## wp1745 — "Export Gains over Ten Years"

## Source details

**Canonical URL:** [wp1745 — "Export Gains over Ten Years"](https://www.imf.org/-/media/files/publications/wp/2017/wp1745.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2017/wp1745.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2017/wp1745.pdf.json)

---

### Abstract and headline result
- Trade agreements boosted exports in Latin America on average by 76.4 percentage points over ten years (cumulative, treated relative to synthetic).
- Sample: sixty-one Latin American country pairs, period 1989-1996 (initially 64 pairs; three removed for data constraints).
- World-average benchmark: 79.7 percentage points over ten years (Hannan 2016, 104 country pairs, 1983-1995).
- JEL Classification Numbers: F1, F13, F14
- Keywords: Trade agreements, international trade flows, synthetic control method.

### Methodology: synthetic control method (SCM) and empirical strategy
- SCM constructs a synthetic control as a weighted average of untreated units; weights chosen to match pre-treatment outcomes and observable covariates.
- Advantages of SCM highlighted:
  - Matches observable covariates and pre-treatment outcomes.
  - Allows unobserved confounders to vary with time (factor-loading model).
  - Addresses endogeneity concerns better than standard difference-in-differences for time-varying unobservables.
- Limitations noted:
  - Standard statistical inference inappropriate; inference via placebo/falsification tests used instead.
  - Requires absence of other events affecting only the treated area.
  - If donor-pool units are affected by the intervention, estimated effects may be underestimated.
  - Good pre-treatment fit between treated and synthetic required.
  - Difficulty finding suitable synthetic if treated region lies at extreme of characteristics.
- Gravity-model determinants used as SCM covariates:
  - Distance; GDP of each country; GDP per capita of each country; population of each country; Bilateral Real Exchange Rate; Remoteness (MTR proxy).
  - Dummy variables: Colonial history; Col to; Col from; Contig; Comleg; Comcur; Common language.
  - Exports lagged by three years.
- Data sources cited: Head, Mayer and Ries (2010) gravity dataset (trade from IMF DOTS), World Bank Development Indicators, Angus Maddison dataset, Shleifer and Glick; Glick and Rose (2002); CEPII; World Development Indicators.
- Donor-pool construction rules:
  - Exclude donor pairs that had a trade agreement in the same year.
  - Exclude donor pairs if the exporting country has any other trade agreement with any other country in the entire sample.

### Sample composition and covered agreements
- Major agreements covered: NAFTA, Mercosur, Group of Three, Andean Community, Central American Common Market, and several bilateral agreements (e.g., Bolivia-Mexico, Costa Rica-Mexico, Mercosur-Chile, Mercosur-Bolivia).
- Distribution of the 61 pairs:
  - Mercosur: 27 pairs
  - Andean Community: 16 pairs
  - NAFTA: 6 pairs
  - Group of Three: 6 pairs

### Results — aggregate and agreement-level averages
- Aggregate average export gain (treated relative to synthetic, cumulative over ten years): 76.4 percentage points.
- Agreement-level average (average treated relative to average synthetic, cumulative over ten years):
  - NAFTA: 80.9 percentage points
  - Mercosur: 23.7 percentage points
  - Group of Three: 9.5 percentage points
- Interpretation: Aggregate export gains in Latin America are driven predominantly by NAFTA when comparing average treated to average synthetic.
- Comparison to world-average (Hannan 2016): Latin America average lower by 3.4 percentage points (76.4 percent vs 79.7 percent).

### Results — individual country-pair outcomes and dispersion
- Of the 61 country pairs, 31 cases have positive export gains (51 percent).
- Per-pair average (average of each country pair’s treated vs its corresponding synthetic, cumulative over ten years):
  - Mercosur (per-pair average): 152.4 percentage points
  - Group of Three (per-pair average): 114.0 percentage points
  - NAFTA (per-pair average): 86.8 percentage points
- NAFTA participant-specific treated vs synthetic gains (ten-year cumulative):
  - United States: 44.5 percentage points
  - Canada: 71.6 percentage points
  - Mexico: 144.3 percentage points
- Cross-sample dispersion:
  - Average of individual country gains: 132.6 percentage points
  - Median of individual country gains: 9.9 percentage points
  - Maximum observed export gain: 4755.5 percentage points for CRI-MEX (Costa Rica’s exports to Mexico, post-1995 agreement).
  - Highest observed export loss: -967.7 percentage points (Bolivia’s exports to Ecuador following the trade agreement in 1995).

