Does globalization alter the monetary transmission mechanism?∗

Tobias Cwik Goethe University Frankfurt

Gernot J. Müller Goethe University Frankfurt

Maik Wolters Goethe University Frankfurt April 2, 2008

Abstract In this paper we quantify effects of globalization - measured by increased goods markets integration - on the monetary transmission mechanism within a two country general equilibrium model. The model is specified to give a quantitatively realistic account of the actual transmission mechanism. Most notably, it features strategic complementarities in price setting such that relative import prices directly impact on domestic inflation. In a first step, we pin down parameter values of the model by matching empirical impulse responses to a monetary policy shock. We compute the latter running vector autoregressions on U.S. time series and an aggregate of industrialized countries. We find that the model is able to mimic the empirical impulse responses quite closely for plausible parameter estimates. In a second step, we use the model to perform counterfactual experiments in order to assess possible changes in the transmission mechanism induced by globalization.

∗

Keywords:

International monetary policy transmission, Globalization, Goods market integration, Openness, Pricing to market

JEL-Codes:

F41, F42, E52

First draft. Comments welcome. We thank Zeno Enders and Keith Kuester for helpful discussions. The usual disclaimer applies. Please address correspondence to [email protected], [email protected] or [email protected]

1 Introduction The consequences of globalization for monetary policy are widely discussed both in academia and among policy makers. In this discussion, globalization is understood as a trend towards increased international integration of goods and financial markets. Several observers have pointed to direct effects of this trend on domestic prices in major industrialized countries, notably disinflationary tendencies due to increased global capacity and competition, see, e.g., Mishkin (2007). In addition, globalization may alter the dynamics of the inflation process which are frequently analyzed on the basis of a Phillips curve relationship, see e.g., Sbordone (2007) and Guerrieri, Gust, and López-Salido (2008). To the extent that globalization changes this relationship at a fundamental level, it alters the way in which monetary policy influences inflation and output, i.e. the monetary transmission mechanism.1 While such changes do not necessarily reduce central banks’ control over the economy, they may nevertheless require, as Yellen (2006) puts it, “some recalibration of policy responses.” In this paper, we try to quantify changes in the monetary transmission mechanism which are likely consequences of further globalization. To do so, we abstract from the effects of increasing financial integration and focus on two specific aspects of increasing goods market integration. First, we consider increasing market integration reflected in increased openness of an economy measured by the average import-to-GDP ratio. If openness is related to ‘home bias’ in preferences and globalization induces an alignment of preferences, as discussed in Corsetti, Meier, and Müller (2007), economies are likely to become more open in due course.2 Second, we consider increasing goods market integration resulting from a declining ability of firms to price discriminate across markets (‘local currency pricing’ or ‘pricing to market’). As pricing to market requires barriers preventing spatial arbitrage, further integration is likely to reduce these barriers, see Senay (1998). We study the consequences of both changes for the international monetary transmission mechanism within a two country dynamic stochastic general equilibrium (DSGE) model of the global economy. In each country a continuum of imperfectly competitive producers specializes in the production of intermediate goods. Monetary policy affects the real economy, because firms are adjusting prices infrequently. Global goods markets are incompletely integrated, because i) a fraction of intermediate goods producers is able to price discriminate across markets and invoices its exports in buyer’s currency; ii) final goods production is biased towards domestic intermediate goods such that the average import-to-GDP ratio falls short of 50 percent which would be observed in the absence of home bias given that countries are symmetric and of equal size. 1

In contrast, the notion that financial globalization via increased capital flows seriously impedes a central bank’s control over interest rates is rejected by most observers, see Bernanke (2007). 2 In the limiting case of full globalization, consumption bundles may be identical across countries, see Corsetti et al. (2007). Alternatively, openness may increase, because trade costs fall. While we assume no trade costs throughout, our results are likely to hold in a setup where trade costs rather than home bias is reduced, see Obstfeld and Rogoff (2000a).

2

As in Gust, Leduc, and Vigfusson (2006), we assume that final goods are assembled on the basis of an aggregation technology which induces demand functions for intermediate goods to display a nonconstant elasticity of substitution. This property induces strategic complementarities in price setting of intermediate goods producers and determines the slope of the Phillips curve for any given level of nominal rigidities. Moreover, as a result of these strategic complementarities import prices become an important factor in the pricing decisions of domestic firms and the more so, the more open an economy. As a consequence, increasing openness is bound to alter inflation dynamics.3 In addition, we assume that prices, if reoptimized, are predetermined relative to monetary policy innovations. Moreover, we assume that private expenditures for consumption and investment are predetermined as well. In both countries monetary policy is characterized by an interest rate feedback rule whereby the nominal short term interest rate is adjusted in response to producer price inflation (PPI) and domestic absorption. Given that we are interested in quantifying the effects which increased globalization exerts on the monetary transmission mechanism, we require our DSGE model to give a quantitatively realistic account of the actual transmission process. As a benchmark, we therefore compute impulse responses to a monetary policy shock within an estimated vector autoregression (VAR) model. Specifically, we use quarterly time series data for the U.S. relative to an aggregate of industrialized countries for the period 1973 to date. Our VAR model includes consumption, investment, PPI-inflation, a short term interest rate, consumer price inflation (CPI) and net export. To identify monetary policy shocks, we assume that - in line with the predictions of our DSGE model - consumption, investment and PPIinflation are predetermined relative to short term interest rates, while consumer price inflation and net exports are allowed to respond to monetary policy innovations. In other words, we employ a recursive identification scheme such that monetary policy shocks are innovations to short term interest rates not accounted for by current realizations of consumption, investment and PPI-inflation as well as by past values of all the other variables included in the VAR model. We treat the impulse responses to monetary policy shocks as a characterization of the actual monetary transmission mechanism. We estimate the structural parameters of the DSGE model employing the minimum distance estimation strategy suggested by Rotemberg and Woodford (1997) and Christiano, Eichenbaum, and Evans (2005). Specifically, we find values for the key parameters of the DSGE model by matching the implied impulse response functions to those obtained from the VAR model. We find that our model, evaluated at the parameter estimates, is able to reproduce quite well the shape and magnitude of the empirical impulse responses. For this to be the case, the model requires parameters to take values 3 Sbordone (2007) focuses on the effects of such a technology on the slope of the New Keynesian Phillips curve and finds that increasing the number of traded goods has a non-monotonic effect on the slope of the Phillips curve. Guerrieri et al. (2008) also focus on the New Keynesian Phillips curve and estimate it on the basis of single equation techniques. They find that incorporating this mechanism improves its empirical performance considerably.

3

which are plausible given the evidence reported by other studies. Specifically, in line with results reported by Bergin (2006) our estimates suggest that local currency pricing (LCP) is pervasive. Similarly, we find evidence that demand functions for intermediate goods are strongly curved and thus quite distinct from what one would obtain under the special case of a constant elasticity of substitution. Given that the model provides an empirically successful account of the actual transmission mechanism, it is well suited to counterfactual experiments. We study how the transmission of an exogenous increase in the short term interest rate changes given an increase in openness and a decrease in the fraction of firms engaged in LCP. Specifically, we consider an increase in the import-to-GDP ratio from 12 percent, which corresponds to the average value for the U.S. in our sample period, to 30 percent, which constitutes a large, yet not implausible increase. The fraction of LCP firms which we estimate to be close to 90 percent is assumed to decline to 60 percent. We find that increasing openness only has little bearing on the effects of the monetary innovation on domestic demand and inflation. This is the result of a high LCP share in the baseline scenario, which isolates import prices from the exchange rate movements triggered by the monetary impulse. Apart from preventing the expenditure switching mechanism from operating, the pervasiveness of LCP also limits the possible effect of relative import prices on inflation dynamics. Consequently, it is only by simultaneously increasing openness and lowering the extent of LCP, that we find fairly strong changes in the monetary transmission mechanism. We find that the effect of the policy shock on domestic absorption is reduced by about 15 percent, while the response of producer price inflation is increased by about 25 percent. In a last experiment, we turn to the recalibration of monetary policy, which would be necessary if monetary policy were to have the same effect on domestic output and inflation in the counterfactual scenarios as in the baseline estimation. Our tentative results point to considerable changes either in the systematic components of monetary policy or the size of the monetary impulse. The remainder of this paper is organized as follows. In section 2 we introduce the details of the model economy. Section 3 presents time series evidence from the estimated VAR model and discusses the estimation of the DSGE model. In section 4 we simulate the effects of globalization on the basis of model-based counterfactual experiments. Section 5 offers a brief conclusion.

