## wpiea2020126-print-pdf

## Source details

**Canonical URL:** [wpiea2020126-print-pdf](https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020126-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2020/english/wpiea2020126-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2020/english/wpiea2020126-print-pdf.pdf.json)

---

### Data collection process
- Information on the invoicing currency in trade is generally recorded and compiled by national customs/revenue authorities and may be disseminated to national statistics offices and central banks.
- Three–pronged data collection approach:
  - For EU countries, reliance on the annual data collection exercise for the ECB’s annual report on the International Role of the Euro (IRE) and non–public ECB archives.
  - For non-EU countries, online searches for publicly available information on trade invoicing currency.
  - For non-EU countries without posted data (the majority), formal requests to national authorities for time-series data on trade invoicing currency shares.
- Contacts and requests:
  - Governors’ offices and senior officials in statistics, payments and international departments of central banks across Europe, the Western Hemisphere, Asia, Africa, and the Middle East were contacted with formal requests for time-series data on their countries’ trade invoicing currency shares.
  - Central banks were contacted first; if no response or no data, ministries of finance, statistics offices, and customs/revenue authorities were then contacted.
  - In many cases authorities were asked to compile invoicing currency data when not readily available, if underlying data were present.
- Outreach scope and yield:
  - National authorities of some 120 countries were contacted between July 2019 and June 2020 with requests for information on their trade’s currency of invoicing.
  - The data for more than half of the countries in the data set are obtained through such requests.

### Data properties and alternative sources
- General data properties and definitions:
  - Ideal data consistency dimensions (often lacking): definitions over time; coverage of currencies, goods, and trading partners; distinction between invoicing versus payment/settlement currency; scope of coverage (universe of trade vs survey subset); aggregation methods; treatment of re-exports.
  - No common international standard ensures harmonised reporting across these dimensions; data can differ across countries and over time.
  - For countries lacking invoicing currency information, information on payment/settlement currency is used following previous research.
  - Country-specific features and deviations are documented where available (referenced as Table A.1).
  - A comparison of invoicing and payments/settlement currency is available for the three countries with both series (referenced as Figure B.1).
- Potential alternative data sources:
  - SWIFT:
    - SWIFT enables financial institutions to send and receive messages on financial transactions in a secure, harmonised manner but neither clears nor settles payments.
    - SWIFT messages are payment orders settled via correspondent accounts; main limitation for trade invoicing research is the difficulty of distinguishing payment orders related to trade from other transactions.
    - Further work is needed to assess how much of the universe of trade transactions SWIFT data capture.
  - Trade finance data:
    - Trade finance data availability is fragmentary and use of trade finance is heterogeneous across countries, potentially introducing selection bias.
    - It is unclear whether the currency in which trade finance is denominated coincides with trade invoicing currency.
    - Literature: optimal trade finance design depends on exporters’ and importers’ relative costs of contract enforcement and financing; different enforcement environments imply different preferred payment modalities (cash in advance vs open accounts vs letters of credit).
  - Methodological caution example:
    - Bahaj and Reis (2020) focus on SWIFT message types MT 103 and MT 202 but do not incorporate MT 400 in their baseline; MT 400 messages are not present in every international trade payment and MT 400 complements MT 202/MT 103 recording.

### EU-specific considerations and handling of vintages
- EU data are often available for three trading partner compositions: trade with the rest of the world (A1), with non-EU countries (V2), and with non-EA countries (J6).
- Preferred variable for cross-country comparability: invoicing currency shares in countries’ trade with the rest of the world, but this series is available for only a few EU countries for the entire period of interest.
- Common past practice combined or adjusted time series across these compositions to maximize coverage; this can produce spurious level shifts when source or composition changes.
- Spain illustrative findings:
  - Spain’s share of exports to the rest of the world invoiced in US dollars averaged 21% over 2009–2013.
  - Using exports to non-EU (non-EA) countries yields shares of only 14% (16.6%) — demonstrating quantitatively significant discrepancies.
  - To extend the limited “rest of the world” series (available only 2009–2013 for Spain), the authors compute percentage-point changes in the share of exports to non-EA countries invoiced in US dollars for 1999–2008 and the shares to non-EU countries for 2014–2018, then apply those changes backward and forward respectively to span 1999–2018.
- Hungary illustrative findings:
  - Two series for Hungary’s share of exports to the rest of the world invoiced in US dollars were provided:
    - a series provided to the ECB by national authorities before the 2020 data collection exercise (based on customs records; coincides with the other series prior to 2003),
    - a series provided during the 2020 exercise (based on surveys; available from 2008 onward).
  - Communication indicated post-2008 data in the 2020 collection are based on surveys rather than customs records (plausible because Hungary joined the EU in May 2004 and intra-EU trade is exempt from customs, eliminating invoicing currency recording).
  - Authors apply yearly changes from extra-EU trade series to extend forward the rest-of-the-world series from customs data when necessary.
- Other EU data handling notes:
  - Time coverage can differ across “vintages”; national authorities sometimes provide only the most recent data collection period.
  - Authors screen internal (non–publicly available) ECB archives and combine recent data with earlier editions of the ECB’s International Role of the Euro records to maximize time-series coverage while attempting to preserve consistency.
  - Eurostat may report data for years when the ECB receives no input; Eurostat data typically pertain to invoicing in goods trade, whereas ECB data can represent payment/settlement currency in goods and services trade.
- Cross-checking:
  - Institutional contacts between the ECB Statistics Department and national central banks were used to cross-check combined time series across vintages and trading partner compositions.
- Conclusion on EU data:
  - Because of the specific handling of vintages, trading-partner compositions, surveys vs customs records, and cross-checking with internal ECB archives, the authors believe their data set reflects invoicing currency patterns in trade with the rest of the world more accurately and more comprehensively than earlier data sets constructed by Goldberg and Tille (2008), Ito and Chinn (2014), and Gopinath (2015).

