Exchange Rate Pass-Through and the Underlying Macro Cause

(1)

1

Exchange Rate Pass-Through and

the Underlying Macro Cause

Harald Wigerstedt

Spring Term 2020

(2)

2

Exchange Rate Pass-Through and the

Underlying Macro Cause

Abstract

Monetary Policy is key in controlling inflation. Yet today we are stuck in a low interest rate environment and unable to boost inflation. Exchange rates has a direct effect on prices and has important implications when forecasting inflation. Previous studies have shown that the underlying cause triggering the exchange rate movement matters. I have created various domestic and global variables to evaluate how these causes effect exchange fluctuations and pass through to consumer prices. To value exchange rate pass-through I constructed structural factor-augmented vector-autoregressions (SFAVAR) for 31 countries. The result indicates that pass-through is approximately zero in regions with credible MP authorities and consequently no conclusions about the underlying cause could be made.

(3)

3

1. Introduction

Below I introduce the purpose and problem definition that led to this thesis.

What is inflation and why is it so important? Inflation is the measure of certain goods’ price increase. To monitor this price development and keeping it relatively low and stable is paramount for the economy we are living in, to prevent income redistribution, negative real interest rate, increased cost of borrowing, diminishing business competitiveness, difficulties to predict inflation expectations, the list goes on. But the single and most important aspect of inflation is that it is an incentive to not postpone consumption. If it is cheaper today than it is tomorrow and your money is worth more today than it is in the future, you will continuously boost the economy by not putting consumption on hold. To ensure that the price development in an economy is stable and at its target, two main players are in control. Fiscal policy and monetary policy (MP).

(5)

5

Investigating the relationship between exchange rates and consumer prices, I hope to shed some light on what might have caused the recent MP to not be as effective as during previous decades.

With this background I hope to answer: Does pass-through from exchange rates to inflation depend on the underlying macro cause for the currency movement?

The phenomena of exchange rates affecting inflation has its own name, exchange rate pass-through (ERPT) which is a measure of how sensitive international prices are to changes in exchange rates. In this paper, ERPT is the change in country-specific inflation to the response of a change in nominal exchange rates. Based on previous studies by Forbes, Hjortsoe and Nenova (2018) and Ha, Stocker and Yilmazkunday (2019) who emphasizes the importance of interpreting the underlying shock causing the exchange rate to move, I have constructed various global and domestic variables to interpret differences in ERPT depending on the macro characteristic hitting the domestic market. See table 1 in Appendix 1. Global variables are inflation, interpreted as global demand, gross domestic product, interpreted as global supply and oil, used to control for changes in global prices. The domestic variables are

inflation representing domestic demand, gross domestic product representing domestic supply and short interest rates representing domestic monetary policy.

Although studies have shown a decline in ERPT because of price discrimination, credible MP, nominal rigidities and cross-border supply chains, ERPT is still researched in recent articles because of the important implications it has on MP. Because currency movements and the expected impact it has on consumer prices is important for central banks in order to implement MP and respond correctly to external shocks.

This study adds to recent strands in the fields of exchange rate movements and pass-through ratios, realising the importance of understanding and interpreting the underlying cause for the currency movements. The empirical strategy and econometric modelling have also been considered and therefore the Factor augmented vector autoregression (FAVAR) is

implemented. I estimate unobserved factors using two-stage principal component analysis and include them in a final structural factor-augmented vector-autoregression (SFAVAR). I run two country-specific regressions for each of the 31 OECD (Organisation for Economic Co-Operation and Development) countries using 79 observations. I analyse ERPT in one

(6)

6

(7)

7

2. Previous studies

In this section, I present some of the articles that worked as a basis for my study and helped me construct and enhance my research in the field of pass-through.

Taylor (2000) investigates the general effect of the inflationary environment on the pricing behaviour of firms. He uses a microeconomic model of price setting and by doing so he examines if lower/stable inflation is a factor behind lower pass-through, a decrease in pricing power. He concludes that lower pass-through should not be taken as exogenous to the

inflationary environment. Because ERPT is or should be endogenous when forecasting inflation, it also effects monetary policy. Taylor (2000) talks about the importance of the degree of pass-through when tightening or expanding monetary policy because of a change in exchange rates and that the coefficient in monetary policy rule depends on the degree of pass-through. He also refers to Ball (1999) who says that the same is true for monetary policy operating procedures that depend explicitly on the forecast of inflation.

Campa and Goldberg (2005) builds on Taylor’s arguments regarding pass-through issues. By researching the role it plays on monetary policy and exchange rate regime optimality in general equilibrium models. Campa et al., (2005) uses a log-linear regression and

distinguishes between producer currency pricing and local currency pricing of imports. The difference was large and whilst the exchange rate movement had close to none effect on retail prices the change in border prices was big even for a large country such as USA. They

concluded that aggregate pass-through rates should be separated into two parts, border- and retail prices1 something that Burstein and Gopinath (2013) also found compelling case for. Burnstein et al., (2013) reiterate that unlike consumer prices, import price indices are

constructed differently across countries. Devreux et al., (2001) stated that if the exporter sets their price in the country’s currency with the most stable MP. Import prices should be more stable in countries with stable MP, Campa et al., (2005) tested this and found it significant. Dornbusch (1987) researched country size and pass-through effect, ERPT should be high if exporters are large in numbers relative to domestic competition. This was concluded to be insignificant by Campa et al., (2005). Future research such as transmission effect from border

(8)

8

to retail and absorption of international fluctuations and other local value-added components that links import to retail prices was encouraged.

ERPT varies depending on the shock hitting the economy which is important for

policymakers when responding accurately to inflationary pressures Corbo and Casola (2018). They start by highlighting limitations of univariate approaches (predominantly explained by its own exogenous shock) when looking at ERPT. Because of that, they construct four domestic shocks (using domestic GDP, consumer prices, policy rate and exchange rate) and two global shocks (foreign GDP and consumer prices) to explain variation in exchange rate movements. Using a structural Vector Autoregression (SVAR) they found that the pass-through to consumer prices from a change in exchange rates caused by a shock was relatively small. The study is implied on a small open economy, Sweden and with estimated ERPT varying from positive to negative values they conclude that the average pass-through is roughly zero.

Forbes, Hjortsoe and Nenova (2018) assume a standard open economy model that expresses the change in exchange rate movement by interpreting the underlying cause for such a movement. Doing so they incorporate the direct shock causing the exchange rate to change and thereby passing through the fluctuations to consumer prices. They use a SVAR

framework for a small open economy and apply it on UK. Their results show that the

underlying cause for exchange rate movements implies differences in pass-through. Domestic demand shocks were accompanied by less pass-through than domestic MP, with such a shock a greater amount was passed through.

(9)

9

indirect channels consisting of mark-ups, wage settings and import prices (Bacchetta and Van Wincoop 2003; Ito and Sato 2008; Burstein and Gopinath 2014; McCarthy 2007).

Empirical analysis of pass-through to consumer prices has shown a decline in ERPT. Further reading on the subject can be done in the articles introduced below. Structural factors

associated with a lower sensitivity of domestic prices to ER movements, degree of

competition importing exporting firms (Amiti, Itskhoki and Konings 2016), frequency price adjustments (Devereux and Yetmen 2003; Corsetti, Dedola and Ledue 2008; Gopinath

and Itskhoki 2010), competition of trade (Campa and Goldberg 2010), level of participation in global value chains (Georgiadis, Gräb and Khalil 2017), the share of trade invoiced in

(10)

10

3. Theoretical framework

In this section, I describe the fundamentals of exchange rates, macroeconomic implications on ERPT and exporting firms pricing behaviour. I move on to the part “Small open

economy” to define the cause of the currency movement and the impact it might have on firms' pricing. Lastly, I highlight some of the reasons for partial exchange rate pass-through.

The theoretical framework in this thesis is based on the work of (Campa and Goldberg 2005; Forbes, Hjortsoe and Nenova 2018; Ha, Stocker and Yilmazkunday 2019).

3.1 Law of one price

The law of one price is an economic concept stating that identical assets or commodities will have the same price globally. It rests upon the assumption that there is a frictionless market, where there are no transportation costs, legal restrictions, or transaction costs, there is no manipulation by buyers or seller, and the currency exchange rates are the same. The law of one price exists to some extent. When arbitrage opportunities arise agents would purchase the asset in a cheaper market and sell it in a market where a higher price could be obtained. Over time, market forces would align the prices and create an equilibrium. The law of one price is the foundation of purchasing power parity.

