5HYLHZRISUHYLRXVZRUNRQSULFHVHWWLQJEHKDYLRXUDQGLQWHJUDWHGPDUNHWV
The law of one price (LOP) is a good place to start.13 If the LOP held for all traded goods and preferences were identical across countries, then absolute14 purchasing power parity (PPP) would hold. In practice, transport and distribution costs drive a wedge between domestic and foreign prices.
But if these were constant, then relative15 PPP holds.16
There is a mixed literature on empirical findings for the LOP and PPP. Research on prices and exchange rate movements has shown that relative prices of goods are systematically related to the exchange rate. For example, following Engel’s (1999) approach, Obstfeld and Rogoff (2000) computed the correlation between the exchange rate and the relative prices of tradables and non-tradables over time in the United States, Germany, France and Japan and showed that the relative prices of tradables exhibited very little mean-reversion, i.e. they could be non-stationary. However, a number of empirical problems emerged because of aggregation issues and the failure to compare like-with-like when comparing prices across countries. Researchers have therefore focussed on industry data. These provided evidence that pass-through was more complete and somewhat faster. In our work using aggregated data reported below, we use sector-specific price indices for producer prices, calculated in domestic price terms.17
0DUNHWVHJPHQWDWLRQDQGSULFLQJWRPDUNHW
The repeated empirical failures of PPP encouraged the development of theories based on strategic interaction and market segmentation. The argument in brief is that firms with monopoly power, selling differentiated products, have an incentive to charge different prices in markets where preferences differ. In a given market, their monopoly power (which depends on the elasticity of demand they face for their products18) and thus their pricing power should be gauged by the price they charge relative to their competitors’ price. Changes in environmental taxes affect these relative prices and therefore the monopoly power and thus firm’s pricing decisions: as a result the ETR pass-through may be full where the firms in the sector are price setters.
In line with the finding by Bergin and Feenstra (2001) for general equilibrium models, even with segmented markets, if preferences are identical and demand elasticities unchanged following a shock to the exchange rate, producers have no incentive to set different prices in different markets. Firms will H[DQWH set prices as a constant mark-up over marginal costs and hence PPP will hold. In the event of a shock, even if prices are sticky, relative prices will change eventually. Deviations from the long-run relationship are likely to be transitory. The desired mark-up is constant and firms will make price adjustments once they have a chance to do so. In this vein, we impose PPP for the long-run structural relationship between exchange rates and foreign prices.
'DWD
There are two basic sources for quarterly data on sectoral output prices, with a sufficient time span.
The OECD compiles one set and the other set is compiled by EUROSTAT. The OECD Statistical Compendium 2004-2 “Indicators of Activities for Industry and Services ISIC Rev.3” (ceased end
13 If agents are profit and utility maximisers and transportation, resale and distribution are costless then, due to arbitrage, identical goods command the same price in common currency terms. If firms are also price takers, this is the perfectly competitive paradigm.
14 Absolute PPP is where all (common currency) prices for identical goods are equal.
15 Relative PPP is where (common currency) prices grow at the same rate.
16 In the estimations that follow, we estimate on the basis of price indices, not on absolute prices, and therefore we can only test for relative PPP instead of absolute PPP.
17 Aggregation issues were not the only arguments against writing off market integration. In open economies, changes in exchange rates impinge on producers’ costs, but cost measures may often fail to pick up exchange rate effects.
18 In a standard Dixit-Stiglitz imperfect competition model, a firm’s mark-up is inversely related to the elasticity of demand it faces. The greater the elasticity of demand, say because of greater availability of substitutes, the lower the monopoly power of a firm and hence the lower is the margin.
2001) was used to extract producer prices (1995=100) for the countries of interest – Denmark, Germany, Finland, Netherlands, Sweden, UK and on the US price as the ‘world price’.19 These were available as a domestic price index constructed in national currency. Corresponding domestic producer price indices at the sectoral level (NACE code) were available from EUROSTAT from 1990 (reference IO7qprin) onwards. The OECD series was used after updating with the appropriate rate of change in the price from the corresponding price series up to quarter 4, 2004.
