F OREIGN SHOCKS IN AN ESTIMATED

MULTI - SECTOR MODEL ∗

D RAGO B ERGHOLT†
February, 2015

Abstract
How are macroeconomic fluctuations in open economies affected by interna-
tional business cycles? To shed some light on this question, I develop and estimate
a medium scale DSGE model for a small open economy. The model incorporates
i) international markets for firm-to-firm trade in production inputs, and ii) producer
heterogeneity where technology and price setting constraints vary across industries.
Using Bayesian techniques on Canadian and US data, I document several macroe-
conomic regularities in the small open economy, all attributed to international dis-
turbances. First, foreign shocks are crucial for domestic fluctuations at all forecast-
ing horizons. Second, productivity is the most important driver of business cycles.
Investment efficiency shocks on the other hand have counterfactual implications for
international spillover. Third, the relevance of foreign shocks accumulates over time.
Fourth, business cycles display strong co-movement across countries, even though
shocks are uncorrelated and the trade balance is countercyclical. Fifth, exchange
rate pass-through to aggregate CPI inflation is moderate, while pass-through at the
sector level is positively linked to the frequency of price changes. Few of these fea-
tures have been accounted for in existing open economy DSGE literature, but all are
consistent with reduced form evidence. The model presented here offers a structural
interpretation of the results.

Keywords: DSGE, small open economy, international business cycles, Bayesian estimation.
JEL Classification: C11, F41, F44.

∗
This work has benefited from discussions with Tommy Sveen, Jesper Lindé, Lars E. O. Svensson, Paul
Levine, Gianluca Benigno, Punnoose Jacob, and Gisle J. Natvik. I am also grateful for comments and dis-
cussions by participants at the 2015 HEC-Montreal-BoC Research Meeting on Dynamic Macroeconomics,
the 46th Money, Macro and Finance (MMF) Conference, the 20th Computing in Economics and Finance
(CEF) Conference, the 2014 NBRE Spring Meeting, the 2013 CAMA Conference on Commodities and
the Macroeconomy, the 36th Annual Meeting of the Norwegian Association of Economists, and seminars
in the Reserve Bank of New Zealand, BI Norwegian Business School, Norwegian School of Economics
(NHH), Norwegian University of Science and Technology (NTNU), and Statistics Norway (SSB). The
views expressed in this paper are those of the author and do not necessarily reflect the views of Norges
Bank. All errors are mine.
†
Norges Bank. E-mail: [email protected].

1
 1 I NTRODUCTION
How and to what extent do international business cycle shocks propagate across coun-
tries? These questions are fundamental in open economy macroeconomics, and of first
order importance for welfare evaluation and policy making. Yet, the jury is still out.
On the one side, a vast number of VAR studies suggest substantial propagation of in-
ternational shocks to open economies.1 But due to it’s reduced form nature, the VAR
framework is largely silent regarding main disturbances and transmission channels at
play. Estimated DSGE models, in contrast, facilitate formal identification of a rich set
of business cycle shocks. But once confronted with data, these models typically have a
hard time accounting for even moderate amounts of international spillover. Perhaps the
most striking example is offered by Justiniano and Preston (2010), who document how
an estimated New Keynesian model attributes virtually all business cycle fluctuations in
Canada to domestic shocks.2
In this paper I revisit the role of international business cycle disturbances within a
multi-sector open economy framework. To this end I develop and estimate a two-country
New Keynesian model, and shed light on how macroeconomic fluctuations are determined
in small open economies (SOEs). Key features of the model are i) international markets
for firm-to-firm trade in production inputs, and ii) producer heterogeneity where firms
operate in segmented markets and face different technological constraints. These mod-
eling choices are guided by data: International input-output matrices reveal substantial
intermediate goods trade, both across diversified industries within countries, and across
country borders. Table 1 reports the intermediate goods share in gross output in all OECD
countries where data were available, as well as in the BRICS economies. About 50% of
gross output in most countries is sold to other firms as production inputs. The interme-
diate goods shares are even higher in exports and imports data – about 60% of all trade
between Canada and US is between firms. Thus, open economy models with only final
goods miss out on more than half of the physical cross-country trade that actually takes
place. Clearly, this might give rise to a mismatch between theory and reality when it
comes to international spillover and synchronization.
The way I model firm-to-firm trade and producer heterogeneity builds on Bouakez,
Cardia, and Ruge-Murcia (2009) and Bergholt and Sveen (2014). Crucially, these fea-
tures create sectoral trade interdependence both within and across economies: First, im-
ported intermediates represent a new cost-channel for spillover of foreign shocks. In
contrast to existing models, where exchange rates only affect domestic firms indirectly
via changes in demand, they also shift supply schedules in the current framework. This
direct exchange rate effect on the domestic production frontier is particularly relevant for
firms who compete in international markets, even more so if these markets are charac-
terized by frequent and large price changes. Second, intersectoral firm-to-firm linkages
induce substantial spillover to relatively non-traded industries. For example, when the
price of manufactured goods deflate, the supply of domestic service firms shifts out. This
1
See e.g. Kose, Otrok, and Whiteman (2003, 2008), Aastveit, Bjornland, and Thorsrud (2011), Crucini,
Kose, and Otrok (2011), Mumtaz, Simonelli, and Surico (2011), and Kose, Otrok, and Prasad (2012).
2
This puzzling result is remarkably robust. Other examples (although they do not necessarily discuss it ex-
plicitly) include Adolfson, Laséen, Lindé, and Villani (2007, 2008), de Resende, Dib, and Kichian (2010),
Rabanal and Tuesta (2010), Dib (2011), and Christiano, Trabandt, and Walentin (2011). Schmitt-Grohé
(1998) demonstrate that also the real business cycle literature fail to account for international spillover.

2
 Table 1: Intermediate trade in OECD and BRICS countries
Country Share Country Share Country Share Country Share

OECD
Australia 0.51 Finland 0.55 Korea 0.59 Slovenia 0.58
Austria 0.49 France 0.51 Luxembourg 0.63 Spain 0.52
Belgium 0.59 Germany 0.50 Netherlands 0.53 Sweden 0.53
Bulgaria 0.57 Greece 0.44 New Zealand 0.55 Switzerland 0.50
Canada 0.50 Hungary 0.62 Norway 0.47 Turkey 0.43
Czech Republic 0.63 Ireland 0.56 Poland 0.55 UK 0.53
Denmark 0.49 Italy 0.54 Portugal 0.54 US 0.46
Estonia 0.60 Japan 0.49 Slovakia 0.63 OECD 0.54
BRICS
Brazil 0.51 China 0.64 India 0.48 Russia 0.49
South Africa 0.50 BRICS 0.52

Note: Intermediate goods share of gross output (OECD data).

is because manufactured goods are important inputs in service production. It follows

that even the supply of completely non-traded firms in general will react to international
shocks. These intersectoral firm-to-firm linkages are crucial as most of aggregate GDP
is produced by domestic service firms. Third, feedback loops in the domestic production
network accelerate initial impulses, resulting in higher order propagation effects.
While Bergholt and Sveen (2014) explain basic mechanisms in a stylized environment,
I extend the setup along several dimensions to facilitate a quantitative assessment, as in
e.g. Adolfson et al. (2007). I estimate structural parameters using Bayesian techniques on
9 Canadian and 8 US time series, but restrict them to fit I-O data in both countries. I then
conduct a broad evaluation of the open economy dimension of macroeconomic fluctua-
tions in Canada (the SOE). Several important results emerge: First, as in wide empirical
literature, foreign shocks account for substantial variation in macroeconomic variables at
all forecasting horizons (20-70%). Second, in a forecasting perspective the role of foreign
shocks tends to build up over time, in line with VAR evidence (see e.g. Cushman and
Zha (1997) and Justiniano and Preston (2010)). Third, while a cocktail of disturbances is
responsible for macroeconomic fluctuations in the very short run, total factor productivity
stands out as the most prominent type of shock over the business cycle. This contrasts
the major role of investment efficiency shocks found in recently estimated DSGE models,
see Justiniano, Primiceri, and Tambalotti (2010, 2011) and Jacob and Peersman (2013).
I argue that these shocks have counterfactual implications for international synchroniza-
tion patterns. Fourth, consistent with the empirical pass-through literature (e.g. Gopinath,
Itskhoki, and Rigobon (2010) and Gopinath and Itskhoki (2010)) I find higher exchange
rate pass-through in sectors with frequent price changes. This feature facilitates interna-
tional spillover of shocks. Finally, when firm-to-firm trade and sectoral heterogeneity are
shut down, the model assigns almost all business cycle fluctuations to domestic events.
The rest of the paper is organized as follows. A multi-sector DGSE model is described
in section 2. Section 3 presents data, calibration and Bayesian estimates. Main empirical
results are reported in section 4. In section 5 I discuss how these results are facilitated by
important transmission channels in the model. Section 6 presents results from a counter-
factual model without intermediate trade and sector heterogeneity. Section 7 concludes.

3
 2 T HE MODEL
I derive a general equilibrium system consisting of two blocks – “home” and “foreign”.
Home is referred to as the domestic economy, while the rest of the world is captured by
the foreign block. My focus is on the limiting case where home is small and has negligible
influence on the world economy. A log-linear approximation around the non-stochastic
steady state is presented below.3 To save space, I restrict attention to the domestic block.

2.1 H OUSEHOLDS
Consider a small open economy (the home economy) with a measure one of symmetric
households. The representative household consists of a continuum of members, with a
P share µj working
fixed  in each production sector j ∈ [1, . . . , J ] in the domestic economy
J
j=1 µj = 1 . Household members consume, work and invest in order to maximize
expected lifetime utility. The maximization problem is subject to a sequence of budget
constraints, with revenues coming from returns on capital, labor income, dividends from
ownership of firms, returns on domestic and foreign bonds, and government transfers.
Optimality conditions for the representative household with respect to consumption, do-
mestic and foreign bond holdings, capital and investment follow below, with prices being
quoted in terms of consumption units:
σ
λt = zU,t − (ct − χC ct−1 ) (1)
1 − χC
λt = Et (λt+1 ) + rt − Et (πt+1 ) (2)
λt = Et (λt+1 ) + rt∗ − Et (πt+1 − ∆et+1 ) − B nf at + zB,t (3)
k


qt = − (rt − Et (πt+1 )) + Et [1 − β (1 − δ)] rt+1 + β (1 − δ) qt+1 (4)
pir,t = qt + zI,t − I [(it − it−1 ) − βEt (it+1 − it )] (5)

The first equation aligns the shadow value of the budget constraint in period t, λt , with
the marginal utility of aggregate consumption ct . σ > 0 and χC ∈ [0, 1] govern the
intertemporal elasticity of substitution and habit persistence in consumption, respectively.
zU,t is a stationary shock to intertemporal preferences. Optimality conditions (2) and (3)
equate the marginal utility of more consumption today with the expected present value of
more future consumption, obtained by investing in domestic and foreign bonds. πt and
∆et are the CPI inflation rate and the nominal depreciation rate, respectively. Nominal
interest rates on domestic and foreign bonds are denoted rt and rt∗ , while nf at is the ratio
of net foreign assets to GDP (measured in absolute deviations from steady state). B > 0
introduces a risk premium on foreign asset returns, as in Adolfson et al. (2007, 2008)
and Christiano et al. (2011). If domestic households are net borrowers, they are charged a
premium. If they are net lenders, they receive a lower return than foreign households. The
risk premium also ensures that steady state is well-defined, see e.g. Schmitt-Grohé and
Uribe (2003). zB,t denotes temporary deviations from interest rate parity, so-called risk
premium shocks. The present value of one more unit of new capital, qt , is characterized
by equation (4). rt − Et (πt+1 ) is the expected real return (real interest rate) foregone
3
A detailed description of the full non-linear model is provided in the online appendix.