Purchasing power parity states that the value of two currencies is equal when a basket of identical goods is priced the same in both countries. It allows comparison of the purchasing power between various world currencies. For example, consider a basket of goods and

(11)

11

3.2 Macroeconomic implications on ERPT

Depending on the shock hitting the domestic market, prices will change more or less easily depending on the characteristics of the market resulting from the shock. Using the aggregate demand-aggregate supply model (AD-AS model) I am able to display national income determination and changes in the price level. By interpreting the AD-AS model, the macroeconomic characteristics and its implications for ERPT is presented.

AD is an aggregate of consumption, investment, government spending and net exports. A change in any of these components will cause the AD curve to shift and create a new short-run macroeconomic equilibrium (see appendix 1). A rise in either component causing AD to increase will result in higher real gross domestic product (rGDP), lower unemployment (UE) and higher price level (PL). A positive AD shock will most likely cause ERPT to be lower than before the shock. The macroeconomic characteristics causes the market to be favourable for an exporter invoicing in the currency of the domestic market where it sells its goods. Higher growth, lower unemployment and higher price levels are all indicators to higher

demand and increasing consumption. If the exporter would experience a weaker exchange rate (home currency is valued lower), they could lower their price less than before the shock because of higher aggregated demand. Consider the opposite and AD decreases. Then rGDP drops, unemployment increases and the price level dampens. With such market characteristics the exporter is more likely to pass through a greater extent of the change in exchange rate (ER), resulting in a higher ERPT.

The second shock in the AD-AS model comes from the supply side of the economy,

represented by the short run aggregate supply curve (SRAS). A shift in SRAS can arise from subsidies for businesses, productivity, input prices, expectations about future inflation and taxes on businesses. However, the result from a supply side shock varies in comparison to a demand shock (see appendix 1). A positive supply side shock, a shift in SRAS, will result in higher rGDP, lower UE and lower PL. In this case higher growth and lower UE would support lower ERPT, but downward pressure on inflation would also make the exporter more exposed to competitive pricing which could increase ERPT. The market characteristics of a negative supply shock is vice versa and a drop in rGDP along with a rise in UE would

(12)

12

3.3 Exporter firm’s pricing dynamics

To understand ERPT and the dynamics of pricing behaviour, the microfoundation of

exporting firms are essential. The import price (𝑃_𝐹,𝑡) is a function of the exchange rate (𝑠_𝑡), mark-up (𝑚𝑘𝑢𝑝_𝑡) and marginal cost (𝑚𝑐_𝑡), using lowercase letters reflecting logarithms:

𝑝𝐹,𝑡= 𝑠𝑡+ 𝑚𝑘𝑢𝑝𝑡+ 𝑚𝑐𝑡 (1)

The price of the good imported is set by the exporter’s cost of making that good and the mark-up which should result in a profit. The price that the importer faces (assuming the price is set in exporter’s currency) is converted using the exchange rate (domestic currency per unit foreign currency). However, depending on market competition and individual pricing

decisions, firms can compete by changing their mark-up to gain market shares (or prevent loss of market shares). Equation (1) illustrates that the correlation between import price and

nominal exchange rate depends on the exporter’s mark-ups’ and marginal costs change when the exchange rate changes. The pass-through to import prices will be complete (100%) if the mark-up and marginal cost do not co-move with the exchange rate. See Appendix 1 for an example.

Firms' pricing is determined from expected future marginal costs, expected demand and predicted future competitive pressures2. These determinants of exporting firms pricing will be affected by various shocks, the extent to which pricing is changed will also vary depending on the shock hitting the market. The degree to which ERPT is reflected in import prices is

thereby determined by the nature of the shock. The price difference is passed through because of the fluctuation in the exchange rate (denoted as 𝑠_𝑡 in equation (1)) is the change in 𝑚𝑘𝑢𝑝_𝑡 and 𝑚𝑐_𝑡. To better understand the changes in these pricing factors when the domestic market is hit by either a domestic or global shock, I move on to the section below and consider a small open economy such as Sweden.

3.4 Small open economy

Suppose Sweden would be hit with a domestic shock, either demand, supply or MP. The marginal cost of a foreign firm who exports to Sweden will not be affected by the domestic

(13)

13

shock (according to the assumptions regarding a small open economy)3. However, the mark-up is sensitive and depending on what currency the good is priced in, the exporting firm could be exposed to exchange rate risk. Let us say the exporter invoice in the market it sells its goods. If faced with price rigidities the mark-up will absorb the exchange rate fluctuations because the price is “sticky” and not likely to change quickly. In a market where the exporter can adjust the price and simultaneously respond to any changes in demand and competitor pricing that might also be caused by ER changes, is a market with greater ERPT.

More explicitly, if the domestic currency, in this case, Swedish krona, appreciates against the exporter’s currency and the good of the exporter is priced in the domestic currency, the exporter carries the exchange rate risk. The cause for the appreciation is important as stated above. Suppose a positive domestic demand shock is responsible for the appreciation. Then the exporter will face a higher demand in the domestic market with increasing prices and wages4_{. In this case, the exporting firm might reduce the price a little if not nothing. Resulting}

in a low ERPT. If the appreciation came from contractionary MP, the domestic market would not have as favourable market characteristics (lower demand, and muted price and wage increases)5. The exporter will reduce the price more and thus passing through a greater extent of the ER change, resulting in a higher ERPT.Lastly, consider a domestic supply shock making the domestic currency appreciate. In this case, the effect on ERPT is ambiguous and dependent on whether the income effect or substitution effect is dominant6. If the positive supply shock alongside a dominant income effect, exporters will see an increase in demand and price reductions will not be necessary. But simultaneously price and wage pressures could decrease and result in exporters having to lower their price too. Generating higher ERPT, the interplay between the two effects decides the amount of pass-through.

3_{A small open economy’s policies does not alter world prices, interest rates or income. Consider a small open}

economy borrowing at the world interest rate (exogenus to the domestic market). The world interest rate will not change, because the domestic market does not hold enough capital to change the equilibrum on the interantional financial market.

4_{In case of a permanent positive demand shock, wages will have to be revised upward.}

5_{Higher interes rates will hault investments, thus having a negativ impact on demand and therfore prices and}

wages want increas as much as befor the shock.

6_{The change in income effect is the change in consumption of a good based only on the income of the}

(14)

14

Next, instead of a domestic shock, a global shockis the cause for the appreciation of the domestic currency. The marginal cost of the foreign firms will be affected this time7. The same still applies regarding the impact the appreciation might have on expected future

demand, future competitive pressures and expected marginal cost. Seeing these factors are the cause to which extent exporters change their mark-ups, pass-through the change in ER and marginal costs to prices. The characteristics of a domestic shock on the economy are short-lived and not as enduring in comparison to a global shock. Therefore, we might expect a higher degree of pass-through with a global shock. Seeing that a global shock is more permanent.

3.5 Partial exchange rate pass-through

Incomplete pass-through8 may arise in internationally traded markets where trade frictions and firms’ ability to price discriminate differ between regions. Nominal rigidities can also help explain the persistence of incomplete pass-through over time and diminishing ERPT through the product chain.

Exporters’ ability to obtain different pricing in different segments of international markets play an important role in ERPT. Pricing-to-market literature has its roots in published papers by Krugman (1987) and Dornbusch (1987). They place monopolistic firms at the centre of international price discrimination. Producers’ ability to alter mark-ups between regions to account for demand functions and price elasticities creates a wedge incomplete pass-through. Also, segments with heterogenic agents (consumers) creates opportunities for price

discrimination, flexible demand systems opens up for optimal pricing.

International trade models of cross-border product chains have further developed partial ERPT. With larger and more competitive firms, gaining greater market shares throughout regions with different currencies, input factors are outsourced. Creating a more complex global value chain and putting downward pressure on average ERPT

(Amiti, Itskhoki and Konings 2014; Gopinath and Neiman 2014).

(15)

15

With local-currency pricing experiencing nominal rigidities the currency of invoice is crucial for ERPT. Producers’ have the choice to invoice either in trade partners’ currency or domestic currency, dependent on competition in the foreign market and credible MP (Devreux and Engel 2001; Gopinath and Itskhoki 2010; Bacchetta and Van Wincoop 2005).

