## wp1908 - conclusion regarding the limited effects of trade costs on CA outcomes remains unchanged.

## Source details

**Canonical URL:** [wp1908 - conclusion regarding the limited effects of trade costs on CA outcomes remains unchanged.](https://www.imf.org/-/media/files/publications/wp/2019/wp1908.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2019/wp1908.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2019/wp1908.pdf.json)

---

### Empirical determinants of estimated nondiscriminatory import barriers
- MFN tariffs play a significant role in accounting for the within-country variation in estimated import barriers in both agriculture and manufacturing.
- GDP per capita is a robust correlate: richer countries tend to have lower import barriers.
- State support (available only for agriculture) is significantly positively associated with the estimated importer-specific barriers, suggesting state support in agriculture impedes imports more than it promotes exports.
- For services, a positive correlation exists between services trade restriction indices and estimated import barriers.
- Overall implication: data support considering nondiscriminatory barriers on the import side rather than on the export side, in line with the paper’s baseline EK framework.

### Consistency with macro-theory and literature
- Empirical findings are consistent with theories that generate small effects of trade costs on current account (CA) balances.
- Mechanisms noted:
  - Trade costs affect expected future profiles of intertemporal prices and income, which interact with central macroeconomic determinants of saving and investment.
  - Obstfeld and Rogoff (2000) show trade costs can create wedges between effective interest rates faced by borrowers and creditors; CA effects in that model are unlikely to be significant except at high levels of trade restriction.
- Empirical result: generally small effects of observed fluctuations in actual trade costs on current account balances.
- Stronger CA–trade restrictions relationship found in earlier sample when trade restrictions were higher.

### Measurement framework and estimation details
- Model framework: multisector generalization of Eaton and Kortum (2002).
- Productivity distribution: Fréchet with F^j_i(z) = exp(−T^j_i z^−θ).
- Key formulae and definitions (as used in estimation):
  - p^j_ni(q) = c^j_i z^j_i(q) d^j_ni.
  - π^j_ni = T^j_i (c^j_i d^j_ni)^−θ / ∑_k T^j_k (c^j_k d^j_nk)^−θ.
  - ln ̃X^j_ni = ln(T^j_i (c^j_i)^−θ) − ln(T^j_n (c^j_n)^−θ) − θ ln d^j_ni.
  - ln d^j_ni = D^j_ni,h + B^j_ni + L^j_ni + Con^j_ni + Col^j_ni + CU^j_ni + RTA^j_ni + im^j_n.
- Geographic distance intervals for D^j_ni,h: [0,350], [350,750], [750,1500], [1500,3000], [3000,6000], [6000,maximum].
- Importer-specific nondiscriminatory component: im^j_n = m^j_n − k^j_n / θ.
- Estimation approach:
  - Gravity equation saturated with exporter fixed effects (k^j_i) and importer fixed effects (m^j_n).
  - Poisson pseudo-maximum likelihood method used to handle heteroskedasticity and zero trade flows (Silva and Tenreyro 2006).
- Parameter choices and robustness:
  - Baseline assumption: θ = 8.28 in all sectors.
  - Robustness check: θ = 4.

### Data description — core inputs and construction
- Bilateral sectoral trade flows:
  - Johnson and Noguera (2017; JN): balanced panel 1970–2009 covering 37 countries and sectors agriculture, manufacturing, services (nonmanufacturing industry excluded).
  - WIOD 2016 release: 2000–2014, 56 disaggregated sectors in 43 countries; aggregated to JN sectors for comparability.
  - Services in JN: bilateral services trade flows imputed by applying bilateral goods trade shares to multilateral services data via constrained least-squares.
- Gravity controls and geographic measures from CEPII (GeoDist and CEPII structural gravity data set); distance measured between most populated cities and converted to miles for interval dummies.
- Tariffs and nontariff barriers:
  - MFN applied tariffs at HS 6-digit from UNCTAD via World Integrated Trade Solution; MFN tariff-by-importer-sector constructed as simple averages across HS 6-digit products within agriculture and manufacturing.
  - Bilateral tariffs complemented with trade-flow information for country pairs with trade agreements; sector-level bilateral tariffs are simple averages across HS 6-digit products with nonzero trade flows.
  - Services Trade Restrictiveness Indices (STRIs): OECD STRIs for 44 countries and 22 service subsectors for 2014–2016; scores in each policy area range from 0 to 1; indicator used is simple average across 22 subsectors.
  - Temporary Trade Barriers (TTB) database: Bown (2016); TTB-affected HS-6 products inferred from higher-digit raw data; bilateral and unilateral TTB shares computed at country-pair–sector and importer-sector levels respectively.
- Other variables:
  - Road density index from Du, Wei, and Xie (2013).
  - Agricultural state support proxied by OECD “total support” estimates.
  - IMF’s PPP-REER series uses 2011 ICP price levels and REER indices spliced onto 2011 exchange rates.
  - GDP per capita from Penn World Table 9.0 (real GDP at “chained” PPPs divided by population).

