Euro area, oil and global shocks: An empirical model-based analysis

Journal of Macroeconomics 46 (2015) 295–314

Contents lists available at ScienceDirect

Journal of Macroeconomics journal homepage: www.elsevier.com/locate/jmacro

Euro area, oil and global shocks: An empirical model-based analysis L. Forni a,1, A. Gerali b,2, A. Notarpietro b,3, M. Pisani b,∗ a International Monetary Fund, 700 19th Street, NW Washington, DC, 20431, US and Department of Economics and Management, University of Padua, Via del Santo 33, 35123 Padova, Italy b Banca d’Italia, Via Nazionale 91, 00184 Rome, Italy

a r t i c l e

i n f o

Article history: Received 19 June 2013 Accepted 25 September 2015 Available online 20 October 2015 JEL Classification: C11 C51 E32 F41 Keywords: Oil shocks DSGE modeling Open-economy macroeconomics Bayesian inference Euro area

a b s t r a c t We assess the impact of oil shocks on euro-area (EA) macroeconomic variables by estimating with Bayesian methods a two-country New Keynesian model of EA and rest of the world (RW). Oil price is determined according to supply and demand conditions in the world oil market. We obtain the following results. First, a 10% increase in the international price of oil generates an increase of about 0.1 annualized percentage points in EA consumer price inflation. Second, the same increase in the oil price generates a decrease in EA gross domestic product (GDP) of around 0.1% and a trade deficit, if it is due to negative oil supply or positive oil-specific demand shocks. Third, it generates a mild EA GDP increase and a trade surplus if due to a positive RW aggregate demand shock. Fourth, the increase in the oil price over the 2004–2008 period did not induce stagflationary effects on the EA economy because it was associated with positive RW aggregate demand shocks. The drop in RW aggregate demand contributes to explain the 2008 fall in oil prices, EA GDP and inflation. © 2015 Elsevier Inc. All rights reserved.

1. Introduction The following facts characterize the role of oil in the euro-area (EA) economy.4 First, oil dependence, defined as oil imports as a percentage of total gross oil consumption, has been close to 100% since the 1960s. Second, oil products are the most important component of final energy consumption, representing 44% of the total. Third, the weight of oil inputs in production (around 5%) and the limited short-term substitutability of oil inputs imply that the rather volatile oil price widely affects production costs and, therefore, prices and output. Fourth, oil price pass-through into fuel price is complete and quick. Fifth, taxes and margins constitute a large share, around 60%, of fuel prices. In this paper we evaluate the macroeconomic effects of oil shocks on the EA by developing and estimating with Bayesian methods a two-country New Keynesian dynamic stochastic general equilibrium (DSGE) model of the EA and the rest of the world ∗

Corresponding author. Tel.: +39 06 4792 3452. E-mail addresses: [email protected], [email protected] (L. Forni), [email protected] (A. Gerali), [email protected] (A. Notarpietro), [email protected] (M. Pisani). 1 Tel.: +1 202 623 76 27. 2 Tel.: +39 06 4792 3620. 3 Tel.: +39 06 4792 2623. 4 See European Central Bank (2010). http://dx.doi.org/10.1016/j.jmacro.2015.09.010 0164-0704/© 2015 Elsevier Inc. All rights reserved.

296

L. Forni et al. / Journal of Macroeconomics 46 (2015) 295–314

(RW) that allows for an explicit role of oil. The model is developed in line with the above stylized facts. Its main features are a world market for crude oil and, as in Finn (2000), the need of fuel to use physical capital in the production of manufacturing goods. The price of crude oil is set in the world oil market, in RW currency, and is the same in both EA and RW, once corrected for nominal exchange rate fluctuations. Oil supply is exogenously set by the RW. For oil demand, in both EA and RW there are firms that, under perfect competition, produce fuel (energy) from crude oil (according to a linear production function). Energy is then sold to firms in the domestic manufacturing sector, that need it as input for production. World demand of oil depends not only on world production of manufacturing goods, but also on an oil-specific demand component. The latter is inversely related to the efficiency in the use of energy, i.e. it is proportional to the amount of energy – and hence oil – needed for running capital at a given capacity. Consistent with empirical evidence, EA imports crude oil. Moreover, the EA fuel price is the sum of the crude oil price, refining margins, distribution margins and (value added and excise) taxes. The margins and taxes are captured in a rather stylized but tractable way by a time-varying term (wedge) between the crude oil price paid at the border and the fuel price paid by firms. Other features of the model are in line with the New Keynesian open-economy framework. We estimate the model on EA and RW quarterly data over the 1995–2012 period and obtain the following results. First, a 10% increase in the international price of oil generates an increase of about 0.1 annualized percentage points (p.p.) in the EA consumer price index (CPI) inflation. Second, the same increase in the oil price generates a decrease in EA gross domestic product (GDP) of around 0.1% and an EA trade deficit if it is due to negative oil supply or positive oil-specific demand shocks. Third, it generates a mild EA GDP increase and a trade surplus if due to a positive RW aggregate demand shock. Fourth, results crucially depend on the magnitude of the wealth effect associated with the oil price change. Specifically, conditional on the estimated parameters, a permanent oil supply shock would induce a relatively large medium-run decrease in EA GDP (−0.3%) and increase in EA inflation (0.2 p.p.), while a RW aggregate demand shock that does not induce an oil price increase would have relatively large effects on EA GDP. Fifth, the increase in the price of oil over the 2004–2008 period did not induce stagflationary effects on the EA economy because it was associated with positive RW aggregate demand shocks. The drop in RW aggregate demand contributes to explain the 2008 fall in oil prices, EA GDP and inflation. Finally, our results are robust to changes in the key parameters related to the transmission mechanism of oil shocks. Our paper is, to the best of our knowledge, the first to assess the macroeconomic effects of different oil price shocks on the EA with an estimated open-economy DSGE model. Jacquinot et al. (2009) use a calibrated large-scale open-economy DSGE model to assess the impact of oil price shocks on EA inflation. Consistent with their approach, we distinguish across the various sources of oil price changes but, different from them, we estimate the model with Bayesian methods. Therefore, we are able to perform a quantitative analysis based on estimated impulse response functions, forecast error variance decomposition and historical decomposition. Using a Structural Vector Autoregression (SVAR) model estimated on EA data, Peersman and Van Robays (2009) distinguish different sources of oil price changes: oil supply shocks, oil-specific demand shocks, and global economic activity shocks. European Central Bank (2010) reports the macroeconomic effects of oil price shocks on EA using a variety of models. Our results are in line with those of the above mentioned contributions. Specifically, they are similar to those obtained with the European Central Bank’s New Area Wide Model (NAWM), an estimated DSGE model of the EA economy, enriched with bridge equations for the energy component of the CPI.5 Our model implies larger macroeconomic effects – in line with those obtained by traditional (estimated) macroeconometric models of the EA economy and by the SVAR of Peersman and Van Robays (2009) – under the common assumption of a permanent oil price shock. Finally, our paper contributes to the strand of the literature that has developed structural models to evaluate the macroeconomic effects of oil shocks. For the US, Bodenstein et al. (2011) use a large-scale two-country open economy DSGE model to assess the impact of different oil shocks on the US trade balance under alternative assumptions on the strength of the relative wealth effect across countries associated with changes in the relative prices of oil. Kilian et al. (2009) provide estimates of the effects of demand and supply shocks in the global crude oil market on several measures of oil exporters’ and oil importers’ external balances. They show that the effect of oil demand and supply shocks on the merchandize trade balance and the current account depend on the source of the shock and critically on the response of the nonoil trade balance. Different from these contributions, we develop and estimate a structural model for the EA economy. The rest of the paper is organized as follows. The next section reports the model setup. Section 3 describes the estimation procedure. Section 4 reports the results. Section 5 concludes. 2. The model We develop a two-country model. One country is labeled as Home (it corresponds to the EA), the other as Foreign (it corresponds to the RW). The size of the world economy is normalized to one. The size of the Home country is n (0 < n < 1), the size of the Foreign country is (1 − n). 6 Main features of the model are a world market for crude oil and the need of oil to use physical capital for producing intermediate goods in the manufacturing sector. Remaining features are in line with the New Keynesian open-economy framework.

5 6

See Christoffel et al. (2008). The NAWM does not explicitly formalize the transmission mechanism of oil shocks. We assume that the size of the country is equal to the number of domestic firms in each sector and domestic households.

