## wpiea2024242-print-pdf — Sections 2–4 and selected 4.x–4.6 content

## Source details

**Canonical URL:** [wpiea2024242-print-pdf — Sections 2–4 and selected 4.x–4.6 content](https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024242-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2024/english/wpiea2024242-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2024/english/wpiea2024242-print-pdf.pdf.json)

---

### How Does Trade in Services Differ from Trade in Goods?
- Key framework:
  - A firm’s invoicing currency choice is determined by its desirable exchange rate pass-through in a staggered price setting; when prices are sticky, firms choose invoicing currency to keep prices close to desired levels during non-adjustment periods (references to Gopinath et al., 2010; Amiti et al., 2022; Mukhin, 2022).
  - If desired pass-through to producer currency is low → producer-currency pricing; if high → local-currency pricing; middle values → dominant-currency pricing.
- Determinants of desired pass-through:
  - Sensitivity of marginal costs to exchange rates, often linked to intensity of imported intermediate inputs. Exporters tend to invoice in the currency that matches their production costs.
  - Sensitivity of desired markup to exchange rates, linked to strategic complementarities in price setting among firms in the destination market. High strategic complementarities can favor dominant currency pricing (DCP) to maintain relative price stability and market shares.
- Sectoral characteristics that affect currency choice:
  - Input structure (WIOD 2016 cross-country averages, year 2014 context):
    - Average share of intermediate input in gross output: manufacturing 26.7%, services 8.7%.
    - Average share of labor input: manufacturing 27.9%, services 57.5%.
    - Implication: higher domestic input intensity in services → lower marginal-cost sensitivity to exchange rates → greater incentives to price in the producer’s currency.
  - Intangibility and nonstorability (proximity burden):
    - When service delivery requires proximity or joint production, choice of currency depends on rivals’ pricing at location of delivery.
    - If exporters mainly compete with local providers at destination (who price in local currency) → local currency pricing; if trade occurs at exporter’s location and exporter also serves local consumers (e.g., Tourism) → producer-currency pricing.
  - Market structure and entry barriers:
    - Natural and behind-the-border policy barriers and network externalities can lead to high concentration in some services sectors, prompting some exporters to adjust markups and adopt LCP or DCP to stabilize market share.
  - Reliance on external finance:
    - Some services rely less on external finance (examples cited: Indian IT sector; generally few service sectors depend on external finance), and short/no lag between delivery and payment reduces working-capital needs.
    - Lower foreign-currency liabilities → less incentive to price in a foreign currency.

### Data Description
- Measurement and mode coverage:
  - WTO GATS modes: mode 1 (cross-border), mode 2 (consumption abroad), mode 3 (commercial presence), mode 4 (movement of natural persons).
  - Balance of Payments (BoP) Statistics measure services provided by modes 1, 2, and 4 only; mode 3 (commercial presence / foreign affiliate sales) are excluded from the analysis.
- Bilateral Services Trade Database (BiTS):
  - Constructed by consolidating, imputing, and reconciling official data sources; primary source WTO-UNCTAD-ITC Annual Trade in Commercial Services Database, supplemented by UN COMTRADE and Eurostat historical records.
  - Concepts/definitions/classification based on EBOPS 2010 and IMF BPM6.
  - Coverage: spans from 1985 to 2022 and covers 245 countries.
  - Uses exporter-reported services trade flows as the base value; employs partner-reported imports to create “mirror” flows to fill missing export values; performs hierarchical consistency checks.
  - Contains exports and imports across 12 broad commercial service sectors (analysis focuses on aggregated and 1-digit services).
- Complementary data:
  - US sectoral trade flows from US BEA for 1999–2022.
  - World Bank WDI (up to 2022) and IMF Trade in Services Database (1970–2014) for multilateral series.
  - Tourism proxies from Eurostat tourism database (1991–2018).

### Basic Patterns of International Trade in Services
- Coverage and growth (WDI, 2022; BiTS coverage):
  - Total services: 7,209.0 (value in billions of US dollars) — Growth: 7.2% — BiTS coverage: 91%
  - Transport: 1,471.9 — Growth: 7.0% — BiTS coverage: 82%
  - Travel: 1,095.5 — Growth: 6.0% — BiTS coverage: 74%
  - Other commercial services: 4,554.4 — Growth: 9.0% — BiTS coverage: 98%
  - Text summary: “Total services grew by 7% annually, with OCS the fastest-growing sector.”
- Rising importance and composition shifts:
  - Commercial services exports have expanded at about 3 times the growth rate of goods exports since 2011.
  - Services now account for nearly a quarter of global exports in gross terms and for about 40 percent in value-added terms.
  - Services contribution to national export revenue:
    - Services amount to 45% of UK and 33% of US gross exports.
    - In value-added terms, the services share exceeds 50% in France, Germany, Italy, the United Kingdom, and the United States.
    - In China, more than a third of the value added in its exports is due to services.
  - Industry composition changes:
    - Mid-1980s: Transport, Travel, and OCS each about a third of world services exports.
    - By 2019: OCS almost 70% of world services exports; Transport and Travel shares declined.
    - Within OCS, fastest-growing segments: Other business services, Financial services, IT services, Charges for use of intellectual property.
    - Top 5 services industries (Transport, Travel, Finance, IT, Other business services) now account for 80% of global services exports.
- Income-group patterns:
  - Most services exports sourced from high-income countries; middle-income countries’ roles have increased.
  - Advanced economies increasingly specialized in services exports; EMDEs more likely to specialize in manufacturing exports.
- Bilateral concentration and zeros:
  - Exporter–importer pairs in the top 1% of trade flows account for 56% of global trade in services; top 10% contribute about 95% in 2021.
  - About 45% of dyads do not trade in services or contain no information on services trade.
- Drivers of services trade growth:
  - Reduced trading costs from information and communication technologies mitigating the proximity burden.
  - Emergence and expansion of global value chains enabling international specialization in service-related tasks and services traded as components within GVCs.

