This paper develops a methodology to estimate the real effective exchange rate (REER) that incorporates two distinctive elements not accounted for in the current literature: (1) product heterogeneity when determining international competitors and their weights, which allows us to identify countries’ direct international competitors more accurately, and (2) a comprehensive treatment of services exports, which allows us to provide a complete view of international competitiveness encompassing the entire export sector.
We apply this methodology to reexamine the evolution of the REER of the Mediterranean quartet (MQ) of Greece, Italy, Portugal, and Spain, and particularly, the evolution of their REER gap with the other euro area members. This case motivates our analysis as the common pattern of real appreciation observed in the MQ countries has created concern in policy and academic circles (European Commission, 2006; Bini-Smaghi, 2007; Roubini, 2007; and Papademos, 2007). Particular attention has been given to the fact that this pattern diverges from the average real depreciation observed in the rest of the euro area (see Figure 1). It is argued that this real appreciation is associated with a loss of international competitiveness in the MQ and that it could lead to a persistent period of slow growth, which has already materialized in the cases of Portugal and Italy (Blanchard, 2006a and 2006b).1
Figure 1.Real Appreciation in the Mediterranean Quartet vs. the Rest of the Euro Area
Note: Unit-labor-cost-based real effective exchange rate index (REER) defined à la IMF (higher means more appreciated); the base year is 1998. EA-8 refers to Austria, Belgium, France, Finland, Germany, Ireland, Luxembourg, and Netherlands. The REER for the EA-8 is estimated aggregating each country’s REER and weighting by their total exports of goods. All countries adopted the euro on January 1, 1999, except for Greece, which adopted the euro on January 1, 2001.
In short, the REER is an aggregated measure of cost competitiveness between countries. It tracks the evolution of cost competitiveness of a particular country with respect to a weighted average of all other countries in the world.2 The methodologies available to calculate the REER have been constantly improving in recent decades as they have been incorporating more realistic assumptions. Table 1 summarizes the existing literature and highlights the approach taken in each study to address the key elements of the REER analysis, that is, the approach used to calculate the importance or weight of each other country and the price used to measure cost competitiveness. Bayoumi, Jayanthi, and Lee (2005, 2006) is the most comprehensive methodology currently available, which includes the latest development in the literature.
|Reference||Importance of Other Countries (Definition of Markets)|
|Partners/competitors||Product and market|
|Services||Local consumption of|
|Bank for International Settlements (Fung and Klau, 2006)||Competitors||Representative-product approach. All manufacturing goods (SITC Rev. 3 5–8) are treated as one identical good; nonmanufacturing goods are not considered. Markets are defined at the country level.||Services are not included in the analysis.||Total manufacturing output.||Aggregate prices. Consumer price index (CPI) and/or manufacturing unit labor cost are used to measure relative cost competitiveness.|
|Bank of Japan (2007)||Partners||Total exports of goods.||Services are not included in the analysis.||Does not apply.||Aggregate prices. CPI and/or manufacturing unit labor cost are used to measure relative cost competitiveness.|
|European Central Bank (Buldorini, Makrydakis, and Thimann, 2002)||Competitors||Representative-product approach. All manufacturing goods (SITC Rev. 3 5–8) are treated as one identical good; nonmanufacturing goods are not considered. Markets are defined at the country level.||Services are not included in the analysis.||Manufacturing output for domestic use.||Aggregate price. CPI, manufacturing unit labor cost, PPI and/or wholesale prices are used to measure relative cost competitiveness.|
|Federal Reserve Board (Loretan, 2005)||Average between competitors and partners||Representative-product approach. All goods are treated as one identical good (except for oil, gold, and military items, which are not considered).||Services are not included in the analysis.||Local production is not considered in the analysis.||Aggregate prices. CPI and/or manufacturing unit labor cost are used to measure relative cost competitiveness.|
|International Monetary Fund (Bayoumi, Jayanthi, and Lee, 2005)||Competitors||Representative-product approach. All manufacturing goods (SITC Rev. 3 5–8, excl. 68) are treated as one identical good. Commodities are disaggregated into 20 categories at 2-dig. SITC Rev.3 level. Markets are defined at the country level for manufactured goods and at the global level for commodities (global goods).||Services are considered in the analysis, but assumed to have the same trade pattern as the observed pattern for manufacturing goods. Tourism is treated separately only for a subset of countries.||Manufacturing output for domestic use.||Aggregate prices. CPI and/or manufacturing unit labor cost are used to measure relative cost competitiveness.|
|Organization for Economic Cooperation and Development (Durand, Simonm, and Webb, 1992; Durand, Madaschi, and Terribile, 1998)||Competitors||Representative-product approach. All manufacturing goods (SITC Rev. 3 5–9) are treated as one identical good. Markets are defined at the country level for individual OECD countries and at the level of country aggregates for six non-OECD country groups.||Services are not included in the analysis.||Total manufacturing output.||Aggregate price. CPI and/or manufacturing unit labor cost are used to measure relative cost competitiveness.|
|Current paper (Bennett and Zarnic)||Competitors||Heterogeneous-product approach. All goods are treated disaggregately at 4-digit ISIC Rev. 3 level. Markets are defined at the country level for all nonglobal goods and at the global level for global goods.||Services are treated disaggregately at 2-digit ISIC Rev. 3 level. Markets are defined at the country level for all services.||Total output for domestic use (at industry-level).||Aggregate and disaggregate prices. CPI, Manufacturing unit labor cost, and sectoral unit labor cost data for goods (1- and 2-digit ISIC Rev. 3) are used to measure relative cost competitiveness.|
Determining the weights by identifying the degree to which countries compete in international markets, as opposed to weighting by trade partnership, is one of the most important characteristics that distinguishes the most up-to-date REER estimations. To illustrate the importance of this feature, consider the extreme case of two countries, A and B, that export mostly to a third country C and have nil bilateral trade between them. If the weight of country B in the calculation of country A’s REER is based on trade partnership, then changes in the exchange rate of country B will not alter the REER of country A. This is not a desirable feature of an index of relative cost competitiveness, because countries A and B compete when exporting to country C and exchange rate movements in either country clearly affect the relative cost competitiveness of the other one. The interrelationship between the cost competitiveness of countries A and of country B is better captured by the REER if the weights are based on a measure of how much these two countries compete in international markets.
