Monetary Policy Spillovers through Invoicing Currencies Appendix

Appendix A Appendix to the Theory Section A.1 Deriving the Nominal Price Indices In this section, I derive expressions for country n price indices for intermediate traded goods n produced in country m, P̃T,m , for aggregate consumption of traded goods, P̃Tn , and for the aggregate consumption bundle of traded and non-traded goods, P̃ n . In country n, the cost of a unit of traded consumption from country m is µm Z n P̃T,m n n P̃T,m,j CT,m,j dj = arg min Z µm
ε n CT,m,j s.t.  1ε dj =1 0 0 All country m intermediate traded goods invoiced in dollars have the same nominal price in country n, and all country m intermediate traded goods invoiced in domestic currency have the same nominal price in country n. Hence, I can re-write the optimization problem as, n n n n n P̃T,m = arg min µm X$m P̃T,m,$ CT,m,$ + µm (1 − X$m ) P̃T,m,h CT,m,h 
ε
ε  1ε n n s.t. µn X$m CT,m,$ =1 + µn (1 − X$m ) CT,m,h n n and CT,m,h and take the ratio to solve for I take first order conditions with respect to CT,m,$ n CT,m,h , n n CT,m,h = CT,m,$ n P̃T,m,$ 1 ! 1−ε . n P̃T,m,h n I solve for CT,m,$ by plugging this equation back into the constraint,  n CT,m,j = µm X$m + µm (1 − X$m ) =  n P̃T,m,$ n P̃T,m,$ n P̃T,m,h − 1ε 1 ! 1−ε   − 1ε 1 ε ε − 1−ε   ε−1   ε−1 m m n m m n µ X$ P̃T,m,$ + µ (1 − X$ ) P̃T,m,h . n n n I use this equation to solve for CT,m,h , plug CT,m,$ and CT,m,h into the objective and derive the n following expression for P̃T,m , n P̃T,m  = µ m X$m  n P̃T,m,$ ε  ε−1 m + µ (1 − X$m )  n P̃T,m,h  ε−1 ε  ε−1 ε . (21) In country n, the cost of a unit of the final traded consumption comprised of intermediate traded goods from each country is, P̃Tn = arg min X n n P̃T,m CT,m s.t. n CT,U S
αnU S n CT,JP
αnJP n CT,EU
αnEU =1 m={U S,JP,EU } First order conditions imply, n P̃ n n P̃ n αEU αJP T,U S n T,U S n n C CT,U S . and C = T,U S T,EU n n αUn S P̃T,JP αUn S P̃T,EU n CT,JP = Plugging these values into the constraint, using the identity P m=U S,JP,EU n = 1 and solving αm n for CT,U S yields, n n P̃T,U S CT,U S = αn US (αUn S )  αnU S  αnJP  αnEU αUn S n n n P̃ P̃ P̃ . n n T,U S T,JP T,EU n αJP n (αJP ) (αEU )αEU Plugging this equation into the objective yields,  P̃Tn = n P̃T,U S αnU S  n P̃T,JP n αnJP  n P̃T,EU n n αnEU n αJP n (αUn S )αU S (αJP ) (αEU )αEU . (22) In country n, the cost of a unit of the aggregate consumption bundle comprised of final traded goods and non-traded goods is, P̃ n = arg min P̃Tn CTn + P̃Nn s.t. (CTn )τ (CNn )1−τ = 1 Repeating the same steps as above yields the following expression for the aggregate nominal price level in country n,  P̃ n = P̃Tn τ  P̃Nn 1−τ (τ )τ (1 − τ )1−τ . (23) A.2 Deriving the Conditions of Optimality In this subsection, I provide additional details about solving the model in Section 3. Households maximize utility (1) subject to their budget constraints (4). Let Ψn denote the Lagrange multiplier on the budget constraint for households in country n. By assumption, the wealth transfer, κn , in equation 4 equalizes Ψn = Ψm for all n, m. For a household in country n, the first order condition with respect to the consumption of intermediate traded good j produced in country n m, CT,m,j,t , yields, n αm τ (Ctn )1−γ n CT,m,t ! n CT,m,t n CT,m,j,t !1−ε n = Ψn Qnt P̃T,m,j,t . The first order condition with respect to the consumption of non-traded goods is, (1 − τ ) (Ctn )1−γ n = Ψn Qnt P̃N,t . n CN,t The first order condition with respect to labor supplied by the household is, Ntn = Ψn Qnt W̃tn . Finally, I take a first order condition with respect to aggregate consumption, (Ctn )−γ = Ψn Qnt P̃tn (24) I derive equations (5) - (7) by taking the ratio between the households’s first order conditions with respect to intermeditate traded consumption, non-traded consumption and labor supply with the household’s first order condition with respect to aggregate consumption. I derive equation (11) by taking the ratio between the first order condition with respect to aggregate consumption for country n and country m, given by equation (24). I can cancel out the Lagrange multipliers, because Ψh = Ψf . Since I do not introduce and frictions into financial markets, Qnt = Ẽtn,m Qm t . I simplify and recover equation (11). In the remainder of this subsection, I solve the firm’s problem. These equations are not shown in the main text. In both periods, the intermediate traded good firm maximizes profits by choosing the price of its variety of intermediary traded good. Firm j in country m faces the following demand function in country n, n CT,m,j,t =  n P̃T,m,j,t 1   1  ε11  αn  ε−1 ε
ε−1 Ctn ε−1 m n C T,m,t τ P̃tn (25) In the second period, nominal prices are flexible and the firm j in country m chooses the nominal price of their traded good in each country and in each state of the world to maximize discounted profits, X max n P̃T,m,j,t   n  CT,m,j,2 n n m µ P̃T,m,j,2 CT,m,j,2 − W̃2 Am 2 n n∈{U S,JP,EU } (26) n , I plug equation (25) into the equation (26), take first order conditions with respect to P̃T,m,j,t n This yields the typical result where the firm sets it’s nominal price as a and I solve for P̃T,m,j,t constant markup over marginal cost in domestic currency, m P̃T,m,j,2 = 1 W̃2n n n and P̃T,m,j,2 = Ẽ2n,m P̃T,m,j,2 n ε A2 In the first period, firms set prices before the realization of shocks to maximize expected discounted first period profits. Furthermore, firms that invoice their exports in a foreign currency need to set a domestic currency price as well as a foreign currency price. If the firm only invoices in domestic currency (producer currency pricing), then it faces the following problem,  X max E Qm 1 m P̃T,m,j,1 n∈{U S,JP,EU }    n  CT,m,j,1 m n . µn P̃T,m,j,1 CT,m,j,1 − W̃1m Am 1 Demand in each country each country is given by, n CT,m,j,1 =  m Ẽ1n,m P̃T,m,j,1 1   1  ε11  αn  ε−1 ε
ε−1 C1n ε−1 m n C . T,m,1 τ P̃1n I plug in the demand function into the firm’s problem, take the first order condition with respect m to P̃T,m,j,1 and solve for the price, m P̃T,m,j,1 = h   i m m E Qm W̃ Y /A T,m,j,1 1 1 1 E [Qm 1 YT,m,j,1 ] where YT,m,j,1 = P n∈{U S,JP,EU } n µn CT,m,j,1 is the total output by firm j located in country m. m If the firm invoices in dollars, then the firm chooses a domestic currency price, P̃T,m,j,1 , as US well as a dollar price, P̃T,m,j,1 , in order to maximize expected discounted first p ( max m US P̃T,m,j,1 ,P̃T,m,j,1 E " Qm 1 µ m m m P̃T,m,j,1 CT,m,j,1 X + µ n US m Ẽ1m,U S P̃T,m,j,1 CT,m,j,1 − W̃1m n6=m P n n µn CT,m,j,1 Am 1 #) . I plug in the demand function into the firm’s problem, take the first order conditions and solve for the optimal prices, m P̃T,m,j,1 i i     h h P m m n n m m m m /A /A µ C C E Q W̃ E Qm W̃ 1 1 1 1 1 1 T,m,j,1 T,m,j,1 n6=m m  m m  i h and P̃T,m,j,1 = = P E Q1 CT,m,j,1 E QU S µn C n 1 n6=m T,m,j,1 m,U S where QU1 S = Qm . 