L. Forni et al. / Journal of Macroeconomics 46 (2015) 295–314

297

In each country, households maximize an intertemporal utility function in consumption and leisure. They are wage-setters and optimally set nominal wages, subject to quadratic adjustment costs. Consumption and investment are non-tradable final goods produced under perfect competition. They are composed of domestic and imported intermediate bundles of (nonfuel) goods. The latter are produced by firms under monopolistic competition according to a Cobb-Douglas technology in labor and “effective” capital, which corresponds to the capacity utilization of the existing physical capital, directly tied to the usage of energy. Intermediate goods are sold domestically and abroad. Firms in the intermediate sector are price-setters and optimally set their prices subject to quadratic adjustment costs. The local currency pricing assumption holds, as firms set prices in the currency of the destination market. Thus, the short-run pass-through of the exchange rate into import and export prices is incomplete. The net foreign asset (NFA) position of one country is composed of a riskless one-period bond denominated in Foreign currency, BF , traded by Home and Foreign households, and paying the Foreign monetary policy rate. When the trade balance shows a surplus (deficit), the NFA position improves (deteriorates). A Taylor rule holds in each country. The monetary policy rate is set to stabilize domestic inflation and real activity. Finally, an uncovered interest parity condition holds, linking the Home and Foreign monetary policy rate differential to the nominal exchange rate. In what follows we describe the world oil market and the main equations for the Home firms, focusing on the role of oil and energy. The complete description of the model is reported in the online Appendix. 2.1. The world oil market The Foreign currency price of crude oil, PtO∗ , is determined in the world oil market. The Foreign country is endowed with an s , governed by a stochastic AR(1) process. The market clearing condition for the world oil market exogenous supply of oil, YO,t reads



s YO,t =

n 0

Ot (he )dhe +



1 n

Ot ( f e )df e ,

(1)

where Ot (he ) and Ot (fe ) are the amounts of oil demanded by the generic firms he and fe in the Home and Foreign fuel sectors, respectively. The law of one price holds at the border.7 The implied crude oil price in Home currency is

PtO = St PtO∗ ,

(2)

where the term S is the nominal exchange rate (number of Home currency units per unit of Foreign currency). 2.2. Firms The fuel sector Firms in the Home fuel sector act under perfect competition. They import crude oil Ot from the RW and transform it into liquid fuel, or “energy” et according to a simple linear technology: et (he ) = Ot (he ).8 Energy is made available to domestic firms producing the Home (nonfuel) intermediate good.9 For Home firms, there is a time-varying wedge ηte between the border price of crude oil expressed in Home currency, PtO , and the retail price of energy Pte :

Pte = PtO + ηte .

(3)

The wedge ηte is a proxy that captures the presence of taxes, refinement and distribution margins in the (retail) price of fuel. It is rebated to Home households in a lump-sum way. Our choice of assuming a simple exogenous process for the margins and taxes guarantees the tractability of the model. Intermediate goods The production function for the generic (nonfuel) intermediate good, produced by firm h, is ξ

1−ξ

Yt (h) = H,t (zt Lt (h)) (ut (h)Kt−1 (h))

,

(4)

where ut (h)Kt−1 (h) denotes physical capital services (Kt−1 (h) is the beginning-of-period t capital stock, ut (h) is its capacity utilization). Lt (h) is labor supplied from domestic households. The parameter ξ (0 < ξ < 1) is the weight of labor in production. The variable zt is a unit-root labor-augmenting technology process common to Home and Foreign firms. Its (gross) growth rate, μz,t ≡ zt /zt−1 , is subject to a shock and has a steady-state value μz . The variable  H, t is a domestic stationary technology process, common to all Home firms. Physical capital evolves according to the following accumulation law



Kt (h) = (1 − δ(ut (h)))Kt−1 (h) + I,t 1 −

ψI 2



It (h) − μz It−1 (h)

2 

It (h),

(5)

7 As we use data about a given quality of oil (Brent), we do not consider that price levels can differ across qualities at least temporarily, as it has been the case for the WTI and other oil qualities since 2010. 8 In the following we use the terms “fuel” and “energy” interchangeably. 9 So firms sell fuel only domestically.

298

L. Forni et al. / Journal of Macroeconomics 46 (2015) 295–314

where investment I is subject to a quadratic adjustment cost (ψ I > 0 is the related parameter). The variable  I, t denotes an investment-specific technology process affecting the efficiency of the newly installed investment good. As in Finn (1995; 2000) capital utilization requires energy et . Specifically, energy complements the service flow from physical capital:

et (h) = a(ut (h))Kt−1 (h),

(6)

and

τ1,t γ1 u , γ1 t

a(ut (h)) =

(7)

where τ 1, t > 0 and γ 1 > 1 are parameters. Eq. (7) states that energy is essential to the utilization u of capital, with increases in utilization requiring more energy usage per unit of capital, at an increasing rate. The parameter τ 1, t is subject to a stochastic shock, labeled as “oil-specific demand”. The shock, common to both Home and Foreign countries, captures changes in oil demand associated with changes in technologies for exploiting energy. Also, the depreciation rate of capital depends on how intensively capital is used in production:

τ 2 γ2 u , γ2 t

δ(ut (h)) =

(8)

where 0 < δ (ut (h)) < 1, τ 2 > 0 and γ 2 > 1 are parameters. The firms’ decisions on capital utilization rate and capital accumulation take into account the two costs of utilization, i.e. depreciation and energy costs, either of which would keep capital from being fully utilized.10 Each firm h in the intermediate sector minimizes its production costs by optimally choosing the amount of inputs given the above technology constraints, the wage rate Wt and the price of energy Pte . We introduce nominal price rigidities by assuming that firms are price-setters and face market-specific costs of adjusting prices à la Rotemberg (1982), when selling their goods domestically and abroad (local currency pricing assumption). Finally, labor L(h) is a composite of a continuum of differentiated labor inputs, each supplied by a different domestic household j under monopolistic competition:

Lt (h) =

 1  1 θL,t n

n 0

Lt ( j)

θL,t −1 θL,t

θL,tθL,t−1 ,

dj

(9)

where 1 < θ L, t < ∞ is the time-varying elasticity of substitution between labor varieties. Households are wage-setters, subject to quadratic wage adjustment costs à la Rotemberg (1982). Final goods Firms in the final goods sector produce three different types of goods under perfect competition. One type is used for private consumption, another for investment and the third one for public sector consumption. The private consumption bundle is produced according to a constant-elasticity-of-substitution (CES) function of nonfuel domestically produced goods CH and imported goods CF



1

η−1

η−1

1

η Ct = aHC CH,tη + (1 − aHC ) η CF,tη

η η−1

,

(10)

where the parameter aHC (0 < aHC < 1) is the share of domestic goods in consumption and η > 0 is the elasticity of substitution between domestic and imported goods. Consumption goods CH and CF are composite baskets of a continuum of, respectively, differentiated intermediate domestic (h) and imported (f) goods, each supplied by a different firm, respectively

CH,t =

CF,t =

 1  1 θH,t n



1 1−n

0

n

CH,t (h)

θF,t1 

1 n

θH,t −1 θH,t

CF,t ( f )

θH,t

θH,t −1

,

dh

θF,t −1 θF,t

θF,tθF,t−1 df

,

(11)

where 1 < θ H, t , θ F, t < ∞ are the time-varying elasticities of substitution among domestic and imported brands, respectively. The production of investment goods I is isomorphic to that of nonfuel consumption (10).11 We allow for (possibly) different import intensities (so for investment we have aHI and 1 − aHI , with 0 < aHI < 1), while the elasticity of substitution between domestic 10 11

See also Katayama (2013) for a similar treatment of the relationship between capital utilization and energy. Specifically,



1

η−1

1

η−1

η It = aHI IH,tη + (1 − aHI ) η IF,tη

η η−1

,

L. Forni et al. / Journal of Macroeconomics 46 (2015) 295–314

299

and imported bundles (η) and across varieties in each bundle (θ H, t , θ F, t ) are the same as those in the corresponding consumption bundles. For the public expenditure, we assume it is fully biased towards domestic nonfuel varieties.12 2.3. Trade balance The Home trade balance, expressed in domestic currency, is

 T Bt =

n 0

∗ ∗ St PH,t (h)YH,t (h)dh −



1 n

PF,t ( f )YF,t ( f )df −



n 0

PtO Ot (he )dhe .

