## 2.1    Data sources; 3.2    Data description; 4.2    Geoeconomic fragmentation in services?

## Source details

**Canonical URL:** [2.1    Data sources; 3.2    Data description; 4.2    Geoeconomic fragmentation in services?](https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025163-source-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2025/english/wpiea2025163-source-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2025/english/wpiea2025163-source-pdf.pdf.json)

---

### Main data sources and scope
- Primary source:
  - OECD-WTO Balanced Trade in Services database (BaTIS; OECD, 2025).
- Supplementary sources:
  - UNCTAD-WTO, Eurostat, and UN Comtrade’s historical records.
- BaTIS current edition:
  - aligned with BPM6 service category classifications
  - covers 202 economies and 26 service categories
  - period 2005 to 2023
- BaTIS earlier edition:
  - based on BPM5
  - includes 191 economies and 11 service categories
  - period 1995-2012
  - two editions are not directly compatible due to differences in classification standards
- BaTIS reports three types of bilateral flows: reported, adjusted/imputed, and balanced; the dataset retains only the original reported flows and excludes adjusted, imputed and balanced values.
- Supplementary source details:
  - UNCTAD database: BPM6, 2005-2022, covering 226 economies and 106 service categories.
  - Eurostat: BPM5, observations for 1985-2003 across 66 economies and 85 categories.
  - UN Comtrade: 2000-2020, covering 244 economies and 101 service categories, under BPM5.
- Outcome:
  - a harmonized dataset of bilateral service-sector trade flows reaching back to 1985, with varying degrees of coverage.
- Units and scope:
  - BiTS contains bilateral services trade flows for up to 245 countries and geographic entities.
  - Data are reported in million US dollars and span 1985–2023.
  - Coverage by category:
    - 12 broad service categories
    - nine of these further divided into 26 subcategories
    - total of 29 distinct service-sector categories

### Dataset construction overview
- Four-step compilation process:
  1. Hierarchical reconciliation of trade flows in each individual source dataset (A).
  2. Use of exporter-reported values where available, and mirrored importer-reported data when necessary (B).
  3. Merge individual datasets that follow the same BPM standard into single combined datasets: BaEuCo for BPM5; BaUN for BPM6 (C).
  4. Use a concordance to convert BPM5 to BPM6 (D), then merge the two BPM6 datasets and perform a geographical consistency check (E).
- The merged result produces the final BiTS dataset and includes a geographical consistency check with respect to reported multilateral trade flows.

### Hierarchical reconciliation (A)
- Vertical hierarchical consistency checks performed within each BPM5 and BPM6 source dataset, proceeding from lowest (more disaggregated) to highest (more aggregated) category levels, consistent with the Extended Balance of Payments Services (EBOPS) hierarchical classification.
- Reconciliation rule (per reporter-partner-year flow):
  - Replace a higher-level category value with the sum of its lower-level components if:
    1) the higher-level category value is missing; or
    2) both higher- and lower-level data are available and the sum of the lower-level value exceeds the reported value of the higher-level category.
- Reconciliation applied separately for each observed flow type and at each category level starting from the most granular.

### Use of mirrored flows (B)
- Reporting principle:
  - BiTS prioritizes exporter-reported flows wherever available; mirrored importer-reported data are used only when exporter-reported data are unavailable.
- Rules to maximize coverage while maintaining consistency:
  - If a higher-level category is populated with exporter-reported data but all its subcategories are missing, mirrored values for subcategories may be introduced and scaled proportionally to preserve the relationship implied by the exporter-reported aggregate.
  - When a category is populated using mirrored data, do not override or augment its subcategories with exporter-reported data. Only one switch in flow type is permitted down the hierarchy: from exporter-reported to mirrored, but never the reverse.
- A companion BiTS version is provided that is entirely based on exporter-reported bilateral trade values without mirrored flows.

### Dataset merging (C)
- Source priority hierarchy (by preference): BaTIS > UNCTAD-WTO > Eurostat > UN Comtrade.
- When reconciling BaUN (BPM6) with BaEUCo (BPM5), BaUN values are prioritized wherever both are available for the same flow.
- If a preferred source provides an aggregate-level flow but lacks subcategory detail, missing subcategories can be filled from lower-priority sources and proportionally re-based to align with the preferred source aggregate to maintain hierarchical consistency.

### Concordance from BPM5 to BPM6 (D)
- IMF Statistics Department concordance (Appendix Table A.2) used to convert BPM5 data into BPM6 categories.
- Concordance converts BPM5 into three BPM6 levels: aggregate services (S), broad services subcategories (SA, SB, SC,...), and immediate subcategories (SC1, SC2,...).
- Accurate mapping may require BPM5 data disaggregated to as many as five levels because some narrow BPM5 subcategories reclassify under different BPM6 categories.
- Three BPM6 destination categories are not mappable from BPM5 and thus have data only from 2005 onward:
  - "Manufacturing services" (SA)
  - "Maintenance and repair services" (SB)
  - "Financial intermediation services indirectly measures" (SG2)
- Note: the concordance is "noisy"; users should be mindful when using pre-2005 trade flows.