### Conceptual Framework and Empirical Strategy
- Parsimonious reduced-form import demand:
  - ∆q_ij,t = −σ_ij(∆p_ij,t − ∆p_j,t) + ∆y_jt.
  - σ_ij is a reduced-form Armington elasticity.
- Invoicing currency composition:
  - θ^i_ij (producer currency PCP), θ^j_ij (local currency LCP), θ^$ _ij (dominant currency DCP), with θ^i_ij + θ^j_ij + θ^$ _ij = 1.
  - Imported price inflation (in importer currency): ∆p_ij,t = Σ_{k=i,j,$} θ^k_ij (∆p^k_ij,t + ∆e_{kj,t}).
- Calvo sticky-price reset and short-run implications (δ → 1):
  - Change in log import value: ∆v_ij,t = (1 − σ_ij) θ^i_ij ∆e_{ij,t} + (1 − σ_ij) θ^$ _ij ∆e_{j$,t} − σ_ij ∆p_j,t + ∆y_j,t.
  - Interpretations:
    - PCP (θ^i_ij = 1): import values reflect bilateral exchange rate changes scaled by (1 − σ_ij).
    - LCP (θ^j_ij = 1): import values independent of exchange rate movements.
    - DCP (θ^$ _ij = 1): import values reflect importer currency–USD changes scaled by (1 − σ_ij).
  - Without knowledge of σ_ij, LCP cannot be tested when σ_ij close to 1.
- Estimation equation (log-linear, motivated by model):
  - ∆v_ij,t = Σ_{τ=0}^L α_τ ∆e_{ij,t−τ} + Σ_{τ=0}^L β_τ ∆e_{j$,t−τ} + Γ′ X_{ij,t} + λ_{ij} + δ_t + ε_{ij,t}.
  - Controls X_{ij,t}: log change of importer real GDP and CPI, exporter PPI, and two lags.
  - Short-term elasticities: α_0 (bilateral) and β_0 (dollar); medium-term: Σ_{τ=0}^L α_τ and Σ_{τ=0}^L β_τ.
  - Standard errors clustered at directed dyadic level.

### Exchange Rate Elasticities of Aggregate Services Trade (Key Empirical Findings)
- Sample excludes U.S. as exporter and importer for main regressions; dependent variable = value of imports of country j from country i (destination currency).
- Contemporaneous bilateral exchange rate elasticities (∆e_{ij,t}):
  - Unweighted: 0.436 ∗∗∗ (0.028).
  - Unweighted with dollar included: 0.223 ∗∗∗ (0.039).
  - Weighted: 0.391 ∗∗∗ (0.078).
  - Weighted with dollar included: 0.176 ∗∗∗ (0.056).
- Contemporaneous dollar exchange rate elasticities (∆e_{j$,t}):
  - Unweighted: 0.453 ∗∗∗ (0.057).
  - Weighted: 0.539 ∗∗∗ (0.097).
- Cumulative (Σ_{l=0}^2) bilateral:
  - Unweighted: 0.264 ∗∗∗ (0.046); with dollar included: 0.019 (0.057).
  - Weighted: 0.222 ∗ (0.133); with dollar: 0.016 (0.102).
- Cumulative dollar (Σ_{l=0}^2): unweighted 0.601 ∗∗∗ (0.087); weighted 0.593 ∗∗∗ (0.141).
- Observations: 79,552; Dyads: 6,708; R^2 reported (unweighted 0.17, weighted 0.32–0.33).
- Interpretations:
  - A 10% depreciation of country j’s currency relative to country i increases import value (in j’s currency) by 4.4% (unweighted) or 3.9% (weighted) when dollar excluded.
  - Including the dollar exchange rate reduces bilateral coefficients substantially.
  - Dollar exchange rate effect larger than bilateral: ≈0.45 (unweighted) and ≈0.54 (weighted).
  - Positive and significant α_0 and β_0 → both PCP and DCP present in total services trade.
  - Positive import value response to both bilateral and dollar depreciation suggests price increases dominate volume declines (consistent with σ < 1 on average).

### Comparison to Manufacturing
- Manufacturing contemporaneous bilateral elasticities:
  - Unweighted: 0.312 ∗∗∗ (0.031); with dollar included: 0.082 ∗ (0.043).
  - Weighted: 0.327 ∗∗∗ (0.044); with dollar included: 0.161 ∗∗∗ (0.038).
- Manufacturing contemporaneous dollar elasticities:
  - Unweighted: 0.488 ∗∗∗ (0.061); weighted: 0.438 ∗∗∗ (0.054).
- Notable contrasts:
  - Manufacturing: dollar exchange rate movements have about six times larger effect than bilateral shocks on manufacturing trade (relative α/β).
  - Services: dollar coefficient only about twice as large as bilateral coefficient.
- Implication: DCP (relative to PCP) is more commonly adopted in manufacturing than in services (if σ similar across transaction currencies).