### Key gravity-estimation findings (selected reported coefficients and diagnostics)
- Distance effects: absolute value of geographic distance coefficients smallest for manufacturing and decreased over time for all three sectors; services show more negative distance coefficients than goods.
- Border and language: common border and common language boost trade; effects became smaller over time only for manufacturing.
- RTA: positively and significantly increase trade flows in agriculture and services; less significant for manufacturing.
- Select coefficients reported (values multiplied by θ = 8.28; standard errors in parentheses; significance: *p <0.1, **p <0.05, ***p <0.01):
  - Distance dummy [0, 375) (1990): Agriculture -3.284*** (0.530), Manufacturing -2.164*** (0.312), Services -4.096*** (0.319).
  - Distance dummy [6000, max) (2005): Agriculture -4.853*** (0.429), Manufacturing -3.513*** (0.192), Services -5.871*** (0.167).
  - Common Border (1990): Agriculture 0.639*** (0.145); (2005) Agriculture 0.838*** (0.126).
  - Common Language (1990): Agriculture 0.276** (0.127); (2005) Agriculture 0.319** (0.157).
  - Common Colonizer (1990): Agriculture -1.894*** (0.326); Manufacturing -3.666*** (0.201); Services -1.169*** (0.246).
  - RTA (1990): Agriculture 0.922*** (0.156); Manufacturing 0.172 (0.157); Services 0.277* (0.148).
  - Currency Union coefficients reported as 0.197, -0.033, 0.022 (standard errors in table).
- Goodness-of-fit and sample:
  - R-squared examples: 1990 agriculture 0.87, manufacturing 0.94, services 0.88; 2005 agriculture 0.88, manufacturing 0.87, services 0.89.
  - N = 1332 (reported in Table I).

### Comparative advantage, aggregation, and effective trade-cost measures
- Export capability (absolute advantage) per sector:
  - A^j_i = exp(k^j_i) / (1/N ∑_{i=1}^N exp(k^j_i)).
- Comparative advantage (relative across sectors for country i):
  - C^j_i = A^j_i / (1/J ∑_{j=1}^J A^j_i), with 1/J ∑_{j} C^j_i = 1 for all i.
- Aggregation across trading partners:
  - Export costs: DX^j_i = ∑_n w^j_ni d^j_ni.
  - Import costs: DM^j_i = ∑_n w^j_in d^j_in.
- Weighting schemes w^j_ni:
  - Lagged trade shares: one-year–lagged shares of exports/imports in partners’ total nominal exports/imports.
  - Frictionless trade counterfactual weights (d_ni = 1), with analytic expressions:
    - Import costs frictionless weights correspond to A^j_n / N.
    - Export costs frictionless weights reduce to partners’ shares in world gross expenditures X^j_n / ∑_n X^j_n.
- Aggregate comparative-advantage–weighted measures:
  - DXC_i = 1/J ∑_j (DX^j_i × C^j_i).
  - DMC_i = 1/J ∑_j (DM^j_i × C^j_i).

### Patterns of sectoral trade costs (reported scaling and stylized facts)
- Units and interpretation:
  - Reported estimates for exporting and importing costs are ad valorem tariff equivalents relative to cost of shipping internally.
  - A value of 2 for DX^j_i or DM^j_i indicates the estimated trade cost is equivalent to a tariff of 100 percent.
  - A value of 0.1 for im^j_i means the tariff equivalent of the nondiscriminatory import cost is 10 percentage points higher than the value for the United States.
- Stylized patterns:
  - Manufacturing trade costs trended down sharply; manufacturing is the sector with the lowest level of trade costs by recent decades.
  - Services trade costs are generally higher than goods irrespective of flow direction.
  - Importing costs are higher in EMDEs than in AEs.
  - EMDEs sharply reduced their costs of importing for all three sectors between the 1970s and 2000s.
- Nondiscriminatory import barriers:
  - Estimated relative to the United States (numéraire); im is zero for the United States by construction.
  - Essentially no country has negative im; the United States has the lowest estimated nondiscriminatory import barriers in the sample.
  - Most EMDEs observed declines in nondiscriminatory barriers between the 1970s and 2000s.

### Main empirical findings on current accounts (CA) and trade barriers
- Estimation approach:
  - Augmented the EBA CA panel regression with measures of aggregate effective trade costs for exporting and importing.
  - Experiments include effective export costs, effective import costs, and both simultaneously.
  - Two samples: 1986–2009 and 2001–2014.
  - Estimation: pooled generalized least squares with panelwide AR(1) correction.
  - Effective trade costs calculated using lagged trade shares; sources: JN and WIOD.
- Key findings:
  - Exporting trade costs:
    - Statistically significant negative impact on the CA.
    - Magnitudes:
      - For 1986–2009: the CA of an average country would improve by 0.5 percent of GDP if the costs of exporting to all trading partners fell by 10 percentage points in all sectors.
      - For 2001–2014: estimated effects of exporting costs are about half of those for 1986–2009.
    - Exporting costs dominate when both export and import costs are included.
  - Importing trade costs:
    - Generally statistically insignificant in 1986–2009.
    - In 2001–2014, sometimes statistically significant but economically small and counterintuitive in sign.
  - Contribution to global imbalances:
    - Decomposition shows effective trade costs made only a minor contribution to global CA imbalances over the last two decades.
- Representative coefficients and diagnostics (reported verbatim):
  - Table II (1986–2009) selected lines:
    - DXC -0.050*** -0.049*** -0.082*** -0.081***
      (0.000)(0.000)(0.000)(0.000)
    - DMC -0.008 0.001 -0.007 -0.004
      (0.359)(0.945)(0.408)(0.606)
    - Observations 761
    - R2 0.613 0.643 0.622 0.647 0.651 0.619 0.656
    - Number of countries 35
  - Table III (2001–2014) selected lines:
    - DXC -0.024** -0.020* -0.035** -0.034**
      (0.019)(0.079)(0.011)(0.012)
    - DMC -0.020* -0.012 -0.021* -0.019*
      (0.059)(0.333)(0.053)(0.090)
    - Observations 465 434 434 434 465 465 465
    - R2 0.739 0.753 0.762 0.763 0.746 0.754 0.758
    - Number of countries 31

