## Box 1: Gravity Model: Explanation of Variables ## Source details **Canonical URL:** [Box 1: Gravity Model: Explanation of Variables](https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024032-print-pdf.pdf) ## Other formats - [Markdown version](/-/media/files/publications/wp/2024/english/wpiea2024032-print-pdf.pdf.md) - [Structured JSON version](/-/media/files/publications/wp/2024/english/wpiea2024032-print-pdf.pdf.json) --- ### Trade variables and regional dummies - LTrTOTij — Log of total trade flows, merchandise and services, in current US dollars between Country i and Country j - LTrMRCij — Log of total merchandise trade flows in current US dollars between Country i and Country j - LTrMANUij — Log of trade flows of manufacturing goods in current US dollars between Country i and Country j - LTrPRIPDij — Log of trade flows of primary commodities in current US dollars between Country i and Country j - LTrSERVij — Log of trade flows of services in current US dollars between Country i and Country j - LAC — Dummy variable taking the value of one if exporting Country i is in the Latin American and Caribbean region, zero otherwise - SA — Dummy variable taking the value of one if exporting Country i is in South America, zero otherwise - CA — Dummy variable taking the value of one if exporting Country i is in Central America, zero otherwise - MEX — Dummy variable taking the value of one if either the exporting Country I or the importing Country j is Mexico - CAR — Dummy variable taking the value of one if exporting Country i is in the Caribbean sub-region, zero otherwise ### Gravity variables (size, distance, commonalities) - LGDPi / LGDPj — Log of GDP, in current US dollars, of Country i / Country j - LPOPi / LPOPj — Log of population of Country i / Country j - LDISTANCE — Log of distance (km) between the capital cities of Countries i and j - LANG — Dummy variable taking the value of one if Countries i and j share a common language, zero otherwise - BORDER — Dummy variable taking the value of one if Countries i and j share a common border, zero otherwise - LANDLLi / LANDLLj — Dummy variable taking the value of one if Country i / Country j is landlocked, zero otherwise ### Trade policy variables (TP) - TRIj — Non-tariff trade barrier index for importing Country j from the Measure of Aggregate Trade Restrictiveness (MATR) database - MFNTj — Most favored nation (MFN) tariff imposed by importing Country j (weighted by trade values) - RTA — Dummy variable taking the value of one if both Countries i and j have a regional trade agreement in place, zero otherwise - NAFTA — Dummy variable taking the value of one if both Countries i and j are members of the NAFTA trade agreement, zero otherwise ### Infrastructure, customs, digital access, and firm constraints - LPITRANSi / LPITRANSj — Quality of trade and transport infrastructure (World Bank LPI) for exporting Country i / importing Country j - LPICUSTi / LPICUSTj — Efficiency of customs and border clearance (World Bank LPI) for exporting Country i / importing Country j - BESTRANSi / BESTRANSj — Percent of firms identifying transport as a major or very severe constraint - BESCUSTi / BESCUSTj — Percent of firms identifying customs regulations and clearance as a major or very severe constraint - MBUSERSi / MBUSERSj — Number of mobile broadband users per 100 inhabitants - BESFINi / BESFINj — Percent of firms identifying access to finance as a major or very severe constraint - BESELECi / BESELECj — Percent of firms identifying electricity as a major or very severe constraint ### Quality, human capital, and governance - BESSKLABi / BESSKLABj — Percent of firms identifying an inadequately educated workforce as a major or very severe constraint - HCIi — Human Capital Index for exporting Country i, from the Penn World Tables - BESCORRi / BESCORRj — Percent of firms identifying corruption as a major or very severe constraint - BESCRIi / BESCRIj — Percent of firms identifying crime, theft and disorder as a major or very severe constraint - BESPOLINSTi / BESPOLINSTj — Percent of firms identifying political instability as a major or very severe constraint - BESGOVAvgi / BESGOVAvgj — Simple average of the three WBES governance indicators for Country i / Country j - WGICORRi / WGICORRj — WGI control of corruption indicator - WGIPOLSTi / WGIPOLSTj — WGI political stability indicator - WGIROLi / WGIROLj — WGI rule of law indicator - WGIGOVAvgi / WGIGOVAvgj — Simple average of the three WGI governance indicators for Country i / Country j --- ### Baseline gravity results — summary and key coefficients (Tables 1–3) - Baseline finding: Significant under-trading only in the case of services for Latin America and the Caribbean (LAC) as a whole (Figure 8). Note: ** statistically significant at the 5 percent level. Otherwise, not statistically significant. - Selected baseline coefficients (coef; tstat) — Merchandise and Services (Table 1): - LGDPi: 0.778***; 18.86 | 0.770***; 19.72 | 1.048***; 29.33 | 1.046***; 29.50 - LGDPj: 0.813***; 14.36 | 0.803***; 15.33 | 1.065***; 36.81 | 1.063***; 36.65 - LPOPi: -0.00525; -0.103 | -0.00619; -0.125 | -0.370***; -8.670 | -0.369***; -8.661 - LPOPj: -0.126***; -2.595 | -0.125***; -2.705 | -0.369***; -10.65 | -0.368***; -10.67 - LDISTANCE: -0.427***; -7.066 | -0.426***; -7.193 | -0.597***; -19.20 | -0.596***; -19.26 - LANG: 0.290***; 2.739 | 0.327***; 3.308 | 0.835***; 10.09 | 0.841***; 10.21 - BORDER: 1.159***; 7.006 | 1.098***; 7.462 | -0.0285; -0.232 | -0.0330; -0.273 - LANDLLi: -0.134; -1.238 | -0.144; -1.321 | -0.284**; -2.445 | -0.286**; -2.462 - LANDLLj: -0.283***; -2.593 | -0.287***; -2.647 | -0.167; -1.301 | -0.168; -1.307 - Observations: 21,010 | 21,010 | 13,687 | 13,687 - R-squared: 0.456 | 0.495 | 0.751 | 0.752 - Gauging intra-regional trade (Table 3): - intra-LAC dummy: -0.426**; -1.991 (merchandise) | -0.521***; -3.149 (services) - Observations: 21,010 | 13,687 - R-squared: 0.459 | 0.751 --- ### Benchmarking constraints (Figure 10; period averages 2015-2019) - Data notes: MFN tariffs are rates, MATR is a count of restrictions, LPI is an index, firm constraints are shares of firms reporting constraints. - LAC median vs EMDE (excluding LAC) 25th and 75th percentiles: - In most measures, LAC countries stand between best and worst EMDE performers, indicating scope for improvement. - MATR: median LAC country stands close to the best performers among non-LAC EMDEs (relatively low non-tariff restrictions). - Electricity as a constraint and governance as a constraint: median LAC country stands closer to the worst performers among EMDEs. --- ### Augmented gravity regressions — how constraint sets affect sub-regions (Figures 11–14; Tables 4–13) - Method: Baseline gravity augmented with four sets of variables — trade policy; transport infrastructure and customs; access to and quality of factor inputs; governance indicators. - Trade policy (Tables 4–6) — selected coefficients: - RTA: 0.662***; 6.793 | 0.0474; 0.560 (Table 4, merchandise and services) - TRIj: -0.292*; -1.671 | -0.0413; -0.385 - MFNTj: -0.0393**; -1.972 | -0.0320*; -1.843 - NAFTA (Table 6): 0.0258; 0.120 | 1.142***; 4.436 (reported in specific columns) - R-squared increases when trade-policy variables included (e.g., Table 4 R-squared: 0.495 | 0.555) - Transport and customs (Tables 7–9) — selected coefficients: - LPITRANSi: 3.602***; 3.666 | -2.225***; -3.466 (Table 7, different columns) - LPITRANSj: 2.292**; 2.533 | -1.398*; -1.690 - LPICUSTi: -1.501**; -2.237 | 2.748***; 5.451 - LPICUSTj: -0.571; -0.811 | 1.312**; 2.010 - MBUSERSi: 0.00287*; 1.648 (Table 8, some columns) - Factor inputs and digital infrastructure (Tables 9–11) — selected coefficients: - BESFINi: 0.00263; 0.511 | -0.00413; -1.221 - BESFINj: 0.00413; 1.124 | 0.00866**; 2.551 - HCIi: 0.483***; 3.730 | -0.401***; -4.272 - BESELECi: -0.00807***; -3.212 | -0.00154; -0.680 - BESELECj: -0.00248; -1.042 | -0.00772***; -2.779 - TABLE 10 selected: BESFINi: -0.0180***; -3.711 | 0.0397***; 5.806 ; HCIi: 0.435***; 4.020 | 0.702**; 2.053 - Governance indicators (Tables 11–13) — selected coefficients: - BESGOVAvgi: -0.00873***; -2.988 (Table 11) - BESGOVAvg i: -0.0157***; -6.370 | 0.00851*; 1.907 (Table 12) - BESTRANSi: -0.0161***; -3.167 (Table 13) - BESCUSTj: 0.0260***; 2.642 (Table 13) #### Sub-region specific augmentation outcomes (summary of Figures 11–14) - South America: - Trade policy variables: negligible impact on regional dummy coefficients. - LPI transport and customs: regional dummy coefficient halves for manufactured goods but significant under-trading remains; under-trading in services fully explained by LPI. - Digital infrastructure (mobile broadband users): marginal impact. - Factor inputs (WBES electricity and access to finance, plus HCI): notable reduction in regional dummy for manufacturing (not services). - WBES governance average: inclusion makes coefficients on regional dummies statistically insignificant across all product groups. - Central America: - Trade policy variables and WBES governance: little impact; regional dummies continue to show significant under-trading for manufactured goods and primary commodities. - LPI index, particularly transport infrastructure: notably reduces regional dummy for manufacturing but under-trading remains significant; marginal impact on primary commodities. - Human Capital Index and WBES electricity/access to finance: little impact on regional dummy values despite many factor variables being statistically significant and of expected sign. - Mexico: - Trade policy variables: fully explain over-trading in merchandise trade (manufactured goods). - Addition of NAFTA dummy: implies significant under-trading in primary commodities. - WBES governance average: reduces extent of under-trading in primary commodities somewhat. - Caribbean: - Trade policy variables and factor input variables: do not explain under-trading in merchandise trade or services. - WBES governance variables: do not explain under-trading. - LPI transport and customs variables: regional dummy for merchandise trade becomes statistically insignificant (helps explain under-trading in manufactured goods); no notable impact on services. --- ### Robustness checks (Tables 14–16) — replacements and findings - Replacements used in robustness checks: - Replace World Bank LPI index variables with WBES transport and customs survey results (Table 14). - Replace Human Capital Index with WBES inadequately educated workforce survey result (Table 15). - Replace WBES governance indicators with WGI governance indicators (Table 16). - Robustness outcomes (selected): - Results generally less satisfactory with alternative variables. - South America: - WBES transport and customs variables fully explain under-trading