2 Model In this section we suggest a two country business cycle model to study monetary policy transmission in open economies. Most of the model features are standard. We assume that in each country there is a continuum of intermediate good producers operating under monopolistic competition and being constrained in price setting à la Calvo. We assume that a fraction of these firms invoice exports in 4

domestic currency (PCP-firms), while the remaining firms invoice exports in foreign currency (LCPfirms). There is a representative household in each country owning the capital stock which is rented together with labor services to intermediate goods producers on a period-by-period basis. Adjusting investment is costly. International financial markets are assumed to be complete.4 In each country final goods firms assemble domestic and imported goods to provide final goods which are used for consumption and investment. At the final good level we assume an aggregation technology which induces home bias in the composition of final goods and a non-constant elasticity of substitution (NCES) in the demand for intermediate goods.5 This technology has recently been put forward by Gust et al. (2006), Sbordone (2007) and Guerrieri et al. (2008) in order to introduce strategic complementarities in price setting arising from the degree of competition in the intermediate goods markets. While these authors stress the importance of this channel in accounting for the consequences of the ongoing globalization process and its impact on inflation dynamics, its implications for monetary policy transmission have not yet been analyzed within a sticky price two country general equilibrium framework. Our model is suited to perform this task. It is outlined in the following by focusing, in turn, on the problems of final goods firms, intermediate good firms and the representative household. We close the model with a characterization of monetary policy in terms of an interest feedback rule. As both countries i ∈ {1, 2} are symmetric, of equal size, and have isomorphic structures, our exposition focuses on country 1, i.e. the ‘home’ country.

2.1 Final Good Firms Let F1t denote final goods in country 1 produced at time t and used for private consumption, investment and government consumption (domestic absorption), i.e. F1t = C1t + X1t + G1t . Final goods firms are perfectly competitive. The problem of a representative final goods firm is to assemble domestically produced intermediate inputs, A1t (j), as well as imported intermediate goods, B1t (j), to produce a given amount of final goods. These inputs are produced by a continuum of intermediate A (j) denote the price in country 1 of goods firms in each country; we assume j ∈ [0, 1]. Letting P1t B (j) denote the price in country 1 of a generic good a generic good produced in country 1 and let P1t

produced in country 2, the final goods firm’s problem is given by Z 1 Z 1 A B min P1t (j)A1t (j)dj + P1t (j)B1t (j)dj 0

(1)

0

4

In setting up the model we draw on earlier work by Chari, Kehoe, and McGrattan (2002), Kollmann (2002), Galí and Monacelli (2005), Corsetti and Pesenti (2005), Bergin (2006) and Schmidt (2006), among others. 5 The second feature is in contrast with earlier open macro models which often employ an Armingtion aggregator as, for instance, in Backus, Kehoe, and Kydland (1994).

5

subject to

· σ h σ−1 σ−1 i σ−1 σ σ Vdt + Vmt −

¸ 1 − 1 = 1, (1 + η)υ

(2)

where Vdt is an aggregate of domestically produced individual goods and Vmt an aggregate of imported individual goods. These aggregates, in turn, are defined as follows ¸υ · Z 1 σ 1 (1 + η) A1t (j) σ−1 − η dj, Vdt = ω (1 + η)υ ω F1t 0 ¸υ · Z 1 σ 1 (1 + η) B1t (j) − η dj. Vmt = (1 − ω) σ−1 (1 + η)υ (1 − ω) F1t 0

(3)

(4)

This structure defined by (2), (3) and (4) specifies the technology available to the final goods firm and follows Gust et al. (2006).6 A few remarks concerning the key parameters are in order. The trade price elasticity, i.e. the elasticity which measures the extent of substitution from goods produced at home to those produced abroad for a given change in relative prices, is a key parameter for the international transmission mechanism. In our setup it is a function of several parameters and given by σ ˜=

−σ . (σ(υ − 1) − υ)(1 + η)

(5)

The elasticity of substitution between goods produced within the same country is time varying. In the steady state this elasticity is given by ²=

1 1 . 1−υ1+η

(6)

Note that the parameter η plays a crucial role for both elasticities. It provides a measure of how strongly our setup deviates from the special case where the elasticity of substitution is constant (CES).7 Finally, the parameter ω is often referred to as providing a measure for ‘home bias’ as it measures the steady state weight of domestically produced goods in final goods. 1 − ω mesures the import-to-GDP ratio in steady state. Solving equation (1) gives rise to domestic demand functions for domestically produced intermediate goods 1 AD 1t (j) = ω 1+η

"µ

A (j) P1t A P1t

1 ¶ υ−1 µ

A P1t Γ1t

#

σ ¶ σ(υ−1)−υ

+ η F1t .

(7)

In the same manner, final goods producers in the foreign country minimize expenditures, which in turn implies the following demand function for domestically produced intermediate goods in the foreign country AD 2t (j) 6 7

1 = (1 − ω) 1+η

"µ

A (j) P2t A P2t

1 ¶ υ−1 µ

A P2t Γ2t

σ ¶ σ(υ−1)−υ

# + η F1t ,

(8)

The original closed economy formulation goes back to Dotsey and King (2005) or more generally to Kimball (1995). The CES case is nested in our setup for η = 0.

6

where Γit are price indices defined below. Note that as in Dotsey and King (2005) and the corresponding open economy model of Gust et al. (2006) the demand curves in equations (7) and (8) include a linear term if η 6= 0 implying the demand elasticities are not constant.8 Global demand for a generic good j produced in country 1 is given by D Y1tD (j) = AD 1t (j) + A2t (j).

(9)

To understand the price indices implicitly defined by the cost minimization problem of the final goods firm, note that pricing behavior at the intermediate good level will be different depending on whether A,LCP (j) denote the price in country 1 of intermediate goods firms engage in LCP or PCP. Let P1t A,P CP a generic good produced in country 1 by a LCP-firm and P1t (j) the price set by a PCP-firm.

Letting α measure the fraction of LCP-firms and (1 − α) the fraction of PCP-firms, then the producer A and the import price index P B in the home country are given by price index (PPI) P1t 1t

µZ A P1t

= 0

and

µZ B P1t

α

= 0

α

Z υ A,LCP P1t (j) υ−1 dj

+ α

Z υ B,LCP P1t (j) υ−1 dj

1

1

+ α

υ A,P CP P1t (j) υ−1 dj

υ B,P CP P1t (j) υ−1 dj

¶ υ−1 υ

(10)

¶ υ−1 υ ,

respectively. Cost minimization also implies the following price index for the final good µZ α ¶ Z 1 1 η A,LCP A,P CP F P1t = Γ1t + ω P1t (j)dj + P1t (j)dj 1+η 1+η 0 α µZ α ¶ Z 1 η B,LCP B,P CP + (1 − ω) P1t (j)dj + P1t (j)dj , 1+η 0 α

(11)

(12) (13)

where Γ1t is given by h i σ(υ−1)−υ (σ−1)υ (σ−1)υ (σ−1)υ A σ(υ−1)−υ B σ(υ−1)−υ . Γ1t = ω(P1t ) + (1 − ω)(P1t )

(14)

Letting St denote the nominal exchange rate (the price of home currency in terms of foreign currency) and assuming that the law of one price holds for PCP-firms, we have B,P CP B,P CP P1t (j) = St P2t (j);

A,P CP A,P CP P1t (j) = St P2t (j).

