0
MSc Economics, Specialisation International Economics &
Globalisation
Master Thesis
Exchange Rate Pass-Through into Import
Prices during Crises
Author:
Reinier H. de Groot
reinier.degroot@student.uva.nl
Student number 10188282
Supervisor:
Second reader:
Dr. K. Mavromatis
Dr. D.J.M. Veestraeten
k.mavromatis@uva.nl
d.j.m.veestraeten@uva.nl
1
Statement of Originality
This document is written by Student Reinier de Groot who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
2
Abstract
This empirical paper studies the extent of exchange rate pass-through into import prices whilst making the distinction between crisis and non-crisis periods. Quarterly data on 17 OECD countries over the period of 1980 to 2016 is used to estimate short and long run pass-through elasticities for each country for crisis and non-crisis observations. This is done by means of OLS regressions using a model which is based on micro-foundations. Short run pass-through elasticities indeed differ between crisis and non-crisis periods, but do not change in the same direction for all countries. Large open economies tend to have increasing ERPTs whilst small open economies tend to have decreasing ERPTs.
3
Table of Contents Page
1. Introduction 4
2. Literature Review 6
2.1 Exchange rate pass-through in general 6
2.2 Distinguishing a crisis 8
2.2.1 Quantitative approach to identify a crisis 8 2.2.2 Qualitative approach to identify a crisis 9
3. Methodology 10
3.1 Empirical model and benchmark regressions 10 3.1.1 Specification of the import price 10
3.1.2 Regression equation 11
3.2 Distinguishing and indicating a crisis 13
3.2.1 Quantitative approach 13
3.2.2 Qualitative approach 14
3.3 Regressions with distinction of crises 15
4. Data 15
5. Results 17
5.1 Benchmark estimates of pass-through 17
5.2 Estimates with distinction of crises 19
4
1. Introduction
In an increasingly interdependent and globalised world such as we live in today,
exchange rates have become of increasing importance for every one of us. The simplest case in which we encounter the effect of exchange rates is our grocery shopping of imported goods or buying gas at the petrol station. Consider the example of grocery shopping: the price of bananas is usually around the same price throughout the year, with the exception of special offers. When looking at petrol prices, one notices that these fluctuate more than the aforementioned bananas, especially during economic crises. Both commodity prices are being affected by import prices and those are in turn affected by exchange rates, among other things. The amount by which retail or consumer prices are being affected by exchange rates is called exchange rate through. The pass-through is generally expressed as an elasticity: the percentage change of import/selling prices in domestic currency as a result of a one percent change in the exchange rate.
This pass-through, however, is not constant and can differ between the short run and long run. During periods of higher exchange rate volatility for instance, retail sellers may not completely charge a depreciation to its customers to keep stable prices and demand. In this sense, there are two extreme forms of exchange rate pass-through: Producer-Currency Pricing and Local-Currency Pricing. The first indicates that the seller charges the entire depreciation to the buyer: the pass-through elasticity is 100 percent. The latter is the opposite: the producer thus absorbs the depreciation in his/her
markup, often referred to as pricing to market when the seller wants to charge different prices in different markets. The pass-through elasticity is 0 percent. This research focuses on the behaviour of exchange rate pass through during currency crises, balance of payments crisis, or debt or banking crises. Hence the main research question is: How is exchange rate pass-through affected by a crisis period?
Earlier research by Campa & Goldberg (2005) provides evidence on the extent of exchange rate pass-through (ERPT) and mentions that ERPT differs between the short run and long run. More importantly, countries that have higher exchange rate volatility seem to have higher pass-through elasticities. A crisis can induce higher exchange rate volatility and thus can influence the ERPT. Currently, however, there is little to no research on the effect of a crisis on ERPT, despite its implications for (monetary) policy and inflation. Monetary policy, for example, is based on exchange rates and inflation. Clearly, it is useful to know how exchange rates affect inflation through the amount of
5
pass-through. On top of that, forecasts with use of models with certain assumptions are perhaps wrong, because these assumptions may not hold during a crisis. A model that assumes that a depreciation makes imports more expensive and thereby increasing inflation can be rendered useless if that depreciation is offset by a change in pass-through. Former studies neglect crises when studying ERPT. My research adds to existing literature by distinguishing between “normal” times and times of crisis.
In this paper I use an empirical model describing a change in ERPT during
periods with higher exchange rate volatility. I take account for possible stochastic trends by checking for stationarity and cointegration and incorporating lags of the variables in the estimation equation. Although I examine this issue primarily from a macroeconomic, international perspective I do pay attention to microeconomic foundations to estimate import prices, accounting for markups on the importer’s marginal costs.
For convenience and completeness of data, I solely use data of developed OECD countries. Quarterly data on the aggregate import price index and data on nominal and real effective exchange rates for the period of 1980-2016 is available at the International Financial Statistics database of the IMF. Quarterly data on the real gross domestic
product for the same period is available at the OECD Statistics database and data on gross capital flows is available at the balance of payments statistics database of the IMF. Gross flows data is used to compute measures to identify crises.
A widely used method in existing literature to estimate exchange rate pass-through elasticites is also used by Campa & Goldberg (2005) and is the same method I apply. I start by constructing the importprice which is influenced by marginal costs and markups, among others. Subsequently, I estimate a model by means of OLS regressions incorporating the import price and the exchange rates and their corresponding lags. Eventually I estimate the pass-through elasticities per country during crisis and non-crisis periods.
