Policy Research Working Paper 10795 Barrier or Opportunity? How Trade Regulations Shape Colombian Firms’ Export Strategies Samuel Rosenow International Finance Corporation June 2024 Policy Research Working Paper 10795 Abstract Firms increasingly must contend with trade regulations to measures and tariffs play a minor role. The technical barriers access foreign markets. This paper quantifies the relative to trade and quantity measures reallocate trade from small importance of trade regulations and their heterogeneous to big firms. The same mechanism benefits firms partici- effects for Colombian firms exporting to Latin America pating in global value chains. However, quantity controls between 2007 and 2017, focusing on specific types and make it more likely that big firms will leave export markets channels. Using panel evidence from a firm-level grav- to the benefit of smaller ones. The results control for the ity model with a difference-in-differences identification endogeneity of trade regulations and are robust to the use strategy, technical barriers to trade and quantity control of different samples and measures of firm size. measures both decrease trade on average. Other non-tariff This paper is a product of the International Finance Corporation. It is part of a larger effort by the World Bank Group 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 author may be contacted at srosenow@ifc.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 Barrier or Opportunity? How Trade Regulations Shape Colombian Firms’ Export Strategies* Samuel Rosenow International Finance Corporation, World Bank Group World Trade Institute, University of Bern JEL Classification: F13, F14, L11, O19. Keywords: Trade Policy, Colombia, Non-Tariff Measures, Firm Heterogeneity *I thank Angella Montfaucin, Ana Fernandes, Miriam Manchin, Joseph Francois, Mahdi Ghodsi, Massimiliano Cali, Ralf Peters, Doug Nelson and participants in the Galbino project summer school for their comments. The findings, interpretations, and conclusions expressed in this paper are entirely those of the author. They do not necessarily represent the views of the World Bank or its affiliated organizations, or those of the Executive Directors of the World Bank, their Managements, or the governments they represent. All errors are mine. 1 Introduction Market access conditions attract policy and academic attention.1 While tariffs have declined over time, trade regulations matter increasingly for firms wishing to access international markets (UNCTAD and World Bank, 2019). Governments impose these regulations to establish the standards and procedures a product must comply with to be sold in their market. Labeling requirements and safety certification are telling examples, assuring safety and product quality for consumers and encouraging trade. However, trade regulations can also increase exporters’ compliance and thus production costs since additional investment in technology and processes may be required. This implies that trade regulations may both reduce supply and increase demand for tradables. Firms also face uneven compliance costs of trade regulations, mostly due to market entry costs (Bernard et al., 2012; Melitz and Redding, 2015).2 Fixed market entry costs imply that the most productive companies are most likely to enter and thrive in export markets (Melitz, 2003; Bernard et al., 2011). They are best equipped to deal with diverse regulations, stan dards, testing, and certification procedures across sales markets. Taken together, the overall impact of trade regulations on trade and its distribution across firms is uncertain. 1In the academic literature, market access is commonly associated with export diversification and firm competitiveness, important precursors to economic development. For references regarding the relationship between market access and export diversification, see, for example, Fugazza and Nicita (2013). For effects of market access for firm capabilities, see Falciola et al. (2020). For important reviews of this policy debate, see Nicita and Melo (2018). 