### Country and time-series coverage
- Dataset scope and structure:
  - Unbalanced panel on invoicing patterns in goods trade of 102 countries from 1990 to 2019.
  - Data record is based on customs/revenue authorities and in general reflects invoicing currency in goods trade; when using payment/settlement data, services trade may be covered as well.
  - Six of the 102 countries are members of the Western African Monetary Union: Benin, Burkina Faso, Guinea-Bissau, Mali, Niger and Togo — for these countries only aggregate currency union–wide data are available (which include intra-currency union trade). For Senegal and Cote d’Ivoire country-specific data are available (also including trade with the rest of the currency union).
  - Data set provides information on the share of trade invoiced primarily in dollars and in euros, and in many cases also on the share invoiced in home currency; information on the share invoiced in other currencies is fairly scattered and excluded from the main data set but is available upon request.
- Coverage limitations and notable country absences:
  - China: authorities do not participate in the data collection exercise; limited information on the share of the renminbi used as settlement currency in goods trade is available from previous data sets.
  - Canada: no longer stores the detailed invoicing currency information used in Devereux et al. (2017); dataset includes one observation for 2001 (obtained from Kamps (2006)).
  - Mexico: disclosure of the currency of invoicing is not mandatory in customs declaration forms; customs authorities do not collect invoicing currency data or do not collect it with sufficient accuracy in some countries.
  - Other missing countries explicitly noted: Singapore, Nigeria, Vietnam. Central America and Sub-Saharan Africa have relatively limited country coverage.
  - Additional data received after the publication cutoff date include extra years for Austria, Georgia, Guyana, Italy, Macao, Malawi, Ghana, Paraguay, and Senegal, and additional countries: Democratic Republic of Congo, Mozambique, the Philippines, and Uganda (to be incorporated in future updates).
- Time-series coverage and balancing approach:
  - Maximum country coverage for data on countries’ exports invoiced in US dollars in any given year is 92 (in 2018).
  - Maximum number of countries for which data is available on dollar import invoicing shares in a given year is 93; it is 90 and 92, respectively, for euro export and import invoicing shares (all in 2018).
  - Raw data coverage: data set covers more than half of world exports since the early 2000s and about two thirds of world trade after 2010.
  - To construct regional and income-level aggregates, the authors interpolate and extrapolate missing data to obtain a balanced panel:
    - For extrapolation, they use the earliest (latest) available data point to extend backward (forward) and hold constant the value of the first (last) available data point.
    - This simple extrapolation will understate any secular trends in invoicing currency share patterns exhibited by regional or income-level aggregates.
  - After interpolation and extrapolation, the data cover around 75% of global exports.
  - EU countries account for roughly one third of the share of global exports covered by the data (after interpolation/extrapolation).

### Stylised facts — the US dollar, the euro, and evolution of invoicing
- The US dollar’s dominant role in global trade:
  - The share of global exports invoiced in US dollars is much larger than the share of exports destined to the US, indicating an outsized role of the dollar in invoicing global exports; patterns for imports are similar.
  - When commodity exports are removed:
    - Dollar share of invoicing is 23%.
    - Share of exports destined for the US is 10%.
  - Euro’s share in global export invoicing is 46% while its share of exports destined to EA countries is 37%.
  - In IMF Direction of Trade Statistics that include all countries, only 24% of global exports were destined for EA countries in 2019.
  - Reasons for large euro export shares noted:
    - Inclusion of intra-EA exports in the data; a large share of intra-EA trade is invoiced in euros.
    - Better coverage for Europe (beyond EA) than for other regions, and large shares of intra-EU/European trade invoiced in euros.
  - Regional patterns:
    - The dollar is globally dominant in trade invoicing.
    - The euro is regionally dominant in Europe and in some parts of Africa; non-EA European countries and several African countries use the euro for invoicing more than just in their exports to the EA.
- Evolution of global invoicing currency patterns:
  - Increasing concentration of invoicing in US dollars and euros over time, occurring despite declining shares of world exports to the US and the EA — implying that vehicle currency use has been on the rise.
  - Regional and group averages (based on interpolated/extrapolated balanced panel and export-weighted averages):
    - US dollar export invoicing shares vary little over time overall.
    - Slight increases in dollar invoicing for Asian emerging market economies; slight decreases for Latin American emerging market economies.
    - Euro invoicing shares are stable overall except for a sharp increase for non-EA EU countries.
    - Extrapolation approach likely understates secular trends.
  - Country-level dynamics (comparing pre-2005 and post-2016 averages):
    - Number of countries whose share of exports invoiced in dollars has decreased: 58.
    - Number of countries whose share of exports invoiced in dollars has increased: 36.
    - Number of countries whose share of exports invoiced in euros has increased: 65.
    - Number of countries whose share of exports invoiced in euros has decreased: 26.
    - Interpretation: dollar declines occurred more often but may be concentrated in countries with smaller trade shares; euro increases are concentrated in European and several African countries.
  - Notable rapid shifts:
    - Selected European countries show striking increases in their export shares invoiced in euros over relatively short periods, typically paralleled by declines in the share invoiced in US dollars.
    - Rise in euro prominence in the EA’s immediate neighbourhood is consistent with models of multiple equilibria and “thickness externalities”.
    - Strategic complementarities in price setting imply that the exporting country’s market share is a crucial factor in invoicing currency choice.

