Exchange Rate Pass-Through into
United Kingdom’s Import Prices:
Evidence from European Union Sectoral Data
Robin Badloe
∗Supervisor: dr. D.J.M. Veestraeten
Second reader: drs. N.J. Leefmans
University of Amsterdam
15 August 2016
Abstract
This thesis empirically examines the effect of the exchange rate of the British pound to other European Union currencies on the United Kingdom’s import prices. Ag-gregate data is used to estimate exchange rate pass-through (EPT) on the country level and disaggregated data is used to estimate EPT on the sector level over the period 1998 to 2015. The findings show that EPT estimates on the country level and the sector level differ from each other and thus that EPT varies between sectors. Also, the results show little evidence of asymmetric EPT; i.e., whether appreciations and depreciations are transmitted in a dissimilar magnitude to im-port prices. Asymmetric EPT only occurs after a half year at the aggregate, food and energy sector. In addition, the results of the model that assumes symmet-ric pass-through reveals that short run EPT is 0,118 on the country level and between 0,089 and 0,372 across sectors. Long run EPT is 0,167 at the country level and between 0,157 and 0,816 across sectors. Furthermore, most sectors are characterised by local currency pricing which implies that transactions of goods are mainly invoiced in the pound sterling.
JEL Classification: F10, F14, F30
Keywords: Exchange Rates, Import Prices, Pass-Through
∗ This thesis is submitted in partial fulfilment of the requirements for the MSc
Eco-nomics degree, specialisation International EcoEco-nomics and Globalisation, at the University of Amsterdam. The author has student number 10124675 and can be contacted via email address robin.badloe@student.uva.nl. He would like to thank dr. Dirk Veestraeten for his comments and assistance during the process of writing this thesis.
Statement of Originality
This document is written by Robin Badloe 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 origi-nal 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.
Contents
1 Introduction 4
2 Exchange Rates and Import Prices 7
2.1 Concept of Exchange Rate Pass-Through . . . 7
2.2 Determinants of Pass-Through . . . 8
2.2.1 Trade Elasticities . . . 9
2.2.2 Market Structure and Product Differentiation . . . 10
2.2.3 Contracts and Currency Choice . . . 11
2.3 Empirical Literature . . . 12
2.3.1 Aggregate and Disaggregated Data . . . 14
2.3.2 Asymmetric Pass-Through . . . 16
3 Empirical Application 18 3.1 Data Description . . . 18
3.1.1 Variables . . . 18
3.1.2 Time Series Graphs . . . 22
3.2 Methodology . . . 26 3.3 Empirical Findings . . . 29 3.3.1 Regression-Based Analysis . . . 29 3.3.2 Discussion . . . 37 3.3.3 Robustness Test . . . 39 4 Concluding Remarks 43 References 45 Appendix: Tables 47
1
Introduction
Within the European Union (EU), the United Kingdom (UK) is one of the few countries that still uses a non-euro currency. The UK uses the pound sterling as single legal tender and it has no plans to replace its currency. This is remarkable, because this means that the UK’s EU trading partners remain subject to currency risk when engaging in trade with UK businesses. Currency risk implies that a change in the exchange rate could have a di-rect effect on the currency converted price that British and other EU firms receive for their goods. As a result, when exchange rates experience large fluctuations, firms’ profits can become very volatile. In order to mitigate this risk, firms can, for instance, hedge their foreign currency exposure on the for-ward exchange market. On this market, they are able to agree to exchange currencies at a specified date in the future. By engaging in these types of contracts, producers can be sure of the amount of revenue they will receive for their products. Alternatively, firms can adjust their prices in response to exchange rate fluctuations. For instance, if the euro appreciates to the pound, a firm in the eurozone may decide to adjust its export price upwards (if invoiced in British pounds) in order to restore its currency converted prof-its. On the other hand, the eurozone firm may decide to change its export price in British pounds only partially and by doing so, it can partially absorb the exchange rate movement in its profit margin. The degree to which these price adjustments occur, is known as exchange rate pass-through (EPT).
In the empirical literature, EPT is studied by estimating the relation be-tween an effective exchange rate, which is the value of a country’s currency relative to a basket of other currencies, and an import price index while con-trolling for other variables, such as exporter’s marginal costs and aggregate demand from the importing country. Most of these studies are conducted on country level data. For the UK, pass-through estimates are typically in the region of 35 per cent for the short run and around 45 per cent for the long run (Campa and Goldberg, 2005; Faruqee, 2006; Mumtaz et al., 2011). These estimates imply that, on average, exchange rate movements are not completely passed through into UK import prices. The estimates also show that long run through tends to be larger than short run pass-through, which means that producers take time to adjust their export prices. Nonetheless, estimates on the aggregate level do not necessarily provide ac-curate information for EPT on the industry level.1 The theoretical literature has proposed several arguments on why EPT could differ across industries.
1In EPT studies, aggregate data studies are based on country level data and
disaggre-gated data studies are based on firm or industry level data. In this thesis, disaggredisaggre-gated data will be employed on the industry level.
Determinants such as the size and openness of an industry and the type of market structure could influence the degree of EPT across industries (Dorn-busch, 1987; Venables, 1990; Webber, 1994). To gain insights in the EPT of different industries, it is therefore necessary to study EPT based on disag-gregated data. To date, only two studies have examined EPT for the UK on the industry level. Estimates from Campa and Goldberg (2005) and Mumtaz et al. (2011) indicate that pass-through is incomplete on the industry level and EPT differs across industries for the short and long run.
Thus far, most studies have examined pass-through into import prices of a particular country in relation to all its trading partners’ exchange rates. Evidence of EPT based on trade between a smaller set of countries (e.g. bilateral trade relations) is limited. Particularly for the UK an interesting avenue would be to investigate EPT on its trade relation with the other EU countries. As stated by the Office for National Statistics, in 2014, more than 55 per cent of the imports originated from its 27 EU trading partners. This high percentage means that EU import prices are a large component of the total import price index of the UK. An EPT study on the relation between the UK and its EU trading partners could give a better understanding of the volatility of import prices to exchange rates of the British pound to the other EU currencies. Furthermore, mainly due to data limitations, fewer empirical studies have studied EPT on disaggregated data. As mentioned above, stud-ies on disaggregated level data provide a more accurate description of EPT as compared to studies on aggregate level data. To address these gaps in the literature, this thesis will study pass-through of the UK by specifically focus-ing on imports from its EU tradfocus-ing partners. Pass-through will be estimated on the aggregate and disaggregated level. To identify the factors influencing EPT, this thesis will put an emphasis on the microeconomic determinants. The research question is as follows:“What is the effect of the exchange rate of the British pound to other European Union currencies on the United King-dom’s import prices?”
In order to provide a concise and coherent answer to the above-mentioned research question, the following data and methodology will be employed. A dataset with quarterly data over the period 1998 to 2015 will be used. To estimate the relation between import prices and exchange rates, country and sector level import price indices from EU exports to the UK will be used and an effective exchange rate based on exchange rates between the UK and its EU trading partners will be constructed. To mitigate omitted variable bias, UK gross domestic product (GDP), UK producer price indices (PPIs) and consumer price indices from EU exporting countries will be employed as control variables. The empirical analysis will first focus on examining the variables for stochastic trends by applying unit root tests. Second, the EPT
relation will be extended to allow for asymmetric pass-through. This implies that EPT is different after a depreciation as opposed to an appreciation. This may, for example, occur if EPT is lower after a disadvantageous movement in the exchange rate (i.e., when it negatively impacts the profit margin), than after a beneficial exchange rate movement (i.e., when the profit margin increases). Finally, the resulting EPT regressions will be estimated on the country and industry level, in which pass-through for the short and long run will be distinguished.