2In general, firms face three types of costs. First, firms incur product-specific production costs that reflect their capabilities. Second, firms incur variable trade costs: shipping costs in the form of iceberg trade costs and ad-valorem tariffs. These costs vary with sales, but do not depend on the exporter’s scope. Both production costs and variable trade costs deter exports at all margins. Third, firms incur fixed exporting costs by product and destination market, reflecting the compliance costs of trade regulations. Thus, the market access cost schedule varies by firm-product-destination. 2 This paper quantifies the heterogeneous effects of Latin America’s trade regulations for Colombia’s exporting firms, concentrating on specific types and channels through which they operate: firms’ value of exports and their probability of participating in and exiting from export markets. There is a focus on Latin America and the Caribbean (LAC) as the region has become Colombia’s most important export market. This, coupled with the unique data of trade regulations and other non-tariff measures (NTMs) in LAC, provides an ideal setup. We use a firm-level gravity model with a difference-in-differences research design that identifies heterogeneity in treatments across NTM types and channels. Our econometric approach accounts for selection and omitted variable bias as well as endogeneity in the timing of NTMs with respect to trade flows into destination markets. Moreover, we test for the robustness of results with stringent fixed effects, different measures of firm size and alternative samples to rule out endogeneity concerns. The impact of trade regulations varies greatly across types and channels. Of all NTM types considered, only technical barriers to trade (TBTs) — a standard-like measure — and quantity controls explain changes in Colombian firm-level exports. On average, TBTs undermine Colombia’s firm- level exports behavior across all three channels considered. These results are consistent with a model in which trade regulations increase firms’ compliance costs relative to consumer benefits. The effect is even more pronounced for quantity control measures which undermine Colombian firms’ exports and their likelihood of remaining in markets. Intriguingly, new quantity controls in product-destination markets lower, on average, firms’ exit rates. Looking at the heterogeneous effects of NTMs, we show that this effect is driven by small firms that benefit from quantity controls. Conversely, big firms are more likely to leave export markets when faced with new quantity controls. We argue that quantity control measures constrain big firms 3 more than small firms and we help to explain this outcome. However, a regressive effect is at play for TBTs for all three channels; big firms benefit from a dampened effect, whereas small firms see their exports and likelihood of remaining in markets significantly reduced. Similarly, firms participating in global value chains (GVC) benefit from a muted trade impact when TBTs or quantity controls are introduced in the sales market. In contrast, non-GVC firms suffer from reduced exports and likelihood of participating in markets. Our paper improves the understanding of the role of NTMs for firm- level trade along three dimensions. First, we quantify the relative importance of NTM types for firm-level export decisions. Some NTM types are by definition trade restrictive, such as import prohibitions and quotas. In contrast, the effect of standard-like measures such as sanitary and phytosanitary (SPS) or TBTs on export dynamics depends on compliance costs relative to information benefits and is, in theory, ambiguous.3 We demonstrate that TBTs and quantity control measures, on average, deter Colombian exports. Our findings are consistent with Adarov and Ghodsi (2023), Disdier et al. (2008), Yousefi and Liu (2013), and Li and Beghin (2012) who find negative TBT effects on trade flows. Second, we explore the channels through which NTM effects operate. To that end, we distinguish three firm-level export decisions: the value of exports to product-destination markets (intensive margin); the probability of participating in them (extensive margin I) and exit from them (extensive margin II). While we document that TBTs and quantity control measures undermine all three margins, firms mostly react by reducing existing exports. This finding challenges Bao and Qiu (2012) who highlight positive firm-level responses at the intensive margin. At the same time, we discover that firms react to additional types of trade regulations at the extensive margin. In addition to TBTs and quantity controls, increases in tariffs and SPS measures in product-destinations make it less likely that Colombian firms 3See Ronen (2017) for the intuition on trade promoting effects of NTMs. Beghin et al. (2015) find that almost 40% of product lines affected by TBT measures yield negative NTM ad valorem equivalents (AVEs), suggesting a net trade-promoting effect of these measures. 