## Long Live Globalization: Geopolitical Shocks and International Trade

## Source details

**Canonical URL:** [Long Live Globalization: Geopolitical Shocks and International Trade](https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023225-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2023/english/wpiea2023225-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2023/english/wpiea2023225-print-pdf.pdf.json)

---

### Abstract — core findings
- Uses an augmented gravity model of bilateral trade flows among 59,049 country-pairs over the period 1948–2021, with a dataset of more than 4 million observations.
- Main empirical finding: geopolitical proximity between countries, measured by similarity of UN voting behavior, has "contradictory and statistically insignificant effects on bilateral trade depending on the level of economic development."
- The economic magnitude of geopolitical alignment is smaller than the effects of income or geographic distance and diminishes significantly when extreme outliers are removed from the sample.
- Standard gravity determinants retain expected signs and significance: income in origin and destination countries positively affects trade; greater distance reduces bilateral trade; cultural similarities and historical ties increase trade; membership in GATT/WTO and presence of FTAs lead to higher international trade.

### Data and methodology
- Coverage and dependent variable:
  - Panel: 59,049 pairs of countries, period 1948–2021; bilateral trade flows for 243 countries and territories from IMF Direction of Trade Statistics; dependent variable = sum of exports and imports between country-pairs expressed in US$.
- Key explanatory variables:
  - Real GDP per capita (World Bank WDI and UNCTAD); population; distance (great-circle kilometers between capital cities); geographic contiguity; common official language; common religion; colonial links (CEPII gravity database); GATT/WTO membership; presence of FTAs (WTO database); geopolitical alignment (ideal point model of UN voting behavior).
- Baseline structural gravity and panel specification:
  - T_ij = B (GDP_i)^α (GDP_j)^γ (Dist_ij)^ϑ U_ij (log-normal error).
  - ln(T_ijt) = β + α ln(GDP_it) + γ ln(GDP_jt) + ϑ ln(Dist_ij) + δ X_ijt + θ Geo_ijt + η_i + φ_j + μ_t + ε_ijt.
  - X_ijt includes population (origin and destination), contiguity, common language, common religion, colonial links, GATT/WTO membership, FTA existence.
  - Geo_ijt is the ideal-point estimate of geopolitical proximity based on UN voting similarity.
  - Country-origin (η_i), country-destination (φ_j) and year (μ_t) fixed effects included.
- Estimators and inference:
  - Poisson Pseudo-Maximum Likelihood (PPML) as recommended by Santos Silva and Tenreyro (2006).
  - Two-stage least squares with instrumental variable (2SLS-IV) for endogeneity checks (lagged geopolitics as instrument).
  - Local projection (LP) method to trace dynamic responses to geopolitical shocks (k = 0, 1, ..., 5 years).
  - Robust standard errors clustered at the country-pair level; results robust to clustering at the country level.

### Data descriptive statistics (preserve reported values)
- Bilateral trade flows: Obs. 913,564; Mean 389,991; Std. Dev. 4,442,396; Min 0; Max 481,000,000.
- Real GDP per capita (Origin and Destination): Obs. 2,902,032; Mean 7,455; Std. Dev. 15,507; Min 261; Max 93,892.
- Distance: Obs. 3,627,844; Mean 8,614; Std. Dev. 4,748; Min 0; Max 19,939.
- Geographical contiguity: Obs. 3,627,844; Mean 0.012; Std. Dev. 0.108; Min 0; Max 1.
- Common language: Obs. 3,324,513; Mean 0.191; Std. Dev. 0.393; Min 0; Max 1.
- Common religion: Obs. 2,699,867; Mean 0.184; Std. Dev. 0.256; Min 0; Max 1.
- Colonial history: Obs. 3,627,400; Mean 0.009; Std. Dev. 0.093; Min 0; Max 1.
- Population (Origin and Destination): Obs. 3,779,748; Mean 24,073,820; Std. Dev. 100,208,324; Min 3,244; Max 1,412,360,000.
- GATT membership (Origin and Destination): Obs. 4,124,232; Mean 0.410; Std. Dev. 0.492; Min 0; Max 1.
- WTO membership (Origin and Destination): Obs. 4,124,232; Mean 0.245; Std. Dev. 0.430; Min 0; Max 1.
- Free trade agreement: Obs. 3,627,844; Mean 0.038; Std. Dev. 0.192; Min 0; Max 1.
- Geopolitical alignment: Obs. 1,849,284; Mean 1.023; Std. Dev. 0.842; Min -5.73; Max 5.73.