The outcomes of this study are relevant for policy purposes, because EPT has an influence on the conduct of monetary policy and the choice between a fixed versus flexible exchange rate regime. Low pass-through implies a small number of price adjustments after an exchange rate shock. Imported goods will then experience less inflation or deflation after shocks to the exchange rate. Thus, there will be less need for contractionary or expansionary mon-etary policy to combat imported inflation or deflation, respectively. For the choice of the exchange rate regime, the size of the effect of the exchange rate shock upon the country’s import prices is important. When EPT is low, an exchange rate shock will result in a small expenditure-switching effect and hence only a small change of the current account. If a current account im-provement is a main goal, then a larger exchange rate shock is necessary and thus a more flexible exchange rate regime is desirable (Engel, 2003). How-ever, a discussion on the proper conduct of central bank policy in relation to EPT is outside the scope of this thesis.
The remainder of this thesis is organised as follows. Section 2 describes the concept of EPT and it gives a brief overview of the literature. This overview contains a summary of the main theoretical explanations and em-pirical findings on EPT with a particular focus on studies on the UK. Sec-tion 3 contains the empirical applicaSec-tion. The methodology and data are described, after which the estimation results are discussed. Finally, Section 4 contains the concluding remarks.
2
Exchange Rates and Import Prices
This section defines exchange rate pass-through and brings an overview of the literature on this subject. Section 2.1 gives a detailed explanation of the notion of exchange rate pass-through. Section 2.2 presents a review of the main theoretical insights on the relation between exchange rates and import prices. Finally, Section 2.3 contains a discussion of the empirical literature. In particular, emphasis is placed on studies related to the empirical approach of this thesis.
2.1
Concept of Exchange Rate Pass-Through
Exchange rate pass-through can be defined as the percentage change in local currency import prices resulting from a one per cent change in the nomi-nal exchange rate between the exporting and importing countries (Goldberg and Knetter, 1997, p. 1248). If initially the price of the imported good is equal to the currency converted price of the exported good, then the situ-ation before the exchange rate movement can be formally expressed as follows
Ptm,i= EtbiPtx,j, (1)
where i and j stand for the relevant countries, t is the time period, Pm is the import price in the importing country’s currency, Ebi is the bilateral
ex-change rate (currency units of the importing country needed to purchase one currency unit of the exporting country) and Px is the export price in the exporting country’s currency. EPT can be determined by investigating how much Ptm,i changes when Ebi
t changes, holding P x,j
t constant.
Whether EPT is an issue for producers depends on the pricing policy followed by these producers. When a producer adopts a pricing policy called local currency pricing (LCP), transactions are invoiced in the currency of the importing country. EPT is then zero per cent. When a producer adopts a policy called producer currency pricing (PCP), transactions are invoiced in the currency of the exporting country. EPT is then hundred per cent and the producer is not exposed to foreign exchange risk. After making a decision on the pricing policy, a firm still has the freedom to adjust its currency con-verted export prices. If a firm desires low exchange rate pass-through before it has a chance to adjust export prices, the firm is better off choosing LCP that results in zero per cent EPT prior to adjustment. On the contrary, if desired exchange rate pass-through is high, the firm should choose PCP that
results in hundred per cent EPT prior to export price adjustment (Gopinath et al., 2010).
As a numerical example, suppose that a British person imports a car from Germany for 90.000 euro. Initially, the pound to euro exchange rate is at 0,5; hence, the car costs 45.000 pounds for the British importer. Assume that a period later, the euro appreciates and as a result the exchange rate changes to 0,6. The car will now cost 54.000 pounds for the British importer. How-ever, the German carmaker may decide to keep its price at 45.000 pounds for the UK importer. In that case, EPT is absent and the producer applies LCP. If the price of the car would fully change with the exchange rate movement (so to 54.000 pounds), EPT is said to be complete and the producer applies PCP. Any adjustment to a price between 45.000 and 54.000 pounds is called incomplete EPT. Furthermore, there are two theoretically plausible cases of perverse pass-through. In the first case, the exporter adjusts its price to more than 54.000 pounds. In the second case, the exporter adjusts its price to less than 45.000 pounds. These cases could occur when marginal costs partially depend on the exchange rate owing to imported inputs (Okuguchi, 2005). In the event of more expensive imported inputs, marginal costs for producing final goods will increase. The increased marginal costs and the exchange rate movement may then both be passed through to the export price of the automobile. In the event of less expensive imported inputs, the opposite mechanism may occur. A different reason for the exporter to lower its car price below 45.000 pounds, could be the strategy to take a large mar-ket share in the current period. The firm aims to obtain more marmar-ket power in the current period and it could take advantage of the gained market power by increasing prices in the next period in order to maximise intertemporal profits (Froot and Klemperer, 1989).
2.2
Determinants of Pass-Through
Several theories have been proposed to explain the extent of exchange rate pass-through. The main theoretical insights try to relate EPT to its mi-croeconomic determinants. They can be divided in the trade elasticities approach, market structure and product differentiation, and contracts and currency choice. These insights are discussed in the following three sections.
2.2.1 Trade Elasticities
Early theoretical approaches on EPT tried to measure the degree of pass-through by using trade elasticities. In these approaches, an import demand function for the importing country and an import supply function for the ex-porting country are determined. The import demand function is dependent on the importing countries’ currency price for the product and the import supply function is dependent on the exporting countries’ currency price for the product multiplied by the nominal exchange rate.2 Both functions are
then differentiated and equalised to determine an equilibrium condition for the market. The resulting expression shows that EPT is dependent on the price elasticities of demand and supply for the imported product. If supply or demand for imports is perfectly elastic (inelastic), then pass-through is going to be complete (absent) (Menon, 1995).
Kreinin (1977) applies this framework to analyse small and large coun-tries. He argues that a small open country faces an infinitely (or very highly) elastic supply of imports from its relatively larger trading partners. As a consequence, it is likely that this small country experiences nearly complete pass-through into import prices. Conversely, a large less open country faces less elastic supply of imports from foreign countries and is therefore expected to experience partial or zero pass-through. By the same reasoning, large countries have a more elastic export supply than small countries. There-fore, exporters from a large country are likely to pass through a greater proportion of an appreciation or depreciation than exporters from a small country. Webber (1994) states that these findings can be generalised further to explain pass-through for small open and large less open industries as well. Measurement of EPT based solely on the elasticities approach has a num-ber of shortcomings. First and foremost, in order to determine EPT the elas-ticities of import demand and supply have to be known and they have to be known precisely. In practice, it is hard to estimate these elasticities (Kreinin, 1977). Second, this approach only shows comparative statics for the short run; it does not show the dynamic effect of exchange rates on prices over mul-tiple periods. It may be possible that EPT runs over mulmul-tiple periods. In order to give a complete description of EPT, long run EPT should then also be calculated by adding up the effects over multiple periods (Menon, 1995). Third, studies of EPT solely based on trade elasticities are restrictive; un-derlying determinants of supply responses of producers are not taken into account. Determinants such as market competition and product
characteris-2 Where the nominal exchange rate is defined as the number of currency units of the
importing country per currency unit of the exporting country. This definition also applies to the remainder of this thesis.
tics could have an impact on the degree of pass-though and could therefore be relevant in explaining EPT. A second strand of literature aims to address these determinants. The next section reviews this literature.