The overall trade balance can be split into nonoil and oil trade balances. The value (in Home currency) of the nonoil trade balance is the difference between the first two integrals on the right-hand side of the equation. The term S is the nominal exchange ∗ h and Y ∗ h represent prices (in Foreign rate (number of Home currency units per unit of Foreign currency). The terms PH,t ( ) H,t ( ) currency) and amounts of exported goods by Home firms (indexed by h), respectively. The terms PF, t (f) and YF, t (f) respectively represent prices (in Home currency) and amounts of Home imported goods (produced by Foreign firms, indexed by f). The last term on the right-hand side is oil-trade balance (value in Home currency), where PtO is the oil price in Home currency and Ot (he ) is the amount of oil demanded by the generic firm he in the Home energy sector. 2.4. Foreign economy The setup of the Foreign economy is similar to that of the Home economy. To keep the model parsimonious, we assume that in the Foreign economy the consumption and investment baskets have the same composition (so they have the same price deflator), wages are fully flexible and there is no distinction between border price of crude oil and the retail price of energy. For the supply of nonoil goods produced in the Foreign country and imported by the Home economy, we assume that the local currency pricing assumption holds (so the prices of Foreign exports are set in Home currency). 2.5. Shocks All shocks follow an exogenous (log-linear) autoregressive process iid Xˆt = ρX Xˆt−1 + εˆX,t , εˆX,t ∼ N(0, σX2 ),

(13)

¯ The only excepwhere 0 ≤ ρ X ≤ 1 and a hat denotes log-deviation from the corresponding steady-state level: Xˆt = ln Xt − ln X. tion is the monetary policy shock, which is iid. The list of Home (EA) shocks includes: consumption preference; neutral technology; investment-specific; markup (domestic tradables, imported tradables, wages); public consumption; monetary policy rate; inflation target; uncovered interest parity; domestic risk premium. Foreign (RW) shocks include: consumption preference; investment-specific technology; growth rate of labor-augmenting stochastic trend (common to the Home country); markup (domestic tradables, imported tradables); monetary policy rate; inflation target. Finally, oil-related shocks include Home fuel distribution margin, worldwide oil supply and worldwide oil-specific demand. The Foreign country faces fewer shocks than the Home country does. This choice is deliberate. It reflects the need to keep the model and the estimation process parsimonious. Since our analysis focuses on the EA economy, we use fewer observable variables for the RW than for the EA. This allows us to identify a relatively small number of structural shocks for the RW. For instance, as we only use one aggregate demand indicator for the Foreign economy, we include a consumption preference shock, but not a government consumption shock. 2.6. Equilibrium We consider a symmetric equilibrium defined as a sequence of quantities and prices such that given initial capital positions and bond positions and the laws of motions for shocks: (i) in each country, the representative household maximizes utility subject to the budget constraint; (ii) in each sector, the representative firm maximizes profits subject to the technology constraint; (iii) all markets clear (including the worldwide oil market). We solve the model by log-linearizing it around the steady-state equilibrium.

and

IH,t = 12

 1  1 θH,t n

n 0

IH,t (h)

θH,t −1 θH,t

θH,t

θH,t −1

, IF,t =

dh

 θF,t1  1 1−n

n

1

IF,t ( f )

θF,t −1 θF,t

θF,tθF,t−1 df

.

The implied basket for public expenditure, Gt , is

Gt =

 1  1 θH,t n

n 0

GH,t (h)

θH,t −1 θH,t

θH,t

θH,t −1

dh

.

For simplicity, we assume that public consumption is financed by lump-sum (nondistortionary) taxes paid by domestic households.

(12)

300

L. Forni et al. / Journal of Macroeconomics 46 (2015) 295–314

3. Estimation We follow the Bayesian approach: the posterior distribution of the estimated parameters is obtained by updating the information in the prior distribution with the information in the data.13 The assumption of a non-stationary technology shock implies a common stochastic trend in the real variables. To make the model stationary, we therefore scale all variables that share the common real trend with the level of productivity, zt . We then proceed with the log-linearization of the model around its non-stochastic steady state. 3.1. Data We use quarterly data for the period 1995:1–2012:4. We are forced to start from 1995 as data for fuel are not available before that date. We match seventeen variables. For the EA (the Home country): real GDP, consumption, investment, exports, imports, employment, CPI deflator (measured by the Harmonized Index of Consumer Prices), investment deflator, fuel relative price (as measured by the corresponding EA HICP component), nominal wage, short-term nominal interest rate, real effective exchange rate. For the RW (Foreign): aggregate demand, CPI deflator, short-term nominal interest rate. For the international oil market: oil supply, crude oil price. All EA data are from the Area Wide Model (AWM) dataset.14 In the AWM dataset export and import series include both intra- and extra-area trade and there is no series on aggregate hours worked.15 The exchange rate is the European Central Bank’s official effective exchange rate for the 12 main trading partners of the EA.16 For the Foreign interest rate, we use the effective Fed funds rate as a proxy. Data for oil supply are from OECD-International Energy Agency. Prior to estimation, we take log first-differences of real GDP, consumption, investment, export and import (consistent with the assumption of balanced growth path), consumption and investment deflators, wage, foreign prices. We remove a linear trend from employment, the oil supply, the RW demand. We also remove an excessive trend of import and export (with respect to EA GDP) series, to make the corresponding shares stationary.17 Interest rates, employment, the real exchange rate, the real price of oil and the real price of fuel are measured as percentage deviations around the mean. The real price of oil (Brent quality) is the US dollar price of oil deflated by the Foreign price index. The real price of fuel is the euro price deflated by the EA CPI index. Finally, we also include measurement errors (ME) in the model. The ME are iid and their standard deviation is calibrated to a rather small value, so as to minimize their impact on the estimation process. Thus, the number of observable variables is lower than the number of structural shocks plus measurement errors and we avoid stochastic singularity. 3.2. Calibrated parameters We calibrate parameters to match the sample mean of observed variables and those that are weakly identified. Values are in line with Adolfson et al. (2007), Christoffel et al. (2008), Jacquinot et al. (2009) and European Central Bank (2010). In Table 1 we report the calibrated parameters. In Table 2 the implied steady-state values of main variables. For households’ preferences, the discount factor β is set to 0.9996 , consistent with an annualized equilibrium nominal interest rate of 4.0% (the sample mean). The inverse of the labor supply elasticity, σ L , is set to 1. The elasticity of substitution between domestic and imported non-oil goods, η, is 1.1. Concerning the energy-related parameters, we set the energy efficiency parameters τ 1 , τ1∗ , τ 2 and τ2∗ to 1, in line with the literature (see e.g. Leduc and Sill, 2004). We set γ 1 to 1.25, in order to match the steady-state ratio of oil imports to output, which amounts to 2% as reported in Table 2. Analogously, we set γ1∗ to 1.2. Finally, we choose γ 2 and γ2∗ to set the depreciation rate (δ and δ ∗ ) of physical capital to 0.025. The corresponding values are 1.16 and 1.19. We set the (steady-state) elasticity of substitution across brands (θH , θF , θH , θF ) to 6, the elasticity of substitution across labor varieties, θ L , to 4.33. They imply steady-state markup values equal to 1.2 and 1.3, respectively. We calibrate the weight of labor, ξ , in the production function to 0.7 (the weight of capital services is 0.3). We set the value of the parameter ηe to 0.2, so that distribution margin, refinement margin and oil taxes, as share of the fuel price, are equal to 0.6, in line with results reported in European Central Bank (2010). The calibration allows us to match all the ratios reported in Table 2. The steady-state NFA position is set to zero, implying that both trade balance and current account are equal to zero. The steady-state growth rate of the world economy is 2.0% per annum. Both EA and RW long-run annualized gross inflation, π¯ and π¯ ∗ , are set to 1.9%.

13 14 15

For a comprehensive discussion on the Bayesian estimation of DSGE models, see Lubik and Schorfheide (2005). For details on the AWM dataset see Fagan et al. (2005). We model the link between employment and hours worked using an auxiliary Calvo-rigidity equation,

Eˆt =

β

1+β



Et Eˆt+1 +

1 (1 − βξE )(1 − ξE )  ˆ ˆ  Eˆt−1 + Lt − Et , 1+β (1 + β)ξE

where β ∈ (0, 1) is the household’s discount factor and ξ E is the fraction of firms that cannot adjust the (log-linear) level of employment Eˆ to the preferred amount of total labor input Lˆ. See Smets and Wouters (2003). 16 See Adolfson et al. (2007). 17 The data treatment is similar to Adolfson et al. (2007) and Christoffel et al. (2008).