### Robustness, goodness-of-fit, and diagnostics
- Goodness-of-fit measure: normalized root-mean-square deviation (NRMSD) — RMSD of pre-treatment fit divided by average exports over ten years prior to intervention.
  - Smaller NRMSD indicates better pre-treatment fit.
  - Truncated plots shown for country pairs with NRMSD < 1 to focus on better-fit cases.
- Reporting of conditional averages:
  - When export gains for cases with NRMSD < 2 are considered, average export increase = 144.6 percentage points.
  - When pairs with export gains less than 1000 percentage points are considered, average export gain = -2.7 percentage points.
- Figures referenced for diagnostics and heterogeneity (enumerated in source): Figures 7–13 show export growth, export gains versus goodness-of-fit, and gains by size and income.

### Placebo (falsification) tests and inference
- Placebo procedure:
  - Ten treated units randomly chosen; for each exporting country A, five country pairs randomly selected showing exports of A to a country without a trade agreement with A (fifty placebo units).
  - SCM applied to these placebo units and evolution of treated relative to synthetic compared to placebo relative to synthetic.
- Placebo results:
  - Of fifty placebo units, five cases where the placebo unit shows a higher gain than the treated unit.
  - Estimated probability from placebo exercise: around 10 percent that a random placebo shows gains as large as treated units.
  - Interpretation: the treated-unit SCM estimates are large relative to randomly chosen pairs, supporting inferential strength via placebo tests.

### Heterogeneity patterns by size and income
- Export gains higher for smaller exporting countries (suggesting larger proportional boosts for smaller economies).
- Suggestive evidence that larger export gains have come from countries with lower income per capita.
- Visualization notes:
  - Figure 11: bubble sizes represent nominal GDP (USD million) of exporting country in the year of agreement.
  - Figure 12: bubble sizes represent nominal GDP per capita (USD) of exporting country in the year of agreement.

### Comparison with related literature
- Corrected-endogeneity parametric studies report substantial impacts:
  - Baier and Bergstrand (2007, 2009): trade agreements increase trade by around 100 percent in the long run (not limited to Latin America).
  - Hannan (2016) SCM on broader agreements yields similar results to parametric corrected-endogeneity estimates.
- Mixed evidence for regional agreements:
  - Kehoe (2003): applied general equilibrium models underestimated NAFTA impacts.
  - Coulibaly (2009): mixed results for Mercosur; positive impact using semi-parametric approach.
- General pattern: more positive results emerge when endogeneity is addressed with parametric or SCM approaches.

### Explanations for lower-than-world export gains in Latin America and policy implications
- Reasons cited for comparatively lower gains:
  - Relative weakness in trade openness.
  - Lower export diversification, product sophistication, and product complexity.
  - Lower participation in global value chains.
  - Need for structural reforms and reduction of nontariff barriers.
  - Lack of large dynamic regional trade hubs analogous to China in emerging Asia.
- Policy recommendations:
  - Undertake structural reforms.
  - Reduce nontariff barriers.
  - Improve infrastructure and quality of education.
  - Lower average tariffs.
  - Pursue initiatives to participate in global value chains where feasible.
  - Increase efforts to penetrate large markets, both advanced economies and regional market hubs.

*Source: wp1745 — "Export Gains over Ten Years" (section from wp1745 PDF).*

### 76.4 percentage points over ten years. However, there is variation across countries and

### wp1745 - 76.4 percentage points over ten years. However, there is variation across countries and

### Abstract
- Export gains reported: 76.4 percentage points over ten years.
- Sample description: world average comprises 104 country pairs in the period 1983-1995.
- JEL Classification Numbers: F1, F13, F14
- Keywords: Trade agreements, international trade flows, synthetic control method.
- Author’s E-Mail Address: sahmed@imf.org

### Acknowledgements and purpose
- Prepared as a background study for the Western Hemisphere Department’s Cluster Report on Trade Integration in Latin America and the Caribbean.
- The author thanks Valerie Cerra, Daniela E. Morgan, Danica C. Owczar, Nan Li, and seminar participants (IMF seminar at Western Hemisphere Department).

### Contents (structure and page pointers)
- Abstract .....................................................................................................................................2
- I. Introduction ..........................................................................................................................4
- II. Trade Agreements in Latin America ................................................................................6
- III. Empirical strategy .............................................................................................................8
  - A. SCM Methodology ....................................................................................................9
  - B. Application of SCM to Bilateral Trade Pairs ..........................................................12
  - The Trade Agreements .....................................................................................13
  - From Gravity Model to SCM ...........................................................................13
- IV. Results ...............................................................................................................................18
  - A. Average of Countries ..............................................................................................18
  - B. Individual Countries ................................................................................................21
  - C. Placebo (or Falsification) Tests ...............................................................................25
- V. Concluding Thoughts – Why Are Export Gains Less Than World? ...........................27
- VI. References ........................................................................................................................28

### Empirical strategy and methods
- Methodology emphasis: synthetic control method (SCM).
- Application focus: SCM applied to bilateral trade pairs.
- Relationship to gravity models: section "From Gravity Model to SCM" included as methodological bridge.