in manufacturing but negligible impact for services. - WBES inadequately educated workforce variable shows statistically significant sign opposite to expectations for South America. - WGI governance indicators help explain some under-trading in services, but not in merchandise trade. - Central America and Caribbean: - WBES transport and customs variables have marginal impacts. - WBES inadequately educated workforce variable shows no notable impact on regional dummies. - WGI governance indicators do not explain under-trading. --- ### Key empirical conclusions (Section VII) - Baseline regressions (no policy/constraint variables): significant evidence of under-trading in LAC only for services. - Disaggregation by sub-region reveals broader under-trading: - South America: significant under-trading in merchandise goods (particularly manufacturing) and services. - Caribbean: significant under-trading across all product groups (merchandise trade and services). - Central America: significant under-trading in merchandise goods (manufactured goods and primary commodities), but not services. - Mexico: significant over-trading in manufactured goods (baseline). - LAC’s under-performance in trade primarily emanates from intra-regional trade rather than inter-regional trade. - Constraint variables explain parts of under-trading: - Poor transport infrastructure and inefficiencies in customs clearance (LPI transport and customs variables) explain a significant part of under-trading in manufactured goods for South America and the Caribbean. - WBES governance indicators significant in explaining under-trading in manufactured goods and services for South America, but not for Central America or the Caribbean. - Access to factor inputs cannot fully explain under-trading for any sub-region, though HCI reduces under-trading in manufactured goods for South America. - For Mexico, trade policy variables fully explain over-trading in manufactured goods; NAFTA dummy introduction implies significant under-trading in primary commodities (possible resource re-allocation effect). - Research gaps identified: - Further research needed to understand (i) factors underpinning Central America’s and the Caribbean’s underperformance in trade, and (ii) the benefits of removing constraints behind LAC’s trade underperformance. --- ### Annex I — Selected model fit and observation counts (Tables 1–16) - Table 1 observations and R-squared: Observations: 21,010 | 21,010 | 13,687 | 13,687; R-squared: 0.456 | 0.495 | 0.751 | 0.752 - Table 2 observations and R-squared: Observations: 21,010 (all columns); R-squared: 0.418 | 0.459 | 0.265 | 0.271 - Table 7 observations and R-squared (transport/customs): Observations: 15,453 | 15,453 | 11,389 | 11,389; R-squared: 0.557 | 0.618 | 0.751 | 0.754 - Table 9 observations and R-squared (digital/infrastructure): Observations: 10,350 | 10,350 | 8,501 | 8,501; R-squared: 0.509 | 0.520 | 0.714 | 0.716 - Table 14 observations and R-squared (robustness): Observations: 12,870 | 12,870 | 9,900 | 9,900; R-squared: 0.507 | 0.521 | 0.713 | 0.711 - Table 15 observations and R-squared (WGIGOV robustness): Observations: 20,460 | 20,460 | 13,687 | 13,687; R-squared: 0.556 | 0.557 | 0.752 | 0.749 *Source: wpiea2024032-print-pdf - Box 1: Gravity Model: Explanation of Variables* ### Box 1: Gravity Model: Explanation of Variables ### Box 1: Gravity Model: Explanation of Variables ### Trade variables - LTrTOTij — Log of total trade flows, merchandise and services, in current US dollars between Country i and Country j - LTrMRCij — Log of total merchandise trade flows in current US dollars between Country i and Country j - LTrMANUij — Log of trade flows of manufacturing goods in current US dollars between Country i and Country j - LTrPRIPDij — Log of trade flows of primary commodities in current US dollars between Country i and Country j - LTrSERVij — Log of trade flows of services in current US dollars between Country i and Country j - LAC — Dummy variable taking the value of one if exporting Country i is in the Latin American and Caribbean region, zero otherwise - SA — Dummy variable taking the value of one if exporting Country i is in South America, zero otherwise - CA — Dummy variable taking the value of one if exporting Country i is in Central America, zero otherwise - MEX — Dummy variable taking the value of one if either the exporting Country I or the importing Country j is Mexico - CAR — Dummy variable taking the value of one if exporting Country i is in the Caribbean sub-region, zero otherwise ### Gravity Variables - LGDPi / LGDPj — Log of GDP, in current US dollars, of Country i / Country j - LPOPi / LPOPj — Log of population of Country i / Country j - LDISTANCE — Log of distance (km) between the capital cities of Countries i and j - LANG — Dummy variable taking the value of one if Countries i and j share a common language, zero otherwise - BORDER — Dummy variable taking the value of one if Countries i and j share a common border, zero otherwise - LANDLLi / LANDLLj — Dummy variable taking the value of one if Country i / Country j is landlocked, zero otherwise ### Trade Policy Variables (TP) - TRIj — Non-tariff trade barrier index for importing Country j from the Measure of Aggregate Trade Restrictiveness (MATR) database computed by Estefania-Flores and others (2022). - MFNTj — Most favored nation (MFN) tariff