L. Forni et al. / Journal of Macroeconomics 46 (2015) 295–314

301

Table 1 Calibrated parameters. Parameter

Description

Value

β β∗ σL σL∗

Home discount factor Foreign discount factor Home inverse of labor supply elasticity Foreign inverse of labor supply elasticity Home domestic trad. cons. share Home domestic trad. inv. share Foreign domestic trad. share Home domestic trad. - nonfuel import substitution Foreign domestic trad. - import substitution Home energy efficiency Foreign energy efficiency Home energy efficiency Foreign energy efficiency Home energy technology Foreign energy technology Home energy technology Foreign energy technology Home trad. brands substitution Home nonfuel import brands substitution Foreign trad. brands substitution Foreign import brands substitution Home labor varieties substitution Home labor weight Foreign labor weight Home size Home distribution margin and taxes

0.9996 0.9996 1.00 1.00 0.84 0.74 0.95 1.1 1.1 1.00 1.00 1.00 1.00 1.25 1.20 1.16 1.19 6 6 6 6 4.33 0.7 0.7 0.3 0.2

aHC aHI a∗F

η η∗ τ1 τ1∗ τ2 τ2∗ γ1 γ1∗ γ2 γ2∗ θH θF θF∗ θH∗ θL ξ ξ∗

n

ηe

Table 2 EA steady state. Variable

Description

Value

π¯ gr R C/Y I/Y PH G/Y (PF (CF + IF ) + POYOD,H )/(PY ) = SPH∗ YH∗ /Y PF CF /Y PF IF /Y P OYOD,H /Y SBF /Y Pe e/(MCY)

Inflation rate (annualized) Growth rate (annualized) Nominal interest rate (annualized) Consumption-to-output ratio Investment-to-output ratio Public expenditure-to-output ratio Import-to-output ratio = Export-to-output ratio Cons. Imp.-to-output ratio Inv. Imp.-to-output ratio Oil imports-to-output ratio Net foreign asset position-to-output ratio Share of fuel in the production cost

1.9 2.0 4.0 0.57 0.21 0.22 0.17 0.09 0.05 0.02 0.00 0.05

3.3. Prior distributions of the estimated parameters We report in Table 3 the prior distributions of the estimated EA parameters (to save on space, prior and posterior distributions of RW estimated parameters are reported in the online Appendix). Their location corresponds to a large extent to that in Adolfson et al. (2007). Parameters bounded between 0 and 1 are distributed according to a beta (B in the table) distribution. Positive parameters other than shocks’ standard deviations have a gamma (G) distribution. Shocks standard deviations have an inverse gamma (IG) distribution. For the monetary policy rule, the prior mean on the lagged interest rate coefficient is set to 0.8, those on inflation and inflation growth coefficients respectively to 1.7 and 0.3. Finally, the prior mean of the coefficient responding to output growth is set to 0.06. For nominal rigidities, we set the prior mean of wages and prices of Home as well Foreign intermediate goods to 250 (if converted in Calvo, 1983 terms, it implies an average contract duration equal to about 8 quarters). The standard deviation is set to 80. For imports and exports, we set the prior mean to 15 (in Calvo, 1983 terms, it corresponds to a contract duration of 2 quarters), so that the exchange rate pass-through into import and export prices is rather quick. We set the standard deviation to 5. The implicit assumption of relatively flexible import and export prices is consistent with estimates by Adolfson et al. (2007), that suggest 2–3 quarters stickiness in these sectors. Finally, we set the mean values of the indexation parameters to 0.5 (standard deviation equal to 0.1). All the autocorrelated shocks have an

302

L. Forni et al. / Journal of Macroeconomics 46 (2015) 295–314 Table 3 Prior and posterior moments of the parameters. Prior

Preferences Habit formation Adjustment costs Investment Monetary policy Interest rate smoothing Resp. to inflation Resp. to change in inflation Resp. to output growth Wage and price setting Dom. prices Exp. Prices Imp. prices Wages Indexation: prices Indexation: exports Indexation: wages Employment Calvo-style parameter Shocks: autoregressive coefficients Transitory techn. Inv.-spec. tech. Permanent tech. Preferences Public exp. Domestic risk premium External risk premium Inflation target Price markup: domestic Price markup: import Price markup: export Wage markup Fuel margin Oil supply Oil-specific demand shock Shocks: standard deviations Monetary policy Transitory techn. Inv.-spec. tech. Permanent tech. Preferences Public exp. Domestic risk premium External risk premium Inflation target Price markup: domestic Price markup: import Price markup: export Wage markup Fuel margin Oil supply Oil-specific demand shock

Posterior

Type

Mean

Std.dev.

Mode

Mean

Median

5%

95%

h

B

0.5

0.15

0.84

0.87

0.88

0.77

0.93

ψ

G

3

1.5

7.40

7.43

7.15

4.69

11.25

ρR ρπ ρ π ρ y

B G G G

0.8 1.7 0.3 0.06

0.05 0.25 0.15 0.05

0.89 1.80 0.15 0.11

0.89 1.91 0.15 0.09

0.89 1.90 0.15 0.09

0.86 1.60 0.08 0.07

0.91 2.26 0.21 0.12

κH κH∗ κF κW αH α∗ αW

G G G G B B B

250 15 15 250 0.5 0.5 0.5

80 5 5 80 0.1 0.1 0.1

213.21 14.84 27.14 287.28 0.44 0.24 0.24

201.35 16.84 27.30 287.19 0.42 0.24 0.22

200.03 16.30 27.03 278.68 0.42 0.23 0.22

79.29 10.62 18.68 164.97 0.29 0.15 0.13

316.28 24.92 36.78 432.77 0.54 0.33 0.34

ξE

B

0.5

0.15

0.80

0.84

0.84

0.81

0.86

ρ ρY ρμ ρζ C ρg ρ RP ρφ˜ ρπ¯ ρθ H ρθ F ρθH∗ ρθL ρηe ρySO ρτ 1

B B B B B B B B B B B B B B B

0.75 0.75 0.75 0.75 0.75 0.75 0.75 0.75 0.75 0.75 0.75 0.75 0.75 0.75 0.75

0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15 0.15

0.76 0.88 0.81 0.89 0.76 0.74 0.93 0.29 0.78 0.97 0.94 0.94 0.94 0.86 0.91

0.85 0.83 0.73 0.81 0.81 0.69 0.93 0.38 0.73 0.97 0.91 0.91 0.95 0.89 0.94

0.85 0.83 0.74 0.82 0.85 0.70 0.93 0.36 0.73 0.97 0.92 0.94 0.95 0.89 0.95

0.73 0.73 0.50 0.65 0.49 0.49 0.87 0.17 0.58 0.94 0.82 0.77 0.91 0.80 0.88

0.98 0.91 0.89 0.91 0.95 0.87 0.97 0.83 0.86 0.99 0.97 0.98 0.98 0.96 0.97

σR σ σY σμ σζ C σg σ RP σφ˜ σπ¯ σ θH σ θF σθH∗ σ θL ση e σySO σ τ1

IG IG IG IG IG IG IG IG IG IG IG IG IG IG IG IG

0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

0.06 0.06 3.45 0.25 2.18 0.06 0.06 0.23 0.31 31.26 15.34 18.10 9.27 5.10 0.91 1.26

0.06 0.70 3.98 0.10 3.26 0.08 0.06 0.25 0.29 32.46 15.47 21.35 12.24 5.16 0.93 1.34

0.06 0.67 3.82 0.08 3.16 0.07 0.05 0.24 0.29 31.16 15.36 21.02 8.40 5.13 0.92 1.33

0.05 0.39 2.58 0.03 1.80 0.03 0.03 0.15 0.15 11.94 12.52 16.60 4.62 4.46 0.81 1.15

0.08 1.10 5.84 0.20 5.24 0.21 0.10 0.38 0.39 56.00 18.76 27.42 32.83 5.97 1.07 1.55

autoregressive coefficient set to 0.75. Innovations to all shocks are assumed to be white noise with standard deviation mean set to 0.1%. 4. Results In what follows we report the estimated values of the parameters, and results from analysis performed on the basis of those estimates, i.e. the responses of the main EA variables to shocks that directly hit the oil market and the contribution of these shocks to the forecast error variance and historical path of EA GDP and inflation. Finally, we report results of the sensitivity analysis, based on appropriately changing the key parameters for the transmission of oil shocks (parameters γ 1 and γ 2 in Eqs. (7) and (8), respectively).