Local input factors that are not subject to ER risk tend to decrease ERPT to consumer prices. Even greater effect bears the distribution channel, creating a significant difference between import price and retail price.

Many studies the relationship between varying ERPT and country characteristics as an underlying factor. Openness to trade and financial transactions, competitiveness of trade, inflation and ER volatility and credible MP. Further reading regarding country characteristics is cited to (Campa and Goldberg 2005,2010; Taylor 2000; Gagnon and Ihrig 2004;

(16)

16

4. Data and method

In this section, I will briefly present the data. Then shortly introduce the Vector

autoregression (VAR) which is the foundation at which the Factor augmented VAR rests upon. Then, the empirical strategy is described and a short description of the tests that were run to ensure the best possible model.

4.1 Data

The sample includes 31 countries over 19 years with 79 quarters, ranging from 2000Q1 to 2019Q3 with a total of 218 variables. The entities chosen are the OECD countries. Five were eliminated because of missing data9.

The following definition of variables were put in the FAVAR model. Country specific inflation is quarter-on-quarter consumer price indices. Country specific output growth is quarter-on-quarter seasonally adjusted real GDP growth. Domestic interest rates are annualized short- term interest rates. Nominal effective exchange rate is quarter-on quarter trade-weighted nominal exchange rate. Global output growth (represented by G20 countries) is quarter-on-quarter seasonally adjusted real GDP growth. Global inflation (represented by G20 countries) is on-quarter consumer price indices. Oil price growth is the quarter-on-quarter growth rate of nominal oil prices average of Dubai, WTI and Brent. See table 1 in Appendix 1.

4.2 VAR model

Vector autoregression (VAR) is the baseline for macroeconomic forecasting. A VAR is a set of k time series regression, where the regressors are lagged values of all k series. It extends the univariate autoregression to a vector of time series variables. With two-time series variables, 𝑍_𝑡 and 𝑉_𝑡, the VAR consists of two equations:

𝑍𝑡 = 𝛽10+ 𝛽11𝑍𝑡−1+ ⋯ + 𝛽1𝑝𝑍𝑡−𝑝 + 𝛾11𝑉𝑡−1+ ⋯ + 𝛾1−𝑝𝑉𝑡−𝑝+ 𝑢1𝑡 (2)

𝑉_𝑡= 𝛽₂₀+ 𝛽₂₁𝑍_𝑡−1+ ⋯ + 𝛽_2𝑝𝑍_𝑡−𝑝+ 𝛾₂₁𝑉_𝑡−1+ ⋯ + 𝛾_2−𝑝𝑉_𝑡−𝑝+ 𝑢_2𝑡 (3) where the coefficients 𝛽 and 𝛾 are unknown and 𝑢_1𝑡 and 𝑢_2𝑡 are error terms. The coefficients are estimated using OLS. To make sure that the time series regression is optimal four

(17)

17

assumptions needs to be satisfied. The first assumption is that the error term (𝑢_𝑡) has a

conditional mean zero given the lags of all regressors. The second one assumes that the data is stationary, which means that the distribution of the data is the same today as in the past. Along with the stationarity assumption, weak dependence between the random variables is regarded. The third assumption is that large outliers are unlikely. The fourth one assumes that there is no perfect multicollinearity present.

4.3 FAVAR model

Bernanke, Boivin and Eliasz (2005) first introduced the FAVAR model after contributing to the work of Sergent and Sims (1997), Geweke (1997) and Stock and Watson (1999). FAVAR eliminates the difficulty of sparse information sets and the degree-of-freedom problem in traditional VAR models. In order to do so, FAVAR includes unobserved low-dimensional factors into the autoregression. Small amounts of factors to summarize useful information in a large number of indicators have been used in many papers, for example, Bernanke and Boivin (2003) and Stock and Watson (2002).

4.3.1 FAVAR methodology

Let 𝑌_𝑡 be a M×1 vector of observable economic variables assumed to drive the dynamics of the economy. Additional economical information not captured by 𝑌_𝑡 might be relevant for modelling this time series. This information is summarized in a K×1 vector of unobserved factors, 𝐹_𝑡, where K is small. Consider 𝐹_𝑡 as capturing fluctuations in unobserved macro influences or reflecting theoretically motivated concepts, such as “economic activity” or “price pressures” that can’t be reflected in one or two series but can be explained by a wide range of economic variables.

Assume that the joint dynamics of (𝐹𝑡′, 𝑌𝑡′) are given by the following transition equation:

𝐴0[𝐹_𝑌𝑡

𝑡] = 𝐴(𝐿) [

𝐹𝑡−1

𝑌_𝑡−1] + 𝑢𝑡 (4)

(18)

18

The difference between the FAVAR and the VAR model is the presence of unobservable factors. Equation (4) is a VAR in (𝐹_𝑡′_{, 𝑌}

𝑡′). This system is reduced to a Standard VAR in 𝑌𝑡 if

the term of A(L) that relate 𝑌_𝑡 to 𝐹_𝑡−1 are all zero: otherwise I will refer to equation (4) as a FAVAR.

𝐹_𝑡 is not observed and cannot be directly estimated. However, we interpret factors as representing forces that potentially effects many economic variables. With the collected variety of economic time series, we hope to infer something about the factor 𝐹𝑡. Because of

the assumed informational time series 𝑋_𝑡 being related to the unobserved 𝐹_𝑡 and observed 𝑌_𝑡 by an observation equation of the form:

𝑋_𝑡= Λ𝑓_𝐹

𝑡′+ Λ𝑦𝑌𝑡′+ 𝑒𝑡 (5)

where Λ𝑓 and Λ𝑦 are factor loadings of dimension N×K and N×M, respectively. The errors are mean zero and largely idiosyncratic. Even though 𝐹𝑡 and 𝑌𝑡 in general might be correlated,

equation two captures the idea that both factors drive the dynamics of 𝑋_𝑡. Conditional on 𝑌_𝑡, 𝑋_𝑡 is a blurred measure of the underlying unobserved factor 𝐹_𝑡.

4.3.2 Why the FAVAR model

Firstly, macroeconomics is blurry, revised multiple times and never free from measurement errors. Secondly, theoretical concepts do not necessarily align with specific data series. In Bernanke et al., (2005) even inflation and output are considered unobserved for the MP authorities and econometricians and only the federal fund rate is observed and contained in 𝑌_𝑡 . The advantage of FAVAR (which is clearly demonstrated in the FAVAR methodology section in Bernanke et al., (2005)) is the possibility to accommodate different assumptions about the information set used.

With the FAVAR model, I can incorporate as much information as possible into the model while keeping the advantage of parasimonius models10. Standard VAR analysis is usually confronted by the degree-of-freedom problem and therefore only consists of six to eight

10_{A parasimonius model is a model that accomplishes the highest degree of prediction or explination with as}

(19)

19

variables. This is avoided with FAVAR and lets us exploit information from our indicators even though N is large in some empirical macro models11.

VAR models tend to predict the opposite of what might be suggested in theory regarding inflation and MP, this is what Sims (1992) discovered when introducing the price puzzle phenomena. Sims (1992) proposed that MP authorities possess more information than what is included in the VAR model. Hence, central banks recognize inflationary pressure in an earlier stage than in the VAR model and apply contractionary MP to dampen inflation. But because the VAR model recognise the increase in inflation with a time lag, the price level increases in the model after the contractionary MP and even less without the monetary authorities.

FAVAR can minimize if not even eliminate the price puzzle completely.

A VAR model consists of a policy variable and a few key variables 12 _{which are supposed to}

explain the policy variable. As opposed to FAVAR, that includes factors representing a much larger data set13_{. Therefore the FAVAR model should be able to perform better than the VAR}

model and give a better interpretation of the economy. Another downside of the VAR model is that only a few set of macroeconomic variables can be calculated from MP shocks. With the FAVAR model, we can calculate all macroeconomic variables contained in 𝑋_𝑡. To do so, all variables in the data set can be represented as linear combinations of the estimated factors14_.

4.3.3 Criticism of FAVAR

The drawbacks of the FAVAR model should not be concealed. Computing the model and its factors goes through lots of stages and I create one model for each country. Computational errors may therefore arise and decisions regarding the number of factors do vary between country models. Information assumptions as to what variable is considered observable and expectations on how each variable reacts to MP shocks also has its implications when estimating 𝐹̂_𝑡. Ending up with biased estimates of the factors estimation will consequently lead to a biased result.