4
by not investing in bonds, while rtk is the rental rate on capital in place. The parameters
β ∈ (0, 1) and δ ∈ [0, 1] denote the time discount factor and the capital depreciation rate,
respectively. Finally, equation (5) determines optimal demand for aggregate investment
goods. It effectively equates the relative investment price pir,t with the marginal gain
of investment – the present value of capital net of investment adjustment costs. The
latter is governed by I ≥ 0, as in Christiano, Eichenbaum, and Evans (2005). zI,t is
a stationary shock to the marginal efficiency of investment, a so-called MEI shock. The
optimality conditions (1)-(5) summarize intertemporal household decisions in goods and
asset markets. They are augmented with a capital accumulation equation of the form
kt+1 = (1 − δ) kt + δ (zI,t + it ) , (6)
where kt is capital operational in period t.
Next I turn to sectoral allocations. ct and it are composite functions of sectoral con-
sumption and investment goods, cj,t and ij,t . In turn, these quantities are combinations
of domestically produced (cHj,t , iHj,t ) and imported (cF j,t , iF j,t ) goods, respectively. At
least some international trade takes place in all sectors. However, the trade intensity is
sector specific, implying that import shares in ct and it depend both on the import shares
in each sector, and on the sector weights in aggregate demand baskets. Cost minimization
gives rise to a set of optimality conditions involving associated (real) price indexes, prj,t ,
prHj,t and prF j,t :
ij,t = −ν prj,t − pir,t + it


cj,t = −νprj,t + ct
cHj,t = −η (prHj,t − prj,t ) + cj,t iHj,t = −η (prHj,t − prj,t ) + ij,t (7)
cF j,t = −η (prF j,t − prj,t ) + cj,t iF j,t = −η (prF j,t − prj,t ) + ij,t
The elasticity of substitution between goods from different sectors is ν > 0, while η >
0 denotes the elasticity of substitution between countries. Thus, households substitute
their demand towards sectors and countries with relatively low prices. Up to first order,
one can express aggregate CPI inflation πt and investment goods inflation πti as linear
combinations of domestic sector prices:4
J
X J
X
πt = ξj πj,t πti = $j πj,t prj,t = αj prHj,t + (1 − αj ) prF j,t
j=1 j=1

The weights ξj , $j and αj represent cost shares in steady state.

Sectoral labor markets are constructed similar to that in Erceg, Henderson, and Levin
(2000), but I add a friction in the sense that labor cannot move freely between sectors or
countries within the business cycle horizon.5 To fix ideas, consider the labor market in
sector j. Firms buy labor services from a sector specific labor union. In turn, the union
provides these services by combining working hours from the µj household members
employed in the sector. Among individual workers, only a randomly drawn fraction 1 −
θwj can adjust nominal wages optimally each period. Remaining workers index their
wages partially to lagged CPI inflation. Nominal wage dynamics follow below:
πwj,t = βEt (πwj,t+1 ) + ιw (πt−1 − βπt ) + κwj (mrsj,t − ωj,t ) (8)
4
Note that sectoral prices are linked to aggregate CPI inflation by the identity πj,t = prj,t − prj,t−1 + πt .
5
Still, workers within each country do not have incentives to change sector occupation over time, as real
wages are equal across sectors in steady state.

5
 (1−θwj )(1−βθwj )
κwj = θwj (1+ 1+
governs the responsiveness of πwj,t to time varying markups in

w
wϕ
)
the real wage ωj,t over mrsj,t , the marginal rate of substitution between hours worked
and consumption. ϕ is the inverse Frisch elasticity of labor supply, while w represents
the steady state markup over competitive wages. ιw ∈ [0, 1] is the degree of indexation
among non-optimizing workers. The marginal rate of substitution is

mrsj,t = zU,t + zN,t + ϕnj,t − λt , (9)

where zN,t is referred to as a labor supply shock.

2.2 F IRMS
Domestic sector j is populated by a continuum of profit maximizing firms. Firms cannot
change sectoral occupation over time, in analogy with labor.6 Each firm produces differ-
entiated consumption, investment and intermediate goods, which are sold in domestic and
foreign markets. Production technology is Cobb-Douglas in materials, labor and capital,
augmented with fixed costs. Aggregate output in sector j becomes

yj,t = (1 + p ) [zAj,t + φj mj,t + ψj nj,t + (1 − φj − ψj ) kj,t ] , (10)

where zAj,t is a sector specific productivity shock, p is the steady state price markup on
differentiated goods, and φj , ψj , (φj + ψj ) ∈ (0, 1).
A defining feature of the model is the presence of segmented markets for firm-to-firm
trade. I follow Bouakez et al. (2009) and Bergholt and Sveen (2014), and let mj,t be a
composite of different materials produced in the different sectors. In principle, domestic
production requires intermediate inputs from all firms in all industries in all countries.
Bergholt and Sveen (2014) show how this setup amplifies the interdependencies between
sectors, and therefore increases the potential role for international shocks in otherwise
closed sectors such as the service industry. Cost minimization implies a set of optimality
conditions for the use of intermediate inputs:

mlj,t = −ν prl,t − pm


rj,t + mj,t
mHlj,t = −η (prHl,t − prl,t ) + mlj,t (11)
mF lj,t = −η (prF l,t − prl,t ) + mlj,t

In analogy with consumption and investment bundles, mlj,t denotes sector j’s demand
for materials from sector l, while mHlj,t and mF lj,t represent the domestic and imported
PJ
components, respectively. pm rj,t = l=1 ζlj prl,t is the composite price index associated
with mj,t . Importantly, the weights ζlj can be found from I-O matrices in each country.
The system in (11) shows that optimal factor demand is directed towards those industries
and countries with relatively low factor prices. Finally, material demand is high when
other factors of production are relatively costly, as seen below:

mj,t − nj,t = ωj,t − pm

rj,t (12)
kj,t − mj,t = pm
rj,t − rtk (13)
6
A free entry condition prevents arbitrage opportunities of changing sectoral occupation in steady state.

6
 Producer prices are sticky á la Calvo (1983). Every period, each individual firm can set
it’s price optimally with probability 1 − θpj . Remaining firms resort to a partial indexation
rule. Nominal inflation dynamics for goods sold domestically and abroad follow:

πHj,t = κ1 Et (πHj,t+1 ) + κ2 πHj,t−1 + κj3 (rmcj,t − prHj,t + zM,t ) (14)

∗ ∗ ∗
+ κj3 rmcj,t − p∗rHj,t + zM,t



πHj,t = κ1 Et πHj,t+1 + κ2 πHj,t−1 (15)
β ιp (1−θ )(1−βθ )
The slope coefficients are defined as κ1 = 1+βι p
, κ2 = 1+βι p
, and κj3 = θpjpj(1+βιp )pj ,
where ιp ∈ [0, 1] is the degree of indexation among non-optimizing price setters. Intu-
itively, inflation comes about from time varying markups in prHj,t and p∗rHj,t , the prices
on domestic goods and exports, over marginal costs rmcj,t . zM,t is referred to as a markup
∗
shock. Equation (15), with πHj,t being expressed in international currency, follows from
the assumption that export prices are set in buyer’s currency – so called local currency
pricing (LCP). I choose LCP rather than producer currency pricing (PCP) for two rea-
sons. First, only 4% of Canadian exports to the US is priced in Canadian dollars. Second,
PCP implies full pass-through from exchange rates into domestic inflation, at odds with
the empirical pass-through literature (Gopinath et al., 2010). Marginal costs are

rmcj,t = −zAj,t + φj pm k
rj,t + ψj ωj,t + (1 − φj − ψj ) rt . (16)

Note for future reference that sector level terms of trade is defined as the domestic cur-
rency export-to-import price ratio, i.e. τj,t = p∗rHj,t − prF j,t .

2.3 D OMESTIC ABSORPTION AND GDP

Aggregate domestic absorption of sector j-goods is defined as the sum of consumption,
investment and material components:
J
X
xj,t = γjc cj,t + γji ij,t + γjlm mjl,t (17)
l=1

The coefficients γjc , γji and γjlm depend on the steady state and are defined in the appendix.
I let xHj,t be domestic absorption of domestically produced j-goods, and xF j,t be the
imported counterpart:

xHj,t = −η (prHj,t − prj,t ) + xj,t (18)

xF j,t = −η (prF j,t − prj,t ) + xj,t (19)

In analogy to domestic producer prices, imported inflation can be written as

∗
πF j,t = κ∗1 Et (πF j,t+1 ) + κ∗2 πF j,t−1 + κ∗j3 rmc∗j,t + st − prF j,t + zM,t


, (20)

β ι∗p (1−θpj
∗
)(1−βθpj∗
)
where κ∗1 = 1+βι∗p
, κ∗2 = 1+βι∗p
, and κ∗j3 =
θpj (1+βιp )
∗ ∗
. st is the real exchange
rate between the two countries, rmc∗j,t represents marginal costs abroad, and zM,t ∗
is an
international markup shock. Similarly to domestic absorption of imports, one can define
x∗Hj,t as global absorption of domestically produced j-goods:

x∗Hj,t = −η p∗rHj,t − st − p∗rj,t + x∗j,t

(21)

7
p∗rj,t and x∗j,t represent sector specific prices and quantities in global markets. Market
clearing implies that yj,t = αxj xHj,t + (1 − αxj ) x∗Hj,t , where αxj is the steady state share
of domestic output that is supplied at home. GDP and trade balances at the sector level
are derived according to the expenditure approach:
gdpj,t = γj1 (prj,t + xj,t ) + tbj,t − γj2 pm


rj,t + mj,t (22)
tbj,t = γjex p∗rHj,t + x∗Hj,t − γjim (prF j,t + xF j,t )


(23)
The trade balance is expressed relative to sector GDP and in absolute deviation from
steady state. γjex and γjim represent sector specific export/import-to-GDP ratios respec-
tively, while γj1 and γj2 are found as solutions to the steady state of the model. Finally, by
aggregating across sectors we get economy wide GDP and trade balance:
J
X J
X
gdpt = γj gdpj,t and tbt = γj tbj,t (24)
j=1 j=1

The parameter γj is the steady state share of sector j in aggregate GDP. From the global
economy’s point of view, their debt is in zero net supply because the home economy
engages in only a negligible part of the financial assets trade. Furthermore, I assume that
foreign investors do not hold financial assets in the home economy.