With respect to the method used to identify international competitors and their weights, the existing literature considers that two countries are international competitors if they both sell products in the same country, that is, in a market defined as a single aggregated sector comprising a representative product category—which we refer to as the representative-product approach (RPA). As a result, the RPA assumes implicitly that all exporters compete with each other in the destination country. In contrast, we take a more micro-based approach that considers product heterogeneity when defining markets and identifying international competitors and their weights. For each product type that we consider, we identify international competitors as competitors competing in the market for that product type in the destination country. This allows us to analyze relative cost competitiveness at disaggregated markets according to the type of product and destination country. We aggregate these market-level REER indices to obtain a country-level REER—which we refer to as the heterogeneous-product approach (HPA). In principle, our methodology can be applied to alternative definitions of market. Based on data availability, however, we define markets at 4-digit ISIC category of goods and at 2-digit ISIC category of services.3
The HPA identifies more precisely a country’s direct international competitors, and thus, their weights. This feature operates at two levels: first, with respect to other exporters, and second, with respect to local producers at the destination of exports. To illustrate the differences, assume that country A exports textiles to country C, but country B exports cars to country C. The RPA focuses on competitors at an aggregate level—at the manufacturing sector for example, the most common case in the literature—suggesting that countries A and B compete in market C, even though exporters of cars are not necessarily competitors of textile exporters. Furthermore, the RPA would imply that all manufacturing goods produced in country C are competitors of exporters to country C, regardless of the type of good that is produced in, and exported to, country C. In contrast, the disaggregated view of the HPA would not consider countries A and B as competitors in this example and would consider only textile producers in country C as competitors of country A.
With respect to services exports, our approach provides a comprehensive view of relative cost competitiveness by incorporating information about exports of services as well as exports of goods. Services exports have become increasingly important and represent 65 percent of total exports in Greece, 19 percent in Italy, 27 percent in Portugal, and 31 percent in Spain. As with the case of goods, we identify competitors in the destination market at disaggregated categories of services. Unfortunately, the available data on disaggregated bilateral trade in services are not as complete as the data for trade in goods, and therefore, our estimates of the REER in services are restricted to the available sample of trade flows. The average coverage of bilateral trade ranges from 89 percent of total services exports for Greece to 59 percent for Spain. The coverage for goods is above 90 percent for all MQ countries.
Our results suggest a modest reduction in the observed REER gap between the MQ countries and the other members of the euro area. Allowing for product heterogeneity (HPA) and services exports implies, compared with the standard results obtained under the RPA, a lower real appreciation from 1998 to 2006 on the order of 2 to 3 percent for all MQ countries—2 percent for Greece, 2.8 percent for Italy, 2.4 percent for Portugal, and 2.3 percent for Spain. These figures are based on difference-in-difference estimates that control for the results obtained for the rest of the euro area countries using the same methodology. As a robustness check, we also show that our results obtained under the RPA are consistent with the ones reported using the methodology presented in Bayoumi, Jayanthi, and Lee (2005 and 2006), the closest methodology to the one presented in this paper that uses RPA.
The above results are based on a single cost measure at the country-level as used in the current literature, namely the unit labor cost (ULC). To the extent that wages and productivity growth vary across exporting sectors, differentiated cost measures at the sector-level would yield a more accurate picture of international competitiveness. We explore this avenue and compute the REER using differentiated ULC measures at the 2-digit level. Unfortunately, the sample of countries with differentiated ULC series (28) is more restricted than the sample with aggregated ULC series (38), particularly regarding Asian countries. Also, the available time span is one year shorter than in the aggregate ULC data. Moreover, the ULC series at the industry-level may be more volatile because of its disaggregated nature, which could be magnifying some of the known problems of the ULC indicator as a measure of cost competitiveness (see footnote 2). Therefore, the results based on differentiated ULC should be taken with caution, given the data limitations detailed above, and should be read as an exploratory effort to determine the effect of differentiated cost measures on the REER.
The existing literature determines the weights used in the estimation of the REER using the most comprehensive bilateral trade data available. These data report the final value of the exported product rather than the value added embodied in it that is domestically produced—that is, the final value net of imported intermediate inputs. As Hummels, Jun, and Yi (2001) and Yi (2003) have pointed out, the latter is a more desirable measure of the economic importance of exports as it accounts for vertical specialization by netting imported intermediate inputs. Unfortunately, and given the nature of the available data, the estimations presented in this paper are not free of this limitation. Having said that, the HPA addresses part of the increase in the variance of the REER estimator due to the proxy used for exports—imported intermediate inputs overestimate the importance of local cost for all countries and not only for the one under analysis. To the extent that vertical specialization for each type of product is relatively similar across competitors within 4-digit level, our methodology is computing a REER that accounts for the difference in the value added of exported products across industries when selecting the relevant set of competitors. The RPA does not account for this factor given its implicit assumption that all exports products compete with each other in the destination country.
The micro-based methodology proposed in this study also allows for a quantitative assessment of each country’s profile of competitors. Such evidence provides information about the exposure of each country to its key competitors around the world; for example, the exposure to emerging competitors like China, which has shown a strong pattern of productivity and trade growth, or the exposure to countries facing significant changes in their cost structure, such as the wage moderation observed recently in Germany and the depreciation of the nominal exchange rate observed in the United States during the recent years. Our findings indicate that the bulk of competition for the MQ still comes from the advanced economies, especially from the euro area—Spain and Portugal are more exposed to euro area competition, and therefore, less exposed to changes in the value of the euro. Nonetheless, there has been a change in the goods sector, as emerging economies have grown in importance since the late 1990s, particularly China in both high- and low-technology sectors.
This section presents the methodology used to estimate the evolution of the REER under the HPA. The first subsection develops a generic framework to aggregate at the country-level the relative cost competitiveness dynamics observed at the country-industry-level. This framework allows us to incorporate into the analysis elements such as global goods—goods whose markets are defined at the world-level rather than at the country-level—and local consumption of local production—the competition that local producers represent at exports’ destination.
A Generic Approach to Aggregate Relative Cost Competitiveness
We construct our index of relative cost competitiveness between country i and the countries in the rest of the world (ROW), denoted by Ri,t, using a geometrical Laspeyres index and the chain link methodology. The subscripts g, d, and t refer to the type of product (including both goods and services), destination (country), and time (year), respectively.
The evolution of the Ri,t index is defined in equation (1), where T denotes the base year, and ΔΩi,t denotes the (natural logarithmic) change of the relative cost competitiveness between country i and the ROW between period t and t–1.4
The change in relative cost competitiveness at the country-level, ΔΩi,t, is constructed as the weighted average of the change in the relative cost competitiveness between country i and the ROW in each market defined by the g,d pair (equation (2)). The change in the relative cost competitiveness in market g,d is denoted by Δθi,g,d,t. The weights are given by the importance of each market g,d in country i’s exports and are denoted by βi,g,d,t−1. The term βi,g,d,t−1 is computed as the share of country i’s total exports represented by its exports of product g to destination d (equation (3)).
where Si,g,d,t represents sales by country i of product g in destination d; that is, sales by country i in market g,d.