4 participate in these markets. This pattern could reflect exporters’ inability to pay the additional compliance costs and trade diversion to less costly product-destination markets. Our findings are consistent with firm-level evidence from France (Fontagné et al., 2015; Fontagné and Orefice, 2018). Third, we extend the analysis of heterogeneity in the effects of NTM types across firm characteristics. Our results align with the predominant finding of the empirical literature, that small firms suffer disproportionate impacts of TBTs (e.g., Arkolakis, 2010; Curzi, 2020; Asprilla et al., 2019; Macedoni and Weinberger, 2022). Moreover, we show that GVC firms, defined as those that both import and export, benefit from the same muted trade effects of TBTs as big firms, even though the two firm characteristics differ substantially in definition. However, we also marshal novel evidence that quantity control measures make it less likely that small firms will exit export markets, with the opposite effect occurring for big firms. We argue that the quantity controls are more binding for big firms, increasing their likelihood of leaving markets and, conversely, helping small firms take over their space. Superior data and methods underpin these three contributions. We combine a panel of Colombian firm-level trade data from 2007 to 2017 with time-varying information on NTMs in regional export markets. Employing a firm-level gravity model with a difference-in-differences research design, we estimate the relative importance of NTM types controlling for unobserved confounders at the firm-product-destination level. In doing so, we address empirical limitations in the literature. This concerns, on the one hand, the use of cross-sectional data to estimate NTM impacts (e.g., Bratt, 2017; Fugazza et al., 2018). On the other hand, many studies evaluate NTMs in isolation or bundled (Disdier and Marette, 2010 or Fontagné et al., 2015), not their joint or relative effects. Moreover, drawing from Kee and Nicita (2022), we also account for the endogeneity of NTMs by predicting NTM selection based on that in neighboring countries. This approach is suitable in our context, given the cultural and legal proximity of Colombia’s regional destination markets and their similarity in introducing trade regulations. 5 The paper is structured as follows. Section 2 explains our data sources and procedures to clean them. Section 3 documents stylized facts about Colombia’s exporting firms and the market access conditions they face. Section 4 explains our empirical identification strategy, defines key variables, and addresses concerns to identify how market access conditions affect Colombian firms trading with the region. The results and robustness tests are presented in Section 5. Sections 5.5 and 5.6 explore heterogeneous NTM effects based on firms’ size and participation in global value chains, respectively. Finally, Section 6 concludes and discusses the implications of our results. 2 Data The empirical investigation is based on three distinct datasets, all covering the period 2007–2017. The first dataset contains information on export and import transactions collected by the Colombian customs. The second provides data on NTM types applied by Latin American importing countries. The third contains bilateral applied ad-valorem tariffs of Latin American importing countries. 