imposed by importing Country j on imports from Countries not having a preferential trade / tariff agreement with Country j (weighted by trade values across import lines) - RTA — Dummy variable taking the value of one if both Countries i and j have a regional trade agreement in place, zero otherwise - NAFTA — Dummy variable taking the value of one if both Countries i and j are members of the NAFTA trade agreement, zero otherwise ### Infrastructure and Customs - LPITRANSi / LPITRANSj — Value of the quality of trade and transport infrastructure component of the World Bank’s International Logistics Performance Index (LPI) for exporting Country i / importing Country j. Higher values denote better infrastructure. - LPICUSTi / LPICUSTj — Value of the efficiency of customs and border clearance component of the World Bank’s International Logistics Performance Index (LPI) for exporting Country i / importing Country j. Higher values denote higher efficiency. - BESTRANSi / BESTRANSj — Percent of firms in Country i / Country j that trade that identify transport as a major or very severe constraint on the operations of their business - BESCUSTi / BESCUSTj — Percent of firms in Country i / Country j that trade that identify customs regulations and clearance as a major or very severe constraint on the operations of their business - MBUSERSi / MBUSERSj — Number of mobile broadband users per 100 inhabitants in Country I Country j ### Quality and Access of Factors of Production - BESFINi / BESFINj — Percent of firms in Country i / Country j that trade that identify access to finance as a major or very severe constraint on the operations of their business - BESELECi / BESELECj — Percent of firms in Country i / Country j that trade that identify electricity as a major or very severe constraint on the operations of their business *Source: wpiea2024032-print-pdf - Box 1: Gravity Model: Explanation of Variables* ### Box 1: Gravity Model: Explanation of Variables (cont.) ### Box 1: Gravity Model: Explanation of Variables (cont.) ### Variables (selected) - BESSKLABi / BESSKLABj: Percent of firms in Country i / Country j that trade that identify an inadequately educated workforce as a major or very severe constraint on the operations of their business - HCIi: Human Capital Index for exporting Country i, from the Penn World Tables - Governance and Institutions: - BESCORRi / BESCORRj: Percent of firms in Country i / Country j that trade that identify corruption as a major or very severe constraint on the operation of their business - BESCRIi / BESCRIj: Percent of firms in Country i / Country j that trade that identify crime, theft and disorder as a major or very severe constraint on the operation of their business - BESPOLINSTi / BESPOLINSTj: Percent of firms in Country i / Country j that trade that identify political instability as a major or very severe constraint on the operation of their business - BESGOVAvgi / BESGOVAvgj: Simple average of the above three governance indicators for Country i / Country j from the World Bank’s Business Enterprise Surveys - WGICORRi / WGICORRj: World Governance Indicator (WGI) aggregate indicator for Country i / Country j on control of corruption - WGIPOLSTi / WGIPOLSTj: WGI aggregate indicator for Country i / Country j on political stability and absence of violence / terrorism - WGIROLi / WGIROLj: WGI aggregate indicator for Country i / Country j on rule of law - WGIGOVAvgi / WGIGOVAvgj: Simple average of the above three WGI governance indicators for Country i / Country j ### Baseline gravity results (regional and sub-regions) - Baseline regressions (no policy variables or proxies for constraints) show: - Significant under-trading only in the case of services for Latin America and the Caribbean (LAC) as a whole (Figure 8). Note: ** statistically significant at the 5 percent level. Otherwise, not statistically significant. - Disaggregated findings: - Intra-regional trade in LAC is lower compared to what would be predicted by economic and geographic factors (Table 3). - Merchandise trade: coefficient for extra-regional trade negative but not statistically significant; intra-regional trade negative and statistically significant. - Services: substantial under-trading both with extra- and intra-regional partners, with more pronounced under-trading intra-regionally. - Sub-region differences (Figure 9): - South America: significant under-trading in manufactured goods and services. - Caribbean: significant under-trading across all product groups (merchandise trade, manufacturing, primary commodities, and services). - Central America: trade flows below gravity benchmark in merchandise goods (both manufactured goods and primary commodities), but not in services. - Mexico: baseline regressions suggest significant over-trading in manufactured goods. ### Benchmarking constraints (summary of Figure 10) - Values are averages for the period 2015-2019. MFN tariffs are rates, MATR is a count of restrictions, the LPI is an index, and constraints are share of firms identifying each factor as a constraint. - LAC median compared to EMDE (excluding LAC) 25th and 75th percentiles: - In most cases LAC countries stand between best and worst performers in EMDEs, indicating scope for improvement. - MATR: median LAC country stands close to the best performers among non-LAC EMDEs (relatively low non-tariff