### World totals trade flows (E)
- Geographical consistency check:
  - For each country, category, and year, compute total bilateral exports and imports. If the total exceeds the country’s reported exports to/imports from WLD, the reported WLD value is replaced with the corresponding sum total of bilateral flows.

### Zero-valued flows and missing values
- All zero-valued bilateral trade flows in BiTS are directly reported in the original source datasets and are not inferred or imputed; users can treat these as "true zeroes".
- Missing values indicate no information available from any source under the described procedure.
- The dataset does not overwrite missing values with zeros; users may infer zeros where reasonable but must make their own inferences.

### Modes of services trade delivery and BiTS coverage
- GATS modes described:
  - Mode 1 – cross-border supply
  - Mode 2 – consumption abroad
  - Mode 3 – commercial presence
  - Mode 4 – presence of natural persons
- Balance of Payments statistics capture only transactions via modes 1, 2 and 4.
- BiTS dataset is limited to Modes 1, 2 and 4.
- Mode 3 services are available from Foreign Affiliates Statistics (FATS) compiled by Eurostat; FATS and BiTS could be combined in principle to cover all four modes.

### Data coverage and temporal patterns
- BiTS coverage over time:
  - Early years: only a handful of major AEs (including the U.S., Germany and Japan) are covered.
  - From mid-1990s: all AEs covered to some extent; EMDE coverage appears regularly.
  - By 2010: BiTS contains at least some bilateral trade data for over 200 countries and territories.
  - Coverage drops off somewhat in 2022 and 2023 due to reporting lags.
- Missing-value patterns:
  - Inclusion of a country does not imply a complete bilateral matrix; most country pairs in any given year have missing bilateral trade values.
  - Among Advanced Economies (AEs), coverage of bilateral services trade is most comprehensive, with non-missing values for just under half of all possible bilateral trade flows from 2010 onwards.
  - Between AEs and EMDEs, even post-2010 three quarters of all possible bilateral trade flows are missing.
  - Bilateral services trade values among EMDEs are almost entirely missing.
- Economic significance of missing flows:
  - For the median country in BiTS, the value of multilateral services exports covered hovers around 30 percent during 1985-2000, rises to about 90 percent by 2010, and remains at this level thereafter.
  - Pattern is the same for total global services exports, indicating missing bilateral flows are likely small toward the end of the period but remain a concern prior to 2000.

### Network structure and composition (insights from 2019 and time series)
- Network visualization (2019):
  - Goods network (DOTS): geography strongly structures clusters; largest nodes include the U.S., China, Germany and Japan.
  - Services network (BiTS): gravity forces less evident; several smaller economies (e.g., the Netherlands and Ireland) are central; China is a minor node in services.
- Composition of services trade over time (1980s–2023):
  - Transport and travel shares have steadily declined over the last four decades.
  - Rising categories by 2023:
    - Other business services: 25 percent of global services trade by 2023.
    - Financial services: 9 percent of global services trade by 2023.
    - Information services: 11 percent of global services trade by 2023.
  - Interpretation: increasing importance of "weightless" services may reduce distance sensitivity and alter empirical properties relative to goods.

### Gravity-related empirical findings (2019 and trends)
- Naïve gravity model (2019) R-squared:
  - R2 for services = 0.49.
  - R2 for goods = 0.63.
- Time-series divergence:
  - Prior to 2007, naïve gravity fit was broadly comparable across goods and services.
  - Since 2007, goods R2 steady around 0.6; services fit has declined.
  - By 2023, model explains only about 2/3 as much variation in services trade as for goods.
- Distance elasticity estimates (PPML with exporter and importer fixed effects and controls):
  - Goods: generally around -0.6.
  - Services: declining over time.
    - Average estimate for 2000-04 = -0.62.
    - Average estimate for 2020-23 = -0.50.
- Category-level distance elasticities (averages reported):
  - Transport (SC): -0.54.
  - Travel (SD): -0.73.
  - Financial services (SG): -0.50.
  - Information services (SI): -0.37.
  - Other business services (SJ): -0.35.
- Aggregation identity and counterfactual:
  - Aggregate parameters approximate a weighted average of category-level estimates with weights equal to category shares.
  - Counterfactual holding category weights at 2000-04 shows that composition change explains most of the decline in the aggregate distance elasticity.

### Geoeconomic fragmentation in services? — empirical approach
- Definition:
  - Geoeconomic fragmentation = geostrategically driven reversal of cross-border economic integration.
- Sample and timing:
  - 76 major economies (same as OECD ICIO coverage).
  - Period: 2004-2023, bilateral trade-flow data averaged within three-year intervals (last interval is a two-year average for 2022-2023).
- Estimation:
  - Panel PPML gravity (equation (3)) with exporter-time (Ω_ot), importer-time (Π_dt) and exporter-importer (∆_od) fixed effects.
  - Time-varying geopolitical distance (geo_odt) measured as ideal point distance from UN General Assembly votes.
  - Controls: time-varying country-pair dummies for RTAs/FTAs and EU membership.
  - Identification of γ_t from within-country-pair over-time changes in ideal point distance.