2.4 M ONETARY AND FISCAL POLICY

The model is closed with a specification of monetary and fiscal policy. I follow previous
work in the DSGE literature (see e.g. Justiniano and Preston (2010); Smets and Wouters
(2007); Lubik and Schorfheide (2007)) and assume that monetary policy can be approxi-
mated by a Taylor-type rule of the form
rt = ρr rt−1 + (1 − ρr ) (ρπ πt + ρy gdpt + ρdy ∆gdpt + ρde ∆et ) + zR,t . (25)
ρr , ρπ , ρy , ρdy and ρde are policy coefficients, and zR,t is a monetary policy shock. Regard-
ing fiscal policy, the government faces a period-by-period budget constraint with Ricar-
dian taxes and newly issued government bonds on the income side, and public spending
and maturing bonds on the expenditure side. Under the assumption that public debt is
zero in steady state, one can write, up to a first order approximation, public spending as
fully financed by (possibly time varying) lump-sum taxes.

2.5 E XOGENOUS DISTURBANCES

I assume that all exogenous disturbances in the model follow a univariate AR(1) repre-
sentation in log-linear form:
i.i.d.
zt = ρz zt−1 + σz εz,t , εz,t ∼ N (0, 1) (26)

zt = [zU,t , zN,t , zB,t , zI,t , zM,t , zR,t , zA1,t , . . . , zAJ ,t ]0 is the vector of exogenous distur-
bances. ρz and σz are diagonal, and all non-zero elements in ρz are bounded between
zero and one. Fluctuations in the foreign economy are subject to a similar set of distur-
bances, except that foreign risk premium shocks are negligible due to the small economy
assumption.

8
 3 E STIMATION
Sector heterogeneity induces a non-symmetric equilibrium across different industries. I
solve for the steady state analytically and use the solution to parameterize a log-linear ap-
proximation of the model. The steady state solution is provided in the appendix. Several
model parameters are estimated using Bayesian techniques. This approach has been popu-
larized by e.g. Geweke (1999), Smets and Wouters (2003, 2007), and An and Schorfheide
(2007). Before discussing the results I describe data, the calibration, and priors.

3.1 DATA
To estimate the model I use HP filtered quarterly data from Canada and US (1982Q4-
2007Q4).7 Canada is treated as the SOE, while US proxies the world economy. This
country-pair has been analyzed in a number of two-country SOE-studies, see e.g. Schmitt-
Grohé (1998) and Justiniano and Preston (2010). I divide each economy into three sectors
– the raw material sector, the manufacturing sector, and the service sector. These are clas-
sified according to the North American Industry Classification System (NAICS). Raw
materials constitute NAICS industries 11-21, manufacturing 22-33, and services 41-56
and 71-72 respectively. The industries are exhaustive in the sense that they aggregate
to privately produced GDP. Sector level GDP series are interpolated as the raw data are
available only at annual frequency. In addition, I use as observables quarterly consump-
tion, investment, hours, CPI inflation, and policy rates from both countries, as well as the
bilateral real exchange. This leaves me with a total of 17 time series used for estimation.8
Details about the data set are relegated to the appendix.

3.2 C ALIBRATION AND PRIORS

A subset of the parameters is calibrated according to data and previous studies. Their
values are reported in Table 2. Parameters not related to the multi-sector setup are set
to common values in the literature (see e.g. Smets and Wouters (2007), Adolfson et al.
(2007, 2008), Justiniano and Preston (2010), and Christiano et al. (2011)). Regarding ν,
I choose a value of 0.5 based on Atalay (2013), who estimate sectoral substitution elas-
ticities between 0.85 and essentially zero. ν = 1.5 gives similar results. The remaining
parameters are sector specific, and these deserve further attention. To parameterize sector
specific steady state ratios I rely on the Canadian and US I-O matrices, obtained from
the Structural Analysis Input Output (Total) Database constructed by OECD. The data
reveal large differences across industries. For instance, while almost 70% of consumption
goods is services, manufacturing firms produce the vast majority of investment goods.
Raw materials, while only accounting for about 2% of aggregate consumption and in-
vestment in Canada, constitutes 16% of GDP because of its exports and large supply of
intermediates. Regarding trade, Canadian export-to-GDP ratio varies from 7% in the ser-
vice sector to about 102% in the manufacturing sector. These sector differences represent
7
Results are similar if data are linearly detrended.
8
Raw data are collected from Federal Reserve Economic Database (FRED), Statistics Canada, and
Bureau of Economic Analysis. They are available to the public and can be downloaded from
http://research.stlouisfed.org/fred2/, http://www.statcan.gc.ca/, and http://www.bea.gov/.

9
 Table 2: Calibration

β Time discount factor 0.99 ν Sectoral elasticity 0.5

σ Inverse intertemporal elasticity 1 δ Capital depreciation 0.025
ϕ Inverse labor supply elasticity 2 B Risk premium elasticity 0.01
w , p Mark-up, labor and goods markets 1/7

SOE ROW
(1) (2) (3) (1) (2) (3)
φj Materials share, gross output 0.37 0.66 0.34 0.35 0.54 0.33
ψj Labor share, gross output 0.12 0.21 0.32 0.10 0.22 0.29
γjex , γjim Trade share, sector GDP 0.67 1.02 0.07 – – –
ξj Sector share, consumption 0.02 0.31 0.67 0.01 0.29 0.70
$j Sector share, investment 0.02 0.85 0.13 0.03 0.77 0.20
0.32 0.21 0.03 0.40 0.18 0.01
ζlj Input-output matrix 0.38 0.61 0.32 0.33 0.58 0.28
0.30 0.18 0.65 0.27 0.24 0.71

Note: Calibrated values in benchmark model. The sectors are (1) raw materials, (2) manufacturing, and (3) ser-
vices. The two I-O matrices at the bottom display the fraction of total materials used in each sector that comes
from each of the other sectors. Columns represent consumption (input), and rows production (output).

a key source of disaggregate heterogeneity in the model. Turning to data on materials, we

see that substantial trade in intermediate goods takes place across sectors, as illustrated by
the non-zero off-diagonal elements of the I-O matrices. For instance, the service sector
in Canada buys about 32% of its materials from the manufacturing sector (which trade
extensively in foreign markets). This is the sense in which trade across sectors provides
indirect import in the model, and thereby serves as a potential amplification mechanism
for foreign shocks.
The remaining parameters are estimated. I choose priors in the mid range of those used
by Adolfson et al. (2007), Justiniano and Preston (2010), and Christiano et al. (2011), with
identical distributions across countries on same parameters.9 The substitution elasticity
between domestic and foreign goods is centered around 1 – above estimates by Heath-
cote and Perri (2002), Corsetti, Dedola, and Leduc (2008), and Gust, Leduc, and Sheets
(2009), but below estimates by Adolfson et al. (2007). Regarding Calvo parameters for
wage stickiness, I am not aware of any studies pointing to substantial sectoral differences.
Thus, θwj is centered around 0.75 ∀ j. Priors on sectoral price stickiness are inspired
by a number of microeconomic studies, who show that raw materials and manufactured
goods change prices much more frequently than service goods. For instance, looking
at disaggregate US data, Bils and Klenow (2004), Nakamura and Steinsson (2008) and
Bouakez et al. (2009) find virtually flexible prices in agricultural and raw materials, while
estimated price durations in services range from 1.6 quarters (Bils and Klenow, 2004) to 9
quarters (Bouakez et al., 2009). I choose priors in the mid range of these estimates: Calvo
parameters are set such that average price durations in raw materials, manufacturing and
services are equal to 1.18, 1.25, and 5 quarters respectively. Priors for the seventeen struc-
tural shocks are comparable with e.g. Adolfson et al. (2007), although technology shocks
9
Justiniano and Preston (2010) on the other hand scale up priors on foreign shocks to twice the size of
domestic shocks. This is done in order to induce a more important role for international business cycles.

10
 Table 3: Prior and posterior distributions
Prior Posterior domestic Posterior foreign
Prior(P1,P2) Mode Mean 5%-95% Mode Mean 5%-95%
χC Habit B(0.50,0.10) 0.61 0.57 0.42-0.71 0.56 0.59 0.49-0.70
I Inv. adj. cost N(5.00,1.00) 0.77 1.07 0.55-1.59 2.59 2.91 1.38-4.37
η H-F elasticity G(1.00,0.15) 0.83 0.83 0.74-0.92 – – –
θw1 B(0.75,0.07) 0.76 0.74 0.62-0.87 0.75 0.73 0.59-0.85
θw2 Calvo wages B(0.75,0.07) 0.38 0.37 0.24-0.49 0.75 0.73 0.61-0.86
θw3 B(0.75,0.07) 0.71 0.69 0.55-0.85 0.72 0.71 0.60-0.83
θp1 B(0.15,0.05) 0.11 0.13 0.06-0.21 0.21 0.21 0.17-0.26
θp2 Calvo prices B(0.20,0.05) 0.14 0.14 0.09-0.20 0.30 0.30 0.25-0.35
θp3 B(0.80,0.07) 0.66 0.64 0.57-0.71 0.80 0.81 0.76-0.85
ιw Indexation, πw B(0.50,0.15) 0.30 0.35 0.13-0.56 0.52 0.57 0.31-0.82
ιp Indexation, πp B(0.50,0.15) 0.17 0.21 0.08-0.34 0.87 0.85 0.76-0.94
ρr Smoothing, r B(0.60,0.05) 0.73 0.74 0.69-0.78 0.76 0.75 0.71-0.80
ρπ Taylor, π N(1.80,0.20) 1.95 2.00 1.72-2.30 1.70 1.73 1.47-1.98
ρy Taylor, gdp N(0.13,0.05) 0.03 0.04 0.02-0.06 0.08 0.08 0.05-0.12
ρdy Taylor, ∆gdp N(0.13,0.05) 0.12 0.12 0.05-0.18 0.15 0.15 0.09-0.20
ρde Taylor, ∆e N(0.10,0.05) 0.10 0.11 0.05-0.16 – – –
ρA Technology B(0.70,0.10) 0.90 0.89 0.84-0.94 0.90 0.90 0.86-0.94
ρR Mon. pol. B(0.70,0.10) 0.29 0.29 0.20-0.38 0.31 0.33 0.23-0.42
ρI Investment B(0.70,0.10) 0.51 0.49 0.36-0.62 0.40 0.40 0.28-0.52
ρU Preferences B(0.70,0.10) 0.41 0.46 0.29-0.62 0.59 0.56 0.42-0.70
ρN Labor supply B(0.70,0.10) 0.72 0.66 0.51-0.81 0.72 0.69 0.52-0.86
ρM Markup B(0.70,0.10) 0.50 0.51 0.36-0.65 0.54 0.55 0.44-0.66
ρB UIP B(0.70,0.10) 0.85 0.84 0.76-0.92 – – –
σA1 IG(0.20,2.00) 0.09 0.48 0.04-1.97 5.49 5.61 4.87-6.31
σA2 Sd technology IG(0.50,2.00) 0.71 0.65 0.28-1.01 1.16 1.16 1.00-1.32
σA3 IG(0.20,2.00) 0.90 0.90 0.73-1.08 0.61 0.61 0.50-0.72
σR Sd mon. pol. IG(0.20,2.00) 0.25 0.26 0.21-0.30 0.11 0.12 0.10-0.13
σI Sd investment IG(0.50,2.00) 2.25 3.12 1.69-4.48 5.61 6.31 3.24-9.24
σU Sd preferences IG(0.20,2.00) 1.66 1.62 1.11-2.12 1.16 1.29 0.94-1.64
σN Sd labor supply IG(0.50,2.00) 0.24 3.93 0.12-10.95 0.23 0.47 0.11-0.85
σM Sd markup IG(0.50,2.00) 0.71 0.70 0.52-0.88 1.36 1.37 1.13-1.60
σB Sd UIP IG(0.20,2.00) 0.42 0.45 0.31-0.59 – – –
σe1 IG(0.20,1.00) 6.48 6.56 5.73-7.40 0.09 0.14 0.05-0.24
σe2 Sd mea. err. IG(0.20,1.00) 3.27 3.15 2.57-3.73 2.98 3.01 2.65-3.37
σe3 IG(0.20,1.00) 0.73 0.76 0.64-0.88 0.13 0.14 0.09-0.20