The thesis is arranged as follows. Section 2 provides a comprehensive literature review, introducing some often used definitions and theories, and methods to identify a crisis. Section 3 gives a detailed description of the methodology. Section 4 elaborates on the data. Section 5 shows the empirical results. Finally, section 6 concludes.
6
2. Literature review
The following literature review comprises two parts. The first part reviews existing studies on exchange rate pass-through. The second part digresses on ways to distinguish or indicate several types of crises.
2.1 Exchange rate pass-through in general
Fortunately, plenty of research has been done on the relation between exchange rates and prices and thus empirical models are available. Goldberg & Knetter (1997) provide a good starting point for a literature review, as they attempt to give a unifying framework for research on the topic prior to 1997. Although they look at the subject more from an industrial organization stance, the paper is certainly useful because they discuss some key components in the theory about ERPT. In the 1970s, after the Bretton-Woods era, interest in prices and exchange rates increased because of both the desire of testing fundamental theories for monetarism such as PPP and the law of one price, and to study effects of a depreciation or appreciation on the external balance.
More specifically, economists wanted to know if a devaluation would improve a country’s trade balance. When one thinks back of the basic models with perfectly elastic export supply functions, this leads to the Marshall-Lerner condition: a devaluation improves the trade balance if the sum of the import and export demand elasticities exceeds one. However, is export supply really perfectly elastic? Would export and/or import prices remain constant in case of a devaluation? To assess the behavior of these prices, knowledge of ERPT behavior is necessary.
Taylor (2000) adds that lower and more stable inflation is a reason for the reduction in the degree importing firms pass through cost increases due to exchange rate fluctuations. This decrease in ERPT can be seen as a decrease in pricing power of firms. Many observers consider this as a reason for the fact that inflation did not increase significantly in the 1990s in the U.S.A, despite strong demand pressure. It is merely a sign that lower ERPT is not exogenous to inflation. Furthermore, the degree of ERPT is dependent on the expected persistence of low and stable inflation. In models of staggered price setting in which prices are set for several periods ahead, pass-through is smaller since a firm or importer will not try to persistently match cost increases due to exchange rate variations.
7
Engel, 2002). Under local-currency pricing, firms set a price in the home currency to sell to the domestic households, but set a price in the foreign currency to sell to foreign households. This leads to a perfectly incomplete exchange rate pass-through, because the prices are unresponsive to nominal exchange rate changes. An implication of this result is that the expenditure switching effect of exchange rate changes might be quite small. In other words, a change in nominal exchange rate will not induce much
substitution between domestically and foreign produced goods because the relative prices do not change that much for the final consumer, as there is little pass-through.
The consequence is that in such a case of local-currency pricing, exchange rate devaluations matter little for the real economy and the (volatility of) macroeconomic aggregates. This potentially leads to a problem for monetary policy. In a small open economy if the country or central bank wishes to stimulate the economy via a
devaluation (for example to combat an economic recession), it is useful for the country or central bank to have knowledge about ERPT.
Devereux, Engel, & Storgaard (2004) digress on the interaction between ERPT and exchange rates in an open economy framework where pass-through is endogenous because firms exercise local-currency pricing. The key result is that ERPT is related to the relative stability of monetary policy. A country with low exchange rate volatility will have a relatively low degree of pass-through, while a country with higher exchange rate volatility will have a relatively high degree of pass-through. Again monetary policy plays a role in the degree of ERPT, but in this case it is endogenous in the degree of exchange rate pass-through. Mishkin (2008) further stresses that exchange rate fluctuations do have an effect on inflation and economic activity and should be taken into consideration in monetary policy.
Junttilla & Korhonen (2012) extend the view on the relation between ERPT and monetary policy. They examine ERPT into aggegate import prices of nine OECD
countries in view of Taylor’s (2000) findings that the degree of pass-through depends on the importing country’s inflation regime. The main result is that the pass-through is highly incomplete and is positively correlated with the importing country’s inflation regime, using nonlinear threshold or smooth transition estimation techniques.
Taking into consideration different pricing methods such as local-currency pricing and producer-currency pricing and the relation between ERPT and monetary policy, Campa & Goldberg (2005) provide a useful model to estimate ERPT. The study
8
makes use of microeconomic fundamentals to estimate the importprice, which depends on ERPT, marginal costs and sellers’ markups. Moreover, they make a distinction
between long run and short run using data of 23 OECD countries. On top of that, Campa & Goldberg (2005) estimated pass-throughs for aggregated as well as disaggregated import price indices. The disaggregated indices were available in five import categories: food, manufacturing, energy, raw materials and nonmanufactured products. This gave them the opportunity to estimate five different ERPTs and therefore find differences in pass-through for different import categories. An important finding is that the move away from a relatively high proportion of energy in the import bundles has been a primary driver of changes in ERPT during the 1990s. Unfortunately, as of 2002 the OECD has stopped publishing the disaggregated import price series which means I cannot incorporate disaggregated prices in my research. Other than that, I perform my
research in a very similar way while adding the distinction between “normal” times and times in crisis. This means that, besides the benchmark estimates, I perform separate regressions for crisis observations and non-crisis observations, indicated by a dummy variable. I further break down the crisis into different types.