### Results summary (as presented)
- Aggregate reported outcome: 76.4 percentage points over ten years (export gains).
- Comparative note: export gains due to trade agreements are lower than the world average (world average based on 104 country pairs, 1983-1995).
- Results organized by:
  - Average of countries (Section IV.A)
  - Individual countries (Section IV.B)
  - Placebo (or falsification) tests (Section IV.C)

### Figures and tables enumerated in the content unit
- Tables:
  - 1. Trade Agreements
  - 2. Country Pairs
- Figures:
  - 1. Number of Country Pairs in Sample for Each Trade Agreement
  - 2. Average Exports
  - 3. Average Exports in NAFTA
  - 4. U.S. Exports to Canada and Mexico
  - 5. Canada Exports to Mexico and U.S.
  - 6. Mexico Exports to Canada and U.S.
  - 7. Export Growth of Average Treated over Ten Years, Relative to Average Synthetic
  - 8. Export Growth of Average Treated in Ten Years, Relative to Average Synthetic
  - 9. Export Gains versus Goodness of Fit (1)
  - 10. Export Gains versus Goodness of Fit (2)
  - 11. Export Gains by Size of Exporting Country
  - 12. Export Gains by Income per Capita of Exporting Country
  - 13. Export Gains over Ten Years, Average of Individual Country Gains

### Key messages and themes to extract for ingestion
- Quantified aggregate impact: "76.4 percentage points over ten years."
- Cross-country heterogeneity: "there is variation across countries and agreements."
- Benchmark/sample context: "world average comprising 104 country pairs in the period 1983-1995."
- Methodological approach: SCM applied to bilateral trade pairs, with links to gravity-model framework.
- Structure for downstream extraction: introduction, trade agreement description, SCM methodology, application to pairs, detailed results (average and individual), placebo tests, concluding interpretation about why export gains are less than world.

*Source: https://www.imf.org/-/media/files/publications/wp/2017/wp1745.pdf*

### 14. Export Gains over Ten Years ........................................................................................

### 14. Export Gains over Ten Years

### Introduction
- Study question: Have trade agreements boosted trade in Latin America?
- Method: Synthetic control method (SCM) applied to bilateral gravity-model-based covariates for country pairs.
- Sample and period: Sixty-one Latin American country pairs, period 1989-1996 (initially 64 pairs; three removed for data constraints).
- Key headline result: Trade agreements boosted exports in Latin America on average by 76.4 percentage points over ten years (cumulative, treated relative to synthetic).

### Rationale for SCM and advantages
- SCM constructs a synthetic (artificial) control unit as a weighted average of untreated units; weights chosen to match the treated unit’s outcome and observable covariates pre-treatment.
- SCM addresses endogeneity concerns by matching observable covariates and pre-treatment outcomes so that unobserved factors (in a factor-loading model) are allowed to vary with time.
- SCM allows the effects of unobserved confounders to vary with time, unlike standard difference-in-differences which only deals with time-invariant unobservables.
- Limitations noted:
  - Standard statistical inference is inappropriate; inference via placebo/falsification tests is used instead.
  - Accuracy requires absence of other events affecting only the treated area.
  - If donor-pool units are affected by the intervention, effects may be underestimated.
  - Good pre-treatment fit between treated and synthetic is needed.
  - If treated region lies at extreme of characteristics, finding a suitable synthetic may be difficult.

### Trade agreements and sample composition
- Major trade agreements covered: NAFTA, Mercosur, Group of Three, Andean Community, Central American Common Market, and several bilateral agreements (e.g., Bolivia-Mexico, Costa Rica-Mexico, Mercosur-Chile, Mercosur-Bolivia).
- Distribution of the 61 pairs in sample:
  - Mercosur: 27 pairs
  - Andean Community: 16 pairs
  - NAFTA: 6 pairs
  - Group of Three: 6 pairs
- Donor pool construction rules:
  - Donor pool excludes country pairs that had a trade agreement in the same year.
  - Donor pool excludes any other trade agreements that the exporting country has with any other country in the entire sample.

