dashed
vertical line is the posterior mode obtained from the posterior kernel maximization. Estimates obtained from
Bayesian estimation of the DSGE model using Slovenian macroeconomic data from 1995Q1-2014Q4.
50

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 193
Figure C.1c: Prior and posterior distributions of the structural parameters, shock pr o-
cesses parameters
0.2 0.4 0.6 0.8 1
0
5
10
15
20
ρµ z
0 0.5 1
0
1
2
3
ρǫ
0 0.5 1
0
0.5
1
1.5
2
2.5
3
ρΥ
0.2 0.4 0.6 0.8 1
0
10
20
30
40
ρ ˜ z ∗
0 0.2 0.4 0.6 0.8
0
1
2
3
4
ρζc
0 0.5 1
0
0.5
1
1.5
2
ρζh
0.2 0.4 0.6 0.8 1
0
5
10
15
20
25
ρ ˜ φ
0 0.2 0.4 0.6 0.8 1
0
1
2
3
ρλd
0 0.5 1
0
0.5
1
1.5
ρλm,c
0 0.5 1
0
0.5
1
1.5
2
ρλm,i
0 0.5 1
0
0.5
1
1.5
2
2.5
ρλx
0 2 4 6 8 10
x 10
−3
0
500
1000
1500
σµ z
Notes: Prior (black) vs. posterior (red) distributions for the estimated structural parameters. The gray dashed
vertical line is the posterior mode obtained from the posterior kernel maximization. Estimates obtained from
Bayesian estimation of the DSGE model using Slovenian macroeconomic data from 1995Q1-2014Q4.
51

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 194
Figure C.1d: Prior and posterior distributions of the structural parameters, shock pr o-
cesses parameters (cont.)
0 0.02 0.04 0.06 0.08
0
50
100
150
200
σǫ
0.01 0.02 0.03 0.04 0.05
0
200
400
600
σΥ
5 10 15
x 10
−3
0
200
400
600
800
σ ˜ z ∗
2 4 6 8 10
x 10
−3
0
200
400
600
σζc
0 2 4 6 8 10
x 10
−3
0
200
400
600
σζh
1 2 3 4 5
x 10
−3
0
500
1000
1500
2000
2500
σ ˜ φ
0 5 10 15
x 10
−3
0
200
400
600
800
1000
1200
σλd
0 5 10 15
x 10
−3
0
200
400
600
800
σλm,c
0 0.2 0.4 0.6 0.8 1
0
100
200
300
400
σλm,i
0 0.01 0.02 0.03 0.04
0
100
200
300
400
σλx
2 4 6 8 10
x 10
−3
0
200
400
600
σg
0 2 4 6 8 10
x 10
−3
0
200
400
600
σtr
Notes: Prior (black) vs. posterior (red) distributions for the estimated structural parameters. The gray dashed
vertical line is the posterior mode obtained from the posterior kernel maximization. Estimates obtained from
Bayesian estimation of the DSGE model using Slovenian macroeconomic data from 1995Q1-2014Q4.
52

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 195
Figure C.1e: Prior and posterior distributions of the structural parameters, policy param-
eters
0 0.5 1
0
0.5
1
1.5
2
2.5
ρ tr
0 0.5 1
0
1
2
3
ρ g
−0.2 0 0.2 0.4 0.6 0.8
0
1
2
3
4
φπ
0 0.2 0.4 0.6 0.8
0
2
4
6
8
φy
0 0.05 0.1
0
10
20
30
40
φb
0 0.05 0.1
0
10
20
30
40
φd
Notes: Prior (black) vs. posterior (red) distributions for the estimated structural parameters. The gray dashed
vertical line is the posterior mode obtained from the posterior kernel maximization. Estimates obtained from
Bayesian estimation of the DSGE model using Slovenian macroeconomic data from 1995Q1-2014Q4.
53