### Fragmentation findings: goods versus services
- Goods trade:
  - Prior to 2016: a one standard deviation increase in ideal point distance associated with about a 2.5 percent decline in bilateral goods trade; evidence borderline statistically significant.
  - Since 2016: effect quadrupled in strength and become unambiguously statistically significant.
- Aggregate services trade:
  - Estimated time-varying coefficient on geopolitical distance remains similar to the pre-2016 goods pattern: a quantitatively weak negative association with limited statistical significance.
  - Conclusion: "At least so far, there is no clear evidence that international services trade is fragmenting along geopolitical fissures."

### Heterogeneity across service categories and geopolitical sensitivity
- Category grouping:
  - "Traditional services": transportation (SC) and travel (SD).
  - "Modern services": other commercial services (examples include "(charges for) intellectual property" (SH) and "telecommunications" (SI)).
- Key heterogeneous effects:
  - Little evidence that ideal point distance reduces bilateral exports of traditional transport and travel services.
  - Strong negative effects for modern services:
    - "(Charges for) intellectual property" (SH): a standard-deviation increase in geopolitical distance reduced intellectual property trade by about 10 percent.
    - "Telecommunications" / information services (SI): a standard-deviation increase in geopolitical distance reduced information services trade by about 8 percent.
- Interpretation:
  - Modern services that involve knowledge and technology sharing are more sensitive to geopolitical distance, likely due to government restrictions or private risk aversion.

### Composition shift and implications
- Composition shift:
  - Four modern services categories accounted for just over 20 percent of cross-border services flows in 1995 and rose to 58 percent by 2023.
- Implication:
  - While technology reduces traditional frictions, the rising share of modern, geopolitically sensitive services implies that growing geopolitical divisions could become an important brake on services trade growth going forward.
  - Empirical regularities for services trade should not be assumed to match those established for goods trade.

### Findings on trade-policy variables and EU membership
- RTAs/FTAs:
  - Promote goods trade on average, but have not had an economically or statistically significant impact on bilateral services trade during the sample period.
- EU membership:
  - Boosted services trade materially; effect on goods trade muted.
  - Identifying accession/exit cases: Bulgaria and Romania accession in 2007, Croatia in 2013, United Kingdom exit in 2020; these events show significant changes in services access to the Single Market even when goods patterns are muted.

### Policy-relevant conclusions and next steps
- Services trade liberalization remains an opportunity to promote economic growth and development given recent ICT improvements and remaining scope for expansion.
- Realizing gains requires:
  - A fuller understanding of frictions that inhibit services trade, especially for modern services sensitive to geopolitical alignment.
  - Targeted policy attention on knowledge- and security-sensitive service sectors (e.g., intellectual property and telecommunications).
- Data contribution:
  - The Bilateral Trade in Services (BiTS) dataset compiles and harmonizes bilateral services trade values under BPM6 across multiple categories and up to 245 economies, enabling broader empirical research on these issues.

*Source: wpiea2025163-source-pdf - Sections 2.1, 3.2, 4.2*

### 2.1    Data sources

### 2.1    Data sources

### Main data sources and scope
- The services trade data in BiTS is primarily sourced from the OECD-WTO Balanced Trade in Services datasbase (BaTIS; OECD, 2025), and it is supplemented by data from UNCTAD-WTO, Eurostat, and UN Comtrade’s historical records.
- BaTIS current edition:
  - aligned with BPM6 service category classifications
  - covers 202 economies and 26 service categories
  - period 2005 to 2023
- BaTIS earlier edition:
  - based on BPM5
  - includes 191 economies and 11 service categories
  - period 1995-2012
  - two editions are not directly compatible due to differences in classification standards
- BaTIS reports three types of bilateral flows: reported, adjusted/imputed, and balanced; the dataset retains only the original reported flows and excludes adjusted, imputed and balanced values.
- Supplementary sources used to address gaps and harmonize to BPM6:
  - UNCTAD database: BPM6, 2005-2022, covering 226 economies and 106 service categories.
  - Eurostat: BPM5, observations for 1985-2003 across 66 economies and 85 categories.
  - UN Comtrade: 2000-2020, covering 244 economies and 101 service categories, under BPM5.
- Outcome: a harmonized dataset of bilateral service-sector trade flows reaching back to 1985, with varying degrees of coverage.

### Dataset construction overview
- Four-step compilation process (illustrated in Figure 3):
  1. Hierarchical reconciliation of trade flows in each individual source dataset (A).
  2. Use of exporter-reported values where available, and mirrored importer-reported data when necessary (B).
  3. Merge individual datasets that follow the same BPM standard into single combined datasets: BaEuCo for BPM5; BaUN for BPM6 (C).
  4. Use a concordance to convert BPM5 to BPM6 (D), then merge the two BPM6 datasets and perform a geographical consistency check (E).
- The merged result produces the final BiTS dataset and includes a geographical consistency check with respect to reported multilateral trade flows.

