## wp18109 ## Source details **Canonical URL:** [wp18109](https://www.imf.org/-/media/files/publications/wp/2018/wp18109.pdf) ## Other formats - [Markdown version](/-/media/files/publications/wp/2018/wp18109.pdf.md) - [Structured JSON version](/-/media/files/publications/wp/2018/wp18109.pdf.json) --- ### 1. Introduction — scope, motivation, and key contributions - Scope and motivation - Studies the WTO Information Technology Agreement (ITA), the sole example of near-global MFN tariff elimination for a substantial number of products. - Identification exploits six ITA characteristics, including coverage of about half of all IT products and exogenous accession of many countries via WTO accession, EU accession, or U.S. FTA negotiation linkages. - Time coverage: 1996-2012. - Key empirical findings — impacts of MFN tariff elimination - MFN tariff reduction elasticity: a percentage point decrease in tariffs raises import value by 0.7 percent. - Tariff elimination (MFN tariff lowered to zero): increases imports by 13 percent within a broad machinery control sector (a sector constituting one third of world trade). - Mechanism for tariff-elimination bonus: zero MFN tariffs reduce frequency and rigor of inspections under risk-based customs systems, lowering time and administrative costs of customs clearance. - Commitment (durability) effects: - Perceived durability via WTO tariff bindings fosters additional “commitment” effects on imports and on exports. - Both tariff-elimination and export-commitment effects are presented as novel contributions to gravity literature. - Production fragmentation and GVCs - Durable MFN tariff elimination increases production fragmentation. - Developed countries specialized upstream; emerging-country members integrated downstream. - Tariff-elimination effect larger for intermediate goods, consistent with GVCs. - Heterogeneous effects across member types - “Passive” signatories (largely developing/emerging; accession driven by WTO/EU/US FTA processes) benefited most. - ITA accession boosted passive signatories’ final goods exports by some 26-40 percent (commitment effect on exports). - Passive signatories’ ITA intermediate-goods imports rose by 26-60 percent relative to other ICT and machinery products after accession. - Passive signatory variation forms a quasi-natural experiment alleviating endogeneity concerns. - Policy implications (summary) - Large effects of MFN tariff elimination where tariffs remain moderately high in many emerging and developing countries. - WTO may be suited to finish tractable parts of the tariff agenda; reviving WTO negotiating function is important. - Past multilateral tariff-elimination success boosted offshoring; reversing zero tariffs risks reversing sizable trade gains. ### 3. The ITA — characteristics, timing, and descriptive patterns - Six unique ITA characteristics (identification advantages) - Rigid positive listing: original ITA lists 190 covered products corresponding to 154 HS1996 subheadings; analysis maps to 106 HS1996 6-digit lines for 1997–2012. - Trade coverage eroded over time for new ICT products (motivating 2016 expansion). - Tariffs on ITA goods eliminated by founding members by 2000; some developing countries implemented up to 2005; most post-1997 joiners had implementation periods generally ≤ three years. - ITA commits members to adjust MFN applied and bound tariffs to zero — making liberalization hard to reverse and enforceable via WTO dispute settlement. - ITA did nothing else but eliminate tariffs, isolating trade policy certainty effects related to locking in tariff elimination. - Late accession of many countries driven by WTO/EU/US FTA processes (passive signatories). - Membership, timing, and critical mass - Negotiations among 34 countries finalized in 1996; agreement into force on July 1, 1997. - Founding members represented >90 percent of world trade in covered products at inception; membership rose to 74 countries by 2012 and 82 by 2017. - ITA applies concessions on an MFN basis to all WTO members; trade diversion among WTO members minimal. - Pre-accession tariff levels and trends (descriptive) - Pre-accession average tariffs on ITA products: - Active signatories: averaged 2.5 percent in the year before accession. - Passive signatories: averaged 3.9 percent in the year before accession. - Average MFN applied tariffs higher: 3.8 and 6.3 percent for initial and late signatories, respectively. - Passive signatories were already reducing tariffs pre-accession; after accession ITA goods’ tariffs decline steeper toward zero than other ICT goods. - Market share and export patterns - Passive signatories’ importance in world ITA goods trade grew immensely over the sample period, at expense of active (predominantly developed) signatories. - China is a dominant player; regressions run with and without Chinese exports — results largely insensitive to exclusion of China. - ITA product exports as share of GDP before accession: - Passive signatories: mean 0.4 percent (median 0.1 percent). - Active signatories: mean 2.8 percent (median 1.0 percent). ### 4. Data — scope, construction, and control groups - Sample period and coverage - Sample period: 1996-2012. - Final dataset size: totals 21.8 million usable observations in primary construction (full combined dataset references reach 28.36 million observations when broad machinery sector included). - Non-zero ITA trade-flow observations: 3.86 million; countries covered: 234. - ITA-related HS1996 lines used: 106; non-ITA ICT control lines: 116. - Tariffs and mapping - Tariffs obtained from UN TRAINS in HS vintages; converted to HS1996 via UNSD conversion tables; missing tariff observations linear interpolated where needed. - Trade flows - Bilateral 6-digit HS1996 trade flows from UN Comtrade; exporter-reported mirror data used when import data missing. - GVC / intermediate vs final classification - Sturgeon and Memedovic (2010) classification used; after conversion covers 47 of 106 ITA lines. - Resulting usable observations when splitting by nature: 680 thousand intermediate-good observations and 825 thousand final-good observations of a total of 2.5 million ITA observations in that split. - Control sectors - Non-ITA ICT goods (OECD definitions): 116 non-ITA ICT lines. - Broad machinery sector: HS sections 84, 85, 87, and 90 covering 995 tariff lines (includes ITA and other ICT products). ### 5. Empirical strategy — product-level structural gravity and two-stage approach - Gravity framework - Four-dimensional structural gravity (exporter, importer, product, time), building on Anderson and Van Wincoop (2003). - Iceberg trade costs specification: T_ijkt = (1 + t_Applied_ijkt) · exp[ζ I_ijt + δ_ijk] = exp[ln(1 + t_Applied_ijkt) + ζ I_ijt + δ_ijk]. - Two-stage estimation strategy - Stage 1: PPML estimation of multiplicative gravity to obtain importer-product-time (α_ikt) and exporter-product-time (α_jkt) fixed effects while controlling for discriminatory trade policy (applied/preferential tariffs ln(1 + t_Applied_ijkt)). - Stage 1 equation: x_ijkt = exp[α_ikt + α_jkt + α_ijk + (1−σ) ln(1 + t_Applied_ijkt) + ψ I_ijt] · ε_ijkt. - Stage 2: OLS regressions using α_ikt (importer-product-time) and α_jkt (exporter-product-time) as dependent variables to estimate effects of MFN tariff reduction, tariff elimination, and ITA membership (commitment) effects. - Importer-stage decomposition: - α_ikt = δ_it + δ_ik + δ_kt + β_1 ln(1 + t_MFN_ikt) + β_2 t0_MFN_ikt + β_3 imita_ikt + ξ_ikt. - t0_MFN_ikt = 1 if MFN tariff is zero. - β_1 captures MFN tariff reduction effect; β_2 captures tariff-elimination bonus; β_3 captures importer ITA membership (commitment) effect (split into active vs passive). - Exporter-stage decomposition: - α_jkt = δ_jt + δ_jk + δ_kt + y_jkt + β_4 exita_jkt + η_jkt. - β_4 captures exporter-side commitment effect. - Identification and endogeneity handling - Heavy use of high-dimensional fixed effects (importer-exporter-product, importer-product-time, exporter-product-time, importer-time, product-time) to control for multilateral resistance and selection. - Quasi-natural experiment: passive signatories’ exogenous accession motivations (WTO accession, EU accession, U.S. FTA negotiation linkages) used to allay remaining endogeneity concerns; restricted-sample estimates on passive signatories presented as robustness. ### 6. Results — layers of ITA trade effects and magnitudes - Overview and structure (Table 2) - Two product splits: ICT goods (narrow ITA and other ICT goods) and broad machinery sector. - Goods categories: All goods, Intermediate, Final. - Stage 1 observations examples: 8,120,335; 32,162,466; stage 2 importer observations examples: 367,521; 1,729,821. - Stage 1 R2 examples: 0.9837; 0.9867. Stage 2 importer adjusted R2 examples: 0.8333; 0.8333. - 6.1 Preferential and MFN tariff reduction effects - Stage 1 (applied tariff ln(1+tariff)): a 1 percent reduction of applied tariffs on ITA products causes a 0.5 to 1.6 percent increase in imports (effects stronger for intermediates). - Stage 2 preferred estimate: imports increase by 0.7 percent in response to a 1 percentage point reduction of MFN tariffs. - 6.2 Tariff elimination effect (t0_MFN) - Tariff elimination estimated to boost eligible imports by about 6-13 percent (preferred machinery-sample estimates toward higher end): - exp(0.063)-1 = 6.5%; exp(0.121)-1 = 13%. - For intermediate goods: estimated impact in the 14-18 percent range. - Final goods: up to a 5 percent trade increase in preferred machinery-sample regression; ICT-sample coefficients smaller/insignificant for finals. - Mechanism: zero MFN tariffs reduce customs inspection incentives and administrative time, lowering trade costs beyond price effects. - Policy implication: eliminating already-low tariffs can yield large trade gains with modest fiscal revenue loss. - 6.3 Commitment effect for importers (β_3) - Passive signatories: - ITA commitment effects boosted their intermediate imports by 26-60 percent (preferred ICT-sample estimate forms higher end). - Final goods imports for passive signatories did not respond strongly; all-goods effect (7-14 percent) driven largely by intermediates. - Active signatories: - Evidence suggests final goods imports increased more in active signatories; ICT-sample indicates exporters increased ITA shipments to active signatories by about 20 percent more than for ICT goods. - Machinery-sample evidence weaker. - Interpretation: aligns with GVC narrative — passive signatories integrated downstream; active signatories shifted downstream stages to passive signatories. - 6.4 Commitment effects for exporters (β_4) - Passive signatories’ ITA exports increased by 17-23 percent more than other ICT or machinery goods after accession. - Final goods drove this overall effect: ITA exports outperformed control groups by 26-40 percent for final goods. - Intermediate exports for passive signatories: estimates positive but low statistical significance. - Active signatories: exporter commitment effects not strong; some weak evidence of contraction in active signatories’ final ITA exports consistent with production relocation to passive signatories. - Results robust to excluding China in most layers; exporter-side effects broadly similar when excluding China. ### 7. Extensions and robustness — extensive margin, lag dynamics, heterogeneity - 7.1 Extensive margin (Table 3) - Extensive margin defined as share of tariff lines with positive trade flows. - Key extensive-margin commitment estimates (selected) - Importer commitment effect, active importers: increases ~0.010 to 0.049 (statistical significance varies by regression). - Importer commitment effect, passive importers: increases ~0.006 to 0.039 (significance varies). - Exporter