### Key estimation results (Preferred specification: column [5], PPML)
- Elasticities (baseline specification, column [5]):
  - Real GDP per capita, origin: 0.088 percent.
  - Real GDP per capita, destination: 0.065 percent.
  - Distance: -0.161 percent.
- Elasticity interpretations:
  - "A 10 percent increase in real GDP per capita in origin is associated with an average increase of 8.8 percent in bilateral trade flows."
  - "A 10 percent increase in real GDP per capita in destination is associated with an average increase of 6.5 percent in bilateral trade flows."
  - "A 10 percent increase in geographic distance lowers bilateral trade flows by more than 16 percent on average."
- Additional coefficients (column [5] highlights):
  - Geographical contiguity: positive effect (not significant in baseline but significant in refined sample).
  - Common language: 0.082*** (baseline column [5], robust).
  - Common religion: 0.031*** (baseline column [5]).
  - Colonial history: 0.126*** (baseline column [5]).
  - Population, destination: 0.067*** (column [5]).
  - GATT, destination: 0.007 (column [5]).
  - WTO, destination: 0.030*** (column [5]).
  - FTA: 0.109*** (column [5]).
- Comparative magnitudes:
  - Impact of WTO (0.023 percent for destination in some specifications) is greater than that of GATT (0.007 percent for destination in some specifications).
  - Presence of FTA increases bilateral trade; magnitude almost twice as large as WTO effect for the full sample and larger for developing countries.

### Geopolitics: estimated effects, heterogeneity, and IV results
- Baseline and trimmed-sample behavior:
  - Preferred specification (column [5]) result: geopolitical alignment coefficient reported as negative and statistically significant in some specifications (earlier columns report -0.026*** for the whole sample).
  - Trimmed sample (5th–95th percentiles): magnitude declines to -0.003 percent and becomes statistically insignificant.
- Heterogeneity by income group:
  - Advanced economies: geopolitical alignment coefficient positive (column [7]); dynamic responses described as negative in some point-estimate descriptions.
  - Developing countries: geopolitical alignment coefficient negative (column [8]); dynamic responses described as positive in some point-estimate descriptions.
  - Decadal analysis: geopolitical alignment remains insignificant in each decade since the 1950s; magnitude increased in developing countries and declined in advanced economies over time.
- 2SLS-IV (lagged geopolitics as instrument) results:
  - IV estimate for geopolitical proximity: 0.082 percent.
  - IV comparative magnitudes reported: income 1.163 percent; geographic distance 1.263 percent.
  - Despite positive IV estimate, economic magnitude of geopolitics (0.082 percent) is smaller than income and distance effects.

### Impulse Response Functions (IRFs) and dynamics
- Method:
  - Local Projections (LP) per Jordà (2005); specification: T_{ij,t+k} - T_{ij,t-1} = β^k Geo_ijt + θ X_ijt + η_i + φ_j + μ_t + ε_ijt; cumulative responses up to k = 0,...,5 years.
  - Controls include income, population, GATT/WTO membership, and FTA existence.
- Findings:
  - IRFs indicate geopolitical shocks do not have a statistically significant effect on bilateral trade flows over the five-year period (95 percent confidence intervals include zero).
  - Dynamic heterogeneity (point estimates with wide confidence bands):
    - Advanced economies: coefficient on geopolitical alignment is negative over the five-year horizon (point estimates).
    - Emerging/developing markets: coefficient on geopolitical alignment is positive over the five-year horizon (point estimates).
  - Overall: "geopolitical proximity as measured by UN voting behavior has no significant effect on trade globalization once conventional determinants are controlled."