The change in relative cost competitiveness at the market-level, Δθi,g,d,t, is given by equation (4)—defined a la IMF, that is, a higher number means more appreciated. It is constructed as the difference between country i’s cost change and a weighted average of the same cost change observed in all other countries competing in market g,d. The variables Pi,g,t and Pc,g,t represent the cost variable used to estimate cost competitiveness, are specific to each industry in each country, and are expressed in the local currency. The subscript c is used to denote a competitor of the country under study—which, as already mentioned, is denoted by the subscript i. The variable Ei,c,t represents the exchange rate between country i and country c defined as units of country i’s currency per unit of country c’s currency.
The weight αi,c,g,d,t–1 is given by the importance of each country (c) as a competitor of country i in market g,d. We relate this weight to the market participation (share) of country c in market g,d, denoted by γc,g,d,t.5
Defining αi,c,g,d,t–1 = γc,g,d,t–1, however, implies that the sum of the weights of all competitors is less than one, that is, Σ∀(c≠i,g,d) βi,g,d,t–1 · αi,c,g,d,t–1 < 1, because country i is not considered as a competitor of itself in equation (4). Alternatively, one could add country i in the latter sum to make it equal to one. However, doing so would violate an important property that an estimator of the REER should have: ceteris paribus, if all competitors of country i depreciate their currency by 10 percent, then, country i’s REER appreciates by 10 percent—the 10 percent property, for short. This property is violated if country i is added to the summation because the total mass of all countries excluding country i is less than one. Another alternative would be to exclude country i when computing each country’s market share γc,g,d,t–1. However, this solution creates a bias, overstating the importance of small competitors relative to the importance of big competitors. To illustrate this point, assume a foreign market with two equally large competitors plus an exporter from country i. Excluding country i from the computation of the market share would imply that the other two competitors would represent 50 percent of the market, regardless of their actual importance as competitors of country i in that market.
We propose an alternative methodology to measure the importance of each country in each market as a competitor of country i, that is, βi,g,d,t–1 · αi,c,g,d,t–1. We rescale the importance of each competitor in each market based on the relative importance that this competitor has in that market vis-à-vis the importance of all other competitors in all other markets (equation (6)). Our adjusted measure satisfies the condition that the weights of all competitors sum up to one, that is Σ∀(c≠i,g,d) βi,g,d,t–1 · αi,c,g,d,t–1 = 1, satisfies the 10 percent property, and does not over or understate the relative importance of each competitor vis-à-vis all other competitors of country i.
Global Goods, Local Consumption of Local Production, Representative-Product Approach, and Aggregated Cost Measures
The more comprehensive REER estimates available in the current literature incorporate two important elements into the analysis (see Table 1 for details): global goods and local consumption of local production. The former refers to goods that can be characterized as commodities (for example, copper) and for which a more appropriate definition of market is at the world-level rather than at the country-level. Regarding consumption of local production, this refers to the competition that local producers represent at exports’ destination, and it is proxied by the difference between local production of a good and the exports of that good from that destination to the rest of the world.
Our generic approach can be used to consider these two elements. First, we define an artificial additional destination d that will not correspond to a particular country but to the world. Therefore, all goods g that are considered global goods (see next section for details) are assumed to compete in the market (g, d = world). Second, we incorporate local consumption of local production by defining the case d = c, which refers to competitor c competing in the market g, d = c.
We also estimate the REER under the RPA to study the marginal effect of the HPA. The methodology presented in the previous section can easily consider this case as well by redefining all goods into a single good g = ḡ. Following Bayoumi, Jayanthi, and Lee (2005, 2006), we treat global goods separately under the RPA.
Finally, the lack of data limits the extent to which differentiated cost measures by sector can be modeled. The generic approach can be adjusted to consider aggregated measures for the corresponding subsectors within the defined aggregation level by simply defining ΔPc,g,t = ΔPc,ĝ,t ∀g s.t. g ⊂ ĝ, where g refers to an aggregated sector.
Goods and Services
Bilateral trade data for goods are compiled from the United Nations Commodity Trade Statistics Database (COMTRADE). The data include 144 different activity classes of goods (4-digit ISIC Rev.3) across 200 countries over the period 1998–2005. We explicitly consider only the exports of domestically produced goods and exclude the exports of foreign goods to ensure that re-export goods are not considered as additional exports when assessing international competitiveness.6
Bilateral trade data for services are compiled from the OECD Statistics of International Trade in Services. The data include nine categories of services, according to the Extended Balance of Payments Services Classification (EBOPS), across 100 countries over the period 1999–2004. The structure of competitors in 1998 and 2005 is extrapolated from the information available for 1999 and 2004, respectively. The average coverage of bilateral trade data for the MQ ranges from 86 percent of total exports of services for Greece to 59 percent for Spain. The same figures for goods are all above 90 percent.
Disaggregated local production series are needed to estimate local consumption of local production. Obtaining consistent and complete production data at 4-digit level is a challenging endeavor because available databases present significant differences in product and time coverage across countries. We approach this difficulty by combining various databases and generating (rough) estimates where possible. Our main source is the United Nations Industrial Demand-Supply Balance database (IDSB), which contains data at the 4-digit level of ISIC Rev.3 classification, which comprises 127 manufacturing commodities and 78 countries. We extend the IDSB database using (1) the annual growth rates of output reported in Eurostat’s Annual Enterprise Statistics database (4-digit NACE Rev.1.1 production data)7; (2) the observed ratio between sectoral output and aggregate manufacturing output in Eurostat’s Annual Enterprise Statistics database; and (3) the observed growth rate of output production in total manufacturing to extend the series for a maximum window of three years—if output production growth in total manufacturing is not available, we use value added growth in total manufacturing. Finally, we consider only series with complete data for the period 1998 to 2005 (original or estimated) to avoid biases/changes in the REER measures due to truncation of series unrelated to changes in relative cost competitiveness.
With respect to production of services, we use the EU KLEMS Growth and Productivity Accounts database. EU KLEMS database reports data for EU25 countries, Australia, Japan, and the United States until 2005. Production data for royalties and license fees are not considered because the match between EBOPS and ISIC Rev.3 classifications has many shared codes that make it impossible to build a consistent correspondence. Time coverage differs across country and sectors, although to a much lower extent than in the case for goods. We extend the series in the same fashion as we do for goods, using production data from the OECD-STAN database and GDP growth rates. We consider only series with complete time series data (original or estimated).8
The coverage for local production of goods, defined as the share of exports represented by the destination markets for which we can construct data on local consumption of local production, ranges from 50 percent for Greece to 70 percent for Portugal (60 percent for Italy and 66 percent for Spain). The coverage for local production of services, defined similarly, is 85 percent for Greece, 70 percent for Italy, 83 percent for Portugal, and 94 percent for Spain. These last figures are not necessarily comparable with the figures for local production of goods because their computation is based on the available bilateral trade data, which has a lower coverage for services than for goods.