11_{Bernanke, Boivin and Eliasz (2005)} 12_{Sims (1992) uses six variables.}

(20)

20

The main drawback of using factors analysis is the economical interpretation. Considering that factors are created from observable variables the information embedded in each factor needs to be interpreted with some sort of economical meaning. When estimating 𝐹̂_𝑡 in my model, I assumed interest rates and oil price growth as observable and contained in 𝑌_𝑡. Thus, the economical interpretation would be an inflation factor or as a cyclical factor taking into consideration the bust or boom of the economy.

4.4 Empirical strategy

4.4.1 Estimation of 𝐹̂

_𝑡

and ultimately a SFAVAR

The FAVAR model consists of the variables 𝑌_𝑡 and the factors 𝐹_𝑡 that summarizes the data contained in 𝑋𝑡, which are all the domestic and global variables. To estimate the FAVAR

model I use SVAR but prior to this the unobserved factor 𝐹̂_𝑡 had to be estimated. Two-stage principal component analysis was implemented to estimate 𝐹̂_𝑡. Then 𝐹̂_𝑡 replaced 𝐹_𝑡 and the model could be estimated.

Assumptions regarding the variables in 𝑋𝑡 had to be made before the estimation. I assumed

that only interest rates and oil growth were contained in 𝑌_𝑡 and therefore observable. The remaining five variables contained in 𝑋_𝑡 was divided into slow- and fast-moving. The difference between the slow- and fast-moving variables are the sensitivity to 𝑌𝑡, economic

activity and monetary policy, slow-moving variables are not assumed to react to shocks in the short-run (based on the work of Fux (2008) and Berggren (2017)). The empirical strategy can be divided into four parts. The first three parts involve the estimation of 𝐹̂𝑡 which was then

used in the last part, estimating the FAVAR model using SVAR.

Firstly, I performed principal component analysis on all variables in 𝑋_𝑡. After deciding the number of vectors that is sufficient to describe the data in all of 𝑋_𝑡 (denoted as 𝐶̂_𝑡), I computed the estimation of the slow-moving factors in 𝑋𝑡 (denoted as 𝐹̂𝑡𝑠). Secondly, I fit a

linear regression model with 𝐶̂𝑡, 𝐹̂𝑡𝑠 and 𝑌𝑡. The third and final step to estimate 𝐹̂𝑡 is to

calculate the factor estimation 𝐹̂_𝑡 = 𝐶̂_𝑡− 𝑏̂_𝑌𝑌_𝑡. The correlation of the principal components included in the 𝐹̂𝑡 factor was very high and therefor only one principal component was needed

in the last step of the empirical work. To make sure that multicollinearity was avoided oil price growth was excluded from the SFAVAR.

(21)

21

4.4.2 Testing the model

Stationarity is an assumption that must be upheld with time-series data. A time series 𝑍𝑡 is

stationary if its probability distribution of (𝑍_𝑠+1+ 𝑍_𝑠+2, … , 𝑍_𝑠+𝑇) does not depend on s regardless of the value of T; otherwise 𝑍_𝑡 is considered nonstationary. In practice this is hard to get, however, weak stationary is acceptable and implies that only the mean and variance is constant over time, and no periodic fluctuations Stock and Watson (2015). To ensure that at least weak stationary was met, I plotted each time series looking for patterns of

autocorrelation, calculated mean, -variance and constructed an autocorrelation plot.

To conclude that the regressors had a predictive power on the dependent variable a Granger test was run. It tests the hypothesis that the coefficients on all the values in a time series variable (for example, the coefficients related to 𝑉_1𝑡−1+ 𝑍_1𝑡−2, … , 𝑉_{1𝑡−𝑞1}) are zero. This null hypothesis implies that the regressors have no predictive content on 𝑍𝑡. The result of the test

had various results and half of the SFAVAR’s showed insignificant results.

The right number of lags that are used in the model is highly important so that overfitting is avoided. I use information criterion to choose the optimal number of lags. Akaike information criterion (AIC) is the most commonly used and is implemented here as well. AIC is a

(22)

22

5. Result

Below, the result from the two country-specific regressions are explained. In the first section, the regression explaining the exchange rate and the underlying cause for the currency movement is presented. In the second section, I present the pass-through to consumer prices. Only the countries having significant effect from exchange rates to consumer prices were analysed. A total of thirteen country-specific regressions showed a significant result of pass-through and twelve out of those had significant cause for the exchange rate movement. See Appendix 2 for all thirteen country-specific regressions.

5.1 Exchange rate and the underlying cause for the currency

movement

Considering that this study focuses on the underlying reason for an exchange rate movement, determining the cause for either the appreciation or depreciation of the domestic currency is described first. To prevent missing any underlying cause for exchange rate movements lags up to 7 (nearly two years) are considered.

MP showed significant result in seven out of the twelve approved country-specific regressions15. In five out of those seven, tightening MP showed significant result for an appreciation of the domestic currency. Domestic demand proved significant only four times and an increase in domestic demand resulted in an appreciation of the currency16. Domestic supply showed a strong result17_{. Out of nine significant results, eight confirmed appreciation}

of the currency.

Global variables had various results. Global demand was the variable with the lowest amount of significant results18. Out of three significant results, two regressions indicated that a higher global demand would appreciate a country’s home currency. Global supply was significant in five cases19. Four out of those five times it had a negative impact on the exchange rate and thereby resulting in depreciation of the home currency. The last global variable (PC1) was calculated using two-stage principal component analysis. This variable is interpreted as a global factor for economic activity. PC1 was significant in six regressions and had a negative

15_{See table 4, 10, 14, 22, 26, and 28 in Appendix 2.} 16_{See table 6, 14 and 22 in Appendix 2.}

17_{See table 4, 6, 10, 12, 14, 16, 18, and 28 in Appendix 2.} 18_{See table 6 and 22 in Appendix 2.}

(23)

23

impact on the exchange rate in five of those20. Thus, higher economic activity (more

consumption, economic growth and increasing interest rates) should result in depreciation of a countries currency.

5.2 Exchange rate pass-through

Thirteen regressions had a significant result of exchange rates affecting inflation21_{. The result}

was scattered. Almost half of the regressions indicated that an appreciation of the home currency leads to higher inflation and the other half led to lower inflation with an appreciating currency. The average value of pass-through from a change in exchange rates to consumer prices was 0,082.

I am only interested in the short-term effect of exchange rate fluctuations on consumer prices. Therefore only the one lagged value is considered. It varies between 0,61 and -0,12 with a 0,05-significance level. It kept varying over zero with eight regressions showing a positive effect and five with a negative effect.

20_{See table 6, 8, 12, 18 and 22 in Appendix 2.}

(24)

24

6. Discussion

Below I compare and discuss the result and how it relates to theory and previous research. I also present thoughts and discussions about potential problems and opportunities for improvement.

Contemplating on the fact that only half of the country-specific regression showed a significant result of pass-through, the effect of the exchange rate might be disappearing in advanced MP regions. The countries that make up the OECD organisation are heavily

concentrated in the Euro area and the EU. A market that facilitates cross-border trade and has implemented one currency to promote trade, encourage investment and enhance mutual support. Creating a highly open market influenced by higher competition and value-chains across borders. With invoicing mostly done in Euro controlled by the creditable ECB. According to previous research, these market characteristics are underlying factors to less ERPT (Amiti, Itskhoki and Konings 2014; Gopinath and Itskhoki 2010; Bacchetta and

Van Wincoop 2005; Taylor 2000; Gagnon and Ihrig 2004; Amiti, Itskhoki and Konings 2016; Forbes, Hjortsoe and Nenova 2017; Corbo and Casola 2018). The high amount of EU

members and trade amongst these countries has probably put pressure on ERPT and is probably the reason why so few regressions show a significant effect of ERPT.

Considering that the result of ERPT was roughly zero, determining the underlying cause for different pass-through ratios became impossible. However, when averaging the significant coefficients of both country-specific regressions, I can analyse pass-through depending on what macroeconomic variable that changes. See Appendix 2, table 2. Studying these averages, domestic variables behave relatively well to theory and previous studies. Both Ha’s et al., (2019) and Forbes’s et al., (2018) result indicated that domestic demand shocks were accompanied by less ERPT. This aligns well with theory; a positive demand shock is

(25)

25

relative to the other domestic variables indicating towards a dominant substitution effect. Contradicting to the result of Ha et al., (2019), who found domestic supply to be scattered and nonconclusive. Studying the global causes, a lower amount of pass-through is present.