### Exchange Rate Elasticities Across Service Categories (Selected Sectoral Estimates)
- Focus sectors: Transport, Travel, Financial services, IT (Telecommunications/computer/information), Other business services, Intellectual Property (IP).
- Short-term (contemporaneous) estimates (elasticity (standard error)):
  - Transport:
    - ∆e_{ij,t} = 0.443 ∗∗∗ (0.048); cumulative Σ_{l=0}^2 = 0.346 ∗∗∗ (0.059).
    - ∆e_{j$,t} = 0.587 ∗∗∗ (0.090); cumulative Σ_{l=0}^2 = 0.345 ∗∗∗ (0.116).
  - Travel:
    - ∆e_{ij,t} = 0.186 ∗∗∗ (0.055); cumulative = 0.208 ∗∗∗ (0.070).
    - ∆e_{j$,t} = 0.217 ∗∗∗ (0.076); cumulative = 0.282 ∗∗ (0.123).
  - Financial services:
    - ∆e_{ij,t} = 0.304 ∗∗∗ (0.039); cumulative = 0.114 ∗ (0.063).
    - ∆e_{j$,t} = 0.356 ∗∗ (0.148); cumulative = 0.76 ∗∗∗ (0.237).
  - IT:
    - ∆e_{ij,t} = 0.225 ∗∗∗ (0.048); cumulative = 0.017 (0.078).
    - ∆e_{j$,t} = 0.430 ∗∗∗ (0.143); cumulative = −0.033 (0.216).
  - Other business services:
    - ∆e_{ij,t} = 0.387 ∗∗∗ (0.077); cumulative = 0.388 ∗∗∗ (0.119).
    - ∆e_{j$,t} = 0.142 (0.110); cumulative = 0.199 (0.135).
  - Intellectual Property (IP):
    - ∆e_{ij,t} = 0.299 ∗∗∗ (0.111); cumulative = 0.154 (0.161).
    - ∆e_{j$,t} = 0.538 ∗∗ (0.212); cumulative = 0.373 (0.299).
- Sectoral interpretation:
  - USD exchange rate more important than bilateral for Transport, Financial services, IT, and IP in the short term.
  - Travel: bilateral and dollar effects similar.
  - Other business services: dollar exchange rate not statistically significant contemporaneously.
  - Over two years, USD effect strengthens for Travel, Financial services, Other business services; weakens for Transport, IT, and IP.
  - Differences reflect sectoral heterogeneity in price stickiness, reliance on imported inputs, and market concentration.

### Exchange Rates and Tourism: Volume Elasticities (Arrivals and Hotel Nights)
- Quantity proxies used: arrivals of nonresidents at hotels and hotel nights spent (Eurostat tourism database).
- Unweighted contemporaneous and cumulative bilateral elasticities (standard errors):
  - Arrivals:
    - ∆e_{ij,t} = −0.369 ∗∗∗ (0.033); with dollar included ∆e_{ij,t} = −0.270 ∗∗∗ (0.049).
    - ∆e_{j$,t} = −0.132 ∗∗ (0.057) when included.
    - Cumulative Σ_{τ=0}^2 ∆e_{ij,t−τ} = −0.493 ∗∗∗ (0.068).
    - Cumulative Σ_{τ=0}^2 ∆e_{j$,t−τ} = −0.054 (0.095).
  - Hotel nights spent:
    - ∆e_{ij,t} = −0.369 ∗∗∗ (0.037); with dollar included ∆e_{ij,t} = −0.232 ∗∗∗ (0.055).
    - ∆e_{j$,t} = −0.184 ∗∗∗ (0.064) when included.
    - Cumulative Σ_{τ=0}^2 ∆e_{ij,t−τ} = −0.459 ∗∗∗ (0.069).
    - Cumulative Σ_{τ=0}^2 ∆e_{j$,t−τ} = −0.185 (0.109).
- Weighted regressions:
  - Dollar exchange rate loses statistical significance both contemporaneously and over two years.
  - Weighted cumulative results show negligible effect of ∆e_{j$,t} on outbound tourism.
- Quantitative implication:
  - A 10% depreciation of a country’s currency against the tourist-hosting country’s currency reduces, over a two-year horizon, tourist arrivals by 4.5% and hotel nights spent by 3.3% (from cumulative bilateral elasticities).
- Interpretation:
  - Unweighted results show evidence of both PCP and DCP in tourism in the short term; inclusion of dollar reduces bilateral coefficient.
  - Dollar exchange rate has much less negative effect on tourism volumes than bilateral exchange rate.
  - Weighted results suggest DCP observed more in smaller economies, while PCP predominates in larger economies.
  - Traditional Mundell-Fleming expenditure-switching effects of exchange rates on export volume remain important for tourism exports.

### Evidence from US Imports and Exports (Section 4.6)
- Methodology:
  - Fixed-effects regressions of log changes of trade value (denominated in the destination country currency) on lagged bilateral exchange rates, importer GDP, importer CPI, exporter PPI and two lags, year fixed effects and dyadic fixed effects.
  - Estimations run separately for US exports (Panel B) and US imports (Panel A).
- Selected coefficient estimates (standard errors in parentheses):
  - Panel A (Imports to the US), Total Services (Column (1)):
    - ∆e_i$,t = 0.094* (0.056)
    - ∑_{l=0}^2 ∆e_i$,t−l = 0.167* (0.089)
  - Panel A (Imports to the US), Manufacturing (Column (9)):
    - ∆e_i$,t = 0.222 (0.191)
    - ∑_{l=0}^2 ∆e_i$,t−l = 0.081 (0.206)
  - Panel B (Exports from the US), Total Services (Column (1)):
    - ∆e$j,t = 0.652*** (0.034)
    - ∑_{l=0}^2 ∆e$j,t−l = 0.686*** (0.053)
  - Panel B (Exports from the US), Manufacturing (Column (9)):
    - ∆e$j,t = 0.737*** (0.207)
    - ∑_{l=0}^2 ∆e$j,t−l = 0.657*** (0.141)
- Key empirical conclusions:
  - US exports respond strongly to bilateral exchange rate fluctuations; large, positive, and statistically significant contemporaneous and cumulative coefficients.
  - US imports (manufacturing) show no significant effect from bilateral exchange rate fluctuations, consistent with dominant currency pricing (USD invoicing).
  - Sectoral heterogeneity: US imports in Transport, Financial services, IT, and Other business services show strong short-run responses to bilateral shocks (implying producer/firm currency pricing in those sectors).
  - US exports across all services categories respond to foreign currency depreciation relative to the USD in both short and medium term.
  - Positive export coefficients imply that when a foreign currency depreciates against the USD, US exporters (with prices denominated in foreign currency) can experience higher export values because volume reduction is small relative to price increase in foreign-currency terms.
- Cross-cutting note:
  - Qualitatively similar conclusions for manufacturing and total services trade, with important heterogeneity across individual services sectors in invoicing and pass-through.