### Hierarchical reconciliation (A)
- Vertical hierarchical consistency checks performed within each BPM5 and BPM6 source dataset, proceeding from lowest (more disaggregated) to highest (more aggregated) category levels, consistent with the Extended Balance of Payments Services (EBOPS) hierarchical classification.
- Reconciliation rule: for each reporter-partner-year flow, replace a higher-level category value with the sum of its lower-level components if:
  1) the higher-level category value is missing; or
  2) both higher- and lower-level data are available and the sum of the lower-level value exceeds the reported value of the higher-level category.
- Example: "travel" (SD) includes "business travel" (SDA) and "personal travel" (SDB). If travel is missing but subcategories exist, travel is replaced by the sum of subcategories; if travel and subcategories exist but subcategories sum to more than travel, travel is replaced by the sum; otherwise data left unchanged.
- Reconciliation is performed separately for each observed flow type (exporter’s exports to and imports from the partner) and at each category level starting from the most granular.

### Use of mirrored flows (B)
- Bilateral services trade flows often reported by both partners; discrepancies occur.
- BiTS prioritizes exporter-reported flows wherever available because services export data are typically gathered more consistently through compulsory surveys; mirrored import data are incorporated only when exporter-reported data are unavailable.
- To maximize coverage while maintaining consistency:
  - If a higher-level category is populated with exporter-reported data but all its subcategories are missing, mirrored values for subcategories may be introduced and scaled proportionally to preserve the relationship implied by the exporter-reported aggregate.
  - When a category is populated using mirrored data, do not override or augment its subcategories with exporter-reported data. Only one switch in flow type is permitted down the hierarchy: from exporter-reported to mirrored, but never the reverse.
- A companion BiTS version is provided that is entirely based on exporter-reported bilateral trade values without mirrored flows.

### Dataset merging (C)
- Merging requires rules for source prioritization and aggregation reconciliation.
- Source priority hierarchy (by preference): BaTIS > UNCTAD-WTO > Eurostat > UN Comtrade.
- When reconciling combined BPM6 dataset (BaUN) with BPM5-based dataset (BaEUCo), BaUN values are prioritized wherever both are available for the same flow.
- If a preferred source provides an aggregate-level flow but lacks subcategory detail, missing subcategories can be filled from lower-priority sources and proportionally re-based to align with the preferred source aggregate in order to maintain hierarchical consistency.
- This proportional rebasing preserves bilateral hierarchical consistency while using high-quality sources for aggregate values and lower-priority sources for disaggregation when necessary.

### Concordance from BPM5 to BPM6 (D)
- A novel concordance table supplied by the IMF’s Statistics Department (Appendix Table A.2) is used to convert BPM5 data into BPM6 categories.
- Concordance restricted to convert BPM5 data into three levels of BPM6 categories: aggregate services (S), broad services subcategories (SA, SB, SC,...), and their immediate subcategories (SC1, SC2,...).
- Accurate concordance may require disaggregated BPM5 data to as many as five levels because some narrow BPM5 subcategories are reclassified under different broader categories in BPM6.
- Almost all destination categories in BPM6 are matched with appropriate origin categories from BPM5 except:
  - "Manufacturing services" (SA)
  - "Maintenance and repair services" (SB)
  - "Financial intermediation services indirectly measures" (SG2)
- Consequently, trade data for these three categories are only available from 2005 onwards.
- Note: transition from BPM5 to BPM6 included methodological changes; the concordance is necessarily "noisy", and users should be mindful when using pre-2005 trade flows.

### World totals trade flows (E)
- Geographical consistency check analogous to hierarchical checks:
  - Many countries report bilateral exports to and imports from the world (WLD).
  - For each country, category, and year, compute the total value of bilateral exports and imports. If the total exceeds the country’s reported exports to/imports from WLD, the reported WLD value is replaced with the corresponding sum total of bilateral flows.

### Zero-valued flows and missing values
- All zero-valued bilateral trade flows in BiTS are directly reported in the original source datasets and are not inferred or imputed; users can treat these as "true zeroes".
- Missing values indicate no information available from any source under the described procedure.
- The dataset does not overwrite missing values with zeros; users may infer zeros where reasonable but must make their own inferences.

### Modes of services trade delivery
- GATS modes described and BiTS coverage:
  - Mode 1 – cross-border supply (e.g., design services via the internet).
  - Mode 2 – consumption abroad (e.g., tourism).
  - Mode 3 – commercial presence (foreign affiliate presence).
  - Mode 4 – presence of natural persons (e.g., consultants working on-site).
- Balance of Payments statistics capture only transactions via modes 1, 2 and 4.
- Consequently, the BiTS dataset is limited to Modes 1, 2 and 4.
- Note: Mode 3 services are available from Foreign Affiliates Statistics (FATS) compiled by Eurostat; in principle FATS and BiTS could be combined to cover all four modes.