(15)

2.2 Intermediate Goods Firms In each country, there is a continuum of firms each of which produces a unique intermediate good and engages in monopolistic competition. Production is Cobb-Douglas: Y1t (j) = K1t (j)θ H1t (j)1−θ ,

(16)

8 As a result, the optimal markup set by intermediate good firms will be time-varying. These strategic complementarities have import implications for inflation dynamics, see Sbordone (2007).

7

where H1t (j) and K1t (j) respectively denote labor and capital employed by firm j . Let W1t and R1t denote the nominal wage rate and the rental rate of capital, respectively. Minimizing the costs of

producing intermediate goods implies for (nominal) marginal costs M C1t (j) =

W1t H1t (j) R1t K1t (j) = , (1 − θ)Y1t (j) θY1t (j)

(17)

where marginal costs are independent of the level of production and identical across firms, because both factors of production can be adjusted freely across firms. We assume that price setting is constrained exogenously by a discrete time version of the mechanism suggested by Calvo (1983). Each firm has the opportunity to change its price with a given probability 1−ξ . Moreover, we assume as, for instance, in Christiano et al. (2005) that when a firm has the oppor-

tunity, it sets the new price in order to maximize the expected discounted value of net profits before the realization of shocks in a given period, i.e. period t prices are set conditional on the information available in period t − 1. Firms that do not reoptimize in a certain period index their price to last period’s producer price inflation, where the degree of indexation is given by the parameter κ ∈ [0, 1]. A,P CP In setting the new price P1t (j), the problem of a generic intermediate good PCP-firm j in country

1 is given by max

∞ X

à ξ k Et−1

" A,P CP P1t (j)

D Qt,t+1 Y1t+k (j)

!

#

k Y

F ΠκP P I,1t+s−1 − M C1t+k /P1t+k

(18)

s=1

k=0

subject to demand functions defined by (9), the production function (16), the optimality condition on factor inputs (17) as well as the constraint that the demand of intermediate good j is satisfied, i.e. A /P A Y1tD (j) = Y1t (j). ΠP P I,1t = P1t 1t−1 denotes producer price inflation. Profits are discounted with

the stochastic discount factor, Qt,t+1 . A,LCP The pricing problem of a generic intermediate good LCP-firm j is twofold. The firm sets P1t (j)

for the domestic market by solving max

∞ X

" A,LCP ξ k Et−1 Qt,t+1 AD (j) 1t+k (j) P1t

k Y

# F ΠκP P I,1t+s−1 − M C1t+k /P1t+k ,

(19)

s=1

k=0

A,LCP subject to the demand function (7). P2t (j) is set to maximize

max

∞ X

" ξ

k

Et−1 Qt,t+1 AD 2t+k (j)

A,LCP St+k P2t (j)

k Y s=1

k=0

subject to the demand function (8).

8

# ΠκP P I,2t+s−1

F − M C1t+k /P1t+k

(20)

2.3 Households A representative household in country 1 allocates consumption expenditures on final goods, C1t , and supplies labor, H1t , to intermediate goods firms. The objective of the household is given by E−1

∞ X [(C1t − bC1t−1 )µ (1 − H1t )1−µ ]1−γ β , 1−γ

(21)

t=0

where β is a time discount factor and b ∈ [0, 1) measures the extent of consumption habits. The parameter γ measures the degree of risk aversion and the parameter µ measures the weight of consumption in the utility function relative to leisure. We assume that decisions concerning period t are conditional on the information available in period t − 1. Labor and capital are internationally immobile; households in country 1 own the capital stock K1t of that country. We follow Christiano et al. (2005) and assume that it is costly to adjust the level of investment, X1t . Specifically, the law of motion for capital is given by K1t+1 = (1 − δ)K1t + [1 − Ψ(Iit /Iit−1 )]Iit ,

(22)

where δ denotes the depreciation rate; restricting Ψ(1) = Ψ0 (1) = 0 and Ψ00 (1) = χ > 0 ensures that the steady state capital stock is independent of investment adjustment costs captured by χ. A complete set of state-contingent securities is traded at an international level. We let Ξ1t+1 denote the period t + 1 payoff of the portfolio held at the end of period t in units of country 1’s currency. For future reference it is useful to define the gross short-term nominal interest rate, it , given as the solution to the following relationship: i−1 t = Et Qt,t+1 ; in other words, it is the inverse of the nominal value of a security paying one unit of account in each state of the world in period t + 1. The budget constraint of the household reads as follows F W1t H1t + R1t K1t + Υ1t + T1t − P1t (C1t + X1t ) = Et {Qt,t+1 Ξ1t+1 } − Ξ1t ,

(23)

where Υ1t denotes nominal profits earned by monopolistic firms and transferred to households and T1t denotes lump-sum taxes. Because we assume that government spending is financed entirely through F G . The household problem is to maximize (21) subject to the lump-sum taxes, we have T1t = P1t 1t

constraints (22) and (23).

2.4 Monetary Policy To close the model, we assume that monetary policy is characterized by an interest rate feedback rule as in Clarida, Galí, and Gertler (2000). Specifically, we assume for the interest rate µ µ ¶¶ F1t − F1 i1t = ρi1t−1 + (1 − ρ) i1 + β −1 φπ (ΠP P I,1t − ΠP P I,1 ) + 0.25β −1 φy + ν1t (24) F1 where ρ ∈ [0, 1] captures interest rate smoothing, φπ captures the long-run adjustment of the interest rate to producer price inflation and φy captures stabilization of domestic absorption. 9

2.5 Model solution We solve the model numerically by applying standard techniques. Specifically, we linearize the equilibrium conditions of the model, which are given by the constraints and the first order conditions which characterize the problems described above, around a deterministic and symmetric steady state. We assign values to the structural parameter of the model and solve the linear model numerically. We then study the transmission of monetary policy shocks, ν1t . To keep the model tractable, we focus on country differences, i.e. the behavior of a domestic variable relative to its foreign counterpart. Our strategy to assign parameter values is described in the following section.

3 Estimation Throughout the paper we focus on impulse responses to monetary policy shocks to characterize the international monetary transmission mechanism. Our objective is that our model provides a quantitatively realistic account of the transmission mechanism as apparent from the data. To achieve this objective, we proceed in two steps. First, we estimate the VAR model on U.S. time series relative to an aggregate of industrialized countries and compute empirical impulse response functions to a monetary policy shock. In a second step, we pin down the values of key parameters of the DSGE model by matching the implied impulse response functions to those obtained from the VAR model.

3.1 Empirical impulse response functions We estimate the VAR model on quarterly time series for the period 1973-2006. We focus on relative variables, i.e. the difference of a variable in the U.S. and its counterpart for an aggregate of industrialized countries, which we treat as the rest of the world (‘ROW’ for short), see also Clarida and Gali (1994) and Rogers (1999).9 This focus allows us to economize on the degrees of freedom, but is also convenient given that we focus on country differences in the DSGE model as well. Specifically, we consider the log of relative consumption, the log of relative investment, the difference in PPI-based inflation rates, the difference in short term interest rates, the difference in CPI-based inflation rates as well as the trade balance for the U.S., where the trade balance is defined as the log difference in deflated exports and imports. Letting Yt denote the vector of endogenous variables, we estimate the structural model A(L)Yt = εt ,

where A(L) =

P4

i i=0 Ai L , LYt

(25)

= Yt−1 and E(εt ε0t ) = I . To achieve identification, we assume

that A0 is lower triangular, i.e. we impose the recursive identification which is frequently employed 9

We provide a detailed description of the data in the appendix. We remove a constant linear trend from consumption and investment before computing relative variables.