2.1 Firm-Level Customs Data We use transaction-level export and import data from Colombia’s customs agency DIAN. The data is part of the expansion to the Exporter Dynamics Database, as described in Fernandes et al. (2016). The data cover the universe of exporting firms in all sectors at the exporter-HS 6-digit product- destination-year level and includes seven variables: country of origin, exporting firm identifier, country of destination, HS 6-digit product, export value, export quantity, and year. Information on export and import values is expressed in US dollars and is FOB (free on board). We processed the raw dataset to a series of cleaning procedures, as detailed in Cebeci and Fernandes (2015). To merge with the tariff variable described below, we convert the HS 6 product nomenclatures of 2007, 2012, and 2017 to the HS combined version. Moreover, we exclude all firm observations from the mining sector (HS chapter 25–27). That helps to avoid potential bias from Latin America’s commodity price cycles during 2007–2017. Thus, our sample frame contains all Colombian exporting firms between 2007 and 2017, except those from the mining sector. 2.2 NTM Data We use NTM data from UNCTAD and the Latin American Integration Association (LAIA) for its 18 core members in LAC between 2007 and 2017.4 The raw data is recorded at the reporter-national tariff line level- destination-NTM, 4-digit, and year level. We apply two steps to process the data, resulting in a novel and unexploited dataset comprising an exhaustive and updated set of trade regulations in Latin America between 2007 and 2017. First, given a change in the classification of NTMs in 2012 we reconcile two NTM classifications pertaining to the periods 2007–2011 and 2012–2017. To that end, we derive a correspondence table between the pre-2012 and post-2012 classification at the 1-digit NTM chapter level using LAIA’s data collection in both classifications in 2011 and 2012. Using the new classification as the reference, we reclassified NTMs between 2007 and 2011 at the chapter level, consistent with UNCTAD’s (2019) Mast Classification M5. As a result, we observe the number of NTMs of each type (e.g. SPS, TBT, PSI, quantity control) for all Latin American reporter-destination- products continuously in each year between 2007 and 2017. Second, we aggregate NTM data from the national tariff line level to the HS 6-digit product level. The background is that NTM data are collected at the national tariff line at 10 digits. However, since we take into account exports to various destination countries that do not harmonize national tariff line classifications, our two datasets cannot be satisfactorily merged 4Argentina, Bolivia, Brazil, Chile, Costa Rica, Colombia, Cuba, Ecuador, El Salvador, Guatemala, Honduras, Mexico, Nicaragua, Panama, Peru, Paraguay, Uruguay and the República Bolivariana de Venezuela. 7 at that level. Aggregation results in little attrition. Moving from 10- to 6-digit product classification implies a reduction of about 6% in the number of observations included in our reference sample. 2.3 Tariff Data Ad-valorem tariffs for Colombia’s 17 regional trading partners come from UNCTAD TRAINS. Measured at the importer-HS 6-digit product level between 2007 and 2017, tariffs reflect the effectively applied rate, which is defined as the lowest available tariff. If a preferential tariff exists, it is used as the effectively applied tariff. Otherwise, the Most-Favored-Nation (MFN) applied tariff is used. Thus, while we measure tariffs in terms of ad valorem price effects, NTMs are measured in count terms. The reason is that bilateral, product-specific ad valorem equivalents (AVEs) of NTMs remain scarce, despite recent progress for standard-like NTMs (Adarov and Ghodsi, 2023). 3 Stylized Facts Before diving into our empirical exercise, we present a series of stylized facts on Colombian exporting firms as well as the tariffs and NTMs they face in their regional sales markets. 3.1 Colombian Firm-Level Exports Table 1 shows the evolution of Colombian exports between 2007 and 2017 and their geographical composition in eight major regions, following the World Bank classification. Three results stand out. First, Colombian exports reached a peak in 2012 (first row) and declined from then until 2016. This decline was driven by the collapse of oil prices, which matter for 35% of Colombian exports related to minerals. Second, Latin American and Caribbean (LAC) countries have become the most important destination market for Colombian exports, surpassing North America, and, in particular, the United States, since 2012. Given the importance of regional exports 8 for Colombian firms, this provides us with sufficient external validity to explore their trade response to market access conditions in LAC. Third, the importance of East Asian and Pacific destination markets for Colombian exporters has