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 196
D DA T A AND ONE-SIDED PREDICTED V ALUES FROM THE MODEL
Figure D.1 : Data (thick black) and one-sided Kalman-filter ed pr edictions (thin r ed)
1995 1998 2001 2004 2007 2010 2013
−0.05
0
0.05
Domestic inflation
1995 1998 2001 2004 2007 2010 2013
−0.1
0
0.1
Consumption
1995 1998 2001 2004 2007 2010 2013
−0.5
0
0.5
Investment
1995 1998 2001 2004 2007 2010 2013
−0.05
0
0.05
Output
1995 1998 2001 2004 2007 2010 2013
−0.2
0
0.2
Export
1995 1998 2001 2004 2007 2010 2013
−0.2
0
0.2
Import
1995 1998 2001 2004 2007 2010 2013
−0.05
0
0.05
Government consumption
1995 1998 2001 2004 2007 2010 2013
−0.05
0
0.05
CPI inflation
1995 1998 2001 2004 2007 2010 2013
−0.05
0
0.05
Real exchange rate
1995 1998 2001 2004 2007 2010 2013
−0.05
0
0.05
Real wage
1995 1998 2001 2004 2007 2010 2013
−0.1
0
0.1
Employment
1995 1998 2001 2004 2007 2010 2013
−0.05
0
0.05
Foreign output
1995 1998 2001 2004 2007 2010 2013
0
0.01
0.02
Foreign inflation
1995 1998 2001 2004 2007 2010 2013
0
0.1
0.2
Foreign interest rate
1995 1998 2001 2004 2007 2010 2013
0
0.2
0.4
Domestic interest rate
 
 
Data Model (one−sided Kalman filtered predictions)
Notes: The plot shows deviations from steady state/trend.
54

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 197
E SMOOTHED SHOCKS
Figure E.1a : Smoothed (two-sided Kalman filter ed) estimates of the structural shocks (de-
viations fr om steady state)
1995 1998 2001 2004 2007 2010 2013
−3
−2
−1
0
1
2
x 10
−3
Unit−root technology
1995 1998 2001 2004 2007 2010 2013
−0.1
−0.05
0
0.05
0.1
Stationary technology
1995 1998 2001 2004 2007 2010 2013
−0.15
−0.1
−0.05
0
0.05
0.1
Investment−specific technology
1995 1998 2001 2004 2007 2010 2013
−0.015
−0.01
−0.005
0
0.005
0.01
Asymmetric technology
1995 1998 2001 2004 2007 2010 2013
−0.015
−0.01
−0.005
0
0.005
0.01
Consumption preference
1995 1998 2001 2004 2007 2010 2013
−5
0
5
x 10
−3
Labour supply
1995 1998 2001 2004 2007 2010 2013
−0.01
−0.005
0
0.005
0.01
0.015
Risk premium
1995 1998 2001 2004 2007 2010 2013
−0.01
−0.005
0
0.005
0.01
0.015
Domestic markup
1995 1998 2001 2004 2007 2010 2013
−0.01
−0.005
0
0.005
0.01
0.015
Imported consumption markup
 
 
Mean Higest posterior density interval (HPDI)
Notes: The plot shows deviations from steady state.
55

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 198
Figure E.1b: Smoothed (two-sided Kalman filter ed) estimates of the structural shocks (de-
viations fr om steady state) (cont.)
1995 1998 2001 2004 2007 2010 2013
−2
−1
0
1
2
3
Imported investment markup
1995 1998 2001 2004 2007 2010 2013
−0.05
0
0.05
Export markup
1995 1998 2001 2004 2007 2010 2013
−0.015
−0.01
−0.005
0
0.005
0.01
Foreign output
1995 1998 2001 2004 2007 2010 2013
−0.01
−0.005
0
0.005
0.01
Foreign inflation
1995 1998 2001 2004 2007 2010 2013
−3
−2
−1
0
1
2
x 10
−3
Foreign interest rate
1995 1998 2001 2004 2007 2010 2013
−0.02
−0.01
0
0.01
0.02
0.03
Government spending
1995 1998 2001 2004 2007 2010 2013
−10
−5
0
5
x 10
−3
Rate of benefits
 
