## _wp0931

## Source details

**Canonical URL:** [_wp0931](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp0931.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp0931.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp0931.pdf.json)

---

### Key conclusions from the introduction and overall argument
- Main result: The WTO does not boost trade among its members; rather, regional and bilateral preferential trade agreements (PTAs) strongly influence trade flows.
- Identification of WTO trade effects requires comprehensive controls for three sources of omitted variable bias:
  - Multilateral resistance (average trade costs faced by a country).
  - Unobserved bilateral heterogeneity (country-pair fixed effects capturing political, personal, civic affinities).
  - Distinct effects across PTAs (avoid aggregating PTAs into one dummy when impacts differ).
- Reconciling prior literature: When all three controls are included, Rose (2004, 2005), Subramanian and Wei (2007, SW), and Tomz, Goldstein and Rivers (2007, TGR) approaches produce a consistent result of no WTO accession effect on bilateral trade.
- Coding convention detail:
  - Rose/TGR coding: mutually inclusive (dummies identify all PTA and WTO memberships).
  - SW coding: hierarchical, mutually exclusive (PTA membership takes precedence over WTO/GSP in dummy coding).

### Data description and dataset modifications (explicit figures and definitions)
- Dataset dimensions and sampling:
  - 55,831 observations for 177 countries and 11,797 bilateral trade pairs in five year intervals from 1950 to 2000.
- Data source and transformations:
  - Bilateral trade values derived from the IMF’s Direction of Trade Statistics, deflated by the U.S. consumer price index.
  - SW’s preferred dataset excludes observations with import values of less than 500 thousand dollars.
- WTO/GSP coding and adjustments:
  - Use of TGR’s WTO membership definition (de facto members included) throughout the principal analysis.
  - GSP: dataset updated to attribute a value of zero to GSP country-pairs that represent an industrialized country exporting to a developing country (so these pairs are identified as “pure” WTO effects).
  - Luxembourg identified as a member of the European Union (EU) in 2000.
  - Minor coding errors identified by TGR corrected; changes do not affect results qualitatively.
- PTA coverage:
  - Initial extension introduces PTAs contained in SW’s and Rose’s (2005) aggregate PTA dummy.
  - Second stage enlarges set of PTAs to include additional agreements suggested by the PTA literature.
  - Introduction of additional PTAs under SW’s hierarchical coding reduces “1” entries in WTO variables.

### Empirical framework and estimation approach (key model features)
- Dependent variable:
  - Bilateral imports, Imports_mxt (time-varying importer and exporter fixed effects require using imports rather than averaged trade flows).
- Multilateral resistance controls:
  - Time-varying importer fixed effects, δ_mt, and exporter fixed effects, λ_xt, capture multilateral resistance (importer/exporter price indices).
- Unobserved bilateral heterogeneity:
  - Country-pair fixed effects are included in extended specifications to control for omitted bilateral affinities.
- PTA coding and disaggregation:
  - SW’s mutually exclusive coding attributes trade creation to PTAs when PTA and WTO/GSP membership overlap.
  - Extension replaces aggregate PTA dummy with a matrix of individual PTA dummies; β4 becomes a vector capturing individual PTA trade impacts.
- Control variables (all 0-1 dummies except log distance):
  - WTOIndustrial_mt, WTODeveloping_mt, GSP_mt, PTA_mxt (or individual PTAs), CurrencyUnion_mx (CU_mxt), CurColony_mxt, EverColony_mx, CommonColonizer_mx, CommonLanguage_mx, CommonNation_mx, Border_mx, and log Distance_mx.
- SW baseline regression structure:
  - Baseline corresponds to equation (1) with mutually exclusive coding of trade agreement dummies and time-varying importer/exporter fixed effects.

### Main empirical findings (selected quantitative results reported exactly)
- SW-analogue (Regression 1, corrected dataset):
  - Implied trade benefit for industrialized countries from WTO membership: 187% (=e^{1.053}-1).
  - No significant WTO impact for developing countries in that specification.
  - Aggregate PTA effect (Regression 1, Table 1a): bilateral imports increase 234% (=e^{1.205}-1) for the average PTA and the average PTA member.
  - “Pure” PTA effects (net of any WTO effects), Regression 1, Table 1:
    - Industrialized country PTA members: bilateral trade increases 16% (=e^{0.152}-1).
    - Developing country PTA members: bilateral trade increases 214% (=e^{1.143}-1).
- Impact of introducing individual PTA dummies:
  - Allowing for individual PTA effects heavily impacts multilateral trade dummies and corrects potential upward bias in prior aggregate specifications.
  - Under mutually exclusive coding, omission of relevant PTAs that are trade creating and whose members are also WTO members can upwardly bias the WTO coefficient.
- Regression comparisons and key coefficients (selected exact numbers preserved):
  - WTO induced trade creation for industrialized countries falls from 187% in regression 1 to 80% (=e^{0.588}-1) in regression 2.
  - Individual PTA magnitudes (examples from extended specifications):
    - NAFTA: 770% (=e^{2.162}-1) in one specification; other regressions show large but varying NAFTA coefficients (e.g., raw NAFTA_mxt: 2.117 *** (0.258) in Table 1a).
    - AFTA: 76% (=exp 0.568 -1) in one reported specification.
    - SPARTECA: 124% (= e^{0.807}-1) reported in regression 12 discussion.
    - EU: 37% (= e^{0.312}-1) increase in trade due to EU accession once unobserved bilateral heterogeneity is controlled for.
    - APEC: Regression 9 with multilateral resistance yields 123% (= e^{0.802}-1); preferred regression 12 reduces APEC to 28% (= e^{0.244}-1).
  - Dataset counts (reported):
    - In dataset 1,593 of the 55,831 observations are country pairs that are contemporaneous members of a common PTA; of these 1,593 observations, in 1,369 cases the importer is also a WTO member.
    - For the broadest PTA set, 3,253 of the 55,813 observations are country pairs that are also members of a common PTA. Of these 3,253 observations, 2,700 of the importers are also contemporaneous WTO members.
- Effects after adding country-pair fixed effects (α_mx):
  - Regression 4 (replication of SW specification with country-pair fixed effects):
    - WTO trade creation for industrialized countries falls from 178% in regression 1 to 48% (=e^{0.393}-1) in regression 4; significance fails to reach the 1% level.
  - Inclusion of specific PTA trade effects in regressions 5 and 6 negates all WTO trade effects for industrialized countries.
  - For developing countries, regressions 4-6 show GSP and WTO effects are hardly impacted by unobservable heterogeneity and remain closely aligned to estimates in regressions 1-3.
- Preferred specifications and summary:
  - Regression 12 (inclusive-coding analogue of regression 6) and regression 6 (exclusive coding) are the authors’ preferred specifications, containing all three key controls.
  - F-tests favor the most extensive specification that controls for multilateral resistance, unobserved bilateral heterogeneity, and individual PTA effects.
  - In preferred specifications, no WTO trade effects are found independent of coding convention.

### Methodological implications and robustness
- Omitting any of the three controls (multilateral resistance, unobserved bilateral heterogeneity, distinct PTA effects) leads to upward-biased WTO coefficient estimates.
- Results are robust to:
  - Using TGR’s de facto WTO membership definition.
  - Both Rose/TGR mutually inclusive coding and SW hierarchical mutually exclusive coding, once comprehensive controls are included.
- SW-style mutually exclusive coding can produce biased WTO estimates under two conditions:
  - (a) industrialized and developing PTAs differ considerably in their trade effects, and
  - (b) PTA effects are constrained to one aggregate PTA coefficient.
- Hierarchical coding mechanics:
  - The industrialized WTO coefficient in SW’s preferred regression functions as an implicit industrialized-PTA dummy due to coding interactions and accession sequencing; explicit partitioning of the aggregate PTA dummy into industrialized and developing PTA dummies produces perfect multicollinearity in regression 1a and forces the industrial PTA dummy to be dropped.
- Country-pair fixed effects:
  - Account for time-invariant bilateral fundamentals that otherwise bias PTA and WTO coefficients.
  - Inclusion substantially reduces many PTA estimates (exceptions where estimates rise include EU, CACM, LAIA).
- Net interpretation of PTA coefficients:
  - PTA coefficients in preferred regressions measure PTA-internal trade increases relative to PTA-external trade after the agreement; these are mixtures of trade creation and trade diversion and cannot be separately identified due to multicollinearity with multilateral resistance controls.