2.2.2 Market Structure and Product Differentiation
More recent theories point to the role of the market structure and prod-uct differentiation. In general, they find a relation between the degree of competition in markets and EPT. There tends to be a different extent of pass-through in industries with different competitive structures. Related to the degree of market competition is product differentiation. The degree of product differentiation determines the degree of substitutability between the domestic and imported product. It increases the price setting power of firms and as such it is positively related to the degree of pass-through (Devereux et al., 2004; Menon, 1995).
Venables (1990) examines EPT in markets with perfect competition. In perfectly competitive markets, there is no mark-up between the price and marginal costs and the supply curve is equal to marginal costs. An exchange rate movement affects the supply curve and this movement is completely or partially passed through to the price of the imported good. It is completely passed through when the supply of the imported good is infinitely elastic and it is partially passed through when the supply curve is less than infinitely elastic. In the later case, a second effect is a decrease in the volume of im-ports supplied.
In practice, many industries do not meet the conditions of perfect com-petition; they are considered to be imperfectly competitive. Fisher (1989) considers the case of a monopoly. If the importing market is sufficiently seg-mented from the exporting market, a monopolist passes through exchange rate movements to importers. His analysis shows that EPT is lower after a disadvantageous movement in the exchange rate (i.e., when it negatively impacts the profit margin), than after a beneficial exchange rate movement (i.e., when the profit margin increases). Thus, asymmetric exchange rate pass-through occurs in his analysis. Dornbusch (1987) considers the outcome under oligopoly. He develops an oligopoly model based on Cournot competi-tion and shows that EPT is higher (closer to completeness) when the number of exporting firms is large relative to the number of firms on the importing market. Furthermore, he finds that incomplete EPT occurs due to changes in mark-ups. Firms with market power earn a profit margin on each unit sold. They can absorb part of the exchange rate movement by varying their mark-ups. In this way, exchange rate fluctuations are only partially passed
through to importers.
Froot and Klemperer (1989) take a different approach than the aforemen-tioned authors. They analyse EPT when firms only care about maintaining or increasing their market shares. In their model, producers have the incen-tive to care about their market shares, because future market shares depend on current market shares. An increase in the current market share will then more likely keep the firm in business for a longer period. To maintain or in-crease a firm’s market share, a producer behaves differently after a permanent exchange rate fluctuation than after a temporary exchange rate fluctuation. A permanent appreciation of the importing country’s currency motivates the exporting firm to price its goods very competitively in order to gain market share. Its current market share may then increase or remain stable depending on the pricing policies of its rivals. The price adjustment after a temporary appreciation or depreciation is ambiguous; the price could either increase or decrease depending on the producer’s expectations of its own market share and its rivals’ prices.
2.2.3 Contracts and Currency Choice
In international trade, it is very common for exporters and importers to con-tractually commit to a fixed price for goods and services. This decreases, among others, uncertainty over the price as a result of currency price move-ments. Firms may also decide to hedge their foreign currency exposure on the foreign exchange market. On this market, they are able to purchase for-eign currency derivatives, which could effectively eliminate all the exchange rate risk. In essence, trade, forward and futures contracts are used to lessen the need for export price adjustments. Thus, these types of contracts could be factors limiting the extent of EPT.
Magee (1973) explores the relation between import prices and the term of contracts. He finds that contracts cause a number of lags in the price adjustment process. First, the new economic situation has to be recognised, second a decision to change prices has to be made and finally, the product has to be delivered. During these time lags, the prices of the products stay unchanged. So the adjustment of import prices will partly depend on the frequency and speed with which contracts are renegotiated. Furthermore, he notes that the currency in which to denominate contracts is an important decision variable for traders. The decision depends on the expectations of the relative strength of both the home and foreign currency. In this decision, importers and exporters have conflicting incentives, because a capital gain for one party due to an exchange rate shock is a capital loss to the other
party. When making this decision, the relative market power of the traders can become an important determinant.
Gopinath and Rigobon (2008) study imports at the US borders. Their observations of micro-level data reveal that the median price duration in the currency of pricing is 11 months for imports. This implies that price stickiness occurs over the period of these contracts. Additionally, their find-ings point to incomplete pass-through and the extent of pass-through differs across the type of goods. Moreover, they argue that non-price aspects can become important for the frequency of price adjustments and the level of im-port prices set in contracts. For instance, the relation and bargaining power between the buyer and seller may influence the resulting import prices.
Allayannis and Weston (2001) investigate firms’ hedging policies. They examine the extent to which firms hedge their exposure to sales in foreign currency and the effect of hedging on the firm value. Their sample consists of 346 US non-financial firms with foreign sales between 1990 and 1995. Firstly, they find that around 55 to 65 per cent of the firms hedge their foreign cur-rency exposure by using derivatives over the years of their sample. Secondly, their empirical results show that firms that adopt hedging policies, increase in value above those firms that choose to remain unhedged and that firms that stop hedging, experience a decrease in firm value relative to those firms that choose to remain hedged. Hence, next to a decrease in currency risk, a gain in firm value is a second incentive for producers to engage in hedg-ing. Based on the outcomes, it can be argued that hedging, which limits the need for export price adjustments, occurs often and as a consequence, a lower number of import price adjustments can be expected in markets with hedging firms.
2.3
Empirical Literature
In the empirical literature, EPT is estimated by rewriting equation (1) in the following manner. Equation (1) is transformed into logarithms as follows
pmt = ebit + pxt, (2)
where lowercase letters reflect logarithms and the country superscripts are dropped for notational brevity. The right hand side consists of the exporter’s marginal costs and a mark-up over the marginal costs
where mkup is the mark-up and mc is the marginal cost (both expressed in logarithms). Substituting expression (3) into expression (2) results in
pmt = mkupxt + mcxt. (4)
Mark-ups depend on a fixed exogenous component related to industry-specific characteristics and a component that moves with the exchange rate
mkupxt = α + βebit. (5)
The exporter’s marginal costs are rising in destination market demand con-ditions ymt and export market wages wtx 3
mcxt = φytm+ δwtx. (6)
Substitution of equation (5) and (6) into equation (4) results in the following final expression
pmt = α + βebit + φytm+ δwtx. (7)
The expression shows that import prices are a function of the exchange rate, aggregate demand of the destination market and wages of the exporter’s mar-ket (Campa and Goldberg, 2005). Based on this expression, the following regression is often estimated in the empirical literature and is considered as the standard pass-through regression4
4pt = α + β 4 et+ γ 4 Dt+ µ 4 ct+ t, (8)
where 4p is the percentage change in the import price into the importing
3As stated at the end of Section 2.1, exporters could use imported inputs for producing
their final goods. The exporter’s cost function will then contain imported input costs and the marginal cost expression should then also include an exchange rate variable. This variable is left out of equation (6) for simplicity.