 
Mean Higest posterior density interval (HPDI)
Notes: The plot shows deviations from steady state.
56

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 199
F IMPULSE RESPONSE FUNCTIONS
Figure F .1: Impulse r esponses to a unit-r oot technology shock
0 5 10 15 20
−2
0
2
4
6
x 10
−4 Domestic inflation
0 5 10 15 20
0
1
2
3
x 10
−3 Real wage
0 5 10 15 20
−2
−1
0
1
2
x 10
−3 Consumption
0 5 10 15 20
−2
0
2
4
x 10
−3 Investment
0 5 10 15 20
−3
−2
−1
0
x 10
−3 Real exchange rate
0 5 10 15 20
0
0.5
1
1.5
2
x 10
−3 Government consumption
0 5 10 15 20
−2
−1
0
1
2
x 10
−3 Employment
0 5 10 15 20
0
0.5
1
1.5
x 10
−3 Output
0 5 10 15 20
0
0.5
1
1.5
2
x 10
−3 Export
0 5 10 15 20
0
1
2
3
x 10
−3 Import
0 5 10 15 20
−2
0
2
4
x 10
−4 CPI inflation
 
 
Mean Higest posterior density interval (HPDI)
Notes: The solid line shows the average impulse responses results over the MCMC parameter draws; the dashed
lines at the 5% and 95% posterior intervals. The impulse horizon is measured in quarters.
57

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 200
Figure F .2: Impulse r esponses to a stationary technology shock
0 5 10 15 20
−2
−1
0
1
x 10
−3 Domestic inflation
0 5 10 15 20
−1
0
1
2
x 10
−3 Real wage
0 5 10 15 20
−5
0
5
10
x 10
−3 Consumption
0 5 10 15 20
−2
0
2
4
x 10
−3 Investment
0 5 10 15 20
−2
0
2
4
x 10
−3 Real exchange rate
0 5 10 15 20
−10
−5
0
5
x 10
−4 Government consumption
0 5 10 15 20
−6
−4
−2
0
2
x 10
−3 Employment
0 5 10 15 20
−1
0
1
2
x 10
−3 Output
0 5 10 15 20
−5
0
5
10
x 10
−4 Export
0 5 10 15 20
−5
0
5
10
x 10
−4 Import
0 5 10 15 20
−15
−10
−5
0
5
x 10
−4 CPI inflation
 
 
Mean Higest posterior density interval (HPDI)
Notes: The solid line shows the average impulse responses results over the MCMC parameter draws; the dashed
lines at the 5% and 95% posterior intervals. The impulse horizon is measured in quarters.
58

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 201
Figure F .3: Impulse r esponses to an investment-specific technology shock
0 5 10 15 20
−15
−10
−5
0
5
x 10
−4 Domestic inflation
0 5 10 15 20
−2
0
2
4
x 10
−3 Real wage
0 5 10 15 20
−5
0
5
10
x 10
−3 Consumption
0 5 10 15 20
−0.05
0
0.05
0.1
0.15
Investment
0 5 10 15 20
0
2
4
6
x 10
−3 Real exchange rate
0 5 10 15 20
−10
−5
0
5
x 10
−4 Government consumption
0 5 10 15 20
−0.01
−0.005
0
Employment
0 5 10 15 20
−5
0
5
10
x 10
−3 Output
0 5 10 15 20
−5
0
5
10
x 10
−4 Export
0 5 10 15 20
−0.02
0
0.02
0.04
Import
0 5 10 15 20
−10
−5
0
5
x 10
−4 CPI inflation
 
 
Mean Higest posterior density interval (HPDI)
Notes: The solid line shows the average impulse responses results over the MCMC parameter draws; the dashed
lines at the 5% and 95% posterior intervals. The impulse horizon is measured in quarters.
59