We focus on REER measures that proxy cost competitiveness using the ULC, as opposed to consumer and producer price indices (CPI and PPI, respectively). The latter variables have the advantage of being available for most countries around the world. However, the ULC measure seems to be more appropriate because it considers changes in productivity. The ULC measure allows us to incorporate, albeit not perfectly, important dynamics when considering cost competitiveness, such as the Balassa-Samuelson effect and the effect of innovation and structural reforms across countries.9
Manufacturing ULC is used as the aggregate ULC measure at the country-level. The data are obtained directly from the OECD Analytic Database and WEO database and are available for 38 countries. Industry-level ULC data at 2-digit level is computed using the EU KLEMS database, which covers 28 countries until 2005. Table A1 in Appendix details the different samples. The industry-level ULC is computed as the ratio of the compensation of employees per hour worked to real gross value added per hour worked. The compensation of employees per hour worked is obtained from the ratio of compensation of employees to total hours worked by employees.
Most of the disaggregated ULC series are complete for the countries covered by the EU KLEMS database, although data for 2006 are not available. On average, 95 percent of manufacturing sectors and 90 percent of all sectors have complete time-series for the period 1995–2005. Disaggregated ULC series with incomplete data are not considered and are replaced by the country’s ULC series for the manufacturing sector computed from the EU KLEMS database. We replace them to avoid biases/changes in the REER measures due to truncations of series unrelated to changes in relative cost competitiveness.
The annual average nominal exchange rates are obtained from the IFS database.
We refer to globally traded goods as those goods whose prices are quoted on organized world exchanges as defined by Rauch (1999), who classifies goods into three categories at the 4-digit SITC Rev.2. classification: commodities, reference-priced goods and differentiated goods. This classification is based on whether a good is traded and priced on organized world exchanges, listed in trade publications but not traded on organized exchanges, or does not posses a reference price, respectively.
In order to identify global goods within the 4-digit ISIC Rev.3 classification, we identify all the ISIC Rev.3 codes associated with each SITC Rev.2 in Rauch (1999) by using the UN correspondence tables. We assign a value of 1 to each good priced in organized world exchanges, a value of 2 to each good listed in trade publications but not traded on organized exchanges, and a value of 3 to each good that does not possess a reference price. We calculate the average value of the associated codes for each ISIC Rev.3 code in a similar way to Jensen (2006). We define a good as globally traded at the 4-digit ISIC Rev.3 level if the value of the calculated average within each ISIC Rev.3 code lies in the interval of [1,2). Goods with values in the interval of [2,3] are not considered as globally traded goods.
Rauch (1999) presents two classifications: the “conservative” and the “liberal.” The liberal version maximizes the number of globally traded goods in the cases where there was room for discretion in the sorting. Table A2 in Appendix presents the resulting group of goods identified as global goods at the 4-digit ISIC Rev.3 classification under both alternatives, conservative and liberal.
For further comparisons, we also report the goods that appear as global when applying a methodology similar to the one applied to Rauch’s list to the list of global (commodity) goods defined by Bayoumi, Jayanthi, and Lee (2005, 2006) at the 2-digit SITC Rev.3 level. The resulting number of global goods from the latter source is higher than the one resulting from our methodology, most likely because in Bayoumi, Jayanthi, and Lee (2005, 2006) the list is defined at a more aggregated category of goods than in Rauch (1999) and this paper. In our analysis, we use Rauch’s (1999) liberal classification following Jensen (2006), which results in a list that is closer to the one implied by the methodology used in Bayoumi, Jayanthi, and Lee (2005, 2006).
This section presents the sensitivity analysis of the REER to the HPA and to the inclusion of services exports. Based on the methodology described in Section II, we estimate the path of the REER indices under the alternative approaches and present comparative statistics.
We first compare the estimated path of the REER under the HPA, denoted by RG, with the equivalent REER index under the RPA, denoted by R1G. The difference represents the effect of relaxing the assumption that all non-global goods are treated as identical goods, and as a result compete in the same market ḡ,d. Second, we compare RG with the estimates obtained for the evolution of the REER that considers only exports of services (under the HPA), denoted by RS. This allows us to have a perspective of how the REER for goods compares with the REER for services. Third, we compute the aggregated REER index for goods and services, denoted by RGS, and we compare it with R1G. This difference represents the sensitivity of the REER under the RPA to both the HPA and a broader coverage of exports that includes services. In addition, we perform different robostness checks.
We also study the sensitivity of the REER for goods to heterogeneous cost dynamics across sectors. We compare RG with the estimated REER for goods under the HPA and the heterogeneous cost dynamics assumption, denoted by RGd, where d stands for differentiated cost measures. As detailed in the introduction, these results should be taken with caution because of data limitations and should be read as an exploratory effort to determine the effect of differentiated sectoral cost measures on the REER.
The contrast between different approaches, for example, the comparison between RG and R1G, is performed in two dimensions. First, we present the difference observed in the appreciation rates from 1998 to the corresponding year shown in the tables for both indices. Using 1998 as the base year for comparison is an ad hoc rule, which has no other merit than being year prior to the adoption of the euro adoption by all the euro area countries (January 1, 1999), except for Greece (January 1, 2001). This difference in levels is computed following equation (7).
where Δt% refers to the growth rate of the index observed from 1998 to year t and i refers to the country whose REER is analyzed.
Second, we study the difference observed between the two estimations of the REER considered, but constructing each estimator relative to the corresponding REER observed in the remaining 11 countries of the euro area. This difference-in-difference estimator is computed following equation (8).
where Δt% refers to the growth rate of the index observed from 1998 to year t, i refers to the country whose REER is analyzed, and EA refers to the remaining 11 countries of the euro area.10
The difference-in-difference (DD) estimator is our preferred estimator for two reasons. First, it allows us to control for methodological issues specific to each type of estimation that could be driving the results without necessarily reflecting changes in relative cost competitiveness. Second, it allows us to control for the equivalent results observed in the rest of 11 euro area countries. Therefore, the DD estimator represents the change in the relative international competitiveness position between the euro area countries. This does not mean that the DD estimator considers only direct competitors from the euro area, but that it compares among euro area countries each country’s position with respect to its direct competitors across the world.