### Exchange rate pass-through — empirical strategy and key results
- Empirical strategy (import price pass-through main regression):
  - ∆pij,t = αij + δt + Σk=0^K βk ∆eij,t−k + θ′Zi,t + εij,t with K = 2.
  - ∆pij,t is log change in importer-currency import price index from i to j.
  - ∆eij,t is log change in bilateral nominal exchange rate (local currency per units of foreign currency; positive = depreciation).
  - Controls Z include exporter PPI (and two lags); regressions include dyad fixed effects and time fixed effects.
  - Extensions include importer’s exchange rate with respect to the US dollar (∆e$jt) and interactions with importer’s dollar invoicing share S$j.
  - Time-invariant invoicing shares constructed as simple averages across available years for robustness.
  - Estimation uses both unweighted and trade-weighted regressions; weights = average share of global non-commodities trade attributable to exports from i to j.
- Pass-through results (import prices):
  - Standard specification (Equation (1)):
    - A 1% depreciation in country j’s nominal bilateral exchange rate with respect to country i is associated with an 0.7% increase in country j’s import prices within the same year.
  - When the US dollar exchange rate is included (Equation (2)):
    - The bilateral exchange rate effect declines sharply to 0.2% (unweighted) and 0.3% (weighted).
    - A 1% depreciation in the importer’s currency vis-à-vis the US dollar is associated with a price increase of 0.8% (unweighted) and 0.6% (weighted).
  - Interactions with dollar invoicing share (Equation (3)):
    - A 1-p.p. increase in the dollar invoicing share leads to an 0.3–0.6-p.p. increase in the dollar pass-through (range depending on weighted vs. unweighted results).
    - A higher dollar invoicing share reduces the bilateral exchange rate pass-through.
  - Role of the US dollar versus the euro:
    - Including the euro reduces the bilateral exchange rate coefficient but the euro’s pass-through to prices is much smaller than the dollar’s.
    - When dollar, euro, and bilateral rates are included jointly, the US dollar dominates both the bilateral and the euro exchange rates; the model fit deteriorates markedly when euro exchange rates are included.
    - Larger shares of dollar and/or euro invoicing decrease bilateral pass-through and increase pass-through of the respective currency exchange rates.
    - Critical observation: the dollar dominates the euro and bilateral exchange rates even after controlling for euro and home-currency invoicing shares; possible explanation is the dollar’s outsized and growing role in firm financing.
- Selected detailed regression coefficients (Tables A.2–A.4 highlights; subsample of Non-Dollar, Euro or Pegged Economies):
  - Table A.2 — Coefficients on ∆eij,t:
    - Column (1): 0.792***; Column (2): 0.162***; Column (3): 0.281***; Column (4): 0.841***; Column (5): 0.359***; Column (6): 0.401***.
    - Coefficients on ∆e$jt: Column (3): 0.805***; Column (4): 0.651***; Column (5): 0.620***; Column (6): 0.379***.
    - Coefficients on ∆e$jt × S$j: Column (5): 0.241***; Column (6): 0.393***.
    - Observations: Columns (1),(2),(4),(5): 34,304; Columns (3),(6): 28,948.
    - R-squared: 0.382, 0.421, 0.476, 0.352, 0.376, 0.605 for Columns (1)–(6) respectively.
  - Table A.3 — Dollar vs. Euro Pass-Through:
    - Coefficients on ∆e$jt: Column (1): 0.964***; Column (2): 0.642***; Column (3): 0.917***; Column (4): 0.396***.
    - Coefficients on ∆eej,t (euro): Column (1): 0.419***; Column (2): -0.237***; Column (3): -0.284***; Column (5): -0.489**; Column (6): -0.371**.
    - Observations: Columns (1),(2),(4),(5): 25,217; Columns (3),(6): 20,945.
    - R-squared: 0.146, 0.166, 0.207, 0.111, 0.130, 0.303 for Columns (1)–(6) respectively.
  - Table A.4 — Trade Elasticities:
    - Coefficients on ∆yij,t regressions for ∆eij,t: Column (1): -0.115***; Column (2): -0.034; Column (3): -0.025; Column (5): -0.091***.
    - Coefficients on ∆e$jt: Column (1): -0.155***; Column (2): -0.685***; Column (3): -0.257**; Column (5): -0.142***; Column (6): -0.723***.
    - Observations: Columns (1)–(4): 36,353; 36,353; 26,898; 22,348 respectively; Columns (5)–(8): same counts.
    - R-squared: Columns (1)–(8): 0.074, 0.076, 0.073, 0.080, 0.159, 0.165, 0.169, 0.209 respectively.
  - Notes: All table notes report standard errors clustered by dyad, inclusion of lags, fixed effects, and significance notation *** p<0.01, ** p<0.05, * p<0.1.
- Robustness:
  - Results robust to excluding all dollar and euro countries as well as economies with pegged currencies.

### Trade volumes and interactions with invoicing shares
- Trade-volume regressions replace ∆pij,t with ∆yij,t (log change in export volumes from i to j) and add importer’s GDP growth (and two lags) as regressors.
- Main results:
  - A depreciation of country j’s currency against country i’s currency is associated with a decrease in country j’s imports from country i (expected sign).
  - Adding the US dollar exchange rate reduces the bilateral exchange rate coefficient substantially; in weighted regressions it becomes statistically insignificant.
  - When dollar, euro, and bilateral rates are included, the dollar exchange rate dominates the other two; in unweighted regressions the euro has no significant effect on trade volumes.
  - Interaction terms: larger shares of dollar or euro invoicing are associated with larger (in absolute terms) pass-through of the dollar or euro exchange rates to trade volumes.

### Key findings, policy-relevant implications, and data applications
- Key findings:
  - Considerable inertia in global trade invoicing currency patterns alongside evidence that invoicing currency choices can change both radically and rapidly.
  - The introduction of the euro led to more use of the euro as an invoicing currency than the sum of the use of all the currencies it replaced.
  - Rapid and sudden changes in invoicing patterns in some European countries; regionally dominant role of the euro outside Europe in parts of Africa; confirmation of the globally dominant role of the US dollar and overall stability of invoicing patterns.
- Policy-relevant implications:
  - Countries invoicing more in US dollars (euros) experience greater US dollar (euro) exchange rate pass-through to import prices and greater sensitivity of trade volumes to these exchange rates.
  - The special role of the US dollar implies potential spillovers for price-setting and trade following dollar movements and highlights complementarities between trade invoicing and firm financing currency choices.
- Data set applications:
  - Useful for further research on invoicing currencies and effects of exchange rate movements, deepening trade integration, global value chains, international currencies, monetary policy conduct, and international spillovers, particularly in a post-COVID-19 context.

### Appendix notes (A.1 and B.1)
- Table A.1 structure and country coverage:
  - Dataset lists country-level invoicing and settlement data with Country, Code, Range, Type, Source, and Comment fields; examples include precise ranges such as:
    - Algeria DZA 2001-10 invoicing Customs Authority — Exports only for 2003-04; 2001 for euro not available due to lack of legacy currency information from Lafarguette (2015) (2003-2004 Exports; Imports: 2001-2010).
    - Egypt EGY 2010-19 invoicing Central Bank of Egypt.
    - Tunisia TUN 1995-2001, 2010-19 invoicing Banque Centrale de Tunisie — 1995-2001 from Kamps (2006), US dollar data until 2018, euro data until 2019.
    - Liberia LBR 2000-19 invoicing Central Bank of Liberia — Liberian trade invoiced exclusively in US dollars according to Central Bank of Liberia. We assume such practice has been the case since 2000.
    - India IND 1991-2000, 2005, 2008-14 invoicing Reserve Bank of India — Invoicing shares are recorded from June to June, hence not clearly attributable to a single year; 1991-2000, 2005, 2008 from Lafarguette (2015).
    - Timor-Leste TLS 2002-2019 invoicing Banco Central de Timor-Leste — Trade invoiced exclusively in US dollars according to the Banco Central de Timor-Leste.
    - Paraguay PRY 2014-2020 invoicing Customs.
    - Canada CAN 2001 invoicing Murray and Powell (2002) from Kamps (2006), only US dollar exports.
    - United States USA 2003-18 invoicing Bureau of Labour Statistics.
  - Methodological note (verbatim): “A1” refers to trade with the rest of the world, “J6” to trade with non-euro area countries, and “V2” to trade with non-EU countries. Priority: A1, then J6, then V2; adjustments assume euro invoicing shares for intra-EU trade typically 90% for euro area countries and 60% for non-euro area EU countries; when necessary perform “continuation-adjustment”, backpolate and extrapolate based on actual changes.
- Figure B.1:
  - Compares invoicing and settlement data for selected countries (UKR, MDA, BLR) for exports and imports; visual encoding: invoicing data = solid lines, settlement data = dashed lines; US dollar = red, euro = blue. Panels labeled: UKR − Exports; UKR − Imports; MDA − Exports; MDA − Imports; BLR − Exports; BLR − Imports.

