EQCHANGE: A World Database On Actual And Equilibrium Effective Exchange .

CEPII Working PaperEQCHANGEThe substantial enhancement introduced byEQCHANGElays in the second sub-databasewhich includes assessments of equilibrium real e ective exchange rates and correspondingcurrency misalignments. More precisely, we estimate ERER based on the BEER approach(Clark and MacDonald, 1998). We have chosen this framework as it allows us to provideseveral ERER estimates for the vast majority of countries in the world (182 countries) andover a reasonably long time-span (from 1973 to 2016). Currency misalignments are thendeduced from the di erence between real e ective exchange rates and their equilibriumvalues; we calculate various misalignments, depending on the de nition of real e ectiveexchange rates indices (weighting schemes di er for example), the speci cation of theequilibrium exchange rate model and the sample of countries.By providing publicly available indicators of equilibrium exchange rates, our databasehelps documenting key debates such as the role of exchange rates on global and/orregional imbalances, the use of exchange rate as a mercantilist tool, the impact ofcurrency misalignments on resources' reallocation between the tradable and the nontradable sectors and on economic growth, to name a few.The primary objective of this paper is to provide a comprehensive account of our methodological framework to calculate its set of e ective exchange rates and of equilibrium realexchange rates from which currency misalignments are derived. To this end, Section 2is devoted to the presentation of the main features of theEQCHANGEdatabase. Asan illustration, Section 3 presents the evolution and essential characteristics of currencymisalignments in the world from 2015 to 2016. Section 4 concludes the paper.2. The EQCHANGE databaseTheEQCHANGEdatabase developed by the CEPII consists of time series of: nominal and real e ective exchange rates (Section 2.1); behavioral equilibrium exchanges rates and resulting currency misalignments (Section2.2).2.1. The e ective exchange rates datasetSeveral institutions, such as the World Bank (WB), the International Monetary Fund(IMF), the Bank of International Settlements (BIS), or think tanks such as Bruegel publish more or less regularly indexes of nominal and real e ective exchange rates. The6

CEPII Working PaperEQCHANGEFigure 1 E ective exchange rates dataset: country coverageEQCHANGEdatabase developed by the CEPII has two distinct characteristics whencompared to those databases.First, theEQCHANGEdatabase has the widest coun-try/time coverage as we compute annual e ective exchange rates (both in nominal andreal terms) for 187 countries over the 1973-2016 period (see the list of countries inAppendix A.1, and Figure 1).Second, by using a simple export weighting method, theEQCHANGEdatabase includese ective exchange rates indices calculated and available along various dimensions, including the weighting scheme and trading partners' coverage. Therefore our database allowsthe adoption of alternate options for the selection of data.rates indices are available for two weighting schemes.Firstly, e ective exchangeThe rst one is time-invariantand consists in two weight sets respectively representative of foreign trade between2008-2012, and(ii )1973-2016.(i )The second weighting system is time-varying and isessentially based on non-overlapping ve-year average weights.8Secondly, e ective ex-change rates rely on broad-based and narrow-based indices which are calculated againsttwo main groups of trading partners, i.e. against 186 trading partners and the top 30trading partners for each country. Di erences between our database and other ones arereported in Table 1.The following periods 1973-1979; 1980-1984; 1985-1989; 1990-1994; 1995-1999; 2000-2004; 20052009; 2010-2016 have been considered.87

CoverageCountries PeriodTypeWeighting systemPeriod CEPII187(i)1973-2016(ii)FixedTime-varying1973 20162008 2012Basket186 and 30Data frequencyUpdatesAnnualRegularly5-year rly andRegularlyAnnualFixedTime varyingFixed1998-20031-year window2011-2013Notes: the column "Basket" indicates the number of trade partners used to calculate the ANGEBIS(i)(ii)(iii)(iv)CEPII Working PaperDatabaseTable 1 Comparison of various datasets