Note: B denotes the beta distribution, N normal, G gamma, IG inverse gamma, P1 prior mean, P2 prior standard de-
viation. Posterior moments are computed from 500000 draws generated by the Random Walk Metropolis-Hastings
algorithm, where the first 200000 are used as burn-in. The volatility of shocks is multiplied by 100 relative to the text.

in services are less volatile than in other sectors. This reflects estimates by Bouakez et al.
(2009), who point to much less volatility in services. TFP differences used here are fairly
conservative compared with their results. Finally, I include a measurement error in each
of the observation equations linking observed GDP series to the model. This is motivated
by the interpolation of sectoral GDP data, which might introduce certain high or low fre-
quency properties not related to the business cycle. The measurement errors are assumed
to be i.i.d. with prior standard deviations centered around 0.2. This is similar to the prior
measurement errors on wages used by Justiniano, Primiceri, and Tambalotti (2013).

11
 3.3 P OSTERIOR ESTIMATES
To build the posterior parameter distribution, I simulate two Random Walk Metropolis-
Hastings chains with 500000 draws per chain, starting at the posterior mode. The first
200000 draws are used as burn-in. I tune the scale of the jumping distribution and ob-
tain acceptance ratios equal to 0.34 and 0.33. Posterior estimates are reported in Table 3.
Most parameters are found to be in line with those found in previous studies, with notable
exceptions discussed below. First, the posterior mode and mean of investment adjustment
costs are significantly smaller in both countries than what is typically found in the DSGE
literature, but still higher than microeconomic estimates (see Groth and Khan (2010)).
This might be due to internal propagation in the model, a point which I will turn back
to later. Second, the estimated price rigidities display large differences across sectors in
both countries, with service sector prices being more sticky than prices in other sectors.
This is consistent with a number of microeconomic studies as discussed earlier (e.g. Bils
and Klenow (2004)), and cannot be accounted for by one-sector models á la Smets and
Wouters (2007). A low Calvo parameter in manufacturing is perhaps also related to the
inclusion of construction firms in that sector, as Bouakez et al. (2009) find that US con-
struction prices are perfectly flexible. Third, there is much less indexation to previous
prices and wages in the Canada than in the US. This might have to do with the open econ-
omy dimension, as other parameters are fairly similar across countries. Also Justiniano
and Preston (2010) report less indexation in Canada compared with the US. Finally, as in
Lubik and Schorfheide (2007), I find some evidence of systematic response by monetary
authorities to exchange rate fluctuations. Turning to the shock processes, we see that tech-
nology shocks are the most persistent, and that the most volatile disturbances in the model
are productivity innovations in raw material sectors and marginal efficiency of investment
shocks. Moreover, productivity is substantially less volatile in the foreign service sector,
in line with results in Bouakez et al. (2009). Finally, note that data are uninformative
about some parameters, in particular those associated with labor supply shocks.

4 E MPIRICAL RESULTS
So far I have presented an estimated multi-sector DSGE model for SOEs. This section
documents the main empirical finding from the estimated model – the significance of for-
eign business cycle shocks for macroeconomic fluctuations in Canada. I restrict attention
to Canadian GDP, consumption, investment, hours, CPI inflation, real wages, net exports,
and the policy rate. Table 4 reports the variance decomposition of domestic forecast er-
rors (FEVDs) at different forecasting horizons. The first column shows the importance of
all foreign innovations combined. Remaining columns report contributions of individual
disturbances.10 Three results stand out. First, at all horizons a substantial fraction of the
forecast error is attributed to foreign shocks. Second, their role in the variance decompo-
sition tends to build up over time. Third, while a cocktail of disturbances is responsible
for macroeconomic fluctuations in the very short run, foreign productivity shocks stand
out as the prominent source of long run volatility. These findings are discussed next.
10
Shocks to the UIP condition are likely a mix of domestic and foreign events. Christiano et al. (2011) label
UIP shocks as foreign, while Justiniano and Preston (2010) include them in the domestic block. I take a
conservative view, and follow the latter definition throughout the analysis.

12
Table 4: Forecast error variance decomposition of foreign shocks (percent)
All foreign
Variable Decomposition
shocks

∗
σA1 ∗
σA2 ∗
σA3 ∗
σR σI∗ σU∗ ∗
σN ∗
σM
Panel A: 1 quarter horizon
GDP 22.01 7.17 0.83 0.00 0.01 7.11 0.08 0.00 6.81
Consumption 11.68 5.74 3.46 0.46 0.09 0.59 0.12 0.00 1.22
Investment 22.30 8.78 7.22 0.73 0.08 2.50 0.25 0.00 2.75
Hours 17.50 1.42 3.71 0.49 0.01 11.46 0.39 0.00 0.01
Interest 37.36 11.53 11.56 1.67 0.80 5.35 0.53 0.00 5.93
Inflation 41.86 15.76 12.36 1.27 0.21 4.49 0.42 0.00 7.35
Wage 45.53 20.82 13.93 1.37 0.24 0.89 0.28 0.00 7.99
Trade balance 37.55 3.01 7.11 1.45 0.22 24.35 1.03 0.00 0.38
Panel B: 4 quarter horizon
GDP 48.30 23.31 14.86 1.46 0.16 1.81 0.12 0.00 6.58
Consumption 20.13 11.41 5.65 0.77 0.05 0.76 0.20 0.00 1.28
Investment 34.88 16.16 12.13 1.11 0.04 2.61 0.33 0.00 2.49
Hours 22.98 6.28 5.69 0.61 0.13 7.73 0.23 0.00 2.31
Interest 38.24 13.69 10.48 1.67 0.38 7.24 0.65 0.00 4.12
Inflation 41.44 15.79 11.87 1.24 0.23 4.64 0.46 0.00 7.21
Wage 49.83 27.12 15.84 1.74 0.10 0.34 0.29 0.00 4.39
Trade balance 33.80 1.91 2.83 0.60 0.15 26.48 0.79 0.00 1.04
Panel C: 8 quarter horizon
GDP 53.17 26.99 18.00 1.94 0.10 1.77 0.14 0.00 4.21
Consumption 19.93 11.90 5.08 0.74 0.04 1.11 0.15 0.00 0.92
Investment 38.45 19.31 13.44 1.09 0.03 2.45 0.27 0.00 1.86
Hours 29.19 10.35 8.72 0.94 0.12 6.60 0.23 0.00 2.23
Interest 36.56 12.46 10.56 1.49 0.37 6.51 0.60 0.00 4.57
Inflation 42.29 15.74 12.15 1.22 0.23 5.03 0.49 0.00 7.43
Wage 50.89 28.53 16.30 1.92 0.06 1.19 0.20 0.00 2.68
Trade balance 31.79 1.78 3.76 0.97 0.16 23.27 0.72 0.00 1.13
Panel D: 20 quarter horizon
GDP 63.17 29.47 22.91 2.07 0.08 5.43 0.10 0.00 3.10
Consumption 39.72 19.43 13.10 1.27 0.05 4.20 0.10 0.00 1.57
Investment 43.56 18.17 14.05 0.98 0.04 8.52 0.23 0.00 1.59
Hours 29.91 10.35 8.91 0.92 0.11 7.22 0.22 0.00 2.17
Interest 35.15 12.08 10.23 1.44 0.35 6.13 0.57 0.00 4.35
Inflation 42.98 16.04 12.46 1.30 0.23 5.14 0.48 0.00 7.33
Wage 65.20 31.99 23.14 2.22 0.06 5.66 0.12 0.00 2.02
Trade balance 33.23 2.12 4.39 1.23 0.14 23.46 0.70 0.00 1.19
Panel E: Long run horizon
GDP 73.77 29.76 30.50 2.27 0.08 8.63 0.10 0.00 2.42
Consumption 75.04 26.98 32.91 2.24 0.06 11.30 0.08 0.00 1.48
Investment 44.88 17.39 15.57 1.11 0.04 9.10 0.21 0.00 1.47
Hours 33.27 11.52 10.62 1.03 0.11 7.62 0.21 0.00 2.16
Interest 43.37 14.85 14.73 1.63 0.31 7.43 0.49 0.00 3.92
Inflation 46.70 17.05 14.74 1.42 0.22 5.92 0.45 0.00 6.91
Wage 82.15 31.63 35.31 2.53 0.06 10.95 0.10 0.00 1.56
Trade balance 32.07 1.99 4.45 1.26 0.13 22.42 0.65 0.00 1.17

Note: Calculated at the posterior mean. Note that when the forecasting horizon s becomes large, the
contribution of a shock to the s step ahead forecast error converges to that shock’s contribution to
the unconditional volatility. Thus, Panel E reports each shock’s contribution to long run volatility.