1 Ẽ1 A.3 Log-Linear Approximation of the Equilibrium I log-linearize the model around the following deterministic steady state: Ant s X = (εµn )−1 µm αnm m n CT,m,j,t YT,m,j,t n P̃T,m,j,t = n αm , µm n CN,t = n YN,t = Y m − (µ ) αn m (1−ε)τ ε(1−τ ) τ n −αn (αm ) m 1−τ  m 1 X n n n n µ αm , Nm,j,t = YT,m,j,t /Am = m t , Nt = µ Nn,j,t µ n  Y αn (1−ε)τ 1 τ −1+ (γ−1)(1−τ n n n m − m n −αn m 1−τ ) ε(1−τ ) = 1, W̃t = εAt , P̃N,t = (1 − τ ) (τ ) (µ ) (αm ) m P̃tn = (τ )γ γ − 1, Ẽtn,m = 1 for all countries n and m and time periods t. A.4 Properties of Constant ν ν = (1−τ )3 +γτ (1−τ ) γτ Ξ1 + (1 − τ )


US
US JP EU 1 − X$EU αEU + 1 − X$JP αJP + αUJPS + αEU + αUEUS + αJP (27) where,
US
JP
EU Ξ1 = 1 − X$JP αJP 1 − X$EU αEU + αUEUS + αJP +


US JP EU 1 − X$EU αEU αUJPS + αEU + 1 − X$JP αJP +


EU JP EU X$JP + X$EU − X$JP X$EU αUEUS + αJP + αEU αUJPS αUEUS + αJP I prove two propositions characterizing the constant ν. Remark. The constant ν > 0. n > 0 for all Proof. This is straightforward to show, because γ > 0, τ ∈ (0, 1), X$n ∈ (0, 1) and αm n, m. Plugging these assumptions into equation (27) shows that ν > 0. n for n 6= m. Remark. ν is increasing in αm Proof. This is straightforward to check. Equation 27 displays ν as the product and sum of n n positive constants and each αm where n 6= m. Hence, all derivatives of ν with respect to αm are positive. A.5 Proof of Prediction 3 The effect of contractionary Japanese monetary policy on aggregate consumption in the U.S. and Europe is γτ 2 (1 + (γ − 1)τ ) 1 − X$JP ∂cn − JP = − ∂ p̃


EU US EU (1 − αUU SS )(1 − αEU ) − αEU αUEUS + αJP X$EU ν (1 − τ )(1 + (γ − 1)τ ) n − MU . ν The first line of the previous equation corresponds with the first line of equation (12). The second line of the previous equation exactly mirrors the second line of equation (12) with MUn taking the place of M$n . Since nominal interest rates are determined by the consumption Euler equation given by equation (14), the effect of of contractionary Japanese monetary policy on nominal interst rates in the U.S. and Europe is   ∂r̃n ∂cn − JP = −γ − JP . ∂ p̃ ∂ p̃ Hence, the effect of Japanese monetary policy on foreign nominal interest rates is symmetric to the effect of U.S. monetary policy due to the symmetry in the monetary policy spillovers to consumption. The corresponding expressions for European monetary policy shocks simply involve replacing the JP superscripts in the expressions above with EU superscripts. A.6 Constant δy n = (1 − α)/2 for n 6= m When µn = 1/3, αnn = α for all countries n and αm δy = A.7 τ (1 − α) ((1 + 3α)(1 − τ ) + γ (1 − 3α(1 − τ ) + τ )) (1 + ((3/2)αγ − 1) τ ) . 16ν Proof of Proposition 1 If τ n = τ is constant across all countries n, then it is straightforward to show that the derivative n ∂yT ∂ p̃n for n = JP, EU is decreases with the share of country n’s dollar invoiced exports, X$n . For n = JP ,



US EU + 1 − αEU −1 τ (1 + (γ − 1)τ ) 1 + γ 1 − X$EU αEU ∂ 2 yTJP =− P m ∂ p̃JP ∂X$JP (1 − τ ) ( m µm αJP ) (Ξ4 )2

US EU × Ξ2 + µU S αJP + µEU αJP Ξ3 < 0, where Ξ2 and Ξ3 are positive constants. Ξ4 is a constant that is not displayed here.

U S EU JP JP Ξ2 =αJP αJP (γ − 1)(1 − τ )(1 − µJP ) 1 + γ(1 − αJP ) − 1 τ + µJP αJP γτ 2
US JP EU JP U S EU EU × (1 − τ )(1 − X$EU ) αUJPS αJP + αEU αJP + (1 − αJP ) γτ αJP αU S + αJP

U S

US + 1 − αUU SS − αEU X$EU αJP (1 − τ )X$EU .