restrictions). - Electricity as a constraint and governance as a constraint: median LAC country stands closer to the worst performers among EMDEs (indicating disproportionate negative effects). ### Augmented gravity regressions — how constraint sets affect sub-regions - Method: Augment baseline gravity with four sets of variables — trade policy; transport infrastructure and customs regulations; access to and quality of factor inputs; governance indicators. Results summarized in Figures 11–14 and Tables 4–13. - South America (Figure 11): - Trade policy variables: negligible impact on regional dummy coefficients. - Transport infrastructure and customs (LPI): regional dummy coefficient halves for manufactured goods but significant under-trading remains; under-trading in services fully explained by LPI. - Addition of digital infrastructure (mobile broadband users per 100 inhabitants): marginal impact. - Factor inputs (WBES electricity and access to finance, plus HCI): notable reduction in regional dummy for manufacturing (not services). - WBES governance average (political instability, corruption, theft/crime): inclusion makes coefficients on regional dummies statistically insignificant across all product groups. - Central America (Figure 12): - Trade policy variables and WBES governance: little impact; regional dummies continue to show significant under-trading for manufactured goods and primary commodities. - LPI index, particularly transport infrastructure: notably reduces regional dummy for manufacturing but under-trading remains significant; marginal impact on primary commodities. - Human Capital Index and WBES electricity/access to finance: little impact on regional dummy values despite many factor variables being statistically significant and of expected sign. - Mexico (Figure 13): - Trade policy variables: fully explain over-trading in merchandise trade (manufactured goods). - Addition of NAFTA dummy: implies significant under-trading in primary commodities (possibly reflecting NAFTA favoring manufactured goods over primary commodities). - WBES governance average: reduces extent of under-trading in primary commodities somewhat. - Caribbean (Figure 14): - Trade policy variables and factor input variables: do not explain under-trading in merchandise trade or services. - WBES governance variables: do not explain under-trading. - LPI transport and customs variables: regional dummy for merchandise trade becomes statistically insignificant (i.e., helps explain under-trading in manufactured goods), but no notable impact on services. ### Robustness checks - Replacements in robustness checks: - Replace World Bank LPI index variables with WBES transport and customs survey results (Table 14). - Replace Human Capital Index with WBES inadequately educated workforce survey result (Table 15). - Replace WBES governance indicators with WGI governance indicators (Table 16). - Robustness outcomes: - Results generally less satisfactory with alternative variables. - South America: - WBES transport and customs variables fully explain under-trading in manufacturing but negligible impact for services. - WBES inadequately educated workforce variable shows statistically significant sign opposite to expectations for South America. - WGI governance indicators help explain some under-trading in services, but not in merchandise trade. - Central America and Caribbean: - WBES transport and customs variables have marginal impacts. - WBES inadequately educated workforce variable shows no notable impact on regional dummies. - WGI governance indicators do not explain under-trading. ### Key empirical conclusions (Section VII) - Baseline regressions (no policy/constraint variables): significant evidence of under-trading in LAC only for services. - Disaggregation by sub-region reveals broader under-trading: - South America: significant under-trading in merchandise goods (particularly manufacturing) and services. - Caribbean: significant under-trading across all product groups (merchandise trade and services). - Central America: significant under-trading in merchandise goods (manufactured goods and primary commodities), but not services. - Mexico: significant over-trading in manufactured goods (baseline). - LAC’s under-performance in trade primarily emanates from intra-regional trade rather than inter-regional trade. - Constraint variables explain parts of under-trading: - Poor transport infrastructure and inefficiencies in customs clearance (LPI transport and customs variables) explain a significant part of under-trading in manufactured goods for South America and the Caribbean. - WBES governance indicators significant in explaining under-trading in manufactured goods and services for South America, but not for Central America or the Caribbean. - Access to factor inputs cannot fully explain under-trading for any sub-region, though HCI reduces under-trading in manufactured goods for South America. - For Mexico, trade policy variables fully explain over-trading in manufactured goods; NAFTA dummy introduction implies significant under-trading in primary commodities (possible resource re-allocation effect). - Research gaps identified: - Further research needed to understand (i) factors underpinning Central America’s and the Caribbean’s underperformance in trade, and (ii) the benefits of removing constraints behind LAC’s trade underperformance. *Source: wpiea2024032-print-pdf - Box 1: Gravity Model: Explanation of Variables (cont.)