### Data coverage and description
- BiTS contains bilateral services trade flows for up to 245 countries and geographic entities.
- Data are reported in million US dollars and span 1985–2023.
- Coverage by category:
  - 12 broad service categories
  - nine of these further divided into 26 subcategories
  - total of 29 distinct service-sector categories (listed in Appendix Table A.1)
- A “Data source” indicator records the source of each bilateral trade-flow observation.
- Data coverage improves over time:
  - Early years: only a handful of major AEs (including the U.S., Germany and Japan) are covered.
  - From mid-1990s: all AEs covered to some extent; EMDE coverage appears regularly.
  - By 2010: BiTS contains at least some bilateral trade data for over 200 countries and territories.
  - Coverage drops off somewhat in 2022 and 2023 due to reporting lags.
- Missing-value patterns:
  - The inclusion of a country does not imply a complete bilateral matrix; most country pairs in any given year have missing bilateral trade values.
  - Among Advanced Economies (AEs), coverage of bilateral services trade is most comprehensive, with non-missing values for just under half of all possible bilateral trade flows from 2010 onwards.
  - Between AEs and EMDEs, even post-2010 three quarters of all possible bilateral trade flows are missing.
  - Bilateral services trade values among EMDEs are almost entirely missing.
- Economic significance of missing flows:
  - Figure 6 comparison: for the median country in BiTS, the value of multilateral services exports covered hovers around 30 percent during 1985-2000, rises to about 90 percent by 2010, and remains at this level thereafter.
  - Pattern is the same for total global services exports, indicating missing bilateral flows are likely small toward the end of the period but remain a concern prior to 2000.

*Source: wpiea2025163-source-pdf - 2.1    Data sources*

### 3.2    Data description

### 3.2    Data description

### Network structure of goods versus services trade (Figure 7)
- Visualization uses BiTS data for services flows and IMF Direction of Trade Statistics (DOTS) for goods flows in 2019.
- Network representation details:
  - Economies more central to the trade network are shown as larger nodes.
  - Thickness of connecting lines reflects the intensity of bilateral trade.
  - Colors indicate modularity groupings generated by the Louvain community detection algorithm: countries in the same color cluster trade more intensively with each other than would be expected by random chance.
  - Note: The sizes of network nodes are comparable within each panel, but not across the two panels.
- Goods trade network (Figure 7(a)) findings:
  - Larger economies—such as the U.S., China, Germany and Japan—represent the most important network nodes.
  - Modularity clusters strongly reflect geography, with North America (blue), East and Southeast Asia (red), and Europe (yellow) forming distinct trading blocks.
  - These patterns reflect the "gravity" forces documented by Tinbergen (1962): variation in bilateral trade flows is well explained by economy size and geographic distance.
- Services trade network (Figure 7(b)) findings:
  - Gravity forces are less evident than in goods.
  - Some larger economies—such as the U.S. and Germany—remain central, but several smaller economies—such as the Netherlands and Ireland—are also central nodes.
  - China is only a minor node in the services trade network, in contrast with its outsized footprint in goods.
  - Geography is less pronounced: one dominant cluster (blue) includes advanced economies from both North America and Europe; the other two clusters mix smaller services traders from different regions.

### Composition of services trade over time (Figure 8)
- Figure 8 shows the evolution in the share of overall services exports of major service categories based on BiTS data.
- Chart elements:
  - Colorful lines: percentage of the value of total bilateral services exports in BiTS accounted for by different subcategories.
  - Grey bars: percentage of overall bilateral services exports (category S) in BiTS accounted for by the sum of bilateral exports across the major service subcategories.
- Key compositional trends:
  - Transport and travel services used to account for the bulk of services flows; these categories reflect the physical movement of goods and people and naturally exhibit gravity-like regularities.
  - The share of transport and travel in services trade has steadily declined over the last four decades.
  - Modern service categories have grown rapidly:
    - Other business services (including activities such as R&D and management consulting) rose from a negligible share in the late 1980s to 25 percent by 2023.
    - Financial services reached 9 percent of global services trade by 2023.
    - Information services reached 11 percent of global services trade by 2023.
  - Interpretation: The rising importance of more "weightless" services (financial, information, other business services) could cause services trade empirical properties to deviate from goods trade over time.