JP Ξ3 = 1 + γ(1 − αJP )−1 τ US )τ 1 + (1 − αJP


1 + αUEUS γ − 1 τ



US EU US X$EU ) − 1 τ . + αEU γ(1 − τ )τ (1 − X EU )$) + αJP τ 1 + γ(1 − αUU SS − αEU The analogous expressions for n = EU are similar except the superscripts and subscripts for EU and JP are reversed. Domestic expansionary monetary policy increases aggregate traded production, n ∂yT . ∂ p̃n Hence, an increase in the share of country n’s dollar invoiced exports decreases the magnitude of the effect of domestic monetary policy on domestic aggregate production. A.8 Model Extension with Incomplete Asset Markets In this section, I extend the model in the paper to include incomplete asset markets. Within each country, only a fixed proportion φ of households (labelled “active”) trade a complete set of statecontigent securities. The remaining 1 − φ of households (labelled “inactive”) are excluded from trading in financial markets. Instead, inactive households cede the claims to their firm profits and their endowments of non-traded goods to active households. In return, inactive households receive a nominal bond that makes a fixed nominal payment B̃tn in each period. Active households maximize expected utility (1) subject to the following modified budget constraint E  X  n n CN,t + Qnt P̃N,t  t=1,2 ( X Z X 1 n n P̃T,i,j,t CT,i,j,t dj 0 i∈{U S,JP,EU }   − 1−φ φ    n = B̃t    )  1 n n E Qnt P̃N,t YN,t + Γ̃nt + W̃tn Ntn + κn . φ t=1,2 The active household’s budget constraint differs from equation (4) in two ways: Active households pay ((1 − φ)/φ) B̃tn units of currency to inactive households in each period, and active households receive 1/φ shares of claims to the non-traded endowment and the profits of the firms in each period. Although the budget constraint has changed from the model Section 3, the first order conditions for active households are unaffected. Equations (5), (6) and (7) still determine the active household’s allocation of intermediate traded goods, non-traded goods and labor supply. The wealth transfer κn equates the marginal utility of wealth across active households such that equation (11) continues to hold. Inactive households also maximize expected utility (1) in each period. However, their problem simply involves choosing how to use their nominal bond payment and their labor income to purchase consumption. Inactive households face the following budget constraint in each period t, n n P̃N,t ĈN,t + X i∈{U S,JP,EU } Z 1 n n ĈT,i,j,t dj P̃T,i,j,t  ≤ B̃tn + W̃tn N̂tn 0 n ĈNn and ĈT,m,j denote the non-traded consumption and intermediate traded consumption of inactive households in country n. N̂ n denotes the labor supplied by inactive households in country n. Since household utility (1) is a Cobb-Douglas aggregate of traded and non-traded goods, it is straightforward to write the inactive household’s consumption of traded and non-traded goods as a function of market prices and their total income. The inactive household’s consumption of intermediate good j produced in country m is n n τ = αm ĈT,m,j n P̃T,m 1 ! 1−ε n P̃T,m,j B̃tn + W̃tn N̂tn n P̃T,m ! . (28) n P̃T,m is the nominal price index in country n for the aggregate traded good produced in country m, which is derived in Appendix A.1. The inactive household’s consumption of the non-traded good is n ĈN,t = (1 − τ ) ! B̃tn + W̃tn N̂tn n P̃N,t . (29) The inactive household’s labor supply is determined by the first order condition with respect to Ñtn  γ W̃ n Ĉtn N̂tn = nt P̃t (30) The resource constraints in the economy need to account for the consumption and labor supply of both active and inactive households. The market clearing condition for non-traded goods is n n n YN,t = φCN,t + (1 − φ)ĈN,t . (31) The market clearing condition for intermediate traded good j produced in country m is YT,m,j,t = m Am t µ   X φNm,j,t + (1 − φ)N̂m,j,t = µ n  n φCT,m,j,t + (1 − n φ)ĈN,t  . (32) n∈{U S,JP,EU } Finally, the labor market clearing conditions for active and inactive households in country m are, Ntm Z = µ Nm,j,t dj and 0 N̂tm Z = µ N̂m,j,t dj (33) 0 An equilibrium in this economy is a set of consumption and labor supply allocations for each n o n n n n household, CT,m,j,t , CN,t , Ntn , ĈT,m,j,t , ĈN,t , N̂tn ; a set of intermediate output, labor demand and n o n n nominal prices for each intermediate firm, YT,m,j,t , Nm,j,t , N̂m,j,t , P̃T,m,j,t ; a set of traded goods n o n n prices and non-traded goods prices, P̃T,t , P̃N,t ; a set of nominal wages for each country and a n o nominal exchange rate, W̃tn , Ẽtn,m ; such that active and inactive households maximize utility subject to their budget constraints and the markets clearing conditions are