### Robustness, diagnostics, and supplementary specifications
- Robustness checks:
  - Alternative specifications including value-added bilateral trade flows (OECD-WTO) for a smaller sample — results consistent with baseline.
  - Deflating trade flows by the US producer price index does not alter results.
  - Results unchanged when using real GDP instead of real GDP per capita, or including both GDP and GDP per capita as implemented by Santos Silva and Tenreyro (2006).
  - Results robust to removal of extreme outliers and to decade-by-decade estimation.
- Residual diagnostics:
  - Residuals distributed symmetrically around zero with no significant skewness or kurtosis.
- Appendix selected regression coefficients (preserve original coefficients and standard errors):
  - Appendix Table A1 (Value-Added):
    - Real GDP per capita, origin 0.161*** [0.015]
    - Real GDP per capita, destination 0.155*** [0.008]
    - Distance -0.195*** [0.011]
    - Geographical contiguity 0.076*** [0.022]
    - Common language 0.097*** [0.017]
    - Common religion 0.055* [0.021]
    - Colonial history 0.108*** [0.021]
    - Population, destination 0.093*** [0.029]
    - FTA 0.042*** [0.010]
    - Geopolitical alignment -0.008 [0.009]
    - Number of observations 118,037; Adjusted R2 0.67
  - Appendix Table A2 (2SLS-IV):
    - Geopolitical alignment 0.082*** [0.013] (whole-sample IV)
    - FTA 0.736*** [0.030] (whole-sample IV)
    - Distance -1.263*** [0.020] (whole-sample IV)
    - Population, destination 0.846*** [0.040] (whole-sample IV)
    - Number of observations 538,193; Adjusted R2 0.62

### Interpretation, policy implications, and recommendations
- Interpretation:
  - "International trade relationships have proven to be, by and large, resilient to changes in the geopolitical landscape over a long period from 1948 to 2021" despite several intervals of heightened geopolitical tensions.
  - Geopolitics, as proxied by UN voting similarity, yields mixed and generally small effects on bilateral trade relative to core gravity determinants.
  - "This does not mean that trade linkages and supply chains remain constant. Global value chains evolve over time with economic and technological developments—and occasionally due to geopolitical considerations."
  - Political economy caveat: "globalization produces losers as well as winners, leading to the inevitable buildup of socioeconomic and political pressures unless corrective policy actions are taken on a timely basis."
- Policy recommendations:
  - Pursue appropriate economic and social policies "aimed at achieving greater openness while reducing the socioeconomic burden of globalization."
  - Avoid nationalist and protectionist policies that could "make the global economy less resilient and more unequal."
  - Ensure domestic social transfers and compensation mechanisms to support adversely affected segments of the workforce to sustain trade globalization.
  - Quotation summarizing the view: "[t]he failure of our world is not that there is too much globalization, but that there is too little." (Wolf (2004))

### Conclusion (concise)
- Empirical evidence from an augmented gravity model with an extensive panel (more than 4 million / more than 4.6 million observations on 59,049 country-pairs, 1948–2021) yields statistically significant and intuitive coefficients for standard gravity determinants: economic size (real GDP per capita and population) and geographic distance are the most important determinants of bilateral trade.
- Cultural and historical ties and FTAs significantly boost bilateral trade; WTO membership has a positive impact larger than GATT in many specifications.
- Geopolitical alignment measured via UN voting similarity: mixed and generally small effects in baseline estimates (negative in some specifications, insignificant when trimming outliers); IV estimates positive but economically small (0.082 percent); dynamic IRFs show no statistically significant impact over a five-year horizon.
- Overall conclusion: geopolitics, as proxied by UN voting similarity, does not appear to be a major determinant of trade globalization relative to income and distance once conventional gravity factors are controlled.