CEPII Working PaperEQCHANGE2.1.1. Methodological featuresAn e ective exchange rate measures the rate at which a country's currency exchangesagainst a basket of other currencies, in either nominal or real terms.iThe nominal e ective exchange rate of countryvalue of the currency of countryiin periodt (NEERi ;t )measures theagainst a weighted average of foreign currencies:NEERi ;t NYjNERwi j;t(1)i j;t 1NER is the index of the nominal bilateral exchange rate between the currency9of country i and the currency of its trade partners j in period t , N denotes the number10of trading partners and wis the trade-based weight associated to the partner j .wherei j;ti j;tThe real e ective exchange rate of countryiin periodt (REERi ;t )weighted average of real bilateral exchange rates against each of itsis calculated as theNtrading partnersj:REERi ;t NYRERi j;tof the country NERP Pi j;tii ;twi j;t(2)i j;tjwhereRER 1is an index of the real exchange rate of the currencyj;tjvis-à-vis the currency of the trading partnerstand respectively for the price index of countryiin periodand of countryt. Pi ;tandPj;tj.With these de nitions, a real (nominal) appreciation of the domestic currency is recordedas an increase in the real (nominal) e ective exchange rate index.Constructing an e ective exchange rate index (both in nominal and real terms) for agiven country then requires determining the trade-weight associated to each of its partners. In many circumstances, the exchange rate plays the role of a relative price of tradedgoods; hence, the relevant weights usually involve trade weights. Moreover, in most openeconomies, policy-makers are more interested in the total international competitive position of their country. In this case, appropriate e ective exchange rate measures shouldinclude both import and export weights.Computing import weights is fairly straight-The nominal exchange rate is expressed as the number of foreign currency units per domestic currency.An increase therefore corresponds to an appreciation of the domestic currencyagainst the foreign currency.PThese weights are normalized so that their sum is equal to one, i.e. 1.910Nj9 1 wi j;t

CEPII Working PaperEQCHANGEforward: import weights are based on bilateral imports, i.e. on the relative importanceof each of the partner countries,j,in total imports of a countryi.Turning to exportweights, various procedures exist including bilateral and double export weighting schemes.We prefer here to follow the bilateral-based approach for its simplicity and transparencyproperties. Indeed, this procedure is based on bilateral exports and therefore takes intoaccount only competition between the domestic country and its direct trading partners.In doing this, we have chosen not to consider the possible indirect competition on thethird markets based on the calculation of a double export weight. Speci cally, international institutions such as the IMF, the WB and the BIS construct their real e ectiveexchange rate series attempting to capture third market e ects. Compared to our simpleweighting procedure, the double export weights method contains additional informationi 's exporters in each given foreign market from ex11 However, whileporters of the countries included in the group of the N trading partners.on the competition faced by countrydelivering a more rigorous assessment of the export competition in foreign markets, thisweighting scheme can lead to a smaller area and time coverage and provides less exibility in updating the dataset, because of limitations in data availability. In particular, theinclusion of the gross value of the domestically produced supply of manufactured goodsin the calculation of the double export weight tends to restrict the number of countriesthat can be considered. In addition, if we compare our calculated overall weights (seeparagraph below) to those of the IMF, we do not observe any signi cant di erences. Wetherefore conclude that our simpler approach is broadly equivalent to that of the IMFwhen considering the overall weights. All these reasons justify our choice to not replicatethe double export weighting scheme.The overall weight of each partnerj in the trade of country iat periodt (wi j;t )is obtainedas the weighted average of the export and import weights:W M M Xi j;ti ;twhereandMWI mpi j;ti ;tand i ;ti ;t WI mpi j;t X M Xi ;ti ;tWXare total imports and total exports of countryi ;tEx pare partner country WEx pi j;tandi j;tj 's i ;t(3)import and export weights, respectivelyi.Indeed, under this approach, the export weight for any country is derived as a combination of two components: a bilateral export weight, which accounts for direct competition between exporters and domesticproducers in a particular export market; and a third-market export weight, which captures competitionbetween exporters from two di erent countries in a third market. The methodology is detailed in Turnerand Van't dack (1993).1110

