Policy Research Working Paper 10951 Heterogeneous Effects of Trade Agreements in Central America The Case of CAFTA-DR Sebastian Franco-Bedoya Woori Lee Macroeconomics, Trade and Investment Global Practice October 2024 Policy Research Working Paper 10951 Abstract This paper studies the effects of the Central America– the effect on US exports to Central America was insignif- Dominican Republic Free Trade Agreement on trade among icant. The paper highlights the importance of controlling member countries. It uses the structural gravity model to for member-specific globalization dynamics as well as using estimate both the aggregate and heterogeneous effects across domestic trade flows to capture the effects of specific trade countries and sectors. The findings show that the agreement agreements. Studying dynamic effects, the paper shows that increased intraregional trade among Central American anticipation effects were important for the Central Amer- countries by 27 percent. The impact on Central American ica–Dominican Republic Free Trade Agreement and trade exports to the United States was smaller but positive, while creation effects strengthened over time. This paper is a product of the Macroeconomics, Trade and Investment Global Practice. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at sfranco2@worldbank.org and wlee10@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Heterogeneous Effects of Trade Agreements in Central America: The Case of CAFTA-DR* Sebastian Franco-Bedoya† Woori Lee‡ Keywords: trade agreements, regional integration, CAFTA-DR, gravity equation. JEL Classifications:: : F100, F140, F160. * We are grateful to Oscar Calvo-Gonzalez, Barbara Cunha, and Doerte Doemeland for useful suggestions and feedback. The paper should not be reported as representing the official views of the World Bank. The opinions expressed and the arguments employed are those of the authors. † World Bank, E-mail: sfranco2@worldbank.org ‡ World Bank, E-mail: wlee10@worldbank.org 1 Introduction Regional trade agreements can play a key role in boosting trade and investment in de- veloping countries. With the lack of progress in multilateral negotiations, regional trade agreements (RTAs) have been the main channel for trade liberalization in the past few decades.1 The rise of RTAs was not only in terms of the number of agreements, but also in terms of the contents as agreements became increasingly "deeper", covering a wide set of border and behind-the-border policy areas beyond preferential tariffs and simple market access (Mattoo et al., 2020). In this paper, we study the impact of the Central America-Dominican Republic Free Trade Agreement (CAFTA-DR) on trade flows between the member countries. CAFTA- DR was one of the first RTAs between the United States and a small group of developing countries. Entering into force between 2006 and 2009, it was one of the most compre- hensive regional integration agreements in the Latin America and the Caribbean (LAC) region. Importantly, the agreement was conceived not only as a trade agreement with the United States, but also among the Central American countries and the Dominican Republic to promote intra-regional trade. This makes it an interesting case to study the heterogeneity in the effects of RTAs, given that it is both an intra-regional and extra- regional agreement. Approaching nearly 20 years since its negotiation, empirical research on the impact of CAFTA-DR is limited. Economic fundamentals help to explain the public’s attitude toward trade policy and trade openness, as shown by research on Costa Rica’s referendum on CAFTA-DR (Urbatsch, 2013; Van Patten and Méndez, 2022). This is why an evidence- based understanding of the impact of the agreement and how the effects may have varied across countries and sectors is important. We use the structural gravity model and the latest estimation techniques to examine the impact of CAFTA-DR on trade flows between member countries. Including domestic trade flows, we investigate the trade-enhancing effect of CAFTA-DR, and how the impact varied across several dimensions: across country pairs (between Central American intra- regional trade and trade with the United States), across sectors, GVC indicators, and over time (anticipation effects). We find that the trade-enhancing average effect of CAFTA- 1 The WTO defines regional trade agreements (RTAs) as any reciprocal trade agreement between two or more partners, not necessarily belonging to the same region. 2 DR among Central American countries is 31 percent. In contrast, the effect on Central American countries’ exports to the US is positive but not statistically significant at the aggregate level. We observe heterogeneous effects across country pairs for trade with Central America and exports to the US. The effect of CAFTA-DR on exports from the US to Central America is not statistically significant. The trade creation effect of RTAs is well established in the literature, with continuous methodological progress to address endogeneity and identify causal effects. Empirical ev- idence suggests that RTAs increase trade flows between member countries by an average of 12 to 24 percent (Larch and Yotov, 2023; Felbermayr et al., 2022).2 The latest contribu- tions to the literature emphasize the significant benefits of estimating gravity equations with domestic (in addition to international) trade flows (Yotov, 2022). The main reasons for this are that this is consistent with trade theory, data on domestic trade flows are now available, and it makes it possible to identify trade diversion effects, non-discriminatory policies, and the impact of country-specific characteristics. However, due to data avail- ability, until recently, methodological advances have typically been made with smaller samples and have focused on the effect for advanced economies. Recent studies on the heterogeneous impact of RTAs underscore the risks of directly applying findings from studies on developed economies to developing countries. The robust empirical support for the positive average effect of RTAs masks large degrees of heterogeneity across and within agreements. In fact, Baier et al. (2019) find that around half of the RTAs included in their study have a statistically insignificant or negative impact on trade. Furthermore, the majority of heterogeneity occurs within RTAs (rather than across different RTAs), with asymmetric effects within pairs (on exports vs. imports) also playing an important role. Given the varying effects of different agreements across different countries, it is important to understand the effects of specific agreements on trade of different countries and country pairs. Literature review - This paper contributes to the literature that examines the effect of individual trade agreements. This literature dates back to the study of Benelux and British Commonwealth preference arrangements by Tinbergen (1963). In the LAC region, the literature has focused mostly on the North American Free Trade Agreement (NAFTA), 2 Larch and Yotov (2023) find an effect of 24 percent with a sample of 70 countries, while Felbermayr et al. (2022) find an effect of 12 percent with 186 countries. 