commitment effect, passive exporters: increases ~0.009 to 0.017 (significant in several specifications); excluding China yields similar magnitudes. - Interpretation: active signatories saw importer-share increases of ~1 percentage point relative to controls (non-negligible given a 0.15 average share in 1996); intermediates drive much of diversification. - Quasi-natural experiment — passive signatories only (Table 4) - Restricted-sample estimates closely mirror baseline: - Tariff elimination effect remains statistically significant (examples: 0.102 ∗∗∗; 0.160 ∗∗∗; 0.312 ∗∗∗ in different regressions). - Commitment effects for passive exporters: example coefficients 0.173 ∗∗; 0.284 ∗∗∗; 0.312 ∗ (varies by split and exclusion of China). - Interpretation: robustness to endogeneity concerns reinforced by passive-sample results. - Lagged tariffs and interaction (Table 5 and Table A10) - Firms react to MFN tariff reductions within about a year; contemporaneous MFN coefficients remain primary, first lag statistically small; some RTA/preferential reductions show significant first-lag responses. - Introducing lagged tariff terms does not materially change tariff-elimination or commitment estimates; tariff-elimination effects slightly reduced to a 5-10 percent range in some specs. - Interaction of tariff reduction and elimination: interaction coefficient negative but generally statistically insignificant — weak evidence tariff-elimination bonus may be higher starting from low pre-elimination “nuisance” tariffs. - Heterogeneity of exporter commitment effects by country characteristics (Table 6; Table A9) - Determinants interacted: remoteness, education (secondary school completion), business environment (control of corruption), rule of law. - Passive signatories’ exporter commitment effects vary by remoteness and institutional/business environment: - Remoteness: passive signatories at higher remoteness percentiles see much smaller exporter commitment effects (upper quartile effects roughly half and often statistically insignificant). - Business environment improvements are associated with larger export commitment gains; rule of law improvements have subtler impact. - Lower-education passive signatories experienced larger export increases — consistent with initial GVC downstream assembly roles. - Policy-relevant insight: complementary reforms (addressing remoteness, business environment, education, institutions) are needed to realize ITA gains fully. ### Appendix highlights — tables and figures (selected numeric facts) - Stage 1 PPML preferential tariff reduction ln(1+tariff) coefficients (Table A1 / A3 examples): - All goods (ICT): -1.150 ∗∗∗ (t = -3.66) - All goods (Machinery): -0.472 ∗ (t = -2.03) - Intermediate (ICT): -1.608 ∗ (t = -2.37) - Extensive-margin mean shares (Table A11 examples) - Active ITA importers in 1996: ITA share = 0.15; ITA (Intermediate) = 0.21; ITA (Final) = 0.22. - Passive ITA exporters in 1996: ITA share = 0.08; by 2012 passive ITA exporters ITA share = 0.13. - Dataset and classification facts (Appendix Figures / Data Appendix) - ITA analysis uses 106 HS1996 6-digit lines (out of 154 product lines originally listed when mapped). - Temporal mapping across HS vintages: consistent HS1996 basis via UNSD conversion; 74 HS1996 lines are included in all years. - Broad machinery sector covers 995 tariff lines; combined dataset reaches up to 28.36 million observations when full machinery sector included. ### 8. Conclusion — synthesis and policy recommendations - Main empirical conclusions - Non-discriminatory, durable MFN tariff elimination (ITA) generated large trade effects via three layered mechanisms: - MFN tariff reduction effect (import responsiveness to MFN tariff changes). - Tariff-elimination bonus when MFN tariff reaches zero (6-13 percent overall; 14-18 percent for intermediates in some specifications). - Commitment effects from WTO tariff bindings (substantial exporter- and importer-side gains, especially for passive signatories). - Commitment effects also influence exports by incentivizing investment, firm location, and sourcing decisions; import tariff elimination can help develop export industries. - Methodological implications for gravity studies - Direct integration of tariff data into gravity analyses reveals non-linear and commitment layers of trade-policy impacts and allows trade-agreement dummies to be interpreted as commitment effects. - Policy recommendations - Strong case for eliminating already-low tariffs: modest fiscal revenue losses but substantial trade gains. - Safeguard zero tariffs where achieved; reversing tariff eliminations could reverse sizable trade gains. - WTO may have comparative advantage in concluding tractable tariff-reduction agendas; pursuing multilateral tariff elimination remains valuable. - Complementary reforms (improving remoteness logistics, business environment, education, institutions) enhance countries’ ability to leverage tariff-elimination and commitment gains. *Source: wp18109 — IMF Working Paper (Extracted sections: 1. Introduction; 3. The Information Technology Agreement (ITA); 4. Data; 5. Empirical Strategy; 6. Results; 7. Extensions and Robustness; 8. Conclusion; Annex I; Appendix Tables and Figures).* ### 1. Introduction ### 1. Introduction ### Scope and motivation - Tariffs are central to trade policy, yet the most wide-ranging and durable elimination of tariffs—Most-favoured-nation (MFN) tariff elimination through World Trade Organization (WTO) tariff bindings—has been comparatively little studied. - The paper studies the WTO Information Technology Agreement (ITA), the sole example of an agreement that instituted near-global MFN tariff elimination for a substantial number of products. - Identification is facilitated by six unique characteristics of the ITA, including that the ITA covered about half of all IT products, allowing non-eligible products to serve as a natural control group, and that many countries joined ITA as an unavoidable part of pursuing larger policy objectives (e.g., WTO membership, EU accession, a Free Trade Agreement with United States), providing largely exogenous selection. ### Key empirical findings — impacts of MFN tariff elimination - MFN tariff reduction itself: a percentage point decrease in tariffs raises import value by 0.7 percent. - Tariff elimination (tariff lowered all the way to zero): increases imports by 13 percent within a broad machinery control sector, which constitutes one third of world trade. - Mechanism: likely linked to border administration practices—zero MFN tariffs reduce frequency of rigorous inspections under risk-based customs systems, lowering time and administrative costs of customs clearance. - Commitment effects (durability via WTO tariff bindings): - Perceived durability of liberalization fosters "commitment" effects on imports. - Commitment effects also manifest on exports: certainty of import availability can support export sectors. - Both the tariff elimination effect and the commitment effect for exports are presented as novel contributions not previously appreciated in gravity literature. ### Production fragmentation and global value chains (GVC) - Durable MFN tariff elimination increases fragmentation of production. - Developed countries specialized in upstream intermediate production; emerging country members integrated downstream. - The tariff elimination effect is larger for intermediate goods, consistent with GVCs where intermediates often cross borders multiple times. ### Heterogeneous effects across member types (active vs passive signatories) - A subgroup of largely developing and emerging countries—those that joined the ITA after inception as an unavoidable by-product of pursuing larger policy objectives (termed passive signatories)—benefitted most. - These passive signatories were often skeptical of the ITA yet reaped larger trade benefits than active signatories. - ITA accession boosted passive signatories’ final goods exports by some 26-40 percent, as illustrated by the commitment effect on exports. - To facilitate export expansion, passive signatories’ ITA intermediate goods imports rose by 26-60 percent relative to other ICT and machinery products after ITA accession. - The passive signatory experience constitutes a quasi-natural experiment that helps allay endogeneity concerns. ### Contributions relative to existing literature - Empirical literatures advanced: - Non-discriminatory trade policy changes: product-level analyses of MFN tariff effects, complementing work by Caliendo et al. (2015) and Heid et al. (2017). - WTO trade effects and trade policy certainty: distinguishing impacts of reducing/eliminating MFN tariffs from impacts of increased trade policy certainty. - International production sharing and GVCs: providing evidence that ITA facilitated deeper GVC integration in covered goods. - ITA-specific literature: extends previous econometric analyses (Mann and Liu (2009); Bora and Liu (2010)) by using detailed product-level data and longer time coverage. ### Policy implications - Tariffs remain at moderately high levels in many emerging and developing countries; MFN tariff elimination has large effects and thus what remains on the negotiating table is sizable. - The WTO may have more promise in finishing tractable parts of the tariff agenda than in addressing complex non-tariff measures (NTMs); reviving WTO negotiating function is important. - Offshoring may become harder to negotiate over time as it deepens, but the paper’s results imply that past multilateral negotiation success boosted offshoring. - Containing populist backlash against trade liberalization is important: reversing previously achieved tariff elimination could reverse sizable trade gains. ### Methodology and paper structure (brief) - Uses product-level data and a structural gravity model with high-dimensional fixed effects to address unobservable characteristics and endogeneity concerns. - Employs the ITA’s six favorable characteristics for identification and exploits natural control groups (non-eligible products) and passive signatory variation. - Time coverage and data span noted: this paper covers the period 1996-2012; earlier ITA studies cited used data through 2003 and found smaller coverage of passive signatories. - Remaining paper sections: Section 2 (related literature), Section 3 (ITA characteristics and initial graphs), Section 4 (dataset), Section 5 (estimation strategy and econometric concerns), Section 6 (baseline results), Section 7 and Appendix (extensions and robustness), Section 8 (conclusion). *Source: wp18109 - 1. Introduction* ### 3. The Information Technology Agreement (ITA) ### 3. The Information Technology Agreement (ITA) ### 3.1 Six unique characteristics and the quasi-natural experiment - ITA liberalized tariffs on a broad set of IT products but used a rigid positive listing; the original ITA lists 190 covered products corresponding to 154 HS1996 subheadings, i.e. 6-digit product codes, with many subheadings only covered partially. - Positive listing implies new ICT products generally are not covered; trade coverage of the 190 ITA products therefore eroded over time, motivating the 2016 expansion of the ITA. - The ITA provides a favorable setting for cross-product identification because: - It covers a broad set of IT products while leaving many other ICT and machinery goods uncovered, providing natural control groups. - Tariffs on ITA goods were eliminated (zero tariffs implemented by founding members by 2000; some developing countries implemented up to 2005; most post-1997 joiners had implementation periods that in most cases did not surpass three years). - The ITA committed members to adjust their MFN applied and bound tariffs to zero, making