2.2 Distinguishing a crisis
In recent years, a great deal of research has been done on the identification, dating and classifying of crises. Still, many questions remain unanswered which shows how complex of a subject a crisis is. Even the seemingly simple task of afterwards indicating when a crisis started is not straightforward. The main theories on explaining or classifying crises have brought some methodologies to fulfil this task, but may not exactly agree due to different measuring thresholds for example. Moreover, the choice between a quantitative or qualitative approach depends on the type of crisis.
2.2.1 Quantitative approach to identify a crisis
Currency crises, for instance, can be identified relatively easy as they usually exhibit large changes in exchange rates. Reinhart & Rogoff (2009) distinguish a currency crisis if the exchange rate depreciates by more than 15 percent per year. Frankel & Rose (1996) on the other hand declare a period as a currency crisis if there is a depreciation of at least 25 percent cumulative over 12 months with at least a 10 percent increase in the rate of depreciation. Obviously, the identified crisis period may differ with different
9
thresholds. Besides, the measure can be adjusted for a country’s usual exchange rate variations. Measurement issues can also arise if there is no significant change in the currency, despite pressure or attacks. Adjustments in international reserves or interest rates can absorb exchange rate pressure and accordingly attenuate currency
fluctuations.
Sudden stops and balance of payments crises can also be identified fairly
straightforward and objectively. Calvo, Izquierdo, & Talvi (2006) define a sudden stop as an event in which there is a sharp decrease in output and large reversals in capital flows. Calvo, Izquierdo & Mejía (2008) add two features to these two criteria. The first
criterium says that the period must cover at least one year-on-year fall in capital flows that is two standard deviations below its sample mean, at minimum. This emphasizes the unexpected aspect of a sudden stop. The second criterium is that the crisis starts when the annual change in capital flows falls one standard deviation below its mean. Forbes & Warnock (2012) develop a new methodology to identify balance of payments crises, using gross capital flows instead of the more traditional net capital flows (or the current account). They use quarterly data to distinguish episodes of extreme capital movements. Their approach with the ability to capture distinctions in the behavior of domestic and foreign investors by using gross flows, presents a more nuanced understanding of extreme capital flow episodes. They specify episodes as surge, stop, flight or retrenchment. Surges and stops are related respectively to periods of large gross capital in- or outflows by foreigners. Flights and retrenchments are related
respectively to periods of large gross capital out- or inflows by domestic residents.
2.2.2 Qualitative approach to identify a crisis
A seemingly easier way to identify crises – in particular for debt crises - is to use the credit ratings of Standard and Poor’s. Broner, Didier, Erce, & Schmukler (2013) have used credit ratings to complement their list of crisis years in each country. They classify a year as a year with a debt crisis if there are downgrades to default levels for sovereign domestic currency debt, for sovereign foreign currency debt or for sovereign foreign currency bank loans. The rest of the crisis years – including banking, currency and debt crises – are identified using methodologies of Laeven & Valencia (2008), Frankel & Rose (1996) and Reinhart & Rogoff (2009).
10
3. Methodology
The methodology section of this thesis consists of three parts. In the first part I explain the empirical model and perform some benchmark regressions to compare with the findings of Campa & Goldberg (2005). In the second part I compute three different crisis indicators and in the third part I incorporate these indicators in the regressions to estimate the additional effect of a crisis.
3.1 Empirical model and benchmark regressions
3.1.1 Specification of the import price
The import price equation is based on micro foundations, which is helpful for
understanding the dynamics of the pricing behavior by exporters. The import prices for a country j at time t, Ptm,j , is the export price, Ptx,j , multiplied by the spot exchange rate
Etj. “Spot” implies that the exchange rate is defined as the amount of domestic currency
per unit of foreign currency. A depreciation (appreciation) of the domestic currency is thus shown as an increase (decrease) in the exchange rate.
𝑃𝑡𝑚,𝑗 = 𝐸𝑡𝑗∗ 𝑃𝑡𝑥,𝑗 (1)
The export prices consist of a markup, MKUPtx, and the marginal costs for the exporter,
MCtx.
𝑃𝑡𝑥,𝑗 = 𝑀𝐾𝑈𝑃𝑡𝑥,𝑗 ∗ 𝑀𝐶𝑡𝑥,𝑗 (2) Hence, the import prices are defined as:
𝑃𝑡𝑚,𝑗 = 𝐸𝑡𝑗∗ 𝑀𝐾𝑈𝑃𝑡𝑥,𝑗∗ 𝑀𝐶𝑡𝑥,𝑗 (3) For convenience, I drop the country superscript j and do a log-linear transformation, with lowercase letters depicting logarithms.
𝑝𝑡𝑚 = 𝑒𝑡+ 𝑚𝑘𝑢𝑝𝑡𝑥+ 𝑚𝑐𝑡𝑥 (4) The markups itself are sensitive to industry-specific fixed effects and to macroeconomic conditions. For simplicity, the macroeconomic conditions are expressed as a function of solely the exchange rates at this point.
11
In which φ indicates the industry-specific effect and Φ indicates the macroeconomic effect.