* ### Annex I. Tables of gravity model results ### Annex I. Tables of gravity model results ### Table 1. Baseline gravity model regressions - Merchandise and Services Trade - Dependent variables: LTrMRCij, LTrSERVij - Key coefficients (coef; tstat): - LGDPi: 0.778***; 18.86 | 0.770***; 19.72 | 1.048***; 29.33 | 1.046***; 29.50 - LGDPj: 0.813***; 14.36 | 0.803***; 15.33 | 1.065***; 36.81 | 1.063***; 36.65 - LPOPi: -0.00525; -0.103 | -0.00619; -0.125 | -0.370***; -8.670 | -0.369***; -8.661 - LPOPj: -0.126***; -2.595 | -0.125***; -2.705 | -0.369***; -10.65 | -0.368***; -10.67 - LDISTANCE: -0.427***; -7.066 | -0.426***; -7.193 | -0.597***; -19.20 | -0.596***; -19.26 - LANG: 0.290***; 2.739 | 0.327***; 3.308 | 0.835***; 10.09 | 0.841***; 10.21 - BORDER: 1.159***; 7.006 | 1.098***; 7.462 | -0.0285; -0.232 | -0.0330; -0.273 - LANDLLi: -0.134; -1.238 | -0.144; -1.321 | -0.284**; -2.445 | -0.286**; -2.462 - LANDLLj: -0.283***; -2.593 | -0.287***; -2.647 | -0.167; -1.301 | -0.168; -1.307 - Observations: 21,010 | 21,010 | 13,687 | 13,687 - R-squared: 0.456 | 0.495 | 0.751 | 0.752 - Note: t-statistics calculated using robust standard errors. *** p<0.01, ** p<0.05, * p<0.1. ### Table 2. Baseline gravity model regressions - Manufacturing and Primary Commodities Trade - Dependent variables: LTrMANUij, LTrPRIPDij - Key coefficients (coef; tstat): - LGDPi: 0.814***; 17.35 | 0.806***; 18.05 | 0.610***; 13.34 | 0.600***; 13.31 - LGDPj: 0.855***; 12.60 | 0.843***; 13.55 | 0.624***; 13.71 | 0.623***; 13.57 - LPOPi: 0.0159; 0.279 | 0.0150; 0.269 | -0.0979*; -1.828 | -0.0991*; -1.869 - LPOPj: -0.181***; -3.246 | -0.180***; -3.405 | 0.133**; 2.276 | 0.132**; 2.264 - LDISTANCE: -0.444***; -6.823 | -0.442***; -6.939 | -0.359***; -5.895 | -0.364***; -6.070 - LANG: 0.291**; 2.424 | 0.334***; 3.008 | 0.213; 1.493 | 0.234*; 1.650 - BORDER: 1.154***; 6.442 | 1.079***; 6.829 | 1.190***; 6.360 | 1.176***; 6.442 - LANDLLi: 0.0312; 0.255 | 0.0256; 0.210 | -0.911***; -5.637 | -0.946***; -5.882 - LANDLLj: -0.235**; -1.987 | -0.238**; -2.021 | -0.751***; -5.316 | -0.760***; -5.374 - Observations: 21,010 (all columns) - R-squared: 0.418 | 0.459 | 0.265 | 0.271 - Note: t-statistics calculated using robust standard errors. *** p<0.01, ** p<0.05, * p<0.1. ### Table 3. Gauging Intra-regional Trade: Baseline regressions - Merchandise and Services - Dependent variables: LTrMRCij, LTrSERVij - Selected coefficients (coef; tstat): - LGDPi: 0.775***; 18.71 | 1.045***; 29.33 - LGDPj: 0.810***; 14.34 | 1.061***; 36.23 - LPOPi: -0.00271; -0.0535 | -0.367***; -8.624 - LPOPj: -0.123**; -2.539 | -0.366***; -10.54 - LDISTANCE: -0.431***; -7.096 | -0.600***; -19.07 - LANG: 0.301***; 2.842 | 0.847***; 10.10 - BORDER: 1.158***; 7.030 | -0.0321; -0.262 - intra-LAC: -0.426**; -1.991 | -0.521***; -3.149 - LANDLLi and j: see coefficients for significance in each column - Observations: 21,010 | 13,687 - R-squared: 0.459 | 0.751 - Note: t-statistics calculated using robust standard errors. *** p<0.01, ** p<0.05, * p<0.1. ### Table 4. Gravity model regressions with Trade Policy Variables - Merchandise and Services Trade - Dependent variables: LTrMRCij, LTrSERVij - Additional trade-policy variables included: RTA, TRIj, MFNTj - Selected coefficients (coef; tstat): - RTA: 0.662***; 6.793 | 0.0474; 0.560 - TRIj: -0.292*; -1.671 | -0.0413; -0.385 - MFNTj: -0.0393**; -1.972 | -0.0320*; -1.843 - LGDPi, LGDPj, LDISTANCE, LANG, BORDER: coefficients and t-stats reported with significance levels (see table) - Observations: 21,010 (both columns) - R-squared: 0.495 | 0.555 - Note: t-statistics calculated using robust standard errors. *** p<0.01, ** p<0.05, * p<0.1. ### Table 5. Gravity model regressions with Trade Policy Variables - Manufacturing and Primary Commodities Trade - Dependent variables: LTrMANUij, LTrPRIPDij - Selected coefficients (coef; tstat): - RTA: 0.756***; 7.125 | 0.286; 1.627 - TRIj: -0.412**; -2.043 | 0.186; 0.913 - MFNTj: -0.0410*; -1.862 | -0.0348; -1.261 - LGDPi, LGDPj, LDISTANCE, LANG, BORDER: coefficients provided with exact numeric values and significance - Observations: 21,010 (all columns) - R-squared: 0.459 | 0.523 | 0.271 | 0.281 - Note: t-statistics calculated using robust standard errors. *** p<0.01, ** p<0.05, * p<0.1. ### Table 6. Gravity model regressions with Trade Policy Variables and NAFTA - Manufacturing and Primary Commodities Trade - Dependent variables: LTrMANUij, LTrPRIPDij (four columns with NAFTA) - Selected coefficients (coef; tstat): - NAFTA: 0.0258; 0.120 | 1.142***; 4.436 (reported in specific columns) - RTA: 0.756***; 7.125 | 0.755***; 7.114 - TRIj and MFNTj: coefficients and t-stats provided - LGDPi and LGDPj: coefficients range (e.g., LGDPi 0.800***; 18.96 ... 