### Appendix A — The Bilateral Trade in Services Database (BiTS): scope, methods, and coverage
- Purpose: Provide detailed international bilateral trade flows in services using original source data (no estimated or extrapolated data).
- Country and sector coverage:
  - 245 countries and a global total plus a residual region for unallocated flows.
  - 12 broad services sectors and an unallocated category; further divided into 30 distinct subcategories (text also refers to 28 subcategories).
  - Time span: 1985-2022.
  - Values reported in million US dollar.
- Primary and supplementary data sources:
  - Primary source: WTO-UNCTAD-ITD Annual Trade in Services Database (referred to as the WTO database), covering 2005-2022.
  - Supplementary: UN COMTRADE covering 2000-2020; Eurostat historical records covering 1985-2003.
  - BaTIS not included because it contains estimated and balanced flows; objective is to aggregate only original data.
- Processing, reconciliation, and mirroring:
  - Vertical reconciliation from subcategories up to aggregate levels under specified rules.
  - Mirroring used to maximize coverage; exporter-reported flows preferred as base; reported imports used to create mirror flows to fill missing exporter values.
  - Conversion from BPM5 to BPM6 follows IMF Statistics Department conversion matrix; specific reallocations applied and hierarchical checks reapplied.
  - Source precedence single-source rule: WTO data preferred, then COMTRADE, then Eurostat; original flows prioritized over mirror flows; single source per country-pair-year.
  - Residual region XRR to account for unallocated exports; world total rebalanced if necessary.
  - All zero trade flows originate from original data sources and are not imputed.
- Distribution of observations by data source (Table 9: Data source Percent):
  - WTO original 43.43
  - WTO mirrored 26.56
  - COMTRADE original 8.88
  - COMTRADE mirrored 7.44
  - Eurostat original 3.63
  - Eurostat mirrored 3.45
  - Rebalanced World 1.14
  - Residual region 5.48

*Source: wpiea2024242-print-pdf (sections 2–4, 4.1, 4.6, and Appendix A) — IMF PDF content provided in the prompt.*

### Section 5 explores the effects of forward GVC participation. We conclude in Section 6 with a brief

### wpiea2024242-print-pdf — Sections 2–4 (How Does Trade in Services Differ from Trade in Goods?; Data Description; Basic Patterns)

### How Does Trade in Services Differ from Trade in Goods?
- Key framework:
  - A firm’s invoicing currency choice is determined by its desirable exchange rate pass-through in a staggered price setting; when prices are sticky, firms choose invoicing currency to keep prices close to desired levels during non-adjustment periods (references to Gopinath et al., 2010; Amiti et al., 2022; Mukhin, 2022).
  - If desired pass-through to producer currency is low → producer-currency pricing; if high → local-currency pricing; middle values → dominant-currency pricing.
- Determinants of desired pass-through:
  - Sensitivity of marginal costs to exchange rates, often linked to intensity of imported intermediate inputs. Exporters tend to invoice in the currency that matches their production costs.
  - Sensitivity of desired markup to exchange rates, linked to strategic complementarities in price setting among firms in the destination market. High strategic complementarities can favor dominant currency pricing (DCP) to maintain relative price stability and market shares.
- Sectoral characteristics that affect currency choice:
  - Input structure (WIOD 2016 cross-country averages, year 2014 context):
    - Average share of intermediate input in gross output: manufacturing 26.7%, services 8.7%.
    - Average share of labor input: manufacturing 27.9%, services 57.5%.
    - Implication: higher domestic input intensity in services → lower marginal-cost sensitivity to exchange rates → greater incentives to price in the producer’s currency.
  - Intangibility and nonstorability (proximity burden):
    - When service delivery requires proximity or joint production, choice of currency depends on rivals’ pricing at location of delivery.
    - If exporters mainly compete with local providers at destination (who price in local currency) → local currency pricing; if trade occurs at exporter’s location and exporter also serves local consumers (e.g., Tourism) → producer-currency pricing.
  - Market structure and entry barriers:
    - Natural and behind-the-border policy barriers and network externalities can lead to high concentration in some services sectors, prompting some exporters to adjust markups and adopt LCP or DCP to stabilize market share.
  - Reliance on external finance:
    - Some services rely less on external finance (examples cited: Indian IT sector; generally few service sectors depend on external finance), and short/no lag between delivery and payment reduces working-capital needs.
    - Lower foreign-currency liabilities → less incentive to price in a foreign currency.

### Data Description
- Modes of supply and BoP measurement:
  - WTO GATS modes: mode 1 (cross-border), mode 2 (consumption abroad), mode 3 (commercial presence), mode 4 (movement of natural persons).
  - Balance of Payments (BoP) Statistics measure services provided by modes 1, 2, and 4 only; mode 3 (commercial presence / foreign affiliate sales) are excluded from the analysis because BoP measures transactions only between resident and nonresident entities.
- Bilateral Services Trade Database (BiTS):
  - Constructed by consolidating, imputing, and reconciling official data sources, with the WTO-UNCTAD-ITC Annual Trade in Commercial Services Database as primary source, supplemented by UN COMTRADE and Eurostat historical records.
  - Concepts/definitions/classification based on EBOPS 2010 and IMF BPM6.
  - Coverage: spans from 1985 to 2022 and covers 245 countries (data coverage in earliest years limited).
  - Methodology notes:
    - Uses exporter-reported services trade flows as the base value due to generally higher reliability of exporter data.
    - Employs partner-reported imports to create “mirror” flows to fill missing export values.
    - Performs systematic hierarchical consistency checks across categories.
  - BiTS contains exports and imports across 12 broad commercial service sectors (further broken down by partner and subcategories). Analysis focuses on aggregated and 1-digit services because finer disaggregation yields fewer observations.
- Complementary and auxiliary data sources:
  - US sectoral trade flows from US BEA for 1999–2022 (used to provide higher-quality sectoral evidence and to test DCP asymmetries).
  - Multilateral services trade: World Bank WDI (up to 2022) and IMF Trade in Services Database (1970–2014); WDI limited to Total services, Other commercial services, Travel, and Transport.
  - Tourism proxies: tourist arrivals and hotel nights from Eurostat tourism database (1991–2018) for nonresident arrivals and nights by country of origin.