Opposing theory which implies that a global shock would be longer-lasting and more effective thus passing through more of the price change. However, the findings of Ha et al., (2019) align well with this, they found scattered effects when analysing global causes. I had hoped for a better result where the underlying cause for the exchange rate movement indicated towards certain amounts of pass-through ratios. The number of significant

regressions were too few and made it impossible to clarify any pattern. I am confident that a larger amount of observations and a greater number of entities with more variety between them would have yielded a better result. The time spanned by the study (2000q1-2019q3) was too short and should have been further, preferably all the way back to the collapse of the Bretton Woods system in 1971. This could have contributed to the study of partial exchange through but also enhanced the study by comparing the effect between external shocks and pass-through. The variety and number of entities were too similar and too few. There were 31 entities, all OECD countries. When half of all country-specific regressions turned out to be insignificant, I was left with a vague and scattered result. Firstly, this could have been avoided with more entities, even though some might be insignificant, the likelihood of a pattern

between pass-through and macroeconomic differences could have been revealed. Secondly, the variety between the OECD countries was minor. Introducing emerging markets could have discovered differences in pass-through between industrial countries and developing countries. Differences in country characteristics between the two such as lower MP

credibility, less openness to trade and higher exchange rate volatility might make emerging markets more exposed to pass-through depending on the shock. All in all, the data used in the research should have been more extensive.

(26)

26

7. Conclusion

In this last section, I make my final remarks and conclusions regarding the result, method and thoughts of the study and subject. Lastly, thoughts on future research are presented.

A flexible and credible MP is important to prevent inflation volatility and anchor inflation expectations. The transmission effect of MP has direct implications from exchange rates and the purpose of this thesis was to identify if the underlying cause for the exchange rate movement had any effect on the amount of price difference that is passed through. Earlier research by Forbes et al., (2017) and Ha et al., (2019), found significant results that the underlying shock matters when studying ERPT. Thereby highlighting the importance of exchange rate fluctuations as a result of MP and different global and domestic shocks. The goal was to find significant proof of differences in pass-through ratios depending on the underlying macro cause. Shedding some light on what might happen in the transmission effect and thus confirming that MP authorities need to account for what type of economic shock that hits the economy before responding accordingly with MP. No such conclusions could be drawn because of the result.

After running country-specific regressions I can conclude that the domestic state of the economy has broader implications for the exchange rate than changes in the global state. Not so surprising considering that with a global shock the appreciation of the home currency is the equivalent depreciation of the foreign currency. Sadly, the different underlying causes for currency movements could not be associated with any low or high pass-through. The values of ERPT kept varying between positive and negative values indicating that pass- through is approximately zero.

I would have liked to make a more thorough study in the field of pass-through if I would have been given more time. Unfortunately, the empirical strategy was much too time-consuming, resulting in over 2000 rows of code and a total of 314 variables when adding my principal components. In hindsight, the study suffered because of this. Instead, time could have been spent increasing the amount of observations and entities, enhancing the probability of a better result where more significant conclusion could be drawn.

It was a shortcoming to not include entities with more profound country characteristic

(27)

27

(28)

28

8. References

Amiti, M., O. Itskhoki, and J. Konings. 2014. “Importers, Exporters, and Exchange Rate Disconnect.” American Economic Review 104 (7): 1942-78.

Amiti, M., O. Itskhoki, and J. Konings. 2016. “International Shocks and Domestic Prices: How Large Are Strategic Complementarities?” NBER Working Paper 22119, National Bureau of Economic Research, Cambridge, MA.

Bacchetta, P., and E. van Wincoop. 2003. “Why Do Consumer Prices React Less Than Import Prices to Exchange Rates?” Journal of the European Economic Association 1 (2-3): 662-70. Bacchetta, P., and E. van Wincoop. 2005. “A Theory of the Currency Denomination of International Trade.” Journal of International Economics 67 (2): 295-319.

Ball, L. 1999. “Efficient Rules for Monetary Policy.” International Finance 2 (1): 63-83. Belviso , F., and F. Milani. 2006. “Structural Factor- Augmented VARs (SFAVARs) and the Effects of Monetary Policy.” The B.E. Journal of Macroeconomics 6 (3): 1-46.

Berggren, E. 2017. “The Estimation of Factors in FAVAR models.” Unpublished, Lund University, Lund.

Bernanke, Ben S., and J. Boivin. 2003. ”Monetary policy in a data- rick environment.”

Journal of Monetary Economics 50: 525-546.

Bernanke, Ben S., J. Boivin,. And Piotr S. Eliasz. 2005. ”Measuring the Effects of Monetary Policy: A Factor- Augmented Vector Autoregressive (FAVAR) Approach.” Quarterly

Journal Of Economics 120 (1): 387-422.

Burstein A., and G. Gopinath. 2013. “International Prices and Exchange Rates”. Handbook of

International Economics 4 (4): 391-451.

Burstein, A., and G. Gopinath. 2014. “International Prices and Exchange Rates.” Handbook of

International Economics , edited by G. Gopinath, E. Helpman, and K. Rogoff, 391-451.

Amsterdam: Elsevier.

Campa, J., and L. Goldberg. 2005. “Exchange Rate Pass-Through into Import Prices.” Review

(29)

29

Campa, J., and L. Goldberg. 2010. “The Sensitivity of the CPI to Exchange Rates: Distribution Margins, Imported Inputs, and Trade Exposure.” Review of Economics and

Statistics 92 (2): 392-407.

Carriere-Swallow, Y., G. Bertrand, E. Magud, and F. Valencia. 2016. “Monetary Policy Credibility and Exchange Rate Pass-Through.” IMF Working Paper 16/240, International Monetary Fund, Washington, DC.

Casas, C., F. Diez, G. Gopinath, and P. O. Gourinchas. 2017. “Dominant Currency Paradigm.” NBER Working Paper 22943, National Bureau of Economic Research, Cambridge, MA.

Charnavoki, V., and J. Dolado. 2014. “The Effects of Global Shocks on Small Commodity-Exporting Economies: Lessons from Canada.” American Economic Journal: Macroeconomics 6 (2): 203-237.

Corbo, V., and P. Casola. 2018. “Conditional Exchange Rate Pass- Through: Evidence from Sweden”. Riksbank

Corsetti, G., L. Dedola, and S. Leduc. 2008. “High Exchange-Rate Volatility and Low Pass- Through.” Journal of Monetary Economics 55 (6): 1113-28.

Devereux, M. B., and C. Engel. 2001. “Endogenus Currency Of Price Setting In a Dynamic Open Economy Model.” NBER Working Paper No. 8559. National Bureau of Economic Research, Cambridge, MA.

Devereux, M. B., and J. Yetman. 2003. “Price Setting and Exchange Rate Pass-Through: Theory and Evidence.” In Price Adjustment and Monetary Policy: Proceedings of a

Conference Held by the Bank of Canada. November 2003, 347-71. Ottawa: Bank of Canada.

Dornbusch, R. 1987. “Exchange Rates and Prices.” American Economic Review 77 (1): 93-106.

Forbes, K., I. Hjortsoe, and T. Nenova. 2017. “Shocks versus Structure: Explaining

Differences in Exchange Rate Pass-Through across Countries and Time.” External Monetary Policy Committee Unit Discussion Paper 50, Bank of England, London.

(30)

255-30

75.

Fux, S. 2008. ” Measuring Monetary Policy - A Theoretical Foundation and an Application of the Factor Augmented Vector Autoregressive Approach”. Unpublished, Lund University, Lund.

Gagnon, J. E., and J. Ihrig. 2004. “Monetary Policy and Exchange Rate Pass-Through.” International Journal of Finance and Economics 9 (4): 315-38.

Georgiadis, G., J. Gräb, and M. Khalil. 2017. “Global Value Chain Participation and Exchange Rate PassThrough.” Unpublished, European Central Bank, Frankfurt.

Geweke, J.F., M. P. Keane, and D. E. Runkle. 1997. “Statistical inference in the multinomial multiperiod probit model”. Journal of Econometrics 80 (1): 125-165.

Gopinath, G., and O. Itskhoki. 2010. “Frequency of Price Adjustment and Pass-Through.”

Quarterly Journal of Economics 125 (2): 675-727.

Gopinath, G., and B. Neiman. 2014. “Trade Adjustment and Productivity in Large Crises.”American Economic Review 104 (3): 793-831.