10

to study the effects of monetary policy shocks, see Kim (2001) for an open economy context. As we only attach a structural interpretation to the innovation in relative short term interest rates, what matters for identification is how the other variables in Yt are ordered relative to this variable, see Christiano, Eichenbaum, and Evans (1999). We order relative consumption, relative investment as well as the inflation differential based on PPI-inflation before and the inflation differential based on CPI-inflation and net exports after the short term interest rate differential. Given this ordering, our identification assumptions are consistent with the implications of the general equilibrium model outline above. Consumption, investment and therefore domestic absorption as well as producer prices are predetermined relative to monetary policy shocks, while consumer prices and the trade balance are free to adjust immediately. Implicitly, as in the theoretical model, we are thus assuming that monetary policy adjusts the short term interest rate in response to contemporaneous changes in producer price inflation and domestic absorption, but not in response to contemporaneous changes in CPI-inflation or the trade balance.10 Figure 1 displays the impulse responses to a (relative) monetary policy shock, i.e. an increase by 100 basis points in U.S. short rates relative to the aggregate of industrialized countries. The solid line shows the point estimate. The shaded area shows 90 percent confidence bounds obtained from bootstrap sampling. The upper row shows the responses of consumption and investment in relative terms; for both we find a protracted and hump-shaped decline. While consumption falls by about 0.3 percent, investment falls about 1.25 percent, with the maximum effect occurring between three to six quarters after the shock. Producer price inflation (in relative terms) responds somewhat sluggishly; the maximum decline of about 8 basis points is observed five quarters after the shock. According to our point estimate, it takes another 3 to 4 years for inflation to return to its pre-shock level. The shock to the short rate is somewhat persistent and the short rate returns to its pre-shock level after about one year. The response of consumer price inflation is remarkably close to that of producer price inflation, both from a quantitative and a qualitative point of view. Finally, the real net exports, measured as the log difference of real exports and imports for the U.S. displays a hump-shaped increase with the maximum effect of about 0.2 percent occurring after about a year. 10 Alternative approaches to identify monetary policy shocks in open economy frameworks focus on monetary aggregates and non-recursive identification schemes, see Eichenbaum and Evans (1995), Cushman and Zha (1997) and Kim and Roubini (2000). More recently, Faust and Rogers (2003) and Scholl and Uhlig (2007) use sign restrictions to achieve identification. A common focus of this literature is the behavior of the real exchange rate in the face of monetary policy shocks and how important are the latter to account for fluctuations in the former. In the present paper, we are not taking up these issues. Instead, we use the VAR responses as a key statistic to pin down parameter values of our general equilibrium model.

11

Consumption

Investment

0.5

3 2 1

0

0 −1

−0.5

−2 −3 −4

−1 0

5

10

15

20

0

5

PPI-inflation

10

15

20

Short rate 2

0.2

1.5 0.1 1 0 0.5 −0.1 0 −0.2

−0.5 0

5

10

15

20

0

CPI-inflation

5

10

15

20

15

20

Real net exports

0.2 0.6 0.1 0.4 0 0.2 −0.1 0 −0.2 −0.2 0

5

10

15

20

0

5

10

Figure 1: Effect of monetary policy shock. Notes: Responses are in relative terms (U.S. vs. ROW), except for net exports which is the log difference of U.S. exports and imports. Solid line: point estimate; shaded areas: bootstrapped 90 percent confidence intervals; dashed-dotted line: responses of estimated general equilibrium model; Vertical axes: percent, except for inflation and interest rate (percentage points). Horizontal axes: quarters.

3.2 Estimation of general equilibrium model The second step of the analysis consists in matching empirical (VAR) and theoretical impulse responses in order to obtain estimates for the parameters of our model. This approach has gained popularity in closed economy studies of monetary policy transmission following the pioneering work of Rotemberg and Woodford (1997) and Christiano et al. (2005), see, for instance, Amato and Laubach

12

(2003), Bovin and Giannoni (2006) and Meier and Müller (2006). To illustrate this approach, define Ψe to be the empirical impulse response function characterizing the data. The model itself, in turn, assigns to each admissible vector of structural parameters θ a theoretical impulse response function Ψt = Ψ (θ). The binding function, Ψ(), must be assumed to be injective to ensure identification. We obtain an estimate for the parameter vector of interest, θb, by minimizing the weighted distance between empirical and theoretical impulse response functions, i.e., Ψe and Ψt : θb = arg min (Ψe − Ψ (θ))0 W (Ψe − Ψ (θ)) ,

(26)

where W represents a positive definite weighting matrix. As the relationship between structural parameters and the implied impulse response functions is nonlinear, we obtain theoretical impulse response functions by applying standard numerical techniques. Note that our procedure only admits saddle path stable solution and thus rules out by construction any parameterization of the model which would give rise to equilibrium indeterminacy. Basically, Ψ (θ) is evaluated repeatedly for different parameter vectors θ until the closest fit with the empirical

impulse responses, Ψe , has been obtained. Our choice of the weighting matrix W is guided by the idea of giving greater weight to impulse responses that are more precisely estimated. Thus we opt for the diagonal matrix W diag whose diagonal entries are the reciprocal values of the variance of the empirical impulse responses. Using this weighting matrix ensures that the theoretical impulse responses are made to be as close to the empirical ones as possible, in terms of point-wise standard deviations. Finally, regarding the length of the impulse response functions, we consider 20 quarters starting from the second quarter. Standard errors for θb are computed using the following expression for the asymptotic variance of our estimator, taken from Wooldridge (2002): ³ ´ ¡ ´¡ ¢ ³ ¢ b G G0 W G −1 . [ θb = G0 W G −1 G0 W ΣW Avar

(27)

where G = ∇θ Ψt represents the Jacobian of the impulse response function generated from the model b denotes the bootstrap-estimated variance matrix of the impulse responses. and Σ In practice, given the number of the structural parameters, it is not possible to identify all of them simultaneously. We therefore fix a number of parameters prior to the estimation which are either given by first moments of the data or largely uncontroversial in the literature. First we set ω = 0.88 which implies an import-to-GDP ratio of 12 percent, about the average value for the U.S. in our sample period. We set the discount factor β to 0.99. Following Backus et al. (1994), we assume γ = 2 and µ = 0.34. Moreover, we set the capital share in intermediate goods’ production θ to 0.36. For the

depreciation rate we assume δ = 0.025. We assume that government spending accounts for 20 percent of GDP, again close to the average in our sample period. Regarding price rigidities, we set ξ = 0.75, 13

Parameter σ ˜

Table 1: Estimated parameter values of DSGE model Description Trade price elasticity 0.50

χ

Investment adjustment costs

κ

Price indexation

1.00

φπ

Inflation coefficient in policy rule

1.00

φy

Output coefficient in policy rule

ρ

Interest rate smoothing

b

Habits

α

Share of firms with local currency pricing

η

NCES-parameter

(0.74)

1.01

(0.64) (−)

(0.52)

0.01

(0.14)

0.67

(0.09)

0.89

(0.05)

0.89

(0.15)

−11.11 (15.23)