### Basic Patterns of International Trade in Services
- Coverage and growth (reported measures and table values):
  - Text statement: “Total services grew by 7% annually, with OCS the fastest-growing sector.”
  - Table 2 values (WDI, 2022):
    - Total services: 7,209.0 (value in billions of US dollars) — Growth: 7.2% — BiTS coverage: 91%
    - Transport: 1,471.9 — Growth: 7.0% — BiTS coverage: 82%
    - Travel: 1,095.5 — Growth: 6.0% — BiTS coverage: 74%
    - Other commercial services: 4,554.4 — Growth: 9.0% — BiTS coverage: 98%
  - Notes: Column [1] represents global service export value in billions of US dollars (as recorded in the WDI for 2022); column [2] is average annual growth rate over 1995–2022; column [3] is average share of world trade covered by BiTS during 2005–2022.
- Rising importance and composition shifts:
  - Commercial services exports have expanded at about 3 times the growth rate of goods exports since 2011; they now account for nearly a quarter of global exports in gross terms and for about 40 percent in value-added terms.
  - Contribution to national export revenue (gross exports):
    - Services amount to 45% of UK and 33% of US gross exports.
    - In value-added terms, the services share exceeds 50% in France, Germany, Italy, the United Kingdom, and the United States.
    - In China, more than a third of the value added in its exports is due to services.
  - Composition by industry:
    - In the mid-1980s, Transport, Travel, and OCS each accounted for about a third of world services exports.
    - By 2019, OCS increased to almost 70% of world services exports; Transport and Travel shares declined.
    - Within OCS, fastest-growing segments: Other business services, Financial services, IT services, Charges for use of intellectual property.
    - Top 5 services industries (Transport, Travel, Finance, IT, Other business services) now account for 80% of global services exports.
  - Income-group patterns:
    - Most services exports are sourced from high-income countries, but middle-income countries’ roles have increased over time.
    - Advanced economies have become increasingly specialized in services exports, whereas EMDEs are more likely to specialize in manufacturing exports.
- Bilateral trade concentration and zeros:
  - Exporter–importer pairs in the top 1% of trade flows account for 56% of global trade in services; top 10% contribute about 95% in 2021.
  - About 45% of dyads do not trade in services or contain no information on services trade (comparable to share of zero goods-trade dyads in other studies).
- Drivers of services trade growth:
  - Reduced trading costs driven by information and communication technologies that mitigate the proximity burden.
  - Emergence and expansion of global value chains (GVCs) enabling international specialization in service-related tasks and services being traded as components within GVCs.

*Source: wpiea2024242-print-pdf (sections 2–4) — IMF PDF content provided in the prompt.*

### 4.1    Conceptual Framework

### 4.1    Conceptual Framework

### Conceptual model of services import demand
- Let q_ij denote the (log) quantity of imports by country j from country i; p_ij − p_j the (log) price of imported services relative to the domestic price in destination country j; y_j (log) real expenditures in importing country j.
- Parsimonious reduced-form import demand:
  - ∆q_ij,t = −σ_ij(∆p_ij,t − ∆p_j,t) + ∆y_jt. (Equation (1))
  - σ_ij is a reduced-form measurement of the elasticity of import quantities with respect to import prices (Armington parameter).
- Invoicing currency composition:
  - A fraction θ^i_ij invoiced in producer i’s currency (PCP), θ^j_ij in importer j’s currency (LCP), θ^$ _ij in dominant currency (DCP), with θ^i_ij + θ^j_ij + θ^$ _ij = 1.
  - Imported price inflation (expressed in importer j’s currency):
    - ∆p_ij,t = Σ_{k=i,j,$} θ^k_ij (∆p^k_ij,t + ∆e_{kj,t}). (Equation (2))
- Calvo sticky price reset (with reset price ̄p^k_ij,t and adjustment probability δ):
  - ∆p_ij,t = θ^i_ij ∆e_{ij,t} + θ^$ _ij ∆e_{j$,t} + (1 − δ) Σ_{k=i,j,$} θ^k_ij (̄p^k_ij,t − p^k_ij,t−1). (Equation (3))

### Short-run (prices rigid, δ → 1) implications for import values
- Change in log value of services imported by j from i (in j’s currency):
  - ∆v_ij,t = ∆p_ij,t + ∆q_ij,t = (1 − σ_ij) θ^i_ij ∆e_{ij,t} + (1 − σ_ij) θ^$ _ij ∆e_{j$,t} − σ_ij ∆p_j,t + ∆y_j,t. (Equation (4))
- Interpretations when controlling for destination prices p_j,t and demand y_j,t (for non-U.S. countries):
  - Producer currency pricing (PCP): θ^i_ij = 1, θ^$ _ij = 0 → import values reflect bilateral exchange rate changes scaled by (1 − σ_ij).
  - Local currency pricing (LCP): θ^i_ij = 0, θ^$ _ij = 0 → exporters fully absorb exchange rate fluctuations; import values independent of exchange rate movements.
  - Dominant currency pricing (DCP): θ^i_ij = 0, θ^$ _ij = 1 → import values reflect changes in importer currency–USD exchange rate scaled by (1 − σ_ij).
- As long as σ_ij ≠ 1, import values will respond to bilateral (producer currency) or dollar (dominant currency) exchange rate changes when the respective θ share is large.
- Without knowledge of σ, LCP cannot be tested (e.g., when σ_ij close to 1).