*Source: wpiea2020126-print-pdf — https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020126-print-pdf.pdf*

### 2.1    Data collection process

### 2.1–2.4 Data collection process; data properties; alternative sources; EU-specific considerations

### Data collection process
- Information on the invoicing currency in trade is generally recorded and compiled by national customs/revenue authorities and may be disseminated to national statistics offices and central banks.
- A three–pronged data collection approach was used:
  - For EU countries, reliance on the annual data collection exercise for the ECB’s annual report on the International Role of the Euro (IRE) and non–public ECB archives.
  - For non-EU countries, online searches for publicly available information on trade invoicing currency.
  - For non-EU countries without posted data (the majority), formal requests to national authorities for time-series data on trade invoicing currency shares.
- Contacts and requests:
  - Governors’ offices and senior officials in statistics, payments and international departments of central banks across Europe, the Western Hemisphere, Asia, Africa, and the Middle East were contacted with formal requests for time-series data on their countries’ trade invoicing currency shares.
  - Central banks were contacted first; if no response or no data, ministries of finance, statistics offices, and customs/revenue authorities were then contacted.
  - In many cases authorities were asked to compile invoicing currency data when not readily available, if underlying data were present.
- Identifying relevant contact points required leveraging formal and informal contacts at the ECB, the IMF, the European Bank for Reconstruction and Development, the African Development Bank, the African Association of Central Banks, the Asian Development Bank, the Bank for International Settlements, and the South East Asian Central Banks Centre.
- Outreach scope and yield:
  - National authorities of some 120 countries were contacted between July 2019 and June 2020 with requests for information on their trade’s currency of invoicing.
  - The data for more than half of the countries in the data set are obtained through such requests.

### General data properties and definitions
- Ideal data consistency dimensions (often lacking): definitions over time; coverage of currencies, goods, and trading partners; distinction between invoicing versus payment/settlement currency; scope of coverage (universe of trade vs survey subset); aggregation methods; treatment of re-exports.
- There is no common international standard ensuring harmonised reporting of trade invoicing currency data across these dimensions.
- As a result, data can differ across countries and over time. Examples and handling:
  - For countries lacking invoicing currency information, information on payment/settlement currency is used following previous research.
  - Country-specific features and deviations are documented where available (referenced as Table A.1 in Appendix A).
  - A comparison of invoicing and payments/settlement currency is available for the three countries with both series (referenced as Figure B.1 in Appendix B).

### Potential alternative data sources
- SWIFT:
  - SWIFT enables financial institutions to send and receive messages on financial transactions in a secure, harmonised manner but neither clears nor settles payments.
  - SWIFT messages are payment orders settled via correspondent accounts; the main limitation for trade invoicing research is the difficulty of distinguishing payment orders related to trade from other transactions.
  - Further work is needed to assess how much of the universe of trade transactions SWIFT data capture.
- Trade finance data:
  - Trade finance data availability is fragmentary.
  - Use of trade finance is heterogeneous across countries and could introduce selection bias at the country level.
  - It is unclear whether the currency in which trade finance is denominated coincides with trade invoicing currency.
  - Literature insight: optimal trade finance design depends on exporters’ and importers’ relative costs of contract enforcement and financing; different enforcement environments imply different preferred payment modalities (cash in advance vs open accounts vs letters of credit).
- Example methodological caution:
  - Bahaj and Reis (2020) focus on SWIFT message types MT 103 and MT 202 but do not incorporate MT 400 in their baseline; MT 400 messages are not present in every international trade payment and MT 400 complements MT 202/MT 103 recording.

### EU-specific considerations
- EU data are often available for three trading partner compositions:
  - invoicing currency shares in countries’ trade with the rest of the world,
  - with non-EU countries,
  - and with non-EA countries.
- The preferred variable for cross-country comparability is invoicing currency shares in countries’ trade with the rest of the world, but this series is available for only a few EU countries for the entire period of interest.
- Common past practice and its problems:
  - Previous data sets often combined or adjusted time series across these three trading partner compositions to maximize coverage; this can produce spurious level shifts when data source or trading partner composition changes.
- Spain example (illustrative findings):
  - The share of Spain’s exports to the rest of the world invoiced in US dollars averaged 21% over 2009–2013.
  - Assuming all intra-regional trade is invoiced in euros and using data on exports to non-EU (EA) countries yields shares of only 14% (16.6%) — demonstrating quantitatively significant discrepancies.
  - To extend the limited “rest of the world” series (available only 2009–2013 for Spain), the authors compute percentage-point changes in the share of exports to non-EA countries invoiced in US dollars for 1999–2008 and the shares to non-EU countries for 2014–2018, then apply those changes backward and forward respectively to span 1999–2018. This approach implicitly assumes yearly changes in invoicing currency shares are the same across EA vs non-EA and EU vs non-EU partners — an assumption judged weaker than assuming all intra-regional trade is invoiced in euros.
- Hungary example (illustrative findings):
  - Distinct vintages and data-source changes can alter data properties. Two series for Hungary’s share of exports to the rest of the world invoiced in US dollars were provided:
    - a series provided to the ECB by national authorities before the 2020 data collection exercise (based on customs records; coincides with the other series prior to 2003),
    - a series provided during the 2020 exercise (based on surveys; available from 2008 onward).
  - Communication with Hungarian authorities indicated post-2008 data in the 2020 collection exercise are based on surveys rather than customs records (plausible because Hungary joined the EU in May 2004 and intra-EU trade is exempt from customs, eliminating invoicing currency recording).
  - The authors apply yearly changes from extra-EU trade series to extend forward the rest-of-the-world series from customs data when necessary.
- Other EU data issues and handling:
  - Time coverage can differ across “vintages” (new vintages may start at later dates), producing missing earlier-year observations.
  - National authorities sometimes provide only the most recent data collection period.
  - The authors screen internal (non–publicly available) ECB archives and combine recent data with earlier editions of the ECB’s International Role of the Euro records to maximize time-series coverage while attempting to preserve consistency.
  - Eurostat may report data for years when the ECB receives no input because national statistics offices are legally obliged to report to Eurostat while the ECB’s exercise involves national central banks on a voluntary best-efforts basis.
  - Eurostat data typically pertain to invoicing in goods trade, whereas ECB data can represent payment/settlement currency in goods and services trade; this can produce differences (illustrated by Spain: settlement currency series for goods and services was recorded through 2013 and then abandoned, whereas Eurostat provides invoicing-for-goods data for 2010–2018).
- Cross-checking:
  - Institutional contacts between the ECB Statistics Department and national central banks were used to cross-check combined time series across vintages and trading partner compositions.
- Conclusion on EU data:
  - Because of the specific handling of vintages, trading-partner compositions, surveys vs customs records, and cross-checking with internal ECB archives, the authors believe their data set reflects invoicing currency patterns in trade with the rest of the world more accurately and more comprehensively than earlier data sets constructed by Goldberg and Tille (2008), Ito and Chinn (2014), and Gopinath (2015).