3 with Caliendo and Parro (2015) being one of the most influential. Other agreements in the region, like the Andean Community, have been studied by Carrillo-Tudela and Li (2004). Examples of studies on MERCOSUR include Nowak-Lehmann and Martinez- Zarzoso (2003) and Baier et al. (2019). We also contribute to the more recent group of studies investigating the heterogeneous effects across and within trade agreements on bilateral trade flows (Baier et al., 2019; Kohl, 2014). CAFTA-DR was not included in the study by Baier et al. (2019) since their sample period ends in 2006. Our study contributes to this literature by investigating the impact of CAFTA-DR in detail, allowing for a broad country and time coverage, and heterogeneity across different dimensions (countries, country pairs, sectors, and over time). There are only two previous empirical studies on the specific effects of CAFTA-DR that we are aware of.3 El Dahrawy Sánchez-Albornoz and Timini (2021) analyze the process of economic integration in Latin America, assessing the effect of the trade agreements signed by at least one Latin American country on international trade. They find that, on average, trade agreements positively affected Latin American trade flows, but CAFTA-DR is part of the 43 percent of agreements with no statistically significant effect on trade. More recently, Rojas Rodríguez and Matschke (2023) found a negative but insignificant average effect of CAFTA-DR on members’ bilateral trade, while the impact was negative and significant for Costa Rica and the Dominican Republic. They explain the negative results by the existence of a common exogenous trend among CAFTA-DR members to diversify trade from the United States to the rest of the world, leading to a shift away from the United States as their main export destination and import source. Our data and empirical strategy differ significantly from Rojas Rodríguez and Matschke (2023). Most importantly, we include intra-national trade flows at the sector level for a wide range of countries, including CAFTA-DR members, which were not included in their paper.4 We see this as a non-trivial improvement, particularly when the main objects of interest are the country- specific estimates of CAFTA-DR.5 Without the necessary domestic observations for the 3 For an analytical study assessing the potential implications of CAFTA-DR before its implementation, see World-Bank (2005). Koehler-Geib et al. (2014a) and Koehler-Geib et al. (2014b) are early country-specific analyses for Costa Rica. 4 Rojas Rodríguez and Matschke (2023) use different data sources for intra-national trade flows and data availability is inconsistent (not available) after 2015. Unfortunately, domestic trade flows for El Salvador, Guatemala, Honduras, Nicaragua, and Dominican Republic, are not included. For Costa Rica, domestic trade flows are only included towards the end of the period. 5 This concern adds to the advice from Baier et al. (2019), saying that the more granular the estimates of trade agreements beyond the world average effect require a more careful analysis. 4 country members, it becomes difficult to build a sound empirical strategy. The rest of the paper is organized as follows. Section 2 provides the history and background of trade integration in Central America. Section 3 describes the data. Section 4 presents the empirical methodology. The results are presented in Section 5, and Section 6 concludes. 2 Trade Integration in Central America: Background and Moti- vation CAFTA-DR is an important but certainly not the first of the region’s efforts to promote economic integration and reduce trade barriers. Prior to CAFTA-DR, there were several trade agreements in the Central American region aimed at promoting economic integra- tion and reducing trade barriers. One of the earliest agreements was the Central Amer- ican Common Market (CACM), which was established in the 1960s and aimed to create a free trade area between member countries. Despite some initial successes, the CACM faced several challenges, including the eruption of armed conflicts and the adoption of protectionist policies by member countries. In the 1990s, the region began to pursue more ambitious integration efforts, such as the establishment of the Central American Integration System (SICA). Dominican Republic joined the trade agreement with Central America a bit later, in 2001. CAFTA-DR members also joined the WTO in the 1990s, subjecting themselves to the rules and obligations set out in the various agreements negotiated at the multilateral level, including those on tariffs, non-tariff measures, and intellectual property rights. These initiatives paved the way for the negotiation of the CAFTA-DR agreement, which entered into force in 2006, with the exception of Dominican Republic and Costa Rica, which joined the agreement in 2007 and 2009, respectively. The agreement represented a significant step forward in the region’s trade integration efforts, as it provided for greater market access, the reduction of tariffs and non-tariff barriers, and the protection of intellectual property rights.6 Preferential access to the US market- Before the implementation of the Central America- Dominican Republic Free Trade Agreement (CAFTA-DR), the United States provided 6 CAFTA-DR negotiation steps can be found at http://www.sice.oas.org/tpd/usa_cafta/usa_cafta_e.asp. 5 preferential access to Central American countries through a series of trade programs. These programs included the Caribbean Basin Initiative (CBI), which was established in 1983 and aimed to promote economic development in the Caribbean region through trade preferences. The CBI provided duty-free treatment for certain products from eligi- ble countries, including many Central American countries. In addition, the Generalized System of Preferences (GSP) provided preferential access for a wide range of products from developing countries, including many products from Central American countries. These trade programs were important for Central American countries, as they provided a competitive advantage for their exports to the United States. These regional dynamics show that while the CAFTA-DR agreement represented a significant step forward in the region’s trade integration efforts, as it provided for greater market access, the reduction of tariffs and non-tariff barriers, and the protection of in- tellectual property rights, it was part of a long-term trade liberalization dynamic in the region. Many initiatives have aimed to boost trade and investment within the region and have also facilitated the participation of Central American countries in the global econ- omy. We aim to study the impact of CAFTA-DR in this context and to solve the empirical identification challenges arriving from those dynamics. 