4 In practice, producers could base their export price adjustments on their
expecta-tions of exchange rate movements. Their export price adjustments to expected temporary exchange rate movements could differ as opposed to expected permanent exchange rate movements. In the exchange rate variable of regression (8), no distinction is made between temporary and permanent exchange rate movements. All observed exchange rate changes are considered to be permanent. This can be considered as a drawback of regression (8).
country at time period t, expressed in the importing country’s currency, 4e is the percentage change in the importing country’s currency per export-ing country’s currency unit, 4D is a vector of variables that affect import demand, such as competitors’ prices or income, in the importing country’s currency and in first differences, 4c is the percentage change in the marginal cost, such as labour costs, in the exporting country’s currency and is the error term. Superscripts are left out for notational brevity. Pass-through is said to be complete if β = 1, incomplete if 0 < β < 1, absent if β = 0 and perverse if β > 1 or β < 0. In recent EPT studies, the exchange rate variable is extended by including a number of time lags. In this way, short run and long run pass-through can be distinguished. Short run EPT is equal to the coefficient on the contemporaneous exchange rate variable and long run EPT is equal to the sum of the coefficients on the contemporaneous exchange rate variable and all the lags of the exchange rate variable.
The next sections are organised as follows. Section 2.3.1 reviews empirical studies which are based on regression (8). Section 2.3.2 presents an overview of studies which are based on an extension of regression (8). The extension allows to study asymmetric exchange rate pass-through; i.e., whether ap-preciations and deap-preciations are transferred in a dissimilar size into import prices.
2.3.1 Aggregate and Disaggregated Data
There are a number of empirical studies that estimate regression (8). Typ-ically, these studies estimate the EPT of a certain country against all of its trading partners. For the import price variable, these studies use an aggre-gate or disaggreaggre-gated import price index. Aggreaggre-gate import price indices consist of the weighted average of the import price indices of all the trading partners. Disaggregated import price indices consist of the weighted average of import price indices of all the trading partners on the industry level. For the exchange rate variable, these studies use an effective exchange rate which consists of a traweighted basket of foreign currencies. For the import de-mand variable, most studies simply employ the real GDP of the importing country. It is found to be more difficult to gather data to control for marginal costs, because data from all trading partners is needed. Countries such as the United States and the United Kingdom have a large number of trading partners which means that marginal cost data from all of these partners is required. In most cases, marginal cost data is not available from all part-ners. A variety of alternative cost variables are utilised to solve this issue, for instance, by employing the wage rate of the importing country or consumer
price indices of the exporting countries. With respect to the methodology employed in the empirical studies, older papers estimate the relation with ordinary least squares (OLS), whereas more recent studies also explore the data for time series properties, such as stochastic trends, before applying OLS (Menon, 1995).
No paper has been written on the research question of this thesis. The most closely related empirical papers are from Campa and Goldberg (2005) and Mumtaz et al. (2011). These two papers estimate EPT for the UK against all of its trading partners on the aggregate and disaggregated level.
Campa and Goldberg (2005) study the EPT relation for the UK for the period 1975 to 2003 on quarterly data. As control variable, they develop a proxy of the exporter’s costs by taking the unit labour cost of the UK and multiplying this by the nominal effective exchange rate divided by the real effective exchange rate. By adjusting the unit labour cost in this way, they state that the variable becomes representative for the marginal costs of the exporting countries. In addition, they control for import demand factors by using the real GDP of the UK. They test their data for non-stationarity and find that the import price, exchange rate and marginal cost variable con-tain stochastic trends. Therefore, they estimate the EPT relation in first differences. Their aggregate data results indicate partial EPT of 36 per cent per quarter (short run) and partial EPT of 46 per cent per year (long run). They test the short and long run coefficients for LCP and PCP. For both time frames, LCP and PCP is rejected at the 5 per cent significance level. Nonetheless, aggregate data estimation results tend to have a shortcoming. Indeed, the authors argue that aggregate EPT estimates differ from disag-gregated EPT estimates due to heterogeneity in market competition across industries. To overcome this drawback, the authors also perform their study on disaggregated data. They use import price information from five sectors: food, energy, raw materials, manufacturing and non-manufacturing. The results indicate that all industries have incomplete pass-through. Further, EPT varies across sectors which means that, on average, producers in dif-ferent industries respond difdif-ferently to exchange rate shocks. Short run and long run estimates are in the range of 7 to 44 per cent and 39 to 52 per cent, respectively. Long run pass-through for each industry is larger than short run pass-through. This signals that it takes more than one quarter for producers to adjust their prices. Overall, the size of EPT on the industry level can be ordered in the following manner (from high to low): food, raw materials, manufacturing, non-manufacturing and energy.
Mumtaz et al. (2011) investigate EPT for the UK over the period 1984 to 2004 based on quarterly data. They employ the same control variables and methodology as Campa and Goldberg (2005). They only make one
alter-ation to the EPT regression by including lagged import prices as independent variables. This aims to capture the reluctance of producers to adjust prices quickly. In other words, the authors assume gradual adjustment of import prices to exchange rate movements. The findings on aggregate data show EPT estimates in the short run of 38 per cent and in the long run of 49 per cent. On the aggregate level, they do not find evidence of LCP or PCP on the 10 per cent significance level. The results are comparable to those of Campa and Goldberg (2005). Next, they perform a detailed study based on disag-gregated data. They use import price indices of the following four SITC5
categories: food, raw materials, energy and manufacturing. Their results on sectoral data show incomplete pass-through for the short run and long run in the range of 2 to 42 per cent and 2 to 54 per cent, respectively. Long run pass-through estimates are larger than short run pass-through estimates, except for the energy sector. For the energy sector, short run and long run pass-through is 2 per cent. Overall, the size of EPT on the industry level can be ordered in the following manner (from high to low): manufacturing, raw materials, food and energy.
2.3.2 Asymmetric Pass-Through
Several authors have extended the usual pass-through regression to allow for asymmetric exchange rate pass-through. Asymmetric pass-through implies that appreciations and depreciations are transmitted in a dissimilar magni-tude to import prices. As a numerical example, assume that a 1 per cent depreciation of the importing country’s currency is followed by a 0,5 per cent increase in import prices. If pass-through is symmetric, a 1 per cent ap-preciation of the importing country’s currency is followed by a 0,5 per cent decrease in import prices, otherwise EPT is asymmetric.
In theory, there are a number of reasons to expect asymmetric EPT. Knet-ter (1994) discusses a scenario in which firms face production bottlenecks. Firms may want to pass-through an exchange rate movement which would lower export prices. But these lower prices will not result in higher sales because producers are constrained in their production capabilities and there-fore, they are not able to deliver. Therethere-fore, producers may want to keep their prices constant or increase their prices given the amount of sales. De-latte and L´opez-Villavicencio (2012) argue that the competitive structure of
5SITC stands for Standard International Trade Classification. This is a classification
of goods used to classify the exports and imports of a country. This enables comparison of trade data from different countries and years in a consistent manner. The classification system is maintained by the United Nations.
an industry is an important determinant in the producer’s pricing decision. On the one hand, in a competitive market, producers that adopt PCP may decide to absorb a depreciation of the importing country’s currency and keep their prices constant to gain or maintain market share, instead of increas-ing prices. In contrast, firms with high market power and market share in the importing country, have less incentive to reduce margins and hence pass through appreciations of the importing country’s currency. An oligopolistic firm will therefore adjust its prices more after a depreciation than after an appreciation.