0.575***; 13.47) - Observations: 21,010 (all columns) - R-squared: 0.523 | 0.524 | 0.281 | 0.377 - Note: t-statistics calculated using robust standard errors. *** p<0.01, ** p<0.05, * p<0.1. ### Table 7. Gravity model regressions with Transport and Customs Variables - Merchandise and Services Trade - Dependent variables: LTrMRCij, LTrSERVij - Transport and customs variables included: LPITRANSi, LPITRANSj, LPICUSTi, LPICUSTj - Selected coefficients (coef; tstat): - LPITRANSi: 3.602***; 3.666 | -2.225***; -3.466 (in different columns) - LPITRANSj: 2.292**; 2.533 | -1.398*; -1.690 - LPICUSTi: -1.501**; -2.237 | 2.748***; 5.451 - LPICUSTj: -0.571; -0.811 | 1.312**; 2.010 - LGDPi, LGDPj, LDISTANCE, LANG, BORDER: coefficients provided with exact values - Observations: 15,453 | 15,453 | 11,389 | 11,389 - R-squared: 0.557 | 0.618 | 0.751 | 0.754 - Note: t-statistics calculated using robust standard errors. *** p<0.01, ** p<0.05, * p<0.1. ### Table 8. Gravity model regressions with Transport and Customs Variables - Manufacturing and Primary Commodities Trade - Dependent variables: LTrMANUij, LTrPRIPDij - Selected coefficients (coef; tstat): - LPITRANSi: 4.453***; 3.809 | 0.358; 0.383 - LPITRANSj: 2.048**; 2.178 | 3.330**; 2.425 - LPICUSTi: -0.0688; -0.0856 | -5.203***; -6.913 - LPICUSTj: -0.763; -1.018 | 0.703; 0.641 - MBUSERSi and MBUSERSj: coefficients included in some columns (e.g., MBUSERSi 0.00287*; 1.648) - Observations: 15,453 (all columns) - R-squared: 0.525 | 0.625 | 0.292 | 0.347 - Note: t-statistics calculated using robust standard errors. *** p<0.01, ** p<0.05, * p<0.1. ### Table 9. Gravity model regressions with Transport and Customs Variables and Digital Infrastructure - Merchandise and Services Trade - Dependent variables: LTrMRCij, LTrSERVij - Digital and infrastructure variables included: BESFINi, BESFINj, HCIi, BESELECi, BESELECj - Selected coefficients (coef; tstat): - BESFINi: 0.00263; 0.511 | -0.00413; -1.221 - BESFINj: 0.00413; 1.124 | 0.00866**; 2.551 - HCIi: 0.483***; 3.730 | -0.401***; -4.272 - BESELECi: -0.00807***; -3.212 | -0.00154; -0.680 - BESELECj: -0.00248; -1.042 | -0.00772***; -2.779 - LGDPi, LGDPj, LDISTANCE, LANG, BORDER: coefficients and significance reported - Observations: 10,350 | 10,350 | 8,501 | 8,501 - R-squared: 0.509 | 0.520 | 0.714 | 0.716 - Note: t-statistics calculated using robust standard errors. *** p<0.01, ** p<0.05, * p<0.1. ### Table 10. Gravity model regressions with Factor Input Variables - Merchandise and Services Trade - Dependent variables: LTrMRCij, LTrSERVij - Factor-input and infrastructure variables included: BESFIN, HCI, BESELEC (various columns) - Selected coefficients (coef; tstat): - BESFINi: -0.0180***; -3.711 | 0.0397***; 5.806 (in different columns) - HCIi: 0.435***; 4.020 | 0.702**; 2.053 - BESELECi: -0.00578**; -2.300 | -0.00984**; -2.184 - BESELECj: -0.000336; -0.134 | -0.0129***; -3.053 - LGDPi, LGDPj, LDISTANCE, LANG, BORDER: coefficients and t-stats reported with exact values - Observations: 10,350 (all columns) - R-squared: 0.553 | 0.564 | 0.163 | 0.289 - Note: t-statistics calculated using robust standard errors. *** p<0.01, ** p<0.05, * p<0.1. ### Table 11. Gravity model regressions with Factor Input Variables - Manufacturing and Primary Commodities Trade - Dependent variables: LTrMANUij, LTrPRIPDij - Selected coefficients (coef; tstat): - BESGOVAvg i and j: coefficients reported (e.g., BESGOVAvg i -0.00873***; -2.988) - LGDPi and LGDPj: coefficients in range (e.g., LGDPi 0.859***; 24.20 ... 0.968***; 31.47) - LDISTANCE: -0.313***; -3.204 | -0.337***; -3.212 | -0.673***; -12.94 | -0.685***; -12.42 - CA, CAR, SA, MEX: region dummies coefficients reported - Observations: 12,638 | 12,638 | 9,704 | 9,704 - R-squared: 0.508 | 0.499 | 0.713 | 0.708 - Note: t-statistics calculated using robust standard errors. *** p<0.01, ** p<0.05, * p<0.1. ### Table 12. Gravity model regressions with Governance Indicators - Merchandise and Services Trade - Dependent variables: LTrMRCij, LTrSERVij - Governance variables included (BESGOVAvg i/j) - Selected coefficients (coef; tstat): - BESGOVAvg i: -0.0157***; -6.370 | 0.00851*; 1.907 (different columns) - BESGOVAvg j: -0.000103; -0.0406 | -0.0144***; -2.967 - LGDPi, LGDPj, LDISTANCE, LANG, BORDER: coefficients and t-stats provided - Observations: 12,638 (all columns) - R-squared: 0.552 | 0.558 | 0.156 | 0.191 - Note: t-statistics calculated using robust standard errors. *** p<0.01, ** p<0.05, * p<0.1. ### Table 13. Gravity model regressions with Governance Indicators - Manufacturing and Primary Commodities Trade - Dependent variables: LTrMANUij, LTrPRIPDij - Selected coefficients (coef; tstat): - BESTRANSi and j, BESCUSTi and j included in robustness specifications - BESTRANSi: -0.0161***; -3.167 | -0.00484; -1.047 - BESCUSTj: -0.00519; -0.685 | 0.0260***; 2.642 - LGDPi, LGDPj, LDISTANCE, LANG, BORDER: coefficients and t-stats recorded in table - Observations: 12,870 | 12,870 | 9,900 | 9,900 - R-squared: 0.507 | 0.504 | 0.713 | 0.709 - Note: t-statistics calculated using robust standard errors. *** p<0.01, ** p<0.05, * p<0.1. ### Table 14. Robustness regressions with Transport and Customs Variables - Merchandise and Services Trade - Dependent variables: LTrMRCij, LTrSERVij - Robustness variables included: BESFIN, BESSKLAB, BESELEC, etc. - Selected coefficients (coef; tstat): - BESSKLABi: 0.0118***; 3.145 | -0.00383; -1.211 - BESFINj: 0.00470; 1.293 | 0.00908***; 2.724 - BESELECi: -0.0150***; -5.326 | 0.00111; 0.418 (in different columns) - BESELECj: -0.00290; -1.263 | -0.00774***; -2.798 - LGDPi, LGDPj, LDISTANCE, LANG, BORDER: coefficients and t-stats provided - Observations: 12,870 | 12,870 | 9,900 | 9,900 - R-squared: 0.507 | 0.521 | 0.713 | 0.711 - Note: t-statistics calculated using robust standard errors. *** p<0.01, ** p<0.05, * p<0.1. ### Table 15. Robustness regressions with Factor Input Variables - Merchandise and Services Trade - Dependent variables: LTrMRCij, LTrSERVij - Robustness variables included: WGIGOVAvg, etc. - Selected coefficients (coef; tstat): - WGIGOVAvgi: 0.0246; 0.454 | 0.110**; 2.399 (in different columns) - WGIGOVAvgj: 0.102*; 1.849 | 0.0877*; 1.787 - LGDPi, LGDPj, LDISTANCE, LANG, BORDER, LANDLL: coefficients and t-stats provided - Observations: 20,460 | 20,460 | 13,687 | 13,687 - R-squared: 0.556 | 0.557 | 0.752 | 0.749 - Note: t-statistics calculated using robust standard errors. *** p<0.01, ** p<0.05, * p<0.1. ### Table 16. Robustness regressions with Governance Indicators - Merchandise and Services Trade - Dependent variables: LTrMRCij, LTrSERVij - Robustness specifications revisiting governance indicators with alternative controls - Selected coefficients and model fit: - LGDPi: 0.764***; 20.99 | 0.742***; 11.08 | 1.047***; 29.66 | 0.938***; 15.31 - LGDPj: 0.637***; 11.26 | 0.544***; 7.05 | 1.001***; 18.33 | 0.916***; 13.58 - LDISTANCE: -0.348***; -5.460 | -0.350***; -5.556 | -0.582***; -13.86 | -0.586***; -13.82 - LANG, BORDER, RTA, TRIj, MFNTj, regional dummies (SA, CA, MEX, CAR), LANDLL, WGIGOVAvg: coefficients and exact t-statistics reported - Observations: 20,460 | 20,460 | 13,687 | 13,687 - R-squared: 0.556 | 0.557 | 0.752 | 0.749 - Note: t-statistics calculated using robust standard errors. *** p<0.01, ** p<0.05, * p<0.1. *Source: Authors' calculations.* ### References ### References (wpiea2024032-print-pdf) ### Core empirical and methodological sources on gravity and trade estimation - Baldwin, R. and D. Taglioni (2006), “Gravity for Dummies and Dummies for Gravity Equations,” NBER Working Paper 12516. - Santos Silva, J.M.C, and S. Tenreyro (2006), “The log of Gravity,” The Review of Economics and Statistics, Vol. 88 (4): 641-658. - Santos Silva, J.M.C, and S. Tenreyro (2011), “Further simulation evidence on the performance of the Poisson pseudo-maximum likelihood estimator,” Economics Letters, Vol. 112 (2): 220-222. - Felbermayr, G., and Y. Yotov (2021), “From Theory to Policy with Gravitas: A Solution to the Mystery of the Excess Trade Balances”, European Economic Review, Vol. 139:103875. - Conte, M. P. Cotterlaz and T. Mayer, 2022. "The CEPII Gravity Database," Working Papers 2022-05, CEPII research center ### Trade costs, logistics, and transport in Latin America and the Caribbean (LAC) - Behar, A., P. Manners, and B.D. Nelson (2013), “Exports and International Logistics”, Oxford Bulletin of Economics and Statistics, Vol. 75 Issue 6, December 2013. - Moreira, M.M., J. Blyde, C. Vulpe, and (2013, eds.), “Too Far to Export: Domestic Transport Costs and Regional Export Disparities in Latin America and the Caribbean”, Inter-American Development Bank, Washington, D.C., U.S.A. - Moreira, M.M., C. Vulpe, and J. Blyde (2008), “Unclogging the Arteries: The Impact of Transport Costs on Latin American and Caribbean Trade”, Inter-American Development Bank, Washington, D.C., U.S.A. - Kahn, T., A. Estevadeordal, and M. M. Moreira (2015), “Bringing Down the Barriers: A Review of IDB Research on Trade Costs in Latin American and the Caribbean”, Inter-American Development Bank, Washington, D.C., U.S.A. - Moreira, M.M., J. Blyde, C. Vulpe, and (2013, eds.), “Too Far to Export: Domestic Transport Costs and Regional Export Disparities in Latin America and the Caribbean”, Inter-American Development Bank, Washington, D.C., U.S.A. (duplicate entry as in source) ### Trade integration, policy, and regional analysis for Latin America and the Caribbean - Bown, C.P., D. Lederman, S. Pienknagura and R. Robertson (2017), “Better Neighbors: Toward a Renewal of Economic Integration in Latin America”, The World Bank Group, Washington, D.C., U.S.A. - Cerra, V. et al. (2017), “Cluster Report – Trade Integration in Latin America and the Caribbean”, IMF Country Report No. 17/66, March 2017, International Monetary Fund, Washington, D.C., U.S.A. - International Monetary Fund (October 2015), Regional Economic Outlook, Western Hemisphere Department, “Trade Integration in Latin America and the Caribbean: Hype, Hope, and Reality”, International Monetary Fund, Washington, D.C., U.S.A. - Mowatt, R. (2017) “Trade Policy Issues in Latin America and the Caribbean: Views from Country Authorities and Current State of Play”, unpublished background paper, International Monetary Fund, Washington, D.C., U.S.A. - Salinas, G. (2021) “Proximity and Horizontal Policies: The Backbone of Export Diversification and Complexity,” IMF Working Paper No. 2021/064. - Campos, R., S. Pienknagura, and J. Timini (2023), “How Far has Globalization Gone? A Tale of Two Regions,” IMF Working Paper WP/23/255. ### Measurement of trade restrictions and services trade databases - Estefania-Flores J., D. Furceri, S.A. Hannan, J.D. Ostry, and A.K. Rose (2022), “A Measurement of Aggregate Trade Restrictions and their Economic Effects”, IMF Working Paper No. 2022/001. - Liberatore, A., and S. Wettstein (2021), “THE OECD-WTO BALANCED TRADE IN SERVICES DATABASE (BaTIS)”: https://www.oecd.org/sdd/its/OECD-WTO-Balanced-Trade-in-Services-database-methodology-BPM6.pdf ### Document identifiers and internal labels observed in the source - IMF WORKING PAPERS Constraint on Trade on the LAC Region - INTERNATIONAL MONETARY FUND 40 *References list as provided in wpiea2024032-print-pdf - References* --- _Source: https://www.imf.org/-/media/files/publications/wp/2024/english/wpiea2024032-print-pdf.pdf_