### Robustness checks and sensitivity analyses
- Importer- versus exporter-specific costs:
  - Baseline follows Eaton and Kortum (2002) with importer-specific imjn; alternative allows exporter-specific exji so trade costs include imjn + exji.
  - Identification: latent factor χjn ≡ mjn − kjn/θ derivable from gravity; imjn and exjn not separately identified without additional data.
  - Re-estimation controlling for bilateral tariffs shows allowing exporter-contingent costs does not alter main results: export costs reduce CA by only a small amount; import costs remain less significant.
- Alternative weights in constructing effective trade costs:
  - Weighting schemes tested: contemporaneous nominal trade shares, nominal GDP, frictionless (counterfactual) trade flows.
  - Findings:
    - Contemporaneous nominal trade shares yield smaller absolute effects of DXC and DMC.
    - Using nominal GDP increases estimated impact: coefficient −0.085 versus −0.05 in baseline with lagged trade shares.
    - Aggregation schemes not dependent on actual trade flows usually indicate larger CA effects.
    - Frictionless aggregation yields slightly larger absolute coefficients for export costs; import costs remain statistically insignificant.
- “Effective” tariffs:
  - Replaced djni with bilateral tariff data for manufacturing and agriculture and aggregated as baseline.
  - Key statistics:
    - Average country had aggregate effective export cost of about 88 percent in 2014 while facing an effective tariff of only 4.7 percent.
    - Effective tariffs about 6 percent in 1995, while effective export costs were estimated at 89 percent that year.
  - Empirical finding: qualitatively similar to baseline; eliminating export tariffs faced by an average country in 2014 in both manufacturing and agriculture would improve that country’s CA balance by 0.8 percent of GDP.
- Lower elasticity of trade with respect to trade costs:
  - Experiment with θ = 4 (versus baseline θ = 8.28) yields higher trade-cost estimates.
  - Result: estimated CA effects are smaller than baseline and not detectable in the WIOD sample.

### Relationship between STRI and estimated nondiscriminatory import barriers (services)
- STRI plotted against gravity-implied nondiscriminatory import barriers, im, both relative to the US for select subsectors (2014):
  - Broadcasting (STRI 10), Telecom (STRI 12), Retail services (STRI 18), Commercial banking (STRI 19), Insurance (STRI 20).
- Evidence: the latent factor χ_jn,t tends to be significantly associated with observed importer-specific barriers such as MFN tariffs, STRI, and quality of institutions.
- Some OECD index measures apply to both domestic and foreign firms; such across-the-board barriers could worsen comparative advantage via productivity impacts rather than directly translating into higher import trade costs.

### Main conclusions — effect of trade policies on external balances
- Methodology: used bilateral trade flows to infer trade costs and sectoral comparative advantage; proposed effective trade cost measure that weights costs by comparative advantage.
- Main empirical conclusions:
  - Modestly robust link between effective trade costs and CA outcomes.
  - Countries facing higher effective costs to export tend to have only somewhat lower CA balances.
  - Overall contribution of trade costs to global imbalances has been small.
  - Conclusion regarding the limited effects of trade costs on CA outcomes remains unchanged.
- Identification caveat:
  - Potential reverse causality from the CA to effective trade costs could bias estimates.
  - In the Eaton and Kortum framework adopted, normalization by the country average of absolute advantage removes price and wage effects common across sectors, mitigating some reverse-causality concerns.

*Source: wp1908 - conclusion regarding the limited effects of trade costs on CA outcomes remains unchanged.*

### conclusion regarding the limited effects of trade costs on CA outcomes remains unchanged.

### wp1908 - conclusion regarding the limited effects of trade costs on CA outcomes remains unchanged.

### Empirical determinants of estimated nondiscriminatory import barriers
- MFN tariffs play a significant role in accounting for the within-country variation in estimated import barriers in both agriculture and manufacturing.
- GDP per capita is a robust correlate: richer countries tend to have lower import barriers.
- State support (available only for agriculture) is significantly positively associated with the estimated importer-specific barriers, suggesting state support in agriculture impedes imports more than it promotes exports.
- For services, a positive correlation exists between services trade restriction indices and estimated import barriers.
- Overall implication drawn: data support considering nondiscriminatory barriers on the import side rather than on the export side, in line with the paper’s baseline EK framework.

### Consistency with macro-theory and literature
- The empirical findings are consistent with theories that generate small effects of trade costs on current account (CA) balances.
- Mechanisms noted: trade costs affect expected future profiles of intertemporal prices and income, which interact with central macroeconomic determinants of saving and investment.
- Cited theoretical insight: Obstfeld and Rogoff (2000) show trade costs can create wedges between effective interest rates faced by borrowers and creditors, affecting consumption and saving; yet CA effects in that model are unlikely to be significant except at high levels of trade restriction.
- Empirical result: generally small effects of observed fluctuations in actual trade costs on current account balances.
- A stronger relationship between CA and trade restrictions is found for the earlier sample when trade restrictions were higher—consistent with Obstfeld and Rogoff (2000).

### Relation to related empirical studies
- Barattieri (2014): also estimates gravity equations for goods and services trade barriers but constructs a world-level home bias index; this paper instead uses an EK-type framework and leverages cross-country heterogeneity.
- Joy et al. (2018): report a strong negative relationship between CA balances and the interaction of revealed comparative advantages and world average tariffs; interpretation is difficult because world average tariff does not separate exporting and importing costs and STRI data are assumed unchanged since the 1990s in that study.
- Both Barattieri (2014) and Joy et al. (2018) rely on the Balassa revealed comparative advantage measure, which confounds effects of trade costs and importer-specific factors.