The euro area countries are all part of the same currency union, and therefore, are a natural benchmark to compare the evolution of the REER in each of the MQ countries and control for potential methodological differences particular to each type of estimation. This approach has also an important economic meaning. The exchange rates between euro area countries are fixed—although the euro is still sensitive to the international competitiveness of the euro area as a whole, in line with the standard exchange rate mechanisms associated with floating currencies. No rebalancing through nominal exchange rate movements is then possible between euro area countries, but only through productivity and wage growth differentials, which tend to take longer to materialize. As a result, divergence in international competitiveness between euro area countries is an important element when assessing the medium-term economic perspective of individual euro area countries.
The Effect of the Heterogeneous-Product Approach and Services
REER for Goods under the HPA
Table 2 presents the estimation of the REER indices RG and R1G for the MQ countries, alongside the estimations for the two main euro area countries, France and Germany, for further comparison. The contrast between RG and R1G is reported in Table 3. The figures for the DD estimator imply that under the HPA Portugal’s, Italy’s, and Spain’s REERs are less appreciated in the range of 2 percent in 2006 (1998 base). The difference is larger in the case of Greece, on the order of 7 percent. This indicates that, relative to the effect on the other 11 euro area countries, the REER under the RPA in Greece is 7 percent more appreciated than what the model assuming the HPA suggests (since 1998).
We study the robustness of our methodology and results by performing three additional contrasts. First, we compare our computation of R1G with the closest measure available in the literature (based on Bayoumi, Jayanthi, and Lee, 2005, 2006); second, we modify the sample of countries considered; and third, we study how our measure of REER changes when domestic market competition is considered.
Table 4 presents the comparison between the R1G and RIMF, where IMF stands for the WEO estimates of the REER based on the methodology proposed by Bayoumi, Jayanthi, and Lee (2005, 2006)—the closest source to our methodology that includes the latest developments in the literature and uses the RPA. The results for the DD estimator are all within the±1 percent range, suggesting that our methodology under the RPA yields similar results to the existing methodologies based on the RPA.11
Our sample of countries is based on the available information for ULC in manufacturing in the OECD and the WEO databases (Table A1). This sample differs from the sample used to compute RIMF, which considers 27 countries. An interesting aspect of the additional 11 countries used in our sample is that they constitute a sample of emerging countries not represented in the sample of 27 advanced economies, with the exception of China.12
We performed a second robustness check to study if our estimates of RG are sensible to including the additional 11 emerging countries. We contrast RG with the REER estimated under the HPA considering only the 27 countries, denoted by RG27. The results, presented in Table 5, indicate that the REER estimated with the sample of 27 countries does not differ substantially from the REER estimated with the sample of 38 countries. The results for the DD estimator are all within the±1 percent range.
Finally, we consider the potential importance of domestic market competition for measuring international competitiveness. Owing to a lack of consistent data on disaggregated internal production across MQ countries, our analysis centers on the external markets where each MQ country competes. Table 6 presents the results of contrasting RG with the REER under HPA including the available information on internal markets, denoted by RGIM. It indicates that the marginal effect of domestic markets is small with differences in the range of ±1 percent.13
Difference in the Structure of Competitors
We complement the results on the effect of the HPA presented in Table 3 with an aggregate view of the difference in the structure of competitors implied by the HPA and the RPA. The larger the difference, the greater the likelihood of finding a large difference between the corresponding REER measures. The actual effect, however, will depend on the interaction between the different weights and the distribution of the change in ULC across countries. In the limit, even large differences in the weights will have no effect if all countries present identical changes in their ULCs, and vice versa, even small differences in the weights can have large effects if changes in ULC are significantly different across countries.
We capture the difference in the structure of competitors implied by each approach using the formula described in equation (9). The variable λi,t aggregates the difference observed in the weight assigned to each competitor of country i in period t.
where χi,c,t is defined by equation (10) and βi,g,d,t · αc,g,d,t refers to the relative importance of each competitor c in market (g,d) with respect to all other competitors in all other markets.14 The variable
Table 7 presents the results for λi,t. Greece, whose sensitivity to the HPA is the highest among all countries, presents the highest level of difference. This was expected, although as mentioned above, the fact that a difference is observed in λi,t does not necessarily imply a difference in the REER measure; it only makes it more likely. In fact, among the rest of the countries, Portugal stands out with the largest value for λi,t, but does not present the highest sensitivity to the HPA.
Table 7 also presents the difference in the structure of competitors when comparing the HPA and the partners-approach, denoted by λp. The partners-approach refers to considering a country’s trade partners as its competitors and it is used by some sources; see Chinn (2006) for more details. The results show that considering the partners-approach yields a stronger difference, two to three times the size of λi,t. A larger difference is expected because considering competitors only on the base of trade partnerships deviates substantially from the concept of competitors used in this paper.
REER for Goods and Services under the HPA
Table 8 presents the estimation of the REER index RS, which includes only services exports. To illustrate the evolution of the services component, we compare RS and RG in Table 9. These results suggest that except for the case of Greece, the services component of the REER has appreciated less than the goods component for the MQ countries. This difference ranges from −3.4 percent for Italy to −0.9 percent for Spain (DD estimator). For Greece, the difference goes in the opposite direction in the range of 7 percent (DD estimator).
Finally, Table 10 presents the estimation of the REER index RGS, which includes both goods and services. We compare RGS with R1G in Table 11, which represents the aggregate sensitivity of the REER under the RPA to both the HPA and a broader coverage of exports that includes services.
The results suggest that these two additional factors together—HPA and services—have had a marginal effect on the REER on the order of −2 percent to −3 percent for all the MQ countries: −2 percent for Greece, −2.3 percent for Spain, −2.4 percent for Portugal, and −2.8 percent for Italy. These numbers are consistent with the previous tables, where the smaller appreciation observed in goods for Italy, Portugal, and Spain under the HPA adds to the smaller appreciation observed in services relative to goods. For the case of Greece, the strongest difference observed under the HPA shrinks significantly when combined with the larger appreciation observed in services relative to goods.
Differentiated ULC by Sector
Product heterogeneity (HPA) and services exports refine the REER as a measure of international competitiveness, but, as detailed in the previous section, these two factors do not change substantially the broad picture of international competitiveness in the MQ. In this section, we explore the sensitivity of the REER to the HPA with differentiated cost measures at the sector-level. Differentiated cost measures would yield a more accurate picture of international competitiveness to the extent that productivity and production costs vary across sectors. The set of results presented in this section, which indicate a higher sensitivity of the REER relative to the assumption of homogenous cost dynamics, should, however, be taken with caution. Given the data limitations detailed before, these results should be read as an exploratory effort to determine the effect of differentiated sectoral cost measures on the REER.