L. Forni et al. / Journal of Macroeconomics 46 (2015) 295–314

303

4.1. Posterior distributions of the estimated parameters We estimate the posterior distributions of the parameters using the Metropolis–Hastings algorithm with 2,000,000 iterations. The results are reported in Table 3, where we show the posterior mode, mean, median and the 5th and 95th percentiles. The posterior means of the degree of habit formation in consumption and the investment adjustment cost parameter are equal to 0.87, and 7.43, respectively, broadly in line with the corresponding values reported in Adolfson et al. (2007). For nominal price rigidities, we find that the degree of domestic price stickiness is equal to be 0.85 in Calvo (1983) terms, which imply an average contract duration of 6 quarters, slightly lower than the findings by Adolfson et al. (2007), Christoffel et al. (2008), and Smets and Wouters (2003); 2005). Nominal wages are sticky as well. The implied duration in Calvo (1983) terms is equal to 10 quarters. The estimated value is higher than those reported by Adolfson et al. (2007) and Christoffel et al. (2008). The likely reason is that high sticky wages contribute to stabilize marginal costs and prices in correspondence of volatile oil prices. For indexation parameters, we find that they are rather low, around 0.3 on average. This finding is common to other contributions (e.g. Adolfson et al., 2007). The posterior means of the shocks persistence parameters are generally above 0.8. Fig. 1 shows the data and the Kalman filtered one-sided estimates of the observed variables, computed at the posterior mode of the estimated parameters. The in-sample fit of the model is particularly good, considering the inclusion of the global financial crisis period in the estimation sample. The model accounts for around a half of the abrupt fall in GDP growth observed in 2009 and almost perfectly tracks the dynamics of CPI inflation in the same period. Finally, parameters seem to be rather well identified. As reported in the online Appendix, data are generally informative about the parameters, since in general the posterior distributions differ from the priors. Moreover, diagnostics suggest that the posterior draws have converged to the true target posterior density for most of the estimated parameters. In particular, the oil shocks parameters (persistence and volatility of oil supply, oil-specific demand and RW aggregate demand shocks) seem to be identified. Oil supply and oil-specific demand shocks are estimated to be rather persistent, with parameter estimates pointing towards the upper bound of the estimates range, equal to one. Finally, the persistence and volatility of the Home wage

EA gdp growth 0.01

EA employment fitted data

0

EA real eff exch 0.1

0.02 0

0

−0.01 −0.02

−0.1

−0.02 −0.04 Q1−00 −3

x 10

Q1−05

Q1−10

Q1−00

Q1−05

Q1−00

Q1−10

EA investment growth

EA consumption growth

Q1−05

Q1−10

EA CPI inflation

−3

x 10 10

10

0.02

5

0

0

−0.02

5 0

−0.04 −5

−5

−0.06 Q1−00 −3

x 10

Q1−05

Q1−10

Q1−00

EA inv price inflation

10

Q1−05

Q1−10

Q1−00

EA wage inflation

−3

x 10

Q1−05

Q1−10

EA imports gr

10

0.02

5

0 −0.02

5

−0.04

0

−0.06

0 Q1−00

Q1−05

Q1−10

Q1−00

EA exports gr

Q1−05

Q1−10

0.02 0 −0.02 −0.04 −0.06 −0.08

0.005 0 Q1−05

Q1−10

Q1−00

RW inflation

x 10

−3

x 10

Q1−05

Q1−10

Q1−00

RW nominal interest rate

Q1−05

Q1−10

EA fuel rel. price

15

8 6 4 2 0 −2

Q1−10

0.04 0.02 0 −0.02 −0.04 −0.06

0.01

Q1−00

Q1−05

RW aggregate demand

0.015

−3

Q1−00

EA nominal interest rate

4

10 2

5 0 Q1−00

Q1−05

Q1−10

0 Q1−00

Q1−05

Q1−10

oil supply

oil internat. rel. price 4 0.02 0

2

−0.02 0 Q1−00

Q1−05

Q1−10

Q1−00

Q1−05

Q1−10

Fig. 1. Data (thick) and one-sided predicted values from the model (thin).

Q1−00

Q1−05

Q1−10

304

L. Forni et al. / Journal of Macroeconomics 46 (2015) 295–314

Oil price

CPI inflation rate 0.2

10

0

0

Output

Interest rate

0.2

0.1

0

0

−10 −0.2 −0.2 −0.1 10 0 0 20 0 10 10 20 0 20 Capacity utilization Consumption Investment 0 0.5 0 0 −0.1

−0.5

−0.2 0 10 20 Hours worked 0.5 0

−1 0

0 10 20 Real wage

−0.5 0

0

0.05

−0.1

0

0.1

−10 0

10

20

0 0

10 Export

20

20

−0.2 0

−1 0 10 20 Nonoil import 0 −0.5

0 10

20

−0.5

−0.5 −0.2 −0.05 0 10 10 20 0 10 20 20 0 Oil import (value) Real exchange rate Trade balance 10 0.2 0.2 0

10 Energy

−1 0 10 20 Nonoil trade balance 0.1 0.05

10

20

0 0

10

20

Fig. 2. Responses of EA variables to a negative oil supply shock. Horizontal axis: quarters. Vertical axis: percentage deviations from the baseline, except for inflation and interest rates (annualized percentage-point deviations), and the trade balance (ratio to GDP, percentage-point deviations). GDP (labeled as “Output” in the charts) and its components are reported in real terms. Real GDP corresponds to real value added.

markup and inflation target shocks seem to be weakly identified, if at all. Indeed, diagnostics suggest that convergence was not fully achieved, as the across-chain and within-chain moments do not perfectly overlap. As a result, the corresponding posterior densities do not coincide with the true target ones and their comparison with the prior distributions is non-informative. The fact that Home wage markup and inflation target shocks are not well identified is however not surprising, since both shocks enter the wage Phillips curve and we do not have data on the inflation target. Thus, many combinations of these shocks can affect the relation between wage inflation and the consumption-leisure marginal rate of substitution in about the same way. Nonetheless, the lack of identification of these shocks does not affect the transmission of the (orthogonal) oil-related shocks, which is the main focus of our analysis. 4.2. Impulse response functions We perform an impulse response analysis to assess how the main EA macroeconomic variables react to shocks affecting the oil market, i.e. the global oil supply shock, the oil-specific demand shock and the RW aggregate demand shock. Oil supply shock We assess the effects on the EA economy of an exogenous reduction in the global oil production that induces on impact an increase in the international real (expressed in foreign consumption terms) price of oil equal to 10% of its steady-state level. Fig. 2 reports the results.18 The CPI inflation rate increases on impact, up to around 0.1 annualized p.p., and then quickly falls, reflecting the rapid and complete pass-through of oil price movements into the price of fuel. Given the increase in CPI inflation, the monetary authority rises the policy rate, in line with results reported in Jacquinot et al. (2009). Consumption persistently declines, because EA households suffer a negative wealth effect relative to the RW households, associated with the higher oil price.19 For the same reason, households increase labor supply, as implicitly suggested by the decline in real wages. Investment decreases, in a hump-shaped and persistent manner. Underlying its decline is a rapid and prolonged fall in capacity utilization and energy, due to the higher energy cost, that more than compensates for the labor increase in the production function. Overall, there is a rather significant and persistent reduction in GDP (around 0.1%). The EA trade balance deteriorates by about 0.1% of GDP. The nonoil component of the trade balance improves (in line with results reported by Kilian et al., 2009), benefiting from the depreciation of the EA real exchange rate, which makes EA goods 18 All figures in this section report the median (solid line) and the 95 percent equal-tail uncertainty bands. The results are based on 5,000 draws from the posterior distribution of the model parameters. 19 There is a cross-country relative wealth effect because international financial markets are incomplete (only one bond is internationally traded). The relative wealth effect is negative for the EA because it is an oil importer. See Bodenstein et al. (2011).

L. Forni et al. / Journal of Macroeconomics 46 (2015) 295–314

305

CPI inflation rate

Oil price 15

0.3 0.2

10

0.1 5 0 0

0 5

10

15

20

−0.1 0

5

Output 0.8

0.2

0.6

0

0.4

−0.2

0.2 5

10

15

20

0 0

Trade balance 0.2

0

0.15

−0.05

0.1

−0.1

0.05 5

10

20

5

10

15

20

Nonoil trade balance

0.05

−0.15 0

15

Real exchange rate

0.4

−0.4 0

10

15

20

0 0

5

10

15

20

Fig. 3. Responses of EA variables to a permanent negative oil supply shock. Horizontal axis: quarters. Vertical axis: percentage deviations from the baseline, except for inflation and interest rates (annualized percentage-point deviations), and the trade balance (ratio to GDP, percentage-point deviations). GDP (labeled as “Output” in the charts) and its components are reported in real terms. Real GDP corresponds to real value added.

cheaper than RW nonoil goods and induces a large drop in EA nonoil imports.20 The real exchange rate depreciates because of the relatively large drop in EA aggregate demand, which is biased towards domestically-produced goods. Aggregate demand in the RW (not reported to save on space) falls relatively less than in the EA, as RW households benefit from a positive (crosscountry) relative wealth effect. EA exports are negatively affected by the drop in RW aggregate demand (the negative effect is only partially offset by the positive one associated with the EA exchange rate depreciation). To further evaluate the wealth effect on the EA economy, we have simulated a permanent negative oil supply shock. Fig. 3 reports the impulse responses. The response of the CPI inflation is larger and more persistent than in the benchmark case, while the short-run response of output does not greatly change relative to the benchmark. In the case of a permanent shock, the negative wealth effect on EA households spending decisions is larger. This implies a larger drop in economic activity. Results are in line with those obtained with other macroeconomic models of the EA when assuming, as we do here, permanent negative oil price shocks.21 Oil-specific demand shock Fig. 4 reports the effects of an oil price rise generated by the oil market-specific demand shock that simultaneously affects EA and RW (it corresponds to an increase in τ 1, t in Eq. (7) in both regions). We calibrate the shock so that the international relative price of oil increases by 10% on impact, to facilitate comparison with the oil supply shock. A positive shock implies that a higher amount of energy is demanded for a given capacity utilization of capital. The implied increase in oil price generates general equilibrium effects similar to those of a negative oil supply shock. In particular, the drop in capacity utilization induces a fall in energy consumption. The shock induces an increase in the EA CPI inflation (0.1 p.p.), a persistent decrease in EA GDP (-0.1%) and a deterioration of the trade balance. Results are in line with Bodenstein et al. (2011).