### Appendix A — Membership in considered Preferential Trading Arrangements (selected entries)
- ANZCERTA — Start: 1983; Members: Australia, New Zealand
- APEC — Start: 1989; Members: Australia, Brunei, Canada, China (1991), Chile (1994), Taiwan (1991), Hong Kong (1991), Indonesia, Japan, South Korea, Malaysia, Mexico (1993), New Zealand, Papua New Guinea (1993), Peru (1998), Philippines, Singapore, Thailand, United States, Vietnam (1998).
- AFTA — Start: 1967; Members include Brunei (1984), Cambodia (1998), Indonesia, Laos (1997), Malaysia, Myanmar (1997), the Philippines, Singapore, Thailand, Vietnam (1995).
- CACM — Start: 1960; Members: Costa Rica (1963), El Salvador, Guatemala, Honduras, Nicaragua.
- CARICOM — Start: 1968; Members listed in Table A1.
- EEA — Start: 1994; Members listed in Table A1.
- EFTA — Start: 1960; Members listed in Table A1.
- EU — Start: 1958; Members listed in Table A1.
- LAIA/LAFTA, MERCOSUR, NAFTA, SPARTECA — Start years and membership coverage listed in Table A1.
- Bilateral PTAs considered: comprehensive list in Table A2 (examples include US - Israel; numerous EC/EST/Poland/Slovenia bilateral links; Canada – Chile; Canada – Israel; various Georgia, Kyrgyz Republic, Bulgaria, Hungary, Mexico bilateral agreements).

*Source: _wp0931 - References..............................................................................................................*

### References..............................................................................................................

### _wp0931 - References..............................................................................................................

### Key conclusions from the introduction and overall argument
- Main result: The WTO does not boost trade among its members; rather, regional and bilateral preferential trade agreements (PTAs) strongly influence trade flows.
- Identification of WTO trade effects requires comprehensive controls for three sources of omitted variable bias:
  - Multilateral resistance (average trade costs faced by a country).
  - Unobserved bilateral heterogeneity (country-pair fixed effects capturing political, personal, civic affinities).
  - Distinct effects across PTAs (avoid aggregating PTAs into one dummy when impacts differ).
- Reconciling prior literature: When all three controls are included, Rose (2004, 2005), Subramanian and Wei (2007, SW), and Tomz, Goldstein and Rivers (2007, TGR) approaches produce a consistent result of no WTO accession effect on bilateral trade.
- Coding convention detail:
  - Rose/TGR coding: mutually inclusive (dummies identify all PTA and WTO memberships).
  - SW coding: hierarchical, mutually exclusive (PTA membership takes precedence over WTO/GSP in dummy coding).

### Data description and dataset modifications (explicit figures and definitions)
- Dataset dimensions and sampling:
  - 55,831 observations for 177 countries and 11,797 bilateral trade pairs in five year intervals from 1950 to 2000.
- Data source and transformations:
  - Bilateral trade values derived from the IMF’s Direction of Trade Statistics, deflated by the U.S. consumer price index.
  - SW’s preferred dataset excludes observations with import values of less than 500 thousand dollars.
- WTO/GSP coding and adjustments:
  - Use of TGR’s WTO membership definition (de facto members included) throughout the principal analysis.
  - GSP: dataset updated to attribute a value of zero to GSP country-pairs that represent an industrialized country exporting to a developing country (so these pairs are identified as “pure” WTO effects).
  - Luxembourg identified as a member of the European Union (EU) in 2000.
  - Minor coding errors identified by TGR corrected; changes do not affect results qualitatively.
- PTA coverage:
  - Initial extension introduces PTAs contained in SW’s and Rose’s (2005) aggregate PTA dummy (examples listed in source).
  - Second stage enlarges set of PTAs to include additional agreements suggested by the PTA literature (examples listed in source).
  - Note: introduction of additional PTAs under SW’s hierarchical coding reduces “1” entries in WTO variables.

### Empirical framework and estimation approach (key model features)
- Dependent variable:
  - Bilateral imports, Imports_mxt (time-varying importer and exporter fixed effects require using imports rather than averaged trade flows).
- Multilateral resistance controls:
  - Time-varying importer fixed effects, δ_mt, and exporter fixed effects, λ_xt, capture multilateral resistance (importer/exporter price indices).
- Unobserved bilateral heterogeneity:
  - Country-pair fixed effects are included in extended specifications to control for omitted bilateral affinities.
- PTA coding and disaggregation:
  - SW’s mutually exclusive coding attributes trade creation to PTAs when PTA and WTO/GSP membership overlap.
  - Extension replaces aggregate PTA dummy with a matrix of individual PTA dummies; β4 becomes a vector capturing individual PTA trade impacts.
- Control variables (all 0-1 dummies except log distance):
  - WTOIndustrial_mt, WTODeveloping_mt (split WTO dummy), GSP_mt, PTA_mxt (or individual PTAs), CurrencyUnion_mx, CurColony_mxt, EverColony_mx, CommonColonizer_mx, CommonLanguage_mx, CommonNation_mx, Border_mx, and log Distance_mx.
- SW baseline regression structure (preferred specification analogue):
  - Baseline corresponds to equation (1) in the source with mutually exclusive coding of trade agreement dummies and time-varying importer/exporter fixed effects.

### Main empirical findings (selected quantitative results reported exactly)
- Regression 1 (SW analogue with dataset corrections):
  - Implied trade benefit for industrialized countries from WTO membership: 187% (=e^{1.053}-1).
  - No significant WTO impact for developing countries in that specification.
  - Aggregate PTA effect (from SW regressions, Regression 1, Table 1a): bilateral imports increase 234% (=e^{1.205}-1) for the average PTA and the average PTA member.
  - “Pure” PTA effects (net of any WTO effects), Regression 1, Table 1:
    - Industrialized country PTA members: bilateral trade increases 16% (=e^{0.152}-1).
    - Developing country PTA members: bilateral trade increases 214% (=e^{1.143}-1).
- Impact of introducing individual PTA dummies:
  - Allowing for individual PTA effects heavily impacts multilateral trade dummies and corrects potential upward bias in prior aggregate specifications.
  - Under mutually exclusive coding, omission of relevant PTAs that are trade creating and whose members are also WTO members can upwardly bias the WTO coefficient.

### Methodological implications and robustness
- Omitting any of the three controls (multilateral resistance, unobserved bilateral heterogeneity, distinct PTA effects) leads to upward-biased WTO coefficient estimates.
- Results are robust to:
  - Using TGR’s de facto WTO membership definition.
  - Both Rose/TGR mutually inclusive coding and SW hierarchical mutually exclusive coding, once comprehensive controls are included.
- The WTO dummy split into industrialized and developing groups (as in SW) does not drive the main result when full controls are applied, although SW-style coding can produce biased WTO estimates under two conditions:
  - (a) industrialized and developing PTAs differ considerably in their trade effects, and
  - (b) PTA effects are constrained to one aggregate PTA coefficient.
- The most comprehensive specification (controls for multilateral resistance, bilateral fixed effects, and individual PTA effects) returns the most reasonable PTA trade effects and reconciles prior divergent findings.

*Source: _wp0931 - References..............................................................................................................*

### 1. Most notably, we find that the economic and statistical significance of WTO membership for

### _wp0931 - 1. Most notably, we find that the economic and statistical significance of WTO membership for

### Key empirical findings on WTO membership and PTA effects
- WTO induced trade creation for industrialized countries falls from 187% in regression 1 to 80% (=e 0.588 -1) in regression 2.
- The sizable trade creation attributed to WTO membership in regression 1 is more accurately associated with individual PTAs; individual PTA estimates are precise and the F-Statistic rejects regression 1 in favor of regression 2.
- Effects of individual PTAs are all trade creating and highly statistically significant (with the exception of the EU, discussed elsewhere).
- Magnitudes of PTA coefficients vary dramatically:
  - 770% (=e 2.162 -1) for NAFTA.
  - 76% (=exp 0.568 -1) for the ASEAN free trade agreement (AFTA).
- Regression 3 (which controls for additional PTAs not included in SW or Rose 2005) shows EFTA, EEA, LAIA, and APEC as highly trade creating.
- Once individual PTA effects are accounted for with the most comprehensive set of PTAs, the SW result of positive WTO effects in industrial countries vanishes completely.
- Dataset counts (as reported):
  - In dataset 1,593 of the 55,831 observations are country pairs that are contemporaneous members of a common PTA; of these 1,593 observations, in 1,369 cases the importer is also a WTO member.17
  - For the broadest PTA set, 3,253 of the 55,813 observations are country pairs that are also members of a common PTA. Of these 3,253 observations, 2,700 of the importers are also contemporaneous WTO members.18

### Findings for developing countries
- Allowing for individual PTA effects yields:
  - WTO membership effects for developing countries estimated to be either non-existent (regressions 1, 2) or negative and statistically significant (-21% = e -0.235 -1 in regression 3).
- PTA coefficients indicate strong trade creation for PTAs involving developing countries.
- Hypothesis: developing countries may have reoriented import activity toward PTA partners after joining PTAs, producing trade creation within PTAs but possibly causing trade diversion away from WTO partners; given SW’s hierarchical, mutually exclusive coding, such reorientation could bias the WTO estimate for developing countries downward.19
- Largest levels of trade creation in regression 3 are observed for PTAs consisting of developing nations (CACM, CARICOM, MERCOSUR, NAFTA-Developing and SPARTECA-Developing), where PTA-internal trade is estimated to be roughly 350% (=e 1.5 -1) greater than PTA-external trade.