3 Data We use the Eora Multi-Region Input-Output (MRIO) database, which provides a panel data set of trade flows, including reliable data on intra-national flows (Lenzen et al., 2012, 2013). This dataset covers bilateral trade flows across sectors for 189 countries over the pe- riod 1990–2015, and has several advantages for our purposes.7 First, the dataset includes intra-national trade flows, which are needed for theory-consistent gravity estimations. As demonstrated by Baier et al. (2019), the estimates of RTAs are affected by the inclusion of domestic trade flows. Second, it has a wide coverage both in terms of countries and time. The large country coverage is essential for analyses focusing on developing coun- tries, which are often left out due to data constraints. The time span of the dataset is also relatively long, and it covers a period of intense globalization efforts with a number of new RTAs. Third, consistent with our focus, the dataset has already been used to analyze 7 The services data tend to be of lower quality than goods data due to the nature of services and how these data are recorded, but we include it for completeness of the analysis and the interest in the service sector. 6 the impact of RTAs.8 We are also interested in quantifying the impact of CAFTA-DR on bilateral cross- border production linkages. For this purpose, we need data that distinguishes between final and intermediate goods on the one hand, and between domestic and foreign-value- added parts of trade flows. Eora itself provides bilateral trade data in final and intermedi- ate goods. Our value-added trade data comes from Borin et al. (2021), which is based on Eora input-output tables, and can be accessed through the World Integrated Trade System (WITS). This data provides different variables that disaggregate total bilateral trade into traditional trade and GVC-related trade. Traditional trade corresponds to the domestic value added embodied either in final or intermediate goods or services that are directly consumed by the importing economy (this value only crosses one border). Pure forward GVC-related trade is the value-added in goods and services entirely generated within the domestic chains (without any border crossing) exported to partners, which in turn reex- port it to other markets. In this case, the exporting sector is engaged in GVC activities at the origin of the chain. Pure backward GVC-related trade is the value of goods and services produced with imported inputs and exported to the final destination market. In this case, the exporting sector engages in GVC activities at the end of the chain. Finally, two-sided or mixed GVC-related trade is the value of goods and services produced with imported inputs, exported by the sector to partners, which in turn re-export it to other markets. The exporting sector is located in a position of the chain that is more central. Our main variable of interest is the domestic value added to exports.9 It determines the contribution of a country’s own economy to the production of exports, rather than simply counting the total gross export value. It is obtained by summing the pure forward GVC trade and the traditional trade flows.10 Data on trade agreements come from the World Bank’s Deep Trade Agreements Database (Mattoo et al., 2020). It maps the coverage of 52 selected policy areas in 279 agreements notified at WTO and signed after 1958. It also includes information about the legal enforceability of each policy area. 8 Other international input-output tables are also available but have either fewer countries or fewer years. The global Eora MRIO tables with a harmonized 26-sector classification across countries (also known as Eora26) can be accessed from http://worldmrio.com/simplified/. For further description of the data, its sources, and construction, see Lenzen et al. (2012, 2013). 9 While the domestic value-added of trade is our main variable of interest, we study the impact of CAFTA- DR on all the GVC variables available from Borin et al. (2021). 10 See the appendix for more details. 7 4 Methodology This section describes how we estimate the effect of CAFTA-DR on trade. We start with the structural gravity model we use to implement our empirical strategy. We explore the impact of the CAFTA-DR trade agreements across different dimensions, disentangling the CAFTA-DR specific effect from the average RTA effect and estimating the effects across members. We rely on a robust specification, based on the latest developments in the literature, that allows us to capture the heterogeneous effects of this trade agreement. 4.1 Structural gravity Our starting point is a general version of the structural gravity equation as derived by Anderson (1979) and as popularized by Eaton and Kortum (2002) and Anderson and van k denote the value of exports from origin country i to destination Wincoop (2003). Let Xij ,t country j in sector k. The gravity equation for these trade flows is as follows: 1− σ k Ek j Yi k k tij k Xij = (1) Yk Pjk Πik k 1− σ k tij Yik ( Pjk ) 1− σ k =∑ (2) i Πik Yk k 1− σ k tij Ek =∑ 1− σ k j (Πik ) (3) j Pjk Yk k includes both international and domestic trade flows, X k . E denotes where Xij ,t ii,t j total expenditure on sector k at destination j, while Yik denotes total output value k from country i to all destinations. Y is the total world output of goods k. σk is the trade elasticity of substitution across origin countries i in goods k. Πi is the outward multilateral resistance, which consistently aggregates the trade costs faced by the producers in each region i as if they ship to a uniform world market. Similarly, the inward multilateral resistance, Pj , consistently aggregates the trade costs of the consumers in each region j as if they buy from a uniform world market. 8 4.2 Empirical strategy We rely on the latest empirical gravity literature developments to specify our econometric model as follows:11 Xij,t = exp β rta RTAij,t + ∑ β t BRDRij,t + γi,t + µij + ω j,t + ε ij,t (4) t Here, Xij,t denotes bilateral trade flows from country i to country j at time t in levels. Importantly, Xij,t includes both international and intra-national trade flows, which is crit- ical to get theory-consistent estimates (Yotov, 2022; Larch and Yotov, 2023). Our analysis uses trade data in total goods (agriculture and manufacturing sectors) as a baseline.12 We also use sector-specific data for a more disaggregated analysis. RTAij,t captures the average world effect of trade agreements and will serve as a ref- erence point to the effect of CAFTA-DR. Equation (4), as is standard in the literature, uses exporter-time fixed effects γi,t and importer-time fixed effects ω j,t to control for the unob- servable exporter and importer multilateral resistances. These fixed effects will also ab- sorb/control for any other country-time-specific characteristics that may impact bilateral trade on the exporter and importer sides. In addition, following the recommendations of Baier and Bergstrand (2007), we also employ country-pair fixed effects (µij ), which control for all time-invariant bilateral trade costs and will mitigate endogeneity concerns with respect to our trade policy variables of interest. Finally, we follow Bergstrand et al. (2015) and account for common globalization effects that could bias our trade agreements variables, with a set of time-varying border dummy variables BRDRij,t , for which a pa- rameter β t is estimated for each year t separately. We estimate our model using the Poisson pseudo maximum likelihood (PPML) estimator, which allows accounting for het- eroscedasticity in trade data and takes advantage of the information that is contained in zero trade flows (Santos Silva and Tenreyro, 2006). Our primary purpose is to estimate the effects of the CAFTA-DR agreement. For this, we introduce the variables CAFTAij,t to capture the specific impact of this trade 11 Larch and Yotov (2023) survey the methods and data for estimating the effects of trade agreements. 