### Preview of gravity-related research applications (Section 4.1)
- Aggregate finding:
  - Economy size and geographic distance can explain a significant share of variation in bilateral services trade, but that share has declined over time and is substantially lower than for goods in recent data.
  - The diminishing explanatory power of gravity in services partly reflects a declining distance elasticity of services trade, much of which is attributable to a structural shift in composition toward less distance-sensitive categories.
- Naïve gravity model (equation (1)) estimates for 2019:
  - Regression R-squared (R2) findings:
    - R2 for services = 0.49.
    - R2 for goods = 0.63.
  - Interpretation: A naïve gravity model explains bilateral services trade significantly less well than bilateral goods trade in 2019.
- Time-series divergence (Figure 10):
  - Prior to 2007, naïve gravity fit was broadly comparable across goods and services.
  - Since 2007, goods R2 has been steady around 0.6, while the fit for services has continued declining.
  - By 2023, the model explains only about 2/3 as much variation in services trade as it does for goods.
- Distance elasticity analysis (equation (2) and Figure 11):
  - Estimation approach: PPML with exporter and importer fixed effects and controls for contiguity, common language, common legal origin, RTA participation, and EU membership.
  - Estimated distance elasticities over 2000 onward:
    - Goods: generally around -0.6 (a one percent increase in distance reduces goods flows by about 0.6 percent).
    - Services: declining over time.
      - Average estimate for 2000-04 = -0.62.
      - Average estimate for 2020-23 = -0.50.
    - The decline in the services distance elasticity aligns with the divergence of goods and services gravity fit.
- Compositional explanation and category-specific distance elasticities (Appendix Table B.1 summary):
  - Transport and travel services (SC and SD), whose shares have been declining, exhibit relatively strong distance elasticities (averaging -0.54 and -0.73, respectively).
  - Financial, information and other business services (SG, SI and SJ), whose shares have been rising, exhibit relatively weak distance elasticities (averaging -0.50, -0.37 and -0.35, respectively).
- Aggregation identity and counterfactual exercise (Figure 12):
  - When estimated with PPML and regressors that do not vary across aggregation levels, aggregate parameter estimates approximate the weighted average of category-level estimates, with weights equal to category shares.
  - Counterfactual holding category weights fixed at 2000-04 levels shows that if composition had remained unchanged, the decline in the aggregate distance elasticity would have been modest.
  - Conclusion: The observed declining services distance elasticity, and the weakening of gravity forces in services trade, is primarily the result of the increasing prominence of less distance-sensitive service categories.

*Source: 3.2    Data description (wpiea2025163-source-pdf)*

### 4.2    Geoeconomic fragmentation in services?

### 4.2    Geoeconomic fragmentation in services?

### Overview and research question
- The IMF defines "geoeconomic fragmentation" as a geostrategically driven reversal of cross-border economic integration.
- There is mounting evidence that geoeconomic fragmentation is underway for goods and capital, but the influence of geopolitics on services trade is less well documented. This section provides a first set of stylized facts on that question using the BiTS dataset.

### Empirical approach and data
- Sample and timing:
  - Sample restricted to 76 major economies (the same as those covered by the OECD Inter-Country Input Output Database).
  - Period: 2004-2023, with bilateral trade-flow data averaged within three-year intervals (the last interval is a two-year average for 2022-2023).
- Estimation:
  - Panel PPML gravity specification (equation (3)) with exporter-time (Ω_ot), importer-time (Π_dt) and exporter-importer (∆_od) fixed effects.
  - Time-varying geopolitical distance (geo_odt) measured as the ideal point distance based on UN General Assembly votes.
  - Controls: time-varying country-pair dummies for RTAs/FTAs and EU membership.
  - Identification of time-varying geopolitics effect γ_t comes from within-country-pair over-time changes in ideal point distance.

### Fragmentation in goods versus fragmentation in services
- Goods trade:
  - Prior to 2016: a one standard deviation increase in ideal point distance was associated with about a 2.5 percent decline in bilateral goods trade; evidence only borderline statistically significant.
  - Since 2016: the effect has quadrupled in strength and become unambiguously statistically significant, consistent with intensified influence of geopolitics on goods trade.
- Services trade (aggregate):
  - The estimated time-varying coefficient on geopolitical distance for aggregate services remains similar to the pre-2016 goods pattern: a quantitatively weak negative association with limited statistical significance.
  - Conclusion: "At least so far, there is no clear evidence that international services trade is fragmenting along geopolitical fissures."

### The role of geopolitical distance across service categories
- Categorization:
  - "Traditional services": transportation (SC) and travel (SD).
  - "Modern services": all other commercial service categories (examples emphasized include "(charges for) intellectual property" (SH) and "telecommunications" (SI)).
- Key estimates and patterns:
  - Heterogeneous sensitivity to foreign policy disagreement across service categories.
  - Little evidence that ideal point distance reduces bilateral exports of traditional transport and travel services.
  - Stronger negative effects for modern services, notably:
    - "(Charges for) intellectual property" (SH): a standard-deviation increase in geopolitical distance reduced intellectual property trade by about 10 percent.
    - "Telecommunications" / information services (SI): a standard-deviation increase in geopolitical distance reduced information services trade by about 8 percent.
- Interpretation:
  - Intellectual property and telecommunications services involve sharing of knowledge and technologies that are business sensitive or security critical; their cross-border supply to geopolitically distant partners may be constrained by government action or private-firm risk aversion.
  - The pattern mirrors evidence from goods trade that foreign policy disagreement is a barrier to high-tech manufacturing.