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 202
Figure F .4: Impulse r esponses to a consumption pr efer ence shock
0 5 10 15 20
−1
0
1
2
3
x 10
−4 Domestic inflation
0 5 10 15 20
−6
−4
−2
0
2
x 10
−4 Real wage
0 5 10 15 20
−5
0
5
10
15
x 10
−3 Consumption
0 5 10 15 20
−3
−2
−1
0
1
x 10
−3 Investment
0 5 10 15 20
−4
−2
0
2
4
x 10
−4 Real exchange rate
0 5 10 15 20
−10
−5
0
5
x 10
−4 Government consumption
0 5 10 15 20
0
2
4
6
x 10
−3 Employment
0 5 10 15 20
−5
0
5
10
x 10
−3 Output
0 5 10 15 20
−10
−5
0
5
x 10
−5 Export
0 5 10 15 20
−1
0
1
2
x 10
−3 Import
0 5 10 15 20
−5
0
5
10
x 10
−5 CPI inflation
 
 
Mean Higest posterior density interval (HPDI)
Notes: The solid line shows the average impulse responses results over the MCMC parameter draws; the dashed
lines at the 5% and 95% posterior intervals. The impulse horizon is measured in quarters.
60

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 203
Figure F .5: Impulse r esponses to a labour supply shock
0 5 10 15 20
−5
0
5
10
x 10
−4 Domestic inflation
0 5 10 15 20
−5
0
5
10
x 10
−3 Real wage
0 5 10 15 20
−6
−4
−2
0
2
x 10
−3 Consumption
0 5 10 15 20
−3
−2
−1
0
1
x 10
−3 Investment
0 5 10 15 20
−3
−2
−1
0
1
x 10
−3 Real exchange rate
0 5 10 15 20
−2
0
2
4
x 10
−4 Government consumption
0 5 10 15 20
−4
−3
−2
−1
0
x 10
−3 Employment
0 5 10 15 20
−10
−5
0
5
x 10
−4 Output
0 5 10 15 20
−6
−4
−2
0
2
x 10
−4 Export
0 5 10 15 20
−10
−5
0
5
x 10
−4 Import
0 5 10 15 20
−2
0
2
4
6
x 10
−4 CPI inflation
 
 
Mean Higest posterior density interval (HPDI)
Notes: The solid line shows the average impulse responses results over the MCMC parameter draws; the dashed
lines at the 5% and 95% posterior intervals. The impulse horizon is measured in quarters.
61

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 204
Figure F .6: Impulse r esponses to a domestic mark-up shock
0 5 10 15 20
−2
0
2
4
6
x 10
−3 Domestic inflation
0 5 10 15 20
−10
−5
0
5
x 10
−3 Real wage
0 5 10 15 20
−0.03
−0.02
−0.01
0
0.01
Consumption
0 5 10 15 20
−10
−5
0
5
x 10
−3 Investment
0 5 10 15 20
−10
−5
0
5
x 10
−3 Real exchange rate
0 5 10 15 20
−1
−0.5
0
0.5
1
x 10
−3 Government consumption
0 5 10 15 20
−6
−4
−2
0
x 10
−3 Employment
0 5 10 15 20
−6
−4
−2
0
2
x 10
−3 Output
0 5 10 15 20
−2
−1
0
1
x 10
−3 Export
0 5 10 15 20
−2
−1
0
1
2
x 10
−3 Import
0 5 10 15 20
−2
0
2
4
x 10
−3 CPI inflation
 
 
Mean Higest posterior density interval (HPDI)
Notes: The solid line shows the average impulse responses results over the MCMC parameter draws; the dashed
lines at the 5% and 95% posterior intervals. The impulse horizon is measured in quarters.
62

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 205
Figure F .7: Impulse r esponses to an imported consumption mark-up shock
0 5 10 15 20
−2
0
2
4
x 10
−4 Domestic inflation
0 5 10 15 20
−5
0
5
10
15
x 10
−4 Real wage
0 5 10 15 20
−0.01
0
0.01
0.02
0.03
Consumption
0 5 10 15 20
−1
0
1
2
3
x 10
−3 Investment
0 5 10 15 20
−3
−2
−1
0
x 10
−3 Real exchange rate
0 5 10 15 20
−10
−5
0
5
x 10
−4 Government consumption
0 5 10 15 20
−1
0
1
2
3
x 10
−3 Employment
0 5 10 15 20
−1
0
1
2
3
x 10
−3 Output
0 5 10 15 20
−4
−2
0
2
x 10
−4 Export
0 5 10 15 20
−2
−1
0
1
2
x 10
−3 Import
0 5 10 15 20
−5
0
5
10
15
x 10
−4 CPI inflation
 