liberalization (i) hard to reverse and (ii) enforceable through the WTO dispute settlement system. - The ITA did nothing else but eliminate tariffs, allowing isolation of trade policy certainty effects related exclusively to locking in tariff elimination. - Endogeneity concerns are allayed by a quasi-natural experiment: latecomers’ accession was often exogenous, driven by broader policy objectives rather than self-selection. Three motivations for accession identified: - Joining ITA as part of WTO accession protocols. - EU accession required adopting EU trade policy, including ITA membership. - U.S. persuasion of RTA partners to join the ITA. - Signatory classification: - “Active” signatories include all founding members and countries that actively sought ITA membership. - “Passive” signatories include countries whose accession was largely driven by WTO accession, EU accession, or U.S. FTA negotiations. Thirteen economies were classified as passive due to WTO accession and fifteen due to EU accession or U.S. RTA relations (classification based on WTO (2012) and additional individual assessment; Annex I summarizes supporting evidence). - Critical-mass and trade diversion: - The ITA is a multilateral “critical mass agreement”: founding members represented, upon inception, more than 90 percent of world trade in products covered by the agreement. - Negotiations among 34 countries were finalized in 1996; nine more countries signed up by a March 1997 deadline, pushing membership past the 90 percent threshold. - Agreement came into force on July 1, 1997. - Membership increased to 74 countries by 2012 (end of sample) and 82 by 2017. - The ITA requires members to apply agreed tariff concessions on an MFN basis to all WTO members, so trade diversion among WTO members is hardly a concern (some trade diversion away from non-WTO members could occur de jure, but in practice importers also apply MFN tariffs to WTO nonmembers). - Reversal of pre-ITA trade diversion is limited: - Where pre-existing RTAs granted preferences, the ITA eliminated those tariff preferences, potentially increasing trade with new ITA partners; estimations explicitly account for this. - Favorable timing: ITA inception in 1997 preceded the true proliferation of RTAs; developing countries covered by pre-existing GSP schemes generally did not have large IT sectors. ### 3.2 A first glance at the ITA’s impact - Pre-accession average tariffs on ITA products: - Active signatories: averaged 2.5 percent in the year before accession. - Passive signatories: averaged 3.9 percent in the year before accession. - Note: these average tariff figures include preferential tariffs. Average MFN applied tariffs are somewhat higher, 3.8 and 6.3 percent for initial and late signatories, respectively. - Tariff trends: - Passive signatories had been reducing tariffs on ITA goods considerably before accession. - Although ITA goods had somewhat lower tariffs than other ICT goods before accession, both followed very similar trends until accession; after accession, ITA goods’ tariffs follow a steeper downward trajectory toward zero than other ICT goods. - Market share and export patterns: - Passive signatories’ importance in world ITA goods trade grew immensely over the sample period at the expense of active signatories (which are predominantly developed countries). - China stands out as a dominant player in IT products; regressions are run both on the whole sample and excluding China’s exports to check sensitivity. - Increases in market shares are more impressive for exports than imports, indicating import demand remained relatively stable while origin of products shifted, consistent with multinational enterprise location and sourcing decisions and GVC integration. - Figures reported (descriptive evidence): - Exports of active signatories and non-members show only moderate growth; passive signatories show a much steeper export growth slope. - When rescaling market shares to 100 in each country’s ITA entry year and focusing on a 7-years-before / 5-years-after window, 12 passive signatories are observed in a “Constant Country Sample.” A broader “Changing Country Sample” includes all passive signatories during this 12-year window. For both samples, export market shares of passive signatories start to increase substantially around ITA accession; results are insensitive to exclusion of China. - Sectoral shares: - ITA product exports amounted to only 0.4 percent of passive signatories’ GDP on average (median 0.1 percent) before accession. - In active signatories, ITA product exports were 2.8 percent of GDP on average (median 1.0 percent) before accession. - Shares of ITA products in countries’ export baskets expanded in response to accession; from the mid-1990s until the dot-com bust in 2001, these shares expanded across all groups. Passive signatories’ mean accession year is computed as late 2002; passive signatories experienced growth of their ITA sector after accession, with shares surging in the mid-2000s and stabilizing subsequently at a higher level. ### 4. Data (dataset construction and scope) - Sample period: 1996-2012. - Dataset matches product-level trade data to tariffs, ITA and WTO membership indicators, and common gravity variables; construction involves conversions across tariff vintages and translation of the ITA agreement into coverage of tariff lines. - Tariff data: - Tariffs at the 6-digit level obtained from UN TRAINS and converted to the HS1996 vintage. - ITA coverage mapping: - The ITA’s list of covered products remained unchanged between 1997 and 2012. - The ITA lists 154 products, translated into a coverage of 106 product lines at the 6-digit HS1996 level (154 products cover 95 of these lines fully and 106 lines almost fully). - Trade data: - Gross imports on these product lines obtained from UN Comtrade; where import data are not available, export data reported by trading partners are filled in. - Value-added / GVC considerations: - Value-added trade data lack the granularity to identify the ITA sector at detailed tariff-line level; to add a value chain dimension, product lines are classified into mostly intermediate and mostly final goods using Sturgeon and Memedovic (2010) at HS 6-digit level: - Classification covers 47 of 106 ITA product lines. - Resulting usable observations when splitting by nature: 680 thousand intermediate-good observations and 825 thousand final-good observations of a total of 2.5 million ITA observations. - Control sectors used: - Other ICT goods not covered by the ITA: - OECD (2003, 2011) definitions cover 193 HS1996 6-digit lines; 77 overlap with the 106-line ITA definition, leaving 116 non-ITA ICT lines. - Broad machinery sector: - HS sections 84, 85, 87, and 90 (electrical and non-electrical machinery, road vehicles, optical/photographic/precision instruments) selected as a larger comparator group; this sector covers 995 tariff lines and includes all those related to ITA and other ICT products. - Final dataset size: totals 21.8 million usable observations. - Additional dataset construction notes: - Non-time-variant variables such as distance are controlled for by country-pair-product fixed effects in specifications. - Appendix provides further construction details, including conversions across tariff vintages and translation of ITA agreement coverage; Appendix Tables and Figures referenced for robustness and additional estimations. *Source: IMF Working Paper "wp18109 - 3. The Information Technology Agreement (ITA)".* ### 5. Empirical Strategy ### 5. Empirical Strategy ### Product-level structural gravity - Estimation strategy based on the structural gravity model of Anderson and Van Wincoop (2003) expanded to exporter, importer, product, and time dimensions to minimize aggregation bias and allow product-level variation in trade costs and elasticities. - Four-dimensional gravity equation (as in source): - x_ijkt = y_jkt x_ikt / y_kt (T_ijkt Π_jkt P_ikt)^{1−σ} - Equivalently: x_ijkt = x_ikt P_ikt^{1−σ} · y_jkt Π_jkt^{1−σ} · T_ijkt^{1−σ} · 1/y_kt - Definitions: - x_ijkt: imports of country i from country j of good k (t denotes years) - y_jkt: total production of good k by country j - x_ikt: total expenditure on good k in country i - y_kt: global production of good k - T_ijkt: bilateral trade costs - Π_jkt: average outward trade barriers faced by country j (multilateral resistance) - P_ikt: average inward trade barriers of country i (multilateral resistance) - Elasticity of substitution between varieties from different exporters is 1−σ, requiring σ > 1. σ is constrained to be the same across all products in this analysis. - Multilateral resistance terms Π_jkt and P_ikt vary over time. - Two-step estimation strategy: - First-stage PPML estimation on bilateral product-level data quantifies discriminatory trade policy (preferential tariffs) and returns consistent importer- and exporter-product-time fixed effects. - Second-stage OLS regressions use those fixed effects as dependent variables to quantify non-discriminatory trade policy effects. - Importer-product-time coefficients (from stage 1) are used to estimate: MFN tariff reduction effect, MFN tariff elimination effect, and the commitment (trade policy certainty) effect on the importer side. - Exporter-product-time coefficients are used to estimate the commitment effect on the exporter side. - Iceberg trade cost specification used: - T_ijkt = (1 + t_Applied_ijkt) · exp[ζ I_ijt + δ_ijk] = exp[ln(1 + t_Applied_ijkt) + ζ I_ijt + δ_ijk] - t_Applied_ijkt: product-level bilateral applied tariff (preferential where present, MFN otherwise) - I_ijt matrix includes gravity regressors varying across importer-exporter-time: dummies for joint currency union membership, common preferential or regional trade agreement (RTA), a dummy capturing exports to EU partners of countries that joined the EU after 1997, a dummy capturing exports to the U.S. of countries that joined an RTA with the U.S. after 1997, and a “both in WTO” variable. - δ_ijk: fixed effect accounting for country-pair-product determinants constant over time (observable or unobservable). - Note: two ITA tariff lines (HS 381800 and HS 950410) are not covered by the four HS sections but retained in the dataset. ### First-stage estimation equation - First-stage gravity equation estimated in multiplicative form by Poisson Pseudo Maximum Likelihood (PPML) to avoid heteroskedasticity bias and retain zero trade flows: - x_ijkt = exp[α_ikt + α_jkt + α_ijk + (1−σ) ln(1 + t_Applied_ijkt) + ψ I_ijt] · ε_ijkt - Fixed effects: - α_ikt (importer-product-time) absorb x_ikt / P_ikt^{1−σ} - α_jkt (exporter-product-time) absorb y_jkt / Π_jkt^{1−σ} - α_ijk (country-pair-product) absorb (1−σ) δ_ijk - ψ = (1−σ) ζ and ε_ijkt is the error term. - Applied tariff term t_Applied_ijkt captures preferential tariff margins over MFN in this equation; MFN tariffs are absorbed by α_ikt because MFN is constant across exporters within an importer-product-time cell. - First-stage equation controls for reversals of trade diversion resulting from ITA implementation (tariff preference pre-ITA and zero preference post-ITA for affected country-pairs). - Estimation approach and implementation notes: - Use of Stata command ppml_panel_sg for first-stage PPML estimations. - Cluster standard errors by country-pair-product combinations in stage 1. - Three sets of high-dimensional fixed effects make estimation computationally challenging. ### Second-stage importer equation - Decomposition of inward multilateral resistance P_ikt: - P_ikt = exp[γ_1 ln(1 + t_MFN_ikt) + γ_2 t0_MFN_ikt + γ_3 imita_ikt] · p_ikt - Interpretation of components: - ln(1 + t_MFN_ikt): captures tariff elimination/reduction effect - t0_MFN_ikt: dummy = 1 if MFN tariff is zero (tariff