20 The EA real exchange rate is defined as ratio of RW to EA consumer prices, both expressed in the EA currency. The increase (decrease) corresponds to the depreciation (appreciation). 21 See European Central Bank (2010).

306

L. Forni et al. / Journal of Macroeconomics 46 (2015) 295–314

Oil price

CPI inflation rate 0.5

20 10 0 0

0 10 20 Consumption

−0.5 0

0

0

−0.1

−0.5

10 20 Investment

Output 0.1

−0.1

0

−0.1 −0.2 0 20 10 0 Capacity utilization 0 0

−0.1

10 Energy

20

−0.1

−0.5

−0.2 −1 −1 10 0 10 20 0 20 0 Real wage Hours worked 0.5 0 0.1 0

Interest rate

0

10 Export

20

−0.2 0

10 20 Nonoil import

0

0

−0.5

−0.5 −0.2 −0.1 −1 0 10 20 0 10 20 0 10 20 0 10 20 Oil import (value) Real exchange rate Trade balance Nonoil trade balance 0.2 0.2 0.1 20 10 0 0

0

0.1 10

20

0 0

10

20

−0.2 0

0.05 10

20

0 0

10

20

Fig. 4. Responses of EA variables to a positive oil-specific demand shock. Horizontal axis: quarters. Vertical axis: percentage deviations from the baseline, except for inflation and interest rates (annualized percentage-point deviations), and the trade balance (ratio to GDP, percentage-point deviations). GDP (labeled as “Output” in the charts) and its components are reported in real terms. Real GDP corresponds to real value added.

Oil price

CPI inflation rate 0.2 0.5

20 0 −20 0

0 10 20 Consumption

−0.5 0

0

0

−0.1

−1

Interest rate 0.2 0

0 10 20 Investment

−0.2 −0.2 0 20 0 10 Capacity utilization 1 1 0

−0.2 −2 −1 10 0 10 20 0 20 0 Hours worked Real wage 0.5 2 0 0

Output

−0.2

10 Energy

20

0 10 Export

20

−1 0

10 20 Nonoil import

1 0

0

−0.5 −0.4 −2 −1 0 0 10 20 10 20 20 0 10 0 10 20 Oil import (value) Real exchange rate Trade balance Nonoil trade balance 20 1 0.2 0.4 0 −20 0

0.1

0 10

20

−1 0

10

20

0 0

0.2 10

20

0 0

10

20

Fig. 5. Responses of EA variables to a positive RW aggregate demand shock. Horizontal axis: quarters. Vertical axis: percentage deviations from the baseline, except for inflation and interest rates (annualized percentage-point deviations), and the trade balance (ratio to GDP, percentage-point deviations). GDP (labeled as “Output” in the charts) and its components are reported in real terms. Real GDP corresponds to real value added.

L. Forni et al. / Journal of Macroeconomics 46 (2015) 295–314

CPI inflation rate

Oil price

307

Interest rate

Output

1

0.1

0.2

0.1

0

0

0.1

0

−1 0

10

20

−0.1 0

10

20

0 0

Investment

Consumption

10

20

−0.1 0

Capacity utilization

0.1

0.5

0.4

0.4

0

0

0.2

0.2

−0.1 0

10

20

−0.5 0

Hours worked 0.2

0.1

0.1

10

20

0 0

Real wage

0.2

0 0

10

20

0 0

Oil import (value)

10

20

0 0

Export

20

1

1

0.5

0

0

−1 0

10

20

10

20

Nonoil import

2

Real exchange rate

1

10

10 Energy

20

−0.5 0

Trade balance

10

20

Nonoil trade balance

0.5

0.2

0.2

0

0

0

0.5 0 −0.5 0

10

20

−0.5 0

10

20

−0.2 0

10

20

−0.2 0

10

20

Fig. 6. Responses of EA variables to a positive RW aggregate demand shock when the oil price does not react. Horizontal axis: quarters. Vertical axis: percentage deviations from the baseline, except for inflation and interest rates (annualized percentage-point deviations), and the trade balance (ratio to GDP, percentagepoint deviations). GDP (labeled as “Output” in the charts) and its components are reported in real terms. Real GDP corresponds to real value added.

RW aggregate demand shock Fig. 5 reports the effects on the EA economy of a positive shock to the RW aggregate demand that increases the international oil price by 10%. 22 The EA CPI inflation rate increases up to the peak of about 0.15 p.p. (annualized). Different from the previous two oil shocks, the positive RW demand shock has an expansionary effect on EA exports and GDP. The latter increases up to 0.1%. Consistent with the increase in inflation and economic activity, the monetary authority rises the policy rate. Importantly, consumption falls, as households suffer the large negative wealth effect associated with the increase in the price of oil. Investment in physical capital also decreases, as the higher price of energy induces a fall in capacity utilization. As in the case of oil supply and demand shocks, hours worked increase, due to the negative wealth effect. Real wages correspondingly fall. Crucially, the trade balance now improves (up to 0.1%, as a ratio to GDP), driven by a large and persistent rise in exports and by a decrease in nonoil imports (due to lower aggregate demand). Thus, the nonoil trade balance surplus (about 0.2% of GDP) more than compensates for the large increase in the value of oil imports, which fully reflects the higher price of oil. The EA real exchange rate depreciates, as the world demand for RW goods increases relatively more than that for EA goods, because of the bias of RW demand for domestically produced goods. The RW demand shock thus affects EA dynamics in two ways: directly, as it increases exports, and indirectly, via its impact on oil demand and hence oil price. To isolate the latter effect, in Fig. 6 we show the responses under the assumption that the price of oil is not affected by the increase in the RW demand. EA inflation is now barely affected by the shock, and it actually slightly falls on impact, albeit the effect is not statistically significant. Consumption and investment slightly decrease, only in the initial periods, while the real wage increases. The reason is that EA households are no longer affected by the negative wealth effects associated with the higher oil price. EA GDP increase is more persistent than in the previous case. Higher consumption drives up oil and nonoil imports. For oil imports, now quantities increase. The increase in oil import value (expressed in EA currency) is much lower, reflecting the unchanged oil price. The overall trade balance improves by around 0.1% (ratio to GDP), more than in previous case.

22

The increase in RW aggregate demand is implemented through a positive consumption preference shock.

308

L. Forni et al. / Journal of Macroeconomics 46 (2015) 295–314

100

Tech Mon pol Dem Mkps Foreign dem Fuel Oil sup Oil−specific Foreign sup Total

50

0

−50

04Q1

05Q1

06Q1

07Q1

08Q1

09Q1

10Q1

11Q1

12Q1

Fig. 7. Historical decomposition of international relative price of oil. Each colored bar shows how the corresponding group of shocks contributes to the percentage deviation from steady state of the international relative price of oil in a given quarter. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

4.3. Forecast error variance decomposition Tables 4–6 decompose the forecast error variance at different horizons of the main EA variables into components attributable to the shocks perturbing the model. We report values obtained using the posterior mode. We group the shocks as follows: technology (Tech), monetary policy (Mon pol), EA demand (Dem), markup (Mkps), RW demand (Foreign dem), fuel margin (Fuel), oil supply (Oil sup), oil-specific demand (Oil-spec), RW supply (Foreign sup). 23 The combination of oil-specific demand and oil supply shocks gives the largest contribution to the international relative price of oil at 1-quarter horizon. Thereafter, the oil price is explained by the combination of shocks to RW aggregate demand (above 40%), oil supply (about 15%) and oil-specific (about 30%). A similar result holds for the EA price of fuel, given that the pass-through of oil price into the fuel price is quick. The impact of oil shocks on EA CPI inflation is not trivial. Oil price shocks explain around 5% at all considered horizons. In particular, oil-specific demand explains above 3%. The impact of oil shocks on EA GDP is virtually nil. GDP is mainly explained by a combination of domestic demand, markup and technology shocks, in line with the evidence reported in Christoffel et al. (2008). Overall, results suggest that oil shocks mainly affect the EA CPI inflation.