Asymmetric pass-through has received little attention in the empirical literature so far. Only one attempt has been made to apply asymmetric EPT to UK data.6 Bussi`ere (2013) sets up a regression similar to equation
(8), so he does not separate pass-through between short run and long run estimates. To study asymmetric EPT, he includes dummy variables which separate appreciations from depreciations. He uses quarterly data on the aggregate level and applies UK GDP and UK PPIs as control variables. The author’s results do not show evidence of asymmetric EPT for the UK over the sample period 1980 to 2006, but he emphasises that his estimates may be exposed to aggregation bias. In the aggregate, cross-sectoral differences in the form of asymmetries may be cancelled out. Therefore, sectoral data may give more insight into asymmetric EPT. Bussi`ere could not study sectoral EPT due to data limitations, but he suggested that it could be an avenue for future research.
This section defined the concept of EPT and also gave an overview of the theoretical and empirical literature. In particular, the theoretical literature points to market competition and trade contracts as factors that influence the extent of EPT. The previous empirical findings mainly show incomplete and heterogeneous estimates of EPT across industries. The following section, Section 3, will present the empirical study on the measurement of EPT and therefore will be able to comment on the empirical validity of the theoretical insights discussed in Section 2.
6See e.g., Pollard and Coughlin (2003) and Yang (2007) for applications to US
3
Empirical Application
The goal of this section is to perform the empirical study. The variables and data are described in detail in Section 3.1. Section 3.2 discusses the methodology used to measure exchange rate pass-through for the UK. Fi-nally, Section 3.3 presents and interprets the estimation results.
3.1
Data Description
Section 3.1.1 will first describe the data employed for each variable. Second, in Section 3.1.2, data of the two most important variables, the import price variable and the exchange rate variable, will be illustrated by use of time series graphs.
3.1.1 Variables
The data employed in this thesis consists of quarterly data from the EU; where the EU consists of 28 countries: Austria, Belgium, Cyprus, Estonia, Finland, France, Germany, Greece, Ireland, Italy, Latvia, Lithuania, Lux-embourg, Malta, Netherlands, Portugal, Slovakia, Slovenia, Spain, Bulgaria, Croatia, Czech Republic, Denmark, Hungary, Poland, Romania, Sweden and the United Kingdom. The data is over the period 1998-2015 which corre-sponds to 18 years of data. The data for the dependent variable (import prices), the data for the independent variables (exchange rates, consumer prices, gross domestic product, competitors’ prices and labour costs) and all other data are discussed below consecutively.7
Import Prices. Import price indices will be used as dependent variable, as stated in regression (8). The import price indices are constructed by the Office for National Statistics by taking an import-weighted average of import prices of all trading partners within the EU. Import price indices are collected on the aggregate and disaggregated level. The import price index on the ag-gregate level is based on the total amount of imports of goods of the UK from all other EU countries. The import price indices on the disaggregated level are based on sector level data. Sector level data indexed according to the SITC8 will be employed, as also examined by Mumtaz et al. (2011). The
following import price indices will be used: import price index SITC level
7Summary statistics of all variables are provided in Table A.1 in the Appendix. 8See Table A.2 in the Appendix for a detailed overview of the SITC levels.
0+1: food, beverages and tobacco; import price index SITC level 3: fuels; import price index SITC level 2+4: basic materials and import price index SITC level 5+6+7+8: manufacturing.9 The data is not seasonally adjusted
and it is available for the whole sample period on quarterly basis and it comes from the Office for National Statistics, UK.
Exchange Rates. An effective exchange rate index will be constructed and used as independent variable. This variable will be used to measure the ef-fect of exchange rates on import prices, which is the main objective of this thesis. Commonly, effective exchange rates are calculated by using imports plus exports from one trading partner divided by the total imports plus ex-ports from all trading partners as weight. However, import price indices are calculated by taking only imports in the calculation of weights. Therefore, an import-weighted effective exchange rate index is more appropriate.10
The effective exchange rate index will be calculated by taking an import-weighted average of spot exchange rates from all EU trading partners. First, each exchange rate will be indexed by taking the first quarter 1998 as base, second the effective exchange rate will be calculated as follows
Et= Et−1∗ N X i=1 Wi,t ∗ Ebi i,t Ebi i,t−1 ! , (9)
where E stands for the value of the effective exchange rate index at time t, N stands for the number of currencies in the index, Ebistands for the exchange
rate of the pound to currency i and W is the weight of each currency, where the weight is equal to UK imports from one trading partner divided by to-tal UK imports from all trading partners.11 The sum of the weights equals
one. Each exchange rate is defined as the pound to currency i exchange rate; where currency i stands for the euro for 19 eurozone countries and a
9The import price index of SITC level 9, which consists of unspecified goods, will not
be included in the analysis, because the composition of this index is unclear. It is therefore difficult to determine which control variables are needed in estimating EPT. Furthermore, the unclear composition of the index will make it difficult to interpret estimation results.
10E.g., the International Financial Statistics (IFS) uses total trade to determine country
weights in the effective exchange rate. However, for a study on EPT, only weights on imports are relevant. So this is one advantage of the approach of this thesis as compared to other EPT studies that use IFS data such as Campa and Goldberg (2005) and Faruqee (2006).
11 See Table A.3 in the Appendix for the weights used in constructing this index. It
can be clearly seen that eurozone countries form the largest component of the effective exchange rate index. Thus, the EPT regressions, which will be estimated in Section 3.3, will mainly measure the effect of the pound to euro exchange rate on UK import prices.
non-euro local currency for 9 other EU countries (assuming that the exchange rate situation in 2015 already applied since 1998). The 19 eurozone countries that have adopted the euro as their official currency are: Austria, Belgium, Cyprus, Estonia, Finland, France, Germany, Greece, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, Netherlands, Portugal, Slovakia, Slovenia and Spain. The 9 other countries each have their local non-euro currency: Bulgaria (BUL) has the lev, Croatia (CRO) has the kuna, the Czech Repub-lic (CZE) has the koruna, Denmark (DEN) has the krone, Hungary (HUN) has the forint, Poland (POL) has the zloty, Romania (ROM) has the leu and Sweden (SWE) has the krona. The spot exchange rates are calculated by taking the exchange rate at the middle of the day, which is at or around 16:00 London time.12 Data is available for the whole sample period on
quar-terly basis. The source of the data is WM/Thomson Reuters.
Consumer Prices. General consumer price levels could rise due to increases in wage and raw material costs. Therefore, inflation can be used as an ap-proximation of marginal cost changes. Eurostat measures consumer price levels with a Harmonised Index of Consumer Prices (HICP) for each EU country. A HICP consisting of 22 EU trading partners will be constructed as follows
HICPt = N
X
i=1
Wi,t∗ HICPi,tcountry, (10)
where HICP stands for the value of the constructed consumer price index at time t, N stands for the number of countries in the index, HICPcountry stands for the HICP, as measured by Eurostat, of country i and W is the weight of each country in the index, where the weight is equal to UK imports from one trading partner divided by the total UK imports from all trading partners. The sum of the weights equals one. Five EU countries (Croatia, Czech Republic, Hungary, Romania and Slovenia) are left out of the calcula-tions due to missing data. Because of their small import sizes, no substantial effects on the final indices are expected. Indices will be used on the aggregate level which covers all items of goods and on the disaggregated level. On the disaggregated level, a HICP for the food sector, energy sector and an index excluding the food and energy sector will be utilised. Data is available for the whole sample period on quarterly basis. The data comes from Eurostat.