The results for the contrast between the REER estimated with an aggregated ULC measure (RG) and the REER estimated with a differentiated ULC by sector (RGd) point to a higher sensitivity of the REER to the assumption of homogenous cost dynamics across sectors—both measures based on the sample for goods because of the excessive volatility found in the data for service. As shown in Table 12, the absolute differences range between 2 and 6 percent. For this contrast, RG is computed using the limited data set available for the calculation of RGd.
These results are not sensitive to outliers. We recalculated the REER eliminating the 0.5 percent tails of the distribution of the annual ULC growth rates observed since 1998. No substantial differences from the results obtained in Table 12 were found. As an additional robustness check, we compared the results obtained for the contrast between RG and R1G (Table 3) with an equivalent contrast using the limited data set available for the calculation of RGd. The differences between both cases are all within the±1 percent range.
IV. The Profile of International Competitors
The HPA proposed in our study also allows a quantitative assessment of each country’s profile of competitors. Such evidence provides information about the exposure of each country to its key competitors around the world—for example, the exposure to emerging competitors like China, a country that has shown a strong pattern of productivity and trade growth, or the exposure to countries facing significant changes in their cost structure, such as the wage moderation observed recently in Germany or the depreciation of the nominal exchange rate observed in the United States during recent years. Our definition of markets also captures the potential vulnerability of each country’s sectors to changing market conditions in competitors’ sectors beyond the country level.
For all six countries, the bulk of competition comes from the advanced and emerging economies, representing on average 95 percent in goods (except for Greece, 92 percent) and 98 percent in services. Since the late 1990s, there has been a change in the composition with emerging economies taking greater importance: they represented in 2005 14 percent of overall exposure to competition in goods for Spain, 19 percent for Italy and Portugal, and 22 percent for Greece (Table 13). China appears as the largest emerging competitor in goods for all four countries, representing at least half of the increase in the importance of emerging economies since 1998.
|Euro area, 1998||47.5||49.7||60.0||59.9||48.9||42.6|
|Euro area, 2005||47.0||48.6||58.6||58.5||48.6||41.1|
|Advanced economies, 1998||72.4||81.4||83.7||85.9||85.7||84.5|
|Advanced economies, 2005||69.9||75.7||77.0||81.5||80.1||77.4|
|Emerging economies, 1998 (1)||18.6||14.7||12.7||11.1||11.7||11.8|
|Emerging economies, 2005 (2)||22.0||19.2||19.0||14.4||15.9||17.7|
|Change in percentage points (2)-(1)||3.4||4.5||6.3||3.3||4.3||5.9|
|Change in percentage points due to China||3.5||3.4||3.4||1.8||2.3||2.6|
Among the advanced economies, the euro area countries represent 59 percent of the competition in goods faced by Spain and Portugal, 49 percent for Italy, and 47 percent for Greece. These data indicate that Spain and Portugal are more exposed to euro area competition and therefore less exposed to changes in the value of the euro. There has been a declining trend since 1998 in the range of 1 percentage point for all countries, which is smaller than the change observed for the aggregate of advanced economies.
From a sectoral point of view, the four MQ countries compete more in low-technology sectors with China (see Table 14). However, the importance of China in high-technology sectors is growing as well (see Figure 2). As a comparison, France and Germany compete more strongly with China in high-technology sectors, suggesting that China should not be seen as a potential competitor in low-tech sectors only.15
|China as competitor|
|Germany as competitor|
|Low-tech sectors||6.29||6.63||4.61||3.90||4.95||…|Figure 2.Importance of China and Germany as Competitors of the Mediterranean Quartet in High- and Low-Tech Sectors in Goods
Note: Technology intensity of sectors is defined according to OECD (2007) classification of industries with respect to intensity of technology used. High-tech (that is, H_tech) industries comprise industries with high and medium-high intensity of technology used and low-tech (that is, L_tech) industries comprise industries with low and medium-low intensity of technology used. Percentages refer to the share of competition from each country.
Figure 2 also show the importance of Germany—the main advanced-country competitor—as a competitor of the MQ. The almost flat or sometimes decreasing importance of Germany highlights the strong growth of China’s importance in both high- and low-technology sectors. Nonetheless, at least until 2005, Germany was still a bigger competitor for the MQ than China in both types of sectors.
In services, emerging markets represent on average about one-third of their importance in goods, showing also, although to a lesser extent, a similar increase in recent years (see Table 15). From 1999 to 2004, the composition shifted to emerging economies in the range of 3 percent for Greece and Italy, 2 percent for Portugal, and 1 percent for Spain. The data suggest that China does not appear as a strong competitor in services.
|Euro area, 1999||37.4||49.6||60.3||54.4||42.8||46.0|
|Euro area, 2004||37.2||50.2||60.3||48.1||42.5||40.7|
|Advanced economies, 1999||95.9||94.1||97.1||95.8||95.1||93.6|
|Advanced economies, 2004||92.0||89.7||94.7||93.5||91.1||87.8|
|Emerging economies, 1999 (1)||3.5||4.9||2.4||3.5||4.2||5.0|
|Emerging economies, 2004 (2)||6.4||7.5||3.9||4.7||6.7||8.6|
|Change in % points (2)–(1)||2.9||2.6||1.5||1.2||2.5||3.6|
|Change in % points due to China||0.3||0.2||0.1||0.1||0.2||0.3|
|Change in % points due to the largest 5 EE||2.0||1.6||0.6||0.5||1.8||2.3|
Among the advanced economies, the euro area countries represent 60 percent of the competition in services faced by Portugal, 50 percent for Italy, 48 percent for Spain, and 37 percent for Greece. There has been a nil trend since 1998 for all countries except for Spain, whose euro area competition has declined by 5 percentage points. These figures for services should be read with caution given the incomplete availability of the data for bilateral trade of services.
We develop a complete methodology to reexamine the evolution of international competitiveness in the MQ, as measured by the REER. In addition to the elements considered in the existing literature, we (1) use a micro-based approach that considers product heterogeneity when identifying each country’s international competitors and their weights and (2) include a comprehensive analysis of the services sector. Our approach enriches the REER analysis by identifying more accurately each country’s direct international competitors and providing an aggregate view of international competitiveness that encompasses the complete export sector.