### Hierarchical, mutually exclusive coding and the implicit industrialized PTA dummy
- PTA trade creation is heterogeneous: bilateral PTA trade is estimated to be significantly larger for PTAs that consist of developing countries than for PTAs among industrialized nations.
- Ignoring these differentials introduces a specific omitted variable bias into SW-style regressions.
- The industrialized WTO coefficient in SW’s preferred regression (regression 1) functions as an error-correction term that implicitly splits SW’s aggregate PTA coefficient into:
  - an industrialized-PTA component (low trade creation)
  - a developing-PTA component (high trade creation)
- Mechanics generating the implicit industrialized PTA dummy:
  - Interaction among SW’s coding convention, industrialized countries’ WTO/PTA accession sequencing, and multilateral resistance controls (time-varying importer dummies).
  - All industrialized importers joined WTO before joining a PTA; under SW’s coding, after WTO accession industrialized importer observations feature “1” for either the WTO/GSP dummy or the PTA dummy, making the linear combination of these two dummies perfectly collinear with the time-varying importer dummy.
- Empirical confirmation:
  - Explicitly partitioning the aggregate PTA dummy into industrialized and developing PTA dummies (regression 1a) produces perfect multicollinearity and forces the industrial PTA dummy to be dropped; regression 1a results are identical to regression 1, but industrialized PTA trade creation can be identified from the WTO importer dummy.20
  - Regression 1’s industrialized WTO coefficient reflects industrialized PTA trade creation, not WTO trade creation.
  - Regression 1a shows that developing countries’ PTA effects exclusively determined the “aggregate PTA” effect in regression 1 (i.e., the aggregate effect was purged of industrialized countries’ effects, which are expressed through the industrialized WTO dummy).

### Unobserved bilateral heterogeneity, multilateral resistance, and PTA estimates
- Trade agreements often form between partners with naturally elevated bilateral trade due to unobserved characteristics; when controls do not account for all unobserved bilateral heterogeneity and omitted variables favor trade, WTO and PTA coefficients are biased upwards.
- Prior literature flagged AP, LAIA and APEC estimates as suspiciously high relative to small tariff reductions; current estimates in regressions 1-3 are similar in magnitude to previous literature, indicating robustness to inclusion of individual PTA effects and multilateral resistance.
- To account fully for unobserved bilateral heterogeneity, the authors introduce country-pair fixed effects (α mx ), replacing the intercept in equation (1) and yielding equation (2) where time-invariant regressors are absorbed into pair-specific fixed effects.

### WTO trade impact after controlling for multilateral resistance and bilateral heterogeneity
- Regressions 4-6 (analogues of regressions 1-3 with country-pair fixed effects) examine:
  - whether WTO trade effects are influenced by unobserved bilateral heterogeneity, and
  - whether PTA effects are reduced to plausible ranges.
- Regression 4 (replication of SW specification with country-pair fixed effects) results:
  - WTO trade creation for industrialized countries falls from 178% in regression 1 to 48% (=e 0.393 -1) in regression 4.
  - The significance of this industrialized WTO effect fails to reach the 1% level.
- Inclusion of specific PTA trade effects in regressions 5 and 6 negates all WTO trade effects for industrialized countries.
- Even the simple disaggregation of SW’s aggregate PTA dummy into individual agreements (regression 5) is sufficient to neutralize industrialized WTO trade effects once bilateral heterogeneity is accounted for.
- For developing countries, regressions 4-6 show that the GSP and WTO effects are hardly impacted by unobservable heterogeneity and remain closely aligned to estimates in regressions 1-3.

*Source: _wp0931 - 1. Most notably, we find that the economic and statistical significance of WTO membership for*

### Section 2. See appendix B for results based on the original SW dataset (without TGR but with country-pair fixed

### Section 2. See appendix B for results based on the original SW dataset (without TGR but with country-pair fixed effects)

### Major methodological extensions and estimation strategy
- Key controls introduced and unified across specifications:
  - Individual PTA effects.
  - Comprehensive multilateral resistance controls.
  - Unobserved bilateral heterogeneity (country-pair fixed effects).
- Two WTO coding conventions compared:
  - Mutually exclusive coding (SW-style).
  - Mutually inclusive coding (Rose-style).
- Purpose of inclusive coding robustness analysis:
  - Determine whether SW hierarchical, mutually exclusive WTO coding underestimates WTO effects by attributing increases to PTAs rather than to both WTO and PTA membership.
  - Extend Rose (2004, 2005) by introducing disaggregated PTAs and comprehensive multilateral resistance controls.

### Statistical diagnostics and specification tests
- F statistics confirm the importance of:
  - Comprehensive country-pair fixed effects.
  - Individual PTA effects.
  - Reported significance levels exceed 0.001%.
- Effect of added controls on aggregate PTA trade creation:
  - Average trade creation across PTAs drops from 234% in regression 1 to 123% in regression 4 (Table 1a).
- Preferred specifications:
  - Regression 12 (inclusive-coding analogue of regression 6) and regression 6 (exclusive coding) are the authors’ preferred specifications, containing all three key controls.
  - F-tests favor the most extensive specification that controls for all three dimensions of omitted variable bias.

### WTO trade effects (mutually inclusive WTO coding) — main findings
- Regression 7 (baseline analogue to Rose, 2004):
  - Trade creation due to WTO membership for both industrialized and developing countries is insignificant.
  - Aggregating industrial and developing WTO dummies into one WTO dummy yields an insignificant combined effect.
  - Rose’s GSP effect is eliminated when multilateral resistance controls are inserted.
- Regressions 8 and 9 (inclusive-coding analogues of regressions 2 and 3):
  - Overturn Rose’s (2005) result of a statistically significant WTO effect once comprehensive multilateral resistance controls and individual PTA effects are included.
- Regressions 10–12 (analogs to regressions 4–6):
  - With inclusive coding no WTO trade effects are found in any of these regressions.
  - Regressions 12 and 6 (preferred specifications) confirm the absence of trade gains from WTO membership independent of the coding convention.
- Synthesis on coding conventions:
  - Mutually inclusive coding yields WTO coefficients that are statistically and economically insignificant even without introducing individual PTA effects.
  - Mutually exclusive coding can yield biased WTO dummies when (1) industrialized and developing PTAs differ considerably in their trade effects and (2) individual PTA effects are constrained to an average coefficient associated with one aggregate PTA dummy.
  - When variables represent net effects (Rose-style), omitted variable bias risk is reduced; mutually exclusive coding may confound PTA and WTO effects.

### Individual PTA trade effects — sensitivity to controls and preferred estimates (regression 12)
- General interpretation of PTA coefficients in regression 12:
  - Percentage increases indicate how much PTA-internal trade increased relative to trade with non-members after the agreement was instituted.
  - These percentage increases are mixtures of trade creation and trade diversion.
  - Multicollinearity with multilateral resistance controls prevents separate identification of trade creation versus diversion.
- Broad patterns and magnitudes (regression 12 preferred):
  - Most PTAs’ net trade creation estimated to range between 30% and 80%.
  - Central and Latin American PTAs (CACM, CARICOM, MERCOSUR and LAIA) show the largest relative trade increases; these are the only PTAs with net increases in trade creation of over 100% (with the exception of SPARTECA).
  - SPARTECA reports an increase of 124% (= e^0.807 -1).
  - EU shows a statistically significant 37% (= e^0.312 -1) increase in trade due to EU accession once unobserved bilateral heterogeneity is controlled for.
  - EU members also automatically reaped another 34% trade benefit when they became members of the EEA in 1994 (since the overwhelming majority of EEA members are EU countries).
  - APEC:
    - Regression 9 (with multilateral resistance) lowers APEC coefficient to 123% (= e^0.802 -1).
    - Preferred regression 12 (controlling for multilateral resistance, unobserved bilateral heterogeneity and individual PTA effects) reduces APEC’s trade creation estimate to 28% (= e^0.244 -1).
- Effects of adding country-pair fixed effects:
  - Comparing regression 9 to regression 12, most PTA estimates are substantially reduced when country-pair fixed effects are included (exceptions: EU, CACM, LAIA where estimated trade creation increases).
  - Example: NAFTA’s suspiciously large net trade creation estimate of 230% in regression 9 is reduced to insignificance after controlling for unobserved bilateral heterogeneity.
- Interpretation of increases for PTA groups with strong bilateral fundamentals:
  - For CACM and some others, common language, common colonizer, and low distance led gravity models to systematically overpredict trade; country-pair fixed effects resolve such overprediction and yield higher estimated PTA impacts for those PTAs.
  - For the EU, despite large absolute trade volumes among members, the net trade creation relative to trade with non-members is estimated at 37% once bilateral heterogeneity is controlled for; harmonization and market size effects are plausible drivers.