Notes: Parameter estimates obtained from matching DSGE and VAR impulse response functions; standard errors are reported in parentheses. Those parameter values which have been estimated to be at their bounds have been assumed to take this value prior to estimation; in this case no standard error is reported.

which implies an average duration of prices of one year which is relatively long given the micro evidence discussed in Nakamura and Steinsson (2008), but still lower than most macroeconometric evidence suggests. We set υ such that the markup earned by intermediate goods firms in steady state is 20 percent. Finally, note that we estimate the trade price elasticity, σ ˜ , by adjusting σ .11

3.3 Results Table 1 provides the results of our estimation exercise, i.e. the solution to (26). We find quite plausible point estimates and fairly narrow confidence bounds implied by the standard errors reported in parentheses. The estimated trade price elasticity is below the values often used or found in the literature. Yet, several recent studies suggest that that a low trade price elasticity may help to account for a larger set of macroeconometric observations, see Lubik and Schorfheide (2006), Kollmann (2006) and de Walque, Smets, and Wouters (2005). Also χ, the parameter capturing investment adjustment costs is somewhat below the value reported in Christiano et al. (2005) for U.S. data. This is likely to be the result of the aggregation function of final goods, see the discussion in Backus et al. (1994). In line with earlier research we also find full indexation of prices, see, for instance, Meier and Müller 11

That is in the estimation we impose the following relationships: υ=

η + (1/1.22) σ ˜ υ(1 + η) and σ = . 1+η 1+σ ˜ (υ − 1)(1 + η)

14

10 0 −10 −20 −30 −40 −50 −60 −3

−2

−1

0

1

2

Figure 2: Demand function for intermediate goods. Notes: Solid line: CES case (η = 0). dashed-dotted line: NCES case (η = −11.1); Vertical axes: relative demand in percent; Horizontal axes: relative price in percent.

(2006) for the U.S. Regarding monetary policy we find parameter values which imply a fairly lose monetary stance and a degree of interest rate smoothing which is in line with previous findings in the literature, see, for instance, Clarida et al. (2000) for the U.S. Note that while we do not impose a bound on the parameter φπ , our solution procedure rules out equilibrium indeterminacy by construction. We find a considerable amount of habits in consumption, somewhat above the values reported in Smets and Wouters (2005) both for the Euro area and the U.S. The share of firms engaged in LCP is estimated to be quite high, in line with results reported in Bergin (2006) who estimates this parameter to be at its upper bound.12 Finally, the estimate for the parameter η provides a measure for the curvature of our demand functions. Our estimate is somewhat higher

than the values used in Gust et al. (2006) and Guerrieri et al. (2008). To assess the implication of this estimate in terms of deviation from the standard CES case (η = 0), figure 2 displays the change in demand for a generic good (vertical axis) resulting from a deviation in its relative price (horizontal axis). The solid line shows the induced change in demand in case η = 0, which is constant. The dashed line shows the induced change in demand for our estimate of η . Relative to the CES case, our estimate implies strong non-linearities in the demand functions. If the relative price increases, demand falls considerably more strongly, while, if the relative price falls, demand increases more mildly. This induces strategic complementarities in price setting, which ceteris paribus, provides firms with an incentive to adjust prices by small amounts if given the opportunity to do so. Given the estimated parameter values, we compute the impulse response of the model and compare them to those obtained from the VAR model. The dashed-dotted lines in the panels of figure 1 show 12

The extent of LCP relative to PCP has been the topic of a considerable controversy which is beyond the scope of the present paper, see Betts and Devereux (2000) and Obstfeld and Rogoff (2000b).

15

0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0

1975

1980

1985

1990

1995

2000

2005

Figure 3: Import-to-GDP ratio in U.S. 1973. Notes: Solid line displays nominal imports divided by nominal GDP; dashed-dotted line gives the average value; dotted line gives counterfactual scenario with import share of 30 percent.

that the model responses track the empirical responses quite closely. All the responses are within the confidence bounds of the VAR responses, except for the impact response of CPI-inflation and net exports. Also the theoretical response of investment is somewhat less pronounced than its empirical counterpart. The response of the consumption differential, as well as those of PPI-inflation and the short rate are matched particularly closely. Overall, we conclude that the DSGE model - if evaluated at the point estimates - can quantitatively account for the international monetary transmission mechanism as apparent for the VAR model estimated on time series from 1973-2006. The DSGE model seems thus well suited to study counterfactual outcomes in a more globalized world.

4 Model-based counterfactual analysis We now turn to the question which motivates our investigation, namely, whether and to what extent globalization alters the international monetary transmission mechanism. Given that our two country general equilibrium model has been shown to give an empirically plausible account of the international transmission mechanism, it is well suited to address these questions. Notably, given that the structural model parameters estimated above are fairly ‘deep’, we can perform counterfactual experiments which are not prone to the Lucas critique. As discussed in the introduction we make the notion of globalization operational by considering two specific parameter variations in the model. First, we consider an increase in trade integration resulting in an increase in the import-to-GDP ratio from 12 to 30 percent. We implement this by lowering the ‘home bias’ parameter ω from 0.88 to 0.7.13 Figure 3 displays the evolution of actual U.S. imports 13

We maintain the assumption that both countries are symmetric.

16

relative to GDP for the period 1973-2006. The average value is close to 12 percent and indicated by the dashed-dotted horizontal line. The import share is increasing over time, from about 6 to about 16 percent. Except for the last 10 years, however, the increase has been rather modest - in line with the common perception that globalization has been accelerating only recently. The horizontal dotted line indicates our counterfactual scenario of an import-to-GDP ratio of 30 percent. Given current trends, this corresponds to a rather strong increase in trade integration, but is not utterly implausible. The second variation concerns the extent of pricing to market (LCP). Currently, in order to mimic the transmission mechanism apparent from the VAR impulse responses the model requires that a considerable fraction of intermediate good firms is price discriminating across markets and invoices exports in the currency of the destination market. As argued in the introduction, one manifestation of globalization is an increase in possibilities to arbitrage spatially. Hence, we may expect that fewer firms will be able to price discriminate across markets; put differently, we may expect the fraction of PCP firms to increase. We therefore consider the counterfactual scenario that the fraction of firms which engages in PCP increases from about 10 to 40 percent. We implement this scenario by lowering α accordingly.

4.1 Counterfactual transmission Figure 4 displays the results of our model-based counterfactual experiments. We focus on the response of key variables in country 1 to an increase in domestic interest rates by 100 basis points.14 As a baseline scenario we plot the response of the baseline economy, i.e. the model evaluated at the parameter estimates obtained above (solid line). We consider three counterfactual scenarios. First, we assume that imports account for 30 percent of GDP (dashed line, CF1); second, we assume that 40 percent of firms engage in PCP rather than LCP (dashed-dotted line, CF2). Finally, we assume that both changes occur at the same time (dotted line, CF3). For these counterfactual simulations we keep all the other parameters at the baseline values. The upper row displays the responses of domestic absorption and output. Note that as government spending is constant, the response of domestic absorption is the weighted sum of the responses of consumption and investment. While we did not include output in the VAR model in order to achieve model consistent identification through a recursive scheme, we can now use the estimated model to assess the impact of monetary policy shocks on output as well. In the second row, we display responses of PPI-inflation as well as the response of consumer prices. In the latter case, we focus on the level response, as CPI inflation responses strongly undershoot on impact thereby preventing a clear-cut interpretation of the counterfactual outcomes. In the third row we display the response of 14

While cross country differentials generally display a very similar pattern, focusing on home variables allows to derive quantitative results which presumably matter most from the perspective of domestic policy making.