Marginal costs for the exporter in turn are specified as rising in both export market wages, wtx, and import market demand conditions, yt. Since I do not have an
appropriate index or variable reflecting the demand of different importing countries, I use an importing country’s GDP as a proxy for market demand conditions. Simply said, higher wages for employees in the exporting country increases the costs of producing the good. In addition, if market conditions of the importing country improve, its demand pressure will increase, which in turn increases the marginal costs. Thus the marginal costs are specified as:
𝑚𝑐𝑡𝑥 = 𝑐
0𝑦𝑡+ 𝑐1𝑤𝑡𝑥 (6)
Combining equations (4), (5) and (6) results in the import price specification: 𝑝𝑡𝑚 = 𝑒
𝑡+ 𝜑 + Φ𝑒𝑡+ 𝑐0𝑦𝑡+ 𝑐1𝑤𝑡𝑥 (7) This specification, however, can be written in a more convenient way in which it elegantly shows a direct relation to the principles of LCP and PCP:
𝑝𝑡𝑚 = 𝜑 + (1 + Φ)𝑒
𝑡+ 𝑐0𝑦𝑡+ 𝑐1𝑤𝑡𝑥 (8)
When using this form of the equation for regressions, with β= (1+Φ), it is immediately clear that the exchange rate through is zero (LCP) if Φ = -1 and exchange rate pass-through is complete if Φ = 0 (PCP). This furthermore shows the relation between ERPT β= (1+Φ) and the structure of competition in the industry.
3.1.2 Regression equation
The arguments in equation (8) are captured in a log-linear regression specification similar to a large part of previous studies on exchange rate pass-through.
𝑝𝑡 = 𝛼 + 𝛿𝑤𝑡+ 𝛽𝑒𝑡+ 𝜑𝑦𝑡+ 𝜀𝑡 (9) In this equation, pt are the import prices in domestic currency, et is the exchange rate, wt
is the primary control variable which represents the costs for the exporter and yt is a
vector of other controls, one of which is the GDP of the importing country. Not including this control variable would lead to biased estimates if the importing country’s GDP and wages are correlated, which is very likely.
12
Finding a control variable that captures the changing exporter costs of an importing country’s trading partners is more difficult. Hence, I construct an export partner trading cost proxy by making use of both the logarithm of the nominal effective exchange rate, neer, and the logarithm of the real effective exchange rate, reer, indexes. This proxy is computed for each importing country as follows:
𝑊𝑡𝑗 = 𝑛𝑒𝑒𝑟𝑡𝑗/𝑟𝑒𝑒𝑟𝑡𝑗 ∗ 𝑃𝑡𝑗 (10)
This way of computing the export partner cost proxy by country gives a measure of all export partners of a country. Each partner is weighted according to its importance in the importing country’s trade.
The first part of the analysis is to estimate short-run and long-run pass-through elasticities. In this case, short-run refers to one quarter and long-run refers to four quarters. Most of the change in prices happens in the first and second quarters after an exchange rate change, thus interpreting four quarters as long-run is empirically
validated. Four quarters also matches the long-run in previous ERPT literature. I furthermore add the lagged exchange rate and lagged export partner cost proxy to incorporate a possible gradual adjustment of import prices to exchange rates. The estimation equation, expressed in first differences and including the lags is as follows:
∆𝑝𝑡𝑗 = 𝛼 + ∑4𝑖=0 𝑎𝑖𝑗∆𝑒𝑡−𝑖𝑗 + ∑𝑖=04 𝑏𝑖𝑗∆𝑤𝑡−𝑖𝑗 + 𝑐𝑗∆𝑔𝑑𝑝𝑡𝑗+ 𝜀𝑡𝑗 (11) I perform ordinary least squares estimation on variables with log differences, in which the estimated coefficient aoj is the short-run pass-through. The benefit of using first
differences is that it avoids problems caused by stochastic trends in time series regressions.
The long-run pass-through elasticity is the sum of the coefficients of the exchange rate at time t and four lags: ∑4𝑖=𝑜𝑎𝑖𝑗. Due to the way the exchange rates are defined – an appreciation (depreciation) is shown by a decrease (increase) in the exchange rate - all resulting elasticities will be negative. The table in the results section, however, shows corrected elasticities and hence a positive elasticity does not have a minus sign and a negative elasticity does have a minus sign.
Other methods such as an error correction model approach are ruled out after testing for stationarity of the import price, exchange rate and export cost series. For the greater part of the Dickey-Fuller unit root tests with trend I fail to reject the null
13
hypothesis of the existence of a unit root. Trying different amounts of lags (0, 2 and 4) yields the same results. Additionally, testing for cointegration revealed that
approximately half of the countries’ time series contains a cointegration relation between the import prices, the exchange rate and the exporter costs. Both the Engle-Granger Augmented Dickey Fuller test and the Johansen cointegration test yield roughly the same result of half of the time series containing a cointegration relation, in contrast with Campa & Goldberg (2005) who find only three cases of cointegration, albeit in a smaller sample of 23 countries. Still, half of the instances does not contain a
cointegration relation and therefore I do not apply an error correction model. Error correction models are also rather uncommon in previous studies on exchange rate pass-through.
3.2 Distinguishing and indicating a crisis
The next step is to distinguish crisis periods and estimate the pass-through in case of a crisis. I use both a quantitative and a qualitative approach to identify crises. This offers possibilities to compare the findings of different methodologies and the possibility to see if the two approaches agree with each other. Moreover, the outcomes of the approaches complement each other well since one or the other does not reveal all relevant crises. I focus on currency, balance of payment and debt crises, as these type of crises are closely related to exchange rate pass-through and will most likely affect pass-through.
3.2.1 Quantitative approach
Using a quantitative approach, I identify two types of crisis with two measures. The first type is the currency crisis. Using the method of Reinhart & Rogoff (2009), these crises can be identified relatively straightforward by making use of nominal exchange rate data. I distuingish a currency crisis if the nominal exchange rate depreciates by more than 15 percent in a year. This results in 98 quarterly observations that are declared as in a currency crisis.