### Empirical strategy: gravity model to SCM covariates
- Gravity-model determinants used as SCM covariates:
  - Distance between bilateral pair
  - GDP of each country in the bilateral pair
  - GDP per capita of each country in the bilateral pair
  - Population of each country in the bilateral pair
  - Bilateral Real Exchange Rate
  - Remoteness of each country (proxy for multilateral trade resistance, MTR)
  - Dummy variables:
    - Colonial history = 1 if pair ever in colonial relationship
    - Col to = 1 if export from hegemon to colony
    - Col from = 1 if export from colony to hegemon
    - Contig = 1 for contiguity
    - Comleg = 1 for common legal origins
    - Comcur = 1 for common currency
    - Common language = 1 for common official language
  - Exports, lagged by three years
- Data sources: underlying gravity dataset from Head, Mayer and Ries (2010) (trade from IMF DOTS), World Bank Development Indicators, Angus Maddison dataset, Shleifer and Glick, Glick and Rose (2002) for legal origin and currency, CEPII for distances and language, World Development Indicators for bilateral real exchange rate.
- Remoteness (MTR proxy): constructed as a weighted average distance from trading partners where weights are partner countries’ shares of world GDP (referenced as equation (10) in source).

### Results — Average of countries (aggregate/average treated vs average synthetic)
- Overall average export gain (treated relative to synthetic, cumulative over ten years): 76.4 percentage points.
- Agreement-level average (average treated relative to average synthetic across trade agreements, cumulative over ten years):
  - NAFTA: 80.9 percentage points
  - Mercosur: 23.7 percentage points
  - Group of Three: 9.5 percentage points
- Comparison with world average (Hannan (2016), 104 country pairs, period 1983-1995):
  - World-average export gain over ten years (Hannan 2016): 79.7 percentage points
  - Latin America average export gain is lower than the world average by 3.4 percentage points.
- Interpretation: Aggregate export gains in Latin America are driven predominantly by NAFTA when comparing average treated to average synthetic.

### Results — Individual country-pair findings (treated vs corresponding synthetic)
- Of the 61 country pairs considered, 31 cases have positive export gains (i.e., not all pairs benefitted).
- Alternative country-pair-level averages (average of each country pair’s treated vs its corresponding synthetic, cumulative over ten years):
  - Mercosur (per-pair average): 152.4 percentage points
  - Group of Three (per-pair average): 114.0 percentage points
  - NAFTA (per-pair average): 86.8 percentage points
- NAFTA participants: all participating countries show positive gains when comparing treated to synthetic:
  - United States: exports increased by 44.5 percentage points over ten years (treated relative to synthetic)
  - Canada: 71.6 percentage points
  - Mexico: 144.3 percentage points
- Cross-sample dispersion:
  - Average of individual country gains: 132.6 percentage points
  - Median of individual country gains: 9.9 percentage points
  - Indicates large variation and skewness in country-pair outcomes.
- Goodness-of-fit and reporting:
  - Results are presented alongside normalized root-mean-square deviation (NRMSD) measures of pre-treatment fit; lower NRMSD indicates better pre-treatment fit.
  - The study shows truncated plots for country pairs with NRMSD < 1 to focus on better-fit cases.

### Cross-cutting patterns and comparisons
- Export gains in Latin America are lower than the world average (Hannan 2016) both in aggregate and in the proportion of country pairs benefitting.
- Possible explanations for comparatively lower gains in Latin America cited in the analysis:
  - Relative weakness in trade openness
  - Lower export diversification
  - Lower participation in global value chains
  - Lagging structural reforms
- Income correlation: countries with lower income received higher export boosts from trade agreements (as suggested by results).

### Presentation and inference
- SCM applied separately to each country pair; results reported as:
  - Averages across pairs and agreements (aggregate perspective)
  - Individual country-pair comparisons (heterogeneity perspective)
- Inference relies on placebo/falsification tests rather than standard parametric statistical tests, reflecting SCM practice.

*Source: "14. Export Gains over Ten Years" (section from wp1745 PDF) — empirical analysis using synthetic control method on 61 Latin American country pairs, period 1989-1996.*

### 4755.5 percentage points for the country pair CRI-MEX (Costa Rica’s exports to Mexico

### wp1745 - 4755.5 percentage points for the country pair CRI-MEX (Costa Rica’s exports to Mexico

### Major empirical findings on export gains
- Maximum observed export gain: 4755.5 percentage points for the country pair CRI-MEX (Costa Rica’s exports to Mexico following the trade agreement in 1995).
- When export gains for cases with less than NRMSD of 2 are considered, the average export increase is 144.6 percentage points.
- When pairs with export gains less than 1000 percentage points are considered, the average export gain is -2.7 percentage points.
- Of the 61 country pairs considered, 31 countries show positive gains from trade agreements (51 percent).
- Highest observed export loss: -967.7 percentage points (representing Bolivia’s exports to Ecuador following the trade agreement in 1995).
- World average of countries gaining from trade agreements: 77 percent (comparison benchmark).