 
Mean Higest posterior density interval (HPDI)
Notes: The solid line shows the average impulse responses results over the MCMC parameter draws; the dashed
lines at the 5% and 95% posterior intervals. The impulse horizon is measured in quarters.
63

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 206
Figure F .8: Impulse r esponses to an imported investment mark-up shock
0 5 10 15 20
−2
0
2
4
x 10
−4 Domestic inflation
0 5 10 15 20
−2
0
2
4
6
x 10
−4 Real wage
0 5 10 15 20
−2
−1
0
1
x 10
−3 Consumption
0 5 10 15 20
−0.03
−0.02
−0.01
0
0.01
Investment
0 5 10 15 20
−4
−3
−2
−1
0
x 10
−3 Real exchange rate
0 5 10 15 20
−2
0
2
4
6
x 10
−4 Government consumption
0 5 10 15 20
−3
−2
−1
0
1
x 10
−3 Employment
0 5 10 15 20
−4
−2
0
2
x 10
−3 Output
0 5 10 15 20
−4
−2
0
2
x 10
−4 Export
0 5 10 15 20
−0.04
−0.02
0
0.02
Import
0 5 10 15 20
−2
0
2
4
x 10
−4 CPI inflation
 
 
Mean Higest posterior density interval (HPDI)
Notes: The solid line shows the average impulse responses results over the MCMC parameter draws; the dashed
lines at the 5% and 95% posterior intervals. The impulse horizon is measured in quarters.
64

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 207
Figure F .9 : Impulse r esponses to an export mark-up shock
0 5 10 15 20
−6
−4
−2
0
2
x 10
−4 Domestic inflation
0 5 10 15 20
−10
−5
0
5
x 10
−4 Real wage
0 5 10 15 20
−1
0
1
2
3
x 10
−3 Consumption
0 5 10 15 20
−4
−2
0
2
x 10
−3 Investment
0 5 10 15 20
−1
0
1
2
x 10
−3 Real exchange rate
0 5 10 15 20
−5
0
5
x 10
−4 Government consumption
0 5 10 15 20
−4
−2
0
2
x 10
−3 Employment
0 5 10 15 20
−4
−2
0
2
x 10
−3 Output
0 5 10 15 20
−0.03
−0.02
−0.01
0
0.01
Export
0 5 10 15 20
−2
−1
0
1
x 10
−3 Import
0 5 10 15 20
−4
−2
0
2
x 10
−4 CPI inflation
 
 
Mean Higest posterior density interval (HPDI)
Notes: The solid line shows the average impulse responses results over the MCMC parameter draws; the dashed
lines at the 5% and 95% posterior intervals. The impulse horizon is measured in quarters.
65

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 208
G HIST ORICAL DECOMPOSITIONS
Figure G.1 : Historical decomposition of consumption gr owth in terms of structural shocks
Notes: The smoothed observed time series is plotted excluding its mean.
66

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 209
Figure G.2: Historical decomposition of investment gr owth in terms of structural shocks
Notes: The smoothed observed time series is plotted excluding its mean.
67

ECONOMIC AND BUSINESS REVIEW | VOL. 21 | No. 2 | 2019 210
Figure G.3 : Historical decomposition of import gr owth in terms of structural shocks
Notes: The smoothed observed time series is plotted excluding its mean.
68

A. KUŠTRIN | A DSGE MODEL FOR THE SLOVENIAN ECONOMY ... 211
Figure G.4 : Historical decomposition of export gr owth in terms of structural shocks
Notes: The smoothed observed time series is plotted excluding its mean.
69