### Empirical strategy (motivation and specification)
- Estimation equation (log-linear, motivated by Equation (4)):
  - ∆v_ij,t = Σ_{τ=0}^L α_τ ∆e_{ij,t−τ} + Σ_{τ=0}^L β_τ ∆e_{j$,t−τ} + Γ′ X_{ij,t} + λ_{ij} + δ_t + ε_{ij,t}. (Equation (5))
- Dependent variable: v_ij,t = value of services imported by j from i (in destination currency).
- Controls X_{ij,t}: log change of importer real GDP and CPI, exporter PPI, and their two lags.
- Fixed effects: directed dyadic fixed effects λ_{ij}; time fixed effects δ_t.
- Short-term elasticities: α_0 (bilateral) and β_0 (dollar); medium-term elasticities: Σ_{τ=0}^L α_τ and Σ_{τ=0}^L β_τ.
- Standard errors clustered at directed dyadic level.
- Authors note identification limitations: exchange rate shocks are macro/financial phenomena; control variables are imperfect—results are reduced-form evidence (referred to as exchange rate elasticities).

### Exchange rate elasticities of aggregate services trade (key empirical findings)
- Regression results (Table 3): country sample excludes U.S. both as exporter and importer; dependent variable = value of imports of country j from country i (destination currency). Columns compare unweighted and trade-weighted regressions.
- Contemporaneous bilateral exchange rate elasticities (∆e_{ij,t}):
  - Unweighted: 0.436 ∗∗∗ (standard error 0.028).
  - Unweighted with dollar included: 0.223 ∗∗∗ (0.039).
  - Weighted: 0.391 ∗∗∗ (0.078).
  - Weighted with dollar included: 0.176 ∗∗∗ (0.056).
- Contemporaneous dollar exchange rate elasticities (∆e_{j$,t}):
  - Unweighted: 0.453 ∗∗∗ (0.057).
  - Weighted: 0.539 ∗∗∗ (0.097).
- Cumulative (Σ_{l=0}^2) bilateral:
  - Unweighted: 0.264 ∗∗∗ (0.046); with dollar included: 0.019 (0.057).
  - Weighted: 0.222 ∗ (0.133); with dollar: 0.016 (0.102).
- Cumulative dollar (Σ_{l=0}^2): unweighted 0.601 ∗∗∗ (0.087); weighted 0.593 ∗∗∗ (0.141).
- Observations: 79,552; Dyads: 6,708; R^2 reported (unweighted 0.17, weighted 0.32–0.33).
- Interpretation:
  - A 10% depreciation of country j’s currency relative to country i increases import value (in j’s currency) by 4.4% (unweighted) or 3.9% (weighted) when dollar excluded.
  - Including the dollar exchange rate reduces bilateral coefficients substantially.
  - Dollar exchange rate effect larger than bilateral: ≈0.45 (unweighted) and ≈0.54 (weighted).
  - Positive and significant α_0 and β_0 → both PCP and DCP present in total services trade.
  - Positive import value response to both bilateral and dollar depreciation suggests price increases dominate volume declines (consistent with σ < 1 on average).
- Comparison to manufacturing (Table 4):
  - Manufacturing contemporaneous bilateral elasticities:
    - Unweighted: 0.312 ∗∗∗ (0.031); with dollar included: 0.082 ∗ (0.043).
    - Weighted: 0.327 ∗∗∗ (0.044); with dollar included: 0.161 ∗∗∗ (0.038).
  - Manufacturing contemporaneous dollar elasticities:
    - Unweighted: 0.488 ∗∗∗ (0.061); weighted: 0.438 ∗∗∗ (0.054).
  - Services contemporaneous (for comparison) mirror earlier numbers: bilateral 0.450 ∗∗∗ (0.027) unweighted; dollar 0.468 ∗∗∗ (0.055) unweighted.
  - Notable contrasts:
    - Manufacturing: dollar exchange rate movements have about six times larger effect than bilateral shocks on manufacturing trade (relative α/β).
    - Services: dollar coefficient only about twice as large as bilateral coefficient.
  - Implication: DCP (relative to PCP) is more commonly adopted in manufacturing than in services (if σ similar across transaction currencies).

### Exchange rate elasticities across service categories (Table 5 key results)
- Focus on five largest service sectors: Transport, Travel, Financial services, Telecommunications/computer/information (IT), Other business services, plus Intellectual Property (IP).
- Short-term (contemporaneous) bilateral (∆e_{ij,t}) and dollar (∆e_{j$,t}) elasticities (selected estimates; standard errors in parentheses):
  - Transport:
    - ∆e_{ij,t} = 0.443 ∗∗∗ (0.048); cumulative Σ_{l=0}^2 ∆e_{ij,t−l} = 0.346 ∗∗∗ (0.059).
    - ∆e_{j$,t} = 0.587 ∗∗∗ (0.090); cumulative Σ_{l=0}^2 ∆e_{j$,t−l} = 0.345 ∗∗∗ (0.116).
  - Travel:
    - ∆e_{ij,t} = 0.186 ∗∗∗ (0.055); cumulative = 0.208 ∗∗∗ (0.070).
    - ∆e_{j$,t} = 0.217 ∗∗∗ (0.076); cumulative = 0.282 ∗∗ (0.123).
  - Financial services:
    - ∆e_{ij,t} = 0.304 ∗∗∗ (0.039); cumulative = 0.114 ∗ (0.063).
    - ∆e_{j$,t} = 0.356 ∗∗ (0.148); cumulative = 0.76 ∗∗∗ (0.237).
  - IT:
    - ∆e_{ij,t} = 0.225 ∗∗∗ (0.048); cumulative = 0.017 (0.078).
    - ∆e_{j$,t} = 0.430 ∗∗∗ (0.143); cumulative = −0.033 (0.216).
  - Other business services:
    - ∆e_{ij,t} = 0.387 ∗∗∗ (0.077); cumulative = 0.388 ∗∗∗ (0.119).
    - ∆e_{j$,t} = 0.142 (0.110); cumulative = 0.199 (0.135).
  - Intellectual Property (IP):
    - ∆e_{ij,t} = 0.299 ∗∗∗ (0.111); cumulative = 0.154 (0.161).
    - ∆e_{j$,t} = 0.538 ∗∗ (0.212); cumulative = 0.373 (0.299).
- Sectoral patterns and interpretation:
  - USD exchange rate more important than bilateral for Transport, Financial services, IT, and IP in the short term.
  - Travel: bilateral and dollar effects similar.
  - Other business services: dollar exchange rate not statistically significant contemporaneously.
  - Over two years, USD effect strengthens for Travel, Financial services, Other business services; weakens for Transport, IT, and IP.
  - Differences likely reflect sectoral differences in price stickiness, reliance on imported inputs, and market concentration.