*Source: wpiea2020126-print-pdf — https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020126-print-pdf.pdf*

### 2.5    Country and time-series coverage

### 2.5 Country and time-series coverage

### Dataset scope and structure
- Unbalanced panel on invoicing patterns in goods trade of 102 countries from 1990 to 2019.
- Data record is based on customs/revenue authorities and in general reflects invoicing currency in goods trade; when using payment/settlement data, services trade may be covered as well.
- Six of the 102 countries are members of the Western African Monetary Union: Benin, Burkina Faso, Guinea-Bissau, Mali, Niger and Togo. For these countries only aggregate currency union–wide data are available (which include intra-currency union trade). For Senegal and Cote d’Ivoire country-specific data are available (also including trade with the rest of the currency union).
- The data set provides information on the share of trade invoiced primarily in dollars and in euros, and in many cases also on the share invoiced in home currency. Information on the share invoiced in other currencies is fairly scattered and excluded from the main data set but is available upon request.

### Coverage limitations and notable country absences
- Several countries are not included for specific reasons:
  - China: authorities do not participate in the data collection exercise; limited information on the share of the renminbi used as settlement currency in goods trade is available from previous data sets.
  - Canada: no longer stores the detailed invoicing currency information used in Devereux et al. (2017); dataset includes one observation for 2001 (obtained from Kamps (2006)).
  - Mexico: disclosure of the currency of invoicing is not mandatory in customs declaration forms; customs authorities do not collect invoicing currency data or do not collect it with sufficient accuracy in some countries.
- Other missing countries explicitly noted: Singapore, Nigeria, Vietnam. Central America and Sub-Saharan Africa have relatively limited country coverage.
- Additional data received after the publication cutoff date include extra years for Austria, Georgia, Guyana, Italy, Macao, Malawi, Ghana, Paraguay, and Senegal, and additional countries: Democratic Republic of Congo, Mozambique, the Philippines, and Uganda (to be incorporated in future updates).

### Time-series coverage and balancing approach
- Maximum country coverage for data on countries’ exports invoiced in US dollars in any given year is 92 (in 2018).
- The maximum number of countries for which data is available on dollar import invoicing shares in a given year is 93; it is 90 and 92, respectively, for euro export and import invoicing shares (all in 2018).
- Raw data coverage (dashed line in Figure 3 left panel): data set covers more than half of world exports since the early 2000s and about two thirds of world trade after 2010.
- To construct regional and income-level aggregates, the authors interpolate and extrapolate missing data to obtain a balanced panel:
  - For extrapolation, they use the earliest (latest) available data point to extend backward (forward) and hold constant the value of the first (last) available data point.
  - This simple extrapolation will understate any secular trends in invoicing currency share patterns exhibited by regional or income-level aggregates.
- After interpolation and extrapolation, the data cover around 75% of global exports.
- EU countries account for roughly one third of the share of global exports covered by the data (after interpolation/extrapolation).

### 2.6 Stylised facts

### 2.6.1 The US dollar’s dominant role in global trade
- The share of global exports invoiced in US dollars is much larger than the share of exports destined to the US, indicating an outsized role of the dollar in invoicing global exports; patterns for imports are similar.
- When commodity exports are removed:
  - Dollar share of invoicing is 23%.
  - Share of exports destined for the US is 10%.
- Euro’s share in global export invoicing is 46% while its share of exports destined to EA countries is 37%.
  - For comparison, in IMF Direction of Trade Statistics that include all countries, only 24% of global exports were destined for EA countries in 2019.
- Reasons for large euro export shares noted:
  - Inclusion of intra-EA exports in the data; a large share of intra-EA trade is invoiced in euros.
  - Better coverage for Europe (beyond EA) than for other regions, and large shares of intra-EU/European trade invoiced in euros.
- Regional patterns:
  - The dollar is globally dominant in trade invoicing.
  - The euro is regionally dominant in Europe and in some parts of Africa; non-EA European countries and several African countries use the euro for invoicing more than just in their exports to the EA.

### 2.6.2 Evolution of global invoicing currency patterns
- Global evolution:
  - Increasing concentration of invoicing in US dollars and euros over time, occurring despite declining shares of world exports to the US and the EA — implying that vehicle currency use has been on the rise.
- Regional and group averages (based on interpolated/extrapolated balanced panel and export-weighted averages):
  - US dollar export invoicing shares vary little over time overall.
  - Slight increases in dollar invoicing for Asian emerging market economies; slight decreases for Latin American emerging market economies.
  - Euro invoicing shares are stable overall except for a sharp increase for non-EA EU countries.
  - Extrapolation approach likely understates secular trends.
- Country-level dynamics:
  - Comparing pre-2005 and post-2016 averages:
    - Number of countries whose share of exports invoiced in dollars has decreased: 58.
    - Number of countries whose share of exports invoiced in dollars has increased: 36.
    - Number of countries whose share of exports invoiced in euros has increased: 65.
    - Number of countries whose share of exports invoiced in euros has decreased: 26.
  - These counts suggest dollar declines occurred more often but may be concentrated in countries with smaller trade shares; euro increases are concentrated in European and several African countries.
- Notable rapid shifts:
  - Selected European countries show striking increases in their export shares invoiced in euros over relatively short periods, typically paralleled by declines in the share invoiced in US dollars.
  - The rise in euro prominence in the EA’s immediate neighbourhood is consistent with models of multiple equilibria and “thickness externalities” (creation of a monetary union can induce one-off increases in network advantages and reductions in menu costs that favor greater use of the euro).
  - Strategic complementarities in price setting imply that the exporting country’s market share is a crucial factor in invoicing currency choice.