The second type of crisis to identify quantitatively is the balance of payments crisis. I implement the methodology of Forbes & Warnock (2012) that incorporates gross capital flows to identify four extreme gross capital flow episodes: a surge, a sudden stop, a sudden flight and a retrenchment. Gross capital inflows is the aggregate of foreign direct investment inflows, portfolio investment inflows and other investment
14
inflows. Gross capital outflows is the aggregate of foreign direct investment outflows, portfolio investment outflows, other investment outflows and the international reserve assets.
For the calculation of a surge or sudden stop, I first compute Ct, the four-quarter
moving sum of gross capital inflows (GINFLOW): 𝐶𝑡 = ∑3 𝐺𝐼𝑁𝐹𝐿𝑂𝑊𝑡−𝑖
𝑖=0 with 𝑡 = 1,2, … , 𝑁. (12)
Then I calculate year-on-year changes in Ct:
∆𝐶𝑡 = 𝐶𝑡− 𝐶𝑡−4 with 𝑡 = 5,6, … , 𝑁. (13) I subsequently compute the rolling means and standard deviations of ΔCt for the last
five years, or 20 quarters. A surge is defined as a period in which (1) ΔCt increases to
more than one standard deviation above the rolling mean of ΔCt and (2) in which there
is at least one quarter t in which ΔCt increases to more than two standard deviations
above the rolling mean of ΔCt. A stop is defined in an analogous way: a period in which
(1) ΔCt decreases to more than one standard deviation below the rolling mean of ΔCt and
(2) in which there is at least on quarter t in which ΔCt decreases to more than two
standard deviations below the rolling mean of ΔCt.
Sudden flight and retrenchment episodes are defined similarly but using gross capital outflows instead of inflows, taking into account that capital outflows by domestic residents are reported with a negative value in balance of payments accounting terms. Thus, a sudden flight is defined as a period in which (1) ΔCt decreases to more than one
standard deviation below the rolling mean of ΔCt and (2) in which there is at least on
quarter t in which ΔCt decreases to more than two standard deviations below the rolling
mean of ΔCt. A retrenchment is defined as a period in which ΔCt increases to more than
one standard deviation above the rolling mean of ΔCt and (2) in which there is at least
one quarter t in which ΔCt increases to more than two standard deviations above the
rolling mean of ΔCt.
The four above described episodes of extreme capital flow movements are all defined as balance of payments crises. This method of identifying crises results in 1539 quarterly observations being declared as in a balance of payments crisis.
15 3.2.2 Qualitative approach
Broner, Didier, Erce, & Schmukler (2013) have produced a list of banking, currency and debt crises in the period of 1970 to 2008. Part of the crises on this list is identified using Standard and Poor’s credit ratings. A year is defined as crisis year if a country’s credit ratings are downgraded to default levels for sovereign domestic currency debt, for sovereign foreign currency debt or for sovereign foreign currency bank loans. I make use of this list to identify any further crises that have not been captured by the previous two quantitative methods. The list of Broner, Didier, Erce, & Schmukler (2013) contains 192 quarterly observations that are declared as crises. This brings the total amount of identified crisis observations to 1673 quarters over 38 years in 33 countries.
3.3 Regressions with distinction of crises
For the following regressions, considering whether there is a crisis or not, I use the same empirical model as before:
∆𝑝𝑡𝑗 = 𝛼 + ∑4𝑖=0 𝑎𝑖𝑗∆𝑒𝑡−𝑖𝑗 + ∑𝑖=04 𝑏𝑖𝑗∆𝑤𝑡−𝑖𝑗 + 𝑐𝑗∆𝑔𝑑𝑝𝑡𝑗+ 𝜀𝑡𝑗 (14) Now, however, I estimate short run and long run exchange rate pass-through for
different subsets of the data. These subsets are: Observations without a crisis
Observations during a currency crisis
Observations during a balance of payments crisis Observations during a debt or banking crisis
Observations during a currency, balance of payments, debt or banking crisis Thus, the last subset covers all types of crises. Comparing the first subset with the last subset shows the difference in pass-through between periods with and without a crisis.
4. Data
For the empirical part of this thesis I make use of quarterly panel data on 35 OECD countries over the period of 1979q1 to 2016q4. Data on the nominal effective exchange rate (NEER), real effective exchange rate (REER) and the aggregate import price index are all available at the International Financial Statistics (IFS) database of the
16
complete and thus exhibits large gaps in the time series. This led me to immediately dropping Estonia, Slovenia and Turkey from the data since even the exchange rate data is missing for these countries. Data on the effective exchange rates is based on the consumer price index (CPI), in contrast with Campa & Goldberg (2005) who use effective exchange rates based on unit labor costs (ULC). I choose for CPI due to the shortage of data on ULC based effective exchange rates. As table 1 below shows, using ULC based exchange rates would leave me with only 1280 observations for which there is data for both the nominal and real effective exchange rates, the real GDP and the aggregate import price index. Data on the quarterly real gross domestic product in national currency is available in the National Accounts Statistics of the OECD database.