Our main findings suggest that the effect of considering both the more micro-based structure of competitors and exports of services implies a modest lower real appreciation from 1998 to 2006 on the order of 2 to 3 percent for all MQ countries—2 percent for Greece, 2.8 percent for Italy, 2.4 percent for Portugal, and 2.3 percent for Spain. These estimates are based on a difference-in-difference estimator that controls for the equivalent effect observed in the rest of 11 euro area countries.
Finally, the methodology proposed in this paper also allows a detailed view of the structure of each country’s competitors. Our findings indicate that the bulk of competition for the MQ still comes from the advanced economies, especially from the euro area. Nonetheless, there has been a change in the composition with emerging economies taking more importance since the late 1990s, particularly China in both the high- and low-technology sectors.
|Country-Level ULC||Differentiated ULC|
|Country||Sample of 38|
|Sample of 27|
|Sample of 28|
|5 China, P.R.: Hong Kong||Yes||Yes||No|
|7 Czech Republic||Yes||No||Yes|
|21 Macedonia, FYR||Yes||No||No|
|24 New Zealand||Yes||Yes||No|
|29 Slovak Republic||Yes||No||Yes|
|31 South Africa||Yes||No||No|
|35 Taiwan POC||Yes||Yes||No|
|37 United Kingdom||Yes||Yes||Yes|
|38 United States||Yes||Yes||Yes|
|ISIC Rev.3 (4-digit)||Activity Description||Conservative||Liberal||IMF|
|0111||Growing of cereals and other crops n.e.c.||Global||Global||Global|
|0112||Growing of vegetables, horticultural specialties and nursery products||—||—||Global|
|0113||Growing of fruit, nuts, beverage and spice crops||Global||Global||Global|
|0121||Farming (cattle, sheep, goats, horses, asses, mules and hinnies; dairy)||Global||Global||Global|
|0122||Other animal farming; production of animal products n.e.c.||—||—||Global|
|0200||Forestry, logging and related service activities||—||—||Global|
|1110||Extraction of crude petroleum and natural gas||—||Global||—|
|1200||Mining of uranium and thorium ores||—||Global||Global|
|1310||Mining of iron ores||Global||Global||Global|
|1320||Mining of non-ferrous metal ores, except uranium and thorium ores||—||Global||Global|
|1410||Quarrying of stone, sand and clay||—||—||Global|
|1421||Mining of chemical and fertilizer minerals||—||—||Global|
|1422||Extraction of salt||—||—||Global|
|1429||Other mining and quarrying n.e.c.||—||—||Global|
|1511||Production, processing and preserving of meat and meat products||Global||Global||Global|
|1512||Processing and preserving of fish and fish products||—||—||Global|
|1513||Processing and preserving of fruit and vegetables||—||—||Global|
|1514||Manufacture of vegetable and animal oils and fats||Global||Global||Global|
|1520||Manufacture of dairy products||—||Global||Global|
|1531||Manufacture of grain mill products||—||—||Global|
|1532||Manufacture of starches and starch products||—||—||Global|
|1533||Manufacture of prepared animal feeds||—||—||Global|
|1541||Manufacture of bakery products||—||—||Global|
|1542||Manufacture of sugar||Global||Global||Global|
|1543||Manufacture of cocoa, chocolate and sugar confectionery||—||—||Global|
|1544||Manufacture of farinaceous products (macaroni and similar)||—||—||Global|
|1549||Manufacture of other food products n.e.c.||—||—||Global|
|1551||Distilling, rectifying and blending of spirits||—||—||Global|
|1552||Manufacture of wines||—||—||Global|
|1553||Manufacture of malt liquors and malt||—||—||Global|
|1554||Manufacture of soft drinks; production of mineral waters||—||—||Global|
|1600||Manufacture of tobacco products||—||—||Global|
|2010||Sawmilling and planing of wood||—||—||Global|
|2411||Manufacture of basic chemicals, exc. fertilizers & nitrogen compounds||Global||Global||—|
|2412||Manufacture of basic precious and non-ferrous metals||—||Global||—|
|2720||Manufacture of basic precious and non-ferrous metals||Global||Global||Global|
|9302||Hairdressing and other beauty treatment||—||—||Global|
AgenorP.R.1995 “Competitiveness and External Trade Performance of the French Manufacturing Industry” IMF Working Paper 95/137 (WashingtonInternational Monetary Fund) pp. 1–23.
Bank of Japan2007 “Explanation of the Effective Exchange Rate (Nominal, Real)” Note of Research and Statistics DepartmentMay pp. 1–8.
BayoumiT.S.Jayanthi and J.Lee2005 “New Rates from New Weights” IMF Working Paper 05/99 (WashingtonInternational Monetary Fund) pp. 1–154.
BayoumiT.S.Jayanthi and J.Lee2006 “New Rates from New Weights” IMF Staff Papers Vol. 53 No. 2 (WashingtonInternational Monetary Fund) pp. 272–305.
Bini-SmaghiL.2007 “Asymmetric Adjustment in Monetary Unions: Evidence from the Euro Area” presentation at the EMU conference “The Eurozone under Stretch? Analyzing Regional Differences in EMU,”BerlinJune19.
BlanchardO.2006a “A Macroeconomic Survey of Europe” seminar presentation for the World Economic Laboratory Massachusetts Institute of TechnologySeptember pp. 1–23. Available via the Internet: www.econ-www.mit.edu/files/761.
BlanchardO.2006b “Portugal, Italy, Spain, and Germany: The Implications of a Suboptimal Currency Area” seminar presentation for the World Economic Laboratory Massachusetts Institute of TechnologyApril pp. 1–26. Available via the Internet: www.econ-www.mit.edu/files/758.
BuldoriniL.S.Makrydakis and C.Thimann2002“The Effective Exchange Rates of the Euro” European Central Bank Occasional Paper No. 2February pp. 1–49.
CataoL.A.V.2007 “Why Real Exchange Rates?” Finance and Development Vol. 44 No. 2 pp. 46–7.
CerraV.J.Soikkeli and S.C.Saxena2003 “How Competitive Is Irish Manufacturing?” Economic and Social Review Vol. 34 No 2 (summer/autumn) pp. 173–93.
ChinnM.D.2006 “A Primer on Real Effective Exchange Rates: Determinants, Overvaluation, Trade Flows and Competitive Devaluation” Open Economies Review Vol. 17 No. 1 pp. 115–43.
DurandM.C.Madaschi and F.Terribile1998 “Trends in OECD Countries’ International Competitiveness: The Influence of Emerging Market Economies” OECD Economic Department Working Paper No. 195 (ParisOrganization for Economic Cooperation and Development) pp. 1–57.
DurandM.J.Simonm and C.Webb1992 “OECD Indicators of International Trade and Competitiveness” OECD Economics Department Working Paper No. 120 (ParisOrganization for Economic Cooperation and Development) pp. 1–51.