23 Following Christoffel et al. (2008), the technology group includes the worldwide permanent technology shock, the EA transitory technology shock and the investment-specific technology shock. The monetary policy group comprises the innovation to EA interest rate and the inflation target shock. The EA demand group includes shocks to consumption preferences, UIP, domestic risk premium, government consumption. The markup group consists of shocks to the EA wage markup, domestic tradables price markup and import price markup. The foreign demand group consists of shocks to RW consumption preferences, investmentspecific technology, interest rate and inflation target. Finally, the foreign supply group comprises the shock to EA export price markup, RW domestic tradables price markup. Moreover, we include the contribution of measurement errors (ME) used in the estimation process. Their contribution becomes not negligible during the recent global financial crisis, as they allow to fit data even in presence of large nonlinearities (that cannot be captured by our linear model).

L. Forni et al. / Journal of Macroeconomics 46 (2015) 295–314

309

Table 4 Forecast error variance decomposition (1-quarter horizon).

Oil internat. rel. price EA fuel rel. price EA CPI inflation EA gdp growth

Tech

Mon pol

Dem

Mkps

Foreign dem

Fuel

Oil sup

Oil-spec

Foreign sup

Meas. Errors

Total

5.8 5.4 0.5 24.0

0.1 0.0 20.0 2.5

0.2 0.2 3.3 22.6

0.1 0.1 60.6 13.6

3.6 3.2 2.0 2.4

0.5 3.8 0.0 0.1

34.5 33.4 1.6 0.0

55.2 53.6 3.7 0.1

0.1 0.0 1.0 26.5

0.0 0.2 7.4 8.1

100.0 100.0 100.0 100.0

Table 5 Forecast error variance decomposition (4-quarter horizon).

Oil internat. rel. price EA fuel rel. price EA CPI inflation EA gdp growth

Tech

Mon pol

Dem

Mkps

Foreign dem

Fuel

Oil sup

Oil-spec

Foreign sup

Meas Errors

Total

5.2 5.0 4.1 27.1

0.0 0.0 16.4 2.3

0.5 0.4 4.3 21.5

1.7 3.3 57.9 16.6

41.4 37.6 5.2 2.6

0.4 4.4 0.1 0.1

17.4 17.0 1.3 0.1

32.7 32.0 3.2 0.3

0.5 0.1 1.9 22.7

0.0 0.1 5.6 6.7

100.0 100.0 100.0 100.0

Table 6 Forecast error variance decomposition (infinite horizon).

Oil internat. rel. price EA fuel rel. price EA CPI inflation EA gdp growth

Tech

Mon pol

Dem

Mkps

Foreign dem

Fuel

Oil sup

Oil-spec

Foreign sup

Meas Errors

Total

5.6 5.4 6.2 30.1

0.0 0.0 10.7 2.2

0.5 0.4 4.3 18.1

3.2 6.2 64.0 25.7

44.8 39.8 5.7 2.3

0.5 5.7 0.1 0.2

15.0 14.3 1.3 0.1

29.3 27.9 3.3 0.3

1.0 0.3 2.0 17.4

0.0 0.0 2.3 3.8

100.0 100.0 100.0 100.0

1.00

0.47 Tech Mon pol Dem Mkps Foreign dem Fuel Oil sup Oil−specific Foreign sup Total

0.00

−0.50

−1.00

04Q1

05Q1

06Q1

07Q1

08Q1

09Q1

10Q1

11Q1

12Q1

Fig. 8. Historical decomposition of EA CPI inflation rate (q/q). Each colored bar shows how the corresponding group of shocks contributes to the absolute deviation from steady state (expressed in percentage points) of the quarterly inflation rate in a given quarter. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

310

L. Forni et al. / Journal of Macroeconomics 46 (2015) 295–314

1.00

0.49

0.00

Tech Mon pol Dem Mkps Foreign dem Fuel Oil sup Oil−specific Foreign sup Total

−0.50

−1.00

−1.50

−2.00

−2.50

−3.00 04Q1

05Q1

06Q1

07Q1

08Q1

09Q1

10Q1

11Q1

12Q1

Fig. 9. Historical decomposition of EA GDP growth rate (q/q). Each colored bar shows how the corresponding group of shocks contributes to the absolute deviation from steady state (expressed in percentage points) of the GDP growth rate in a given quarter. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

4.4. Historical decomposition We next show the historical decomposition (generated at the posterior mode) of international relative price of oil, EA GDP growth and EA CPI inflation over the 2004–2012 period. We group the shocks in the same way as for the forecast error variance decomposition.24 Fig. 7 shows the international relative price of oil. The steady increase over the 2004–08 period is attributed to a large and growing contribution of RW (Foreign) demand shocks and, only up to 2006, to the contribution of oil-specific demand shocks. After reaching a historical peak in the first half of 2008, the oil price sharply falls in 2009, reflecting a sizeable contraction in RW aggregate demand. It starts growing again in 2010 and 2011, driven by both oil supply and oil-specific demand shocks. Results are in line with Badel and McGillicuddy (2015). They find that during the 2008 financial crisis, negative oil-specific demand shocks together with negative aggregate demand shocks caused a sharp decline in the price of oil. Following the financial crisis, oil-specific demand shocks have been positive. Fig. 8 reports the EA CPI inflation rate. Given the quick pass-through of oil prices into fuel and CPI inflation, oil-related shocks give contributions similar to those observed for the international relative price of oil. Over the 2004–08 period, the contributions of the RW demand and oil-specific demand shocks are mainly positive. The drop in 2009 partly reflects the drop in oil prices due to the fall in RW demand. From 2010, oil supply and oil-specific demand provide positive contributions, by favoring the increase in oil prices. Fig. 9 reports EA GDP growth rate. The prolonged growth between 2005 and 2008 can be mainly attributed to domestic shocks. RW aggregate demand helps explain the sharp decline in GDP growth observed since the second half of 2008, while oil supply and oil-specific demand shocks do not seem to play a relevant role. GDP growth subsequently under-performs in 2011 and 2012, reflecting a weakening domestic demand and an increase in oil prices, as testified by the non-negligible negative contribution of oil shocks. Overall, the analysis suggests that the increase in the oil price between 2004–2008 period did not have stagflationary effects on the EA economy, as it was associated with expansionary RW demand shocks. Similarly, the oil price drop in 2009 did not 24

We do not report neither the contributions of the initial conditions nor those of the ME as we focus on the contributions of structural shocks.

L. Forni et al. / Journal of Macroeconomics 46 (2015) 295–314

311

Oil price

CPI inflation rate

10

0.3 0.2

5

0.1 0

0 0

5

10

15

20

−0.1 0

5

0.2

0

0.15

−0.05

0.1

−0.1

0.05 5

10

15

20

15

20

Real exchange rate

Output 0.05

−0.15 0

10

15

20

0 0

5

10

Nonoil trade balance

Trade balance 0.1

0.06

0

0.04

−0.1 0.02

−0.2 −0.3 0

5

10 Benchmark

15

20 Max. effect

0 0 Min. effect

5

10

15

20

Estimated

Fig. 10. Sensitivity analysis. Responses of EA variables to a negative oil supply shock for alternative values of oil-related parameters. Horizontal axis: quarters. Vertical axis: percentage deviations from the baseline, except for inflation and interest rates (annualized percentage-point deviations), and the trade balance (ratio to GDP, percentage-point deviations). GDP (labeled as “Output” in the charts) and its components are reported in real terms. Real GDP corresponds to real value added.

have expansionary effects as it was due to a drop in RW demand. Finally, the recent increase in oil prices can have stagflationary effects as it is due to positive oil-specific demand shocks. 4.5. Sensitivity analysis We have evaluated the performance of the model under alternative assumptions on the values of the two oil-related Home (EA) parameters, γ 1 and γ 2 , which are crucial for the transmission of oil shocks. Following a given increase in oil prices, firms reduce energy demand. Eq. (7) suggests that the larger the value of γ 1 , the larger the implied reduction in capacity utilization and, thus, in economic activity. If the decrease in capacity utilization implies a larger decrease in the depreciation rate (relatively small value of γ 2 in Eq. (8)), then firms have a larger incentive to reduce capacity utilization and to decrease investment in physical capital to smooth consumption, as the existing stock of capital is affected to a lower extent by the oil price shock. By the same token, decreasing γ 1 and increasing γ 2 imply a smaller transmission of the same shock. In order to quantify these effects, the model has been estimated by alternatively: (i) calibrating γ 1 and γ 2 to values respectively 10% larger and 10% smaller than in the benchmark calibration, so as to magnify the transmission of the oil price shocks (“Max. effects”); (ii) calibrating γ 1 and γ 2 to values respectively 10% smaller and 10% larger than in the benchmark calibration, so as to limit the transmission of the oil price shocks (“Min. effects”). Finally, γ 1 and γ 2 have also been estimated.25 Results are reported in Table 7. The largest log-likelihood (in absolute terms) is obtained when parameters γ 1 and γ 2 are calibrated to obtain relatively low effects of the oil price shocks. The log-likehood value in the benchmark case is closer to these two values than the one corresponding to the “Max. effects” case. When estimated, the values of γ 1 and γ 2 are between those of 25 The prior distribution for γ 1 and γ 2 is a gamma. In the estimation process, we have transformed the parameters so that they are bounded below by zero. In particular, their prior means are set to 0.25 and 0.16, respectively, and their standard deviations are set to 0.05.