### Measurement framework and estimation details
- The paper adopts a multisector generalization of Eaton and Kortum (2002).
- Productivity distribution: Fréchet with F^j_i(z) = exp(−T^j_i z^−θ).
- Key definitions and relationships:
  - Market price from country i to destination n for sector j: p^j_ni(q) = c^j_i z^j_i(q) d^j_ni.
  - Probability country i supplies the lowest price in n for sector j:
    π^j_ni = T^j_i (c^j_i d^j_ni)^−θ / ∑_k T^j_k (c^j_k d^j_nk)^−θ.
  - Gravity equation (normalized exports):
    ln ̃X^j_ni = ln(T^j_i (c^j_i)^−θ) − ln(T^j_n (c^j_n)^−θ) − θ ln d^j_ni.
  - Bilateral trade costs decomposition:
    ln d^j_ni = D^j_ni,h + B^j_ni + L^j_ni + Con^j_ni + Col^j_ni + CU^j_ni + RTA^j_ni + im^j_n.
- Geographic distance intervals used in D^j_ni,h: [0,350], [350,750], [750,1500], [1500,3000], [3000,6000], [6000,maximum].
- Importer-specific nondiscriminatory component:
  im^j_n = m^j_n − k^j_n / θ.
- Estimation approach:
  - Gravity equation saturated with exporter fixed effects (k^j_i) and importer fixed effects (m^j_n).
  - Poisson pseudo-maximum likelihood method is used to handle heteroskedasticity and zero trade flows (Silva and Tenreyro 2006).
- Parameter choices and robustness:
  - Baseline assumption: θ = 8.28 in all sectors.
  - Robustness check: θ = 4.

### Comparative advantage and aggregation
- Export capability (absolute advantage) per sector:
  A^j_i = exp(k^j_i) / (1/N ∑_{i=1}^N exp(k^j_i)).
- Comparative advantage (relative across sectors for country i):
  C^j_i = A^j_i / (1/J ∑_{j=1}^J A^j_i), with 1/J ∑_{j} C^j_i = 1 for all i.
- Bilateral trade costs estimated via Equation (6) produce d^j_ni; nondiscriminatory importer component derived from fixed effects as above.
- Aggregation across trading partners to construct country-sector–level costs:
  - Export costs: DX^j_i = ∑_n w^j_ni d^j_ni.
  - Import costs: DM^j_i = ∑_n w^j_in d^j_in.
- Two weighting schemes for w^j_ni:
  - Lagged trade shares: one-year–lagged shares of exports/imports in partners’ total nominal exports/imports.
  - Frictionless trade counterfactual weights (d_ni = 1), yielding analytic expressions where:
    - For import costs, frictionless weights correspond to source countries’ absolute advantage A^j_n / N.
    - For export costs, frictionless weights reduce to partners’ shares in world gross expenditures X^j_n / ∑_n X^j_n.
- Aggregate comparative-advantage–weighted measures:
  - DXC_i = 1/J ∑_j (DX^j_i × C^j_i).
  - DMC_i = 1/J ∑_j (DM^j_i × C^j_i).

### Main conclusions
- Empirical evidence points to generally small effects of observed trade-cost fluctuations on current account balances.
- The inferred importer-side nondiscriminatory barriers correlate with MFN tariffs, GDP per capita, state support in agriculture, and services trade restriction indices—supporting focus on import-side nondiscriminatory barriers in the EK baseline.
- The conclusion regarding the limited effects of trade costs on CA outcomes remains unchanged.

*Source: wp1908 - conclusion regarding the limited effects of trade costs on CA outcomes remains unchanged.*

### 2.2    Data Description

### 2.2 Data Description

### Bilateral sectoral trade flows: sources and construction
- Primary data sources:
  - Johnson and Noguera (2017; hereafter JN): balanced panel for 1970–2009 covering 37 countries and the sectors of agriculture, manufacturing, services, and nonmanufacturing industry. The nonmanufacturing industry is excluded from analysis.
  - World Input-Output Database (WIOD), 2016 release: covers 2000–2014, 56 disaggregated sectors in 43 countries; 29 countries overlap with JN.
- Treatment of services trade in JN:
  - Bilateral services trade flows are imputed by applying bilateral trade shares for goods to multilateral data on services.
  - Imputation implemented via a constrained least-squares estimation that minimizes deviations between imputed services trade shares and target shares (based on goods trade) while constraining, for each country, the sum of bilateral flows to equal multilateral exports and imports.
- Aggregation for comparability:
  - WIOD’s 56 sectors are aggregated to match the JN sectors to avoid zero trade-flow problems arising from finer disaggregation.

### Gravity controls and geographic measures
- Gravity controls are drawn from the CEPII database (GeoDist and CEPII structural gravity data set).
- Geographic distance:
  - Measured as simple distance in kilometers between countries’ most populated cities; converted to miles to generate six interval-based distance dummies (see Section 2.1).
- Dummy variables and controls:
  - Landlocked indicator and common-continent indicator from CEPII GeoDist.
  - Regional trade agreement (RTA) indicator from WTO-based information.
  - Dummies for ever-in-colonial-relationship, same colonizer after 1945, and common official language.
- Other CEPII structural gravity controls used as available.