*IMF Working Paper WP/23/225, "Long Live Globalization: Geopolitical Shocks and International Trade," Serhan Cevik, October 2023.*

### Section 1

### Long Live Globalization: Geopolitical Shocks and International Trade

### Abstract — core findings
- Uses an augmented gravity model of bilateral trade flows among 59,049 country-pairs over the period 1948–2021, with a dataset of more than 4 million observations.
- Main empirical finding: geopolitical proximity between countries, measured by similarity of UN voting behavior, has contradictory and statistically insignificant effects on bilateral trade depending on the level of economic development.
- The economic magnitude of geopolitical alignment is smaller than the effects of income or geographic distance and diminishes significantly when extreme outliers are removed from the sample.
- Standard gravity determinants retain expected signs and significance: income in origin and destination countries positively affects trade; greater distance reduces bilateral trade; cultural similarities and historical ties increase trade; membership in GATT/WTO and presence of FTAs lead to higher international trade.

### Context and motivations
- Motivating question: "Are we really witnessing the death of globalization?"
- Recent shocks (COVID-19 pandemic; Russia’s invasion of Ukraine) have raised concerns about geoeconomic fragmentation, but trade openness (exports + imports divided by GDP) shows no sign of structural retreat and rebounded after pandemic-induced disruptions.
- Annual trade restrictions worldwide increased from about 1,000 in 2019 to over 2,500 in 2022; more recent data show a notable decline in post-pandemic trade restrictions in 2023 (source: Global Trade Alert noted in text).
- Geopolitical Risk Index (Caldara and Iacoviello, 2022) rose after Russia’s invasion of Ukraine but has not reached spikes seen after 9/11 or during the Cuban Missile Crisis and has since dissipated below the historical average.

### Data overview
- Panel: 59,049 pairs of countries, period 1948–2021; bilateral trade flows for 243 countries and territories from IMF Direction of Trade Statistics; dependent variable = sum of exports and imports between country-pairs expressed in US$.
- Key explanatory variables:
  - Real GDP per capita in origin and destination (World Bank WDI and UNCTAD).
  - Distance (great-circle distance in kilometers between capital cities), geographic contiguity, common official language, common religion, colonial links (CEPII gravity database).
  - Population in origin and destination (WDI).
  - Membership in GATT and the WTO; presence of bilateral or regional FTAs (WTO database).
  - Geopolitical alignment measured by an ideal point model of UN voting behavior (Voeten, 2013; Bailey, Strezhnev, and Voeten, 2017).
- Notes on measurement:
  - Distance proxies transportation costs and other trade barriers.
  - Binary variables (language, religion, colonial history, contiguity) take value 1 if shared, 0 otherwise.
  - About a quarter of UN resolutions are adopted without a roll call and thus do not provide information for ideal point estimation.

### Empirical strategy and methods
- Estimation approaches:
  - Poisson Pseudo-Maximum Likelihood (PPML) regression for the augmented gravity model.
  - Two-stage least squares with instrumental variable (2SLS-IV) to control for potential endogeneity.
  - Local projection (LP) method to trace dynamic responses to geopolitical shocks.
- Robustness checks:
  - Alternative specifications including value-added bilateral trade flows (OECD-WTO) for a smaller sample — results consistent with baseline.
  - Deflating trade flows by the US producer price index does not alter results.
  - Results unchanged when using real GDP instead of real GDP per capita, or including both GDP and GDP per capita as implemented by Santos Silva and Tenreyro (2006).
  - Results robust to removal of extreme outliers and to decade-by-decade estimation.

### Detailed empirical results (summary)
- Geopolitical alignment:
  - Exhibits contradictory and statistically insignificant effects on bilateral trade overall; effects vary by level of economic development.
  - Economic magnitude is smaller than that of income or distance and attenuates when outliers are removed.
  - Dynamic responses over the long run are similar in pattern to baseline findings.
- Gravity variables and other determinants:
  - Real GDP per capita (origin and destination): positive impact on bilateral trade.
  - Distance: negative association with bilateral trade (higher trade costs).
  - Geographical contiguity: increases trade among adjacent countries.
  - Cultural and historical ties (common official language, common religion, colonial relations): positive effects on trade.
  - Population (origin and destination): positive effects; population in destination countries has a greater impact.
  - International trade agreements (GATT/WTO membership, FTAs): associated with higher levels of trade.