The domestic manufacturing wage for the entire period is available from the OECD and is calculated as a quarterly index (2000=100) of hourly earnings in all manufacturing for each country. Sectoral specific wage rates were not available.
The exchange rates used were obtained from EUROSTAT (Ameco) and are represented as a quarterly average where one DM, US dollar or SEK is equal to so many units of domestic currency.
Post Euro values were converted back to domestic currencies existing prior to the introduction of the Euro in order to achieve a consistent exchange rate time series.
The following time series are analysed for the 30-year period Q1 1975 to Q4 2004:
XXCHEMPR : domestic producer price for Chemicals (1990=1) XXBASMETPR : domestic producer price for Basic Metals (1990=1)
XXFBTPR : domestic producer price for Food, Beverages and Tobacco (1990=1) XXNMETPR : domestic producer price for Non-metallic Mineral Products (1990=1) XXPAPPR : domestic producer price for Paper and Paper Products (1990=1)
XXWOODPR : domestic producer price for Wood and Wood Products (1990=1) XXUSD : 1 US dollar = units of domestic currency
XXDE : 1 German Deutschmark = units of domestic currency XXSW : 1 Swedish Kroner = units of domestic currency Where XX = DE, DK, FI, IE, NL, SW, UK, US
DEMANW : All manufacturing manual wage index for Germany DKMANW : All manufacturing manual wage index for Denmark FIMANW : All manufacturing manual wage index for Finland IEMANW : All manufacturing manual wage index for Ireland20 NLMANW : All manufacturing manual wage index for the Netherlands SWMANW : All manufacturing manual wage index for Sweden
UKMANW : All manufacturing manual wage index for the United Kingdom All of the above wage rates are in index form, 2000 = 1.000
The data employed in this study are graphically displayed in Appendix Charts 2.5.1. For each sector there is a graph of a logarithmic transformation of the time series data and a graph of the first differences of the logarithmic transformations.
The prefix ‘L’ stands for the natural logarithm of the time series and ‘D’ denotes differencing of the relevant time series. All econometric estimations in this section have been carried out using Eviews 5.0.
19 Slovenian price data were not available from the OECD.
20 Ireland was included in the dataset reported here.
(FRQRPHWULFLVVXHV
The first step is to test for a unit root. In essence this is to show whether the data conform to the requirements for the relationship not to be spurious. Appendix Table 2.5.1 summarises the results for the price variables of unit root tests on levels and in first differences of the data. They are tested for unit roots and their order of integration using an ADF test. Strong evidence emerges that the series are generally I(1). The relationship we seek is a cointegrating one.
It is well known that systems in which non-stationary variables are cointegrated can be described by error correction mechanisms (e.g. Granger, 1986; Engle and Granger, 1987). The Granger representation theorem states that if a set of variables are cointegrated, then there exists a valid error-correction representation of the data. Thus, if <W and ;W are both I(1) and have a cointegrating vector(1,−β)’, there exists an error-correction representation
∑
∆ +∑
∆ ++
− +
=
∆<W α1 λ
(
<W−1 β1;W−1)
α2(
L)
\W−L α3(
L)
;W−L ε\Wwhere β = the parameters of the cointegrating vector and ε\W= white-noise disturbance with no moving average part and αLare all parameters.
If <Wand ;Ware both I(1) and have a long-run relationship, there must be some force that pulls the variables back in line. The error correction model does exactly this: it describes how the variables behave in the short-run consistent with a long-run relationship. The systematic dynamics are kept as simple as possible in our estimation with just one lag.
Further if ;W does not adjust to the equilibrium error (has a zero adjustment parameter), it is weakly exogenous for β (as defined by Engle, Hendry and Richard 1983). This means that we can include
;W
∆ in the estimated equation without affecting the error correction term −λ
(
<W−1−β;W−1)
and therefore the long run relationship.21 Lambda is the speed of adjustment parameter where a higher value indicates a faster convergence from short-run dynamics to the long-run situation. We include both the domestic cost and foreign price effect within the ECM term as members of the ; vector.Estimation of the nonlinear parameterisation of the ECM results in super-consistent estimators of the long-run coefficients.22 We do not consider dynamic homogeneity i.e. that the long-run domestic price level is affected by the growth rate of prices and costs. We employ a specification that includes one lag in the dynamic terms, where the long-run margin is driven by the levels of domestic costs variable and the competitor price level (PPP imposed23). The econometric tests are shown in Appendix Table 2.5.1.