312

L. Forni et al. / Journal of Macroeconomics 46 (2015) 295–314

CPI inflation rate

Oil price 10

0.3 0.2

5

0.1 0

0 0

5

10

−0.1 0

20

15

5

Output 0.2

−0.1

0.15

−0.2

0.1

−0.3

0.05 5

10

15

20

15

20

Real exchange rate

0

−0.4 0

10

0 0

20

15

Trade balance

5

10

Nonoil trade balance

0.1

0.1

0 −0.1

0.05

−0.2 −0.3 0

5

10

15

Benchmark

0 0

20 Max. effect

5

10

15

20

Estimated

Min. effect

Fig. 11. Sensitivity analysis. Responses of EA variables to a positive oil-specific demand shock for alternative values of oil-related parameters. Horizontal axis: quarters. Vertical axis: percentage deviations from the baseline, except for inflation and interest rates (annualized percentage-point deviations), and the trade balance (ratio to GDP, percentage-point deviations). GDP (labeled as “Output” in the charts) and its components are reported in real terms. Real GDP corresponds to real value added. Table 7 Sensitivity analysis: oil-related parameters (Home).

γ1 γ2 Log-likelihood

Benchmark

Max. effects

Min. effects

Estimated

1.25 1.16 −4217.16

1.12 1.27 −4205.1

1.37 1.04 −4222.95

1.31 1.13 −4220.82

the benchmark case and those of the “Min. effects” case. To assess the extent to which the alternative considered values of the parameters affect the macroeconomic effects of the oil shocks, Figs. 10–12 report the responses of the EA main macroeconomic variables to the three oil-related shocks. In all cases, including the one in which the two parameters are estimated, the responses are qualitatively and quantitatively similar. Responses are, as expected, larger in the “Max. effects” case than in the other cases, but not extremely so. Overall, sensitivity analysis suggests that our main results do not greatly change when considering alternative values of γ 1 and γ 2 , and in particular when the latter two parameters are estimated. 5. Conclusions In this paper we have empirically analyzed the macroeconomic effects of oil price shocks on the EA economy by estimating a two-country DSGE model where the oil price is endogenously determined by oil demand and supply shocks in the global oil market. According to our results, EA GDP and CPI inflation increase, and the trade balance shows a surplus, when the higher oil price is due to an increase in RW aggregate demand. To the opposite, global oil supply shocks and oil-specific demand shock induce stagflationary effects on the EA economy and have a negative effect on the trade balance. Overall, and in line with Kilian et al. (2009), results point out the need of identifying the (demand and supply) shocks that drive the changes in oil prices, to fully and correctly assess the impact of these changes on macroeconomic variables.

L. Forni et al. / Journal of Macroeconomics 46 (2015) 295–314

313

Oil price

CPI inflation rate

10

0.3 0.2

5

0.1 0

0 0

5

10

15

20

−0.1 0

5

0.6

0.1

0.4

0

0.2

−0.1

0 5

10

15

20

−0.2 0

0.4

0.1

0.3

0.05

0.2

0

0.1 5

10 Benchmark

20

5

10

15

20

Nonoil trade balance

Trade balance 0.15

−0.05 0

15

Real exchange rate

Output 0.2

−0.2 0

10

15

20 Max. effect

0 0 Min. effect

5

10

15

20

Estimated

Fig. 12. Sensitivity analysis. Responses of EA variables to a positive RW aggregate demand shock for alternative values of oil-related parameters. Horizontal axis: quarters. Vertical axis: percentage deviations from the baseline, except for inflation and interest rates (annualized percentage-point deviations), and the trade balance (ratio to GDP, percentage-point deviations). GDP (labeled as “Output” in the charts) and its components are reported in real terms. Real GDP corresponds to real value added.

Our contribution can be improved along several dimensions. First, by inserting microfoundations of the supply of oil, as done in Nakov and Pescatori (2009, 2010), that endogenize OPEC decisions. On a different but complementary route, taking into account the role of oil inventories could be relevant for clearly distinguishing between precautionary demand and supply. Also, we have not fully specified the refining and distribution margins. Hence, we cannot fully capture the role of refining and distribution sector for the propagation of oil supply and demand shocks. Moreover, we do not explicitly formalize value added and excise taxes, that could have relevant implications for relative prices and, hence, welfare and optimal policy. We leave these issues for future research. Acknowledgments We thank R. Alquist, V. Di Nino, A. Locarno, R. Mendes, P. Pagano, A. Pescatori, F. Venditti, two anonymous referees and participants at 2009 Bank of Canada Workshop “Understanding Economic Outcomes in Uncertain Times”, 2010 Bank of Philippines Central Bank Macroeconomic Modeling Workshop, 2012 European Economic Association Annual Congress, 2013 International Conference on Computing in Economics and Finance, for useful suggestions. We thank F. Coluzzi for excellent research assistance. This is a revised version of the paper previously circulated under the title “Euro area and global oil shocks: an empirical model-based analysis”. All errors are ours. Usual disclaimers hold. Supplementary material Supplementary material associated with this article can be found, in the online version, at doi:http://dx.doi.org/10.1016/j. jmacro.2015.09.010.

314

L. Forni et al. / Journal of Macroeconomics 46 (2015) 295–314

References Adolfson, M., Laséen, S., Lindé, J., Villani, M., 2007. Bayesian estimation of an open economy DSGE model with incomplete pass-through. J. Int. Econ. 72, 481–511. Badel, A., McGillicuddy, J., 2015. Oil prices: is supply or demand behind the slump? Economic Synopses, Federal Reserve Bank of St. Louis (8). Bodenstein, M., Erceg, C.J., Guerrieri, L., 2011. Oil shocks and external adjustment. J. Int. Econ. 83 (2), 168–184. Calvo, G., 1983. Staggered prices in a utility maximizing framework. J. Monet. Econ 12, 383–398. Christoffel, K., Coenen, G., Warne, A., 2008. The new area-wide model of the euro area a micro-founded open-economy model for forecasting and policy analysis ECB Working Paper no. 944. European Central Bank, 2010. Structural Issues Report on Energy Markets and the Euro Area Macroeconomy. European Central Bank. Fagan, G., Henry, J., Mestre, R., 2005. An area-wide model for the euro area. Econ. Model. 22 (1), 39–59. Finn, M.G., 1995. Variance properties of solow’s productivity residual and their cyclical implications. J. Econ. Dyn. Control 19, 1249–1282. Finn, M.G., 2000. Perfect competition and the effects of energy price increases on economic activity. J. Money Credit Bank 32, 400–416. Jacquinot, P., Kuismanen, M., Mestre, R., Spitzer, M., 2009. An assessment of the inflationary impact of oil shocks in the euro area. Energy J. 30 (1), 49–84. Katayama, M., 2013. Declining effects of oil price shocks. J. Money Credit Bank 45 (6), 977–1016. Kilian, L., Rebucci, A., Spatafora, N., 2009. Oil shocks and external balances. J. Int. Econ. 77, 181–194. Leduc, S., Sill, K., 2004. A quantitative analysis of oil-price shocks, systematic monetary policy, and economic downturns. J. Monet. Econ. 51 (4), 781–808. Lubik, T., Schorfheide, F., 2005. A Bayesian look at new open economy macroeconomics. In: Gertler, M., Rogoff, K. (Eds.), In: NBER Macroeconomics Annual. MIT Press. Nakov, A., Pescatori, A., 2009. Oil and the great moderation. Econ. J. 120 (543), 131–156. Nakov, A., Pescatori, A., 2010. Monetary policy trade-offs with a dominant oil producer. J. Money Credit Bank 42 (1), 1–32. Peersman, G., Van Robays, I., 2009. Oil and the euro area economy. Econ. Policy 24 (60), 603–651. Rotemberg, J.J., 1982. Monopolistic price adjustment and aggregate output. Rev. Econ. Stud. 49, 517–531. Smets, F., Wouters, R., 2003. An estimated dynamic stochastic general equilibrium model of the euro area. J. Eur. Econ. Assoc. 1 (5), 1123–1175. Smets, F., Wouters, R., 2005. Comparing shocks and frictions in us and euro area business cycles: a Bayesian DSGE approach. J. Appl. Econom. 20 (2), 161–183.