12 We focus on total goods data for our baseline due to data quality issues in the services trade flows. 9 agreement, as follows: Xij,t = exp β rta RTAij,t + β ca f ta CAFTAij,t + ∑ β t BRDRij,t + γi,t + µij + ω j,t + ε ij,t (5) t where the CAFTAij,t indicator switches to one on the year when the CAFTA-DR agree- ment is enforced in both countries i and j, which ranges between 2006 and 2009. In some specifications, we also interact the CAFTAij,t variable with country dummies for captur- ing country or country-pair-specific effects. As shown by Baier and Bergstrand (2007), the effects of trade agreements tend to evolve over time, phasing in over several years. For this reason, we include a set of lags, CAFTAij,t+s . In the same way, the literature calls for the inclusion of leads to capture the anticipation effects of trade agreements (Egger et al., 2022). Also, we use consecutive-year data instead of estimating gravity equations with interval or averaged data. Egger et al. (2022) show that relative to interval or averaged data, the use of consecutive-year data avoids downward-biased effect estimates due to the distribution of trade-policy events during an event window as well as due to anticipation (pre-interval) and delayed (post-interval) effects, and it improves the efficiency of effect estimates due to the use of more data. These adjustments yield the following model: Xij,t = exp β rta RTAij,t + β ca f ta CAFTAij,t + ∑ β s CAFTAij,t+s + ∑ β s CAFTAij,t−k s k (6) + ∑ β t BRDRij,t + γi,t + µij + ω j,t + ε ij,t t where ∑k β s CAFTAij,t−k and ∑s β s CAFTAij,t+s capture the lags and leads of the CAFTA- DR effect, respectively. We also have additional specifications. First, while Equation (6) controls for world average globalization dynamics, it does not control for those that are country-specific. Given that these controls may be important for obtaining unbiased estimates, we use the following specification: Xij,t = exp β rta RTAij,t + β ca f ta CAFTAij,t + ∑ β s CAFTAij,t+s + ∑ β s CAFTAij,t−k s k (7) + ∑ β t BRDRij,t × 1(i = c) + γi,t + µij + ω j,t + ε ij,t t 10 where the globalization coefficients are interacted with a country c specific dummy 1(i = c) to capture for those country-specific globalization dynamics. This follows Franco- Bedoya (2023) analysis of country-specific globalization dynamics and its impact on the estimation of RTA coefficients. Second, as the global economy has become increasingly structured around global value chains (GVCs), intermediate inputs often cross borders several times along the production process. The implication is that gross trade flows from gross trade statistics no longer reflect the domestic value-added content trade. For this reason, we use other dependent variables that allow us to decompose bilateral trade flows into domestic value- added and GVC-related trade terms, as follows: GVCij,t = exp β rta RTAij,t + β ca f ta CAFTAij,t + γi,t + µij + ω j,t + ε ij,t (8) where GVCij,t are different GVC trade flows. Total trade, as used so far in our em- pirical strategy, is the sum of traditional and GVC trade. Traditional trade is the value of trade flows that only cross one border, being absorbed by the direct trade partner. GVC related-trade is the value of bilateral trade that crosses at least two borders, flowing along a supply chain. GVC-related trade can be further decomposed into three components, pure forward, pure backward, and two-sided, as explained by Borin et al. (2021). "Pure backward GVC related-trade" is the value of goods and services produced with imported inputs and exported to the final destination market. "Pure forward GVC related-trade" is the value-added in goods and services entirely generated within the domestic economy (by different sectors) exported to partners, which in turn, re-exports it to other mar- kets. "Two-sided GVC-related trade" is the value of goods and services produced with imported inputs and exported to partners, which in turn re-export them to other mar- kets. Finally, note that the domestic value-added content of trade is the summation of traditional trade and pure forward GVC-related trade. These distinctions are important to understand the impact of a trade agreement on economic activity and will allow for gaining deeper insights regarding the specific effects of CAFTA-DR.13 13 We access this data generated by Borin et al. (2021), based on Eora, through the WITS website. 11 5 Results In this section, we estimate the effects of CAFTA-DR. We start with a comparison to the average RTA effect available in the literature, and then explore the heterogeneous effects of CAFTA-DR at the sector, country, and country-pair levels. We also examine the dynamics of these effects and consider the effect on GVCs indicators, including domestic value-added in trade. 5.1 The importance of including domestic trade flows In this section, we compare the estimated average effect of RTAs using our data, with and without including domestic trade flows, to the standard estimates established in the literature. The first column in Table (1) shows the average RTA effect using our data on only international trade flows. We identify a [exp(0.02) − 1] × 100 = 2.02 percent positive and significant effect. However, as discussed before, the inclusion of domestic trade flows in estimating the gravity equation provides significant benefits of adhering to the theory. Column 2 includes domestic trade flows and obtains a much larger effect, estimated at [exp(0.29) − 1] × 100 = 33.64 percent. But this estimation is biased upward due to not controlling for globalization effects (Bergstrand et al., 2015). Column 3 controls for globalization effects and estimates the average effect of RTA to be [exp(0.12) − 1] × 100 = 12.75 percent. The finding that intra-national trade data are important for estimating the effects of RTAs is in line with several recent contributions in the literature (Dai et al., 2014; Baier et al., 2019). More importantly for us, the estimated coefficient in column 3, our bench- mark average RTA effects, is in line with recent studies that include a large number of countries (including many developing countries), as we do. Very close to our benchmark result, Felbermayr et al. (2022) find an effect of 12 percent while covering 186 countries. Other recent studies that use a smaller sample of countries find slightly larger coefficients. Larch and Yotov (2023) find an effect of 24 percent with a sample of 70 countries. 