### Exchange rates and tourism: volume elasticities (Table 6)
- Use higher-quality quantity proxies: arrivals of nonresidents from country j at hotels in country i (and hotel nights spent). Regression specification as in Equation (5).
- Unweighted contemporaneous elasticities (∆e_{ij,t} and ∆e_{j$,t}; standard errors in parentheses):
  - Arrivals:
    - ∆e_{ij,t} = −0.369 ∗∗∗ (0.033); with dollar included ∆e_{ij,t} = −0.270 ∗∗∗ (0.049).
    - ∆e_{j$,t} = −0.132 ∗∗ (0.057) when included.
    - Cumulative Σ_{τ=0}^2 ∆e_{ij,t−τ} = −0.493 ∗∗∗ (0.068).
    - Cumulative Σ_{τ=0}^2 ∆e_{j$,t−τ} = −0.054 (0.095).
  - Hotel nights spent:
    - ∆e_{ij,t} = −0.369 ∗∗∗ (0.037); with dollar included ∆e_{ij,t} = −0.232 ∗∗∗ (0.055).
    - ∆e_{j$,t} = −0.184 ∗∗∗ (0.064) when included.
    - Cumulative Σ_{τ=0}^2 ∆e_{ij,t−τ} = −0.459 ∗∗∗ (0.069).
    - Cumulative Σ_{τ=0}^2 ∆e_{j$,t−τ} = −0.185 (0.109).
- Weighted regressions:
  - Dollar exchange rate loses statistical significance both contemporaneously and over two years.
  - Weighted cumulative results show negligible effect of ∆e_{j$,t} on outbound tourism.
- Quantitative implication:
  - A 10% depreciation of a country’s currency against the tourist-hosting country’s currency reduces, over a two-year horizon, tourist arrivals by 4.5% and hotel nights spent by 3.3% (from cumulative bilateral elasticities).
- Interpretation:
  - Unweighted results show evidence of both PCP and DCP in tourism in the short term; inclusion of dollar reduces bilateral coefficient.
  - Dollar exchange rate has much less negative effect on tourism volumes than bilateral exchange rate.
  - Weighted results suggest DCP observed more in smaller economies, while PCP predominates in larger economies.
  - Traditional Mundell-Fleming expenditure-switching effects of exchange rates on export volume remain important for tourism exports.

_Italic: Content summarized from "4.1 Conceptual Framework" and subsequent sections of the supplied IMF chapter PDF._

### 4.6    Evidence from US Imports and Exports

### 4.6    Evidence from US Imports and Exports

### Methodology
- Fixed-effects regressions of log changes of trade value (denominated in the destination country currency) on:
  - lagged (log changes of) bilateral exchange rates,
  - importer GDP (log changes),
  - importer CPI (log changes),
  - exporter PPI and two lags,
  - year fixed effects and dyadic fixed effects.
- Estimations run separately for US exports (Panel B) and US imports (Panel A).

### Key empirical findings — Aggregate and manufacturing comparisons
- Consistent with prior work (Gopinath et al., 2010, 2020), results indicate:
  - US exports respond strongly to bilateral exchange rate fluctuations.
  - US imports (manufacturing) show no significant effect from bilateral exchange rate fluctuations.
- Select coefficient estimates from Table 7 (standard errors in parentheses):
  - Panel A (Imports to the US), Total Services (Column (1)):
    - ∆e_i$,t = 0.094* (0.056)
    - ∑_{l=0}^2 ∆e_i$,t−l = 0.167* (0.089)
  - Panel A (Imports to the US), Manufacturing (Column (9)):
    - ∆e_i$,t = 0.222 (0.191)
    - ∑_{l=0}^2 ∆e_i$,t−l = 0.081 (0.206)
  - Panel B (Exports from the US), Total Services (Column (1)):
    - ∆e$j,t = 0.652*** (0.034)
    - ∑_{l=0}^2 ∆e$j,t−l = 0.686*** (0.053)
  - Panel B (Exports from the US), Manufacturing (Column (9)):
    - ∆e$j,t = 0.737*** (0.207)
    - ∑_{l=0}^2 ∆e$j,t−l = 0.657*** (0.141)
- Interpretation:
  - For exports: large, positive, and statistically significant contemporaneous and cumulative coefficients indicate that depreciation of a foreign currency relative to the US dollar increases US export values (denominated in foreign currency), implying export volumes decline little and export values rise.
  - For imports: manufacturing imports to the US are largely insensitive to bilateral exchange rate fluctuations in the short run, supporting dominant currency pricing (DCP) behavior where import prices are USD-denominated and do not move with bilateral rates.

### Sectoral heterogeneity in US services trade
- Short-run evidence that US imports in certain services sectors respond to bilateral exchange rate shocks, implying producer/firm currency pricing in these sectors:
  - Transport, Financial services, IT, and Other business services show strong short-run responses in import values to bilateral exchange rate shocks.
- US exports across all services categories (denominated in foreign currency) respond to foreign currency depreciation relative to the USD in both short and medium term.
- Exception noted for Travel and Other business services where export results are different; travel shows relatively stronger evidence of producer currency pricing based on tourist flow measures (discussed elsewhere in the chapter).

### Implications for invoicing and pricing behavior
- Results for the United States generally support pervasiveness of DCP in total services trade, mirroring manufacturing evidence that export prices and volumes respond to bilateral exchange rates while import values (USD-denominated) do not.
- Sectoral differences highlight that some service providers price in their own (producer) currencies, particularly in Transport, Financial services, IT, and Other business services for US imports.
- The positive export coefficients imply that when a foreign currency depreciates against the USD, US exporters (with prices denominated in foreign currency) can experience higher export values because the volume reduction is small relative to the price increase in foreign-currency terms.