13
 4.1 O N THE ROLE OF FOREIGN SHOCKS
Are foreign shocks important for macroeconomic volatility in small open economies?
The model presented here answers “yes” when it is confronted with Canadian data. This
is in line with ample empirical evidence. For instance, Kose et al. (2003) estimate a
FAVAR model with separate world, region, and country specific factors. They report
that the world and region factors combined explain about 45-75% of the volatility in
Canadian GDP, consumption and investment. Similar results are obtained in VAR studies
of different countries and sample periods, and with alternative identifying assumptions
regarding shocks. Recent examples include Kose et al. (2008), Crucini et al. (2011),
Mumtaz et al. (2011), and Kose et al. (2012). Estimated SOE-DSGE models, in contrast,
have a hard time accounting for international business cycle transmission. Let us take
GDP as an example: Justiniano and Preston (2010), using a benchmark model, find that
foreign shocks explain about 1% of the fluctuations in Canadian GDP at all forecasting
horizons. Adolfson et al. (2007) estimate a medium scale model on Swedish data, and
report that foreign shocks explain between 9% (1 quarter) and 1% (20 quarters) of Swedish
GDP.11 Christiano et al. (2011) extend the Swedish model to include financial frictions and
unemployment, and find that 8% of GDP is explained by a set of five foreign disturbances
(including UIP shocks) within the 8 quarters horizon. The limited role for foreign shocks
seems to hold also in DSGE models for large economies (see Jacob and Peersman (2013)).
The second result, that foreign variance shares are increasing in the forecasting hori-
zon, is consistent with a number of empirical studies as well. Justiniano and Preston
(2010) estimate a SUR model and report that foreign shocks explain 22% of Canadian
GDP at the 1 quarter horizon, 44% at the 4 quarter horizon, and 76% in the long run.
The numbers in Table 4 closely resemble those findings. Also Cushman and Zha (1997)
and Aastveit et al. (2011) document higher foreign variance shares at longer horizons.
However, the DSGE model allows us to take one step further and ask, within a structural
framework, why foreign variance shares rise over time. The clue lies in estimated proper-
ties of TFP. Table E.3 in the appendix reports the FEVD of domestic shocks. In the very
short run (1 quarter), no single shock is the major driver of the selected set of macroeco-
nomic variables. Rather, innovations to different variables are caused by different distur-
bances. For example, GDP is driven both by shocks to service productivity, the interest
rate, and the marginal efficiency of investment (MEI). Consumption and investments are
explained well by preference and MEI shocks respectively, while the trade balance is
captured by risk premium and MEI shocks, as in Jacob and Peersman (2013). For the
unconditional volatility of macroeconomic variables (the stationary forecast error), it is
clear that productivity plays a major role. All foreign and domestic TFP shocks combined
are responsible for about 70-80% of aggregate volatility in GDP, consumption and wages,
and about half of the movements in inflation and interest rates.12 Arguably, the increasing
importance of foreign productivity over time can be traced to the estimated AR(1) process
for TFP. Productivity innovations are relatively long lasting events, explaining why they
account for substantial shares of the forecasting errors at longer horizons. The fundamen-
tal question is why data prefer productivity driven business cycles. Section 5 sheds light
on the issue by inspecting transmission channels and propagation mechanisms at play.
11
These numbers are found in a working paper version, see Adolfson, Laséen, Lindé, and Villani (2005).
12
Similar results are found for the US variables (not shown).

14
 Figure 1: GDP in data and in the model with only foreign shocks
A. AGGREGATE B. RAW MATERIALS

4
5
2

0 0

−2
−5
−4

−6
−10
1983 1986 1989 1992 1995 1998 2001 2004 2007 1983 1986 1989 1992 1995 1998 2001 2004 2007

C. MANUFACTURING D. SERVICES

5 2

0 0

−5
−2

−10
−4
1983 1986 1989 1992 1995 1998 2001 2004 2007 1983 1986 1989 1992 1995 1998 2001 2004 2007

Note: GDP in data (blue) andNBER RECESSIONS

in smoothed DATA ONLY FOREIGN SHOCKS
counterfactual series excluding domestic shocks (gray). Coun-
terfactuals are calculated based on the posterior mean. Shaded bars denote NBER defined US recessions.

As an illustration of the importance of international business cycle spillover, Figure 1

plots quarterly Canadian GDP in data and in the model when only foreign shocks are
included. Consider first aggregate GDP. A significant share of the movements is explained
by foreign shocks, and their importance rise over the sample as the initial discrepancy
caused by pre-sample conditions dies out. Further decomposition into sectoral variables
suggests a tendency of more variation being explained by foreign disturbances in the raw
material sector than in manufacturing, while manufacturing seems more prone to foreign
shocks than services.13

4.2 A NOTE ON EXCHANGE RATE PASS - THROUGH

Exchange rate pass-through is defined as the response of prices to a change in the nominal
exchange rate. High pass-through implies transmission of international business cycle
shocks via the exchange rate channel. However, empirical studies suggest only a weak
link between domestic prices and exchange rate fluctuations – a partial exchange rate
disconnect. This is also true in the model presented here. Yet, I will argue that sectoral
differences in exchange rate pass-through, coupled with the use of intermediate goods in
production, greatly facilitates international spillover.
This section documents the degree of pass-through implied by the estimated model.
To this end I calculate price responses to a UIP shock. Arguably this shock is exogenous
13
The figure indicates that the two largest recessions in the sample, those in 81-82 and 90-92, had little to do
with international events. Foreign shocks in the first case are probably hidden in pre-sample conditions.
In the latter case the recession was indeed far more severe in Canada, see Voss (2009) and Cross (2011).

15
 Table 5: Short run pass-through rates (scaled by 100)
Price measure πt πti πj,t πF j,t πHj,t ∗
πHj,t m
πj,t τj,t
Aggregate 12.70 26.48 – – – – – –
Raw materials – – 35.64 65.15 14.08 31.99 23.96 -33.17
Manufacturing – – 29.79 53.91 16.99 36.12 26.40 -17.79
Services – – 4.10 7.85 3.92 78.41 13.27 70.56
∗
Note: Pass-through rates scaled by 100, calculated based on the posterior mean. πHj,t in the
table is expressed in terms of domestic currency, in contrast to the text.

to model fundamentals, making it comparable with exchange rate fluctuations analyzed

in reduced form studies. Results are shown in Table 5, where price responses are ex-
pressed relative to the exchange rate innovation (i.e. ∆pricet /∆et ). On impact, a one
percent shock to the risk premium causes the nominal exchange rate to depreciate by about
3.41%, while inflation rises by 0.43%. Short run pass-through to CPI inflation follows as
∆πt
∆et
= 12.7%. The corresponding pass-through to aggregate investment prices is 26.5%.
Aggregate results mask significant sector heterogeneity: Pass-through to sectoral market
prices (third column) is 35.6% in raw materials and 29.8% in manufacturing, compared
with only 4.1% in services. In turn, these numbers are weighted averages of inflation in
imported and domestically produced goods’ prices. The exchange rate depreciation re-
duces the mark-up of importing firms. Their desire to stabilize the mark-up causes higher
imported inflation (fourth column). This is standard. What is less standard, is the signif-
icant pass-through to prices on domestically produced goods (fifth column). This comes
about from trade in materials, as some of the exchange rate depreciation passes through to
material prices (seventh column). The resulting drop in domestic firms’ mark-up leads to
higher domestic inflation. The exchange rate effect on πHj,t adds to overall pass-through
and can be high if domestic prices are changed frequently. More generally, the model
predicts an inverse relationship between price stickiness and pass-through because firms
with frequent price changes are more likely to respond to the exchange rate. This explains
the bulk of pass-through variation across sectors. High pass-through to flexible prices is
also a robust feature in data, as shown by Gopinath and Itskhoki (2010).
Without stretching the results too far, a brief comparison with empirical pass-through
regressions is warranted. Campa and Goldberg (2005) estimate the short run (quarterly)
exchange rate pass-through to aggregate import prices in Canada, and find an elasticity of
75%. Gopinath et al. (2010) report a lower number, about 35%. Table 5 provides results
somewhere in between, about 54-65% in the trade intensive sectors manufacturing and
raw materials. Finally, Goldberg and Campa (2010) estimate pass-through elasticities of
the CPI in 21 developed economies (Canada is not included) and find an average elasticity
of about 15%, close to the value of 13% in Table 5.

5 T RANSMISSION CHANNELS
In this section I describe the mechanisms leading to transmission of business cycle fluc-
tuations across countries, and analyze the role of different shocks. First I point out an
important feedback loop that comes about from the intersectoral linkages. It’s main impli-
cation is synchronization of firms’ mark-up across sectors and countries. In turn, mark-up

16
synchronization leads to co-movement of domestic and foreign prices and quantities, a
key feature of data. Shocks that generate mark-up synchronization are good candidates
for international business cycle co-movement, and favored by the likelihood based esti-
mation procedure. First, I describe these mechanisms in more detail. Second, I study
impulse responses to shed light on the dynamic effects of different shocks.

5.1 T HE ROLE OF INTERMEDIATE TRADE AND SECTOR

HETEROGENEITY
To better understand how intermediate trade and sector heterogeneity change spillover
from international business cycles to the SOE, I proceed in three steps. First, note that
laws of motion for domestic producer prices (prHj,t and p∗rHj,t ) in sector j can be written
as below (prices on goods sold domestically and exported, respectively):
prHj,t = θpj (prHj,t−1 − πt + ιp πHj,t−1 ) + (1 − θpj ) p̄rHj,t
(27)
∗
p∗rHj,t = θpj p∗rHj,t−1 + ∆et − πt + ιp πHj,t−1 + (1 − θpj ) p̄∗rHj,t

Both prices above are quoted in domestic currency and in terms of consumption goods.
The two equations state that prices on domestically produced goods are linear combina-
tions of the lagged price level (and some terms associated with indexation and exchange
rate changes) and the new prices set by firms who re-optimize in the current period, p̄rHj,t
and p̄∗rHj,t . If optimal new prices rise, we get inflationary pressure on the sector averages
prHj,t and p∗rHj,t as well. The second step is to note that the forward-looking nature of the
dynamics described above is captured by two optimality conditions for newly set prices
(mark-up shocks are abstracted from):
∞
X
p̄rHj,t = prHj,t + (1 − βθpj ) Et (βθpj )s (rmcj,s − prHj,s )
s=t
∞ (28)
X s
p̄∗rHj,t p∗rHj,t p∗rHj,s


= + (1 − βθpj ) Et (βθpj ) rmcj,s −
s=t

These two equations show that the profit maximizing price, from the individual firm’s
point of view, is a linear combination of i) the sector specific average and ii) current
and expected future deviations in the price mark-up over marginal costs. In the limit as
θpj goes to zero, the expectation sums disappear.14 However, when θpj > 0, then all
innovations that increase (decrease) real marginal costs relative to producer prices cause
temporary upward (downward) pressure on p̄rHj,t and p̄∗rHj,t . This takes us to the third
step, the introduction of intermediate trade and sector heterogeneity. The linearized real
marginal cost in sector j can be written as follows:
J
X
rmcj,t = −zAj,t + φj ζlj prl,t + ψj ωj,t + (1 − φj − ψj ) rtk
l=1
XJ
= −zAj,t + φj ζlj [αl prHl,t + (1 − αl ) prF l,t ] + ψj ωj,t + (1 − φj − ψj ) rtk
l=1
(29)
14
Optimal prices without price setting rigidities and mark-up shocks are prHj,t = p∗rHj,t = rmcj,t ∀ t.