### Tariffs and nontariff barriers
- Tariff data sets:
  - Most-favored nation (MFN) applied tariffs at HS 6-digit level from UNCTAD Trade Analysis Information System via the World Integrated Trade Solution of the World Bank.
    - MFN tariffs used to disambiguate estimated nondiscriminatory import barriers; MFN tariff-by-importer-sector constructed as simple averages of MFN applied tariffs across HS 6-digit products within agriculture and manufacturing sectors.
    - MFN tariff data available for nearly all WTO members starting from the late 1980s.
  - Bilateral tariffs: constructed by complementing applied effective tariff rates with bilateral trade-flow information for country pairs with a trade agreement; sector-level bilateral tariffs are simple averages across HS 6-digit products with nonzero trade flows between each country pair.
- Services Trade Restrictiveness Indices (STRIs):
  - OECD STRIs for 44 countries and 22 service subsectors for 2014–2016 (Grosso et al. 2015).
  - Five policy areas: restrictions on foreign entry, restrictions on movement of individuals, other discriminatory measures, barriers to competition, regulatory transparency.
  - Scoring in each area ranges from 0 (completely open) to 1 (completely closed); subsector scores aggregated into a single value; indicator used here is the simple average across 22 subsectors.
- Temporary Trade Barriers (TTB) database:
  - Source: Bown (2016), published by the World Bank.
  - Includes bilateral measures (anti-dumping, countervailing duties, China-specific safeguards) and unilateral global safeguards.
  - Raw data highly disaggregated (HS-8 or higher) → assume an HS 6-digit product is TTB-affected when a higher-digit product is affected.
  - For bilateral TTBs: calculate number of TTB-affected 6-digit products at country pair–sector level and divide by number of nonzero 6-digit trade flows to estimate share of TTB-affected products.
  - For unilateral TTBs: same exercise at importer-sector level.

### Other variables and indices
- Domestic transport infrastructure:
  - Road density index from Du, Wei, and Xie (2013): total length of railroads and paved roads and railroads per square kilometer.
  - Time-invariant geographic/topological features addressed by controlling for all time-invariant features in specifications.
- Agricultural state support:
  - OECD “total support” estimate used as proxy for state support of agriculture (annual monetary value of gross transfers to agriculture from consumers and taxpayers arising from government policies).
- Real effective exchange rates and PPP:
  - IMF’s PPP-REER series uses 2011 ICP price levels and REER indices from IMF’s Information Notice System; REER indices are “spliced” onto 2011 exchange rates after expressing ICP levels relative to trading partners.
- GDP per capita:
  - Computed as real GDP (at “chained” PPPs) divided by population; both series from Penn World Table 9.0.

### Key gravity-estimation findings (summary)
- Gravity regressions estimated for 1990 and 2005, separately for agriculture, manufacturing, and services, based on JN data.
- Distance effects:
  - Absolute value of geographic distance coefficients are smallest for manufacturing and decreased over time for all three sectors.
  - Services show more negative distance coefficients than goods, indicating higher geographic barriers.
- Border and language:
  - Common border and common language boost trade by roughly similar magnitudes.
  - Effects of common border or language have become smaller over time only for manufacturing.
- Other controls:
  - Common continent and common currency show little effect on trade once other factors controlled for (note: eurozone is the sample’s only currency union).
  - Regional trade agreements (RTA) positively and significantly increase trade flows in agriculture and services; effect is relatively less significant for manufacturing.

- Table I specific metadata and summary statistics (as reported):
  - Reported values are the estimated coefficients multiplied by θ = 8.28.
  - Select reported coefficient examples (1990 and 2005 columns shown in table; standard errors in parentheses):
    - Distance dummy [0, 375) (1990): Agriculture -3.284*** (0.530), Manufacturing -2.164*** (0.312), Services -4.096*** (0.319).
    - Distance dummy [6000, max) (2005): Agriculture -4.853*** (0.429), Manufacturing -3.513*** (0.192), Services -5.871*** (0.167).
    - Common Border (1990): Agriculture 0.639*** (0.145); (2005) Agriculture 0.838*** (0.126).
    - Common Language (1990): Agriculture 0.276** (0.127); (2005) Agriculture 0.319** (0.157).
    - Common Colonizer (1990): Agriculture -1.894*** (0.326); Manufacturing -3.666*** (0.201); Services -1.169*** (0.246).
    - RTA (1990): Agriculture 0.922*** (0.156); Manufacturing 0.172 (0.157); Services 0.277* (0.148).
    - Currency Union coefficients reported as 0.197, -0.033, 0.022 (standard errors in table).
  - Goodness-of-fit and sample:
    - R-squared values reported: e.g., 1990 agriculture 0.87, manufacturing 0.94, services 0.88; 2005 agriculture 0.88, manufacturing 0.87, services 0.89.
    - N = 1332 (reported in Table I).
  - Significance notation: *p <0.1, **p <0.05, ***p <0.01.
  - Note: Estimates for both 1990 and 2005 are based on the Johnson and Noguera (2017) data set.

### Comparative advantage patterns (brief)
- Comparative advantage estimates plotted separately for JN and WIOD (differences arise from country coverage and service data construction).
- Observed country-specific trends (1970–2014):
  - China: sharp rise in manufacturing comparative advantage after the 1990s; agriculture comparative advantage falling.
  - United States: gained comparative advantage in services starting in 1980s; manufacturing comparative advantage weaker and deteriorated during the 2000s.
  - Japan and United Kingdom: manufacturing comparative advantage declining since early 1990s.
  - China, Germany, Japan, Mexico: greater comparative advantages in manufacturing than in services.
  - United States and United Kingdom: greater comparative advantages in services than in manufacturing.
- Correlation with Balassa-type revealed comparative advantage:
  - Pooled correlations nearly 0.8 using either data set.
  - Exception: services during 1970–2009 with correlation 0.4 (suggests trade costs and importer-specific factors relatively more important for services in that period).
  - Specific contrasts: gravity-based estimates show a steeper upward trend in China’s manufacturing comparative advantage after the mid-1980s relative to Balassa; Balassa-type measure shows stronger agricultural improvement for Brazil coinciding with early 1990s liberalization.