*Source: wpiea2020126-print-pdf (Sections 2.5–2.6).*

### introduction of the euro led to more use of the euro as an invoicing currency than the

### wpiea2020126-print-pdf - introduction of the euro led to more use of the euro as an invoicing currency than the

### Key findings on invoicing currency patterns
- Considerable inertia in global trade invoicing currency patterns alongside evidence that invoicing currency choices can change both radically and rapidly.
- Evidence consistent with theoretical predictions emphasizing nonlinearities, path dependence, and history (e.g., currency unions or episodes of deep institutional integration).
- The introduction of the euro led to more use of the euro as an invoicing currency than the sum of the use of all the currencies it replaced.
- The data set:
  - covers more than 100 countries (invoicing data).
  - trade price and volume data are restricted: the trade data set used for estimations includes 56 countries, and there are dollar and euro invoicing shares for 48 of them.
  - the extension of the price-volume data to include 2016–2018 generates more than 10,000 additional dyad-year observations, which is equivalent to nearly a 25% increase in the sample size.

### Exchange rate pass-through: empirical strategy
- Objective: relate fluctuations in trade prices and volumes to fluctuations in exchange rates and to invoicing shares.
- Main regression (import price pass-through, dynamic lag specification):
  - ∆pij,t = αij + δt + Σk=0^K βk ∆eij,t−k + θ′Zi,t + εij,t with K = 2 (two-year lag).
  - ∆pij,t is log change in importer-currency import price index from i to j.
  - ∆eij,t is log change in bilateral nominal exchange rate (local currency per units of foreign currency; positive = depreciation).
  - Controls Z include exporter PPI (and two lags).
  - Regressions include dyad fixed effects and time fixed effects.
- Extensions:
  - Include importer’s exchange rate with respect to the US dollar (e$j) and interactions with importer’s dollar invoicing share S$j.
  - Time-invariant invoicing shares constructed as simple averages across available years to address data gaps, mechanical variation from exchange rate/composition changes, and overall stability of shares.
- Estimation uses both unweighted and trade-weighted regressions; weights = average share of global non-commodities trade attributable to exports from i to j.

### Pass-through results (import prices)
- Standard specification (Equation (1)):
  - A 1% depreciation in country j’s nominal bilateral exchange rate with respect to country i is associated with an 0.7% increase in country j’s import prices within the same year.
- When the US dollar exchange rate is included (Equation (2)):
  - The bilateral exchange rate effect declines sharply to 0.2% (unweighted) and 0.3% (weighted).
  - A 1% depreciation in the importer’s currency vis-à-vis the US dollar is associated with a price increase of 0.8% (unweighted) and 0.6% (weighted).
- Interactions with dollar invoicing share (Equation (3)):
  - A 1-p.p. increase in the dollar invoicing share leads to an 0.3–0.6-p.p. increase in the dollar pass-through (range depending on weighted vs. unweighted results).
  - A higher dollar invoicing share reduces the bilateral exchange rate pass-through.
  - Including these interactions improves the fit of trade-weighted regressions.
- General note: lagged exchange rate coefficients are statistically significant but economically small.

### Role of the US dollar versus the euro
- Adding the euro exchange rate to the regressions:
  - Including the euro reduces the bilateral exchange rate coefficient but the euro’s pass-through to prices is much smaller than the dollar’s.
  - In weighted regressions, the bilateral exchange rate can dominate the euro.
- When dollar, euro, and bilateral rates are included jointly:
  - The US dollar dominates both the bilateral and the euro exchange rates.
  - The model fit deteriorates markedly when euro exchange rates are included.
- Interaction results:
  - Larger shares of dollar and/or euro invoicing decrease bilateral pass-through and increase pass-through of the respective currency exchange rates.
- Critical observation:
  - The dollar dominates the euro and bilateral exchange rates even after controlling for euro and home-currency invoicing shares.
  - Possible explanation: the dollar’s outsized and growing role in firm financing may create complementarity between currency choice for trade and for finance.
  - Implication: the US dollar plays a special role in the international price system that is not fully explained by its role in invoicing alone.

### Trade volumes
- Trade-volume regressions replace ∆pij,t with ∆yij,t (log change in export volumes from i to j) and add importer’s GDP growth (and two lags) as regressors.
- Main results:
  - A depreciation of country j’s currency against country i’s currency is associated with a decrease in country j’s imports from country i (expected sign).
  - Adding the US dollar exchange rate reduces the bilateral exchange rate coefficient substantially; in weighted regressions it becomes statistically insignificant.
  - When dollar, euro, and bilateral rates are included, the dollar exchange rate dominates the other two; in unweighted regressions the euro has no significant effect on trade volumes.
  - Interaction terms: larger shares of dollar or euro invoicing are associated with larger (in absolute terms) pass-through of the dollar or euro exchange rates to trade volumes.
- Robustness: results are robust to excluding all dollar and euro countries as well as economies with pegged currencies.

### Conclusion and implications
- The paper provides a comprehensive panel on invoicing currency patterns and establishes novel stylised facts:
  - Rapid and sudden changes in invoicing patterns in some European countries.
  - Regionally dominant role of the euro outside Europe in parts of Africa.
  - Confirmation of the globally dominant role of the US dollar in invoicing and overall stability of invoicing patterns.
- Policy-relevant implications illustrated via pass-through and trade-volume exercises:
  - Countries invoicing more in US dollars (euros) experience greater US dollar (euro) exchange rate pass-through to import prices and greater sensitivity of trade volumes to these exchange rates.
  - The special role of the US dollar implies potential spillovers for price-setting and trade following dollar movements and highlights complementarities between trade invoicing and firm financing currency choices.
- Data set applications:
  - Useful for further research on the relationship between invoicing currencies and effects of exchange rate movements, deepening trade integration, global value chains, international currencies, monetary policy conduct, and international spillovers, particularly in a post-COVID-19 context.