17
The first line shows that costs are directly affected by market prices prl,t in all domestic
industries l ∈ J , because intermediate trade takes place across sectors. The second line
demonstrates that costs depend on import prices prF l,t , set by firms in foreign sectors. This
is true as long as the domestic absorption parameters αj are less than one. Importantly,
prF l,t can be represented by a system similar to (27)-(29). Thus, shocks that affect sectoral
marginal costs in the foreign economy will in principle show up in equation (29). Three
important observations immediately follow from the system (27)-(29). First, intermediate
trade introduces co-movement between domestic and foreign producer prices. That is, a
rise (fall) in any import price prF l,t directly reduces (increases) the mark-up of domestic
firms (equation (29)), resulting in rising (declining) domestic producer prices prHj,t and
p∗rHj,t . Second, the model features an important feedback loop. That is, the first round rise
(fall) in prHj,t further increases (reduces) domestic sector j’s costs, because prHj,t shows
up in (29). There is a similar feedback loop involving foreign producer prices and foreign
marginal costs. Third, the initial impulse propagates across sectors as long as ζlj > 0 for
some l 6= j. Thus, foreign shocks can hit some industries in the SOE, notably those with
high trade intensity, and then propagate to others via intermediate trade. The latter kind
of spillover is governed by the off-diagonals of the I-O matrix, and allows even relatively
non-traded sectors to be affected by international disturbances. The setup presented here
nests as special cases some common approaches in the literature, including models with i)
one sector (J = 1),15 ii) no intermediate trade (φj = 0), and iii) no foreign trade (αj = 1).
However, all these dimensions matter for the transmission of foreign shocks. Obviously,
if αj = 1 ∀ l ∈ J , then economic activity in the SOE is completely unrelated to the
rest of the world. If φj = 0, then there is one less source of co-movement in producer
prices (the one described above), and hence one less mechanism that induces business
cycle spillover. If J = 1, then the entire transmission has to take place at the aggregate
level without sectoral reallocations. In contrast, the multi-sector model presented here
allows industries with limited trade to be affected via cross-sectoral intermediate market
linkages. Thus, even fluctuations in completely non-traded industries can in principle be
driven by business cycle shocks abroad.

5.2 T HE SCOPE FOR INTERNATIONAL SYNCHRONIZATION

To better understand the role of foreign disturbances for domestic business cycles, I an-
alyze the impulse responses of domestic variables to selected international shocks. I ex-
plain why foreign productivity shocks generate business cycle co-movement, how foreign
investment shocks create a wedge between domestic and foreign investment, and argue
that the multi-sector setup facilitates higher pass-through than what is typically found in
models with LCP.

5.2.1 DYNAMIC EFFECTS OF PRODUCTIVITY SHOCKS

A productivity shock in the foreign manufacturing sector is plotted in Figure 2. Productiv-
ity innovations in the US manufacturing sector are essential for understanding macroeco-
nomic fluctuations in the SOE, according to Table 4. One thing to note from Figure 2 and
15
Or alternatively, no (ex ante) sector heterogeneity (φj = φ, γlj = γ, αj = α).

18
 Figure 2: A productivity shock in foreign manufacturing
GDP GDP RAW MATERIALS GDP MANUFACTURING GDP SERVICES
1 1 0.7
1
0.8 0.6
0.5
0.6 0.5 0.5
0.4 0 0.4
0.2 0.3
−0.5 0
5 10 15 20 5 10 15 20 5 10 15 20 5 10 15 20

CONSUMPTION HOURS TRADE BALANCE INFLATION

0.4
0.5 0.1
0
0.2 0
0.4
0.3 −0.1 −0.1
0
−0.2
0.2
−0.2 −0.3 −0.2
0.1
5 10 15 20 5 10 15 20 5 10 15 20 5 10 15 20

INVESTMENT REAL WAGE REAL EXCHANGE RATE INTEREST RATE

0.05
−0.1
2.5 0.4
0
2 0.35 −0.2

1.5 0.3 −0.05

−0.3
1 0.25
−0.1
−0.4
0.5 0.2
5 10 15 20 5 10 15 20 5 10 15 20 5 10 15 20

Note: Foreign (red) and domestic (blue) Bayesian impulse responses to a productivity shock in for-
eign manufacturing (one standard deviation). Posterior mean (solid) and 90% highest posterior proba-
bility intervals (dotted). The trade balance is in absolute deviations from steady state relative to GDP.

figures in the appendix is the striking co-movement in aggregate responses across coun-
tries in both GDP, consumption and investment. What is going on here? Consider first the
foreign IRFs. As expected, higher foreign productivity raises foreign GDP, consumption,
and investment. The set of frictions in the model, in particular sticky prices and mo-
nopolistic competition, also implies less working hours and a lower foreign interest rate
(see ??). All these effects are well known from the textbook one-sector, closed economy
model. However, cheaper manufactured goods in the foreign economy not only lead to
expenditure switching towards that sector, but also to cheaper manufactured intermedi-
ates. This latter effect reduces costs and prices in the other foreign industries as well, and
therefore creates the feedback loop emphasized by e.g. Acemoglu, Carvalho, Ozdaglar,
and Tahbaz-Salehi (2012). Regarding spillover to the SOE, note first that lower prices
on manufacturing imports induce expenditure switching in that sector towards imports.
While this kind of expenditure switching helps in generating co-movement between do-
mestic and foreign absorption of manufactured (sector j) goods, it is contractionary from
the point of view of domestic sector j firms. In a one-sector world, the substitution to-
wards imports is basically be the main spillover effect. This is why previous models find
little co-movement in GDP, hours, and other supply side variables across countries.
In contrast, the multi-sector structure presented here provides us with a rich story
about additional mechanisms at work. First, lower imported inflation in the domestic
manufacturing sector implies lower overall inflation in manufactured prices, relative to
prices from other industries. This creates domestic substitution towards all manufactured
goods, including those that are produced domestically. The sectoral substitution effect
is expansionary from the point of view of domestic manufacturing firms. Second, the

19
 Figure 3: Foreign demand shocks
GDP TRADE BALANCE GDP REAL EXCHANGE RATE
1 0.08
0.6 0.2
0.8 0.06
0.4
0.6 0.1
0.04
0.2
0.4 0.02
0
0.2 0
0
−0.2 −0.1
5 10 15 20 5 10 15 20 5 10 15 20 5 10 15 20

CONSUMPTION MPV OF CAPITAL CONSUMPTION INFLATION

0.04
0.4
−0.1 0.4
0.2 0.02
−0.2 0.2
0
0
−0.2 −0.3 0
5 10 15 20 5 10 15 20 5 10 15 20 5 10 15 20

INVESTMENT INVESTMENT PRICE INVESTMENT INTEREST RATE

0.1
0.03
2 0
0 0.02
1
−0.2
0.01
0 −0.1
−0.4 0
−1
5 10 15 20 5 10 15 20 5 10 15 20 5 10 15 20

Note: Foreign (red) and domestic (blue) Bayesian impulse responses to i) a foreign MEI shock (columns
1-2) and ii) a foreign preference shock (columns 3-4). Posterior mean (solid) and 90% highest posterior
probability intervals (dotted). The trade balance is in absolute deviations from steady state relative to GDP.

cheaper manufactured goods also reduce domestic firms’ expenditures on intermediate

goods. This is seen from equation (29). In fact, producer costs decline in all domestic in-
dustries because also non-manufacturing producers use manufactured goods extensively
as input. Profit maximizing behavior then induces domestic firms across the economy
to reduce their prices, and overall domestic inflation declines even further. That triggers
another round of cheaper intermediate goods, and another round of price reductions, and
so on. The interest rate naturally declines as well. Domestic prices are not perfectly ad-
justed, so some of the increased productivity is materialized as lower demand for labor –
a well known outcome in New Keynesian models. Thus, total hours decline in the domes-
tic economy the first periods. Cheaper domestic goods also limit the initial expenditure
switching towards imports, implying further domestic expansion in demand for consump-
tion and investment goods. For domestic producers, substitution towards imports and cost
reduction are two forces that push in opposite directions. Their relative importance in
each sector depends on the sectoral trade intensity, the degree of price stickiness, and the
share of intermediate inputs in production. In total, the increase in foreign manufacturing
productivity generates relatively large dynamics in the SOE. Part of the reason is the high
trade intensity in the manufacturing sector. Another point is that manufactured goods
prices are much less sticky than service prices, implying that they react more following
the shock. Also, manufactured goods are important inputs in services, the largest sector
in the economy.

20
 5.2.2 DYNAMIC EFFECTS OF INVESTMENT SHOCKS
Next I describe effects of a shock to the marginal efficiency of investment (MEI) in the
foreign economy. The goal is to understand why estimated DSGE models for closed
economies have attributed larger macroeconomic fluctuations to this shock than what I
find here. Figure 3 plots the IRFs. The MEI shock temporarily increases the amount of
capital transformed from each investment unit, and thereby raises the relative return to
capital investments. This induces foreign households to invest more, and cut back on con-
sumption the first periods due to resource constraints (see Furlanetto, Natvik, and Seneca
(2013) for an analysis of this issue). The net effect is still a positive shift in aggregate
demand, leading to upward inflation pressure. After some periods the investments start
to pay off in form of capital abundance in the foreign economy, leading to a prolonged
period with higher consumption demand as well.
In the SOE, the foreign MEI shock generates responses in GDP, consumption, hours,
interest rate and inflation that are qualitatively similar to those in the foreign economy.
That is, due to higher imported inflation, overall price level and the interest rate in the
SOE increase. This reduces domestic consumption demand and makes production more
expensive. Yet, high foreign investment demand is expansionary for domestic raw mate-
rial and manufactured goods producers, who export investment goods intensively.
Still, the MEI shock cannot explain all international synchronization patterns – it
causes strong divergence between investment in the two countries. To see why, note
that domestic absorption of investment goods from sector j can be written as follows:
∞
1 X s−t
pir,t qs − pir,s



ij,t = −ν prj,t − + it−1 + Et β (30)
I s=t
Equation (30) shows that ij,t is linked to the relative sector price prj,t , and
 via aggregate
∞
investment demand, to the expected path for real returns to investments, qs − pir,s s=t .16
Intuitively, when the value of current and future capital exceeds the cost of capital accu-
mulation, sectoral and aggregate investment demand is high. Investment co-movement
across countries is therefore stimulated by synchronization of expected capital returns.
However, in the case of a foreign MEI shock, higher import prices in the SOE raises pir,t ,
while higher real interest rates reduce the present value of installed capital.17 This low-
ers domestic sectoral and aggregate investment demand, and generates a wedge between
domestic and foreign investment activity that is not typically seen in the data. At the
same time, some of the increase in foreign demand is targeted towards domestic goods,
in particular those who can be transformed into capital abroad. The result is a large im-
provement in trade balances, and higher aggregate GDP. Also, domestic demand towards
domestic goods goes up, due to relatively cheaper home products. Taken together, the
foreign MEI shock is able to generate co-movement between several domestic and for-
eign variables, but not between domestic and foreign investment. This latter point implies
that the posterior weight on foreign MEI shocks is smaller, although they still explain
important parts of several domestic variables.
16
Investment adjustment costs are priced into pir,t . Without adjustment costs, equation (30) collapses to
qs + zI,s = pir,s ∀ s.
17
The link between qt and real interest
P∞rates is found s−t by solving the linearized optimality condition
 for
k
capital forward to obtain qt = Et s=t (β (1 − δ)) − (rs − πs+1 ) + (1 − β(1 − δ)) rs+1 . Thus,
an increase in current or future expected real interest rates reduce the value of capital.