17

Domestic absorption

Output

0.05

0

0

−0.05

−0.05

−0.1

−0.1

−0.15

−0.15

−0.2

−0.2

−0.25

−0.25

−0.3

−0.3 0

5

10

15

−0.35 0

20

PPI-inflation 0

−0.02

−0.2

−0.04

−0.4

−0.06

−0.6

−0.08

−0.8

5

10

15

−1 0

20

Real exchange rate 0.2

2.5

0

2

−0.2

1.5

−0.4

1

−0.6

0.5

−0.8

0

−1 5

10

15

20

15

5

10

15

20

15

20

Real net exports

3

−0.5 0

10

Consumer price level

0

−0.1 0

5

−1.2 0

20

5

10

Figure 4: Effect of monetary policy shock in theoretical model.

Notes: Impulse response of domestic variables to exogenous increase in short term interest rate (100 bp). Solid line: baseline estimate; dashed line: import-toGDP ratio is 30 percent (CF1); dashed-dotted line: 40 percent PCP (CF2); dotted line: import-to-GDP ratio 30 percent and 40 percent PCP (CF3); Vertical axes: percent, except for inflation and interest rate (percentage points) and real net exports (percentage points of GDP). Horizontal axes: quarters.

the real exchange rate for which an increase indicates an appreciation as well as real net exports.15 From a qualitative point of view, the counterfactual scenarios imply a transmission of monetary policy shocks which is fairly close to the one implied by the estimated model. Yet, important quantitative 15

We do not report the response of the short interest rate, because it is virtually unaffected by the counterfactual experiments.

18

Domestic absorption

Ex ante real interest rate

−0.18

0.5

−0.2

0.45

−0.22 0.4 −0.24 0.35 −0.26 0.3

−0.28 −0.3 0

0.2

0.4

0.6

0.8

1

0.25 0

0.2

0.4

0.6

0.8

1

Figure 5: Peak responses of domestic absorption to monetary policy shock and corresponding response of ex ante real interest rate. Notes: vertical axis measures peak response of domestic absorption (left) and response of ex ante rate rate in third quarter (right); solid line corresponds to import share of 12 percent; dashed line corresponds to import share of 30 percent; dashed dotted line corresponds to import share of 40 percent; horizontal axes measures extent of PCP; dashed vertical line corresponds to estimated extent of PCP; dotted vertical line corresponds to extent of PCP assumed in counterfactuals computed above.

differences can be observed. Considering the first counterfactual, one observes small differences relative to the baseline responses. With the exception of output, which falls less than in the baseline case, increasing trade openness has little bearing on the transmission mechanism. Similarly, if only the extent of LCP is reduced the quantitative changes are contained, except for net exports (CF2). In contrast, under the third counterfactual scenario, we observe considerable changes in the response of domestic absorption; its peak response is reduced by some 15 percent. Also the response of PPIinflation increases by some 25 percent (similarly the consumer price level falls much more strongly). Finally, net exports tend to fall, rather than rise in the counterfactual scenarios (CF2 and CF3). To build up intuition for the mechanisms which bring about these changes and to assess the quantitative differences more systematically, we plot in figure 5 the peak responses of domestic absorption against the increasing fraction of PCP firms (left panel) together with the corresponding response of the ex ante real interest rate. More specifically, on the horizontal axis we increase PCP from zero to one. On the vertical axis we display the peak responses for three different assumptions on the importto-GDP ratio. The solid line corresponds to our baseline case with imports accounting for 12 percent of GDP, while the dashed (dashed-dotted) line displays results obtained assuming 30 (40) percent. We observe that the higher the fraction of PCP firms, the lower the increase in the ex ante real rate. One way to think about this finding is by recalling that the ex ante real interest rate differential corresponds to the expected change in the real exchange rate, see Clarida and Gali (1994). As apparent from figure 4, the real exchange rate appreciates less strongly, the more PCP and the more open the economy. Intuitively, the more open the economy and the more responsive prices are in buyers’ currency, the less strong the response of international relative prices to bring about an reallocation of

19

expenditure. Also the change in the real exchange rate and, hence, the response of the ex ante real rate is smaller in the counterfactuals with more PCP and a higher import-to-GDP ratio. Intuitively, the appreciation lowers the price of imports and thus, ceteris paribus, lowers the price of consumption today relative to tomorrow. Corsetti et al. (2007) provide a lengthy discussion of this mechanism in the context of fiscal policy transmission. As a consequence, monetary policy’s leverage over real interest rates decreases with PCP and the more so, the more open the economy. In other words, it appears that with the ongoing globalization process, the effects of monetary policy on domestic absorption become more contained. A similar consideration serves to rationalize the change in the response of real net exports displayed in the lower right panel of figure 4. In the period of the shock, domestic absorption is predetermined such that any change in real net exports reflects substitution from domestic and foreign goods triggered by an exchange rate appreciation. However, with PCP being pervasive, the exchange rate pass-through and, hence, expenditure switching is limited. Only if PCP increases, one observes substantial changes in the international transmission mechanism as far as the response of net export is concerned.16 A key relationship taht shows how inflation dynamics are triggered by the monetary shock is given by the Phillips curve. It is also pivotal for interpreting the results of our counterfactual analysis. Drawing on earlier work by Guerrieri et al. (2008), we start from the price stetting problem of a generic intermediate good firm and derive a dynamic relationship between PPI-inflation, marginal cost and relative import prices (with first order accuracy expressed in deviations from steady state). In other words, our model gives rise to the following variant of the the new Keynesian Phillips curve ¸ · σ ˜ B (28) πt − κπt−1 = βEt−1 [πt+1 − κπt ] + λEt−1 (1 − Ψ)st + Ψ(1 − ω) qt ² where λ=

(1 − βξ)(1 − ξ) −ηµ and Ψ = . ξ 1 − ηµ

Here πt , st and qtB measure PPI-inflation, real marginal costs and the relative import price relative to their steady state values. As stressed by Guerrieri et al. (2008), if η > 0, i.e. in case the price elasticity of substitution implied by the demand functions (7) is non-constant, relative import prices directly determine inflation in addition to expected future inflation rates and real marginal costs. Intuitively, with η > 0 if imported goods are relatively cheap those domestic firms which are able to adjust their prices will tend to reduce them because of strategic complementarities. In contrast, in the CES case, i.e. if η = 0 we have Ψ = 0, such that inflation dynamics are governed by marginal costs only, while the slope coefficient λ depends only on the extent of price rigidities, ξ , as in the standard new Keynesian Phillips curve. As a consequence, changing the import share, i.e. lowering ω alters 16 B F Formally, we observe that the relative price of imports, P1t /P1t , falls much more strongly if LCP is reduced; it falls less, the more open an economy (quantitatively the first effect is much stronger. Results are available on request).

20

the slope coefficients in the Phillips curve only in the NCES case. Against, this background, a large value of η as implied by our estimates is likely to enhance the possible effects of increased openness, i.e. lower values of ω , on inflation dynamics. PPI inflation

Relative import price term −3

−0.04

0

−0.05

−1

−0.06

−2

−0.07

−3

−0.08

−4

−0.09

−5

−0.1

−6

−0.11

−7

−0.12 0

0.2

0.4

0.6

0.8

1

x 10

−8 0

0.2

0.4

0.6

0.8

1

Figure 6: Peak responses of inflation (PPI) to monetary policy shock and corresponding response of relative import price term. Notes: vertical axis measures peak response of inflation (left) and response of relative import price in fifth quarter (right); solid line corresponds to import share of 12 percent; dashed line corresponds to import share of 30 percent); dashed dotted line corresponds to import share of 40 percent horizontal axes measures extent of PCP; dashed vertical line corresponds to estimated extent of PCP; dotted vertical line corresponds to extent of PCP assumed in counterfactuals computed above.