satisfied. I log-linearize the model around the same deterministic steady state described in Appendix A.3. The addition of market segmentation significantly complicates the analytic expressions in the model. Hence, I make the simplifying assumptions in order to derive expressions analogous to the results from Section 4. I assume households consume an equal share of traded goods from n all countries, αm = α = 1/3 and countries are all the same size, µn = 1/3, for all countries n and m. Furthermore, I assume all firms in the Europe invoice in Euros, X$EU = 0. I derive the effects of a contractionary U.S. monetary policy shock on nominal exchange rates, consumption and production and show that Predictions 1 through 4 hold in the incomplete markets model. First, I define the constant ν̂ ν̂ = (1 − τ )(1 + (γ − 1)τ )Ξ̂1 Ξ̂2 3Ξ̂3 Ξ̂4 , where Ξ̂1 = φ(1 − τ ) + γτ (1 + ε(1 − τ )(2 − φ)) + εγ 2 τ 2 (2φ − 1) Ξ̂2 = 2γτ (1 + ε(1 − τ )) + γτ Ξ̂4 (1 − X$JP ) + 3φ(1 − τ ) + εγτ (φ(1 − τ ) + 2γτ (2φ − 1)) Ξ̂3 = φ(1 − τ ) + γ(1 − (1 − τ )φ) + γετ (1 + (γ − 1)τ )(2φ − 1) Ξ̂4 = 1 + ε (4(1 − τ )(1 − φ) + γτ (2φ − 1)) It is straightforward to show that ν is positive whenever φ > 1/2. Hence, I assume φ > 1/2 for the remainder of this section. Equation (34) provides the incomplete markets analog to equation (13) in Section (4), − ∂ẽn,U S γ(1 − τ )(1 + (γ − 1) τ ) n = δ̂e − M$ . U S ∂ p̃ ν̂ (34)
δ̂e is constant that is common across countries, M$JP = ατ and M$EU = α + X$JP α τ . Equation (35) provides the incomplete markets analog to equation (12) in Section (4), − ∂cn (1 − τ )(1 + (γ − 1)τ ) n = δ̂c − M$ . U S ∂ p̃ ν̂ (35) Active household consumption in Europe decreases relative to active household consumption in Japan, because Europe has a larger share dollar invoiced consumption than Japan. Equation (36) provides the incomplete markets analog to equation (16) − ∂yTn = δ̂y (36) ∂ p̃U S α(1 − τ )(1 + (γ − 1)τ ) (γτ (γετ − 1)(2φ − 1) + 2φ(1 + (γ(1 + ε(1 − τ )) − 1)τ )) JP X$ . − ν̂ Ξ̂4 Traded production in Japan decreases more than traded production in Europe, because Japan has a higher share of dollar invoiced exports. Finally, equation (37) shows the effect of a contractionary Japanese monetary policy shock on an active household’s consumption in the U.S. and Europe, − (1 − τ )(1 + (γ − 1)τ ) n ∂cn MU . = δ̂cJP − JP ∂ p̃ ν̂ (37) In the incomplete markets extension, monetary policy shocks from Japan have the same magnitude at those from the U.S. after controlling for the share of consumption invoiced in each currency. Finally, I show that the incomplete markets model breaks the tight link between a country’s aggregate consumption and changes in its exchange rates and interest rates. Aggregate consumption in country n (denoted by C̄ n ) is the weighted sum of active household consumption and inactive household consumption, C̄ n = φC n + (1 − φ) Ĉ n Equation (38) shows the effect of a contractionary U.S. monetary policy shock on the difference in aggregate consumption between Japan and Europe −
(1 − τ )(1 + (γ − 1)τ ) (2γετ (1 − 2φ) + φ(γετ − 1)) ατ X$JP ∂ JP EU c̄ − c̄ = − . ∂ p̃U S ν Ξ̂4 (38) It is straightforward to find an example where the right-hand side of equation 38 is negative. For example, this is true whenever φ ∈ [1/2, 2/3] and the coefficient of risk aversion is sufficiently large (γ > φ/(ετ (1 − (3/2)φ))). Then, the decrease in Japanese aggregate consumption is greater than the decrease in European aggregate consumption even though European households consume more dollar invoiced traded goods as a share of total consumption. B Appendix to the Empirical Section This section of the Appendix provides summary statistics provides summary statistics for the data, empirical results and robustness checks for the main empirical results. B.1 Summary Statistics Table 9: Central Banks Central Bank Federal Reserve Bank European Central Bank Bank of Japan Bank of England Bank Code 3 Month Interest Rate Future FRB Eurodollar ECB Euribor BOJ Tibor BOE Short Stirling Table 10: Data Availability Region Alpha3 Australia AUS Canada CAN Denmark DNK Eurozone Iceland ISL Israel ISR Japan JPN Norway NOR South Korea KOR Sweden SWE Switzerland CHE United Kingdom GBR United States USA Austria Belgium Cyprus Estonia Finland France Germany Greece Ireland Italy Latvia Lithuania Luxembourg Malta Netherlands New Zealand Portugal Slovakia Slovenia Spain AUT BEL CYP EST FIN FRA DEU GRC IRL ITA LVA LTU LUX MLT NLD NZL PRT SVK SVN ESP Exchange Rates Interest Rates 1999 - 2014 1999 - 2014 1999 - 2014 1999 - 2014 1999 - 2014 1999 - 2014 1999 - 2014 1999 - 2014 1999 - 2014 2004 - 2014 1999 - 2014 2004 - 2014 1999 - 2014 1999 - 2014 1999 - 2014 1999 - 2014 1999 - 2014 2002 - 2014 1999 - 2014 1999 - 2014 1999 - 2014 1999 - 2014 1999 - 2014 1999 - 2014 1999 - 2014 1999 - 2014 Eurozone Countries Ind. Prod. 1999 - 2014 1999 - 2014 1999 - 2014 1999 1999 1999 1999 1999 1999 1999 1999 1999 - 2014 2014 2014 2014 2014 2014 2014 2014 2014 1999 - 2014 1999 - 2014 1999 - 2007 1999 - 2010 2004 - 2007 2004 - 2010 1999 - 2000 1999 - 2000 1999 - 2013 2002 - 2014 2004 - 2013 2004 - 2014 1999 - 2007 2004 - 2007 1999 - 2006 2004 - 2006 1999 1999 1999 1999 1999 1999 1999 2000 1999 1999 - 2014 2014 2014 2014 2014 2014 2014 2014 2014 2014 1999 1999 1999 1999 1999 1999 - 2014 2014 2014 2014 2014 2014 Table 11: Summary Statistics with Nakamura and Steinsson (2015) Data Obs Mean Std Dev Monetary Policy Shocks (bps) Federal Reserve Bank 115 0.331 3.609 Country Characteristics Dollar Imports / Imports 19 0.444 0.206 Imports / Consumption 19 0.846 0.330
19 0.349 0.186 Dollar Imports / Consumption M$n
n 19 0.409 0.235 Dollar Exports / Consumption X$ Max Min 9.863 -13.736 0.808 1.441 0.843 0.852 0.125 0.308 0.134 0.000 Notes: The table summarizes monetary policy shocks on regularly scheduled policy announcements days for each central bank between 1999 to 2014. I drop days where more than one central bank issued a monetary policy announcement, and I drop the height of the financial crisis (July 2008 - June 2009). Table 12: Summary Statistics with Multiple Central Banks Obs Mean Std Dev Max Min Monetary Policy Shocks (bps) Federal Reserve Bank 108 -0.644 4.943 20.500 -20.000 European Central Bank 98 0.612 3.742 15.000 -8.500 Bank of Japan 200 -0.145 1.253 3.000 -11.000 Bank of England 66 -0.606 5.217 23.000 -19.000 Country Characteristics Currency f Imports / Imports 67 0.176 0.233 0.808 0.000 Imports / Consumption 67 0.846 0.308 1.441 0.308
n Currency f Imports / Consumption Mf
67 0.144 0.198 0.843 0.000 Currency f Exports / Consumption Xfn 67 0.159 0.227 0.852 0.000 Notes: The table summarizes monetary policy shocks on regularly scheduled policy announcements days for each central bank between 1999 to 2014. I drop days where more than one central bank issued a monetary policy announcement, and I drop the height of the financial crisis (July 2008 - June 2009). C Data Appendix C.1 Monetary Policy Shocks Table 13 summarizes the scheduled monetary policy announcement dates in my sample. I list the initial date, the final date, the number of monetary policy announcement days per year and link to the data. Table 13: Scheduled Monetary Policy Announcement Days Central Bank FRB ECB BOJ BOE Data Availability Announcements per Year 1990 - 2015 For the relevant part of the sample (1999 2014) there have been eight meetings a year. However, during the Great Recession, we also observe “Intermeeting Press Releases” on: Aug 10, 2007; Aug 17, 2007; Jan 22, 2008; Oct 8, 2008 and May 9, 2010. 1999 - 2015 For the most part, the ECB meets monthly. However, the number of meetings increased when the Euro was first created, and during the Eurozone Crisis. 1998 - 2015 Fourteen to twenty 1997 - 2015 Twelve monthly meetings Source Link Link Link Link I gather daily measures of three month interest rate futures from Bloomberg. The series symbols for three month interest rate futures for the dollar, euro, yen and pound are “ED1 Comdty”, “ER1 Comdty”, “YE1 Comdty” and “L 1 Comdty”, respectively. For each series, I pull the opening price (PX OPEN) and the closing price (PX CLOSE). My measure of monetary policy shock is the change between the openning price and closing price on days of monetary policy announcements. C.2 Import and Export Invoicing Currencies I use data on import and export invoicing currencies from Gopinath (2015). The data are available on her website: http://scholar.harvard.edu/gopinath/publications/international-price-system The data provide the average share of each countries imports and exports that are invoiced in each currency over the period 1999 - 2014. I need to perform additional calculations for the Eurozone, because I only observe one nominal interest rate for the Eurozone as a whole but I obseve currency invoicing data for the countries of the Eurozone individually. Hence, I need to estimate the currency composition of the Eurozone’s exports and the Eurozone’s imports with respect to the rest of the world. I use bilateral goods trade data to estimate the share of each country’s imports and exports with respect to other Eurozone countries, and I assume that all trade conducted within the Eurozone is done using Euros. These assumptions allow me to estimate each country’s imports and exports with respect to non-Eurozone