### Patterns of sectoral trade costs (summary)
- Definitions and scaling:
  - Reported estimates for exporting and importing costs are ad valorem tariff equivalents relative to cost of shipping internally.
  - A value of 2 for DX^j_i or DM^j_i indicates the estimated trade cost is equivalent to a tariff of 100 percent.
  - A value of 0.1 for im^j_i means the tariff equivalent of the nondiscriminatory import cost is 10 percentage points higher than the value for the United States.
- Sectoral and temporal patterns:
  - Manufacturing trade costs trended down sharply; manufacturing is the sector with the lowest level of trade costs by recent decades.
  - Services trade costs are generally higher than goods irrespective of flow direction.
  - Importing costs are apparently higher in EMDEs than in AEs.
  - EMDEs sharply reduced their costs of importing for all three sectors between the 1970s and 2000s (observations mostly below the 45-degree line in analyses).
- Nondiscriminatory import barriers:
  - Estimated relative to the United States (numéraire); im is zero for the United States by construction.
  - Essentially no country has negative im; the United States has the lowest estimated nondiscriminatory import barriers in the sample.
  - Most EMDEs observed declines in nondiscriminatory barriers between the 1970s and 2000s.

*Source: 2.2 Data Description, wp1908 - 2.2 Data Description (IMF working paper).*

### 3. Services2. Manufacturing1. Agriculture

### 3. Services2. Manufacturing1. Agriculture

### Methodology and data
- Augmented the EBA current account (CA) panel regression with measures of aggregate effective trade costs for exporting and importing.
- Experiments proceed by including effective export costs, effective import costs, and both simultaneously (reported in columns (2)–(7) of Tables II and III).
- Two samples analyzed: 1986–2009 (Table II) and 2001–2014 (Table III).
- Estimation: pooled generalized least squares with a panelwide first-order autoregressive correction.
- Effective trade costs calculated using lagged trade shares. Sources: Johnson and Noguera (2017) and the 2016 WIOD.

### Key empirical findings
- Exporting trade costs
  - Statistically significant negative impact of effective exporting trade costs on the CA.
  - Magnitudes:
    - For 1986–2009: the CA of an average country would improve by 0.5 percent of GDP if the costs of exporting to all trading partners fell by 10 percentage points in all sectors.
    - For 2001–2014: the estimated effects of exporting costs are about half of those for 1986–2009.
  - Country heterogeneity: effective trade costs can account for a nonnegligible portion of the CA balance in countries with higher-than-average effective costs to export (example cited: the US).
- Importing trade costs
  - In general, effective costs to import are statistically insignificant in the earlier sample (1986–2009).
  - In the later sample (2001–2014), coefficients for import costs are sometimes statistically significant but are economically small and in a counterintuitive direction.
  - When both export and import costs are included, exporting costs dominate in statistical and economic significance.
- Contribution to global imbalances
  - Figure V decomposition: over the last two decades, effective trade costs have made only a minor contribution to global CA imbalances.
  - Predicted CA values were decomposed into contributions from effective trade costs (exporting and importing) and from other EBA model variables; trade costs’ share is small.

### Representative coefficients and model diagnostics (as reported)
- Table II: Current Account and Trade Barriers, 1986–2009 (selected lines reproduced verbatim)
  - DXC -0.050*** -0.049*** -0.082*** -0.081***
    (0.000)(0.000)(0.000)(0.000)
  - DMC -0.008 0.001 -0.007 -0.004
    (0.359)(0.945)(0.408)(0.606)
  - Observations 761
  - R2 0.613 0.643 0.622 0.647 0.651 0.619 0.656
  - Number of countries 35
- Table III: Current Account and Trade Barriers, 2001–2014 (selected lines reproduced verbatim)
  - DXC -0.024** -0.020* -0.035** -0.034**
    (0.019)(0.079)(0.011)(0.012)
  - DMC -0.020* -0.012 -0.021* -0.019*
    (0.059)(0.333)(0.053)(0.090)
  - Observations 465 434 434 434 465 465 465
  - R2 0.739 0.753 0.762 0.763 0.746 0.754 0.758
  - Number of countries 31

### Interpretation and contextualization
- The larger and more robust effect of export costs (relative to import costs) is consistent with theoretical mechanisms that generate transitional and usually small effects of trade costs on CA balances, including habit formation in consumption and “time to build” in investment.
- Other empirical work cited (Barattieri, Cacciatore and Ghironi (2017); Joy et al. (2018)) finds small trade-balance responses to increased protectionism driven by expenditure switching and reduced import demand from lower real incomes.
- A detailed interpretation of why export and import cost effects differ would require deeper exploration of underlying mechanisms; that exploration is beyond the scope of this paper.

*Italic: Source: wp1908 - 3. Services2. Manufacturing1. Agriculture (IMF working paper content provided).*

### 3.2    Robustness Checks

### 3.2 Robustness Checks

### Importer- versus Exporter-Specific Costs
- Baseline formulation follows Eaton and Kortum (2002) with an importer-specific component imjn in trade costs. Alternative formulation allows exporter-specific component exji so that trade costs include imjn + exji (Equation (18)).
- Identification:
  - Only the latent factor χjn ≡ mjn − kjn/θ can be derived from the gravity estimation (Equation (19)).
  - imjn and exjn cannot be simultaneously identified without additional data.
- Re-estimation approach:
  - Gravity equation re-estimated controlling for bilateral tariffs and incorporating exji (Equation (20)), where τjni = ln(1 + tariffjni).
  - Including bilateral tariffs is especially important when assuming exjn = χjn because nondiscriminatory importer-specific tariff effects would otherwise be misattributed to estimated comparative advantage and domestic competition toughness.
- Empirical findings:
  - Allowing trade costs to differ contingent on the exporter does not alter main results (Table IV, columns (1) and (2)).
  - Column (1): results with imjn = χjn but controlling for bilateral tariffs are similar to baseline (column (4) of Tables II and III).
  - Column (2): setting exjn = χjn yields fairly similar results: aggregate effective export costs reduce the current account by only a small amount, and effects of import costs are less significant both economically and statistically.