### Interpretation and implications
- The world economy remains deeply interconnected; trade relationships have been resilient to episodic geopolitical shifts across the long sample 1948–2021, which includes Cold War and other major geopolitical events (e.g., Russia–Georgia 2008; Crimea 2014).
- Geopolitical shocks raised risks but have not produced structural deglobalization in bilateral trade when accounting for core gravity determinants.
- Policy challenge: manage socioeconomic burdens of globalization with sound economic and social policies and avoid nationalist and protectionist measures that could reduce resilience and increase inequality.

### Structure of the paper (as presented)
- Section II: overview of data used in empirical analysis.
- Section III: description of augmented gravity model and econometric results.
- Section IV: summary and concluding remarks.

*IMF Working Paper WP/23/225, "Long Live Globalization: Geopolitical Shocks and International Trade," Serhan Cevik, October 2023.*

### Section 2

### Section 2

### Methodology: Augmented Gravity Framework
- Structural gravity baseline: T_ij = B (GDP_i)^α (GDP_j)^γ (Dist_ij)^ϑ U_ij (log-normal error).
- Panel specification estimated (Equation (2)):
  - ln(T_ijt) = β + α ln(GDP_it) + γ ln(GDP_jt) + ϑ ln(Dist_ij) + δ X_ijt + θ Geo_ijt + η_i + φ_j + μ_t + ε_ijt.
  - X_ijt includes: population (origin and destination), geographical contiguity, common official language, common religion, colonial links, GATT and WTO membership, existence of an FTA.
  - Geo_ijt is the ideal-point estimate of geopolitical proximity based on UN voting similarity.
  - Country-origin (η_i), country-destination (φ_j) and year (μ_t) fixed effects included.
- Estimator choices:
  - Poisson Pseudo Maximum Likelihood (PPML) as recommended by Santos Silva and Tenreyro (2006) to accommodate zero trade flows, heteroskedasticity, and correlated errors.
  - Robust standard errors clustered at the country-pair level.
  - OLS (random-effects) used as robustness; fixed-effects OLS yield comparable results when applicable.

### Data
- Dataset coverage:
  - More than 4 million observations on 59,049 pairs of countries during the period 1948–2021.
  - Later phrased as more than 4.6 million observations on 59,049 country-pairs over 1948–2021.
- Descriptive statistics sample excerpts (preserve reported values):
  - Bilateral trade flows: Obs. 913,564; Mean 389,991; Std. Dev. 4,442,396; Min 0; Max 481,000,000.
  - Real GDP per capita (Origin and Destination): Obs. 2,902,032; Mean 7,455; Std. Dev. 15,507; Min 261; Max 93,892.
  - Distance: Obs. 3,627,844; Mean 8,614; Std. Dev. 4,748; Min 0; Max 19,939.
  - Geographical contiguity: Obs. 3,627,844; Mean 0.012; Std. Dev. 0.108; Min 0; Max 1.
  - Common language: Obs. 3,324,513; Mean 0.191; Std. Dev. 0.393; Min 0; Max 1.
  - Common religion: Obs. 2,699,867; Mean 0.184; Std. Dev. 0.256; Min 0; Max 1.
  - Colonial history: Obs. 3,627,400; Mean 0.009; Std. Dev. 0.093; Min 0; Max 1.
  - Population (Origin and Destination): Obs. 3,779,748; Mean 24,073,820; Std. Dev. 100,208,324; Min 3,244; Max 1,412,360,000.
  - GATT membership (Origin and Destination): Obs. 4,124,232; Mean 0.410; Std. Dev. 0.492; Min 0; Max 1.
  - WTO membership (Origin and Destination): Obs. 4,124,232; Mean 0.245; Std. Dev. 0.430; Min 0; Max 1.
  - Free trade agreement: Obs. 3,627,844; Mean 0.038; Std. Dev. 0.192; Min 0; Max 1.
  - Geopolitical alignment: Obs. 1,849,284; Mean 1.023; Std. Dev. 0.842; Min -5.73; Max 5.73.