### Conclusions and implications
- Main empirical conclusions:
  - No robust evidence of trade-creating effects from WTO membership once comprehensive multilateral resistance, unobserved bilateral heterogeneity, and individual PTA effects are controlled for.
  - Inclusion of individual PTA trade effects, multilateral resistance, and country-pair fixed effects is essential to obtain unbiased estimates of trade agreement impacts.
  - Many previously reported large trade effects for certain non-tariff-reducing agreements (e.g., APEC) are greatly attenuated when the most comprehensive controls are applied.
- Broader implications:
  - The unified approach reconciles previously diverging results in the WTO/PTA literature (e.g., Rose, 2004, 2005; Subramanian and Wei, 2007; Tomz, Goldstein and Rivers, 2007) by showing omitted variable bias as a likely driver of differences.
  - Country-pair specific unobservables are an important determinant of bilateral trade and are not fully addressed in current trade theories; accounting for them materially changes estimated PTA effects.
  - Multilateral resistance controls are theoretically motivated and necessary but by themselves are insufficient; country-pair fixed effects remain crucial.

*Source: _wp0931 - Section 2. See appendix B for results based on the original SW dataset (without TGR but with country-pair fixed effects).*

### References

### _wp0931 - References

### References (selected)
- Aitken, N.D., 1973. “The Effect of the EEC and EFTA on European Trade: A Temporal Cross-Section Analysis,” American Economic Review, 63 (5), pp. 881-892.
- Anderson, J., 1979. “A theoretical foundation for the gravity equation,” American Economic Review, 69 (1), pp. 106-116.
- Anderson, J., van Wincoop, E., 2003. “Gravity with Gravitas: a solution to the border puzzle,” American Economic Review, 93 (1), pp. 170-192.
- Baier, S.L., Bergstrand, J.H., 2001. “The growth of world trade: tariffs, transport costs, and income similarity,” Journal of International Economics, 53 (1), pp. 1-27.
- Baier, S.L., Bergstrand, J.H., 2002. “On the endogeneity of international trade flows and free trade agreements,” Manuscript, August 2002.
- Baier, S.L., Bergstrand, J.H., 2007. “Do free trade agreements actually increase members’ international trade?” Journal of International Economics, 71, pp. 72-95.
- Baldwin, R., 2006. “The euro’s trade effects”, Working Paper prepared for ECB Workshop “What effects is EMU having on the euro area and its member countries?” ECB Working Paper 594.
- Baldwin, R., Taglioni, D., 2006. “Gravity for Dummies and Dummies for Gravity Equations”, NBER Working Paper 12516.
- Bergstrand, J.H., 1985. “The gravity equation in international trade: some microeconomic foundations and empirical evidence” Review of Economics and Statistics, 67 (3), pp. 474-481.
- Carrere, C., 2006. “Revisiting the effects of regional trade agreements on trade flows with proper specification of the gravity model,” European Economic Review, 50, pp. 223-247.
- Cheng, I., Wall, H.J., 2005. “Controlling for Heterogeneity in Gravity Models of Trade and Integration”, Federal Reserve Bank of St. Louis Review, 87 (1), pp. 49-63.
- Deardorff, A.V., 1998. “Determinants of bilateral trade: does gravity work in a classical world”, in: Frankel, J. (Ed.), The Preferentialization of the World Economy. University of Chicago Press, Chicago, pp. 7–22.
- Eaton, J., Kortum, S., 2002. “Technology, geography, and trade,” Econometrica, 70 (5), pp. 1741-1779.
- Egger, P., 2000. “A note on the proper econometric specification of the gravity equation,” Economics Letters, 66, pp. 25-31.
- Egger, P., Pfaffermayr, M., 2003. “The proper panel econometric specification of the gravity equation: A three-way model with bilateral interaction effects,” Empirical Economics, 28, pp. 571-580.
- Eicher, T. S., Henn, C., Papageorgiou, C., 2007. “Trade Creation and Diversion Revisited: Accounting for Model Uncertainty and Unobserved bilateral heterogeneity,” University of Washington Working Paper 2007-18
- Frankel, J., 1992. “Is Japan Creating a Yen Bloc in East Asia and the Pacific?” NBER Working Paper 4050.
- Frankel, J., 1997. Preferential Trading Blocs in the World Trading System, Institute for International Economics, Washington, D.C.
- Frankel, J., Stein, E., Wei, S.J., 1995. “Trading blocs and the Americas: the natural, the unnatural and the supernatural?” Journal of Development Economics, 47, pp. 61-95.
- Freund, C., 2000. “Different paths from free trade: the gains from regionalism,” The Quarterly Journal of Economics, 115, pp. 1317–1341.
- Ghosh, S., Yamarik, S., 2004. “Are preferential trade agreements trade creating? An application of extreme bounds analysis,” Journal of International Economics, 63, pp. 369-395.
- Glick, R., Rose, A.K., 2002. “Does a currency union affect trade? The time-series evidence,” European Economic Review, 46, pp. 1125-1151.
- Greene, W.H., 2003. Econometric Analysis, 5th Edition, Prentice-Hall, Princeton, NJ.
- Helpman, E., Melitz, M., Rubinstein, Y., 2007. “Estimating Trade Flows: Trading Partners and Trading Volumes,” NBER Working Papers No. 12927
- Felbermayr, G. J., Kohler, W., 2006. "Exploring the Intensive and Extensive Margins of World Trade," Review of World Economics (Weltwirtschaftliches Archiv), vol. 127(4), pages 642-674, December.
- Liu, S., 2007, “GATT/WTO Promotes Trade Strongly: Sample Selection and Model Specification” Kennesaw State University, mimeo.
- Melitz, M., 2003. “The impact of trade on intra-industry reallocations and aggregate industry productivity,” Econometrica, 71 (6), pp. 1695-1725.
- Rose, A.K., 2000. “One money, one market? The Effects of Common Currencies on International Trade,” Economic Policy, 15, pp. 7-46.
- Rose, A.K., 2004. “Do We Really Know That the WTO Increases Trade?” American Economic Review, 13 (4), pp. 682-698.
- Soloaga, I., Winters, L.A., 2001. “Regionalism in the nineties: what effect on trade?” North American Journal of Economics and Finance, 12, pp. 1-29.
- Subramanian, A., Wei, S.J., 2007. “The WTO Promotes Trade Strongly, But Unevenly,” Journal of International Economics, forthcoming.
- Tomz, M., Goldstein, J. L., Rivers, D., 2007. “Do We Really Know that the WTO Increases Trade? Comment,” American Economic Review, 97 (5), pp. 2005-2018.