### Cross-cutting note
- The chapter reports qualitatively similar conclusions for manufacturing and for total services trade, but emphasizes important heterogeneity across individual services sectors in currency invoicing and exchange rate pass-through.

*Source: IMF staff analysis (section 4.6, wpiea2024242-print-pdf).*

### References

### References

### Key bibliographic sources (selection)
- Citations include academic articles, working papers, technical reports, and major data/overview publications covering trade in goods and services, invoicing currency, exchange rate pass-through, trade elasticities, and global value chains. Representative authors and outlets in the list include Adler; Amador et al.; Amiti, Itskhoki, and Konings; Ariu; Atkeson and Burstein; Auer and Schoenle; Baldwin; Berman, Martin, and Mayer; Bernard et al.; Bils and Klenow; Borchert and Mattoo; Boz et al.; Bussière et al.; Calvo; Chen, Chung, and Novy; Chung; Devereux et al.; Duygan-Bump et al.; Eaton and Kortum; Francois and Hoekman; Gervais and Jensen; Goldberg and Tille; Gopinath (multiple contributions); Haveman and Hummels; Head, Mayer, and Ries; Helpman et al.; Heuser and Mattoo; Hill; Hoekman and Shepherd; IMF; Kimball; Klenow and Malin; Kugler and Verhoogen; Leigh et al.; Lenzen et al.; Li et al. (forthcoming); Lyonnet, Martin, and Mejean; Manova and Zhang; Mayer and Steingress; Miroudot, Sauvage, and Shepherd; Mukhin; Nakamura and Steinsson; Soyres et al.; Spilimbergo and Vamvakidis; WB (World Bank); WTO.

### Appendix A — The Bilateral Trade in Services Database (BiTS): scope and construction
- Purpose: Provide detailed international bilateral trade flows in services using original source data (no estimated or extrapolated data).
- Country and sector coverage:
  - 245 countries and a global total plus a residual region for unallocated flows.
  - 12 broad services sectors and the unallocated category.
  - Further divided into 30 distinct subcategories (text also refers to 28 subcategories in one place).
  - Time span: 1985-2022.
  - Values reported in million US dollar.

### A.1 Data sources (facts preserved)
- Primary source: WTO-UNCTAD-ITD Annual Trade in Services Database (referred to as the WTO database), covering 2005-2022 and updated annually in July. It draws on Eurostat, the OECD, the IMF, and various national statistical sources and reports exports and imports aligned with EBOPS 2010 (based on BPM6).
- Supplementary sources:
  - UN COMTRADE (COMTRADE) spanning 2000-2020 (used to extend earlier years).
  - Eurostat historical records covering 1985-2003 (used to extend earlier years).
- Rationale for exclusions: UNCTAD BPM6 annual statistics largely overlap with WTO database; BaTIS contains estimated and balanced flows using gravity-based models and symmetry-index weighted averages and was therefore not included because the objective is to aggregate only original data.

### A.2 Methods — processing, reconciliation, and mirroring
- Hierarchical consistency checks:
  - Vertical reconciliation applied from most-disaggregated levels up to aggregated levels to impute missing higher-level branches and to replace higher-level values with summed lower-level entries when:
    1) all subcategories are reported and the summed value is not zero, or
    2) subcategories are partially reported and the summed value is greater than the reported higher-level value.
  - Example: travel sector reconciliation between business travel and personal travel; apply replacement when subcategory sums exceed reported travel totals; do not replace when reported personal travel and business travel are both zeros.
  - Special handling: other modes of transport (transport sector) requires additional comparison due to alternative extended classifications in BPM5 and BPM6.
  - Reconciliation rules are based on general practices including TSD, ITPD-E, and BaTIS.
- Mirroring process:
  - Performed to maximize coverage and address exporter-importer reporting discrepancies arising from different data collection methods, reporting standards, and classification systems.
  - Preference established: exporter-reported services trade flows used as the base value; reported imports are used to create mirror flows to fill missing exporter values.
  - A companion dataset provides importer-reported trade with mirrored exporter-reported data filling missing values.
- Conversion from BPM5 to BPM6:
  - Follow IMF Statistics Department conversion matrix.
  - Most BPM5 categories matched to BPM6 with exceptions for insurance services and charges for the use of intellectual property.
  - After conversion, hierarchical consistency checks are reapplied for BPM6-aligned categories.
  - Specific reallocations noted (examples): Postal services moved into Transport services; Telecommunication services moved into Telecommunications, computer, and information services; merchanting removed to be consistent with BPM6.
- Source precedence and single-source rule:
  - When multiple sources cover a country-pair-year, preferential order is: WTO data, followed by COMTRADE data, and then Eurostat data.
  - Within each source, original flows are prioritized over mirror flows.
  - For consistency, a single source is used per country-pair-year; different sources are not combined to maximize subcategory coverage within a country-pair-year.
- Unaccounted values and zero flows:
  - Geographic consistency checks performed after selection preferences.
  - A residual region designated as XRR created to account for unallocated exports when original world total exceeds sum of bilateral flows.
  - World total rebalanced if it is lower than the sum of bilateral flows.
  - All zero trade flows in the database originate from original data sources and are not the result of imputation.

### A.3 Data coverage and provenance (exact distribution from Table 9)
- Database composition and metadata:
  - Includes a variable “Data source” to indicate provenance of each observation.
  - Reported values in million US dollar.
- Distribution of observations by data source (Table 9: Data source Percent)
  - WTO original 43.43
  - WTO mirrored 26.56
  - COMTRADE original 8.88
  - COMTRADE mirrored 7.44
  - Eurostat original 3.63
  - Eurostat mirrored 3.45
  - Rebalanced World 1.14
  - Residual region 5.48

*Bilateral Trade in Services and Exchange Rates: Evidence of Dominant Currency Pricing Working Paper No. WP/2024/242*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024242-print-pdf.pdf_