### Alternative Weights in Constructing Effective Trade Costs
- Purpose: test sensitivity of aggregate effective trade costs to different weighting schemes when aggregating bilateral sectoral trade costs djni across partners and sectors.
- Weighting schemes examined:
  - Contemporaneous nominal trade shares to aggregate across trading partners (Table IV, column (2)).
  - Nominal GDP to aggregate across trading partners (Table IV, column (3)).
  - Frictionless (counterfactual) trade flows to aggregate across both trading partners and sectors (Table IV, column (5)).
- Empirical findings:
  - Using contemporaneous nominal trade shares yields smaller absolute effects of DXC and DMC on current account balances (column (2)).
  - Using nominal GDP increases the estimated impact on the current account: coefficient is −0.085 versus −0.05 in the baseline regression with lagged trade shares (column (3) vs baseline).
  - Aggregation schemes that do not depend on actual trade flows—whether contemporaneous or lagged—usually indicate larger current account effects.
  - Using frictionless trade flows across sectors:
    - Import-side aggregation with frictionless flows is equivalent to using sectoral comparative advantage: DMCi = ∑j DMji × Cji.
    - Export-side aggregation: DXCi = ∑j DXji × (∑n Xjn)/(∑j ∑n Xjn), i.e., sector weights equal to each sector’s share in total world production.
    - This alternative yields slightly larger (in absolute value) coefficients for costs to export; costs to import remain statistically insignificant (Table IV, column (5)).

### “Effective” Tariffs
- Extension: replace djni with bilateral tariff data for manufacturing and agriculture, then aggregate as in baseline to construct DXji, DMji, DXCi and DM Ci.
- Limitations:
  - Ignores services sector barriers, where barriers are not in the form of tariffs.
  - Tariffs constitute only a small fraction of total trade costs in the sample.
- Key statistics from estimations:
  - Average country in sample had aggregate effective export cost of about 88 percent in 2014 while facing an effective tariff of only 4.7 percent.
  - Effective tariffs were about 6 percent in 1995, while effective export costs were estimated at 89 percent in that year.
- Empirical finding (Table IV, column (6)):
  - Results qualitatively similar to baseline, though effective tariffs have quantitatively larger effects.
  - Estimated coefficient implies that eliminating export tariffs faced by an average country in 2014 in both manufacturing and agriculture would improve that country’s current account balance by 0.8 percent of GDP.
  - Overall implication: trade policies (tariff elimination) have a limited effect on the current account.

### Lower Elasticity of Trade with Respect to Trade Costs
- Experiment: set θ = 4 for all three sectors (lower elasticity than baseline θ = 8.28).
  - Motivation: recent work using tariff data reports aggregate elasticities of about 4 (e.g., Caliendo and Parro 2015).
- Consequence:
  - Lower θ translates into higher trade cost estimates.
- Empirical finding (Table IV, column (7)):
  - Estimated current-account effects are smaller than under the baseline scenario with θ = 8.28.
  - These effects are not detectable in the WIOD sample.

*Source: IMF Working Paper — 3.2 Robustness Checks (wp1908).*

### 0.8 for subsectors—such as telecommunications,  broadcasting,  retail and financial services—that

### wp1908 - 0.8 for subsectors—such as telecommunications,  broadcasting,  retail and financial services—that

### Relationship between STRI and estimated nondiscriminatory import barriers
- The relationship between the Services Trade Restrictiveness Index (STRI) and estimated nondiscriminatory import barriers is plotted in Figure VI for select subsectors in services for 2014.
- Subsections plotted in Figure VI:
  - Broadcasting (STRI 10)
  - Telecom (STRI 12)
  - Retail services (STRI 18)
  - Commercial banking (STRI 19)
  - Insurance (STRI 20)
- The figure plots the STRI against gravity-implied nondiscriminatory import barriers, im, both relative to the US.
- Evidence in this section suggests the estimated latent factor, χ_jn,t, tends to be significantly associated only with observed importer-specific barriers, for instance:
  - MFN tariffs
  - Services trade restrictiveness (STRI)
  - Quality of institutions
- Even state support in agriculture, commonly viewed as an export distortion, appears to have a more significant effect on importers (by enhancing their import competitiveness) than on exporters, thereby acting more as an import barrier than an export subsidy.
- Note: Some measures underlying the OECD’s indices apply to both domestic and foreign firms; such across-the-board barriers could worsen a country’s comparative advantage through adverse impacts on productivity of domestic firms and may not necessarily translate into higher trade costs to import. They would, however, constitute nondiscriminatory trade barriers to the extent that they affect foreign firms disproportionately more.
- To be consistent with the construction of estimated nondiscriminatory trade barriers, the analysis uses each country’s average STRI relative to that of the US.