doubled between 2007 and 2017 to 10% of total exports now. Table 1: Colombian exports (2007 base) and selected destinations Destination 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 World 1.00 1.25 1.10 1.36 1.97 2.08 2.00 1.85 1.22 1.04 1.27 East Asia & Pacific 0.05 0.04 0.05 0.08 0.06 0.09 0.11 0.13 0.11 0.08 0.10 Eastern Europe 0.01 0.01 0.01 0.01 0.00 0.00 0.00 0.00 0.01 0.01 0.01 Latin America & Caribbean 0.37 0.39 0.31 0.29 0.33 0.32 0.32 0.31 0.36 0.34 0.37 Middle East & N. Africa 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 North America 0.37 0.40 0.41 0.44 0.40 0.37 0.33 0.28 0.29 0.35 0.30 South Asia 0.00 0.00 0.02 0.02 0.01 0.02 0.05 0.05 0.02 0.01 0.01 Sub-Saharan Africa 0.00 0.00 0.00 0.01 0.01 0.01 0.00 0.01 0.01 0.01 0.00 Western Europe 0.19 0.16 0.19 0.15 0.18 0.17 0.18 0.20 0.20 0.20 0.19 Source: Author’s calculation based on the original Colombian customs data. Note: This table displays Colombian exports as a share of total exports and is normalized in 2007. Since NTM data are only available for LAC countries, the focus of the paper is on Colombian exports to LAC countries. Thus, the following tables explore how export margins have evolved, taking LAC countries as the destination. Table 2 reveals that the number of firms exporting to LAC countries followed a U-shaped pattern over the period under investigation. The number of firms decreased by about 10% during the global financial crisis of 2007–2010 and then steadily increased to reach a new peak in 2017. In relative terms, however, the importance of LAC firms remained stable at around 68% of all exporting firms over the whole period. Moreover, the importance of firms only exporting to LAC destination markets remained stable between 2007– 17. Indeed, both the number and the corresponding share of firms exporting to LAC countries exclusively wavered at around 3,500 firms or 44% of all exporting firms in Colombia during the whole period, respectively. Looking at the product diversification of Colombian firms yields three insights. First, multi-product firms dominate Colombian exporters. They make up 91% of all exporting firms, a share which remained remarkably stable between 2007–17. Second, Table 3 reveals that Colombian exporters increased the diversification of their product portfolio from an average of 9 Table 2: Number of exporting firms and destination markets Destination 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 Number of exporting firms to LAC 5,615 5,815 5,696 5,098 5,255 5,520 5,508 5,450 5,604 5,612 5,738 Number of exporting firms to LAC only 3,383 3,713 3,669 3,114 3,286 3,526 3,535 3,462 3,518 3,523 3,621 Share of LAC firms in total firms 0.68 0.69 0.70 0.69 0.71 0.73 0.71 0.70 0.69 0.68 0.68 Share of LAC only firms in total firms 0.41 0.44 0.45 0.42 0.44 0.46 0.46 0.45 0.43 0.43 0.43 Source: Author’s calculation based on Colombian customs data. Note: This table displays Colombian firm-level export dynamics to Latin America and the Caribbean (LAC), measured in absolute (number of firms) and relative terms (share in total number of exporting firms). The third and fourth rows report the number and the share of firms among those exporting to some LAC country that exports exclusively to the region. 19 HS 6-digit products in 2007 to 26 products in 2017. Third, the increase in product diversification is driven by the expansion of regional export relationships, rather than extraregional ones. Put in perspective, Colombian exporting firms show a relatively high level of product diversification; Peruvian firms, for example, only export, on average, 8.5 HS 6-digit products (Fugazza et al., 2018). Table 3: Number of exported products per firm Mean Median Maximum LAC All LAC All LAC All 2007 19.5 17.4 10.0 9.0 128 131 2008 20.5 18.3 10.0 9.0 133 133 2009 22.8 20.1 11.0 10.0 180 181 2010 23.3 20.8 11.0 10.0 175 175 2011 26.0 22.9 12.0 11.0 193 195 2012 27.0 23.7 13.0 11.0 202 204 2013 29.0 25.7 13.0 11.0 207 209 2014 29.4 26.1 13.0 12.0 202 204 2015 29.6 26.0 13.0 11.0 207 208 2016 28.6 25.4 13.0 11.0 182 186 2017 26.6 23.6 12.0 11.0 153 155 Source: Author’s calculation based on Colombian customs data. Note: This table reports the number of exported HS 6-digits products by Colombian firms over time and by LAC and world destination markets. Colombian exporting firms are increasingly diversified in their sales markets. Table 4 shows that the average share of the sales market grew from 6.4 in 2007 to 7 in 2017. What is more, Colombian firms export to an additional five destination markets beyond the region, resulting in a total of 12 destination markets globally (columns 2 and 3). 