CEPII Working PaperEQCHANGEThe import and export weights (Equations (4) and (5)) of each partneras its simple share of countryI mpi j;tWEx pi j;tXji ;tMji ;tare calculatedi 's imports and exports:Wwherej M Mi ;t(4) X Xi ;t(5)ji ;tji ;tdenotes import ows into the countrydenotes export ows of countryito marketjifrom countryduring periodjduring periodt,andt.2.1.2. DataE ective exchange rates are computed from the bilateral exchange rate with the USdollar of each country. The primary data source for bilateral exchange rates is the WorldBank's World Development Indicators (WDI). Other alternative sources such as nationalcentral bank statistics or the Penn World tables (9.0)12 have also been used in the caseof incomplete data (see Table A.2.1 in Appendix A.2).The price series used to de ate the bilateral nominal exchange rates are the most commonly used price series, i.e.Consumer Price Index (CPI). Although there are theo-retical reasons to prefer other types of price indexes when measuring competitiveness(Rosensweig, 1987), CPIs have the advantage of being timely and available for a widearray of countries over a long time period (see Table A.2.2 in Appendix A.2).2.1.3. Empirical comparisons for robustness checksAs e ective exchange rates can be calculated by using di erent weighting schemes,we report the correlations between the nominal (resp.real) e ective exchange ratesof our database and those provided by other databases in Table A.3.1 (resp.in Appendix A.3.A.3.2)Table A.3.1 shows that in most cases, there is a strong correlationbetween the di erent databases' nominal e ective exchange rates (NEERs). In contrast,the correlations for real e ective exchange rates (REERs) are weaker but, compared tothe Bruegel's REERs, the CEPII's REERs present higher correlations with those providedby the IMF and the WB (Table A.3.2).However, the statistical reliability of thosePWT 9.0 can be downloaded via the University of Groningen webpage athttp://www.rug.nl/ggdc/productivity/pwt/1211

CEPII Working PaperEQCHANGEcorrelations is limited, due to the relatively small number of common observations. Inorder to get a proper/meaningful comparison of the di erent series and for illustrativepurposes, we have reported in Figures C.1 and C.2 in Appendix C the evolution ofthe NEERs and REERs provided by the di erent datasets.As shown, we validate ourcalculations as the dynamics of our exchange rates indices quite follow those issued fromthe di erent databases.2.2. The database on equilibrium real exchange rates and currency misalignmentsIn this section we outline the construction of this second database and also discuss theunderlying methodology and data sources. The dataset computes equilibrium real e ective exchange rates and currency misalignments from 1973 to 2016, for 182 economies,i.e. the largest coverage with available data (Figure 2).13Figure 2 The equilibrium real exchange rates dataset: country coverageNote: the di erent shades of grey indicate the temporal coverage of the series: the darker, the longer is the temporalcoverage.This database has three distinct characteristics. First, it is the rst database on equilibrium real e ective exchange rates and currency misalignments with a global coverage.Second, we rely on the BEER approach rather than the purchasing power parity (PPP)approach. Our measures are then less restrictive than those derived from the PPP only(as the Big Mac Index for example). Moreover, BEER estimates are a good alternative toFor few economies such as countries of the former Soviet Union, series of equilibrium exchange ratesstart later, due to data availability.1312