### Key Estimation Results (Preferred specification: column [5], PPML)
- Elasticities (baseline specification, column [5]):
  - Real GDP per capita, origin: 0.088 percent.
  - Real GDP per capita, destination: 0.065 percent.
  - Distance: -0.161 percent.
- Interpretation in elasticity terms:
  - A 10 percent increase in real GDP per capita in origin is associated with an average increase of 8.8 percent in bilateral trade flows.
  - A 10 percent increase in real GDP per capita in destination is associated with an average increase of 6.5 percent in bilateral trade flows.
  - A 10 percent increase in geographic distance lowers bilateral trade flows by more than 16 percent on average.
- Other estimated effects (column [5] highlights):
  - Geographical contiguity: positive effect (not significant in baseline but significant in refined sample).
  - Common language: 0.082*** (baseline column [5], robust).
  - Common religion: 0.031*** (baseline column [5]).
  - Colonial history: 0.126*** (baseline column [5]).
  - Population, origin and destination: positive; population in destination has a greater impact.
    - Population, destination: 0.067*** (column [5]).
  - GATT, destination: 0.007 (column [5]).
  - WTO, destination: 0.030*** (column [5]).
  - FTA: 0.109*** (column [5]).
- Comparative magnitudes:
  - Impact of WTO (0.023 percent for destination in some specifications) is greater than that of GATT (0.007 percent for destination in some specifications).
  - Presence of FTA increases bilateral trade; magnitude almost twice as large as WTO effect for the full sample and larger for developing countries.

### Geopolitics: Estimated Effects and Robustness
- Preferred specification (column [5]) result:
  - Geopolitical alignment coefficient: negative and statistically significant in some specifications; reported baseline in earlier columns as -0.026*** (whole sample, column [1] context) and -0.026 percent mentioned for whole sample.
  - In trimmed sample (5th–95th percentiles) magnitude declines to -0.003 percent and becomes statistically insignificant.
- Heterogeneity by country group:
  - Advanced economies: geopolitical alignment coefficient positive (column [7]); dynamic responses negative in one description—see heterogeneity below.
  - Developing countries: geopolitical alignment coefficient negative (column [8]); dynamic responses positive in one description—see heterogeneity below.
- Endogeneity check (2SLS-IV using lagged geopolitics as instrument):
  - IV estimates indicate geopolitical proximity has a positive impact on bilateral trade: 0.082 percent.
  - IV magnitudes compared: income 1.163 percent; geographic distance 1.263 percent.
  - Despite positive IV estimate, economic magnitude of geopolitics (0.082 percent) is smaller than income and distance effects.

### Sample Partitioning and Time Dynamics
- Decadal analysis (1948–2021):
  - Real GDP per capita in origin and destination consistently significant for bilateral trade flows in advanced and developing countries across decades.
  - Geopolitical alignment remains insignificant in each decade since the 1950s.
  - Over time: statistically insignificant magnitude of geopolitical alignment increased in developing countries but declined in advanced economies.
- Partition by income groups (columns [7] and [8]):
  - Advanced economies: coefficient on geopolitical alignment positive (statistically insignificant in some decadal estimates).
  - Developing countries: coefficient on geopolitical alignment negative.
- Table 2 summary (selected metrics across specifications):
  - Number of observations varies by specification (e.g., 807,129; 770,194; 680,714; 613,257; 220,810; 459,904 reported across columns).
  - Adjusted R^2 reported across specifications: 0.66, 0.67, 0.67, 0.67, 0.67, 0.65, 0.48, 0.37.
  - Origin FE, Destination FE, Year FE included in all displayed specifications.

### Impulse Response Functions (IRFs) to Geopolitical Shocks
- Method:
  - Local Projections (LP) per Jordà (2005) applied to panel to estimate cumulative responses up to k = 0, 1, ..., 5 years.
  - Specification: T_{ij,t+k} - T_{ij,t-1} = β^k Geo_ijt + θ X_ijt + η_i + φ_j + μ_t + ε_ijt.
  - Controls include income, population, GATT and WTO membership, and FTA existence.
- Findings:
  - IRFs indicate geopolitical shocks do not have a statistically significant effect on bilateral trade flows over the five-year period (95 percent confidence intervals include zero).
  - Dynamic heterogeneity:
    - Advanced economies: coefficient on geopolitical alignment is negative over the five-year horizon (point estimates), with wide confidence bands.
    - Emerging/developing markets: coefficient on geopolitical alignment is positive over the five-year horizon (point estimates), with wide confidence bands.
  - Overall conclusion from IRFs: geopolitical proximity as measured by UN voting behavior has no significant effect on trade globalization once conventional determinants are controlled.