Ha, J., M. Stocker., and H. Yilmazkunday. 2019. “Inflation and Exchange Rate Pass- Trough”. World Bank Policy Research Working Paper 8780 (1)

Ito, T., and K. Sato. 2008. “Exchange Rate Changes and Inflation in Post-Crisis Asian

Economies: Vector Autoregression Analysis of the Exchange Rate Pass-Through.” Journal of

Money, Credit and Banking 40 (7): 1407-38.

Krugman, P. 1987. “Pricing to Market When the Exchange Rate Changes.” NBER Working Paper 1926, National Bureau of Economic Research, Cambridge, MA.

McCarthy, J. 2007. “Pass-Through of Exchange Rates and Import Prices to Domestic Inflation in Some Industrialized Economies.” Eastern Economic Journal 33 (4): 511-37.

(31)

31

Sargent, T. J., and C. A. Sims. 1977. “Business Cycle Modeling Without Pretending to Have Too Much A Priori Economic Theory”. Working Papers 55, Federal Reserve Bank of

Minneapolis.

Sims, Christopher A. (1992). ”Interpreting the macroeconomic time series facts: The Effects of Monetary Policy.” European Economic Review 36: 975-1000.

Stock, J., and M. Watson. 1999.”Forecasting Inflation”. Journal of Monetary Economics 44: 293-335.

Stock, J.H. and Watson, M.W. 2015. Introduction to econometrics. 3. rev. ed., Global ed. Harlow: Pearson Education Limited.

Takhtamanova, Y. 2008. “Understanding Changes in Exchange Rate Pass-Through”. Journal

of Finance and Economics. 2016, 4 (6): 191-198.

Taylor, J. 2000. “Low Inflation, Pass-Through, and the Pricing Power of Firms.” European

Economic Review 44 (7): 1389-1408.

(32)

32

9. Appendix 1

9.1 Data description

Table 1. Definition of data

Variable Variable name Source Description

Domestic inflation DPCI OECD Quarter-on-quarter consumer price indices Domestic output growth DGDP OECD Quarter-on-quarter seasonally adjusted real GDP growth Domestic nominal effective exchange rate

DNEER IFS Quarter-on quarter

trade- weighted nominal exchange rate

Domestic interest rates

DIR OECD Annualized short-

term interest rates Global inflation GCPI OECD G20 countries,

quarter-on-quarter consumer price indices Global output growth GGDP OECD G20 countries, quarter-on-quarter seasonally adjusted real GDP growth Oil price growth GOIL World bank Quarter-on-quarter

(33)

33

9.2 Structural model

In the FAVAR methodology section the model was presented. In its structural form the FAVAR model is represented by:

𝐻₀𝑌_𝑡 = 𝛼 + ∑ 𝐻_𝑖𝑌_𝑡−𝑖

𝐿 𝑖=1

+ 𝜀_𝑡

where 𝜀_𝑡 is a vector of orthogonal structural innovations; 𝑦_𝑖 consists of country specific inflation (𝜋𝑖,𝑑𝑜𝑚𝑒𝑠𝑡𝑖𝑐), country specific changes in nominal effective exchange rate(𝐸𝑅𝑖), country specific output (𝑌𝑖,𝑑𝑜𝑚𝑒𝑠𝑡𝑖𝑐), country specific interest rates(𝐼𝑖), global inflation (𝜋𝑔𝑙𝑜𝑏𝑎𝑙_{), global output (𝑌}𝑔𝑙𝑜𝑏𝑎𝑙_{) and oil price growth (∆𝑜𝑝). The vector 𝜀}

𝑖 consists of seven

global and domestic shocks. Assuming in the econometric model that 𝐻₀−1 has a recursive structure that the reduced-form errors (𝑢𝑖) can be decomposed according to 𝑢𝑖 = 𝐻0−1𝜀𝑖. The

impelled sign and short-term restriction is analogous to Charnavoki and Dolado (2014), Forbes, Hjortsoe and Nenova (2018) and Ha, Stocker and Yilmazkunday 2019:

[ 𝑢𝑡 𝑌,𝑔𝑙𝑜𝑏𝑎𝑙 𝑢_𝑡𝑂𝑃 𝑢_𝑡𝜋,𝑔𝑙𝑜𝑏𝑎𝑙 𝑢_𝑡𝑌,𝑑𝑜𝑚𝑒𝑠𝑡𝑖𝑐 𝑢_𝑡𝜋,𝑑𝑜𝑚𝑒𝑠𝑡𝑖𝑐 𝑢_𝑡𝐼,𝑑𝑜𝑚𝑒𝑠𝑡𝑖𝑐 𝑢_𝑡𝐸𝑅 _] = [ + + + ∗ ∗ ∗ ∗ − + + ∗ ∗ ∗ ∗ + + − ∗ ∗ ∗ ∗ 0 0 0 + + ∗ ∗ 0 0 0 + − ∗ ∗ 0 0 0 − − + + 0 0 0 ∗ ∗ ∗ +] [ 𝜀_𝑡𝐺𝑙𝑜𝑏𝑎𝑙𝐷𝑒𝑚𝑎𝑛𝑑 𝜀_𝑡𝑂𝑖𝑙𝑃𝑟𝑖𝑐𝑒 𝜀_𝑡𝐺𝑙𝑜𝑏𝑎𝑙𝑆𝑢𝑝𝑝𝑙𝑦 𝜀𝑡𝐷𝑜𝑚𝑒𝑠𝑡𝑖𝑐𝐷𝑒𝑚𝑎𝑛𝑑 𝜀_𝑡𝐷𝑜𝑚𝑒𝑠𝑡𝑖𝑐𝑆𝑢𝑝𝑝𝑙𝑦 𝜀_𝑡𝑀𝑜𝑒𝑛𝑡𝑎𝑟𝑦𝑃𝑜𝑙𝑖𝑐𝑦 𝜀_𝑡𝐸𝑥𝑐ℎ𝑎𝑛𝑔𝑒𝑅𝑎𝑡𝑒 ]

According to theory this is the initial expected response of each of the shocks. Where * stand for an unrestricted initial response. Imposing that global shocks can affect country-specific variables and that domestic shocks cannot affect global variables in the first four quarters.

9.3 ERPT example

(34)

34

American importer. Selling the car at a lower price, for example 42’000$. The appreciation of the Japanese Yen is still 12,5% but the price increase is only 5%. The pass-through is

incomplete, not 100%.

9.4 AD-AS model

Figure 1. Positive AD shock

Figure 2. Negative AD shock

(35)

35

(36)

36

10. Appendix 2

10.1 Average values of pass-through ratios and effect on

exchange rates

Table 2. Calculated averages

Underlying cause Calculated average Value Domestic demand Pass-through ratio 0,064

Effect on ER 0,768

Domestic supply Pass-through ratio 0,129

Effect on ER 0,461

Domestic interest rate Pass-through ratio 0,105

Effect on ER 1,585

Global demand Pass-through ratio 0,006 Effect on ER -0,371 Global supply Pass-through ratio 0,034

Effect on ER -1,771 Global economic activity Pass-through ratio 0,130

Effect on ER -4,224

10.2 Country-specific regressions

Table 3. Chile regression 1 “Exchange rate pass- through”

Variable Estimate Std. Error T value Pr (>|T|) Sign. Level

(37)

37

Table 4. Chile regression 2 “Cause for exchange rate movement”

dataF_t.DCPICHL.l1 -0.8060 0.8781 -0.918 0.3627 dataF_t.DNEERCHL.l1 0.8538 0.3226 2.647 0.0106 * dataF_t.DGDPCHL.l1 0.9436 0.6914 1.365 0.1780 dataF_t.DIRCHL.l1 0.7399 0.7964 0.929 0.3570 dataF_t.GCPICHL.l1 -1.1527 1.1358 -1.015 0.3147 dataF_t.GGDPCHL.l1 2.8832 1.6584 1.739 0.0878 . dataF_t.PC1CHL.l1 -3.2003 9.4726 -0.338 0.7368 dataF_t.DCPICHL.l2 1.6825 1.5536 1.083 0.2836 dataF_t.DNEERCHL.l2 -0.3425 0.3588 -0.955 0.3440 dataF_t.DGDPCHL.l2 -0.1497 0.6647 -0.225 0.8227 dataF_t.DIRCHL.l2 1.2354 0.8269 1.494 0.1410 dataF_t.GCPICHL.l2 1.3681 1.5349 0.891 0.3767 dataF_t.GGDPCHL.l2 -2.5134 1.8060 -1.392 0.1697 dataF_t.PC1CHL.l2 -7.3557 9.5440 -0.771 0.4442 dataF_t.DCPICHL.l3 -1.4855 1.0695 -1.389 0.1705 dataF_t.DNEERCHL.l3 -0.1092 0.3316 -0.329 0.7431 dataF_t.DGDPCHL.l3 -0.3136 0.6187 -0.507 0.6143 dataF_t.DIRCHL.l3 -1.3442 0.6687 -2.010 0.0494 * dataF_t.GCPICHL.l3 0.4575 1.0854 0.422 0.6751 dataF_t.GGDPCHL.l3 -0.4474 1.5031 -0.298 0.7671 dataF_t.PC1CHL.l3 -3.3659 8.9139 -0.378 0.7072 const 49.6312 38.4239 1.292 0.2020