CEPII Working PaperEQCHANGEFEER assessments such as those provided by the Peterson Institute for InternationalEconomics as the underpinning methodologythe internal and external balances, andnet foreign asset position.(i ) does not require assumptions about(ii ) takes into account stock e ects through theThird, we compute various series of equilibrium exchangerates and currency misalignments along a large set of assumptions used for the equilibrium exchange rate model, for the measure of REER, and for the coverage of partners'countries.2.2.1. Equilibrium exchange ratesThe BEER approach.We rely on the BEER approach to assess equilibrium real ex-change rates (see Clark and MacDonald, 1998).Indeed, one of the di culties whencomputing equilibrium exchange rates is to identify the long-run equilibrium path of theeconomy. We have opted for the BEER approach which is more pragmatic, as it doesnot require to estimate or to make assumptions on the long-run values of the economicfundamentals (such as current account norms for instance) as in the macroeconomic14balance approach.Instead, the BEER approach consists in assessing directly the equi-librium level of real exchange rates through the estimation of a long-run relationshipbetween real exchange rates and their fundamentals.simple and easy to implement and requires few data.This approach is then relativelyThis point is crucial given thepaucity of reliable data in several developing and emerging economies.Analyses di er in the choice of real exchange rates' fundamentals, in part because ofcountries' sample considerations. Edwards (1988), Elbadawi (1994), Hinkle and Montiel(1999) and Elbadawi and Soto (2008) among others have provided suitable theoretical frameworks that identify potential drivers of real exchange rates in developing andemerging countries. In particular, the major fundamentals behind long-run movementsof the real exchange rate identi ed for these countries are usually the terms of trade, therelative productivity of the tradable sector and the net foreign asset position, as we willdetail below.15For the advanced economies, Faruqee (1995), followed by Alberola et al.We do not postulate that the BEER methodology achieves superior performance against other equilibriumexchange rate approaches. Indeed, all these approaches, far from being opposed to each other, are rathercomplementary insofar as they assess equilibrium exchange rates over di erent time horizons (BénassyQuéré et al., 2010).For the sake of completeness and as a robustness check, we have also included in our analysis othervariables considered to be of importance to the exchange rate determination for developing countriessuch as government spending, remittances, nancial development, aid ows, and openness. However,employing such variables dramatically reduce the size of the sample (both in terms of temporal andindividual dimensions). Furthermore, the inclusion of these additional variables did not change the results141513

CEPII Working PaperEQCHANGE(1999) and Alberola (2003), have proposed a stock- ow model for the exchange ratedetermination in the long run, based on the balance-of-payments approach which emphasizes the accumulation of external disequilibria and the Balassa-Samuelson hypothesiswhich underscores relative productivity trends (Balassa, 1964; Samuelson, 1964).The empirical framework.In line with this literature, the three following variables areincluded in our regressions as fundamentals of the real exchange rate:(i)a measure ofsectoral productivity relative to trading partners to account for the Balassa-Samuelsone ect,(ii)the economy's net foreign asset position, and(iii)the economy's terms oftrade.However, rather than pooling directly together in the same equation these threefundamentals, we adopt a sequential approach by starting with the most parsimoniousregression model including only one variable and then by adding successively the otherfundamentals in the regression model.Then, the rst model (Model 1) we considerincludes only a measure of relative productivity and is drawn on the ndings that theBalassa-Samuelson e ect (BS, thereafter) is the most important economic force shapingreal exchange rate behavior. Equilibrium exchange rates derived from this rst model canbe interpreted as PPP-based equilibrium exchange rates adjusted by the BS e ect. Wealso estimate two additional models which augment Model 1 by addingasset position (Model 2), and(i ) the net foreign(ii ) the net foreign position and the terms of trade (Model3):reeri ;treeri ;treeri ;t i1 BS " iBSi ;t i1i ;t BSi ;t2nf ai ;t2 nf ai ;t3(Model 1)i ;ttoti ;t "i ;t "i ;t(Model 2)(Model 3)reer denotes the real e ective exchange rate expressed in logarithm; BS ,nf a and tot stand respectively for the BS e ect proxy (expressed in logarithm), thewherei ;ti ;ti ;ti ;tnet foreign asset position (in percentage of GDP) and the terms of trade (expressed inlogarithm). irepresents country- xed e ects and"distributed error term.in any substantial way. The results are available upon request.14i ;tis an independent and identically

CEPII Working PaperEQCHANGEThe improvement of these three fundamentals is expected to appreciate the equilibriumreal exchange rate. First, the Balassa-Samuelson e ect describes the convergence process of an economy which results in an appreciation of its real exchange rate through alarger productivity growth in the domestic traded goods.16If the real appreciation is inline with the stage of development, the currency of the country will then not necessarilybe overvalued.Second, the connection between real exchange rates and net foreign assets derives fromth

e ective exchange rates and (ii) estimates of equilibrium real exchange rates and of currency misalignments can be freely downloaded. The rst sub-database, on e ective exchange rates, provides indices of nominal and real e ective exchange rates for 187 economies over the 1973-2016 period. The ef-