### Robustness and Diagnostics
- PPML chosen for handling zeros, heteroskedasticity, and misspecification; OLS random-effects used as additional robustness.
- Country fixed effects for exporters and importers capture multilateral resistance terms.
- Results robust to clustering standard errors at country-pair level; remain broadly unchanged when clustering at country level.
- Residual diagnostics:
  - Residuals distributed symmetrically around zero with no significant skewness or kurtosis, implying the gravity model adequately captures main patterns.

### Conclusion (summary of substantive findings)
- Empirical evidence from an augmented gravity model with an extensive panel (more than 4 million / more than 4.6 million observations on 59,049 country-pairs, 1948–2021) yields statistically significant and intuitive coefficients for standard gravity determinants.
- Key drivers of bilateral trade:
  - Economic size (real GDP per capita and population) and geographic distance are the most important determinants.
  - Cultural and historical ties (common language, common religion, colonial history) and FTAs significantly boost bilateral trade.
  - WTO membership has a larger positive impact than GATT; FTAs have an even larger effect.
- Geopolitical alignment measured via UN voting similarity:
  - Mixed and generally small effects on bilateral trade in baseline estimates (negative in some specifications, insignificant when trimming outliers).
  - IV estimates produce a positive but economically small effect (0.082 percent).
  - Dynamic IRFs show no statistically significant impact over a five-year horizon.
- Overall: geopolitics, as proxied by UN voting similarity, does not appear to be a major determinant of trade globalization relative to income and distance once conventional gravity factors are controlled.

*Source: IMF; CEPII; World Bank; WTO; Voeten (2013); author's calculations.*

### Section 3

### Section 3

### Findings on geopolitical alignment and trade
- Geopolitical alignment, measured by the similarity of voting behavior at the UN, has "contradictory and statistically insignificant effects on trade, depending on the level of economic development: positive in advanced economies and negative in developing countries."
- "The economic magnitude of this effect is not as important as income or distance between the countries and it diminishes significantly when extreme outliers are removed from the sample."
- The statistical pattern remains similar when estimated by decades and when focusing on the dynamic response of bilateral trade flows to geopolitical shocks over a five-year horizon using the LP method.
- "International trade relationships have proven to be, by and large, resilient to changes in the geopolitical landscape over a long period from 1948 to 2021" despite "several intervals of heightened geopolitical tensions during the Cold War and various conflicts and wars, including Russia’s invasion of Georgia in 2008 and annexation of Crimea in 2014."

### Gravity variables and determinants of bilateral trade
- Income and market size:
  - "The level of income in both origin and destination countries has a positive impact on bilateral trade flows."
  - Population in origin and destination countries contributes positively to trade, with population in destination countries having a greater impact.
- Distance and geography:
  - "Distance between the countries ... is negatively associated with bilateral trade flows."
  - Geographical contiguity increases trade.
- Cultural and historical ties:
  - Linguistic similarities, common religion, and colonial relations have positive effects on bilateral trade flows.
- Trade agreements and institutional membership:
  - Membership to the GATT and the WTO and presence of FTAs "tend to have higher level of international trade."

### Robustness, dynamics, and interpretation
- Results are "broadly unchanged when I introduce additional variables, including a measure of geopolitical alignment between countries."
- "Taken together, these findings—robust to different specifications—show that there is no systemic retreat in trade globalization due to geopolitical developments."
- Caveat: "This does not mean that trade linkages and supply chains remain constant. Global value chains evolve over time with economic and technological developments—and occasionally due to geopolitical considerations."
- Historical perspective and implication: "no country—or region—in the world can be completely self-sufficient" and "global trade integration has continued to advance—with occasional setbacks—and brought prosperity to a growing number of people across the world."
- Political economy note: "globalization produces losers as well as winners, leading to the inevitable buildup of socioeconomic and political pressures unless corrective policy actions are taken on a timely basis."