elimination effect beyond reduction) - imita_ikt: importer ITA membership dummy capturing commitment/trade policy certainty effect (split into active and passive signatories in estimations) - p_ikt: unexplained remainder of inward multilateral resistance - Second-stage OLS importer regression using first-stage α_ikt as dependent variable: - α_ikt = δ_it + δ_ik + δ_kt + β_1 ln(1 + t_MFN_ikt) + β_2 t0_MFN_ikt + β_3 imita_ikt + ξ_ikt - β_l = (1−σ) γ_l for l = 1,2,3 - Fixed effects: - δ_it: importer-time (controls for country-time heterogeneity such as national business cycles and accession-related reforms) - δ_ik: importer-product (time-constant part of country-product heterogeneity, e.g., expenditure patterns and pre-existing industry structure) - δ_kt: product-time (global product-specific shifts over time) - Three effects of non-discriminatory trade policy identified: - MFN tariff reduction effect β_1: import demand elasticity in response to MFN tariff changes (preferential and RTA effects cleaned in stage 1). - Tariff elimination effect β_2: additional gains when MFN tariffs are eliminated completely (t0_MFN_ikt = 1). Rationale: zero MFN tariffs reduce customs inspection frequency and scrutiny, lower customs clearance costs, and thus increase trade beyond pure price effects. Risk-based customs systems reduce inspections for low-risk declarations; zero MFN tariffs remove incentives to under- or mis-declare and to misrepresent origin, lowering inspection probability and rigor. - Supporting considerations: customs systems (e.g., ASYCUDA) and risk-based selection; references to customs literature and “time in trade” impacts on trade. - Importer ITA membership (commitment) effect β_3: higher trade policy certainty from ITA commitments (bound WTO tariffs reduced to zero) lowers risk of reversal and can incentivize investment and trade relationships; split into active and passive signatory dummies to capture heterogeneity. - Identification: relies on cross-product variation (how ITA imports performed relative to other ICT products within same year after controlling for product-time and country-time effects). - Notes on other variables: - First-stage α_ikt had absorbed ln(x_ikt) − (1−σ) ln(P_ikt); product-time fixed effects δ_kt later pick up 1/y_kt. ### Second-stage exporter equation - Decomposition of outward multilateral resistance Π_jkt: - Π_jkt = exp(γ_4 exita_jkt) · π_jkt - exita_jkt: dummy = 1 when exporter is an ITA member; captures exporter-side commitment effect through location and sourcing decisions, FDI flows, and technology transfer influenced by import availability and institutional assurances. - Second-stage OLS exporter regression using first-stage α_jkt as dependent variable: - α_jkt = δ_jt + δ_jk + δ_kt + y_jkt + β_4 exita_jkt + η_jkt - β_4 = (1−σ) γ_4 - Fixed effects: - δ_jt: exporter-time - δ_jk: exporter-product (absorbs country-product determinants like industrial structure) - δ_kt: product-time - π_jkt and y_jkt are the unexplained remainder of outward multilateral resistance and unobservable total production of product k in country j, respectively; any variation not absorbed by fixed effects enters η_jkt. - Note: α_jkt is the natural log of the first-stage fixed effect coefficients. ### Concerns: Omitted variables and endogeneity - Fixed effects mitigate omitted variable bias substantially, but residual concerns: - First-stage: omitted country-pair-specific determinants that vary over time and correlate with regressors could bias estimates, but judged low risk given pair-specific controls; importer- and exporter-product-time effects (primary interest) likely least affected. - Second-stage: omitted variable bias requires ikt-specific variation correlated with the three layers of effects. Example concern: adoption of non-tariff measures (NTMs) that increase after tariff reductions (Kuenzel, 2018). For serious bias of ITA commitment effect, NTMs would need to be targeted exclusively at ITA goods while omitting similar ICT goods — considered unlikely. If narrowly targeted NTMs existed, bias would be toward zero. - Endogeneity concerns: - Empirical evidence that sectors with higher import penetration receive greater protection; high import penetration could influence tariff levels, likelihood of zero tariffs, and ITA accession decisions, and negotiators’ product selection. - In absence of instruments, strategy follows literature standard relying on detailed fixed effects (Baier and Bergstrand, 2004, 2007). Importer-exporter-product fixed effects account for largely time-invariant bilateral characteristics that influenced selection into ITA membership; importer- and exporter-product-time fixed effects help control for endogenous selection of products into the agreement and time-varying changes (e.g., import penetration) that may have swayed decisions. - Any remaining endogeneity is argued to bias estimates toward zero, so estimates can be interpreted as lower bounds if endogeneity persists. - Quasi-natural experiment using passive signatories: - Limiting estimation sample to passive signatories (whose ITA accessions were largely exogenous to ITA product list decisions) provides a specification largely devoid of endogeneity concerns; comparison with full-sample results suggests the estimation strategy addresses endogeneity well in the full sample. *Source: IMF Working Paper — 5. Empirical Strategy (wp18109 - 5. Empirical Strategy).* ### 6. Results ### 6. Results ### Overview: layers of ITA trade effects (Table 2) - Baseline uses the full sample (Table 2). Results from a sample including only passive signatories and non-members closely mirror the baseline. - Table 2 distinguishes layers of impact and covers: - Two product splits: ICT goods (narrow ITA and other ICT goods) and machinery (broader sector including non-IT machinery). - Three goods categories: All goods, Intermediate, Final. - Stage structure: - Stage 1 (PPML): dependent variable bilateral product-level imports; key regressor: ln(1+tariff) capturing preferential tariff reduction (implied tariff elasticity 1−σ). - Stage 2 (OLS) for importers: dependent variable importer-product-time fixed effects; key regressors include MFN tariff reduction, tariff elimination dummy, and ITA commitment effects for active and passive ITA importers. - Stage 2 (OLS) for exporters: dependent variable exporter-product-time fixed effects; key regressors include ITA commitment effects for active and passive ITA exporters and for exporters other than China (coefficients from an exact-analog regression excluding China). - Observations and fit (selected examples from Table 2): - Stage 1 observations: 8,120,335; 32,162,466; 1,583,430; 3,329,109; 2,466,917; 3,069,076. - Stage 1 R2: 0.9837; 0.9867; 0.9730; 0.9784; 0.9926; 0.9914. - Stage 2 importer observations: 367,521; 1,729,821; 75,069; 151,923; 105,377; 148,161. Adjusted R2: 0.8333; 0.8333; 0.8266; 0.8507; 0.8548; 0.8306. - Stage 2 exporter observations: 335,896; 1,389,361; 63,661; 133,096; 97,005; 130,472. Adjusted R2: 0.7336; 0.7012; 0.7508; 0.7405; 0.7237; 0.7234. - Significance notation: +, *, **, *** denote 10, 5, 1, 0.1 per cent significance levels. - China sensitivity: impacts at most layers are insensitive to excluding China; export commitment effects show some limited variation when China is excluded. ### 6.1. Preferential and MFN tariff reduction effects - Stage 1 (applied tariff variable ln(1+tariff)): - Estimates imply requirement of σ >1 is fulfilled. - A one percent reduction of tariffs on ITA products would cause a 0.5 to 1.6 percent increase in imports. - Effects are stronger for intermediate than for final goods. - Coefficient magnitudes vary between ICT and machinery samples; machinery-sample estimates are lower in absolute value. - Interpretation and context: - Lower estimated elasticities compared with some prior literature likely due to high concentration in the IT sector. - Examples of high concentration and market power cited (DRAM cartel fines for 1999 to 2002). - Stage 2 (importer-level, MFN tariff reductions): - Preferred estimate: imports increase by 0.7 percent in response to a 1 percentage point reduction of MFN tariffs. - Impact larger on intermediate goods; ICT-sample estimates are smaller and less precise. ### 6.2. Tariff elimination effect - Novel non-linear effect: zero-tariff dummy captures additional impact of MFN tariff reaching zero. - Estimated impact: - Tariff elimination is estimated to boost eligible imports by about 6-13 percent; preferred estimate from the machinery sample forms the higher end of the range. - Note: exp(0.063)-1=6.5%; exp(0.121)-1=13% (range obtained from Regressions 1 and 2). - For intermediate goods: impact estimated in the 14-18 percent range. - For final goods: up to a 5 percent trade increase in the preferred machinery-sample regression; coefficients smaller and insignificant on the ICT sample. - Mechanisms: - Zero MFN tariffs reduce frequency and rigor of border inspections, expedite customs clearance, and reduce time in transit, generating an additional trade-cost reduction beyond tariff-rate declines. - Policy implications: - Strong case for eliminating already-low tariffs: modest fiscal revenue loss but substantial trade gains. - Sectoral tariff elimination agreements at the WTO can provide credibility to secure elimination. - Tariff elimination in other sectors (machinery, chemicals) likely to produce trade expansion beyond tariff reduction alone. - Reintroducing moderate tariffs on products currently benefiting from zero tariffs could disproportionately hurt trade; trade-war analyses should account for tariff-impact non-linearities. - Literature connection: - Explains Feenstra (2008)’s large estimates of import-price pass-through for ITA goods as potentially driven by non-linearity around tariff elimination rather than only low pre-accession tariffs. - Connects to Yi (2003): tariff elimination incidence could explain higher potency of tariff reductions during the 1980s and ’90s versus earlier decades. ### 6.3. Commitment effect for importers - Concept: whether tariff elimination implemented through hard-to-reverse WTO bindings under ITA delivers larger benefits than unilateral liberalization. - Findings for passive signatories: - ITA commitment effects boosted their intermediate imports by 26-60 percent; preferred estimate from the ICT sample forms the higher end of the range. - Final goods imports for passive signatories did not respond strongly; 7-14 percent impact in the all goods sample driven largely by intermediates. - Findings for active signatories: - Importer commitment effect for active signatories suggests final goods imports increased more in active ITA signatories; coefficient from the ICT sample indicates exporters increased their ITA goods shipments to active signatories by 20 percent more than those of ICT goods. - Evidence weaker on the machinery sample. - Interpretation: - GVC narrative: passive signatories integrated downstream; active signatories moved downstream production stages to passive signatories, increasing final goods imports in active signatories. - Importer commitment effects for active signatories capture relative shifts between destinations; exporter commitment effects capture trade increases across destinations. ### 6.4. Commitment effects for exporters - Overall exporter-side impacts for passive signatories: - Passive signatories’ ITA exports increased by 17-23 percent more than those of other ICT or machinery goods after ITA accession. - This overall effect is largely driven by final goods, where ITA exports outperformed control groups by 26-40 percent. - Effects remain significant and broadly similar when Chinese exports are excluded. - Intermediate exports: - Intermediate goods exports among passive signatories did not expand strongly; estimates positive but with low statistical significance. - Active signatories: - Commitment effects for active signatories’ exports are not very strong; coefficients statistically insignificant, suggesting ITA exports evolved broadly in line with control product groups. - Weak evidence of a contraction in active signatories’ final ITA exports, consistent with production shifting to passive signatories and expansions centered in passive signatories’ downstream activities. - Interpretation: - Consistent with value-chain literature: initial GVC entrants typically join in downstream activities (assembly, final-goods production) which require more readily acquirable capabilities. ### 7. Extensions and Robustness (intro and extensive margin) - Section 7 presents four extensions and robustness checks: 1. Explore the extent to which trade gains were driven by the extensive margin. 