*Source: https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020126-print-pdf.pdf*

### References

### References

### Key cited works
- Adler, G., Casas, C., Cubeddu, L., Gopinath, G., Li, N., Meleshchuk, S., Osorio-Buitron, C., Puy, D., Timmer, Y., 2020. Dominant currencies and external adjustment. IMF Staff Discussion Note.
- Bacchetta, P., van Wincoop, E., 2005. A Theory of the Currency Denomination of International Trade. Journal of International Economics 67, 295–319.
- Bahaj, S., Reis, R., 2020. Jumpstarting an International Currency. CEPR Discussion Paper 14793.
- Burstein, A., Gopinath, G., 2014. International Prices and Exchange Rates. Handbook of International Economics 4, 391–451.
- Committee on the Global Financial System, 2014. Trade Finance: Developments and Issues. Bank for International Settlements.
- Devereux, M., Dong, W., Tomlin, B., 2017. Importers and Exporters in Exchange Rate Pass-through and Currency Invoicing. Journal of International Economics 105, 187–204.
- Devereux, M., Shi, S., 2013. Vehicle Currency. International Economic Review 54, 97–133.
- European Central Bank, 2019. The International Role of the Euro. European Central Bank, Frankfurt.
- Goldberg, L., Tille, C., 2008. Vehicle-currency Use in International Trade. Journal of International Economics 76, 177–192.
- Gopinath, G., 2015. The International Price System. NBER Working Paper 21646.
- Gopinath, G., Casas, C., Diez, F., Gourinchas, P.O., Plagborg-Moller, M., 2020. Dominant Currency Paradigm. American Economic Review 110, 677–719.
- Gopinath, G., Itskhoki, O., Rigobon, R., 2010. Currency Choice and Exchange Rate Pass-Through. American Economic Review 100, 304–336.
- Gopinath, G., Stein, J., 2018. Banking, Trade, and the Making of a Dominant Currency. NBER Working Paper 24485.
- Ito, H., Chinn, M., 2014. The Rise of the “Redback” and the People’s Republic of China’s Capital Account Liberalization: An Empirical Analysis of the Determinants of Invoicing Currencies. ADBI Working Paper 473.
- Kamps, A., 2006. The Euro as Invoicing Currency in International Trade. ECB Working Paper 665.
- Lafarguette, R., 2015. Update of Kamps (2006). ECB mimeo.
- Mukhin, D., 2018. An Equilibrium Model of the International Price System. mimeo.
- Murray, J., Powell, J., 2002. Dollarization in Canada: The Buck Stops There. Bank of Canada Technical Report 90.
- Portes, R., Rey, H., 1998. The Emergence of the Euro as an International Currency. Economic Policy 13, 305–343.
- Rey, H., 2001. International Trade and Currency Exchange. Review of Economic Studies 68, 443–464.
- Schmidt-Eisenlohr, T., 2013. Towards a Theory of Trade Finance. Journal of International Economics 91, 96–112.

*Additional bibliographic entries and country-specific data sources are listed in the source section.*

---

### A. Invoicing currency data overview (Table A.1)

### Structure and country coverage
- The dataset lists country-level invoicing and settlement data with Country, Code, Range, Type, Source, and Comment fields.
- Examples of country entries with exact ranges and types:
  - Algeria DZA 2001-10 invoicing Customs Authority — Exports only for 2003-04; 2001 for euro not available due to lack of legacy currency information from Lafarguette (2015) (2003-2004 Exports; Imports: 2001-2010).
  - Egypt EGY 2010-19 invoicing Central Bank of Egypt.
  - Tunisia TUN 1995-2001, 2010-19 invoicing Banque Centrale de Tunisie — 1995-2001 from Kamps (2006), US dollar data until 2018, euro data until 2019.
  - Liberia LBR 2000-19 invoicing Central Bank of Liberia — Liberian trade invoiced exclusively in US dollars according to Central Bank of Liberia. We assume such practice has been the case since 2000.
  - India IND 1991-2000, 2005, 2008-14 invoicing Reserve Bank of India — Invoicing shares are recorded from June to June, hence not clearly attributable to a single year; 1991-2000, 2005, 2008 from Lafarguette (2015).
  - Timor-Leste TLS 2002-2019 invoicing Banco Central de Timor-Leste — Trade invoiced exclusively in US dollars according to the Banco Central de Timor-Leste.
  - Paraguay PRY 2014-2020 invoicing Customs.
  - Canada CAN 2001 invoicing Murray and Powell (2002) from Kamps (2006), only US dollar exports.
  - United States USA 2003-18 invoicing Bureau of Labour Statistics.

### Methodological note on series concepts and adjustments (verbatim)
- Note: “A1” refers to trade with the rest of the world, “J6” to trade with non-euro area countries, and “V2” to trade with non-EU countries. When data for more than one concept is available for the same time period, priority is given to the A1 series, followed by the J6 series and lastly the V2 series. In these cases, J6 and V2 series are adjusted to refer to trade with the rest of the world assuming that a certain share of intra-EU and intra-euro area trade is invoiced in euros, typically 90% for euro area countries and 60% for non-euro area EU countries. When data are available for different concepts for different, non-overlapping time periods, we perform “continuation-adjustment”. In particular, we adjust the V2/J6 series by assuming a euro invoicing share for intra-EU trade such that the transition between the time series is smooth. Finally, when data are available for overlapping time periods but also cover different sub-periods we backpolate and extrapolate based on actual changes, again giving priority to A1, J6 and then V2.

---

### A. Exchange rate pass-through and trade elasticities (Tables A.2–A.4)

### Table A.2 — Exchange Rate Pass-Through into Import Prices:
Subsample of Non-Dollar, Euro or Pegged Economies
- Coefficients on ∆eij,t (first row):
  - Column (1): 0.792*** (p-value (0.00))
  - Column (2): 0.162*** (p-value (0.00))
  - Column (3): 0.281*** (p-value (0.00))
  - Column (4): 0.841*** (p-value (0.00))
  - Column (5): 0.359*** (p-value (0.00))
  - Column (6): 0.401*** (p-value (0.00))
- Coefficients on interaction ∆eij,t × S$j:
  - Column (1): -0.173*** (p-value (0.00))
  - Column (2): -0.115* (p-value (0.09))
- Coefficients on ∆e$j,t:
  - Column (3): 0.805*** (p-value (0.00))
  - Column (4): 0.651*** (p-value (0.00))
  - Column (5): 0.620*** (p-value (0.00))
  - Column (6): 0.379*** (p-value (0.00))
- Coefficients on ∆e$j,t × S$j:
  - Column (5): 0.241*** (p-value (0.00))
  - Column (6): 0.393*** (p-value (0.00))
- Observations:
  - Columns (1), (2), (4), (5): 34,304
  - Columns (3), (6): 28,948
- R-squared:
  - Columns (1)–(6): 0.382, 0.421, 0.476, 0.352, 0.376, 0.605 respectively
- Dyads:
  - Columns (1), (2), (4), (5): 1,741
  - Columns (3), (6): 1,449
- Notes (verbatim): The first (last) three columns present unweighted (trade-weighted) regressions. All regressions include two ∆ER and ∆PPI lags, and time and dyad fixed effects. Standard errors clustered by dyad. p-values in parentheses. *** p<0.01, ** p<0.05, *p<0.1.