Table 5. France regression 1 “Exchange rate pass- through”

(38)

38

Table 6. France regression 2 “Cause for exchange rate movement”

Variable Estimate Std. Error T vlaue Pr (>|T|) Sign. Level

dataF_t.DCPIFRA.l1 -1.1486 0.6361 -1.806 0.075260 . dataF_t.DNEERFRA.l1 1.7412 0.4234 4.112 0.000105 *** dataF_t.DGDPFRA.l1 1.3132 0.7306 1.798 0.076566 . dataF_t.DIRFRA.l1 -0.2032 0.2014 -1.009 0.316469 dataF_t.GCPIFRA.l1 0.4742 0.2652 1.788 0.078061 . dataF_t.GGDPFRA.l1 -1.6229 0.8017 -2.024 0.046757 * dataF_t.PC1FRA.l1 -8.3246 4.1955 -1.984 0.051158 . const -5.0874 8.0025 -0.636 0.527028

Table 7. Hungary Regression 1 “Exchange rate pass- through”

dataF_t.DCPIHUN.l1 1.502103 0.133592 11.244 3.37e-16 *** dataF_t.DNEERHUN.l1 -0.116593 0.062218 -1.874 0.06598 . dataF_t.DGDPHUN.l1 -0.032776 0.100160 -0.327 0.74467 dataF_t.PC1HUN.l1 -1.580274 1.190720 -1.327 0.18966 dataF_t.DCPIHUN.l2 -1.030120 0.230979 -4.460 3.82e-05 *** dataF_t.DNEERHUN.l2 0.039425 0.074481 0.529 0.59860 dataF_t.DGDPHUN.l2 -0.027523 0.106341 -0.259 0.79669 dataF_t.PC1HUN.l2 1.196368 1.308619 0.914 0.36439 dataF_t.DCPIHUN.l3 0.991706 0.235371 4.213 8.90e-05 *** dataF_t.DNEERHUN.l3 0.002754 0.075920 0.036 0.97119 dataF_t.DGDPHUN.l3 0.006278 0.106347 0.059 0.95313 dataF_t.PC1HUN.l3 -1.046345 1.228720 -0.852 0.39795 dataF_t.DCPIHUN.l4 -0.488499 0.143148 -3.413 0.00118 ** dataF_t.DNEERHUN.l4 0.029746 0.060465 0.492 0.62462 dataF_t.DGDPHUN.l4 -0.154406 0.101228 -1.525 0.13261 dataF_t.PC1HUN.l4 0.432245 0.977831 0.442 0.66010 const 6.986669 4.665538 1.498 0.13968

Table 8. Hungary regression 2 “Cause for exchange rate movement”

(39)

39

Table 9. Iceland regression 1 “Exchange rate pass- through”

dataF_t.DCPIICE.l1 1.26424 0.16113 7.846 4.17e-11 *** dataF_t.DNEERICE.l1 -0.02990 0.01712 -1.747 0.0852 . dataF_t.DGDPICE.l1 -0.06644 0.02842 -2.338 0.0223 * dataF_t.PC1ICE.l1 2.31512 1.26457 1.831 0.0715 . dataF_t.DCPIICE.l2 -0.25888 0.16056 -1.612 0.1115 dataF_t.DNEERICE.l2 0.03289 0.01674 1.964 0.0536 . dataF_t.DGDPICE.l2 0.03297 0.02926 1.127 0.2637 dataF_t.PC1ICE.l2 -1.38956 1.26904 -1.095 0.2774 const -0.25621 1.02443 -0.250 0.8033

Table 10. Iceland regression 2 “Cause for exchange rate movement”

dataF_t.DCPIICE.l1 23.42992 20.90758 1.121 0.26676 dataF_t.DNEERICE.l1 3.74408 2.54715 1.470 0.14664 dataF_t.DGDPICE.l1 2.04200 0.76185 2.680 0.00941 ** dataF_t.DIRICE.l1 -0.02741 2.38379 -0.011 0.99086 dataF_t.GCPIICE.l1 -29.47743 23.34179 -1.263 0.21137 dataF_t.GGDPICE.l1 -13.50261 15.09146 -0.895 0.37440 dataF_t.PC1ICE.l1 227.57172 206.35836 1.103 0.27438 dataF_t.DCPIICE.l2 19.06295 18.18077 1.049 0.29847 dataF_t.DNEERICE.l2 2.50723 2.33034 1.076 0.28614 dataF_t.DGDPICE.l2 1.33412 0.74302 1.796 0.07744 . dataF_t.DIRICE.l2 4.28448 2.34621 1.826 0.07265 . dataF_t.GCPIICE.l2 -20.52390 21.86813 -0.939 0.35161 dataF_t.GGDPICE.l2 -16.96491 14.69086 -1.155 0.25261 dataF_t.PC1ICE.l2 216.81489 189.78032 1.142 0.25766 const 144.82760 81.18158 1.784 0.07932 .

Table 11. Italy regression 1 “Exchange rate pass- through”

dataF_t.DCPICHL.l1 1.28950 0.13031 9.896 6.16e-15 ***Estimate Std. Error T value Pr (>|T|) Sign. Level

(40)

40

Table 12. Italy regression 2 “Cause for exchange rate movement”

dataF_t.DCPIITA.l1 -0.28580 0.22689 -1.260 0.2120 dataF_t.DNEERITA.l1 1.40828 0.25235 5.581 4.23e-07 *** dataF_t.DGDPITA.l1 1.09913 0.50285 2.186 0.0322 * dataF_t.DIRITA.l1 -0.13104 0.18615 -0.704 0.4838 dataF_t.GCPIITA.l1 0.08815 0.09835 0.896 0.3732 dataF_t.GGDPITA.l1 -1.72706 0.72363 -2.387 0.0197 * dataF_t.PC1ITA.l1 -5.02999 2.57712 -1.952 0.0550 . const -20.06129 13.17700 -1.522 0.1324

Table 13. Japan regression 1 “Exchange rate pass- through”

dataF_t.DCPIJAP.l1 1.04610 0.04449 23.515 < 2e-16 *** dataF_t.DNEERJAP.l1 0.02569 0.01453 1.769 0.08109 . dataF_t.DGDPJAP.l1 0.11474 0.05190 2.211 0.03017 * dataF_t.PC1JAP.l1 -23.31285 8.13839 -2.865 0.00545 ** const -6.78386 5.35779 -1.266 0.20948

(41)

41 Variable Estimate Std. Error T vlaue Pr (>|T|) Sign. Level

(42)

42

Table 15. Luxembourg regression 1 “Exchange rate pass- through”

dataF_t.DCPILUX.l1 0.569961 0.166571 3.422 0.001098 ** dataF_t.DNEERLUX.l1 0.340347 0.134442 2.532 0.013863 * dataF_t.DGDPLUX.l1 0.045838 0.036432 1.258 0.212968 dataF_t.PC1LUX.l1 -2.028246 0.796033 -2.548 0.013288 * dataF_t.DCPILUX.l2 0.757675 0.196215 3.861 0.000268 *** dataF_t.DNEERLUX.l2 -0.315066 0.174648 -1.804 0.076012 . dataF_t.DGDPLUX.l2 0.001738 0.036993 0.047 0.962669 dataF_t.PC1LUX.l2 1.861129 0.861515 2.160 0.034561 * dataF_t.DCPILUX.l3 -0.336971 0.161120 -2.091 0.040527 * dataF_t.DNEERLUX.l3 0.012579 0.127991 0.098 0.922019 dataF_t.DGDPLUX.l3 0.032249 0.036749 0.878 0.383518 dataF_t.PC1LUX.l3 0.168592 0.687752 0.245 0.807149 const -2.536450 8.339714 -0.304 0.762023