Data on gross capital flows for 35 OECD countries over the period of 1979q1 to 2016q4 is available at the Balance of Payments Statistics (BoPS) database of the IMF. More precisely, data on foreign direct investment (FDI), portfolio investment, other investments and international reserve assets is used to compute the gross capital inflows (by foreigners) and the gross capital outflows (by domestic residents) in the following way:
- Gross Capital Inflows (CIF) = FDI incurrence of liabilities + Portfolio Investment
incurrence of liabilities + Other Investment incurrence of liabilities
- Gross Capital Outflows (COD) = FDI acquisition of assets + Portfolio Investment
acquisition of assets + Other Investment acquisition of assets + International Reserve Assets
The CIF and COD variables are used to identify balance of payments crises. The frequency of these and other crises are presented in table 2.
Eventually, of the initial dataset covering 35 OECD countries, there is complete data for only 17 countries to estimate exchange rate pass-through elasticities.
17 Table 1: Summary statistics
This table presents an overview of the dataset used in the research. NEER (CPI) is the Nominal Effective Exchange Rate based on the Consumer Price Index. REER (ULC) is the Relative Effective Exchange Rate based on Unit Labor Costs. CIF is the Gross Capital Inflows by Foreigners. COD is the Gross Capital Outflows by Domestic Residents.
VARIABLES # of obs. Mean St. Deviation Minimum Maximum
Real GDP 4,428 1.871e+07 8.025e+07 9,498 5.327e+08
Aggr. Import price index 2,698 81.61 26.08 1.610 185.8
NEER (CPI) 4,568 702.1 5,854 18.51 179,906
NEER (ULC) 2,112 99.03 10.95 70.05 156.9
REER (CPI) 4,422 97.76 15.85 38.38 190.1
REER (ULC) 2,112 98.86 16.97 50.67 173.3
Export Cost Proxy (ULC) 1,280 94.24 19.18 43.22 159.5
Export Cost Proxy (CPI) 2,448 168.9 902.3 5.515 14,056
CIF 3,845 22,669 67,493 -700,287 870,875
COD 3,845 20,991 61,212 -810,513 866,501
Table 2: Frequencies of crisis observations
The “Benchmark” type of observation refers to the observations in which there is made no distinction between a crisis or not.
Type of observation Number of quarterly observations
Benchmark 4,732
No Crisis 3,059
Crisis 1,673
Currency Crisis 98
Balance of Payments Crisis 1,539
Debt or Banking Crisis 192
5. Results
5.1 Benchmark estimates of pass-through
The estimates of short run and long run exchange rate pass-through for different subsets of data are presented below in table 3. The table contains short run and long run pass-through elasticities for 17 OECD countries for the period of 1980 to 2016. Columns (1) and (2) show the short and long run ERPT per country without making the distinction whether observations are during a crisis or not. In essence, this is the typical way of
18
estimating the pass-through in previous studies and serves as a benchmark.
One thing to notice is that all estimated elasticities are considerably lower than the estimates by Campa & Goldberg (2005). Part of this could be due to the use of a different estimation period, but that would most likely not lead to such big differences. Another reason for the difference could be the aggregate import price index data I have used. Campa & Goldberg (2005) use the index published by the OECD, whereas I use the index published by the IMF since it isn’t available at the OECD anymore. However,
although the estimated elasticities are of a smaller magnitude, the relative differences between countries are roughly the same. For example, Austria, the U.K. and the U.S. have some of the smallest elasticities, whilst Hungary and Japan have some of the largest elasticities. Hence, inferences about relative differences may still be considered valid.
The unweighted averages of the short run - one quarter - and the long run - four quarters - are respectively 0.13 and 0.28. These averages, however, do not tell about cross-country differences in pass-through into import prices. The United States and United Kingdom, for instance, show relatively low elasticities of 5 and 3 percent in the short run. France’s short run pass-through is also quite low with 6 percent. This suggests that large open economies exhibit lower pass-through than small open economies. In general, smaller countries tend to have less stable and higher pass-through but a systematic relation has not been empirically established (Campa & Goldberg, 2005). Germany and Japan, however, show somewhat larger ERPTs with Germany being in the middle range with 15 percent and Japan being in the higher range with 29 percent. The differences between countries seem to correspond with previous pass-through studies. The fact that Japan displays this rather high pass-through may have to do with its period of economic stagnation during the 1990s, also known as the “Lost Decade”. This period most likely involved a more unstable monetary policy which would lead to a higher exchange rate pass-through (Devereux & Engel, 2001).
The long run elasticities are generally higher than the short run elasticities, consistent with previous literature. However, only one instance of the long run
elasticities is significant at the 5 percent level, namely Greece, and one is significant at the 10 percent level, Hungary. This is due to the strict requirement that for a long run pass-through to be considered significant at the 5 percent level, for instance, the
19
percent level at minimum. Relaxing this requirement to only the first two lags being significant would increase the number of significant long run elasticities to 7.
5.2 Estimates with distinction of crises
Below, the pass-through estimates when making a distinction between crisis and non-crisis observations are presented in columns (3) to (6) of table 3. Furthermore, the estimates per type of crisis are presented in columns (7) to (12).
The average short run elasticities of the benchmark, non-crisis and crisis
estimates do not differ very much, with elasticities of respectively 13, 13 and 12 percent. Thus, these do not tell much about differences between the benchmark, non-crisis and crisis estimates. Some differences do appear when looking at specific countries. It appears that roughly half of the countries’ pass-through elasticities increase during a crisis and roughly half of them decrease compared with non-crisis observations. Denmark, France, Germany, Japan, Poland and the U.S increase, the rest decreases or stays the same. It seems that, with the exception of the U.K., all large open economies have increasing exchange rate pass-through. This may have to do with the “power” and influence large open economies have to be able to set prices. Importers in smaller economies, on the other hand, might have to decrease the pass-through to maintain competitiveness during a crisis.