### Distance elasticities across service categories (from Table B.1)
- Persistent and sizable differences in distance elasticities (β̂) by service category (averages reported in text):
  - Transport (SC): -0.54 on average.
  - Travel (SD): -0.73 on average.
  - Financial services (SG): -0.50 on average.
  - Information services (SI): -0.37 on average.
  - Other business services (SJ): -0.35 on average.

### Composition shift in services trade and implications
- Composition change:
  - The four modern services categories analyzed accounted for just over 20 percent of cross-border services flows in 1995 and rose to 58 percent by 2023.
- Implication:
  - While technology reduces traditional frictions, the rising share of modern services—those more sensitive to geopolitics—means growing geopolitical divisions could become an important brake on services trade growth going forward.
  - Empirical regularities for services trade should not be assumed to match those established for goods trade.

### Findings on trade-policy variables and EU membership
- Trade agreements (RTAs/FTAs):
  - Promote goods trade on average, but have not had an economically or statistically significant impact on bilateral services trade during the sample period.
- EU membership:
  - Boosted services trade materially; effect on goods trade is muted given pre-existing good trade agreements in accession/exit cases.
  - Sample includes EU accessions and exit that identify EU effect via changes in membership:
    - Accessions: Bulgaria and Romania in 2007, Croatia in 2013.
    - Exit: United Kingdom in 2020.
  - EU entry/exit in these instances did not show significant changes in goods trade patterns but did change access to the Single Market for services, explaining the differential effects.

### Policy-relevant conclusions and next steps
- Services trade liberalization remains an opportunity to promote economic growth and development given recent ICT improvements and remaining scope for expansion.
- However, realizing gains from services trade requires:
  - A fuller understanding of the frictions that inhibit services trade, especially for modern services sensitive to geopolitical alignment.
  - Targeted policy attention on knowledge- and security-sensitive service sectors (e.g., intellectual property and telecommunications).
- Data contribution:
  - The Bilateral Trade in Services (BiTS) dataset compiles and harmonizes bilateral services trade values under BPM6 across multiple categories and up to 245 economies, enabling broader empirical research on these issues.

*Source: IMF Working Paper — Section 4.2 "Geoeconomic fragmentation in services?"*

### References

### References

### Geopolitics, fragmentation, and multilateralism
- AIYAR, S., CHEN, J., EBEKE, C. H., GARCIA-SALTOS, R., GUDMUNDSSON, T., ILYINA, A., KAN-GUR, A., KUNARATSKUL, T., RODRIGUEZ, S. L., RUTA, M., SCHULZE, T., SODERBERG, G. and TREVINO, J. P. (2023). Geoeconomic fragmentation and the future of multilateralism.IMF Staff Discussion Note 2023/001.
- —, MALACRINO, D. and PRESBITERO, A. F. (2024). Investing in friends: the role of geopolitical alignment in fdi flows.European Journal of Political Economy,83, 102508.
- BOSONE, C. and STAMATO, G. (2024). Beyond borders: how geopolitics is reshaping trade.ECB Working Paper 2024/196.
- CATALÁN, M., FENDOGLU, S. and TOMOHIRO, T. (2024). A gravity model of geopolitics and financial fragmentation.IMF Working Paper 2024/196.
- GOPINATH, G., GOURINCHAS, P.-O., PRESBITERO, A. F. and TOPALOVA, P. (2024). Changing global linkages: a new cold war?Journal of International Economics,153, 104042.
- HAKOBYAN, S., MELESHCHUK, S. and ZYMEK, R. (2023). Divided we fall: differential exposure to geopolitical fragmentation in trade.IMF Working Paper 2023/270.
- JAKUBIK, A. and RUTA, M. (2023). Trading with friends in uncertain times.Journal of Policy Modelling,45(4), 768–780.
- KESHK, O., POLLINS, B. and REUVENY, R. (2004). Trade still follows the flag: the primacy of politics in a simultaneous model of interdependence and armed conflict.Journal of Politics,66(4), 1155–1179.
- MARTIN, P., MAYER, T. and THOENIG, M. (2008). Make trade not war?Review of Economic Studies,75(3), 865–900.

### Gravity models, trade costs, and methodological foundations
- ANDERSON, J. E. (1979). A theoretical foundation for the gravity equation.American Economic Review,69(1), 106–116.
- — and VANWINCOOP, E. (2004). Trade costs.Journal of Economic Literature,42(3), 691–751.
- ANNE-CÉLIA, D. and HEAD, K. (2008). The puzzling persistence of the distance effect on international trade.Review of Economics and Statistics,90(1), 37–48.
- HEAD, K. and MAYER, T. (2014). Gravity equations: workhorse, toolkit, and cookbook. In G. Gopinath, E. Helpman and K. Rogoff (eds.),Handbook of International Economics, vol. 4, Elsevier, pp. 131–195.
- LEAMER, E. E. and LEVINSOHN, J. (1995). International trade theory: the evidence. In G. M. Grossman and K. Rogoff (eds.),Handbook of International Economics, vol. 3, Elsevier, pp. 1330–1294.
- SANTOSSILVA, J. M. C. and TENREYRO, S. (2006). The log of gravity.The Review of Economics and statistics,88(4), 641–658.
- YOTOV, Y. V. (2022). On the role of domestic trade flows for estimating the gravity model of trade.Contemporary Economic Policy,40(3), 526–540.
- BREINLICH, H., NOVY, D. and SANTOSSILVA, J. M. C. (2024). Trade, gravity, and aggregation.Review of Economics and Statistics,106(5), 1418–1426.
- CONTE, M., COTTERLAZ, P. and MAYER, T. (2022). The CEPII gravity database.CEPII Working Paper 2022-05.
- BOZ, E., LI, N. and ZHANG, H. (2019). Effective trade costs and the current account: an empirical analysis.IMF Working Paper 2019/008.