### Table A.3 — Dollar vs. Euro Exchange Rate Pass-Through into Prices:
Subsample of Non-Dollar, Euro or Pegged Economies
- Coefficients on ∆eij,t (first row):
  - Column (1): 0.369*** (p-value (0.00))
  - Column (2): 0.195*** (p-value (0.00))
  - Column (3): 0.604*** (p-value (0.00))
  - Column (4): 0.587*** (p-value (0.00))
  - Column (5): 0.400*** (p-value (0.00))
  - Column (6): 0.442** (p-value (0.04))
- Coefficients on ∆eij,t × (S$j + Sej):
  - Column (1): -0.466*** (p-value (0.00))
  - Column (2): -0.125 (p-value (0.58))
- Coefficients on ∆e$jt:
  - Column (1): 0.964*** (p-value (0.00))
  - Column (2): 0.642*** (p-value (0.00))
  - Column (3): 0.917*** (p-value (0.00))
  - Column (4): 0.396*** (p-value (0.00))
- Coefficients on ∆e$j,t × S$j:
  - Column (5): 0.452*** (p-value (0.00))
  - Column (6): 0.677*** (p-value (0.00))
- Coefficients on ∆e ej,t:
  - Column (1): 0.419*** (p-value (0.00))
  - Column (2): -0.237*** (p-value (0.00))
  - Column (3): -0.284*** (p-value (0.00))
  - Column (4): 0.091 (p-value (0.53))
  - Column (5): -0.489** (p-value (0.02))
  - Column (6): -0.371** (p-value (0.03))
- Coefficients on ∆e ej,t × Sej:
  - Column (5): 0.584*** (p-value (0.00))
  - Column (6): 0.579** (p-value (0.01))
- Observations:
  - Columns (1), (2), (4), (5): 25,217
  - Columns (3), (6): 20,945
- R-squared:
  - Columns (1)–(6): 0.146, 0.166, 0.207, 0.111, 0.130, 0.303 respectively
- Dyads:
  - Columns (1), (2), (4), (5): 1,689
  - Columns (3), (6): 1,397
- Notes (verbatim): The first (last) three columns present unweighted (trade-weighted) regressions. All regressions include two ∆ER and ∆PPI lags, dyad fixed effects and global controls. Standard errors clustered by dyad. p-values in parentheses. *** p<0.01, ** p<0.05, *p<0.1.

### Table A.4 — Trade Elasticity with respect to Exchange Rates:
Subsample of Non-Dollar, Euro or Pegged Economies
- Coefficients on ∆yij,t regressions:
  - Column (1) ∆eij,t: -0.115*** (p-value (0.00))
  - Column (2) ∆eij,t: -0.034 (p-value (0.13))
  - Column (3) ∆eij,t: -0.025 (p-value (0.64))
  - Column (4) ∆eij,t: -0.117 (p-value (0.60))
  - Column (5) ∆eij,t: -0.091*** (p-value (0.00))
  - Column (6) ∆eij,t: -0.008 (p-value (0.82))
  - Column (7) ∆eij,t: -0.045 (p-value (0.25))
  - Column (8) ∆eij,t: 0.175 (p-value (0.36))
- Coefficients on ∆eij,t × (S$j + Sej):
  - Column (2): 0.103 (p-value (0.71))
  - Column (4): -0.229 (p-value (0.30))
- Coefficients on ∆e$jt:
  - Column (1): -0.155*** (p-value (0.00))
  - Column (2): -0.685*** (p-value (0.00))
  - Column (3): -0.257** (p-value (0.05))
  - Column (5): -0.142*** (p-value (0.00))
  - Column (6): -0.723*** (p-value (0.00))
  - Column (7): -0.226* (p-value (0.08))
- Coefficients on ∆e$j,t × S$j:
  - Column (6): -0.364** (p-value (0.02))
  - Column (7): -0.414** (p-value (0.03))
- Coefficients on ∆e e j,t:
  - Column (1): 0.315*** (p-value (0.00))
  - Column (2): 0.130 (p-value (0.18))
  - Column (3): 0.551*** (p-value (0.00))
  - Column (4): 0.252* (p-value (0.08))
- Coefficients on ∆e e j,t × Sej:
  - Column (7): -0.090 (p-value (0.65))
  - Column (8): -0.474** (p-value (0.01))
- Observations:
  - Columns (1)–(4): 36,353; 36,353; 26,898; 22,348 respectively
  - Columns (5)–(8): 36,353; 36,353; 26,898; 22,348 respectively
- R-squared:
  - Columns (1)–(8): 0.074, 0.076, 0.073, 0.080, 0.159, 0.165, 0.169, 0.209 respectively
- Dyads:
  - Columns (1)–(4): 1,755, 1,755, 1,755, 1,452
  - Columns (5)–(8): 1,755, 1,755, 1,755, 1,452
- Notes (verbatim): The first (last) four columns present unweighted (trade-weighted) regressions. All regressions include two ∆ER and importer’s ∆GDP lags and dyad fixed effects. Columns (1)–(2) and (5)–(6) include time fixed effects, and columns (3)–(4) and (7)–(8) include global controls. Standard errors clustered by dyad. p-values in parentheses. *** p<0.01, ** p<0.05, *p<0.1.

---

### B. Additional figure note

- Figure B.1 compares invoicing and settlement data for selected countries (UKR, MDA, BLR) for exports and imports over the period shown. Visual encoding:
  - Invoicing data are depicted by solid lines; settlement data by dashed lines.
  - Data for the US dollar are depicted in red, while data for the euro are depicted in blue.
- Figure panels (verbatim labels) include:
  - UKR − Exports
  - UKR − Imports
  - MDA − Exports
  - MDA − Imports
  - BLR − Exports
  - BLR − Imports

*Source: wpiea2020126-print-pdf - References (IMF).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020126-print-pdf.pdf_