All in all, estimates of pass-through during crisis periods do seem to differ from estimates during non-crisis periods. In which direction the ERPT changes, depends on the country, however. An estimation of the short run pass-through elasticity with the data pooled for all countries for non-crisis observations and crisis observations yields elasticities of respectively 38 and 49 percent, statistically significant at the 1 percent level. The benchmark estimate with data pooled for all countries yields 43 percent for the short run pass-through, statistically significant at the 1 percent level. Although these elasticities do not give much insight in absolute values of ERPT, they do suggest that there are differences in pass-through between non-crisis and crisis periods.
Lastly, columns (7) to (12) of table 3 are meant to show how different crises affect the pass-through elasticities in different ways. The large amount gaps in the table indicates that there are too little observations to produce estimations for a large amount of countries. Only for Greece and New Zealand are there significant estimates of the short run elasticities for both the BoP crisis and the Debt or Banking crisis. Greece’s
20
short run pass-through is approximately the same, whether it is in a balance of payments crisis or not in a crisis at all. New Zealand’s short run ERPT, however, decreases from 12 to 7 percent, when in a BoP crisis, whilst it remains the same at 12 percent when in a debt or banking crisis. Both are not considered large open economies, yet they do not have the same direction of changes in pass-through. A possible
explanation for this is the type of balance of payment crisis. The BoP crisis comprises four different extreme capital movements: surge, sudden stop, sudden flight and retrenchment. A surge may for instance have a different effect on ERPT than a sudden stop, which might explain the difference, in this case, between Greece’s and New Zealand’s change in short run pass-through.
21
Table 3: Elasticities of exchange rate pass-through into aggregate import prices
The elasticities below “Benchmark” are the estimates in which there is made no distinction between non-crisis and crisis periods. “BoP crisis” refers to a Balance of Payments crisis. Statistical significance at the 10, 5 and 1 percent level are indicated by *, ** and ***, respectively.
Exchange Rate Pass-Through Elasticities
Benchmark No Crisis Crisis Currency crisis BoP crisis Debt or banking
crisis
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Country Short run Long run Short run Long run Short run Long run Short run Long run Short run Long run Short run Long run
Australia 0.20*** 0.30 0.19*** 0.25 0.19*** 0.19 - - 0.19*** 0.19 - - Austria 0.05 0.00 0.05 0.04 0.00 0.00 - - 0.10 0.00 - - Belgium 0.13*** 0.24 0.13*** 0.24 - - - - Denmark 0.13*** 0.30 0.11** 0.20 0.15*** 0.43 - - 0.16** 0.43 - - Finland 0.05* 0.05 0.09** 0.09 -0.04 0.00 - - -0.04 0.00 - - France 0.06** 0.06 0.01 0.00 0.19*** - - - 0.19*** 0.20 - - Germany 0.15*** 0.15 0.13*** 0.13 0.16*** - - - 0.16*** 0.16 - - Greece 0.42*** 0.85** 0.43*** 0.78 0.40*** 0.79 - - 0.42*** 0.83 0.27*** - Hungary 0.30*** 0.82* 0.33*** 0.49 0.24** 0.58 - - 0.18 0.35 - - Japan 0.29*** 0.29 0.21*** 0.21 0.40*** 0.40 0.99 - 0.17** 0.15 - - Mexico -0.02 0.00 0.00 0.00 -0.01 -0.01 - - -0.01 -0.01 0.01 0.04 New Zealand 0.10*** 0.28 0.12*** 0.34 0.06*** 0.18 - - 0.07*** 0.20 0.12*** - Poland 0.05 0.44 0.09 0.83 0.11 0.04 - - 0.14 0.07 - - Portugal 0.11* 0.69 0.14 0.29 0.03 0.60 - - 0.06 0.04 - - Sweden 0.14*** 0.12 0.14*** 0.14 0.12** 0.12 - - 0.06 0.00 0.14*** 0.56 U.K. 0.03** -0.01 0.05** 0.05 0.03 -0.07 - - 0.03 -0.08 - - U.S. 0.05** 0.23 0.03 0.14 0.06 0.11 0.09 - 0.03 0.08 - - Average 0.13 0.28 0.13 0.25 0.12 0.24 0.54 - 0.12 0.16 0.14 0.30
22
6. Conclusion
In this thesis I provide evidence on the pass-through of exchange rates into import prices in crisis periods as well as in non-crisis periods. The aim is to be able to reveal differences in pass-through, depending on whether there is a crisis or not. This is done using quarterly data on 17 OECD countries over the period of 1980 to 2016. I implement an empirical model based on micro-foundations similar to the model used by Campa & Goldberg (2005) to estimate pass-through elasticities per country for both the short run – one quarter – and the long run – four quarters. The estimation method is OLS
regression, incorporating data on the aggregate import price index, nominal effective exchange rates, gross domestic product and an exporter cost proxy. The first estimation of pass-through elasticities is without making a distinction between crisis and non-crisis observations and serves as a benchmark. The benchmark estimations yield average pass-through elasticities of 13% on the short run and 28% on the long run. Note that these estimates are considerably lower than the ones in previous research (Campa & Goldberg, 2005). This might be due to a different time period or a different import price index being used. The proportions between countries’ pass-through elasticities are roughly the same, however, and inferences about the differences in pass-through between countries thus remain valid. The estimations after the benchmark estimations do distinguish between crisis and non-crisis observations and even between different types of crisis: Currency, Balance of Payments and debt or banking crisis. The types of crises are identified using methods of Reinhart & Rogoff (2009), Forbes & Warnock (2012) and Broner, Didier, Erce, & Schmukler (2013).