Our counterfactual responses suggest that the effects of monetary policy shocks on PPI-inflation tend to increase in all three scenarios. The left panel of figure 6 illustrates this more systematically. It shows that the peak effects become stronger in the fraction of PCP starting from the point estimate (dashed vertical line) and more so, the more open an economy. It is noteworthy, however, that increasing openness only matters for the peak response if there is a considerable amount of PCP firms. Increasing only the import-to-GDP ratio while maintaining the assumption on the fraction of PCP firms (the first counterfactual), induces little change in the peak effect on inflation. The Phillips curve relationship (28) provides some guidance in interpreting these findings. Specifically, investigating how the responses of marginal costs and relative import price changes in PCP and openness (not shown) reveal that they respond less strongly and are thus of little help in accounting for the stronger inflation response.17 Therefore, we investigate the behavior of the relative import price term, i.e. the response of relative import prices multiplied with the slope coefficient in the Phillips curve. The result which is displayed in the right panel of figure 6 supports the notion that increasing openness induces a stronger response of PPI-inflation to monetary policy shocks, because the cheaper import prices exert a disinflationary pressure on domestic prices in the presence of strategic complementarities. It is interesting to note that as relative import prices per se do not change much 17

A suggestive explanation for the weaker effects on marginal costs is that the more open the economy, the more strongly tends the real appreciation to lower real marginal costs because domestic factor prices are deflated with the CPI.

21

in the PCP share or the degree of openness, this finding is driven by the slope coefficient on relative import prices in the Philips curve relationship. Moreover, note that the effect of increasing openness, i.e. lowering ω , is magnified by the parameter Ψ the absolute size of which depends on η . Hence, it appears that relative import prices will gain increasing importance in our globalization scenario and alter inflation dynamics, because we find strong complementarities in prices setting resulting from the NCES specification.

4.2 Recalibrating monetary policy We finally address the issue whether in the counterfactual globalization scenario monetary policy can maintain control over key variables and to what extent monetary policy needs to be recalibrated if it wishes to induce the same effects on these variables. Clearly, our framework allows us to study only the transmission of monetary policy shocks and not how monetary policy impacts on the transmission of other sources of business cycle fluctuations. Note also that we take an entirely positive perspective, because we currently limit simulations to have first order accuracy only and are therefore not in the position to study questions of optimal policy.18 Nevertheless, with the limitations of our framework in mind, we can assess quantitatively, which adjustment in the monetary policy rule would bring about the same effects of a monetary policy shock by 100 basis points in our counterfactual scenarios as in the baseline parametrization. We focus on the maximum effect on inflation and domestic absorption by a monetary policy shock under the baseline scenario, which is at -0.07 percentage points for inflation -0.29 percent for domestic absorption. For all three counterfactual scenarios we compute parameter values for the interest rate feedback rule (24), φπ and φy , such that the monetary impulse brings about the same effect on inflation and domestic

absorption as in the baseline scenario. In other words, we ask what changes in the systematic part of monetary policy would ensure monetary policy shocks to have the same effects in a more globalized world as in the baseline scenario.19 Columns 2 and 3 of table 2 report the results. We find that the coefficient on inflation in the interest rate feedback rule, φπ increases strongly, while the coefficient on domestic absorption φy declines. Intuitively, as inflation tends to respond more strongly in our counterfactuals monetary policy has to be more aggressively in order to bring about the same effect on inflation. In contrast, domestic absorption is more stable in the globalized economy compared to the baseline case due to reasons highlighted above. Therefore, the monetary authority even needs to act destabilizing to achieve a domestic absorption reaction as in the baseline case. 18

Note, moreover, that to take up questions of optimal monetary policy we would need to take a stance on the precise nature of business cycle fluctuations which we can currently avoid. 19 Our procedures is similar in spirit to our estimation exercise discussion in section 3. We minimize the distance between the baseline impulse response function peaks and the respective counterfactual impulse response peaks for PPI-inflation and domestic absorption. Holding fix the interest rate smoothing parameter ρ at the baseline estimation we achieve a unique solution for φπ and φy that recovers the peaks of the baseline responses.

22

Baseline CF1 CF2 CF3

φπ 1.00 1.86 2.41 5.76

Table 2: Monetary Control Parameters φy σν1t (PPI-Peak) 0.01 1 −0.28 0.97 −0.45 0.91 −1.39 0.80

σν1t (DA-Peak) 1 1.03 1.04 1.13

Notes: CF1: import share is 30 percent; CF2: 40 percent PCP; CF3: 30 percent import share and 40 percent PCP. Column 2 and 3 report variation in policy rule coefficients to bring about

The parameter values confirm that globalization requires adjustments (‘recalibration’) of monetary policy, see, for example Yellen (2006). Our results complement the findings by Woodford (2007) who analyzes three additional features of globalization and concludes that monetary authorities do not loose any control over inflation. In a sense, our finding suggest that central bank obtain better control over inflation. Another way to assess this change, we report in column 4 of table 2: the size of the monetary policy shock which induces the same peak response of PPI-inflation in a globalized economy as a 100 basis points shock in the baseline economy. We find that the size of the necessary monetary impulse declines. Note, however, that as domestic absorptions responds less in the counterfactual scenario, monetary policy would need to engineer a larger stimulus in order to generate the same effect as in the baseline case, see column 5 of table 2.

5 Conclusion In this paper, we ask whether globalization alters the monetary transmission mechanism and if so, to what extent monetary policy responses need to the recalibrated in order to engineer the same effects on domestic absorption and inflation for a given monetary policy shock. We proceed as follows. First, we suggest a general equilibrium model featuring two symmetric countries and several frictions which recent business cycle research has found to key to account for several macroeconometric observations. Most importantly, we embed a final good aggregation technology into our general equilibrium model, which induces non-constant demand elasticities for intermediate goods. This feature has been put forward by Gust et al. (2006), Sbordone (2007) and Guerrieri et al. (2008), drawing on earlier closed economy work by Kimball (1995) and Dotsey and King (2005). In contrast to these studies, we explore its consequences in a full-fledged open economy general equilibrium model with sticky prices. To pin down parameter values of the key parameters of the model, we first estimate a VAR model on U.S. time series data relative to an aggregate of industrialized countries for the post Bretton-Woods period 1973-2006. We identify monetary policy shocks in the VAR model imposing an identification

23

scheme which is consistent with our theoretical model and trace out the transmission mechanism by studying impulse response functions. In a second step, we find parameter values of the general equilibrium model by matching its implied impulse responses to those obtained from the VAR. We find that the estimated model is generally able to mimic the empirical response functions quite closely and conclude that it is able to give an empirically plausible account of the actual transmission mechanism. Most importantly, we find that the model requires that most of the firms engage in LCP and that demand elasticities are very responsive to deviations from steady state. Hence, the model is well suited to perform counterfactual experiments. Specifically, we study three experiments. First, we increase the import-to-GDP ratio from 12 percent (the assumption maintained during the estimation) to 30 percent, as increased trade integration is a likely manifestation of globalization. Second, we lower the fraction of firms engaged in PCP, thus reducing the extent of price discrimination across countries. Our estimates suggest that currently LCP is quite pervasive; however, globalization through reducing barriers which prevent arbitrage may induce a rise in the number of PCP firms. In a third counterfactual scenario we study the effect of both changes occurring at the same time. Focusing on the effects of a monetary policy shock on domestic absorption and PPI-inflation in the counterfactual scenarios relative to the estimated model, we find that globalization alters the transmission mechanism quantitatively. Most strongly so in our third counterfactual, i.e. if both LCP is reduced and the openness increased simultaneously. Intuitively, only if the exchange rate pass-through into import prices increases (as LCP falls), the extent of trade openness becomes more important for monetary transmission. Most interestingly, if a considerable amount of firms engages in PCP, the role of relative import prices for inflation dynamics increases considerably with openness. Finally, we show that monetary policy may still induce the same response of inflation and domestic absorption to a given monetary impulse (in terms of peak responses). First, by adjusting its endogenous response to the economy, monetary policy may at the same time achieve the same effects on inflation and domestic absorption in the counterfactual scenarios as in the baseline scenario. Alternatively, it may alter the size of the monetary impulse and thus achieve the same effect on either inflation or output. Clearly, our investigation is limited to the transmission of monetary policy shocks and globalization is likely to alter the transmission of other sources of business cycle fluctuations as well. This may, in turn, require other adjustments in the conduct of monetary policy. We leave these questions as well as questions of optimal monetary policy for future research.