12 This time is considered to be the middle of the ‘global day’ and the time of highest
Gross Domestic Product. GDP of the importing country will be used to control for import demand. GDP is defined as the total value of output in the economic territory. An increase in GDP is expected to raise marginal costs for industries with decreasing economies of scale in the following manner. A higher GDP could result in a higher level of demand for imports. Firms could increase their production levels in order to meet the higher level of demand. When increasing the production levels, it is possible that firms will have to incur higher marginal costs, for example, due to increases in wage levels or due to less efficient production. The increased marginal costs could then be transferred into import prices. Nevertheless, some industries are characterised by increasing economies of scale. For these industries higher production could result in lower marginal costs (Ethier, 1982). In short, changes in GDP could lead to changes in import price levels and this re-lation has to be controlled for when studying EPT. The UK’s GDP will be employed and it is denominated in sterling pounds. GDP is calculated in real terms by keeping 2012 prices constant. GDP is seasonally adjusted which makes it possible to control for the underlying trend of import demand be-tween periods. Data is available for the whole sample period on quarterly basis. The data comes from the Office for National Statistics, UK.
Competitors’ Prices. It is likely that export prices of the 27 EU exporting firms are affected by prices from competing firms in the UK. Two reasons for this relation are price competition and collusion between the UK and other EU firms. Therefore to control for competitors’ prices, a PPI from the UK will be used as control variable. A PPI measures the average selling prices received by domestic producers for their output. PPIs will be utilised on the aggregate and disaggregated level. On the aggregate level, the PPI is com-posed of selling prices from domestic producers of the whole UK economy. On the disaggregated level, PPIs will be used for the following sectors: in-termediate goods, non-durable consumer goods, manufacturing and energy. Data is available for the whole sample period on quarterly basis. The data is not seasonally adjusted and comes from Eurostat.
Labour Costs. To approximate marginal costs, a nominal unit labour cost index of all 27 EU trading partners will be constructed. Where the nominal unit labour cost is defined as the ratio of labour costs to labour productivity. Eurostat has nominal unit labour cost indices available for each EU trading partner. However, labour costs as an approximation for marginal costs has a drawback. Typically, marginal costs of firms contain raw material costs. This part is not included in the labour cost index. An import-weighted average of all 27 EU trading partners will be calculated to end up with a representative
labour cost index. The formula employed is similar to equation (10), but now using labour cost indices instead of HICP indices. Data is missing for Croatia, Malta and Poland for the years 1998 and 1999. Therefore, these countries are left out of the calculation of the index for the years 1998 and 1999. For all other countries, data is available for the whole sample period. Data on labour costs is only available on aggregate level. Therefore, this variable can only be used in the estimation on the country level. This es-timation is provided in the section with the robustness test, Section 3.3.3. The data is on quarterly basis and comes from Eurostat.
Imports. In order to make indices for exchange rates, consumer prices and labour costs, the weights of each EU trading partner in the indices will have to be determined. Data on imports will be used to calculate the country weights. Annual data is available and the imports are from goods and ser-vices. More appropriate would be to use quarterly data of imports from goods-only. However, this data is not available. Hence, the calculated coun-try weights will be assumed to equal weights based on data from goods-only and the weights are considered to remain stable during the year. The data covers the period 1999 to 2014. For the years 1998 and 2015, data is not available on imports. To solve this issue, imports for the year 1998 will be assumed to equal the import data from the year 1999; and imports for the year 2015 will be considered to equal the import data from the year 2014. The source of the data is the Office for National Statistics, UK.
3.1.2 Time Series Graphs
Data from the import price variable and the exchange rate variable are plot-ted in Figure 1, 2, 3 and 4. For each data series, Q1 1998 is taken as base quarter and all series equal 100 at Q1 1998. Setting Q1 1998 as base quarter makes it easy to compare the series with each other and to follow the path of each series. In particular, it is interesting to see the movements of sector level import price indices compared to the aggregate import price index, which is equal to the weighted average of all sector level import price indices. The comparison gives an indication of how related the import price indices are to each other. Also, the nominal exchange rates and the effective exchange rate index are plotted in order to discuss some interesting features of the data.
Figure 1: UK Import Price Indices from EU Trade Only on Country Level and Industry Level SITC 0+1, 2+4 and 5+6+7+8, 1998-2015
Source: Calculations based on data from the Office for National Statistics, UK.
Figure 2: UK Import Price Indices from EU Trade Only on Country Level and Industry Level SITC 3, 1998-2015
Source: Calculations based on data from the Office for National Statistics, UK.
Import price index data is plotted in Figure 1 and 2. Figure 1 shows that the paths of disaggregated data and aggregate data differ from each other.
The import price index of SITC Level 2+4 (raw materials sector) and the import price index of SITC Level 0+1 (food sector) are more volatile as compared to the whole economy data and the import price index of SITC Level 5+6+7+8 (manufacturing sector). The import price index of SITC 5+6+7+8 (manufacturing sector) follows the aggregate import price index closely. The import price index of the whole economy, the import price index of SITC level 0+1 (food sector) and the import price index of SITC level 2+4 (raw materials sector) rise up to 2012, after which they moderately start to decline. The import price index of SITC level 5+6+7+8 (manufacturing sector) first decreases up to 2007, after which it starts to increase up to 2012. After 2012, it starts to experience a modest decline similar to the aggregate import price index.
Figure 3: Five Bilateral Exchange Rates of Currencies Relative to the Pound and the Effective Exchange Rate Index, 1998-2015
Source: Calculations based on data from WM/Thomson Reuters. Note: The effective exchange rate index, the pound to lev, the pound to euro and the pound to krone look
the same in the figure.
Also, in Figure 2, the paths of disaggregated data and aggregate data differ from each other. The figure shows that the import price index of SITC level 3 (energy sector) is highly volatile. The data shows a number of relatively large upward and downward fluctuations. A comparison of the index levels
from the aggregate import price index and the import price index of SITC level 3 (energy sector), reveals that the disaggregated data series move to considerably higher index levels than the aggregate import price index.
Figure 3 and 4 show exchange rate data of the British pound to the UK’s EU trading partner currencies. In general, most currencies tend to move relatively closely to the effective exchange rate index. Furthermore, the effective exchange rate index is very highly related to the pound to euro exchange rate. This is as expected, because the pound to euro exchange rate forms a large component of the effective exchange rate index. Compared to 1998 Q1, the pound has depreciated with more than ten per cent to the import-weighted average of the other EU currencies. Two notable exchange rates are the pound to leu from Romania and the pound to korune from the Czech Republic. Since 1998 Q1, the Romanian currency has clearly depreciated to the British pound with approximately 80 per cent. The Czech currency, on the other hand, has clearly appreciated to the British pound with approximately 50 per cent over the same period. These two exchange rates do not impact the effective exchange rate index highly, because the exchange rates form a relatively small part of the effective exchange rate index.
Figure 4: Four Bilateral Exchange Rates of Currencies Relative to the Pound and the Effective Exchange Rate Index, 1998-2015
3.2
Methodology
The purpose of this thesis is to study the effect of the exchange rate of the British pound to other EU currencies on the UK’s import prices. The re-search will be performed on data over the period 1998 to 2015 as specified in Section 3.1. To study EPT, a regression-based analysis will be adopted. Estimations will be performed on country and sector level data. Further-more, pass-through estimates will be broken down into short and long run estimates, and the analysis includes tests in which the presence of asymmet-ric exchange rate pass-through will be examined.