5.2 Heterogeneous impact of CAFTA-DR This section presents the results of our analysis focusing on the impact of CAFTA-DR. In Table (2), we disentangle the additional effect (to the average RTA effect) of CAFTA-DR, 12 Table 1: Comparison to the literature in the manufacturing sector (1) (2) (3) trade trade trade RTA 0.02∗∗ 0.29∗∗∗ 0.12∗∗∗ (0.01) (0.03) (0.03) Domestic trade NO YES YES Globalization control NO NO YES Observations 422656 425984 425984 Note: All regressions use the PPML estimator with exporter-time, country-pair and importer-time fixed effects. Robust standard errors, clustered at the country-pair level, are in parentheses. ∗ p < 0.10,** p < 0.05,*** p < 0.01 and report the total effect of CAFTA-DR at the bottom of the table, by estimating equation (5). Results in column 1 suggest that CAFTA-DR had a negative effect on bilateral trade in goods among the member countries. While an average RTA increased trade in goods by 13 percent, the impact of CAFTA-DR is estimated to have been negative, decreasing trade by an average of 12 percent. The negative impact is observed consistently across sectors, despite differences in magnitudes ranging from 6.2 to 12.7 percent. The negative impact of CAFTA-DR on trade between member countries is puzzling but consistent with earlier studies that also fail to find a trade-promoting effect of the CAFTA-DR agreement (El Dahrawy Sánchez-Albornoz and Timini, 2021; Rojas Rodríguez and Matschke, 2023). In fact, a substantial share of RTAs have insignificant or negative impacts on trade flows, as shown in Baier et al. (2019). To better understand the negative effect of CAFTA-DR, we investigate potentially het- erogeneous impacts within the agreement. CAFTA-DR was one of the first RTAs between the U.S. and a small group of developing countries, acting as an agreement not only be- tween the U.S. and Central America, but also among Central American countries and the Dominican Republic.14 We, therefore, first disentangle the effect of CA trade with the U.S. and the intra-regional trade within Central America. Results in Column 2 of Table (3) show that the negative average effect of CAFTA-DR identified in Table (2) is driven by CA trade flows with the U.S. In fact, intra-regional trade is estimated to have increased 14 In the remainder of the paper, Central America (CA) refers to signatories of CAFTA-DR excluding the U.S.: Costa Rica, El Salvador, Guatemala, Honduras, Nicaragua, and the Dominican Republic. 13 Table 2: CAFTA-DR results (1) ( 2) (3) ( 4) All goods Manufacturing Agriculture Services RTA 0.13∗∗∗ 0.13∗∗∗ 0.12∗∗∗ 0.06∗∗∗ (0.03) (0.03) (0.03) (0.01) RTA × CAFTA -0.25∗∗∗ -0.25∗∗∗ -0.18∗∗∗ -0.13∗∗∗ (0.04) (0.04) (0.04) (0.02) Total CAFTA -0.12∗∗∗ -0.12∗∗∗ -0.06∗∗ -0.07∗∗∗ (0.03) (0.04) (0.03) (0.02) Observations 425984 425984 425984 425984 Note: All goods refer to manufacturing and agricultural products. All regressions use the PPML es- timator with exporter-time, country-pair and importer-time fixed effects. Border-time controls are also included in all regressions. Robust standard errors, clustered at the country-pair level, are in parentheses. ∗ p < 0.10,** p < 0.05,*** p < 0.01 by 31 percent as a result of the agreement. This is an important finding since trade lib- eralization within the region is critical to enhance regional integration and support the development of regional supply chain linkages. Even within a given country pair, an RTA does not necessarily affect trade in both directions symmetrically (Baier et al., 2019). We, therefore, further differentiate the effects of CAFTA-DR on Central America’s exports to and imports from the U.S. Column 3 shows that the negative effect is much stronger and more robust for Central American imports from the U.S., which decreased by 25.9 percent. On the other hand, the CAFTA- DR impact on Central America’s exports to the U.S. is statistically insignificant. While the baseline results are presented for trade in goods, columns 4-6 show the estimates by broad sectors. The positive effect of CAFTA-DR for intra-regional trade is largely driven by the increased trade in the manufacturing sector, which also represents a large share of trade flows for Central American countries. The effect on agriculture and services is not statistically significant. We now turn to explain these results in which the intra-CAFTA effect is positive, but the one for Central American exports to the US is not significant and negative for US exports to Central America. Rojas Rodríguez and Matschke (2023) discussed the existence of a common trend among CAFTA-DR members to diversify trade from the United States to the rest of the world, explaining their shift away from the U.S. as their main export destination and import source that may have contributed to these results. From our point 14 Table 3: CAFTA-DR effects disentangling specific effects with the US (1) (2) (3) (4) ( 5) (6) All goods All goods All goods Manufacturing Agriculture Services RTA 0.13∗∗∗ 0.13∗∗∗ 0.13∗∗∗ 0.13∗∗∗ 0.12∗∗∗ 0.06∗∗∗ (0.03) (0.03) (0.03) (0.03) (0.03) (0.01) RTA × CAFTA -0.25∗∗∗ 0.13∗∗∗ 0.13∗∗∗ 0.16∗∗∗ -0.07∗∗ -0.01 (0.04) (0.03) (0.03) (0.04) (0.03) (0.03) RTA × CAFTA × US -0.44∗∗∗ (0.05) RTA × CAFTA × US as Exp -0.57∗∗∗ -0.58∗∗∗ -0.29∗∗∗ -0.13∗∗∗ (0.06) (0.06) (0.07) (0.04) RTA × CAFTA × US as Imp -0.32∗∗∗ -0.33∗∗∗ -0.07 -0.15∗∗∗ (0.06) (0.06) (0.05) (0.05) TOTAL CAFTA -0.12∗∗∗ 0.26∗∗∗ 0.27∗∗∗ 0.29∗∗∗ 0.05 0.05 (0.03) (0.05) (0.05) (0.05) (0.04) (0.03) TOTAL CAFTA - US -0.17∗∗∗ (0.03) TOTAL CAFTA - US as Exp -0.30∗∗∗ -0.30∗∗∗ -0.24∗∗∗ -0.08∗∗∗ (0.05) (0.05) (0.05) (0.03) TOTAL CAFTA - US as Imp -0.05 -0.04 -0.01 -0.09∗∗ (0.04) (0.04) (0.03) (0.04) Observations 425984 425984 425984 425984 425984 425984 Note: All goods refer to manufacturing and agricultural products. All regressions use the PPML es- timator with exporter-time, country-pair and importer-time fixed effects. Border-time controls are also included in all regressions. Robust standard errors, clustered at the country-pair level, are in parentheses. ∗ p < 0.10,** p < 0.05,*** p < 0.01 15 of view, this potential effect and the fact that Central American countries already enjoyed market access to the US before signing CAFTA-DR explains the insignificant effect of exports from Central America to the US. A possible explanation for the negative impact of CAFTA-DR on U.S. exports to Cen- tral America is that we are not controlling for US-specific globalization effects that might be biasing our CAFTA × USasExp coefficient. The literature has widely studied the neg- ative effects on manufacturing employment in the US starting in the early 2000s when globalization arguably intensified. For instance, Autor et al. (2019) study the large-scale labor demand shocks stemming from rising international manufacturing competition dur- ing 1990–2014 and its consequences. Among their conclusions, they find that this trade shock significantly reduced the employment and earnings of young adult males. There is also evidence that deep trade agreements (or other type of trade barrier reductions) can reduce exports of advanced economies to developing countries due to reducing the fixed costs of Foreign Direct Investment (FDI) significantly more than trade barriers. This could also have been the case with the trade agreements signed by Japan with Asian countries, as shown by Baek and Hayakawa (2022). Franco-Bedoya (2023) estimates a unique decline in the US manufacturing sector