References J. D. Amato and T. Laubach. Estimation and control of an optimization-based model with sticky prices and wages. Journal of Economic Dynamics and Control, 27:1181–1215, 2003.

24

D. Backus, P. Kehoe, and F. Kydland. Dynamics of the trade balance and the terms of trade: the J-curve? American Economic Review, 84(1):84–103, March 1994. P. R. Bergin. How well can the new open economy macroeconomics explain the exchange rate and current account? Journal of International Money and Finance, 25:675–701, 2006. B. S. Bernanke. Globalization and monetary policy. Speech at the Fourth Economic Summit, Stanford Institute for Economic Policy Research, 2007. C. Betts and M. B. Devereux. Exchange rate dynamics in a model of pricing-to-market. Journal of International Economics, 50:215–244, 2000. J. Bovin and M. Giannoni. Has monetary policy become more effective? The Review of Economics and Statistics, 88:445–462, 2006. G. Calvo. Staggered prices in a utility maximizing framework. Journal of Monetary Economics, 12: 383–398, 1983. V. V. Chari, P. J. Kehoe, and E. R. McGrattan. Can sticky price models generate volatile and persistent real exchange rates? Review of Economics Studies, 69:533–563, 2002. L. J. Christiano, M. Eichenbaum, and C. L. Evans. Monetary policy shocks: What have we learned and to what end? In J. B. Taylor, editor, Handbook of Macroeconomics, pages 319–347. Elsevier B.V., 1999. L. J. Christiano, M. Eichenbaum, and C. L. Evans. Nominal rigidities and the dynamic effects of a shock to monetary policy. Journal of Political Economy, 113:1–45, 2005. R. Clarida and J. Gali. Sources of real exchange-rate fluctuations: How important are nominal shocks? Carnegie-Rochester Conference Series on Public Policy, 41:1–56, 1994. R. Clarida, J. Galí, and M. Gertler. Monetary poliy rules and macroeconomic stability: Evidence and some theory. Quarterly Journal of Economics, CXV:147–180, February 2000. G. Corsetti and P. Pesenti. International dimenasions of optimal monetary policy. Journal of Monetary Economics, 52:281–305, 2005. G. Corsetti, A. Meier, and G. J. Müller. International dimensions of fiscal policy transmission. mimeo Goethe University Frankfurt, 2007. D. O. Cushman and T. Zha. Identifying monetary policy in a small open economy under flexible exchange rates. Journal of Monetary Economics, 39:433–448, 1997. G. de Walque, F. Smets, and R. Wouters. An estimated two-country DSGE model for the Euro area 25

and the US economy. mimeo, 2005. M. Dotsey and R. G. King. Implications of state-dependent pricing for dynamic macroeconomic models. Journal of Monetary Economics, 52:213–242, 2005. M. Eichenbaum and C. L. Evans. Some empirical evidence on the effects of shocks to monetary policy on exchange rates. The Quarterly Journal of Economics, 110(4):975–1009, 1995. J. Faust and J. H. Rogers. Monetary policy’s role in exchange rate behavior. Journal of Monetary Economics, 50:1403–1424, 2003. J. Galí and T. Monacelli. Monetary policy and exchange rate volatility in a small open economy. Review of Economic Studies, 72:707–734, 2005. L. Guerrieri, C. Gust, and D. López-Salido. International competition and inflation: A new keynesian perspective. International Finance Discussion Papers, 920, 2008. C. Gust, S. Leduc, and R. J. Vigfusson. Trade integration, competition, and the decline in exchangerate pass-through. International Finance Discussion Papers, Number 864, Board of Governors of the Federal Reserve System, 2006. S. Kim. International transmission of U.S. monetary policy shocks: Evidence from VAR’s. Journal of Monetary Economics, 48:339–372, 2001. S. Kim and N. Roubini. Exchange rate anomalies in the industrial countries: A solution with a structural VAR approach. Journal of Monetary Economics, 45:561–586, 2000. M. Kimball. The quantitative analytics of the basic monetarist model. Journal of Money, Credit and Banking, 27(4):1241–1277, 1995. R. Kollmann. Monetary policy rules in the open economy: Effects on welfare and business cycles. Journal of Monetary Economics, 49:989–1015, 2002. R. Kollmann. International portfolio equilibrium and the current account. CEPR Discussion Paper 5512, 2006. T. Lubik and F. Schorfheide. A bayesian look at new open economy macroeconomics. In M. Gertler and K. Rogoff, editors, NBER Macroeconomics Annual 2005, Volume 20, pages 313–366. MIT Press, Cambridge MA, 2006. A. Meier and G. J. Müller. Fleshing out the monetary transmission mechanism: Output composition and the role of financial frictions. Journal of Money, Credit and Banking, 38:2099–2134, 2006. F. S. Mishkin. Globalization, macroeconomic performance, and monetary policy (speech). BIS 26

Review, 108, 2007. E. Nakamura and J. Steinsson. Five facts about prices: A reevaluation of menu cost models. mimeo Columbia University, 2008. M. Obstfeld and K. Rogoff. The six major puzzles in international macroeconomics: Is there a common cause? In B. Bernanke and K. Rogoff, editors, NBER Macroeconomics Annual 2000, pages 339–390. MIT Press, Cambridge, MA, 2000a. M. Obstfeld and K. Rogoff. New directions for stochastic open economy models. Journal of International Economics, 50:117–153, 2000b. J. H. Rogers. Monetary shocks and real exchange rates. Journal of International Economics, 49: 269–288, 1999. J. R. Rotemberg and M. Woodford. An optimization-based econometric framework for the evaluation of monetary policy. In J. B. Taylor, editor, NBER Macroeconomics Annual, pages 297–346. Cambridge, MA.: MIT Press, 1997. A. M. Sbordone. Globalization and inflation dynamics: the impact of incnreased competition. Federal Reserve Bank of New York, 2007. C. Schmidt. International transmission effects of monetary policy shocks: Can asymmetric price setting explain the stylized facts? Internation Journal of Finance and Economics, 11:205–218, 2006. A. Scholl and H. Uhlig. New evidence from the puzzles: Results from agnostic identification on monetary policy and exchange rates. forthcoming: Journal of International Economics, 2007. O. Senay. The effects of goods and financial market integration on macroeconomic volatility. The Manchester School, 66:39–61, 1998. F. Smets and R. Wouters. Comparing shocks and frictions in US and euro area business cycles: a Bayesian DSGE approach. Journal of Applied Econometrics, 20:161–183, 2005. M. Woodford. Globalization and monetary control. In International Dimensions of Monetary Policy. Jordi Gali and Mark Gertler, editors, 2007. NBER Books in Progress. J. L. Yellen. Monetary policy in a global environment. Speech at The Euro and the Dollar in a Globalized Economy Conference, U.C. Santa Cruz, Stanta Cruz, CA, 2006.

27