### Empirical findings — Table 1: WTO and PTA Effects (Hierarchical, Mutually Exclusive Coding)
- Dependent variable: bilateral imports.
- Regression specifications: columns labeled Regression # 1, 1a, 2, 3, 4, 5, 6; Country-Pair FE: No for regressions 1–4, Yes for regressions 5–7.
- Adj R2 values:
  - 0.7411, 0.7411, 0.7415, 0.7430, 0.8747, 0.8751, 0.8760.
- F Stat vs. Regr.# reported (Prob>F): 0.00 for reported comparisons.
- Selected coefficients (coefficient (standard error)) — preserve sign and significance markers:
  - GSP_mt (ind. imp. grants):
    - -0.209 *** (0.041)
    - -0.209 *** (0.041)
    - -0.202 *** (0.041)
    - -0.202 *** (0.040)
    - -0.242*** (0.049)
    - -0.233*** (0.049)
    - -0.181 *** (0.047)
  - WTODev_mt (dev. imp.):
    - 0.062 (0.134)
    - 0.062 (0.134)
    - -0.045 (0.131)
    - -0.235 *** (0.075)
    - 0.210 (0.147)
    - -0.207 (0.148)
    - -0.254 *** (0.082)
  - WTOInd_mt (ind. imp.):
    - 1.053 *** (0.141)
    - -0.152 *** (0.056)
    - 0.588 *** (0.195)
    - -0.035 (0.092)
    - 0.393 ** (0.165)
    - 0.124 (0.189)
    - -0.068 (0.092)
  - PTA_mxt ind. composite (aggregate):
    - 0.152 *** (0.056)
    - 0.361 *** (0.055)  [appears as composite reported]
  - PTA_mxt dev. composite (aggregate):
    - 1.143 *** (0.073)
    - 1.143 *** (0.073)
    - 0.545 *** (0.082)  [composite reported]
  - BilateralPTA_mxt (ind. importer / dev. importer):
    - ind. importer: 0.072 (0.109); 0.351 *** (0.085); -0.169 * (0.101); -0.054 (0.084)
    - dev. importer: 0.705 *** (0.085); 0.551 *** (0.080); 0.161 * (0.086); 0.132 * (0.180)
  - NAFTA_mxt (ind. importer / dev. importer):
    - ind. importer: 1.529 *** (0.188); 1.276 *** (0.161); 0.155 (0.139); 0.257 (0.169)
    - dev. importer: 2.162 *** (0.216); 1.476 *** (0.172); 0.485 *** (0.158); 0.443 ** (0.176)
  - Regional PTAs (selected):
    - EU_mxt: 0.065 (0.056); -0.139 ** (0.067); 0.473 *** (0.062); 0.306 *** (0.068)
    - CACM_mxt: 1.346 *** (0.191); 1.509 *** (0.173); 1.907 *** (0.253); 1.938 *** (0.242)
    - CARICOM_mxt: 1.437 *** (0.143); 1.446 *** (0.143); 1.071 *** (0.253); 1.070 *** (0.253)
    - MERCOSUR_mxt: 1.530 *** (0.197); 1.552 *** (0.202); 0.982 *** (0.227); 1.068 *** (0.243)
    - AFTA_mxt: 0.568 *** (0.159); 0.391 ** (0.172); -0.035 (0.196); 0.049 (0.207)
    - ANZCERTA_mxt: 1.875 *** (0.123); 1.647 *** (0.135); 0.747 ** (0.293); 0.770 *** (0.297)
    - SPARTECA_mxt (ind. importer / dev. importer):
      - ind. importer: 0.824 *** (0.236); 1.238 *** (0.224); 0.587 *** (0.207); 0.742 *** (0.178)
      - dev. importer: 1.457 *** (0.210); 1.438 *** (0.217); 0.917 *** (0.176); 0.927 *** (0.170)
    - EFTA_mxt: 0.491 *** (0.091); 0.200 *** (0.075)
    - EEA_mxt: 0.508 *** (0.096); 0.342 *** (0.105)
    - AP_mxt: 0.638 *** (0.183); 0.677 *** (0.223)
    - LAIA_mxt: 0.267 ** (0.105); 1.277 *** (0.191)
    - APEC_mxt (ind. importer / dev. importer):
      - ind. importer: 0.651 *** (0.086); 0.107 (0.092)
      - dev. importer: 0.851 *** (0.070); 0.293 *** (0.079)
  - CU_mxt (Currency union):
    - 0.615 *** (0.082); 0.615 *** (0.082); 0.616 *** (0.082); 0.643 *** (0.082); 0.291 ** (0.113); 0.310 *** (0.114); 0.317 *** (0.114)
  - Historical/geographical controls (selected):
    - CurColony_mxt (Current colony): 0.760 *** (0.145); repeated coefficients 0.760 *** (0.145); 0.765 *** (0.145); 0.765 *** (0.145); 0.376 *** (0.137); 0.356 *** (0.137); 0.377 *** (0.137)
    - EverColony_mx (Ever colony): 1.164 *** (0.065); 1.164 *** (0.065); 1.156 *** (0.065); 1.161 *** (0.065)
    - ComColonizer_mx (Common colonizer): 0.578 *** (0.050); 0.578 *** (0.050); 0.567 *** (0.051); 0.573 *** (0.051)
    - ComLang_mx (Common lang.): 0.142 *** (0.031); 0.142 *** (0.031); 0.141 *** (0.031); 0.125 *** (0.032)
    - ComNat_mx (Same nation): -0.522 (0.544); -0.522 (0.544); -0.529 (0.545); -0.544 (0.544)
    - Border_mx (Common border): 0.024 (0.065); 0.024 (0.065); 0.027 (0.065); 0.040 (0.065)
    - Dist_mx (Log of distance): -0.965 *** (0.017); -0.965 *** (0.017); -0.963 *** (0.017); -0.932 *** (0.017)
- Notes: *, **, *** are 10%, 5%, 1% significance levels; standard errors in parenthesis, fixed effect coefficients are suppressed. GSP and PTA coefficients are composite coefficients that report GSP and PTA effects net of WTO effects and are calculated by the Delta method (the original regression output is in Table 1a). All regressions include time-varying importer and exporter fixed effects.

### Empirical findings — Table 1a: Raw Regression Output (including WTO effects)
- Dependent variable: bilateral imports. Country-Pair FE: No for regressions 1–4, Yes for regressions 5–7.
- Selected raw coefficient estimates (coefficient (standard error)):
  - GSP_mt (ind. imp. grants):
    - 0.844 *** (0.142)
    - -0.360*** (0.115)
    - 0.385 ** (0.194)
    - -0.237 ** (0.093)
    - 0.151 (0.166)
    - -0.110 (0.187)
    - -0.249 *** (0.095)
  - PTA_mxt Ind./Dev (aggregate):
    - 1.205 *** (0.131)
    - 0.755 *** (0.157)  [reported in corresponding column]
  - PTA_mxt Dev. only (aggregate):
    - 1.205 *** (0.131)
  - BilateralPTA_mxt:
    - 0.660 *** (0.164)
    - 0.316 *** (0.095)
    - -0.045 (0.167)
    - -0.122 (0.097)
  - NAFTA_mxt:
    - 2.117 *** (0.258)
    - 1.241 *** (0.148)
    - 0.279 (0.215)
    - 0.189 (0.152)
  - EU_mxt:
    - 0.652 *** (0.199)
    - -0.174 * (0.097)
    - 0.596 *** (0.196)
    - 0.238 ** (0.107)
  - CACM_mxt:
    - 1.301 *** (0.168)
    - 1.274 *** (0.166)
    - 1.700 *** (0.244)
    - 1.684 *** (0.239)
  - CARICOM_mxt:
    - 1.391 *** (0.209)
    - 1.211 *** (0.165)
    - 0.864 *** (0.288)
    - 0.817 *** (0.264)
  - MERCOSUR_mxt:
    - 1.485 *** (0.246)
    - 1.317 *** (0.198)
    - 0.776 *** (0.271)
    - 0.814 *** (0.233)
  - AFTA_mxt:
    - 0.523 *** (0.188)
    - 0.156 (0.154)
    - -0.241 (0.237)
    - -0.204 (0.194)
  - ANZCERTA_mxt:
    - 2.463 *** (0.227)
    - 1.612 *** (0.132)
    - 0.870 ** (0.347)
    - 0.702 ** (0.295)
  - SPARTECA_mxt:
    - 1.412 *** (0.199)
    - 1.203 *** (0.207)
    - 0.710 *** (0.178)
    - 0.674 *** (0.164)
  - EFTA_mxt (inclusive columns):
    - 0.455 *** (0.122); 0.132 (0.114)
  - EEA_mxt (inclusive columns):
    - 0.472 *** (0.069); 0.275 *** (0.065)
  - AP_mxt:
    - 0.404 ** (0.166); 0.423 ** (0.214)
  - LAIA_mxt:
    - 0.032 (0.110); 1.023 *** (0.185)
  - APEC_mxt:
    - 0.616 *** (0.090); 0.039 (0.098)
- Notes: *, **, *** are 10%, 5%, 1% significance levels. Standard errors clustered by country-pairs in parentheses. Coefficients of Fixed Effect controls are suppressed. Remaining coefficients (WTO regressors and historical/geographical controls) are as in corresponding regressions in Table 1.

### Empirical findings — Table 2: WTO and PTA Effects (Inclusive Coding)
- Dependent variable: bilateral imports. Regression columns # 7, 8, 9, 10, 11, 12. Country-Pair FE: No for 7–9, Yes for 10–12.
- Adj R2 values:
  - 0.7401, 0.7415, 0.7431, 0.8747, 0.8751, 0.8760.
- F Statistic vs. Regr.# comparisons: Prob>F: 0.00 for reported comparisons.
- Selected coefficients (coefficient (standard error)):
  - GSP_mt (ind. imp. grants):
    - -0.127 *** (0.040)
    - -0.183 *** (0.040)
    - -0.181 *** (0.039)
    - -0.252 *** (0.048)
    - -0.228 *** (0.049)
    - -0.187 *** (0.048)
  - WTODev_mt (dev. imp.):
    - -0.103 (0.065)
    - -0.118 * (0.065)
    - -0.109 * (0.065)
    - -0.051 (0.069)
    - -0.054 (0.069)
    - -0.035 (0.068)
  - WTOInd_mt (ind. imp.):
    - 0.058 (0.065)
    - 0.069 (0.065)
    - 0.071 (0.065)
    - -0.028 (0.068)
    - -0.026 (0.068)
    - -0.002 (0.066)
  - Ind./dev. PTA_mxt (aggregate):
    - 0.629 *** (0.052)
    - 0.473 *** (0.045)
  - BilateralPTA_mxt:
    - 0.453 *** (0.069)
    - 0.470 *** (0.070)
    - 0.070 (0.068)
    - 0.059 (0.068)
  - NAFTA_mxt:
    - 1.755 *** (0.145)
    - 1.194 *** (0.153)
    - 0.323 ** (0.155)
    - 0.150 (0.171)
  - EU_mxt:
    - 0.072 (0.056)
    - -0.146 ** (0.064)
    - 0.506 *** (0.060)
    - 0.312 *** (0.068)
  - CACM_mxt:
    - 1.316 *** (0.163)
    - 1.372 *** (0.164)
    - 1.788 *** (0.238)
    - 1.789 *** (0.238)
  - CARICOM_mxt:
    - 1.447 *** (0.144)
    - 1.454 *** (0.143)
    - 1.065 *** (0.254)
    - 1.061 *** (0.254)
  - MERCOSUR_mxt:
    - 1.540 *** (0.197)
    - 1.402 *** (0.196)
    - 0.984 *** (0.228)
    - 0.879 *** (0.229)
  - AFTA_mxt:
    - 0.551 *** (0.156)
    - 0.189 (0.154)
    - -0.043 (0.195)
    - -0.161 (0.194)
  - ANZCERTA_mxt:
    - 1.877 *** (0.124)
    - 1.514 *** (0.124)
    - 0.747 ** (0.292)
    - 0.623 ** (0.291)
  - SPARTECA_mxt:
    - 1.310 *** (0.206)
    - 1.350 *** (0.208)
    - 0.814 *** (0.168)
    - 0.807 *** (0.167)
  - EFTA_mxt:
    - 0.473 *** (0.091); 0.199 *** (0.075)
  - EEA_mxt:
    - 0.489 *** (0.067); 0.294 *** (0.069)
  - AP_mxt:
    - 0.385 ** (0.167); 0.424 * (0.217)
  - LAIA_mxt:
    - 0.189 * (0.098); 1.190 *** (0.182)
  - APEC_mxt:
    - 0.798 *** (0.063); 0.244 *** (0.071)
  - CU_mxt (Currency union):
    - 0.618 *** (0.082); 0.627 *** (0.082); 0.654 *** (0.082); 0.290 ** (0.113); 0.312 *** (0.114); 0.318 *** (0.114)
  - Historical/geographical controls (selected):
    - CurColony_mxt (Current colony): 0.755 *** (0.144); 0.762 *** (0.145); 0.762 *** (0.145); 0.370 *** (0.137); 0.358 *** (0.137); 0.382 *** (0.137)
    - EverColony_mx (Ever colony): 1.173 *** (0.066); 1.159 *** (0.065); 1.162 *** (0.064)
    - ComColonizer_mx (Common colonizer): 0.605 *** (0.050); 0.584 *** (0.051); 0.589 *** (0.054)
    - ComLang_mx (Common language): 0.148 *** (0.031); 0.137 *** (0.031); 0.121 *** (0.032)
    - ComNat_mx (Same nation): -0.503 (0.545); -0.532 (0.547); -0.547 (0.546)
    - Border_mx (Common border): 0.014 (0.066); 0.027 (0.065); 0.042 (0.065)
    - Dist_mx (Log of distance): -0.968 *** (0.017); -0.961 *** (0.017); -0.930 *** (0.017)
- Notes: *, **, *** are 10%, 5%, 1% significance levels. Standard errors clustered by country-pairs in parentheses. Coefficients of Fixed Effect controls are suppressed. All regressions include time-varying importer and exporter fixed effects.