21
 Figure 4: A risk premium shock
GDP GDP RAW MATERIALS GDP MANUFACTURING GDP SERVICES
0.4
0.6 0
0 0.2
0.4 0
−0.2 −0.2
−0.2
0.2
−0.4 −0.4 −0.4
0 −0.6
−0.6 −0.2 −0.8 −0.6
5 10 15 20 5 10 15 20 5 10 15 20 5 10 15 20

CONSUMPTION HOURS TRADE BALANCE INFLATION

0 0.8
0.2 0.2
0.6
−0.1 0.15
0.4
0 0.1
−0.2 0.2
0.05
0
−0.3 −0.2
0
−0.2
5 10 15 20 5 10 15 20 5 10 15 20 5 10 15 20

INVESTMENT REAL WAGE REAL EXCHANGE RATE INTEREST RATE

1 1.5
0 0.15
0 1
−1 −0.1 0.1
0.5
−2 0.05
−0.2 0
−3
0
5 10 15 20 5 10 15 20 5 10 15 20 5 10 15 20

Note: Domestic Bayesian impulse responses to a UIP shock (one standard deviation). See Figure 2 for de-
tails.

5.2.3 R ISK PREMIUM SHOCKS AND EXCHANGE RATE PASS - THROUGH

The empirical section established that exchange rate pass-through to import prices and
the CPI is moderate, with large sectoral differences depending on the degree of price
flexibility in each industry. Next I investigate these results in more detail. Insights can
be found if we study the responses of a shock to the risk premium on foreign bonds.
As argued earlier, this shock should be exogenous to model fundamentals, making it
comparable to exchange rate shocks analyzed in empirical studies. Impulse responses to
the risk premium shock are provided in Figure 4. On impact, a shock which raises the risk
premium causes the nominal exchange rate to depreciate, while inflation rises moderately
at the aggregate level, with sectoral price responses depending on the frequency of price
adjustment.
The intuition is as follows: Higher risk premium raises demand for foreign bonds, and
thereby reduces the value of domestic currency in nominal terms. This is seen from the
linearized no arbitrage condition, which writes as
rt = rt∗ + Et (∆et+1 ) − B nf at + zB,t .
Clearly, following a rise in zB,t there must be a contemporaneous depreciation unless
the monetary authority pegs the exchange rate.18 Due to LCP, weaker currency in the
SOE then directly reduces foreign firms’ export income and profits. This is seen from the
marginal income of foreign exporters under LCP, which is pF j,t − et at the sector level.
These firms react by raising their export prices. The result is higher imported, and thus
higher overall domestic inflation. This is what we define as exchange rate pass-through.
18
The SOE assumption implies that foreign variables, including rt∗ , do not change.

22
 Two forces in the model limit the pass-through to domestic prices. First, the presence
of price stickiness reduces pass-through to import prices, as only a subset of the foreign
∗
exporters can adjust their prices optimally. Indeed, when θpj → 1 the pass-through be-
comes zero. In one-sector New Keynesian models with LCP, the presence of price sticki-
ness typically leads to less pass-through than suggested by empirical literature (Gopinath
et al., 2010). Second, as households now find it more profitable to save abroad, they lower
consumption and investment demand to reallocate resources towards foreign bonds. The
decline in domestic absorption is seen in Figure 4. Lower consumption demand is partic-
ularly relevant for service firms, who supply most domestic consumption goods. The drop
in investment demand on the other hand affects GDP in the manufacturing sector, which
produce most investment goods. In fact, the aggregate decline in domestic absorption is
large enough to lower GDP in these two sectors.19 Most importantly, it puts downward
pressure on both domestic producer prices and import prices, and thus limits exchange
rate pass-through. Taken together, the combination of LCP and contraction in domestic
absorption should lead to small or even negative pass-through rates in the SOE.
Sector heterogeneity modifies the pass-through story outlined above. As seen in Ta-
ble 5, we have relatively high pass-through in the sectors with frequent price changes.
In the model, this relationship comes about from the simple observation that firms who
re-optimize prices frequently, have higher probability of responding optimally to the ex-
change rate depreciation. As the optimal sector price equates pF j,t − et with marginal
costs, pF j,t will rise more aggressively when price stickiness is low. Moreover, CPI mea-
sures in raw materials and manufacturing put high weights on the import price pF j,t ,
adding to the positive pass-through in these industries. The presence of intermediate trade
further increases pass-through rates: Higher imported inflation drives up producer costs
among those firms who import intermediate goods, and thus puts upward pressure on do-
mestic producer prices. Again, this cost channel is particularly important for the trade
intensive raw material and manufacturing firms. This explains why pass-through to both
imported and domestic producer price inflation is relative high in these sectors. Moreover,
cross-sectoral linkages in domestic intermediate markets allow exchange rate fluctuations
to spill over to marginal costs of domestic service firms as well. Thus, both the existence
of firms with relatively flexible prices, and the presence of intermediate trade channels in
the model, help to increase the pass-through to domestic prices. This is consistent with
the empirical work by Goldberg and Campa (2010), who find that intersectoral linkages
are the most important source of exchange rate pass-through to the domestic CPI.

6 T HE COUNTERFACTUAL MODEL ECONOMY

This paper analyzes the role of foreign shocks in a model with intermediate trade be-
tween heterogeneous firms. In this section I ask whether the role of international business
cycles survives in a context without these extensions to the conventional open economy
DSGE model. To this end, I estimate the particular version of the model when J = 1
and φ = 0. The model now becomes a fairly standard DSGE model for a small open
19
Yet, the drop in total GDP is muted, because the accumulation of foreign assets is financed by a trade
surplus driven by exports of raw materials and manufactured goods.

23
 Table 6: Counterfactual model – Business cycle predictions
Panel A: Panel B: Panel C:
100ρy,y∗ 1Q 4Q 8Q 20Q LR ∆π/∆e
GDP 27.4 1.89 3.11 3.52 6.22 10.33 πt 7.86
Consumption 34.8 0.94 1.22 1.93 5.47 16.43 πF,t 36.77
Investment 3.2 1.37 1.37 1.28 2.69 3.42 πH,t -2.27
Hours 9.1 1.72 2.15 2.18 2.26 2.36 ∗
πH,t 57.55
Interest 17.6 3.73 3.08 3.04 3.00 4.02 τt 20.78
Inflation 14.9 2.76 2.60 2.64 2.70 3.03
Wage 39.3 4.50 4.30 5.31 10.34 19.17
Trade balance – 8.14 10.03 10.17 10.21 10.16

Note: See Table 4 and Table 5 for details.

economy.20 Calibrated values are set as follows: First, I rescale labor and capital shares in
both economies to keep the constant returns to scale assumption based on the numbers in
Table 2. This gives ψ = 0.543. Second, I calibrate trade shares in GDP by subtracting the
intermediate input share of imports in each sector, and then calculating aggregate (sector
GDP weighted) import share in the economy. The resulting import share in GDP is 0.26
(α = 0.7405). The remaining calibrated values are chosen as before. Also the prior dis-
tributions are as in the baseline model, except that price and wage stickiness have prior
modes equal to 0.7, while the aggregate TFP shock has a mode equal to 0.2.
Business cycle predictions from the counterfactual model (based on posterior mean
estimates) are provided in Table 6. Parameter estimates are reported in the appendix.
Consider first model implied cross-country correlations (Panel A). For all variables under
consideration, they fall to less than half of those in the baseline model. The decline is
particularly large for investment. Still, the degree of co-movement is higher than that
found by Justiniano and Preston (2010), and international consumption synchronization
actually comes fairly close to that in data. Part of difference from Justiniano and Preston
(2010) is attributed to the inclusion of investment, which is abstracted from in their study.
When higher foreign productivity takes down international prices, domestic investment
(and capital) is stimulated by cheaper imports.
Turning to the decomposition of shocks (Panel B), we see that foreign shocks become
nearly irrelevant for most domestic variables within the business cycle. They explain less
than 7% of the variation in all variables except wages and the trade balance within the
5-year horizon. This is bout one tenth of the shares attributed to foreign shocks in the
baseline model (Table 4). In the long run, foreign shocks account for about 2-20% of the
macroeconomic volatility in the SOE, far below typical estimates in the VAR literature.
Implied pass-through rates are provided in Panel C. The short run pass-through to CPI
drops from 12.4% to 7.9% – still a fairly high number given that exporters in the model
price their goods in local currency. The main reason is the estimated low degree of price
stickiness, with a posterior centered around 0.5. In order to illustrate the limited role
of foreign shocks in the counterfactual model, Figure 5 plots the impulse responses to a
foreign MEI shock. As before we get a drop in domestic consumption and investment.
20
Obviously, one counterfactual implication of this model version is the symmetric response of all firms
within countries to all kinds of business cycle shocks. Bouakez et al. (2009) analyze implications of
imposing such symmetry in a closed economy setting.

24
 Figure 5: Impulse responses to a foreign MEI shock in the counterfactual model
GDP CONSUMPTION INVESTMENT
0.4
0.1 1.5
0.05
0.3
0 1
%−deviation

0.2 −0.05
−0.1 0.5

0.1 −0.15
0
−0.2
0
−0.25
−0.5
−0.3
5 10 15 20 5 10 15 20 5 10 15 20

DOMESTIC ECONOMY REST OF THE WORLD

Note: See Figure 2 for details.

However, unlike before we also get a drop in domestic GDP. The intuition is straight
forward: Without intermediate trade, the increase in foreign investment demand does not
call for more exports of materials. Instead, the main transmission channel to domestic
GDP is via a higher domestic real interest rate, which lowers domestic consumption and
investment demand. Thus, without intermediate trade, the foreign MEI shock cannot
even explain international co-movement in GDP. Finally, note that the negative effect on
foreign consumption is amplified in the one-sector model, a reasonable result given that
consumption and investment now are close substitutes.