With average short run estimates of ERPT of 12 percent with a crisis and 13 percent without a crisis, there does not appear to be a large difference. However, a look at the individual pass-through elasticities reveals that primarily the large open
economies show an increase in ERPT during a crisis whilst small open economies show a decrease in ERPT. This may have to do with the influence large open economies posess to be able to set prices, in comparison with smaller economies. This finding also answers the main research question: How is exchange rate pass-through affected by a crisis period? ERPT thus changes, but the changes differ per country. Mainly the large open countries experience an increase in short run pass-through, whilst the small open countries experience a decrease in short run pass-through. Long run pass-through estimates are not significantly different during a crisis, compared to a non-crisis period.
23
Knowledge on exchange rate pass-through is useful for monetary policy makers, among others. One of their goals is to keep prices stable. In this view, it is beneficial to know how a change in exchange rate affects the import prices and indirectly consumer prices and whether this differs during a crisis. In other words, the effect of exchange rate fluctuations on inflation during crisis periods and non-crisis periods becomes clearer. Furthermore, findings of this thesis might bring the academic world a small step closer to understanding the price puzzle of the insensitivity of consumer prices to exchange rate changes. Consumer prices are indirectly determined by import prices, which react differently during crisis times. This most likely requires further research in that area.
Besides, there is more room for improvement in future research on the topic of exchange rate pass-through, whether distinguishing crises or not. One rather large limitation of my research is the dataset. Especially data on the import price index is difficult to find and incomplete in the International Financial Statistics database of the IMF. Improving the data would be beneficial for the research on ERPT. Moreover, it would be useful to have data on the disaggregated imported price indices to be able to capture trends of changing import baskets that influence the aggregated price index but in fact not the pass-through. This data is not available at the OECD Statistics database anymore, but could perhaps be generated by hand or by collecting data from different statistics bureaus such as the Bureau of Labor Statistics (U.S.), the Office for National Statistics (U.K.) and the Central Bureau of Statistics (the Netherlands).
Moreover, finding one or multiple measures to indicate a crisis proves to be a difficult task and may benefit from extra research on the subject. Also, with the
estimations of pass-through elasticities during crises, one of the difficulties is that there are not a lot of crisis observations, which makes it hard to produce significant estimates of ERPT. A future search for an alternative way to estimate the ERPT during crises would therefore be highly appreciated.
Lastly, this research is done only on a dataset containing developed OECD countries. The results might be different for developing countries. Future research on developing countries would help to expand knowledge on the topic of exchange rate pass-through.
24
References
Broner, F., Didier, T., Erce, A., & Schmukler, S. L. (2013). Gross capital flows: Dynamics and crises. Journal of Monetary Economics, 60(1), 113-133.
Calvo, G. A., Izquierdo, A., & Mejía, L.-F. (2008). Systemic Sudden Stops: The Relevance of Balance-Sheet Effects and Financial Integration. NBER Working Papers, 14026. Calvo, G. A., Izquierdo, A., & Talvi, E. (2006). Phoenix Miracles in Emerging Markets:
Recovering Without Credit from Systemic Financial Crises. NBER Working Papers,
12101.
Campa, J. M., & Goldberg, L. S. (2005). Exchange Rate Pass-through into Import Prices.
The Review of Economics and Statistics, 87(4), 679-690.
Devereux, M. B., & Engel, C. (2001). Endogenous currency of price setting in a dynamic open economy model. NBER Working papers, 8559.
Devereux, M. B., & Engel, C. (2002). Exchange rate pass-through, exchange rate. Journal
of Monetary Economics, 49(1).
Devereux, M. B., Engel, C., & Storgaard, P. E. (2004). Endogenous exchange rate pass-through when. Journal of International Economics, 63(1), 263-291.
Forbes, K. J., & Warnock, F. E. (2012). Capital flow waves: Surges, stops, flight, and entrenchment. Journal of International Economics, 88, 235-251.
Frankel, J. A., & Rose, A. K. (1996). Currency crashes in emerging markets: An empirical.
Journal of International Economics, 41(3-4), 351-366.
Goldberg, P. K., & Knetter, M. M. (1997). Goods Prices and Exchange Rates: What Have We Learned? Journal of Economic Literature, 35(3), 1243-1272.
Junttilla, J., & Korhonen, M. (2012). The role of inflation regime in the exchange rate pass-through to. International Review of Economics and Finance, 24(1), 88-96. Laeven, L., & Valencia, F. (2008). Systemic Banking Crises: A New Database. IMF Working
Paper, 224.
Mishkin, F. S. (2008). Exchange rate pass-through and monetary policy. NBER working
paper, 13889.
Reinhart, C. M., & Rogoff, K. S. (2009). This Time Is Different: Eight Centuries of Financial
Folly. Princeton Press.
Taylor, J. B. (2000). Low in#ation, pass-through,. European Economic Review, 44(1), 1389-1408.