European Commission2006 “Adjustment Dynamics in the Euro Area: Experiences and Challenges” in The EU Economy Review 2006ed. byM.McCarthy and M.Watson (BrusselsEuropean Commission) pp. 79–107.
FungS.S. and M.Klau2006 “The New BIS Effective Exchange Rate Indices” BIS Quarterly ReviewMarch pp. 51–65Available via the Internet http://ideas.respec.org/albis/kisqtr/oboze.html.
HatzichronoglouT.1997 “Revision of the High-Technology Sector and Product Classification” Science Technology and Industry Working Paper No. 216 (ParisOrganization for Economic Cooperation and Development) pp. 1–25.
HummelsD.I.Jun and K.Yi2001 “The Nature and Growth of Vertical Specialization in World Trade” Journal of International Economics Vol. 54 No. 1 pp. 75–96.
JensenP.E.2006 “Trade, Entry Barriers, and Home Market Effects” Review of International Economics Vol. 14 No. 1 pp. 104–18.
LipschitzL. and D.McDonald1992 “Real Exchange Rates and Competitiveness: A Clarification of Concepts, and Some Measurements for Europe” Empirica Vol. 19 No. 1 pp. 37–69.
LoretanM.2005 “Indexes of the Foreign Exchange Value of the Dollar” Federal Reserve Bulletin Vol. 91 No. Winter pp. 1–8.
MarshI.W. and S.P.Tokarick1996 “An Assessment of Three Measures of Competitiveness” Weltwirtschaftliches Archiv/Review of World Economics Vol. 132 No. 4 pp. 700–22.
NearyP.2006 “Measuring Competitiveness” IMF Working Paper 06/209 (WashingtonInternational Monetary Fund) pp. 1–19.
Organization for Economic Cooperation and Development (OECD)2007 “Classification of Manufacturing Industries Based on Technology” in OECD Science Technology and Industry Scoreboard 2007 overall co-ordinator of publicationZunigaP. (ParisOECD) Annex 1 pp. 219–21.
PapademosL.2007“Inflation and Competitiveness Divergences in the Euro Area Countries: Causes Consequences and Policy Responses” speech European Central Bank FrankfurtSeptember7. Available via the Internet: www.ecb.int/press/key/date/2007/html/sp070907_2.en.html.
RauchJ.E.1999 “Networks versus Markets in International Trade” Journal of International Economics Vol. 48 No. 1 pp. 7–35.
RogoffK.1996 “The Purchasing Power Parity Puzzle” Journal of Economic Literature Vol. 34 No. 2 pp. 647–68.
RoubiniN.2007“Le Differenze che Frenano l’Europa dell’Euro”La Repubblica (Milan) August8 pp. 1 and 21. Available via the Internet: www.ricerca.repubblica.it/repubblica/archivio/repubblica/2007/08/08/le-differenze-che-frenano-europa-dell.html.
TurnerP. and J.Van t’Dack1993 “Measuring International Competitiveness” BIS Economic Paper No. 39 (GenevaBank for International Settlements) pp. 1–152.
Herman Bennett is Visiting Lecturer in Public Policy at the Harvard Kennedy School of Government and Ziga Zarnic is a researcher and PhD candidate in economics at LICOS Centre for Institutions and Economic Performance at Catholic University of Leuven. Proprietary data used in this paper come from the United Nations Statistics Division and the Organization for Economic Cooperation and Development. The authors thank James Daniel, Michael Deppler, Julio Escolano, Eva Gutierrez, Alex Hoffmaister, Deniz Igan, Emilia Jurzyk, Bogdan Lissovolik, Stephen Tokarick, an anonymous referee, and seminar participants at IMF, LICOS—Katholieke Universiteit Leuven, and Université Catholique de Louvain for valuable comments and discussions.
See, for example, “The Quest for Prosperity,” Special Report, The Economist, March 15, 2007, which stated: “In particular, the Mediterranean quartet of Italy, Spain, Portugal and Greece has suffered a huge loss of competitiveness in a relatively short time…. This loss is reflected in colossal current-account deficits … or pitifully slow growth.”
We initially attempted to base our definition of international competitors on the degree of substitution between goods. We intended to include producers of other 4-digit level industries as competitors, weighting their importance by the degree of substitution between the corresponding goods, as measured by the cross elasticity of substitution. Available volume indices, however, present important measurement error. This affects the estimation of price indices, necessary for the estimation of the cross elasticities, and therefore, the whole methodology would have resulted in a significant increase in the variance of the REER estimates. Also, the presence of monopolistic competition at different intensities across industries and countries suggests that the estimation of cross elasticities is subject to potential identification problems. We have, therefore, assumed the simplifying assumption that the structure of international competitors within 4-digit sectors provides a good representation of the structure of competitors that exporters within those sectors face. The same is assumed for services within the 2-digit industry-level.
Throughout the paper, the notation Δ denotes the natural logarithmic difference between t and t–1 of the corresponding variable.
The methodology presented in this paper can also be applied in the ideal case of having a database with bilateral trade of value added, that is, total export value net of imported intermediate inputs. In this case, βi,g,d,t–1 and γc,g,d,t–1 would be computed based on value added figures rather than on final sale figures.
The lack of comparable data on value added of bilateral trade flows prevent us from basing our measure of international competitiveness on value added of exports. We try to address this issue by accounting for re-exports and defining detailed markets for traded products. By doing so, we address part of the increase in the variance of the REER estimator due to the fact that the share of value added of exports may differ across products. Ideally, one would like to have a database with value of exports net of imported intermediate inputs.
Eurostat’s Annual Enterprise Statistics database has good coverage of production data, but includes only members of European Union. With respect to the correspondence used, we consider data that (1) corresponds to only one type of product at the 4-digit ISIC Rev.3 level and (2) does not share the image with data points that correspond to more than one type of product at the 4-digit ISIC Rev.3 level.
The combined information on bilateral trade and consumption of local production accounts for more than 36 million observations.
See Lipschitz and McDonald (1992); Marsh and Tokarick (1996); Agenor (1995); Turner and Van t’Dack (1993); and Cerra, Soikkeli, and Saxena (2003) for more on the advantages and disadvantages of using ULC as a measure of cost competitiveness.
Our estimates as well as the IMF estimates reported are based on ULC data as of August 2007.
The sample of 27 countries covers on average 70 to 85 percent of MQ’s competitors, while the sample of 38 countries covers 80 to 90 percent.
We cannot perform this analysis for Greece because of insufficient data on its disaggregated structure of production.
See Section I for more details.