2. Confirm quasi-natural experiment sample results mirror the baseline. 3. Test robustness to different ways of introducing tariff reduction and elimination effects. 4. Illustrate how country characteristics influenced passive signatories’ gains from ITA. - 7.1 Extensive margin (approach summary): - Define extensive margin as the share of tariff lines registering positive trade flows (following Martincus et al. (2010) and Silva et al. (2014)). - Calculate, for each importer-exporter-year, the number of positively traded tariff lines for ITA goods and for each control sector; obtain shares by dividing positively traded lines by total tariff lines for the sector. - Explore two estimation samples: ICT goods and machinery goods; produce two observations (ITA and control) for each importer-exporter-year. - Also calculate shares for intermediate and final goods analogously. Appendix Table A11 presents sample means for various subsamples. *Italic source: wp18109 - 6. Results (PDF chapter/section) — canonical source: https://www.imf.org/-/media/files/publications/wp/2018/wp18109.pdf* ### Section II.C in WTO (2014) reviews this literature. ### Section II.C in WTO (2014) reviews this literature. ### Extensive margin results (Table 3) - Method: - Two-stage estimation technique; Stage 1 for extensive margin estimated by OLS (note: other tables use Poisson). - Dependent variable (stage 2): bilateral share of products with positive imports in a sector; stage 2 regressions include importer-product-time or exporter-product-time fixed effects. - Key estimated commitment effects (relative changes in trade share of ITA goods vs control sectors): - Importer commitment effect, active ITA importers: 0.010 ∗∗∗, 0.012 ∗∗, 0.037 ∗∗∗, 0.049 ∗∗∗, -0.014 ∗, -0.014 ∗ (t-statistics in parentheses shown in source: (3.93)(3.06)(3.96)(5.06)(-2.01)(-2.27)). - Importer commitment effect, passive ITA importers: 0.006 +, 0.014 ∗∗, 0.035 ∗∗∗, 0.039 ∗∗∗, -0.010, 0.003 (t-statistics: (1.80)(2.61)(4.46)(3.68)(-1.38)(0.44)). - Exporter commitment effect, active ITA exporters: 0.016 ∗∗∗, 0.027 ∗∗∗, 0.054 ∗∗∗, 0.061 ∗∗∗, -0.009, 0.010 ∗ (t-statistics: (6.19)(6.23)(5.26)(5.97)(-1.40)(2.03)). - Exporter commitment effect, passive ITA exporters: 0.009 ∗∗, 0.015 ∗∗, 0.017 ∗, 0.014, 0.007 +, 0.012 ∗ (t-statistics: (2.97)(2.99)(2.29)(1.56)(1.93)(2.43)). - Exporter commitment effect excluding China: 0.009 ∗∗, 0.015 ∗∗, 0.016 ∗, 0.010, 0.009 ∗, 0.013 ∗∗ (t-statistics: (2.94)(2.96)(2.09)(1.21)(2.45)(2.59)). - Magnitude and interpretation: - Active signatories: importer share of ITA products from an average trade partner increased by around 1 percentage point relative to control sectors (non-negligible given a 0.15 average share for import activity of active signatories in 1996). - Diversification driven by intermediates: increase of around 4 percentage points relative to a 1996 share of 0.21 on the sourcing side. - Active signatories’ exporter commitment effects: boost in share of products exported by around 2 percentage points more than control group; intermediates registered a relative increase of about 6 percentage points. - Passive signatories: importer share rose by about 1 percentage point in ITA goods relative to control sectors; intermediates increased close to 4 percentage points. Exports broadened by about 1 percentage point, driven more by final goods for passive signatories. - Observations and fit: - Observations reported in Table 3 (varies by regression set): e.g., 7,558; 7,616; 7,090; 7,428; 7,350; 7,420 (importer-stage regressions) with Adjusted R2 values such as 0.93490, 0.83000, 0.75680, 0.66280, 0.86410, 0.8411. - Notes: - Extensive margin regressions eliminate tariff reduction and elimination effects because dependent variable is share of traded goods. - Silva et al. (2014) find OLS performs better than log-linear, Poisson and negative binomial for fractional response extensive margin; alternative estimators not implementable with three sets of high-dimensional fixed effects. ### Quasi-natural experiment of passive signatories (Table 4) - Sample: Restricted sample excluding active ITA signatories in the second stage. - Stage 2 importer-side results: - MFN tariff reduction: -0.212, -0.747 ∗∗∗, -0.214, -1.034 ∗∗, -0.687 +, -0.255 (t-statistics provided in source: (-0.98)(-7.82)(-0.32)(-2.96)(-1.79)(-1.03)). - Tariff elimination effect: 0.102 ∗∗∗, 0.160 ∗∗∗, 0.312 ∗∗∗, 0.243 ∗∗∗, -0.0240, 0.0799 ∗ (t-statistics: (4.16)(14.16)(4.67)(5.17)(-0.64)(2.29)). - Commitment effect, passive ITA importers: 0.0904 ∗, 0.0133, 0.411 ∗∗∗, 0.221 ∗∗, 0.0161, 0.0286 (t-statistics: (2.26)(0.42)(4.07)(2.96)(0.24)(0.45)). - Observations: 238,930; 1,130,605; 47,297; 99,525; 70,792; 98,968. Adjusted R2 examples: 0.7434, 0.7217, 0.7277, 0.7673, 0.7568, 0.7343. - Stage 2 exporter-side results: - Commitment effect, passive ITA exporters: 0.173 ∗∗, 0.284 ∗∗∗, 0.312 ∗, 0.324 ∗∗, 0.136, 0.330 ∗∗∗ (t-statistics: (2.82)(6.34)(2.01)(3.08)(1.24)(3.37)). - Commitment effect excluding China: 0.144 ∗, 0.281 ∗∗∗, 0.289 +, 0.334 ∗∗, 0.0968, 0.291 ∗∗ (t-statistics: (2.27)(6.02)(1.80)(3.04)(0.85)(2.93)). - Observations: 194,665; 760,091; 34,548; 77,512; 59,277; 77,708. Adjusted R2 examples: 0.6759, 0.6427, 0.6987, 0.6879, 0.6571, 0.6629. - Interpretation: - Results closely mirror baseline, reinforcing robustness to endogeneity concerns. - For importer-side, MFN tariff reduction and elimination effects remain statistically significant and for intermediate goods may be stronger. - Exporter-side commitment effects on the all goods sample slightly stronger and appear more driven by intermediate goods in this restricted sample. - Caution: these regressions have fewer observations due to classification and sample restrictions. ### Lagged tariffs and interaction of tariff reduction and tariff elimination effects (Table 5 and discussion) - Lagged tariff reduction analysis: - Stage 1 (Preferential tariff reduction regressors across lags) reported coefficients: - Preferential tariff reduction: -1.150 ∗∗∗, -0.472 ∗, -0.808 ∗∗, -0.288 +, -0.841 ∗∗, -0.308 ∗ (t-statistics shown: (-3.66)(-2.03)(-2.85)(-1.76)(-2.92)(-2.04)). - Preferential tariff reduction,t-1: -0.676 ∗, -0.294, -1.058 ∗∗∗, -0.398 ∗∗ (t-statistics: (-2.15)(-1.43)(-3.78)(-2.63)). - Preferential tariff reduction,t-2: -0.822 ∗∗, 0.240 (t-statistics: (2.88)(1.02)). - Sum t to t-2 examples: -1.150 ∗∗∗, -0.472 ∗, -1.484 ∗∗, -0.582 +, -1.077 +, -0.466 (t-statistics: (-3.66)(-2.03)(-2.91)(-1.74)(-1.66)(-1.05)). - Stage 2 importer-side: - MFN tariff reduction contemporaneous and lagged coefficients reported, e.g., MFN tariff reduction: -0.265, -0.687 ∗∗∗, -0.189, -0.666 ∗∗∗, -0.227, -0.638 ∗∗∗ (t-statistics: (-1.47)(-8.81)(-0.67)(-5.78)(-0.71)(-4.86)). - MFN tariff reduction,t-1 and t-2 coefficients also reported; Sum t to t-2 examples: -0.265, -0.687 ∗∗∗, -0.095, -0.547 ∗∗∗, -0.801 ∗∗∗, -0.750 ∗∗∗ (t-statistics: (-1.47)(-8.81)(-0.46)(-6.10)(-3.41)(-7.30)). - Tariff elimination effect: 0.063 ∗∗∗, 0.121 ∗∗∗, 0.049 ∗∗, 0.092 ∗∗∗, 0.045 ∗∗, 0.086 ∗∗∗ (t-statistics: (4.04)(14.75)(3.16)(11.09)(2.87)(10.27)). - Commitment effect, active ITA importers: -0.024, 0.019, -0.046, 0.022, -0.079, -0.005 (t-statistics: (-0.50)(0.50)(-0.88)(0.54)(-1.51)(-0.12)). - Commitment effect, passive ITA importers: 0.131 ∗∗∗, 0.0689 ∗, 0.116 ∗∗, 0.083 ∗∗, 0.112 ∗∗, 0.093 ∗∗ (t-statistics: (3.38)(2.23)(2.95)(2.62)(2.75)(2.80)). - Stage 2 exporter-side: - Commitment effect, active ITA exporters: 0.003, -0.049, 0.019, -0.015, 0.022, -0.003 (t-statistics: (0.05)(-0.98)(0.31)(-0.30)(0.37)(-0.06)). - Commitment effect, passive ITA exporters: 0.157 ∗∗, 0.210 ∗∗∗, 0.124 ∗, 0.193 ∗∗∗, 0.118 +, 0.184 ∗∗∗ (t-statistics: (2.70)(4.95)(2.04)(4.36)(1.92)(4.07)). - Commitment effect excluding China: 0.126 ∗, 0.216 ∗∗∗, 0.092, 0.209 ∗∗∗, 0.078, 0.199 ∗∗∗ (t-statistics: (2.06)(4.82)(1.42)(4.42)(1.18)(4.11)). - Interpretation: - Firms typically react to MFN tariff reductions within about a year; contemporaneous MFN coefficients virtually unchanged and first lag statistically insignificant. - Preferential tariff reductions exhibit a significant first lag similar in magnitude to contemporaneous regressor, suggesting some adjustment time for firms regarding eligibility and documentation for new RTA preferences; adjustments appear complete by the second year (second lag coefficients counterintuitively positive). - Estimates of tariff elimination and ITA commitment effects are not much affected by introducing lagged tariff reduction effects; tariff elimination effects slightly reduced to a 5-10 percent range (preferred estimate at top end of range). - Passive signatories’ ITA commitment effects virtually unchanged on import side; slight reduction on export side. - Interaction of tariff elimination with pre-elimination average tariff: - Interaction coefficient negative but generally statistically insignificant (Appendix Table A10). - Weak evidence that tariff elimination effect may be higher for countries starting from low “nuisance” tariffs. ### Heterogeneity of ITA export commitment effects alongside country characteristics (Table 6) - Objective: assess whether passive signatories with certain characteristics gained more ITA-related export benefits; determinants considered: - Geographical remoteness (custom ITA-specific measure, with penalty for landlocked countries via export costs per container relative to nearest coastal country). - Education proxied by secondary school completion rates. - Business environment proxied by control of corruption (World Bank WGI). - Institutions proxied by rule of law (World Bank WGI). - Method: - Interact determinants with ITA export commitment effects for active and passive exporters; add interactions to Regression 1 in Table 2. - Combine regression coefficients and percentile values of determinants for passive signatories in year 2010 to compute Table 6. - Table 6 (passive signatories, exporter commitment effect by percentile of determinant): - Remoteness: - 10th percentile: 0.181 ∗∗ - 25th percentile: 0.176 ∗∗ - Median: 0.162 ∗∗ - 75th percentile: 0.092 - 90th percentile: 0.084 - Education: - 10th percentile: 0.157 + - 25th percentile: 0.156 + - Median: 0.124 ∗ - 75th percentile: 0.092 - 90th percentile: 0.078 - Business environment: - 10th percentile: 0.118 - 25th percentile: 0.124 + - Median: 0.136 ∗ - 75th percentile: 0.161 ∗ - 90th percentile: 0.186 + - Rule of law: - 10th percentile: 0.134 - 25th percentile: 0.136 + - Median: 0.139 ∗ - 75th percentile: 0.146 ∗ - 90th percentile: 0.151 - Interpretation and policy-relevant insights: - Geographical remoteness is the strongest impediment to leveraging ITA membership for developing IT exports; for the upper quartile of most remote countries, the ITA exporter commitment effect roughly halves and becomes statistically insignificant. - Improvements in the business environment help leverage ITA benefits; rule of law improvements have a more subdued impact in this analysis. - Lower-education passive signatories experienced larger ITA export increases, possibly reflecting initial GVC integration in downstream assembly stages that rely on unskilled labor. Higher education likely needed later for quality upgrading within GVCs. - Overall interpretation consistent with WTO (2014) survey: broad reforms (addressing remoteness, education, business environment, institutions) are needed to create a favorable environment for achieving and deepening GVC participation. *Source: wp18109 - Section II.C in WTO (2014) reviews this literature.