glob- alization dynamics since 2000, in line with the evidence available in the literature. More- over, he shows how country-specific globalization dynamics can affect the estimation of RTA coefficients. For this reason, we use the specification in Equation (7). Table (4) fol- lows Equation (7) and introduces the controls for globalization dynamics specific to CA, the US, and the rest of the world. In comparison to columns (1) to (4), where these spe- cific globalization dynamics are not included, columns (5) to (8) show that there is no statistically significant effect on exports from the US to Central America. In intraregional trade, the effect on manufacturing is very similar, but in agriculture and services, it is now positive with 21 and 14 percent, respectively. Finally, CAFTA has a positive effect ok 13 percent on agricultural exports from Central America to the US. These results show that including the specific globalization trends is important in obtaining robust RTA effects at disaggregate levels like country pairs. Given the importance of the manufacturing sector in the context of CAFTA-DR and the availability of data, we go one step further to explore the effect across different sub- sectors. We find that the positive effect of CAFTA-DR on intra-regional trade is observed 16 Table 4: CAFTA-DR effects controlling for Globalization Dynamics (1) ( 2) (3) ( 4) ( 5) ( 6) (7) ( 8) All goods Manufacturing Agriculture Services All goods Manufacturing Agriculture Services RTA 0.13∗∗∗ 0.13∗∗∗ 0.12∗∗∗ 0.06∗∗∗ 0.11∗∗∗ 0.10∗∗∗ 0.13∗∗∗ 0.06∗∗∗ (0.03) (0.03) (0.03) (0.01) (0.02) (0.02) (0.03) (0.01) RTA × CAFTA 0.13∗∗∗ 0.16∗∗∗ -0.07∗∗ -0.01 0.13∗∗∗ 0.14∗∗∗ 0.06 0.07∗∗ (0.03) (0.04) (0.03) (0.03) (0.04) (0.05) (0.05) (0.03) RTA × CAFTA × US as Exp -0.57∗∗∗ -0.58∗∗∗ -0.29∗∗∗ -0.13∗∗∗ -0.28∗∗∗ -0.27∗∗∗ -0.21∗∗∗ -0.14∗∗∗ (0.06) (0.06) (0.07) (0.04) (0.06) (0.06) (0.07) (0.04) RTA × CAFTA × US as Imp -0.32∗∗∗ -0.33∗∗∗ -0.07 -0.15∗∗∗ -0.16∗∗∗ -0.16∗∗∗ -0.07 -0.08 (0.06) (0.06) (0.05) (0.05) (0.06) (0.06) (0.06) (0.06) TOTAL CAFTA 0.27∗∗∗ 0.29∗∗∗ 0.05 0.05 0.24∗∗∗ 0.24∗∗∗ 0.19∗∗∗ 0.13∗∗∗ (0.05) (0.05) (0.04) (0.03) (0.05) (0.06) (0.07) (0.04) TOTAL CAFTA - US as Exp -0.30∗∗∗ -0.30∗∗∗ -0.24∗∗∗ -0.08∗∗∗ -0.03 -0.03 -0.03 -0.01 (0.05) (0.05) (0.05) (0.03) (0.04) (0.04) (0.07) (0.03) TOTAL CAFTA - US as Exp -0.05 -0.04 -0.01 -0.09∗∗ 0.09∗ 0.08 0.12∗∗∗ 0.05 (0.04) (0.04) (0.03) (0.04) (0.04) (0.05) (0.03) Group globalization NO NO NO NO YES YES YES YES Observations 425984 425984 425984 425984 425958 425958 425958 425958 Note: All goods refer to manufacturing and agricultural products. All regressions use the PPML es- timator with exporter-time, country-pair and importer-time fixed effects. Border-time controls are also included in all regressions. Robust standard errors, clustered at the country-pair level, are in parentheses. ∗ p < 0.10,** p < 0.05,*** p < 0.01 across all sub-sectors, except transport, where the result is not statistically significant (Figure (1)) Regarding Central American exports to the U.S., the statistically insignificant results are also consistent across sub-sectors. We now focus on manufacturing and agricultural products ("all goods") to estimate the country and country-pair-specific effects. Figure (2) shows that at this level of disag- gregation, most effects are positive. The largest effects are for intra-regional trade for El Salvador’s exports. The lowest, negative, effects seem to be on exports to the US from Costa Rica and Honduras. The case of Costa Rica is partly explained by the decline in electronics around 2013.15 In spite of some negative effects, no country in Central America displays a negative total effect of CAFTA-DR on trade, and several estimates, in particular in intra-regional trade, are positive. 5.3 GVC dimension Following the growing importance of Global Value Chains (GVCs), we now focus on the CAFTA-DR effect on GVC trade flows. We do it for two reasons. First, it allows us to understand whether CAFTA-DR has contributed to supply chain integration. Second, it offers a way to estimate the effect on economic activity through the effect on the domestic 15 This is related to the exit of Intel as an assembler of microchips in Costa Rica. 17 Figure 1: CAFTA-DR effect by manufacturing sub-sectors Note: This figure distinguishes the intra-CAFTA effect and the effect o exports to the US. All regression use the PPML estimator. The dependent variable is bilateral exports. Fixed effects and constants not reported for the sake of simplicity. t statistics in parentheses. ∗ p < 0.10,** p < 0.05,*** p < 0.01 Figure 2: Heterogeneous impact of CAFTA by country and country pairs (a) By country: Total and intra-regional trade (b) By country pair (directional) Note: This figure disentangles the country and country-pair effects of CAFTA-DR. All regressions use the PPML estimator with exporter-time, country-pair and importer-time fixed effects. Border-time controls are also included in all regressions. 18 value-added content of trade. We first look into supply chain integration by distinguishing between final and inter- mediate goods. Columns 2 and 3 of table (5) show that the positive effect on intra-regional trade is stronger in intermediate goods. This suggests a positive effect on intra-regional production networks. The negative effect on US exports to Central America is stronger on final goods. The rest of table (5) uses GVC trade flows as the dependent variable, and in this case, domestic bilateral trade flows are not included in the analysis.16 In column 4, the variable of interest is the domestic value-added content of trade. It determines the contribution of a country’s own economy to the production of exports, rather than simply counting the total gross export value.17 Traditional trade, in column 5, corresponds to the domestic value added embodied either in final or intermediate goods or services that are directly consumed by the importing economy (this value only crosses one border). Pure forward GVC-related trade (column 6) is the value-added in goods and services entirely generated within the domestic chains, without any border crossing, exported by the sector to partners, which in turn re-export it to other markets. The exporting sector is engaged in GVC activities at the origin of the chain. Pure backward GVC-related trade in column 7 is the value of goods and services produced with imported inputs and exported by the sector to the final destination market, with the exporting sector engaged in GVC activities at the end of the chain. Finally, two-sided or mixed GVC-related trade, in column 8, is the value of goods and services produced with imported inputs, exported by the sector to partners, which in turn re-export it to other markets. The exporting sector is located in a position of the chain that is more central. The results in table (5) show an interesting outlook of the effect of CAFTA-DR on GVC-trade flows and the positioning of Central American countries along the value chains. First, intra-regional trade benefited from a trade-enhancing effect on domestic value added to exports, explained by the effect on traditional trade. It also benefited from a positive effect on pure backward GVC-related trade, confirming the effect on intra- regional production networks. Interestingly, similar effects are observed on exports from Central America to the US. And the negative effect on US exports to Central America is observed in all GVC-trade flows, except for the pure forward component. 