### Policy recommendations and prescriptions
- Pursue appropriate economic and social policies "aimed at achieving greater openness while reducing the socioeconomic burden of globalization."
- Avoid nationalist and protectionist policies that could "make the global economy less resilient and more unequal."
- On compensation/adjustment: literature cited argues "the persistence of trade globalization depends on both global economic conditions and domestic social transfers to compensate the adversely affected segments of the workforce."
- Quotation underlining the view: "[t]he failure of our world is not that there is too much globalization, but that there is too little." (Wolf (2004))

### Appendix: Selected regression results (preserve original coefficients and standard errors)

- Appendix Table A1. Gravity Model of Bilateral Trade in Value-Added (Dependent variable: Bilateral trade in value-added; Robust standard errors clustered at the country-pair level)
  - Real GDP per capita, origin 0.161*** [0.015]
  - Real GDP per capita, destination 0.155*** [0.008]
  - Distance -0.195*** [0.011]
  - Geographical contiguity 0.076*** [0.022]
  - Common language 0.097*** [0.017]
  - Common religion 0.055* [0.021]
  - Colonial history 0.108*** [0.021]
  - Population, origin 0.019 [0.052]
  - Population, destination 0.093*** [0.029]
  - GATT, origin 0.050 [0.031]
  - GATT, destination 0.004 [0.036]
  - WTO, origin 0.004 [0.022]
  - WTO, destination 0.004 [0.006]
  - FTA 0.042*** [0.010]
  - Geopolitical alignment -0.008 [0.009]
  - Number of observations 118,037
  - Origin FE Yes
  - Destination FE Yes
  - Year FE Yes
  - Adjusted R2 0.67

- Appendix Table A2. Gravity Model of Bilateral Trade Flows: 2SLS-IV Estimation (Dependent variable: Bilateral trade flows; Robust standard errors clustered at the country-pair level)
  - Real GDP per capita, origin 0.634*** [0.182] 0.720*** [0.036] 0.623*** [0.023]
  - Real GDP per capita, destination 0.529*** [0.164] 0.718*** [0.020] 0.461*** [0.023]
  - Distance -1.263*** [0.020] -1.395*** [0.035] -1.245*** [0.024]
  - Geographical contiguity 0.608*** [0.080] 0.017 [0.342] 0.645*** [0.082]
  - Common language 0.605*** [0.033] 0.523*** [0.057] 0.652*** [0.040]
  - Common religion 0.150*** [0.045] 0.069 [0.080] 0.171** [0.057]
  - Colonial history 1.063*** [0.074] 1.216*** [0.097] 0.861*** [0.114]
  - Population, origin 0.149*** [0.046] -0.009 [0.103] 0.151* [0.060]
  - Population, destination 0.846*** [0.040] 0.454*** [0.050] 1.199*** [0.057]
  - GATT, origin 0.147*** [0.041] 0.438*** [0.089] 0.095 [0.045]
  - GATT, destination 0.085* [0.032] -0.013 [0.704] 0.167*** [0.051]
  - WTO, origin 0.104* [0.038] -0.035 [0.090] 0.130** [0.041]
  - WTO, destination 0.150*** [0.028] 0.181*** [0.033] 0.151*** [0.042]
  - FTA 0.736*** [0.030] 0.158*** [0.030] 0.962*** [0.040]
  - Geopolitical alignment 0.082*** [0.013] 0.154*** [0.019] 0.113*** [0.018]
  - Number of observations 538,193 184,024 354,169
  - Origin FE Yes Yes Yes
  - Destination FE Yes Yes Yes
  - Year FE Yes Yes Yes
  - Adjusted R2 0.62 0.78 0.55

*Source: wpiea2023225-print-pdf - Section 3*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2023/english/wpiea2023225-print-pdf.pdf_