* ### 8. Conclusion ### 8. Conclusion ### Main findings on the ITA and tariff elimination - Tariff elimination under the ITA had large trade effects because it was instituted in a non-discriminatory and durable fashion. - Non-discriminatory trade policies are harder to study than discriminatory ones because identifying their trade effects requires comprehensively controlling for multilateral resistance and endogeneity. - The WTO Information Technology Agreement (ITA) is a unique setting that helps overcome these challenges. - A quasi-natural experiment of a subset of “passive” signatories helps allay endogeneity concerns, because these countries joined the agreement as an unavoidable by-product of pursuing a larger policy objective. - Use of product-level data and the ITA’s partial coverage of the IT sector provides a control group of very comparable ICT products for identification. ### Mechanisms and heterogeneity of impacts - Impacts of tariff changes are non-linear: - There is a sizable bonus to a complete elimination of tariffs, as it makes customs inspections less frequent and burdensome, reducing traders’ costs. - Certainty of trade policies, as achieved through WTO tariff bindings, results in an additional layer of benefits (a “commitment effect”). - Commitment effects also affect exports, with investment and firm location/sourcing decisions likely playing an important role. - Import tariff elimination can help develop export industries—an angle largely neglected in existing gravity literature. - Commitment effects for exports are most important for a subset of largely developing countries (the passive signatory group), which were able to leverage ITA membership to develop downstream export activities. - Whether similar impacts obtain outside the IT and machinery sectors is left for future research, though durable, non-discriminatory tariff elimination is plausibly beneficial across sectors. ### Implications for gravity methodology - Tariff data should be integrated directly into gravity analyses. - Doing so reveals multiple layers of impacts and produces new policy-relevant information. - It shows a premium to complete tariff elimination, especially for goods with high value chain integration potential, consistent with border administration and time-in-trade literatures. - The layered structure allows dummy variables of trade agreement membership to be interpreted as commitment effects, an advantage over traditional gravity studies where a single membership dummy captures many policy changes. ### Policy implications and recommendations - Tariff elimination can cause considerable trade expansion, particularly when backed by strong commitments that convince agents of its durability. - In cases where tariffs are already low, their elimination could return large trade benefits with only small fiscal revenue losses. - Synergies exist between the overall business environment and the gains from durable tariff elimination; complementary reforms are often necessary to fully reap benefits. - It is important to safeguard zero tariffs wherever they have been achieved. *Source: wp18109 - 8. Conclusion* ### Annex I: Evidence on the quasi-natural experiment nature of “passive” signato- ### Annex I: Evidence on the quasi-natural experiment nature of “passive” signatories’ ITA accessions ### Motivation: ITA membership as a prerequisite for WTO accession - All 15 countries that acceded to the WTO since ITA came into force had ITA commitments added to their Protocols of Accession. - All 15 countries were given the option to join the ITA already during but before conclusion of their WTO accession negotiations, as reported in WTO (2012). - Since all these 15 countries were in WTO accession negotiations in 1996, they could therefore have become founding members of the ITA, but only Estonia and Chinese Taipei used this opportunity. - The remaining 13 countries joined the ITA between 1999 and 2006, around the time of their WTO accession. - WTO (2012) highlights that becoming ITA signatories for these countries has been “mostly the result of a process where [ITA] participants have encouraged those in the process of acceding [to the WTO] to join the ITA as a part of their accession package”. - Country-specific evidence: - China: Liang (2007) argues industry was skeptical about ITA membership when President Jiang Zemin announced at the 1997 APEC summit that China would join the ITA; the announcement served to improve trade relations with the United States and other trading partners and effectively accelerated China’s WTO membership. China ultimately joined the ITA only in 2003 upon its accession to the WTO as required by the accession protocols. ### Motivation: ITA membership driven by EU accession and harmonization - ITA membership includes the 12 countries that joined the EU in 2004 and 2007. - Five countries are listed as active signatories: - Estonia, the Czech Republic, Poland, Romania, and the Slovak Republic chose to become ITA founding members long before EU membership. - Seven countries are classified as passive signatories with ITA accession driven by EU-related requirements: - Latvia and Lithuania joined the ITA due to commitments in their WTO Accession Protocols in 1999. - Bulgaria, Cyprus, Hungary, Malta, and Slovenia joined the ITA as part of harmonizing trade policy with the EU during the EU accession process. - Timing examples: - Hungary and Malta joined the ITA in 2004 and Bulgaria in 2007, upon their EU accession. - Slovenia and Cyprus joined the ITA in 2000, four years before the EU enlargement. - Documentary corroboration: - A representative of the Permanent Mission of Cyprus stated: “Indeed, in as far as Cyprus is concerned, accession to the ITA was definitely related to its accession process to the EU. During this process, Cyprus had to fully harmonise with the ’acquis communautaire’, just like all the other candidate countries. Thus, becoming an ITA member was strongly encouraged and supported by the EU, both during the preparation of the tables/lists and the last stages of the accession itself in the WTO context.” - The Commission of the European Communities (2001) report on Bulgaria’s progress: “In February 2001 Bulgaria submitted its list of commitments concerning tariff reductions on IT products and the list was approved by the WTO. At present the formal decision by the committee on the accession of Bulgaria to the ITA is expected.” ### Motivation: ITA membership as a precondition for U.S. free trade agreements - WTO (2012) notes: “Several WTO members have also joined the ITA, which can partly be explained by a US policy that has systematically encouraged partners negotiating a free-trade agreement with the United States to also join the ITA. These include, for example, the Dominican Republic, Guatemala, Honduras, Nicaragua and Panama.” - Additional countries for which evidence indicates ITA accession was tied to U.S. FTA negotiations: Bahrain, Colombia, Morocco, and Peru. - Documentary evidence and examples: - USITC (2006): “As a result of the TPA [Trade Promotion Agreement with the United States], Peru will become a signatory to the WTO ITA, which removes tariff and nontariff barriers to information technology (IT) products”. - ACTPN (2004) on the Dominican Republic: “The ACTPN is particularly pleased that the Dominican Republic, as part of these negotiations, will become a member of the WTO Information Technology Agreement, which provides for duty-free treatment of information technology products. This has been a key objective of U.S. high-tech industries.” - Guatemala, Honduras and Nicaragua committed to join the ITA under the CAFTA-DR (USITC, 2004, p. 133). - Colombia agreed to join the ITA under the U.S.-Colombia RTA provisions (Villarreal, 2014). - Bahrain: USTR (2004) indicates ITA membership of Bahrain was an early negotiation outcome of the U.S.-Bahrain FTA. - Morocco: the International Trade Administration (2004) lists Morocco’s commitment to join the ITA under the U.S.-Morocco FTA as a key opportunity from the agreement. - Panama: the Assistant USTR for the Americas stated before the U.S. Senate Committee on Finance that Panama’s agreement to join the ITA during FTA negotiations benefits the IT products sector of the U.S. Panama joined the ITA in June 1998 while it signed its FTA with the U.S. in June 2007; ITA membership was one of the first conditions to advance FTA negotiations. *Source: Annex I of the referenced IMF working paper (wp18109) on ITA signatory motivations and timing.* ### Appendix Tables ### Appendix Tables ### Table A1: Complete first-stage regression of Tables 2 and 4 - Regression type: Stage 1 (PPML). Dependent variable: bilateral product-level imports. - Key coefficient: Preferential tariff reduction ln(1+tariff) - All goods (ICT): -1.150 ∗∗∗ (t = -3.66) - All goods (Machinery): -0.472 ∗ (t = -2.03) - Intermediate (ICT): -1.608 ∗ (t = -2.37) - Intermediate (Machinery): -0.582 + (t = -1.66) - Final (ICT): -0.601 (t = -1.57) - Final (Machinery): -0.533 (t = -1.15) - Other controls (examples of coefficient estimates and t-statistics): - Exporter late EU signatory: 0.081 (0.83); 0.068 (1.09); 0.059 (0.30); 0.163 (1.50); 0.166 (1.33); 0.353 ∗∗ (2.58) - Exporter late US-FTA signatory: -0.002 (-0.03); -0.018 (-0.38); -0.144 (-1.12); -0.025 (-0.21); -0.000 (-0.00); -0.111 (-1.64) - RTA: 0.022 (0.39); 0.022 (0.50); -0.029 (-0.41); -0.007 (-0.10); 0.050 (0.97); 0.054 (1.13) - Common currency: 0.067 (1.49); 0.044 (1.43); 0.092 (1.21); 0.074 (1.35); 0.114 + (1.66); 0.124 + (1.81) - Both in WTO: -0.096 (-0.71); 0.004 (0.03); -0.040 (-0.22); 0.024 (0.14); -0.154 (-0.54); -0.129 (-0.56) - Observations: - Columns (1)–(6): 8,120,335; 32,162,466; 1,583,430; 3,329,109; 2,466,917; 3,069,076 - R2: - 0.9837; 0.9867; 0.9730; 0.9784; 0.9926; 0.9914 - Notes: - +, *, **, *** denote 10, 5, 1, 0.1 per cent significance levels, based on robust standard errors clustered by country-pair-product combinations. - Regression includes importer-exporter-product, importer-product-time and exporter-product-time fixed effects. - Exporter late EU / US-FTA signatory: Takes value “1” for intra-EU trade (after accession) of all countries that joined the EU after 1997. Analogously for US FTA. ### Table A2: Complete first-stage regression of Table 3 - Regression type: Stage 1 (OLS). Dependent variable: bilateral share of products with positive imports in a sector. - Key coefficients: - Exporter late EU signatory: - All goods (ICT): 0.020 ∗∗∗ (5.74) - All goods (Machinery): 0.009 ∗∗ (3.02) - Intermediate (ICT): 0.016 ∗∗ (2.94) - Intermediate (Machinery): 0.007 (1.45) - Final (ICT): 0.014 ∗∗ (2.69) - Final (Machinery): 0.011 ∗ (2.38) - Exporter late US-FTA signatory: 0.004 (0.40); 0.003 (0.29); 0.003 (0.23); 0.007 (0.67); 0.012 (0.99); 0.010 (0.94) - RTA: - 0.009 ∗∗∗ (7.41); 0.010 ∗∗∗ (8.82); -0.0002 (-0.08); 0.003 (1.48); 0.011 ∗∗∗ (5.95); 0.011 ∗∗∗ (6.55) - Common currency: - 0.016 ∗∗ (2.99); 0.011 ∗ (2.12); 0.021 ∗∗∗ (3.35); 0.014 ∗ (2.19); 0.029 ∗∗∗ (4.48); 0.028 ∗∗∗ (4.61) - Both in WTO: 0.003 (1.00); 