### Heterogeneity by country size and income
- Export gains are higher for countries with smaller size, indicating that smaller countries can get a disproportionately higher boost from trade agreements, possibly due to the greater opportunity of trade integration and lower base effect.
- Relationship between exporting country’s income per capita and export gains is less clear-cut; however, there is suggestive evidence that larger export gains have come from countries with lower income per capita.
- Size of bubbles in associated figures represents:
  - Figure 11: nominal GDP (USD million) of exporting country during the year of trade agreement.
  - Figure 12: nominal GDP per capita (USD) of exporting country during the year of trade agreement.

### Aggregate results by trade agreement and region
- Figure 13 (average of individual country gains over ten years across major trade agreements) reports:
  - Mercosur: 152.4 percentage points.
  - Group of Three: 114.0 percentage points.
  - NAFTA: 86.8 percentage points.
  - Average of export gains of trade agreements in Latin America: 132.6 percentage points.
  - World average of export gains over ten years: 190.2 percentage points.
- Individual country gains in Latin America are markedly lower than the world average of 190.2 percentage points over ten years.

### Methodology and goodness-of-fit diagnostics
- Export gains are defined as export growth of treated over ten years relative to synthetic, in cumulative percentage points.
- Goodness of fit is measured by the normalized root-mean-square deviation (NRMSD), computed by dividing the RMSD of each country pair by the average exports over ten years prior to intervention. A smaller NRMSD indicates a better fit.
- Figures 9 and 10 illustrate export gains versus goodness of fit; smaller NRMSD corresponds to better pre-treatment fit between treated and synthetic.

### Placebo (falsification) tests and inference
- Placebo test procedure:
  - Ten treated units are randomly chosen from the sample.
  - For each chosen exporting country A, five country pairs are randomly selected showing exports of A to a country that does not have a trade agreement with A (generating fifty placebo units).
  - SCM is applied to these control units and the evolution of treated unit relative to synthetic is compared to placebo units relative to their synthetics.
- Results of placebo exercise (Figure 14):
  - For the fifty placebo units considered, there are only five cases where the placebo unit shows a higher gain than the treated unit.
  - There is around 10 percent probability that the placebo test shows that the export gains suggested by the treated unit are also manifested in other units chosen at random.
  - Interpretation: the effect estimated by SCM for a country pair affected by the trade agreement is large relative to the effect estimated for a country pair chosen at random, supporting the inferential strength of the approach.

### Comparison with existing literature
- Prior corrected-endogeneity studies find substantial impacts:
  - Baier and Bergstrand (2007, 2009) find trade agreements increase trade by around 100 percent in the long run (studies not limited to Latin America).
  - SCM applied across a broader number of trade agreements (Hannan 2016) yields similar results.
- Mixed evidence for NAFTA and Mercosur:
  - Kehoe (2003) finds applied general equilibrium models drastically underestimated the impact of NAFTA on North American trade.
  - Coulibaly (2009) summarizes studies with both positive and negative trade creation for Mercosur and finds a positive impact using a semi-parametric approach.
- General pattern: more positive results have emerged when endogeneity concerns are addressed with parametric or SCM approaches.

### Explanations and policy implications for lower-than-world export gains in Latin America
- Reasons for lower export gains in Latin America:
  - Trade agreements are not sufficient alone; they need to be accompanied by structural reforms and reduction of nontariff barriers.
  - Latin America is comparatively less trade open, with less export diversification, product sophistication, and product complexity than Asian and Eastern European emerging markets.
  - The region is less integrated in global value chains.
  - Larger economies in Latin America are not acting as dynamic emerging market trade hubs analogous to China in emerging Asia.
- Recommended policy actions:
  - Undertake structural reforms.
  - Reduce nontariff barriers.
  - Improve infrastructure and quality of education.
  - Lower average tariffs.
  - Take initiatives to participate in global value chains where feasible.
  - Increase efforts to penetrate large markets, both through advanced economies and regional market trade hubs.

*Source: wp1745 - 4755.5 percentage points for the country pair CRI-MEX (Costa Rica’s exports to Mexico)*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2017/wp1745.pdf_