*Italic: Source document content as provided in _wp0931 - References.*

### APPENDIX A

### _wp0931 - APPENDIX A

### Membership in considered Preferential Trading Arrangements (Table A1)
- ANZCERTA (Closer Economic Relations Trade Agreement) — Start: 1983; Members: Australia, New Zealand
- APEC (Asia Pacific Economic Community) — Start: 1989; Members: Australia, Brunei, Canada, China (1991), Chile (1994), Taiwan (1991), Hong Kong (1991), Indonesia, Japan, South Korea, Malaysia, Mexico (1993), New Zealand, Papua New Guinea (1993), Peru (1998), Philippines, Singapore, Thailand, United States, Vietnam (1998).
- AP (Andean Community / Andean Pact) — Start: 1969; Members: Bolivia, Colombia, Ecuador, Peru, Venezuela (1973). Former: Chile (1969-76)
- AFTA (ASEAN Free Trade Area) — Start: 1967; Members: Brunei (1984), Cambodia (1998), Indonesia, Laos (1997), Malaysia, Myanmar (1997), the Philippines, Singapore, Thailand, Vietnam (1995).
- CACM (Central American Common Market) — Start: 1960; Members: Costa Rica (1963), El Salvador, Guatemala, Honduras, Nicaragua.
- CARICOM (Caribbean Community/Carifta) — Start: 1968; Members: Antigua and Barbuda, Bahamas (1983), Barbados, Belize (1995), Dominica (1974), Guyana (1995), Grenada (1974), Jamaica, Montserrat (1974), St. Kitts and Nevis, St. Lucia (1974), St. Vincent and the Grenadines, Suriname (1995), Trinidad and Tobago.
- EEA (European Economic Area) — Start: 1994; Members: Austria, Belgium, Denmark, Finland, France, Germany, Greece, Luxembourg, Iceland, Italy, Ireland, Liechtenstein, Netherlands, Norway, Portugal, Spain, Sweden, United Kingdom.
- EFTA (European Free Trade Association) — Start: 1960; Members: Iceland, Liechtenstein (1991), Norway (1986), Switzerland. Former: Denmark (1960-72), United Kingdom (1960-72), Portugal (1960-85), Austria (1960-94), Sweden (1960-94), Finland (1986-94).
- EU (European Union) — Start: 1958; Members: Austria (1995), Belgium, Denmark (1973), Finland (1995), France, Germany, Greece (1981), Luxembourg, Ireland (1973), Italy, Netherlands, Portugal (1986), Spain (1986), Sweden (1995), United Kingdom (1973).
- LAIA/LAFTA (Latin America Integration Agreement) — Start: 1960; Members: Argentina, Bolivia (1967), Brazil, Chile, Colombia (1961), Ecuador (1961), Mexico, Paraguay, Peru, Uruguay, Venezuela (1966).
- MERCOSUR (Southern Cone Common Market) — Start: 1991; Members: Argentina, Brazil, Paraguay, Uruguay.
- NAFTA (Canada-US Free Trade Arrangement / North America Free Trade Agreement) — Start: 1988; Members: Canada, United States, Mexico (1994).
- SPARTECA (South Pacific Regional Trade and Economic Cooperation Agreement) — Start: 1981; Coverage: trade relations between the Cook Islands, Fiji, Kiribati, Marshall Islands, Micronesia, Nauru, Niue, Palau, Papua-New Guinea, Salomon Islands, Samoa, Tonga, Tuvalu, Vanuatu, and Australia and New Zealand.
- BilateralPTA — All bilateral agreements considered are listed in Table A2.

(Source: Eicher, Henn and Papageorgiou (2007).)

### Bilateral Preferential Trade Agreements considered (Table A2)
- Comprehensive list of bilateral PTAs considered in BilateralPTA including (examples extracted from table):
  - US - Israel
  - Turkey - Slovenia
  - Papua New Guinea - Australia Trade & Commercial Relations Agreement (PATCRA)
  - EC - Slovenia; EC - Tunisia; EC - Lithuania; EC - Estonia; EC - Latvia; EC - Israel; EC - Mexico; EC - South Africa; EC - Morocco; EC - Israel
  - Estonia - Faroe Islands; Estonia - Ukraine; Estonia – Turkey
  - Poland - Turkey; Poland - Latvia; Poland - Faeroe Islands; Poland – Lithuania
  - Canada – Chile; Canada – Israel
  - EFTA - Morocco; EFTA – Estonia; EFTA – Latvia; EFTA – Lithuania; EFTA – Estonia (Faeroe Islands)
  - CEFTA accession of Romania; CEFTA accession of Slovenia; CEFTA accession of Bulgaria
  - Various Georgia bilateral agreements: Georgia - Armenia; Georgia - Azerbaijan; Georgia - Kazakhstan; Georgia - Turkmenistan; Georgia - Ukraine
  - Kyrgyz Republic - Moldova; Kyrgyz Republic - Ukraine; Kyrgyz Republic - Uzbekistan
  - Bulgaria - Turkey; Bulgaria - former Yugoslav Rep. of Macedonia
  - Hungary - Latvia; Hungary - Lithuania; Hungary – Turkey; Hungary – Israel
  - Slovenia – Israel; Slovenia – Croatia; Slovenia – Estonia; Slovenia – Lithuania
  - Slovakia/Slovak Republic links: Slovak Republic – Turkey; Slovak Republic – Estonia; Slovak Republic – Latvia; Slovak Republic – Lithuania; Slovak Republic – Israel
  - Romania – Turkey; Czech Republic – Estonia; Czech Republic – Israel; Czech Republic – Latvia; Czech Republic – Lithuania
  - Mexico - Israel; Chile - Mexico
  - Additional bilateral links involving Papua New Guinea, Turkey, Israel, and others.
- Source: Subramanian and Wei (2007).