21 That is, we can condition upon ;W in the error correction model for <W.
22 In investigating the price to market hypothesis, the traditional approach of first differencing disregards potentially important equilibrium relationships amongst the levels of the series to which the hypotheses of economic theory usually apply (Engle and Granger 1987). In this single equation ECM, this information is maintained.
23 Multi-cointegration applies here where the combination of foreign price and exchange rate is cointegrated with domestic wage costs.
$SSHQGL[&KDUWV&RLQWHJUDWLRQDQDO\VLVIRUSULFHYDULDEOHV
Prices are expressed in levels (first graph) and as first differences in the second graph, logged.
The charts’ legends are as follows:
D = First differences L = logged
DE = Germany DK = Denmark FI = Finland IE = Ireland NL = Netherlands SW = Sweden
UK = United Kingdom PR = price
CHEM = Chemicals BASMET = Basic metals
FBT = Food, beverages and tobacco NMET = Non-metallic mineral products PAP = Paper and paper products
WOOD = Wood and wood products
Ireland is included in the data-set, though not being an ETR country it is not included in the analysis.
&KHPLFDOV
-1.5 -1.0 -0.5 0.0 0.5
1975 1980 1985 1990 1995 2000 LD EC HEMP R
LD KC HEMP R LFIC HEMP R
LIEC HEMPR LNLC HEMPR LSWC HEMPR
LUKC HEMPR LUSC HEMPR
-.15 -.10 -.05 .00 .05 .10 .15
1975 1980 1985 1990 1995 2000 DLDECHEMPR
DLDKCHEMPR DLFICHEMPR
DLIECHEMPR DLNLCHEMPR DLSWCHEMPR
DLUKCHEMPR DLUSCHEMPR
%DVLF0HWDOV
-1.2 -0.8 -0.4 0.0 0.4 0.8
1975 1980 1985 1990 1995 2000 LDEBASMETPR
LDKBASMETPR LFIBASMETPR
LIEBASMETPR LNLBASMETPR LSWBASMETPR
LUKBASMETP R
-.12 -.08 -.04 .00 .04 .08 .12
1975 1980 1985 1990 1995 2000 DLDEBASMETPR
DLDKBASMETPR DLFIBASMETPR
DLIEBASMETPR DLNLBASMETPR DLSWBASMETPR
DLUKBASMETPR DLUSBASMETPR
)RRG%HYHUDJHVDQG7REDFFR
-1.6 -1.2 -0.8 -0.4 0.0 0.4
1975 1980 1985 1990 1995 2000 LDEFBTPR
LDKFBTPR LFIFBTPR
LIEFBTPR LNLFBTPR LSWFBTPR
LUKFBTPR LUSFBTPR
-.08 -.04 .00 .04 .08 .12
1975 1980 1985 1990 1995 2000 DLDEFBTPR
DLDKFBTPR DLFIFBTPR
DLIEFBTPR DLNLFBTPR DLSWFBTPR
DLUKFBTPR DLUSFBTPR
1RQPHWDOOLFPLQHUDOSURGXFWV
-1.6 -1.2 -0.8 -0.4 0.0 0.4 0.8
1975 1980 1985 1990 1995 2000 LDENMETPR
LDKNMETPR LIENMETPR
LNLNMETPR LS WNMETPR LUKNMETPR
LUSNMETPR
-.08 -.04 .00 .04 .08 .12
1975 1980 1985 1990 1995 2000 DLDENMETPR
DLDKNMETPR DLFINMETPR
DLIENMETPR DLNLNMETPR DLSWNMETPR
DLUKNMETPR DLUSNMETPR
3DSHUDQG3DSHU3URGXFWV
-1.5 -1.0 -0.5 0.0 0.5
1975 1980 1985 1990 1995 2000 LDEPAPPR
LDKPAPPR LFIPAPPR
LIEPAPPR LNLPAPPR LSWPAPPR
LUKPAPPR LUSPAPPR
-.08 -.04 .00 .04 .08 .12
1975 1980 1985 1990 1995 2000 DLDEPAPPR
DLDKPAPPR DLFIPAPPR
DLIEPAPPR DLNLPAPPR DLSWPAPPR
DLUKPAPPR DLUSPAPPR
:RRGDQG:RRG3URGXFWV
-1.6 -1.2 -0.8 -0.4 0.0 0.4
1975 1980 1985 1990 1995 2000 LDEWOODPR
LDKWOODPR LFIWOODPR
LIEW OODPR LNLWOODPR LSWWOODPR
LUKWOODPR LUSWOODPR
-.08 -.04 .00 .04 .08 .12
1975 1980 1985 1990 1995 2000 DLIEWOODPR
DLDEWOODPR DLDKWOODPR
DLFIWOODPR DLNLWOODPR DLSWWOODPR