### Services trade, measurement, and datasets
- ARNOLD, J. M., JAVORCIK, B. S. and MATTOO, A. (2011). Does services liberalization benefit manufacturing firms?: Evidence from the Czech Republic.Journal of International Economics,85(1), 136–146.
- FIORINI, M., HOEKMAN, B. and QUINN, D. (2023). Services trade policy and industry performance in African economies.The World Economy,46(2), 382–395.
- FRANCOIS, J. and HOEKMAN, B. (2010). Services trade and policy.Journal of Economic Literature,48(3), 642–692.
- — and PINDYUK, O. (2013). Consolidated data on international trade in services.IIDE Discussion Paper 20130101.
- HOEKMAN, B. and SHEPHERD, B. (2019). Services trade policies and economic integration: new evidence for developing countries.Robert Schuman Centre for Advanced Studies Research Paper, RSCAS(2019/57).
- MIROUDOT, S., SAUVAGE, J. and SHEPHERD, B. (2013). Measuring the cost of international trade in services.World Trade Review,12(4), 719–735.
- LOUNGANI, P., MISHRA, S., PAPAGEORGIOU, C. and WANG, K. (2017). World trade in services: evidence from a new dataset.IMF Working Paper 2017/077.
- LI, N. and MELESHCHUK, S. (2024). Bilateral trade in services and exchange rates: Evidence of dominant currency pricing.IMF Working Paper 2024/242.
- OECD (2025).The OECD-WTO Balanced Trade in Services Database (BaTIS). OECD.
- OECD-WTO-IMF-UN (2025).Handbook on Measuring Digital Trade, Second Edition. OECD.
- OECD (2023). Development of the OECD Inter Country Input-Output Database 2023.OECD Science, Technology and Industry Working Papers 2023/08.
- ONS (2020). Asymmetries in trade data: updating analysis of UK bilateral trade data. Office for National Statistics, accessed June 10, 2025.

### Political economy, conflict, and state preferences
- BAILEY, M., STREZHNEV, A., and VOETEN, E. (2017). Estimating state preferences from United Nations voting data.Journal of Conflict Resolution,61(2), 430–456.
- KESHK, O., POLLINS, B. and REUVENY, R. (2004). Trade still follows the flag: the primacy of politics in a simultaneous model of interdependence and armed conflict.Journal of Politics,66(4), 1155–1179.
- POLLINS, B. (1989a). Conflict, cooperation and commercie: the effect of international political interactions on bilateral trade flows.American Journal of Political Science,33, 737–761.
- — (1989b). Does trade still follow the flag?American Political Science Review,83, 465–480.
- JOY, M., LISACK, N., LLOYD, S., REINHARDT, D., SAJEDI, R. and WITAKER, S. (2018). Mind the (current account) gap.Bank of England, Financial Stability Paper,43.

### Conceptual, definitional, and broader economic context
- HILL, T. P. (1977). On goods and services.Review of Income and Wealth,23(4), 315–338.
- COSTINOT, A. and RODRÍGUEZ-CLARE, A. (2014). Trade theory with numbers: quantifying the consequences of globalization. In G. Gopinath, E. Helpman and K. Rogoff (eds.),Handbook of International Economics, vol. 4, Elsevier, pp. 197–261.
- TINBERGEN, J. (1962). Shaping the world economy: Suggestions for an international economic policy.Twentieth Century Fund, New York.
- BALDWIN, R. (2016).The Great Convergence: Information Technology and the New Globalization. Belkna p Press: An Imprint of Harvard University Press.
- —, FREEMAN, R. and THEODORAKOPOULOS, A. (2024). Deconstructing deglobalization: the future of trade is in intermediate services.Asian Economic Policy Review,19(1), 18–37.
- BARATTIERI, A. (2014). Comparative advantage, service trade, and global imbalances.Journal of International Economics,92(1), 1–13.
- WORLDBANK (2024). Falling Long-Term Growth Prospects. World Bank Publications.
- YOTOV, Y. V. (2022). On the role of domestic trade flows for estimating the gravity model of trade.Contemporary Economic Policy,40(3), 526–540.

*Bilateral Trade in Services: Insights from A New Research Dataset Working Paper No. WP/2025/163*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025163-source-pdf.pdf_