### List of countries in sample and year of WTO accession (Table A3)
- Industrialized Countries (with accession years):
  - Australia (1948), Austria (1951), Belgium (1948), Canada (1948), Denmark (1950), Finland (1950), France (1948), Germany (1951), Greece (1950), Iceland (1968), Ireland (1967), Italy (1950), Japan (1955), Luxembourg (1948), Netherlands (1948), New Zealand (1948), Norway (1948), Portugal (1962), Spain (1963), Sweden (1950), Switzerland (1966), United Kingdom (1948), United States (1948)
- Developing Countries and Territories (selection, with accession years where provided):
  - Albania (2000), Algeria, Angola (1994), Antigua and Barbuda (1987), Argentina (1967), Armenia, Azerbaijan, Bahamas, The, Bahrain, Kingdom of (1993), Bangladesh (1972), Barbados (1967), Belarus, Belize (1983), Benin (1996), Bermuda (1948), Bhutan, Bolivia (1990), Botswana (1987), Brazil (1948), Bulgaria (1996), Burkina Faso (1963), Burundi (1965), Cambodia, Cameroon (1963), Cape Verde, Central African Rep. (1963), Chad (1963), Chile (1949), China (2001), China, Hong Kong SAR (1986), Colombia (1981), Comoros (1948), Congo, Dem. Rep. of (Zaire) (1971), Congo, Republic of (1963), Costa Rica (1990), Côte d'Ivoire (Ivory Coast) (1963), Croatia (2000), Cyprus (1963), Czech Republic (1993), Djibouti (1994), Dominica (1993), Dominican Republic (1950), Ecuador (1996), Egypt (1970), El Salvador (1991), Equatorial Guinea, Estonia (1999), Ethiopia, Fiji (1993), Gabon (1963), Gambia, The (1965), Georgia (2000), Ghana (1957), Grenada (1994), Guatemala (1991), Guinea (1994), Guinea-Bissau (1994), Haiti (1950), Honduras (1994), Hungary (1973), India (1948), Indonesia (1950), Iran, Islamic Republic of, Iraq, Ireland appears in industrialized list, Israel (1962), Jamaica (1963), Japan appears in industrialized list, Jordan (2000), Kazakhstan, Kenya (1964), Kiribati, Korea (1967), Kuwait (1963), Kyrgyz Republic (1998), Lao People's Dem.Rep, Latvia (1999), Lesotho (1988), Liberia, Libya, Lithuania (2001), Luxembourg appears in industrialized list, Macedonia, Madagascar (1963), Malawi (1964), Malaysia (1957), Maldives (1983), Mali (1993), Malta (1964), Mauritania (1963), Mauritius (1970), Mexico (1986), Moldova (2001), Mongolia (1997), Morocco (1987), Mozambique (1992), Myanmar (1948), Namibia (1992), Nepal, Netherlands appears in industrialized list, New Zealand appears in industrialized list, Nicaragua (1950), Niger (1963), Nigeria (1960), Norway appears in industrialized list, Oman (2000), Pakistan (1948), Panama (1997), Paraguay (1994), Peru (1951), Philippines (1979), Poland (1967), Portugal appears in industrialized list, Qatar (1994), Romania (1971), Russia, Rwanda (1966), São Tomé & Príncipe, Saudi Arabia, Senegal (1963), Seychelles, Sierra Leone (1961), Singapore (1973), Slovak Republic (1993), Slovenia (1994), Solomon Islands (1994), Somalia, South Africa (1948), Sri Lanka (1948), St. Kitts and Nevis (1994), St. Lucia (1993), St.Vincent & The Grenadines (1993), Sudan, Suriname (1978), Swaziland (1993), Sweden appears in industrialized list, Switzerland appears in industrialized list, Syria, Tajikistan, Tanzania (1961), Thailand (1982), Togo (1964), Tonga, Trinidad and Tobago (1962), Tunisia (1990), Turkey (1951), Turkmenistan, Uganda (1962), Ukraine, United Arab Emirates (1994), Uruguay (1953), Uzbekistan, Vanuatu, Venezuela, Rep. Bol. (1990), Vietnam, Yemen, Republic of, Yugoslavia, Soc. Fed. R. of (1966), Zambia (1982), Zimbabwe (1948).
- Source: Subramanian and Wei (2007).

### De jure coding: Regression summaries and key coefficient estimates (Appendix B: Tables B1, B1a, B2)
- Regression setup notes (common across tables):
  - Dependent variable: bilateral imports.
  - All regressions include time-varying importer and exporter fixed effects.
  - Standard errors clustered by country-pairs (reported in parentheses).
  - Notes: *, **, *** denote 10, 5, 1% significance levels.
  - Fixed effect coefficients are suppressed in reported tables.
  - Coefficients for GSP and PTAs in Table B1 are composite coefficients that report GSP and PTA effects net of WTO effects and are calculated by the Delta method.

- Table B1: De jure coding. WTO and PTA Effects (Hierarchical, Mutually Exclusive Coding)
  - Regressions #1–#6 reported; Country-Pair FE included in regressions #4–#6 (Yes).
  - Adj R2:
    - Regression #1: 0.7411
    - Regression #2: 0.7415
    - Regression #3: 0.7430
    - Regression #4: 0.8747
    - Regression #5: 0.8751
    - Regression #6: 0.8760
  - F Statistic vs. Regr.# reported with Prob>F: 0.00 for comparisons shown.
  - Selected coefficient estimates (coefficient (standard error)):
    - WTOInd_mt (WTO indust. imp.): Regression #1: 0.962 *** (0.138); #2: 0.398 ** (0.177); #3: -0.080 (0.088); #4: 0.264 * (0.142); #5: 0.084 (0.165); #6: -0.096 (0.086).
    - WTODev_mt (WTO devel. imp.): Regression #1: -0.057 (0.125); #2: -0.286 ** (0.131); #3: -0.281 *** (0.075); #4: 0.068 (0.117); #5: -0.259 ** (0.118); #6: -0.269 *** (0.076).
    - GSP_mt (indust. imp. grants GSP): Regression #1: -0.209 *** (0.041); #2: -0.202*** (0.041); #3: -0.199 *** (0.040); #4: -0.242*** (0.049); #5: -0.233*** (0.049); #6: -0.178 *** (0.047).
    - PTA_mxt (catch-all, ind. imp.): Regression #1: 0.152 *** (0.055); Regression #2: 0.361 *** (0.055).
    - PTA_mxt (catch-all, devl. imp): Regression #1: 1.171 *** (0.082); Regression #2: 0.557 *** (0.082).
    - BilateralPTA_mxt (indust. importer): Regression #2: 0.052 (0.114); #3: 0.356 *** (0.091); #5: -0.174 * (0.101); #6: -0.044 (0.083).
    - BilateralPTA_mxt (devel. importer): Regression #2: 0.736 *** (0.091); #3: 0.558 *** (0.088); #5: 0.169 * (0.086); #6: 0.129 (0.079).
    - NAFTA_mxt (indust. importer): Regression #2: 1.513 *** (0.128); #3: 1.317 *** (0.127); #5: 0.149 (0.140); #6: 0.280 * (0.167).
    - NAFTA_mxt (devel. importer): Regression #2: 2.197 *** (0.160); #3: 1.518 *** (0.139); #5: 0.492 *** (0.159); #6: 0.453 *** (0.174).
    - EU_mxt: Regression #2: 0.064 (0.069); #3: -0.127 (0.080); #5: 0.472 *** (0.062); #6: 0.314 *** (0.067).
    - CACM_mxt: Regression #2: 1.479 *** (0.210); #3: 1.535 *** (0.198); #5: 1.937 *** (0.248); #6: 1.947 *** (0.242).
    - CARICOM_mxt: Regression #2: 1.501 *** (0.159); #3: 1.506 *** (0.157); #5: 1.094 *** (0.248); #6: 1.093 *** (0.248).
    - MERCOSUR_mxt: Regression #2: 1.531 *** (0.251); #3: 1.582 *** (0.243); #5: 0.983 *** (0.227); #6: 1.073 *** (0.243).
    - AFTA_mxt: Regression #2: 0.631 *** (0.179); #3: 0.431 ** (0.188); #5: -0.032 (0.196); #6: 0.065 (0.205).
    - ANZCERTA_mxt: Regression #2: 1.876 *** (0.123); #3: 1.678 *** (0.129); #5: 0.747 *** (0.292); #6: 0.785 *** (0.295).
    - SPARTECA_mxt (indust. importer): Regression #2: 0.937 *** (0.234); #3: 1.318 *** (0.201); #5: 0.636 *** (0.208); #6: 0.800 *** (0.179).
    - SPARTECA_mxt (devel. importer): Regression #2: 1.621 *** (0.210); #3: 1.519 *** (0.197); #5: 0.979 *** (0.180); #6: 0.973 *** (0.174).
    - EFTA_mxt: Regression #5: 0.497 *** (0.103); Regression #6: 0.206 *** (0.075).
    - EEA_mxt: Regression #5: 0.534 *** (0.103); Regression #6: 0.359 *** (0.101).
    - AP_mxt: Regression #5: 0.690 *** (0.204); Regression #6: 0.695 *** (0.222).
    - LAIA_mxt: Regression #5: 0.288 *** (0.177); Regression #6: 1.288 *** (0.191).
    - APEC_mxt (indust. importer): Regression #3: 0.656 *** (0.094); Regression #5: 0.116 (0.092).
    - APEC_mxt (devel. importer): Regression #3: 0.857 *** (0.076); Regression #5: 0.289 *** (0.078).
    - CU_mxt (Currency union): Regression #1: 0.615 *** (0.091); #2: 0.620 *** (0.091); #3: 0.647 *** (0.091); #4: 0.291 *** (0.113); #5: 0.312 *** (0.114); #6: 0.319 *** (0.114).
    - CurColony_mxt (Current colony): Regression #1: 0.760 *** (0.181); #2: 0.763 *** (0.181); #3: 0.764 *** (0.182); #4: 0.377 *** (0.137); #5: 0.352 *** (0.137); #6: 0.374 *** (0.137).
    - EverColony_mx (Ever colony): Regression #1: 1.164 *** (0.079); #2: 1.157 *** (0.079); #3: 1.161 *** (0.079).
    - ComColonizer_mx (Common colonizer): Regression #1: 0.577 *** (0.054); #2: 0.567 *** (0.055); #3: 0.573 *** (0.054).
    - ComLang_mx (Common language): Regression #1: 0.142 *** (0.035); #2: 0.140 *** (0.035); #3: 0.125 *** (0.036).
    - ComNat_mx (Same nation): Regression #1: -0.522 (0.586); #2: -0.527 (0.587); #3: -0.542 (0.585).
    - Border_mx (Common border): Regression #1: 0.024 (0.078); #2: 0.027 (0.077); #3: 0.040 (0.077).
    - Dist_mx (Log of distance): Regression #1: -0.965 *** (0.019); #2: -0.964 *** (0.019); #3: -0.932 *** (0.019).