16 Domestic bilateral trade flows are not available. See (Borin et al., 2021) for more details. 17 Asexplained before, total trade equals traditional plus GVC-related trade. Domestic value added in exports equals traditional trade plus pure forward GVC trade. 19 Table 5: CAFTA GVCs (1) (2) (3) ( 4) (5 ) ( 6) ( 7) (8) Trade goods Final goods Inputs DVA Trad. trade GVCFP GVCBP GVCMIX RTA 0.11∗∗∗ 0.10∗∗∗ 0.10∗∗∗ 0.02∗∗∗ 0.02∗∗∗ 0.01 0.01 -0.00 (0.02) (0.02) (0.02) (0.01) (0.01) (0.01) (0.01) (0.01) RTA × CAFTA 0.13∗∗∗ 0.11∗∗ 0.15∗∗∗ 0.07∗∗∗ 0.08∗∗∗ 0.01 0.07∗∗ 0.02 (0.04) (0.05) (0.04) (0.03) (0.03) (0.02) (0.03) (0.04) RTA × CAFTA × US as Exp -0.28∗∗∗ -0.23∗∗∗ -0.27∗∗∗ -0.16∗∗∗ -0.17∗∗∗ -0.03 -0.16∗∗∗ -0.10∗∗ (0.06) (0.07) (0.06) (0.03) (0.03) (0.03) (0.04) (0.04) RTA × CAFTA × US as Imp -0.16∗∗∗ -0.10 -0.16∗∗∗ -0.03 -0.03 0.01 -0.01 -0.00 (0.06) (0.06) (0.06) (0.02) (0.02) (0.05) (0.03) (0.06) TOTAL CAFTA 0.24∗∗∗ 0.20∗∗∗ 0.25∗∗∗ 0.09∗∗∗ 0.11∗∗∗ 0.01 0.08∗∗ 0.02 (0.05) (0.06) (0.06) (0.03) (0.03) (0.03) (0.04) (0.04) TOTAL CAFTA - US as Exp -0.03 -0.03 -0.02 -0.07∗∗∗ -0.07∗∗∗ -0.01 -0.08∗∗∗ -0.07∗∗∗ (0.04) (0.05) (0.04) (0.02) (0.02) (0.02) (0.02) (0.03) TOTAL CAFTA - US as Imp 0.09∗ 0.11∗∗ 0.09∗ 0.06∗∗∗ 0.08∗∗∗ 0.03 0.07∗∗ 0.02 (0.04) (0.05) (0.05) (0.02) (0.02) (0.05) (0.03) (0.06) Observations 425984 425984 425984 421009 421009 421009 421009 421009 Note: All regressions use the PPML estimator with exporter-time, country-pair and importer-time fixed effects. Border-time controls are included in columns (1) to (3). Robust standard errors, clustered at the country-pair level, are in parentheses. ∗ p < 0.10,** p < 0.05,*** p < 0.01 5.4 Dynamic effects Our last analysis focuses on the CAFTA-DR dynamics since the trade agreement entered into force, studying the phase-in effect on trade. The literature emphasizes that RTAs’ effects evolve over time. The literature notes that the RTA anticipation effects are im- portant, and RTAs have an impact on trade between partners even before entering into force. There are two possible explanations for this result. One is that once an agree- ment is announced, some firms start to adjust in anticipation of the implementation of the agreement. In addition, it is possible that the potential member countries are already relaxing some administrative measures to reduce trade costs even before the agreement enters into force. Therefore, it is important for econometric models to explicitly allow for anticipation effects, and lags, of RTAs with data that are measured at a sufficiently fine granularity in the time dimension. In the context of CAFTA-DR, as discussed in section 2, member countries already had RTAs in place and started negotiating the agreement in 2001. Figure (3) shows the dynamics of the CAFTA-DR effect on intra-regional trade, using 3-year leads and the lags up to 9 years after entering into force.18 Figure (3) shows that by 18 Thenumber of lags in our case is constrained by the fact that CAFTA-DR entered into force in 2006, and our data covers until 2015. Also, note that the agreement entered into force in 2007 for the Dominican Republic and 2009 for Costa Rica, so we build the leads and lags accordingly, being country-specific. 20 the time the agreement entered into force, there was already a significant positive effect on intra-regional trade of around 21%. From that moment, the positive effect increases, and nine years after the agreement, the total effect is above 36%. These dynamics are in line with the literature and show that the anticipation and lag effects are important to understand the total effect of the agreement. Figure 3: CAFTA-DR intra-regional dynamics Note: This figure considers total bilateral trade, both manufacturing and agriculture, of Central American countries and Dominican Republic (the US is excluded). All regression use the PPML estimator. Figure (4) shows the dynamics for exports from Central America to the US. The dy- namics describe an evolution in which the total effect after 9 years is not statistically significant, but the 2 years before and after seem to have a statistically negative effect. We see this as an adjustment period to the new trade agreement that replaced the previous preferential access that the countries in Central America had prior to CAFTA-DR. Finally, figure (5) plots the dynamics for exports from the US to Central America, showing that the effect is not statistically significant before and after CAFTA-DR entered into force. 21 Figure 4: CAFTA-DR dynamics of US imports from the rest of CAFTA-DR members Note: This figure considers total bilateral trade, both manufacturing and agriculture, of US imports from the rest of CAFTA-DR members. All regression use the PPML estimator. Figure 5: CAFTA-DR dynamics of US exports to the rest of CAFTA-DR members Note: This figure considers total bilateral trade, both manufacturing and agriculture, of US imports from the rest of CAFTA-DR members. All regression use the PPML estimator. 22 6 Concluding remarks Regional trade agreements can play a key role in boosting trade and investment in de- veloping countries. With the lack of progress in multilateral negotiations, regional trade agreements (RTAs) have been the main channel for trade liberalization in the past few decades. In Central America, CAFTA-DR is the deepest agreement between regional members and the US. We study the trade effects of CAFTA-DR across different dimen- sions. The agreement was conceived not only as an RTA with the United States but also among the Central American countries and the Dominican Republic to promote intra- regional trade. This makes it an interesting case to study the heterogeneity in the effects of RTAs, given that it is both an intra-regional and extra-regional agreement. Entering into force between 2006 and 2009, it is one of the most comprehensive regional integra- tion agreements in LAC and is the latest effort for regional integration. Recent studies on the heterogeneous impact of RTAs underscore the risks of directly applying findings from studies on developed economies to developing countries. The robust empirical support for the positive average effect of RTAs masks large degrees of heterogeneity across and within agreements. In fact, Baier et al. (2019) find that nearly half of the RTAs included in their study have either statistically insignificant or nega- tive impacts on trade flows. Furthermore, the majority of heterogeneity occurs within RTAs (rather than across different RTAs), with asymmetric effects within pairs (on ex- ports vs. imports) also playing an important role. In our analysis of CAFTA-DR, we focus on obtaining very detailed estimations to understand the heterogeneous effects across countries. The are only two previous empirical studies on the specific effects of CAFTA-DR that we are aware of. The most recent one of these is done by Rojas Rodríguez and Matschke (2023). The main methodological and empirical differences in our study are that we stick to the theoretical structural gravity model and consider all intra-national trade flows for our countries of interest in Central America. We see the omission of these flows as a serious caveat when the main objects of interest are the country-specific estimates of CAFTA-DR. We see our results as an important contribution to the literature. We focus on the country and country-pair effects of a specific trade agreement. When comparing our 23 results to previous studies of CAFTA-DR, we prove the importance of including not only domestic trade flows in general but including those of the countries of interest. Future studies might find it interesting to continue the analysis after 2015, once the data becomes available to estimate the effect of CAFTA-DR until the COVID-19 pandemic. 