DLUKWOODPR DLUSWOODPR
COMETR (FP6 prop. 501993) 19/9/03 (FRQRPHWULF7HVWV
A Johansen test strongly rejects the null of no cointegration of the dependent and key independent variables at the 5% level, that is, they tested satisfactorily. Given the result that a unique cointegrating relationship exists, a single equation ECM offers a robust alternative to the Johansen method. Validity is conditional on the regressors being weakly exogenous1, but we show that this condition is satisfied. Thus the estimation can proceed in the single-equation framework outlined above. Single equation estimates should be reliable and a well-determined t statistic on the ECM term is further evidence of cointegration. Additional tests for unit roots were undertaken for the price variables, interest focusing on foreign prices that the regressions tested for influence on the domestic price. Results are shown in Appendix Table 2.5.1.
1 In a cointegrated system, if a variable does not respond to the discrepancy from the long-run equilibrium, it is weakly exogenous, that is, the speed of adjustment parameter is 0.
COMETR (FP6 prop. 501993) 19/9/03
$SSHQGL[7DEOH8QLW5RRW7HVWV
Level First Differences ADF test statistic ADF test statistic
LDEBASMETPR -3.315497 -5.986174***
LDKBASMETPR -2.689924* -8.359629***
LFIBASMETPR -1.568949 -6.104499***
LIEBASMETPR -0.343549 -5.547648***
LNLBASMETPR -2.598586* -6.701238***
LSWBASMETPR -1.871067 -5.493550***
LUKBASMETPR -3.237086** -4.188560***
LUSBASMETPR -3.189334** -3.401767***
LDECHEMPR -1.330966 -4.915669***
LDKCHEMPR -2.962546** -7.000909***
LFICHEMPR -2.212823 -6.790495***
LIECHEMPR -3.452354*** -6.803949***
LNLCHEMPR -3.159378** -5.440110***
LSWCHEMPR -1.996497 -6.833328***
LUKCHEMPR -3.897648*** -7.34932***
LUSCHEMPR -0.439414 -5.204615***
LDEFBTPR -1.855558 -5.190388***
LDKFBTPR -4.054082*** -7.484247***
LFIFBTPR -4.326288*** -3.774157***
LIEFBTPR -4.104027*** -2.838678**
LNLFBTPR -1.469798 -4.694399***
LSWFBTPR -3.247706** -2.765823*
LUKFBTPR -4.929782*** -2.587562*
LUSFBTPR -2.012392 -4.839287***
LDENMETPR -1.954511 -2.406641
LDKNMETPR -3.013959** -1.791162
LFINMETPR -6.842790*** -4.144532***
LIENMETPR -6.652777*** -2.941475**
LNLNMETPR -2.355987 -2.109424
LSWNMETPR -3.669447*** -2.504838*
LUKNMETPR -5.059891*** -2.759671**
LUSNMETPR -1.992435 -1.908696
LDEPAPPR -2.093627 -5.394902***
LDKPAPPR -2.953416** -3.220020***
LFIPAPPR -2.021902 -5.405187***
LIEPAPPR -3.470909** -5.316844***
LNLPAPPR -1.118458 -4.005505***
LSWPAPPR -1.641998 -4.828684***
LUKPAPPR -4.850024*** -4.230459***
LUSPAPPR -2.462221 -4.873529***
LDEWOODPR -3.409769** -2.037768
LDKWOODPR -3.083240***
LFIWOODPR -3.095870** -5.883673***
LIEWOODPR -4.413811*** -5.580623***
LNLWOODPR -2.435048* -3.394687**
LSWWOODPR -3.588209*** -4.895384***
LUSWOODPR -2.135125 -3.561840***
LUKWOODPR -4.660079*** -3.139922**
Critical values: 1% level = -3.48655; 5% level = -2.886074; 10% level = -2.579931.