0.007 ∗∗ (3.03); 0.001 (0.23); 0.007 + (1.71); -0.005 (-1.03); -0.004 (-0.93) - Observations: - 527,823; 623,089; 366,864; 452,240; 416,251; 435,742 - Adjusted R2: - 0.9432; 0.9487; 0.9219; 0.9241; 0.9206; 0.9232 - Notes: - +,*, **, ∗∗∗ denote 10, 5, 1, 0.1 per cent significance levels, robust standard errors clustered by country-pair-product combinations. - Stage 1 includes importer-exporter-product, import-product-time and exporter-product-time fixed effects. - Exporter late EU / US-FTA signatory: Takes value “1” for intra-EU trade (after accession) of all countries that joined the EU after 1997. Analogously for US FTA. ### Table A3: Complete first-stage regression of Table 5 - Regression type: Stage 1 (PPML). Dependent variable: bilateral product-level imports. - Key coefficients for Preferential tariff reduction ln(1+tariff): - (1): -1.150 ∗∗∗ (-3.66) - (2): -0.472 ∗ (-2.03) - (3): -0.808 ∗∗ (-2.85) - (4): -0.288 + (-1.76) - (5): -0.841 ∗∗ (-2.92) - (6): -0.308 ∗ (-2.04) - Dynamic terms (where reported): - Preferential tariff reduction,t-1: - -0.676 ∗ (-2.15); -0.294 (-1.43); -1.058 ∗∗∗ (-3.78); -0.398 ∗∗ (-2.63) - Preferential tariff reduction,t-2: - -0.822 ∗∗ (2.88); 0.240 (1.02) - Sum t to t-2: - -1.150 ∗∗∗ (-3.66); -0.472 ∗ (-2.03); -1.484 ∗∗ (-2.91); -0.582 + (-1.74); -1.077 (-1.66); -0.466 (-1.05) - Other controls and examples: - Exporter late EU signatory: 0.081 (0.83); 0.068 (1.09); 0.065 (0.65); 0.057 (0.94); 0.089 (0.88); 0.065 (1.05) - RTA: 0.022 (0.39); 0.022 (0.50); 0.023 (0.43); 0.020 (0.49); 0.043 (0.90); 0.027 (0.71) - Common currency: 0.067 (1.49); 0.044 (1.43); 0.047 (1.07); 0.031 (1.03); 0.075 (1.66); 0.041 (1.33) - Both in WTO: -0.096 (-0.71); 0.004 (0.03); -0.115 (-0.89); -0.007 (-0.06); -0.153 (-1.28); -0.011 (-0.10) - Observations: - Columns (1)–(6): 8,120,335; 32,162,466; 7,232,615; 28,414,441; 6,697,866; 26,269,922 - R2: - 0.9837; 0.9867; 0.9840; 0.9870; 0.9842; 0.9872 - Notes: - +,*, **, *** denote 10, 5, 1, 0.1 per cent significance levels; robust standard errors clustered by country-pair-product combinations. - Regression includes importer-exporter-product, importer-product-time and exporter-product-time fixed effects. - Exporter late EU / US-FTA signatory as defined in notes. ### Table A4: Robustness — Analogue to Table 2 with positive trade values only - Stage 1 (PPML). Dependent variable: bilateral product-level imports. - Preferential tariff reduction ln(1+tariff): - All goods (ICT): -1.083 ∗∗∗ (-3.40) - All goods (Machinery): -0.444 + (-1.91) - Intermediate (ICT): -1.614 ∗ (-2.34) - Intermediate (Machinery): -0.568 (-1.62) - Final (ICT): -0.566 (-1.45) - Final (Machinery): -0.530 (-1.14) - Observations (Stage 1): 5,635,912; 22,105,280; 1,100,335; 2,489,738; 1,683,823; 2,113,942 - R2: 0.9838; 0.9867; 0.9731; 0.9784; 0.9926; 0.9914 - Stage 2 for importers (OLS). Dependent variable: importer-product-time FEs. - MFN tariff reduction: - -0.038 (-0.23); -0.507 ∗∗∗ (-7.07); 0.251 (0.51); -0.497 ∗ (-2.09); -0.203 (-0.67); 0.076 (0.40) - Tariff elimination effect: - 0.012 (0.87); 0.023 ∗∗ (3.08); 0.067 + (1.83); 0.052 + (1.80); -0.056 ∗ (-2.36); -0.020 (-0.88) - Commitment effect: active ITA importers: - -0.024 (-0.53); 0.063 + (1.74); 0.063 (0.59); 0.129 (1.62); 0.120 (1.43); 0.076 (0.94) - Commitment effect: passive ITA importers: - 0.117 ∗∗∗ (3.32); 0.076 ∗∗ (2.69); 0.342 ∗∗∗ (3.78); 0.196 ∗∗ (2.87); 0.033 (0.56); 0.047 (0.83) - Observations (Stage 2 importers): 367,521; 1,729,821; 75,069; 151,923; 105,377; 148,161 - Adjusted R2: 0.8276; 0.7908; 0.8154; 0.8445; 0.8488; 0.8229 - Stage 2 for exporters (OLS). Dependent variable: exporter-product-time FEs. - Commitment effect: active ITA exporters: - 0.080 (1.48); 0.026 (0.62); 0.152 (1.02); 0.083 (0.73); 0.027 (0.30); 0.041 (0.49) - Commitment effect: passive ITA exporters: - 0.159 ∗∗ (3.19); 0.114 ∗∗ (3.10); 0.179 (1.47); 0.106 (1.23); 0.244 ∗∗ (2.64); 0.240 ∗∗ (2.82) - Commitment effect other than China: - 0.098 + (1.90); 0.082 ∗ (2.14); 0.122 (0.95); 0.108 (1.18); 0.173 + (1.85); 0.165 ∗ (1.96) - Observations (Stage 2 exporters): 335,896; 1,389,361; 63,661; 133,096; 97,005; 130,472 - Adjusted R2: 0.7752; 0.7473; 0.7830; 0.7756; 0.7733; 0.7633 - Notes: - +,*, **, ∗∗∗ denote 10, 5, 1, 0.1 per cent significance levels; robust standard errors clustered appropriately. - Stage 1 regressions include importer-exporter-product, import-product-time and exporter-product-time fixed effects and additional dummies (Both in WTO, late joiners, RTA, common currency). Given fixed effects, those coefficients are insignificant. - Stage 2 importer regressions include importer-time, importer-product and product-time fixed effects. - Stage 2 exporter regressions include exporter-time, exporter-product and product-time fixed effects. - Coefficients excluding China’s exports available upon request; commitment effect for passive ITA exporters substantially varies when excluding China. ### Table A5: Robustness — Analogue to Table A4 with OLS estimation in first stage - Stage 1 (OLS). Dependent variable: bilateral product-level imports. - Preferential tariff reduction ln(1+tariff): - -0.572 ∗∗∗ (-5.29); -0.293 ∗∗∗ (-6.22); -1.144 ∗∗∗ (-3.97); -0.276 ∗ (-2.00); -0.247 (-1.29); -0.475 ∗∗ (-2.94) - Observations (Stage 1): 5,207,713; 20,282,225; 1,023,268; 2,327,039; 1,565,702; 1,950,770 - Adjusted R2 (Stage 1): 0.7695; 0.7640; 0.7859; 0.7997; 0.7633; 0.7748 - Stage 2 for importers (OLS). Dependent variable: importer-product-time FEs. - MFN tariff reduction: - -0.323 ∗ (-2.41); -0.414 ∗∗∗ (-7.06); 0.056 (0.14); -0.298 (-1.61); -0.182 (-0.81); 0.240 (1.59) - Tariff elimination effect: - 0.025 ∗ (2.30); 0.021 ∗∗∗ (3.73); 0.091 ∗∗∗ (3.49); 0.037 + (1.85); -0.008 (-0.41); 0.002 (0.08) - Commitment effects: - Active ITA importers: -0.057 (-1.57); 0.020 (0.67); -0.062 (-0.72); 0.096 (1.49); 0.058 (0.94); -0.003 (-0.05) - Passive ITA importers: 0.062 ∗ (2.18); 0.023 (1.00); 0.145 ∗ (2.01); 0.055 (1.05); 0.026 (0.55); 0.049 (1.06) - Observations (Stage 2 importers): 317,017; 1,468,418; 61,991; 132,276; 94,644; 130,233 - Adjusted R2 (Stage 2 importers): 0.2085; 0.1664; 0.1794; 0.2212; 0.2317; 0.2210 - Stage 2 for exporters (OLS). Dependent variable: exporter-product-time FEs. - Commitment effect: active ITA exporters: - 0.039 (0.95); 0.001 (0.03); 0.036 (0.34); -0.037 (-0.49); 0.068 (0.95); 0.074 (1.12) - Commitment effect: passive ITA exporters: - 0.061 (1.42); 0.050 (1.56); -0.052 (-0.47); -0.020 (-0.29); 0.231 ∗∗ (2.92); 0.226 ∗∗ (3.10) - Commitment effect other than China: - 0.008 (0.18); 0.011 (0.33); -0.187 (-1.52); -0.057 (-0.75); 0.216 ∗∗ (2.66); 0.172 ∗ (2.34) - Observations (Stage 2 exporters): 220,154; 904,298; 42,524; 89,570; 63,401; 84,377 - Adjusted R2 (Stage 2 exporters): 0.0752; 0.0473; 0.0899; 0.1091; 0.1008; 0.0779 - Notes: - +,*, **, ∗∗∗ denote 10, 5, 1, 0.1 per cent significance levels. - Fixed effects and robustness details as in Table A4 notes. ### Table A6: Robustness — Analogue to Table 2 with sector-specific tariff effects - Stage 1 (PPML). Dependent variable: bilateral product-level imports. - Preferential tariff reduction (ITA goods): - -1.252 ∗∗ (-2.59); -1.258 ∗∗ (-2.61); -1.966 ∗ (-2.05); -2.172 ∗ (-2.30); -0.419 (-0.42); -0.315 (-0.31) - Preferential tariff reduction (other goods): - -1.035 ∗ (-2.42); -0.425 (-1.58); -1.374 (-1.48); -0.334 (-0.91); -0.638 (-1.59); -0.605 (-1.20) - Observations (Stage 1): 8,120,335; 32,162,466; 1,583,430; 3,329,109; 2,466,917; 3,069,076 - R2: 0.9837; 0.9867; 0.9730; 0.9784; 0.9926; 0.9914 - Stage 2 for importers (OLS). Dependent variable: importer-product-time FEs. Sector-specific results shown for ITA goods and other goods. - MFN tariff reduction (ITA goods): - 0.042 (0.18); 0.293 (1.39); 0.588 (0.93); 1.735 ∗∗∗ (3.44); -0.713 (-1.43); -0.526 (-1.09) - MFN tariff reduction (other goods): - -0.452 ∗ (-2.10); -0.742 ∗∗∗ (-9.17); -1.246 + (-1.95); -1.213 ∗∗∗ (-4.30); -0.450 (−1.24); 0.083 (0.40) - Tariff elimination effect (ITA goods): - 0.033 + (1.83); 0.031 + (1.82); 0.089 ∗ (2.10); 0.094 ∗ (2.47); -0.075 ∗ (-2.24); -0.064 + (-1.92) - Tariff elimination effect (other goods): - 0.122 ∗∗∗ (4.54); 0.146 ∗∗∗ (15.79); 0.386 ∗∗∗ (4.69); 0.213 ∗∗∗ (4.35); 0.031 (0.83); 0.140 ∗∗∗ (3.96) - Commitment effect: active ITA importers: - -0.007 (-0.13); 0.048 (1.22); 0.103 (0.89); 0.168 + (1.95); 0.196 ∗ (2.24); 0.115 (1.37) - Commitment effect: passive ITA importers: - 0.141 ∗∗∗ (3.64); 0.085 ∗∗ (2.73); 0.505 ∗∗∗ (5.10); 0.252 ∗∗∗ (3.45); 0.078 (1.19); 0.114 + (1.82) - Observations (Stage 2 importers): 367,521; 1,729,821; 75,069; 151,923; 105,377; 148,161 - Adjusted R2: 0.8334; 0.7984; 0.8260; 0.8496; 0.8548; 0.8302 - Stage 2 for exporters (OLS). Dependent variable: exporter-product-time FEs. - Commitment effect: active ITA exporters: - 0.004 (0.05); -0.051 (-1.00); -0.006 (-0.03); 0.092 (0.63); -0.157 (-1.46); -0.034 (-0.35) - Commitment effect: passive ITA exporters: - 0.156 ∗∗ (2.70); 0.212 ∗∗∗ (4.91); 0.208 (1.41); 0.188 + (1.89); 0.229 ∗ (2.18); 0.333 ∗∗∗ (3.52) - Commitment effect other than China: - 0.126 ∗ (2.06); 0.218 ∗∗∗ (4.80); 0.174 (1.13); 0.202 + (1.92); 0.198 + (1.80); 0.306 ∗∗ (3.23) - Observations (Stage 2 exporters): 335,896; 1,389,361; 63,661; 133,096; 97,005; 130,472 - Adjusted R2: 0.7335; 0.6980; 0.7496; 0.7391; 0.7232; 0.7225 - Notes: - +,*, **, ∗∗∗ denote 10, 5, 1, 0.1 per cent significance levels; robust standard errors clustered appropriately. - Fixed effects and exclusions (China) notes as in prior tables. ### Table A7: Robustness — Analogue to Table 2 excluding China’s exports - Stage 1 (PPML). Dependent variable: bilateral product-level imports. - Preferential tariff reduction ln(1+tariff): - -1.060 ∗∗ (-3.11); -0.460 (-1.81); -1.324 (-1.95); -0.399 (-1.17); -0.599 (-1.42); -0.624 (-1.29) - Observations (Stage 1): 7,854,306; 30,991,385; 1,531,220; 3,224,329; 2,385,817; 2,962,625 - R2: 0.9790; 0.9855; 0.9558; 0.9735; 0.9779; 0.9900 - Stage 2 for importers (OLS). Dependent variable: importer-product-time FEs. - MFN tariff reduction: - -0.406 ∗ (-2.23); -0.648 ∗∗∗ (-8.33); -0.686 (-1.25); -0.803 ∗∗ (-3.16); -0.774 ∗ (-2.49); -0.0699 (-0.35) - Tariff elimination effect: - 0.0716 ∗∗∗ (4.54); 0.126 ∗∗∗ (15.36); 0.160 ∗∗∗ (4.00); 0.119 ∗∗∗ (3.80); 0.0246 (0.94); 0.0855 ∗∗∗ (3.45) - Commitment effect: active ITA importers: - -0.0095 (-0.19); 0.0565 (1.45); 0.0288 (0.25); 0.132 (1.53); 0.286 ∗∗ (3.19); 0.197 ∗ (2.33) - Commitment effect: passive ITA importers: - 0.0947 ∗ (2.40); 0.0442 (1.40); 0.424 ∗∗∗ (4.29); 0.201 ∗∗ (2.77); -0.0136 (-0.20); 0.0253 (0.39) - Observations (Stage 2 importers): 364,525; 1,714,610; 74,378; 150,971; 104,543; 146,773 - Adjusted R2: 0.8208; 0.7908; 0.8213; 0.8464; 0.8327; 0.8146 - Stage 2 for exporters (OLS). Dependent variable: exporter-product-time FEs. - Commitment effect: active ITA exporters: - -0.0127 (-0.20); -0.0494 (-1.00); 0.00413 (0.02); 0.0844 (0.59); -0.148 (-1.39); -0.0120 (-0.13) - Commitment effect: passive ITA exporters: - 0.126 ∗ (2.06); 0.216 ∗∗∗ (4.82); 0.173 (1.13); 0.202 (1.94); 0.198 (1.80); 0.303 ∗∗ (3.15) - Observations (Stage 2 exporters): 332,426; 1,373,696; 62,947; 131,758; 96,137; 129,233 - Adjusted R2: 0.7281; 0.6979; 0.7469; 0.7366; 0.7119; 0.7174 - Notes: - +,*, **, ∗∗∗ denote 10, 5, 1, 0.1 per cent significance levels; robustness and fixed-effect structure as in prior tables. - Coefficients excluding China’s exports: commitment effect for passive ITA exporters is the only one to substantially vary; full restricted-sample results available upon request. ### Table A8: Robustness — Analogue to Table 2 excluding country pairs with RTAs by 1996 - Stage 1 (PPML). Dependent variable: bilateral product-level imports. - Preferential tariff reduction ln(1+tariff): -1.062 + (-1.84); -0.457 (-1.28); -1.734 (-1.33); -0.497 (-0.69); -0.806 (-1.31); -0.866 (-1.18) - Observations (Stage 1): 5,701,870; 2,250,1666; 1,106,049; 2,372,692; 1,742,614; 2,190,899 - R2: 0.9881; 0.9891; 0.9768; 0.9771; 0.9951; 0.9933 - Stage 2 for importers (OLS). Dependent variable: importer-product-time FEs. - MFN tariff reduction: - -0.222 (-1.17); -0.714 ∗∗∗ (-8.66); -0.077 (-0.14); -0.616 ∗ (-2.23); -0.174 (-0.54); 0.383 + (1.94) - Tariff elimination effect: - 0.047 ∗∗ (2.79); 0.122 ∗∗∗ (13.54); 0.153 ∗∗∗ (3.53); 0.130 ∗∗∗ (3.86); -0.048 + (-1.67); 0.049 + (1.82) - Commitment effect: active ITA importers: - -0.017 (-0.34); 0.003 (0.08); 0.125 (1.07); 0.200 ∗ (2.30); 0.126 (1.38); 0.042 (0.48) - Commitment effect: passive ITA importers: - 0.121 ∗∗ (2.94); 0.091 ∗∗ (2.71); 0.430 ∗∗∗ (4.11); 0.206 ∗∗ (2.64); 0.007 (0.10); 0.066 (0.99) - Observations (Stage 2 importers): 348,656; 1,629,802; 70,679; 144,657; 100,847; 141,432 - Adjusted R2: 0.8254; 0.7868; 0.8211; 0.8432; 0.8431; 0.8188 - Stage 2 for exporters (OLS). Dependent variable: exporter-product-time FEs. - Commitment effect: active ITA exporters: - -0.034 (-0.53); -0.072 (-1.42); -0.012 (-0.07); -0.011 (-0.08); -0.145 (-1.42); -0.092 (-0.97) - Commitment effect: passive ITA exporters: - 0.124 ∗ (2.04); 0.158 ∗∗∗ (3.48); 0.144 (0.97); 0.156 (1.50); 0.258 ∗ (2.27); 0.362 ∗∗∗ (3.49) - Observations (Stage 2 exporters): 279,417; 1,136,040; 53,096; 111,123; 80,994; 109,302 - Adjusted R2: 0.8018; 0.7495; 0.7965; 0.7890; 0.8102; 0.7850 - Notes: - +,*, **, ∗∗∗ denote 10, 5, 1, 0.1 per cent significance levels; robust clustering; fixed-effect structure as prior. ### Table A9: Extension — Heterogeneity of ITA commitment effects on exports - Stage 2 for exporters (OLS). Dependent variable: exporter-product-time FEs. - Interaction variables analyzed: None; Remoteness; Education; Business environment; Rule of law. - Coefficients (interaction specification reported as pairs: main dummy, interaction term): - Active ITA Exporter: - None: 0.003 (0.05) - Remoteness: 0.102 (1.27) and interaction 0.037 ∗ (2.13) - Education: 0.201 ∗ (1.97) and interaction -0.063 ∗ (-2.23) - Business environment: 0.087 (1.11) and interaction -0.110 ∗∗ (-3.11) - Rule of law: 0.134 (1.56) and interaction -0.167 ∗∗∗ (-3.27) - Passive ITA Exporter: - None: 0.157 ∗∗ (2.70) - Remoteness: 0.103 (1.18) and interaction -0.021 (-0.76) - Education: 0.199 (1.43) and interaction -0.026 (-0.66) - Business environment: 0.150 ∗ (2.54) and interaction 0.041 (0.47) - Rule of law: 0.142 ∗ (2.44) and interaction 0.009 (0.12) - Observations: 335,896; 330,137; 243,187; 323,254; 323,936 - Adjusted R2: 0.7336; 0.7322; 0.7288; 0.7277; 0.7283 - Memorandum item: Interaction variable values in 2010 by percentile for passive ITA signatories (used to compute Table 6) - Remoteness percentiles: - 10th: -3.766 - 25th: -3.560 - 50th: -2.864 - 75th: 0.520 - 90th: 0.910 - Education percentiles: - 10th: 1.607 - 25th: 1.652 - 50th: 2.870 - 75th: 4.048 - 90th: 4.604 - Business environment percentiles: - 10th: -0.773 - 25th: -0.628 - 50th: -0.352 - 75th: 0.249 - 90th: 0.854 - Rule of law percentiles: - 10th: -0.885 - 25th: -0.600 - 50th: -0.329 - 75th: 0.480 - 90th: 0.982 - Notes: - Stage 1 regression same as Regression 1 of Table 2. - +,*, **, ∗∗∗ denote significance levels; robust standard errors clustered as in prior tables. - Definition notes: - Remoteness: sum over importers in any given year of Distance ijt*(ITA Imports it/ITA World Imports t), adjusted for landlocked countries by container export cost multiplier. Distance measured in thousands of kilometers and centered. - Education proxied by secondary school completion rate (Prati et al, 2013), divided by 10. - Business environment proxied by World Bank control of corruption variable. - Rule of law from World Bank World Governance Indicators. - Percentile values for 2010 calculated using sample of passive signatories only. ### Table A10: Extension — Interaction of tariff reduction and tariff elimination effects - Stage 1 (PPML): same as in Table 2. - Stage 2 for importers (OLS). Dependent variable: importer-product-time FEs. - MFN tariff reduction: - -0.262 (-1.39); -0.769 ∗∗∗ (-9.26); -0.838 (-1.51); -0.882 ∗∗ (-3.07); -0.718 ∗ (-2.11); -0.252 (-1.17) - Tariff elimination effect: - 0.0612 ∗∗ (3.07); 0.121 ∗∗∗ (12.17); 0.136 ∗∗ (2.74); 0.106 ∗∗ (2.88); -0.012 (-0.37); 0.068 ∗ (2.24) - Tariff reduction x elimination effect: - -0.247 (-0.78); -0.572 ∗∗∗ (-3.76); -0.049 (-0.07); -0.165 (-0.35); -0.240 (-0.46); -0.501 (-1.24) - Observations: 304,034; 1,466,540; 61,180; 132,314; 86,944; 124,603 - Adjusted R2: 0.8274; 0.8028; 0.8145; 0.8475; 0.8522; 0.8273 - Notes: - +,*, **, *** denote 10, 5, 1, 0.1 per cent significance levels; robust standard errors clustered as in prior tables. - Fixed effects structure and additional dummies as in Table A1; first-stage coefficients available in Table A1. ### Table A11: Average share of products trade with bilateral partners - This table presents sample means by product categories of the dependent variable in the first stage of Table 3 (average share of products traded with bilateral partners). - Values reported (shares): - Active ITA importers in 1996: - ITA: 0.15; ICT: 0.13; Machinery: 0.10; ITA (Intermediate): 0.21; ICT: 0.16; Machinery: 0.16; ITA (Final): 0.22; ICT: 0.18; Machinery: 0.15 - Active ITA importers in 2012: - 0.16; 0.14; 0.09; 0.25; 0.18; 0.14; 0.25; 0.22; 0.18 - Active ITA exporters in 1996: - 0.21; 0.20; 0.15; 0.29; 0.23; 0.23; 0.34; 0.28; 0.22 - Active ITA exporters in 2012: - 0.25; 0.22; 0.15; 0.38; 0.25; 0.21; 0.39; 0.36; 0.28 - Passive ITA importers in 1996: - 0.17; 0.16; 0.11; 0.24; 0.19; 0.20; 0.30; 0.24; 0.19 - Passive ITA importers in 2012: - 0.14; 0.11; 0.07; 0.20; 0.13; 0.11; 0.21; 0.19; 0.15 - Passive ITA exporters in 1996: - 0.08; 0.07; 0.04; 0.13; 0.09; 0.09; 0.15; 0.11; 0.08 - Passive ITA exporters in 2012: - 0.13; 0.11; 0.06; 0.21; 0.14; 0.11; 0.24; 0.19; 0.14 - Notes: - A value of 0.20 for ITA products implies that 20 percent of all ITA products exhibit positive trade, while 80 percent exhibit zero trade flows. - In this table, ICT (Machinery) refers to non-ITA ICT goods (non-ITA Machinery goods), unlike in the regression tables. *Italic line: Source: wp18109 - Appendix Tables (PDF).* ### Appendix Figures ### Appendix Figures ### Data Appendix — ITA product coverage and vintages - The ITA contains a rigid positive listing across 6-digit HS1996 classification: in total it affects 154 product lines, but only 95 product lines are covered fully. - For analysis the authors consider all fully covered lines plus another 11 lines identified by WTO (2012) as including a high proportion of ITA products, yielding a total of 106 lines. - Coverage must be reassessed across HS vintages (HS1996, HS2002, HS2007) because product-line composition changes over time as some ITA products become technologically obsolete. - Separate lists of ITA-covered lines were obtained for: 1996–2001 in HS1996, 2002–06 in HS2002, and 2007–12 in HS2007. - Mapping to a consistent HS1996 basis used UNSD conversion tables: - HS2002 lines map into the exact same set as HS1996. - HS2007 mapping differs, producing a dataset with different numbers of tariff lines across periods: 23 (9) HS1996 lines appear only during the former (latter) period, while 74 HS1996 lines are included in all years, resulting in the total of 106 lines. ### Trade flow and sample construction - 6-digit HS1996 bilateral trade flow data for 1996–2012 were obtained from UN Comtrade (import flows plus exporter-reported mirror data where importer did not report). - The resulting dataset contains 3.86 million observations of non-zero ITA trade flows covering 234 countries (not all observations usable in all regressions due to missing tariff values). ### Tariff data, conversion, and imputation - Tariffs obtained from UN Trains in HSCombined for 1996–2012 (HSCombined reports: 1996–2001 in HS1996, 2002–06 in HS2002, 2007–12 in HS2007). - Conversion to HS1996 used UNSD conversion tables; when multiple HS2007 or HS2002 lines correspond to a HS1996 line, a simple average across the HS2007 or HS2002 lines was taken to obtain the tariff for the HS1996 line. - To fill some missing observations, tariffs were linearly intrapolated between years for which observations exist. - EU is presented as a single country in TRAINS; the dataset was appended to include all its members in various years to achieve consistent coverage of active signatories throughout the sample period. ### Right-hand-side controls and auxiliary datasets - Standard gravity variables varying across time within country or country-pair were collected. - GDP and GDP per capita taken from Penn World Table Version 8.0. - RTA and currency union membership data taken from De Sousa (2012); De Sousa (2012) data cover currency union relationships up to 2009. To extend the data, Estonia joining the Euro in 2011 was added; no other changes assumed after 2009 and before 2013. - The currency-union definition is transitive (Glick and Rose (2002) style). - Remoteness measure computed by weighting bilateral distances by trading partner shares in world GDP and combining importer and exporter remoteness by multiplication before taking the natural logarithm. - WTO membership data collected from the WTO website. ### Classification for GVC/ICT/machinery analysis - Sturgeon and Memedovic (2010) classification (HS2007 basis) used to distinguish primarily final vs. intermediate goods for electronics; converted to HS1996 using UNSD conversion tables. - After conversion, this classification covers 47 of the 106 ITA product lines. - When the ICT (machinery) control sector is added, 102 of 202 (165 of 995) lines are covered by the classification (authors’ note). - ITA products classified into 7 broad product categories (number of 6-digit HS1996 lines in parentheses): - Computers (14) - Instruments and apparatus (17) - Parts and accessories (32) - Semiconductor manufacturing equipment (10) - Semiconductors (15) - Data-storage media and software (9) - Telecommunications equipment (9) - Computers, semiconductors, and parts and accessories together make up around 80 percent of ITA product trade flows. ### Control sectors used for comparison - Two control sectors: - Non-ITA ICT goods (OECD definition) — OECD covers 193 product lines in HS1996; 77 of these are also covered by the ITA under the 106-line definition, leaving 116 non-ITA ICT lines as the ICT control sector, adding another 3.70 million observations to the dataset. - Broad machinery sector — HS sections 84, 85, 87, and 90 (electrical and non-electrical machinery, road vehicles, optical/photographic/precision instruments); comprises all ITA and OECD-ICT tariff lines and brings the full dataset to a total of 28.36 million observations. - Notes: 29 lines are covered by the ITA that are not considered OECD ICT goods (examples include printing machinery, electric typewriters and optical photocopiers; laser discs and magnetic tapes; electric and power capacitors; equipment for measuring liquid or gas; parts and accessories). Two ITA tariff lines (HS 381800; HS 950410) are not covered by the four HS sections but remain in the dataset. ### Dataset scale and coverage summary - Total non-zero ITA trade-flow observations: 3.86 million. - Countries covered: 234. - Total ITA-related HS1996 lines used: 106. - Additional non-ITA ICT control lines: 116 (adding 3.70 million observations). - Final combined dataset size including broad machinery sector: 28.36 million observations. - Temporal coverage: 1996–2012 (tariffs and trade flows), with vintages HS1996, HS2002, HS2007 reconciled to HS1996 via conversion tables. *Source: wp18109 - Appendix Figures* --- _Source: https://www.imf.org/-/media/files/publications/wp/2018/wp18109.pdf_