Table 16. Luxembourg regression 2 “Cause for exchange rate movement”

dataF_t.DCPILUX.l1 -0.04761 0.40668 -0.117 0.9072 dataF_t.DNEERLUX.l1 1.54403 0.30477 5.066 3.91e-06 *** dataF_t.DGDPLUX.l1 -0.11157 0.08467 -1.318 0.1924 dataF_t.DIRLUX.l1 0.36569 0.26138 1.399 0.1668 dataF_t.GCPILUX.l1 -0.24586 0.28050 -0.877 0.3841 dataF_t.GGDPLUX.l1 -0.12841 0.39455 -0.325 0.7459 dataF_t.PC1LUX.l1 -2.41957 1.94977 -1.241 0.2193 dataF_t.DCPILUX.l2 0.12696 0.33135 0.383 0.7029 dataF_t.DNEERLUX.l2 -0.75524 0.32918 -2.294 0.0252 * dataF_t.DGDPLUX.l2 -0.14069 0.07919 -1.777 0.0805 . dataF_t.DIRLUX.l2 -0.29571 0.29663 -0.997 0.3227 dataF_t.GCPILUX.l2 0.21269 0.27087 0.785 0.4353 dataF_t.GGDPLUX.l2 0.18726 0.28506 0.657 0.5137 dataF_t.PC1LUX.l2 2.77964 1.79543 1.548 0.1267 const 16.89941 30.34077 0.557 0.5795

Table 17. Netherland regression 1 “Exchange rate pass- through”

(43)

43

Table 18. Netherland regression 2 “Cause for exchange rate movement”

dataF_t.DCPINLD.l1 -0.4640 0.4794 -0.968 0.3364 dataF_t.DNEERNLD.l1 1.4286 0.2883 4.956 4.82e-06 *** dataF_t.DGDPNLD.l1 0.8162 0.4803 1.699 0.0937 . dataF_t.DIRNLD.l1 -0.5060 0.3918 -1.291 0.2008 dataF_t.GCPINLD.l1 0.1656 0.2180 0.760 0.4500 dataF_t.GGDPNLD.l1 -1.4427 0.7653 -1.885 0.0635 . dataF_t.PC1NLD.l1 -5.9464 2.9567 -2.011 0.0482 * const -12.1751 7.2977 -1.668 0.0997 .

Table 19. New Zealand regression 1 “Exchange rate pass- through”

dataF_t.DCPINZL.l1 0.977405 0.008646 113.045 <2e-16 *** dataF_t.DNEERNZL.l1 0.020119 0.010117 1.989 0.0505 . dataF_t.DGDPNZL.l1 -0.182535 0.086289 -2.115 0.0378 * dataF_t.PC1NZL.l1 -0.521435 0.336314 -1.550 0.1254 const 0.573692 0.534283 1.074 0.2865

Table 20. New Zealand regression 2 “Cause for exchange rate movement”

(44)

44

Table 21. Spain regression 1 “Exchange rate pass- through”

dataF_t.DCPIESP.l1 0.34433 0.15429 2.232 0.029198 * dataF_t.DNEERESP.l1 0.60850 0.18481 3.293 0.001631 ** dataF_t.DGDPESP.l1 -0.11617 0.21858 -0.531 0.596951 dataF_t.PC1ESP.l1 -4.54063 1.25594 -3.615 0.000597 *** dataF_t.DCPIESP.l2 1.14421 0.12347 9.267 2.24e-13 *** dataF_t.DNEERESP.l2 -0.52557 0.23665 -2.221 0.029962 * dataF_t.DGDPESP.l2 0.14293 0.25420 0.562 0.575923 dataF_t.PC1ESP.l2 3.88957 1.36937 2.840 0.006061 ** dataF_t.DCPIESP.l3 -0.49429 0.15406 -3.209 0.002099 ** dataF_t.DNEERESP.l3 -0.13984 0.19888 -0.703 0.484564 dataF_t.DGDPESP.l3 -0.08592 0.18832 -0.456 0.649760 dataF_t.PC1ESP.l3 1.32616 1.20891 1.097 0.276822 const 6.68047 15.99926 0.418 0.677697

(45)

45 Variable Estimate Std. Error T vlaue Pr (>|T|) Sign. Level

(46)

46

Table 23. Sweden regression 1 “Exchange rate pass- through”

dataF_t.DCPISWE.l1 0.717156 0.128019 5.602 4.98e-07 *** dataF_t.DNEERSWE.l1 0.091346 0.038732 2.358 0.02147 * dataF_t.DGDPSWE.l1 0.087583 0.070411 1.244 0.21815 dataF_t.PC1SWE.l1 4.048539 1.945926 2.081 0.04155 * dataF_t.DCPISWE.l2 0.494394 0.148091 3.338 0.00142 ** dataF_t.DNEERSWE.l2 -0.066286 0.051948 -1.276 0.20664 dataF_t.DGDPSWE.l2 -0.051572 0.071012 -0.726 0.47038 dataF_t.PC1SWE.l2 -0.005113 2.138114 -0.002 0.99810 dataF_t.DCPISWE.l3 -0.211536 0.124772 -1.695 0.09494 . dataF_t.DNEERSWE.l3 0.025488 0.037085 0.687 0.49443 dataF_t.DGDPSWE.l3 0.162070 0.072799 2.226 0.02958 * dataF_t.PC1SWE.l3 0.558479 1.652521 0.338 0.73652 const -4.898091 3.858999 -1.269 0.20902

Table 24. Sweden regression 2 “Cause for exchange rate movement”

(47)

47

Table 25. Great Britain regression 1 “Exchange rate pass- through”

dataF_t.DCPIGBR.l1 0.76420 0.11926 6.408 2.13e-08 *** dataF_t.DNEERGBR.l1 -0.06552 0.02839 -2.308 0.02431 * dataF_t.DGDPGBR.l1 0.42797 0.15150 2.825 0.00633 ** dataF_t.PC1GBR.l1 -0.87574 0.63354 -1.382 0.17176 dataF_t.DCPIGBR.l2 0.48260 0.15152 3.185 0.00225 ** dataF_t.DNEERGBR.l2 0.05584 0.03647 1.531 0.13072 dataF_t.DGDPGBR.l2 -0.09823 0.16636 -0.590 0.55700 dataF_t.PC1GBR.l2 1.06153 0.66876 1.587 0.11745 dataF_t.DCPIGBR.l3 -0.22931 0.13407 -1.710 0.09212 . dataF_t.DNEERGBR.l3 0.02090 0.02312 0.904 0.36962 dataF_t.DGDPGBR.l3 -0.12021 0.12537 -0.959 0.34132 dataF_t.PC1GBR.l3 1.14233 0.59535 1.919 0.05955 . const -2.48806 2.14358 -1.161 0.25014

Table 26. Great Britain regression 2 “Cause for exchange rate movement”

(48)

48

Table 27. USA regression 1 “Exchange rate pass- through”

dataF_t.DCPIUSA.l1 1.032444 0.134194 7.694 1.23e-10 *** dataF_t.DNEERUSA.l1 -0.068807 0.034890 -1.972 0.0530 . dataF_t.DGDPUSA.l1 -0.086321 0.127114 -0.679 0.4996 dataF_t.PC1USA.l1 0.195970 1.138946 0.172 0.8639 dataF_t.DCPIUSA.l2 -0.826361 0.196489 -4.206 8.39e-05 *** dataF_t.DNEERUSA.l2 -0.009836 0.049162 -0.200 0.8421 dataF_t.DGDPUSA.l2 0.050560 0.119439 0.423 0.6735 dataF_t.PC1USA.l2 -0.790915 1.376524 -0.575 0.5676 dataF_t.DCPIUSA.l3 0.794280 0.137255 5.787 2.44e-07 *** dataF_t.DNEERUSA.l3 0.082528 0.032548 2.536 0.0137 * dataF_t.DGDPUSA.l3 0.270499 0.121293 2.230 0.0293 * dataF_t.PC1USA.l3 1.654106 1.121352 1.475 0.1452 const 0.233962 1.191319 0.196 0.8449

Table 28. USA regression 2 “Cause for exchange rate movement”