Note: ADF is the Augmented Dickey-Fuller Test for unit roots. The null hypothesis that the series is not stationery is rejected if the test statistic exceeds the critical value in absolute terms. The lag length is based on the Schwarz Information Criterion.
$SSHQGL[ 5DQNLQJE\XQLWHQHUJ\FRVWVDQGPDUNHWSRZHU
This appendix sets out the ranking by decreasing vulnerability. Chart 2.7.1 showed this for ETR countries combined, and here it is shown for individual ETR countries by looking at HQHUJ\H[SHQGLWXUHVKDUHVRI YDOXHDGGHG alongside PDUNHWSRZHUas measured by foreign price influence.
COMETR (FP6 prop. 501993) 19/9/03
The model results on market power were used for ranking vulnerability, though this should be viewed as approximate because it is not an exact method. The ranking method was based mainly on the significance and size of the coefficient on the variable Foreign Price in Table 2.5.1(b), with consideration being given to Table (a). After variables with significant foreign price coefficients were exhausted, those with significant domestic costs were used to inform the ranking.
'(10$5. (the most vulnerable sectors are at the top of each column) ([SHQGLWXUHVKDUHV )RUHLJQSULFHLQIOXHQFH
Basic metals Food beverages and tobacco
Non-metallic mineral products Basic metals Food beverages and tobacco Paper
Wood + Paper Non-metallic minerals
1RWHWood and wood products and Chemicals are not well-modelled and are therefore omitted. In the expenditure shares Wood is included with Paper.
:(67*(50$1<
([SHQGLWXUHVKDUHV )RUHLJQSULFHLQIOXHQFH
Basic chemicals Paper
Pharmaceuticals Chemicals
Wood + Paper Wood
1RWH: The shaded sectors are those where the issue of different classification between the two columns arises. Where the sectors are contiguous consistency can be maintained by amalgamation.
),1/$1'
([SHQGLWXUHVKDUHV )RUHLJQSULFHLQIOXHQFH
Basic metals Basic metals
Wood + Paper Non-metallic mineral products
Non-metallic mineral products Wood 1(7+(5/$1'6
([SHQGLWXUHVKDUHV )RUHLJQSULFHLQIOXHQFH
Basic chemicals Basic metals
Basic metals Chemicals
Pharmaceuticals Non-metallic mineral products
Non-metallic minerals Wood
Wood + Paper Food beverages and tobacco
Food beverages and tobacco 6:('(1
([SHQGLWXUHVKDUHV )RUHLJQSULFHLQIOXHQFH
Basic metals Paper
Basic chemicals Basic metals
Non-metallic mineral products Chemicals
Pharmaceuticals Wood
Wood + Paper Non-metallic mineral products
8.([SHQGLWXUHVKDUHV )RUHLJQSULFHLQIOXHQFH
Basic chemicals 1 Basic metals 1
Non-metallic mineral products 2 Paper 2
Basic metals 3 Wood
Wood + Paper 4 Chemicals 3
Pharmaceuticals Food beverages and tobacco 4
Food beverages and tobacco 5 Non-metallic mineral products 5
COMETR (FP6 prop. 501993) 19/9/03 (8(75
([SHQGLWXUHVKDUHV )RUHLJQSULFHLQIOXHQFH
Basic chemicals 1 Basic metals 1
Pharmaceuticals Paper 2
Non-metallic mineral products 2 Wood
Basic metals 3 Chemicals 3
Wood + Paper 4 Food beverages and tobacco 4
Food beverages and tobacco 5 Non-metallic mineral products 5
It is the ranking from this last table that is used for the vulnerability charts in the main text, in the concluding section.