countries that are invoiced in each currency. Finally, I aggregate the country level data to calculate the invoicing shares of the Eurozone’s imports and exports with respect to the rest of the world. The details of my procedure for estimating currency invoicing shares for the Eurozone are as follows. For each country n in the Eurozone in year t, I observe the share of the country’s n imports invoiced in currency c. Denote this as SM,c . I denote the total current dollar value of country n’s imports as VMn , and the total current dollar value of country n’s imports invoiced in n currency c as VM,c . Furthermore, denote the value of country n’s imports from the Eurozone as VMn,EU R and the value of country n’s imports with the non-Eurozone countries as VMn,RoW . All values are denoted in current U.S. dollars and these values are calculated annually. I have chosen to drop the time t notation to simplify the exposition. I estimate the value of each country’s imports from non-Eurozone countries that are invoiced n,RoW in currency c is, VM,c as n SM,c = n,RoW VM,c VMn,RoW ! VMn,RoW VMn ! + n,EU R VM,c ! VMn,EU R VMn,EU R VMn ! . (39)   n,RoW Where VM,c /VMn,RoW is the share of country n’s imports from non-Eurozone countries that   n,RoW n are invoiced in currency c and VM /VM is the share of country n’s total imports that   n,EU R come from non-Eurozone countries. Analogously, VM,c /VMn,EU R is the share of country n’s   imports from Eurozone countries that are invoiced in currency c and VMn,EU R /VMn is the share of country n’s total imports that come from Eurozone countries. I use equation (39) to solve for the share of country n’s imports from non-Eurozone countries   n,RoW /VMn,RoW . I observe the left-hand side, and I assume that are invoiced in currency c, VM,c that for all countries inside the Eurozone, all intra-Eurozone trade is conducted in Euros. I need to estimate the share of each country’s total imports that originate from other Eurozone countries. To do so, I use bilateral bilateral goods trade data from the OECD Library. I use “Bilateral Trade in Goods by Industry and End-use (BTDIxE), ISIC Rev. 4” data. See Appendix C.5 for additional details. For each Eurozone country and year, I calculate the share of that country’s goods imports originating from other Eurozone countries. I do not have bilateral services trade data. Hence, I assume that the share of total imports originating from Eurozone countries is equal to the share of total goods imports originating from Eurozone countries.   n,RoW These assumptions allow me to estimate VM,c /VMn,RoW for each country n. For a small number of observations this value is negative, which means I over-estimated the share of that   n,RoW country’s total imports from the Eurozone that were invoiced in Euros. If VM,c /VMn,RoW < 0, I set it to zero. Next, I estimate the share of the Eurozone’s imports invoiced in currency c by taking a   n,RoW weighted average of VM,c /VMn,RoW over the Eurozone countries, EU R,RoW VM,c VMEU R,RoW   n,RoW n,RoW n,RoW V V /V M,c M n M . P n,RoW V M n P ! = where VMn,RoW is still the value of country n’s imports that originate from non-Eurozone countries. I estimate VMn,RoW using bilateral goods import data by multiplying the share of country n’s goods imports from non-Eurozone countries with the total value of country n’s imports. I use import data from the OECD Library. See Appendix C.7 for more details. Thus, I estimate the share of the Eurozone’s imports from the rest of the world that are invoiced in currency c. I repeat the same calculations for exports. C.3 Spot and Forward Rates I use data for dollar-based spot and forward exchange rates from Datastream to construct nominal interest rates for each country. I use the World Markets PLC/ Reuters (WM/R) data provider within Datastream for my spot and forward exchange rate data. Identifiers for the forward rate data usually follow the form U Sisomm, where the U.S. dollar is the base currency, “iso” is the other currency’s ISO code, and “mm” is a two letter identifier for the maturity of the forward rate. The identifiers for the three month, six month, one year and two year forward rates are “3F”, “6F”, “1Y” and “2Y”, respectively. The identifiers for the spot rate data are of the form cccccc$, which is a six digit currency identifier followed by a dollar sign. C.4 Industrial Production Data I use the OECD index for Industrial Production, which is part of their Key Economic Indicators (KEI) series. The data are available at the monthly level. The indicators are prepared by national statistical agencies. Measures of industrial production typically