- Table B1a: Raw Regression Output De jureCoding. GSP and PTA Coefficient Estimates including WTO Effects (Mutually exclusive coding)
  - Regressions #1–#6; Country-Pair FE: Yes for #4–#6.
  - Adj R2:
    - Regression #1: 0.7411; #2: 0.7415; #3: 0.7430; #4: 0.8747; #5: 0.8751; #6: 0.8760.
  - Selected coefficient estimates:
    - GSP_mt (indust. country granting): Regression #1: 0.754 *** (0.139); #2: 0.196 (0.177); #3: -0.279 *** (0.091); #4: 0.022 (0.143); #5: -0.149 (0.163); #6: -0.274 *** (0.089).
    - PTA_mxt (catch-all dummy): Regression #1: 1.114*** (0.124); Regression #2: 0.625*** (0.132).
    - BilateralPTA_mxt: Regression #1: 0.451 *** (0.155); #2: 0.277 *** (0.098); #3: -0.090 (0.140); #4: -0.139 (0.093).
    - NAFTA_mxt: Regression #1: 1.912 *** (0.204); #2: 1.237 *** (0.112); #3: 0.233 (0.195); #4: 0.184 (0.153).
    - EU_mxt: Regression #1: 0.463 *** (0.186); #2: -0.207 ** (0.104); #3: 0.556 *** (0.172); #4: 0.218 ** (0.103).
    - CACM_mxt: Regression #1: 1.193 *** (0.196); #2: 1.253 *** (0.192); #3: 1.678 *** (0.242); #4: 1.678 *** (0.239).
    - CARICOM_mxt: Regression #1: 1.215 *** (0.189); #2: 1.225 *** (0.167); #3: 0.835 *** (0.272); #4: 0.824 *** (0.258).
    - MERCOSUR_mxt: Regression #1: 1.245 *** (0.283); #2: 1.300 *** (0.243); #3: 0.724 *** (0.256); #4: 0.804 *** (0.234).
    - AFTA_mxt: Regression #1: 0.345 * (0.199); #2: 0.150 (0.173); #3: -0.291 (0.223); #4: -0.203 (0.194).
    - ANZCERTA_mxt: Regression #1: 2.274 *** (0.214); #2: 1.599 *** (0.129); #3: 0.831 *** (0.335); #4: 0.690 *** (0.294).
    - SPARTECA_mxt: Regression #1: 1.335 *** (0.187); #2: 1.238 *** (0.181); #3: 0.721 *** (0.164); #4: 0.705 *** (0.163).
    - EFTA_mxt (in other columns): 0.417 *** (0.124); 0.111 (0.108).
    - EEA_mxt: 0.454 *** (0.078); 0.264 *** (0.065).
    - AP_mxt: 0.409 ** (0.197); 0.426 ** (0.214).
    - LAIA_mxt: 0.006 (0.117); 1.020 *** (0.185).
    - APEC_mxt: 0.576 *** (0.093); 0.020 (0.095).

- Table B2: De jure coding. WTO and PTA Effects (Inclusive Coding)
  - Regressions #7–#12; Country-Pair FE: Yes for #10–#12.
  - Adj R2:
    - Regression #7: 0.7402; #8: 0.7417; #9: 0.7432; #10: 0.8748; #11: 0.8751; #12: 0.8760.
  - Prob>F: 0.00 for tests reported.
  - Selected coefficient estimates:
    - WTOInd_mt: #7: 0.091 (0.057); #8: 0.109 * (0.057); #9: 0.117 ** (0.057); #10: -0.033 (0.053); #11: -0.029 (0.054); #12: -0.021 (0.053).
    - WTODev_mt: #7: -0.102 * (0.056); #8: -0.110 ** (0.056); #9: -0.102 * (0.056); #10: 0.074 (0.055); #11: 0.068 (0.055); #12: 0.068 (0.054).
    - GSP_mt (indust. imp. grants GSP): #7: -0.123 *** (0.043); #8: -0.179 *** (0.042); #9: -0.177 *** (0.041); #10: -0.250 *** (0.048); #11: -0.227 *** (0.049); #12: -0.185 *** (0.048).
    - PTA_mxt (catch-all dummy): #7: 0.627 *** (0.064); #8/#9/#10 etc. show further catch-all estimates (0.470 *** (0.045) in one column).
    - BilateralPTA_mxt: #7: 0.453 *** (0.079); #8: 0.470 *** (0.079); #9: 0.074 (0.068); #10: 0.063 (0.068).
    - NAFTA_mxt: #7: 1.756 *** (0.112); #8: 1.192 *** (0.129); #9: 0.324 ** (0.160); #10: 0.153 (0.177).
    - EU_mxt: #7: 0.066 (0.069); #8: -0.156 ** (0.078); #9: 0.501 *** (0.060); #10: 0.312 *** (0.068).
    - CACM_mxt: #7: 1.315 *** (0.187); #8: 1.371 *** (0.187); #9: 1.789 *** (0.237); #10: 1.790 *** (0.238).
    - CARICOM_mxt: #7: 1.449 *** (0.163); #8: 1.457 *** (0.162); #9: 1.056 *** (0.254); #10: 1.055 *** (0.254).
    - MERCOSUR_mxt: #7: 1.549 *** (0.251); #8: 1.407 *** (0.242); #9: 0.969 *** (0.225); #10: 0.866 *** (0.227).
    - AFTA_mxt: #7: 0.553 *** (0.179); #8: 0.190 (0.175); #9: -0.048 (0.196); #10: -0.165 (0.194).
    - ANZCERTA_mxt: #7: 1.876 *** (0.123); #8: 1.512 *** (0.119); #9: 0.741 *** (0.292); #10: 0.617 ** (0.291).
    - SPARTECA_mxt: #7: 1.316 *** (0.187); #8: 1.357 *** (0.191); #9: 0.813 *** (0.168); #10: 0.807 *** (0.167).
    - EFTA_mxt: #9/#10: 0.469 *** (0.101); 0.199 *** (0.075).
    - EEA_mxt: 0.496 *** (0.078); 0.294 *** (0.069).
    - AP_mxt: 0.382 * (0.200); 0.424 ** (0.216).
    - LAIA_mxt: 0.195 * (0.110); 1.190 *** (0.182).
    - APEC_mxt: 0.802 *** (0.072); 0.244 *** (0.071).
    - CU_mxt (Currency union): #7: 0.624 *** (0.091); #8: 0.635 *** (0.091); #9: 0.662 *** (0.091); #10: 0.284 *** (0.113); #11: 0.306 *** (0.114); #12: 0.312 *** (0.114).
    - CurColony_mxt: #7: 0.763 *** (0.180); #8: 0.772 *** (0.180); #9: 0.773 *** (0.181); #10: 0.368 *** (0.137); #11: 0.357 *** (0.138); #12: 0.383 *** (0.138).
    - EverColony_mx: #7: 1.175 *** (0.081); #8: 1.161 *** (0.079); #9: 1.164 *** (0.078).
    - ComColonizer_mx: #7: 0.594 *** (0.054); #8: 0.571 *** (0.054); #9: 0.577 *** (0.054).
    - ComLang_mx: #7: 0.149 *** (0.036); #8: 0.138 *** (0.035); #9: 0.121 *** (0.036).
    - ComNat_mx (Same nation): #7: 0.517 (0.584); #8: -0.549 (0.586); #9: -0.566 (0.584).
    - Border_mx: #7: 0.013 (0.079); #8: 0.025 (0.077); #9: 0.040 (0.076).
    - Dist_mx (Log of distance): #7: -0.966 *** (0.019); #8: -0.960 *** (0.019); #9: -0.928 *** (0.019).

- Notes across regression tables:
  - Coefficients marked with significance stars indicate levels: * 10%, ** 5%, *** 1%.
  - Standard errors are in parentheses and clustered by country-pairs.
  - Table B1 reports composite GSP and PTA coefficients net of WTO effects (Delta method); Table B1a reports raw coefficients including WTO effects.
  - Mutually exclusive coding vs. inclusive coding distinctions are used across tables to differentiate how WTO and PTA dummies are constructed and interpreted.

*Source: Subramanian and Wei (2007); Eicher, Henn and Papageorgiou (2007).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp0931.pdf_