7 C ONCLUSIONS
I ask how and to what extent international business cycle disturbances cause macroeco-
nomic fluctuations in small open economies. To shed some light on these questions, I
construct and estimate a medium scale small open economy model with several shocks
and frictions typically used in the DSGE literature. The model is embedded with i) trade
in intermediate goods between firms, and ii) sectoral producer heterogeneity. These exten-
sions to the workhorse one-sector open economy model are sufficient to reconcile DSGE
theory with data along international dimensions. When the model is fitted to Canadian
and US data, a set of important empirical results emerge: First foreign shocks explain
a major share of macroeconomic fluctuations in the SOE. Second, posterior estimates
emphasize the role of productivity, in the sense that technology shocks, not investment
efficiency fluctuations, are the major drivers of business cycles. Third, foreign shocks
become increasingly important over longer forecasting horizons. Fourth, the model gen-
erates substantial business cycle synchronization, even though trade balances are counter-
cyclical and shocks are uncorrelated. Fifth, exchange rate pass-through is moderate, with
sectoral pass-through depending on the frequency of price changes. While these results
are consistent with reduced form literature such as VAR and FAVAR studies, they are
typically not found in the literature using open economy DSGE models.
The model presented here allows us to gain insight about the mechanisms that cause
these results. An important implication of intermediate trade is that it synchronizes pro-
ducer prices and costs in the cross-section of firms, both within and across borders. This
helps in generating co-movement in an environment with producer heterogeneity and oth-

25
erwise segmented markets. Foreign shocks in particular can enter the SOE through some
industries exposed to international trade, and then propagate to others via domestic factor
markets. Synchronized producer prices across sectors and countries generate substantial
international co-movement in i) current and future real interest rates, which determines
consumption, and ii) the expected path of capital returns, a key statistic for investment
decisions. However, synchronization of real interest rates comes at the cost of too high
consumption co-movement across countries. I find that foreign technology shocks are
particularly well suited for international business cycle synchronization. These are also
relatively persistent, an important reason why foreign shocks become essential at longer
forecasting horizons. Foreign investment efficiency shocks on the other hand cause in-
ternational divergence in the present value of capital and investment. Investment is posi-
tively correlated across countries, implying that the likelihood based estimation procedure
attributes a smaller role to investment efficiency shocks.
One obvious limitation with the present model is the lack of meaningful interactions
between financial markets and the macroeconomy. Indeed, the recent financial crisis has
demonstrated the potential importance of financial frictions for international business cy-
cles. By now, there is a large (an growing) literature on financial frictions in closed
economies. Yet, for many, if not most small open economies, the recent financial crisis
was associated with foreign events. Therefore, a topic for future research is the prop-
agation of financial distress across countries, e.g. an open economy extension of the
market frictions studied by Christiano, Motto, and Rostagno (2014). However, for such
an analysis to make sense, one should be equipped with a model that can account for
macroeconomic spillover as well. This paper offers a preliminary, but instructive step
towards that end.

26
 R EFERENCES
Aastveit, K. A., H. C. Bjornland, and L. A. Thorsrud (2011). The world is not enough! Small
open economies and regional dependence. Working Paper 2011/16, Norges Bank.
Acemoglu, D., V. M. Carvalho, A. Ozdaglar, and A. Tahbaz-Salehi (2012). The network origins
of aggregate fluctuations. Econometrica 80(5), 1977–2016.
Adolfson, M., S. Laséen, J. Lindé, and M. Villani (2005). Bayesian estimation of an open economy
DSGE model with incomplete pass-through. Working Paper Series 179, Sveriges Riksbank
(Central Bank of Sweden).
Adolfson, M., S. Laséen, J. Lindé, and M. Villani (2007). Bayesian estimation of an open economy
DSGE model with incomplete pass-through. Journal of International Economics 72(2), 481–
511.
Adolfson, M., S. Laséen, J. Lindé, and M. Villani (2008). Evaluating an estimated New Keynesian
small open economy model. Journal of Economic Dynamics and Control 32(8), 2690–2721.
An, S. and F. Schorfheide (2007). Bayesian analysis of DSGE models. Econometric reviews 26(2-
4), 113–172.
Atalay, E. (2013). How important are sectoral shocks? Manuscript.
Bauwens, L., M. Lubrano, and J. F. Richard (1999). Bayesian inference in dynamic econometric
models. Oxford University Press.
Bergholt, D. and T. Sveen (2014). Sectoral Interdependence and Business Cycle Synchronization
in Small Open Economies. Norges Bank Working Paper Series 2014(04).
Bils, M. and P. J. Klenow (2004). Some evidence on the importance of sticky prices. Journal of
Political Economy 112(5), 947–985.
Bouakez, H., E. Cardia, and F. J. Ruge-Murcia (2009). The transmission of monetary policy in a
multisector economy. International Economic Review 50(4), 1243–1266.
Calvo, G. A. (1983). Staggered prices in a utility-maximizing framework. Journal of Monetary
Economics 12(3), 383–398.
Campa, J. M. and L. S. Goldberg (2005). Exchange rate pass-through into import prices. The
Review of Economics and Statistics 87(4), 679–690.
Canova, F. (2007). Methods for applied macroeconomic research, Volume 13. Princeton Univer-
sity Press.
Carvalho, V. and X. Gabaix (2013). The great diversification and its undoing. American Economic
Review 103(5), 1697–1727.
Christiano, L. J., M. Eichenbaum, and C. L. Evans (2005). Nominal rigidities and the dynamic
effects of a shock to monetary policy. Journal of political Economy 113(1), 1–45.
Christiano, L. J., R. Motto, and M. Rostagno (2014). Risk shocks. American Economic Re-
view 104(1), 27–65.
Christiano, L. J., M. Trabandt, and K. Walentin (2011). Introducing financial frictions and un-
employment into a small open economy model. Journal of Economic Dynamics and Con-
trol 35(12), 1999–2041.
Corsetti, G., L. Dedola, and S. Leduc (2008). International risk sharing and the transmission of
productivity shocks. The Review of Economic Studies 75(2), 443–473.
Cross, P. (2011). How did the 2008-2010 recession and recovery compare with previous cycles?
Technical Report 1.
Crucini, M., A. Kose, and C. Otrok (2011). What are the driving forces of international business
cycles? Review of Economic Dynamics 14(1), 156–175.
Cushman, D. O. and T. Zha (1997). Identifying monetary policy in a small open economy under
flexible exchange rates. Journal of Monetary Economics 39(3), 433–448.

27
de Resende, C., A. Dib, and M. Kichian (2010). Alternative optimized monetary policy rules in
multi-sector small open economies: The role of real rigidities. Working Papers 10-9, Bank of
Canada.
Dib, A. (2011). Monetary policy in estimated models of small open and closed economies. Open
Economies Review 22(5), 769–796.
Erceg, C. J., D. W. Henderson, and A. T. Levin (2000). Optimal monetary policy with staggered
wage and price contracts. Journal of Monetary Economics 46(2), 281–313.
Eyquem, A. and G. Kamber (2013). A note on the business cycle implications of trade in interme-
diate goods. Macroeconomic Dynamics 1, 1–15.
Furlanetto, F., G. J. Natvik, and M. Seneca (2013). Investment shocks and macroeconomic co-
movement. Journal of Macroeconomics, Elsevier 37(C), 208–216.
Galı́, J. and T. Monacelli (2005). Monetary policy and exchange rate volatility in a small open
economy. Review of Economic Studies 72(3), 707–734.
Geweke, J. (1999). Using simulation methods for Bayesian econometric models: inference, de-
velopment,and communication. Econometric Reviews 18(1), 1–73.
Goldberg, L. S. and J. M. Campa (2010). The sensitivity of the CPI to exchange rates: Distribution
margins, imported inputs, and trade exposure. The Review of Economics and Statistics 92(2),
392–407.
Gopinath, G. and O. Itskhoki (2010). Frequency of price adjustment and pass-through. The
Quarterly Journal of Economics 125(2), 675–727.
Gopinath, G., O. Itskhoki, and R. Rigobon (2010). Currency choice and exchange rate pass-
through. American Economic Review 100(1), 304–36.
Groth, C. and H. Khan (2010). Investment adjustment costs: An empirical assessment. Journal of
Money, Credit and Banking 42(8), 1469–1494.
Gust, C., S. Leduc, and N. Sheets (2009). The adjustment of global external balances: Does partial
exchange-rate pass-through to trade prices matter? Journal of International Economics 79(2),
173–185.
Heathcote, J. and F. Perri (2002). Financial autarky and international business cycles. Journal of
Monetary Economics 49(3), 601–627.
Jacob, P. and G. Peersman (2013). Dissecting the dynamics of the US trade balance in an estimated
equilibrium model. Journal of International Economics 90(2), 302–315.
Justiniano, A. and B. Preston (2010). Can structural small open-economy models account for the
influence of foreign disturbances? Journal of International Economics 81(1), 61 – 74.
Justiniano, A., G. Primiceri, and A. Tambalotti (2011). Investment shocks and the relative price
of investment. Review of Economic Dynamics 14(1), 101–121.
Justiniano, A., G. E. Primiceri, and A. Tambalotti (2010, March). Investment shocks and business
cycles. Journal of Monetary Economics 57(2), 132–145.
Justiniano, A., G. E. Primiceri, and A. Tambalotti (2013, April). Is there a trade-off between
inflation and output stabilization? American Economic Journal: Macroeconomics 5(2), 1–31.
Klein, P. (2000). Using the generalized Schur form to solve a multivariate linear rational expecta-
tions model. Journal of Economic Dynamics and Control 24(10), 1405–1423.
Koop, G. (2003). Bayesian econometrics. John Wiley & Sons.
Kose, M. A., C. Otrok, and E. Prasad (2012). Global business cycles: Convergence or decoupling?
International Economic Review 53(2), 511–538.
Kose, M. A., C. Otrok, and C. H. Whiteman (2003). International business cycles: World, region,
and country-specific factors. American Economic Review 93(4), 1216–1239.
Kose, M. A., C. Otrok, and C. H. Whiteman (2008). Understanding the evolution of world business
cycles. Journal of International Economics 75(1), 110–130.

28
Lubik, T. A. and F. Schorfheide (2007). Do central banks respond to exchange rate movements?
A structural investigation. Journal of Monetary Economics 54(4), 1069–1087.
Mumtaz, H., S. Simonelli, and P. Surico (2011). International comovements, business cycle and
inflation: a historical perspective. Review of Economic Dynamics 14(1), 176–198.
Nakamura, E. and J. Steinsson (2008). Five facts about prices: A re-evaluation of menu cost
models. The Quarterly Journal of Economics 123(4), 1415–1464.
Rabanal, P. and V. Tuesta (2010). Euro-dollar real exchange rate dynamics in an estimated two-
country model: An assessment. Journal of Economic Dynamics and Control 34(4), 780–797.
Schmitt-Grohé, S. (1998). The international transmission of economic fluctuations:: Effects of
U.S. business cycles on the Canadian economy. Journal of International Economics 44(2),
257–287.
Schmitt-Grohé, S. and M. Uribe (2003). Closing small open economy models. Journal of Inter-
national Economics 61(1), 163 – 185.
Smets, F. and R. Wouters (2003). An estimated dynamic stochastic general equilibrium model of
the Euro area. Journal of the European Economic Association 1(5), 1123–1175.
Smets, F. and R. Wouters (2007). Shocks and frictions in US business cycles: A Bayesian DSGE
approach. American Economic Review 97(3), 586–606.
Sudo, N. (2012). Sectoral comovement, monetary policy shocks, and inputoutput structure. Jour-
nal of Money, Credit and Banking 44(6), 1225–1244.
Voss, G. (2009). Aspects of Canadian and US business cycles. Bank of Canada Working Paper
Series.
Yun, T. (1996). Nominal price rigidity, money supply endogeneity, and business cycles. Journal
of Monetary Economics 37(2-3), 345–370.

29