The first part of the analysis consists of estimations of a regression model based on regression (8) and after that, in the second part of the analysis, the same model will be estimated, but then with a different covariate to approx-imate marginal costs.
The first part of the analysis is as follows. First, all variables from the dataset will be transformed using logarithms. This will remove the expo-nential trends from the time series data. Subsequently, all variables will be tested on the presence of stochastic trends. Stochastic trends are random trends that vary over time. The problem with this, is that t-statistics can become non-normally distributed. A non-normal distribution means that under the assumption of a standard normal distribution, confidence intervals are not valid and hypothesis tests cannot be conducted (Stock and Wat-son, 2012). For this reaWat-son, variables with stochastic trends will have to be transformed so that they do not have a trend any more. In order to test for stochastic trends, the Augmented Dickey Fuller (ADF) test will be ap-plied. To determine the optimal lag length for each variable in the ADF test, the Akaike Information Criterion (AIC) will be used.13 If test results show
that stochastic trends are present, the variables will be transformed into first differences. Still, even variables in first differences could contain stochastic trends. Hence, also differenced variables will be tested on stochastic trends up to the point that the variables do not contain stochastic trends any more. The hypothesis of asymmetric pass-through will be tested by including two dummy variables in the estimation of the model: the first dummy variable equals one when the pound depreciates (i.e., the effective exchange rate index increases) and equals zero otherwise; the second dummy variable equals one when the pound appreciates (i.e., the effective exchange rate index decreases) and equals zero otherwise. Furthermore, as proposed by Campa and
Gold-13There exists a probability that the AIC overestimates the correct lag length.
Nonethe-less, Stock and Watson (2012, p. 594) recommend the AIC over other information criteria, such as the Bayesian Information Criterion, to estimate the lag length of variables for the ADF regression.
berg (2005), four lags of the exchange rate variable will be included. They implicitly assume that it takes up to four quarters for the majority of the firms to adjust their import prices in response to an exchange rate movement. The model will then be estimated with OLS and, to account for possible het-eroskedasticity and/or autocorrelation, Newey-West standard errors will be applied on aggregate and disaggregated data over the whole sample period. If the results do not show significant evidence of asymmetric EPT, a second regression will be estimated without dummy variables. The final regressions with and without dummies are as follows
4qp t= α + 4 X j=0 βj1Dj14qe t−j+ 4 X j=0 βj2D2j 4qe t−j+ µ 4qhicpt +φ 4qyt+ γ 4qppit+ t, (11) and 4qp t = α + 4 X j=0 βj 4qet−j+ µ 4qhicpt+ φ 4qyt+ γ 4qppit+ t, (12)
where q stands for the number of differences to take for all variables, p is the aggregate or disaggregated import price index at quarterly period t, j stands for the time lag, e is the effective exchange rate index, D1is a dummy variable that equals 1 when the effective exchange rate increases and equals 0 other-wise and D2 is a dummy variable that equals 1 when the effective exchange
rate decreases and equals 0 otherwise, hicp is the aggregate or disaggregated consumer price index, y is real GDP, ppi is the aggregate or disaggregated PPI, and is the error term. All variables are expressed in lowercase letters which denotes that logarithms are taken for these variables.
As mentioned, regressions will be performed on aggregate and disaggre-gated data. For the exchange rate and output variable, only aggregate data will be used. For the import price index, consumer price index and producer price index variable, aggregate or disaggregated data will be utilised. Re-gressions on aggregate and disaggregated level data will be estimated, which will include the following 5 sets of variables
Set A = { pagg, e, hicpall, y, ppiindust } ⇒ country level,
Set B = { pS0+1, e, hicpf ood, y, ppinondur } ⇒ food sector,
Set C = { pS2+4, e, hicpexcl, y, ppiinterm } ⇒ raw materials sector,
Set D = { pS3, e, hicpenergy, y, ppienergy } ⇒ energy sector,
In all sets of variables, e and y stand for the effective exchange rate index and real GDP, respectively. In Set A, pagg is the aggregate import price index,
hicpall is the consumer price index that includes all items and ppiindust is the
producer price index of the industry. In Set B, pS0+1 is the import price index of SITC levels 0+1 (food sector), hicpf ood is the consumer price index
that includes items from the food sector and ppinondur is the producer price
index of the non-durables sector. In Set C, pS2+4 is the import price index of SITC levels 2+4 (raw materials sector), hicpexcl is the consumer price index
that includes all items except the food and energy sector and ppiinterm is the
producer price index of the intermediate goods sector. In Set D, pS3 is the import price index of SITC level 3 (energy sector), hicpenergy is the consumer
price index of the energy sector and ppienergy is the producer price index of
the energy sector. In Set E, pS5−8is the import price index of SITC levels 5 to 8 (manufacturing sector), hicpexcl is the consumer price index that includes
all items except the food and energy sector and ppimanu is the producer price
index of the manufacturing sector.
The most important estimated coefficients will be the ones on the ex-change rate variable. These coefficients show to what extent prices are ad-justed after an exchange rate movement. The coefficients will be tested on whether they significantly differ from zero (LCP) and one (PCP). Asymmet-ric pass-through will be determined by statistically testing the restAsymmet-riction
β1 j = βj2
on the EPT coefficients of regression (11), where j is time lag 0, 1, 2, 3 or 4. If the coefficients significantly differ from each other, then pass-through is asymmetric, otherwise not. Furthermore, short run and long run pass-through will be separated. Short run pass-pass-through is equal to the coefficient on the contemporaneous exchange variable. Long run pass-through is equal to the sum of the coefficients on the contemporaneous exchange rate variable and all lags of the exchange rate variable.
In the second part of the analysis, a robustness check is performed on the country level estimation. Set A will be adjusted to include a different covariate for the marginal cost variable and the resulting estimation will be compared to the estimation results from Set A. The estimated sets will in-clude the following variables
Set F = { pagg, e, labourcost, y, ppiindust } ⇒ country level,
Set A = { pagg, e, hicpall, y, ppiindust } ⇒ country level,
respec-tively. The aggregate import price index is denoted by pagg, the consumer price index that includes all items is denoted by hicpall and the producer
price index of the industry is denoted by ppiindust. The variable labourcost is
an aggregate index that measures the unit labour costs of the EU exporting countries. This variable is expressed in logarithms. In the next section, the results of the empirical study are provided.
3.3
Empirical Findings
In this section, the empirical findings will be presented. The results of the regression-based analysis will be shown in Section 3.3.1. The EPT estima-tion results will be discussed in Secestima-tion 3.3.2 and a robustness test will be presented in Section 3.3.3.