### Conclusions — effect of trade policies on external balances
- Methodology and measures:
  - Used bilateral trade flows to infer trade costs and sectoral comparative advantage for a globally representative set of countries over varying time periods.
  - Proposed an effective trade cost measure that weights trade costs by comparative advantage to better capture costs pertaining to sectors where countries have greater underlying potential to export and import.
- Main empirical findings:
  - Found a modestly robust link between effective trade costs and current account (CA) outcomes.
  - Countries facing higher effective costs to export tend to have only somewhat lower CA balances.
  - The overall contribution of trade costs to global imbalances has been small.
  - These results are consistent with previous theoretical predictions suggesting limited effects of trade costs on current account balances.
- Limitation and identification concern:
  - Estimates could be biased by potential reverse causality from the CA to effective trade costs.
  - Dekle, Eaton, and Kortum (2008) note that movements in the current account can have general equilibrium effects on sectoral prices and wages, with implications for comparative advantage.
  - In the Eaton and Kortum (2002) framework adopted, such effects would be eliminated from the comparative advantage measure via normalization by the country average of absolute advantage (`a la Hanson et al. 2018). This normalization removes price and wage effects to the extent they are common across sectors.

### Appendix: Description of Other Variables in the CA Regression
- L.NFA/Y:
  - Lagged net foreign asset to GDP ratio from Lane and Milesi-Ferretti (2018).
- L.NFA/Y * (dummy if NFA/Y<–60%):
  - The preceding variable interacted with a dummy that takes the value of 1 when a country’s NFA-to-GDP ratio is less than –60 percent.
- L.Output per worker, relative to top 3 economies:
  - Lagged PPP GDP divided by working age population (from the IMF’s World Economic Outlook database) relative to the average of the United States, Japan, and Germany.
- L.Relative output per worker * K openness:
  - The preceding variable interacted with capital account openness (from the Quinn database).
- Oil and natural gas trade balance * resource temporariness:
  - Difference between exports and imports of oil and natural gas as a percentage of GDP; enters the regression only when the balance is positive.
  - The balance is interacted with a measure of temporariness, calculated as the ratio of current extraction to proven reserves (i.e., the inverse of “years until exhaustion”) scaled by the same ratio applicable to Norway in 2010. Higher values indicate the resource is expected to be exhausted sooner.
  - Data sources: the IMF’s World Economic Outlook database, the World Bank, and the BP Statistical Review of World Energy.
- GDP growth, forecast in 5 years:
  - IMF’s World Economic Outlook’s forecast of the five-year-ahead real GDP growth.
- L.Public health spending/GDP:
  - Ratio of public health spending to GDP, based on data from the OECD, the World Bank’s World Development Indicators, the UN Economic Commission for Latin America and the Caribbean, the IMF’s Financial Affairs Department, and the Asian Development Bank.
- L.demeaned VIX * K openness:
  - Chicago Board Options Exchange Volatility Index (from Haver Analysis database) interacted with capital account openness.
- Own currency’s share in world reserves:
  - A country’s currency share in world reserves, obtained from IMF and Currency Composition of Official Foreign Exchange Reserves.
- L.demeaned VIX * K openness * share in world reserves:
  - Interaction between the preceding two variables.
- Output gap:
  - Output gap as estimated by the IMF’s World Economic Outlook.
- Commodity ToT gap * trade openness:
  - Commodity terms-of-trade (ToT) index is the ratio of a geometric weighted average price of 43 commodity export categories to a geometric weighted average price of 43 commodity imports, both computed relative to manufactured goods prices in advanced economies; weights are given by commodities’ export and import shares.
  - To derive the cyclical gap, the ToT series is extended into the medium term (using commodity prices projected by the most recent IMF World Economic Outlook) and then Hodrick–Prescott filtered for each country. The resulting gap is interacted with trade openness, defined as the share in GDP of exports plus imports of goods and services.
- Institutional/political environment (ICRG-12):
  - Indicator used to gauge institutional and political risk based on factors including socioeconomic conditions, investment profile, corruption, religious tensions, democratic accountability, government stability, law and order, and bureaucratic quality (from the PRS Group’s International Country Risk Guide).
- Detrended private credit/GDP:
  - Private credit detrended using the BIS methodology, with sources: BIS credit statistics and the World Bank’s World Development Indicators.
- Cyclically adjusted fiscal balance, instrumented:
  - Instrument generated using a first-stage regression that incorporates the lagged and cyclically adjusted global fiscal balance, a time trend, lagged world GDP growth, lagged domestic and world output gaps, US corporate credit spreads, exchange rate regimes, the polity index, and the average cross-sectional fiscal balance.
- (∆Reserves)/GDP * K controls, instrumented:
  - Change in central bank foreign exchange reserves scaled by nominal GDP, both in US dollars.
  - First-stage regression includes M2/GDP, US interest rates, and global reserve accumulation with country-specific slopes.
  - Sources: IMF’s World Economic Outlook, Lane and Milesi-Ferretti (2018), and IMF’s Data Template on International Reserves and Foreign Currency Liquidity.
- Dependency ratio:
  - Old-age dependency ratio (ages 65+/ages 30–64), from UN World Population Prospects.
- Population growth:
  - From UN World Population Prospects.
- Prime savers share:
  - Share of prime savers (ages 45–64) as a proportion of the total working-age population (ages 30–64), from UN World Population Prospects.
- Life expectancy at prime age:
  - Life expectancy of a current prime-aged saver, from UN World Population Prospects.
- Life expectancy at prime age * future dep. ratio:
  - Interaction between future old age dependency ratio (computed as a moving average of the ratio 15-25 years forward) and life expectancy at prime age.

*Sources: WIOD and OECD Services Trade Restrictiveness Index; additional data sources as cited in the appendix descriptions above.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2019/wp1908.pdf_