24 References Anderson, J. E. (1979). A Theoretical Foundation for the Gravity Equation. The American Economic Review, 69(1):106–116. Anderson, J. E. and van Wincoop, E. (2003). Gravity with Gravitas: A Solution to the Border Puzzle. American Economic Review, 93(1):170–192. Baek, Y. and Hayakawa, K. (2022). Fixed Costs in Exporting and Investing. Baier, S. L. and Bergstrand, J. H. (2007). Do free trade agreements actually increase mem- bers’ international trade? Journal of International Economics, 71(1):72–95. Baier, S. L., Yotov, Y. V., and Zylkin, T. (2019). On the widely differing effects of free trade agreements: Lessons from twenty years of trade integration. Journal of International Economics, 116:206–226. Bergstrand, J. H., Larch, M., and Yotov, Y. V. (2015). Economic integration agreements, border effects, and distance elasticities in the gravity equation. European Economic Re- view, 78:307–327. Borin, A., Mancini, M., and Taglioni, D. (2021). Measuring Exposure to Risk in Global Value Chains. Working Paper, World Bank, Washington, DC. Caliendo, L. and Parro, F. (2015). Estimates of the Trade and Welfare Effects of NAFTA. The Review of Economic Studies, 82(1 (290)):1–44. Carrillo-Tudela, C. and Li, C. A. (2004). Trade Blocks and the Gravity Model: Evidence from Latin American Countries. Journal of Economic Integration, 19(4):667–689. Dai, M., Yotov, Y. V., and Zylkin, T. (2014). On the trade-diversion effects of free trade agreements. Economics Letters, 122(2):321–325. Eaton, J. and Kortum, S. (2002). Technology, Geography, and Trade. Econometrica, 70(5):1741–1779. Egger, P. H., Larch, M., and Yotov, Y. V. (2022). Gravity Estimations with Interval Data: Revisiting the Impact of Free Trade Agreements. Economica, 89(353):44–61. 25 El Dahrawy Sánchez-Albornoz, A. and Timini, J. (2021). Trade agreements and Latin American trade (creation and diversion) and welfare. The World Economy, 44(7):2004– 2040. Felbermayr, G., Larch, M., Yalcin, E., and Yoto, Y. (2022). On the Heterogenous Trade and Welfare Effects of GATT/WTO Membership. FIW Working Paper series 195, FIW. Franco-Bedoya, S. (2023). Measuring Globalization When It Is Needed the Most : A Long-Run Analysis. Koehler-Geib, F., Sanchez, S. M., Calvo-Gonzalez, O., Kouame, A. T., Lopez, J. H., Jaramillo, C. F., Lachy Castillo, D. M., Ulku, H., Thorburn, C. W., and Vidal, E. E. (2014a). Costa Rica - Five years after CAFTA-DR assessing early results for the Costa Rican economy. World Bank. Koehler-Geib, F., Sanchez, S. M., Calvo-Gonzalez, O., Kouame, A. T., Lopez, J. H., Jaramillo, C. F., Lachy Castillo, D. M., Ulku, H., Thorburn, C. W., and Vidal, E. E. (2014b). Costa Rica - Five years after CAFTA-DR assessing early results for the Costa Rican economy (Vol. 2). World Bank. Kohl, T. (2014). Do we really know that trade agreements increase trade? Review of World Economics / Weltwirtschaftliches Archiv, 150(3):443–469. Larch, M. and Yotov, Y. (2023). Estimating the Effects of Trade Agreements: Lessons From 60 Years of Methods and Data. School of Economics Working Paper Series. Lenzen, M., Kanemoto, K., Moran, D., and Geschke, A. (2012). Mapping the Structure of the World Economy. Environmental Science & Technology, 46(15):8374–8381. Lenzen, M., Moran, D., Kanemoto, K., and Geschke, A. (2013). Building Eora: A Global Multi-Region Input–Output Database at High Country and Sector Resolution. Economic Systems Research, 25(1):20–49. Mattoo, A., Rocha, N., and Ruta, M. (2020). Handbook of Deep Trade Agreements. Washing- ton, DC: World Bank. Nowak-Lehmann, F. and Martinez-Zarzoso, I. (2003). Augmented Gravity Model: An 26 Empirical Application to Mercosur-European Union Trade Flows. Journal of Applied Economics: Vol 6, No 2. Rojas Rodríguez, J. R. and Matschke, X. (2023). The CAFTA-DR Free Trade Agreement — Analyzing its effects in a modern gravity framework. International Economics and Economic Policy, 20(1):27–93. Santos Silva, J. M. C. and Tenreyro, S. (2006). The Log of Gravity. Tinbergen, J. (1963). Shaping the world economy. The International Executive, 5(1):27–30. Urbatsch, R. (2013). A Referendum on Trade Theory: Voting on Free Trade in Costa Rica. International Organization, 67(1):197–214. Van Patten, D. and Méndez, E. (2022). Voting on a Trade Agreement: Firm Networks and Attitudes Toward Openness. World-Bank (2005). DR-CAFTA: challenges and opportunities for Central America. Yotov, Y. V. (2022). On the role of domestic trade flows for estimating the gravity model of trade. Contemporary Economic Policy, page coep.12567. 27 Appendix A GVC data from Borin et al. (2021) As explained by Borin et al. (2021), “GVC-related trade” measures the value of goods and services exported by a sector or a country that cross more than one border. The difference between gross trade and GVC-related trade is defined as "traditional trade", i.e. the value of goods and services that cross just one border. “GVC-related trade” presents two desirable features: (i) once expressed as a share of gross trade, it is bounded between 0 and 1; (ii) it is additive at any level of aggregation/disaggregation of trade flows; thus, data can be summed at any level – total country exports/world exports/world sector exports/country groups and so on – in order to obtain the proper GVC participation measures at the desired level of aggregation “GVC-related trade” is always traced in the exporting sector. The overall “GVC-related trade” encompasses three different types of GVC linkages. • "Pure backward GVC related-trade": value of goods and services produced with imported inputs and exported by the sector to the final destination market. The exporting sector is engaged in GVC activities at the end of the chain. • "Pure forward GVC related-trade": value-added in goods and services entirely gen- erated within the domestic chains – without any border crossing – exported by the sector to partners which, in turn, re-exports it to other markets. The exporting sector is engaged in GVC activities at the origin of the chain. • "Two-sided GVC related-trade": value of goods and services produced with im- ported inputs, exported by the sector to partners which, in turn, re-exports it to other markets. The exporting sector is located in a position of the chain that is more central. 28