3.3.1 Regression-Based Analysis
In Section 3.1.2, the import price and exchange rate variable showed signs of non-stationarity. Therefore, these variables and also the other variables from the dataset are inspected for stochastic trends by performing ADF tests. The results of the ADF tests and the number of lags employed in these tests are reported in Table 1, 2 and 3. The test results reported in Table 1 show that most variables are at least integrated of order one. These variables are then transformed into first differences in order to evaluate whether they are integrated of order two. In Table 2, the first differenced variables are tested for stochastic trends. The ADF tests reject non-stationarity at least at the ten per cent level for all variables, except for the PPI intermediate goods sector, the PPI non-durable goods sector, the PPI energy sector and the HICP all items. In Table 3, the non-stationary variables from Table 2 are evaluated on whether they are integrated of order three. All ADF tests reject the null hypothesis of non-stationarity. Overall, the ADF test results indicate that the HICP all items excluding the food and energy sector is integrated of order zero; the import price index all sectors, the import price index SITC level 0+1, the import price index SITC level 2+4, the import price index SITC level 3, the import price index SITC level 5+6+7+8, the real gross domestic product, the labour cost index, the effective exchange rate index, the PPI industry, the PPI manufacturing sector, the HICP food sector and the HICP energy sector are integrated of order one; the PPI intermediate goods sector, the PPI non-durable goods sector, the PPI energy sector and
the HICP all items are integrated of order two.14 The results are not as expected as an order of integration of one was expected for all variables. An order of integration of zero is not expected for HICP, because prices tend to increase in the medium to long run (with increases in the money supply). An order of integration of two for the PPI and HICP could, for instance, mean that both variables have an increasing mean of inflation over time. Intuitively, this seems unlikely to be the case for the countries under study, because most of these countries have a central bank that adopts an inflation targeting policy. Also taking account of the caveat that the ADF tests have weak power, the variables in the following regressions will be estimated in first differences; so in regression (11) and (12) q equals one.
Next, regression (11) is estimated over different sets of variables based on aggregate and disaggregated data, as stated in Section 3.2. The complete regression results are stated in Table 4, the tests for asymmetric EPT are in Table 5 and the most important features of these two tables are summarised in Table 6.
Table 4 shows that, in general, the coefficient on the contemporaneous and first lag of the exchange rate variable are larger than the coefficients on the other lags of the exchange rate variable. Noteworthy are the relatively large coefficients on the second and third lag of the exchange rate variable in the regression of Set D (energy sector). The effect of hicp is significant and positive in the regression of Set A (country level) and Set D (energy sec-tor) at the one per cent level. This implies that import prices increase when marginal costs increase, as expected. The effect of a one per cent increase in the hicp variable in the regression of Set C (raw materials sector) and Set D (energy sector) is more than proportional with 1,707 per cent and 3,389 per cent, respectively. In the regression of Set E (manufacturing sector), the coefficient on the hicp variable is negative. This indicates that an increase in hicp (marginal cost) with one per cent implies a decrease of 0,478 per cent in the export price of a manufacturing producer. This effect is difficult to comprehend, but the effect is inaccurately estimated, because the effect is not statistically significant. The output variable is not significantly different from zero at any level in all the regressions. Based on this result, it seems that output has no significant effect on import prices. This means that an increase in import demand does not lead to a change in import prices. Campa and Goldberg (2005) and Mumtaz et al. (2011) found the same result for im-port demand in their analyses. The outcome could indicate that firms with
14 In Table A.4, A.5 and A.6 in the Appendix, ADF tests are performed in which an
intercept and a linear time trend are included in the ADF regression. Overall, the results indicate similar orders of integration for the variables as in the ADF tests in which only an intercept is included in the ADF regression.
Table 1: ADF tests for Non-Stationarity (Intercept Only)
Variable Lag length
according to AIC
ADF Test Statistic Import Price Index All Sectors 2 -1,168 Import Price Index SITC Level 0+1 5 -1,157 Import Price Index SITC Level 2+4 8 -1,548 Import Price Index SITC Level 3 1 -2,036 Import Price Index SITC Level 5+6+7+8 2 -1,082 Real Gross Domestic Product (UK) 5 -1,845 Labour Cost Index (27 Countries of EU) 8 -1,509 Effective Exchange Rate Index 1 -1,375
PPI Industry (UK) 3 -1,055
PPI Intermediate Goods Sector (UK) 8 -1,310 PPI Non-Durable Goods Sector (UK) 2 -0,724 PPI Manufacturing Sector (UK) 4 -1,031 PPI Energy Sector (UK) 7 -0,660 HICP All Items (22 Countries of EU) 7 -2,351 HICP Food Sector (22 Countries of EU) 7 -1,458 HICP Energy Sector (22 Countries of EU) 1 -1,613 HICP All Items Excluding the Food
and Energy Sector (22 Countries of EU) 8 -4,740***
Note: All variables are expressed in logarithms. For the ADF test, critical values from a non-normal distribution are used. ***,** and * denote the significance level at one, five and ten per cent, respectively. These correspond to the critical values -3,43, -2,86, -2,57,
respectively. The t-test is a one sided test on the respective coefficient (it is a test for a coefficient smaller than zero). H0: non-stationary, H1: stationary. All numbers are
rounded to three decimals.
increasing economies of scale and decreasing economies of scale, on average, have an equal impact on import prices. An other interpretation could be that firms do not alter their production when UK consumers’ demand for imports rises and hence no economies of scale can arise for these firms. The producer price indices are significant at the one per cent level for all regressions, except in the regression of Set B (raw materials). This means that producer price indices have a strong amount of explanatory power in the EPT regressions. The effect is largest in regression of Set C (raw materials sector), with an almost one-to-one effect of the producer price index on the import price in-dex. This means that exporters increase their price with almost one per cent, if their competitors in the importing country increase their prices with one percent. The constant in all the regressions prove to be small and mostly
Table 2: ADF tests for Non-Stationarity on First Differenced Variables (In-tercept Only)
Variable (in First Differences) Lag length according to AIC
ADF Test Statistic Import Price Index All Sectors 1 -4,449*** Import Price Index SITC Level 0+1 4 -4,177*** Import Price Index SITC Level 2+4 5 -2,864** Import Price Index SITC Level 3 0 -7,195*** Import Price Index SITC Level 5+6+7+8 1 -5,448*** Real Gross Domestic Product (UK) 4 -2,847* Labour Cost Index (27 Countries of EU) 8 -2,722* Effective Exchange Rate Index 0 -8,497*** PPI Industry (UK) 2 -3,698*** PPI Intermediate Goods Sector (UK) 7 -1,820 PPI Non-Durable Goods Sector (UK) 8 -1,450 PPI Manufacturing Sector (UK) 3 -3,522*** PPI Energy Sector (UK) 8 -2,068 HICP All Items (22 Countries of EU) 4 -2,526 HICP Food Sector (22 Countries of EU) 6 -3,808*** HICP Energy Sector (22 Countries of EU) 0 -7,838***
See notes Table 1.
Table 3: ADF tests for Non-Stationarity on Second Differenced Variables (Intercept Only)
Variable (in Second Differences) Lag length according to AIC
ADF Test Statistic PPI Intermediate Goods Sector (UK) 6 -5,111*** PPI Non-Durable Goods Sector (UK) 8 -2,779* PPI Energy Sector (UK) 8 -4,459*** HICP All Items (22 Countries of EU) 4 -5,705***
See notes Table 1.
insignificant. This implies that the fixed exogenous component of the mark-up, as indicated in equation (5), is not significantly different from zero. This could indicate that most industries are characterised by perfect competition, but the type of market structure is also dependent on a firm’s freedom to absorb exchange rate movements into its profit margin. Lastly, the adjusted R2s show that between 21,9 and 63,8 per cent of the variance in the import price indices is explained by the independent variables in the regressions.