include output from manufacturing, mining, electricity, gas and water industries. C.5 Bilateral Trade Data I use bilateral trade data to aggregate currency invoicing data, as well as other measure trade intensity, for the Eurozone countries. I use data from the “Bilateral Trade in Goods by Industry and End-use (BTDIxE), ISIC Rev. 4” database in the OECD Statistics Library. I use data from 1999 to 2014 for all countries within the Eurozone. Unforturnately, these data only cover goods trade and I would ideally like to observe bilateral trade in both goods and services. These data are mostly drawn from the UN Comtrade database. Their documentation is located: http://www.oecd.org/trade/bilateraltradeingoodsbyindustryandend-usecategory.htm C.6 Exchange Rate Peg Data I use exchange rate peg data using classification mechanism from Shambaugh (2004). The data are available from Shambaugh’s website and define whether or not countries have fixed exchange rates from 1960 to 2014. The data are available at the annual frequency. Among other variables, the data classify whether countries engage in hard exchange rate pegs or soft exchange rate pegs. C.7 Imports, Exports and GDP I construct measures of the import share of GDP and export share of GDP for each country in my sample. I use annual import, export, and GDP data from the Penn World Tables version 9. I use the following series: cgdpo for GDP, csh x for the export share of GDP, and csh m for the import share of GDP. For each country in the Eurozone, I estimate the share of the country’s imports (exports) with non-Eurozone countries using bilateral goods trade data. Implicitly, I assume that the pattern of each country’s trade in goods and services matches that country’s pattern of trade in goods alone. These calculations are analogous to those for the currency invoicing data, described in Appendix C.2. For each country, I calculate the share of its imports origination from each country, using bilateral trade data from the OECD. Next, I multiply these shares by the total value of goods and services imports in each country, from the Penn World Tables data, to estimate the value of each country’s total imports from each country. Using these data, I calculate the total value of imports into the Eurozone originating from non-Eurozone countries. I normalize the Eurozone’s imports by the total GDP of all countries within the Eurozone in each year. Finally, I average the annual data over all years in the sample period (1999 - 2014). I perform the analogous calculations for exports. C.8 Trade Costs I use measures of trade costs from Mayer and Zignago (2011). The data are located in the GeoDist database on the CEPII website: http://www.cepii.fr/CEPII/en/bdd modele/presentation.asp?id=6 I use the following variables from this database: distw and contig. I use Germany for the geographic region for the Euro. See the documentation in Mayer and Zignago (2011) for additional details. C.9 Price Stickiness Data I download the Rauch (1999) classification of goods from the author’s website: http://econweb.ucsd.edu/ jrauch/rauch classification.html The classification distinguishes between traded goods that are traded on an organized exchange, reference priced or differentiated products. The classification is provided at the 4-digit SITC rev. 2 level. I merge the Rauch classification with UN Comtrade data. From the UN Comtrade database, I download import and export data for each country in my sample for each year in my sample period (1999 - 2014) at the 4-digit SITC rev. 2 level. For each country, I calculate the value of each country’s imports classified as differentiated products and normalize by the country’s total imports. The “Diff. Share” variable is the average share of each country’s imports classified as differentiated products, where I take the average across years in the sample (1999 - 2014). For the Eurozone, I first calculate the share of all Eurozone imports from outside of the Eurozone that are differentiated products for each year in the sample, and then I average across years. C.10 Foreign Monetary Policy Rates I use foreign central bank policy rates from Global Financial Data. The data are provided by the Central Bank Interest Rate series. The data are found under the ticker code “IDcccD”, where “ccc” represents a three digit country code. The data are provided at the daily frequency. I observe central bank policy rate data for the following countries: Australia, Canada, Switzerland, Cyprs, Denmark, Estonia, the United Kingdom, Greece, Iceland, Israel, Japan, Korea, Lithuania, Latvia, Malta, Norway, Slovenia, Sweden and the Eurozone.