10 Table 4: Number of destinations per firm Mean Median Maximum LAC All LAC All LAC All 2007 6.4 10.1 6.0 8.0 16 45 2008 6.5 10.3 6.0 8.0 16 48 2009 6.6 10.7 6.0 9.0 16 48 2010 6.7 11.2 7.0 9.0 17 51 2011 6.8 11.5 7.0 9.0 17 51 2012 6.9 11.6 7.0 9.0 17 49 2013 7.0 12.0 7.0 9.0 17 51 2014 7.0 12.3 7.0 9.0 17 61 2015 7.0 11.9 7.0 9.0 17 61 2016 7.0 11.9 7.0 9.0 17 58 2017 6.9 11.5 7.0 9.0 16 60 Source: Author’s calculation based on Colombian customs data. Note: This table reports the number of destination markets by Colombian exporters over time. 3.2 Tariffs in LAC Colombian exporters saw declining tariffs in their regional destination markets between 2007 and 2017. The applied tariff decreased steadily from 6.7% in 2007 to only 4.1% in 2017, as seen in Table 5, as Colombia ratified numerous preferential trade agreements with regional partners over the period under investigation.5 This is most evident for 2017 when tariff reductions of the Pacific Alliance between Colombia and Mexico, Peru and Chile, came into effect. At the same time, Most-Favored-Nation (MFN) tariffs negotiated at the World Trade Organization (WTO) declined only marginally from 9.1% in 2007 to 8.7% in 2017, reflecting the persistent deadlock in multilateral trade negotiations. 3.3 Non-Tariff Measures in Latin America The NTM dataset contains 6,211 regulations for Colombia’s 17 destination markets in LAC between 2007 and 2017. To study the variation of regu- 5The Pacific Alliance, signed in 2013 between Colombia and Mexico, Peru and Chile, reduced 92% of tariffs between members in 2016. Moreover, the free trade agreement (FTA) between Colombia and Costa Rica entered into force on August 1, 2016. As a result, 74% of industrial products became duty-free. 11 Table 5: Simple average tariff (applied and MFN) to Colombian exports in Latin American countries, 2007–2017 Applying country Tariff 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 ARG Applied 4.5 3.6 3.4 2.5 2.0 1.4 0.8 1.2 1.1 1.1 1.0 MFN 11.1 11.0 12.3 12.3 12.3 12.3 13.2 13.4 13.4 13.4 13.5 BOL Applied 7.9 7.9 9.6 10.3 10.3 10.3 10.6 10.6 10.6 10.6 10.5 MFN 8.7 8.7 10.4 11.1 11.1 11.1 11.5 11.4 11.4 11.5 11.7 BRA Applied 4.0 3.0 2.3 1.5 0.7 2.5 13.0 2.0 2.0 2.1 2.1 MFN 12.1 13.0 13.3 13.4 13.4 13.4 13.4 13.5 13.5 13.5 13.5 CHL Applied 5.5 5.4 5.5 5.4 5.3 5.3 5.9 5.9 0.6 5.9 0.7 MFN 6.0 6.0 6.0 6.0 6.0 6.0 6.0 6.0 6.0 6.0 6.0 CRI Applied 6.1 6.6 5.7 5.7 5.6 5.7 5.7 5.7 5.7 5.7 5.7 MFN 6.1 6.6 5.7 5.7 5.6 5.7 5.7 5.7 5.7 5.7 5.7 CUB Applied 10.6 10.6 10.6 10.6 10.6 10.6 10.2 10.2 10.2 10.2 10.2 MFN 10.6 10.6 10.6 10.6 10.6 10.6 10.2 10.2 10.2 10.2 10.2 ECU Applied 11.4 11.2 11.2 11.7 10.2 10.2 10.2 10.5 10.7 11.0 11.0 MFN 12.2 11.9 11.8 11.8 10.8 10.8 10.8 11.1 11.4 11.8 11.8 GTM Applied 5.5 5.5 5.5 4.5 4.2 4.0 3.7 3.4 3.3 3.3 3.3 MFN 5.8 5.8 5.8 5.8 5.8 5.8 5.7 5.7 5.7 5.7 5.7 HND Applied 5.9 5.8 5.9 5.9 5.9 5.8 5.8 5.8 2.1 2.1 2.1 MFN 5.9 5.9 5.9 5.9 5.9 5.8 5.8 5.8 5.8 5.8 5.8 MEX Applied 2.6 2.4 2.3 8.9 8.2 7.8 7.5 1.6 6.9 6.8 0.8 MFN 12.9 12.8 11.8 9.8 9.1 8.7 8.3 8.1 7.8 7.7 7.6 NIC Applied 5.9 5.8 5.8 5.9 5.8 5.8 5.8 5.8 5.8 5.8 5.8 MFN 5.9 5.8 5.8 5.9 5.8 5.8 5.8 5.8 5.8 5.8 5.8 PAN Applied 7.5 7.5 7.4 7.2 7.1 7.1 7.0 7.0 7.0 7.0 7.0 MFN 7.5 7.5 7.4 7.2 7.1 7.1 7.0 7.0 7.0 7.0 7.0 PER Applied 9.3 6.1 5.5 5.5 4.0 4.0 3.8 3.8 3.0 3.0 0.9 MFN 9.4 6.1 5.5 5.5 4.1 4.1 3.9 3.9 3.0 3.0 3.0 PRY Applied 6.3 9.5 4.8 4.2 3.6 3.0 2.2 1.7 1.1 1.0 0.9 MFN 10.3 10.3 10.3 10.3 10.3 10.3 10.2 10.1 10.1 10.1 10.0 SLV Applied 5.7 5.8 5.7 5.8 5.8 4.5 4.2 3.9 3.8 3.8 3.4 MFN 6.1 6.2 6.1 6.1 6.1 6.1 6.1 6.1 6.1 6.1 6.1 URY Applied 4.8 4.2 3.6 3.0 2.5 2.0 1.5 1.1 0.8 0.7 0.7 MFN 10.5 10.5 10.5 10.5 10.5 10.5 10.6 10.6 10.6 10.5 10.5 VEN Applied 10.8 10.9 10.9 10.7 10.9 10.9 10.9 7.5 4.1 3.9 4.0 MFN 13.3 13.4 13.4 13.2 13.4 13.4 13.4 13.0 13.1 12.8 13.9 LAC average Applied 6.7 6.6 6.2 6.4 6.0 5.9 6.4 5.2 4.6 4.9 4.1 MFN 9.1 8.9 9.0 8.9 8.7 8.7 8.7 8.7 8.6 8.6 8.7 Source: Author’s calculation based on UNCTAD TRAINS tariff data. Note: This table displays Colombia’s most-favored nation (MFN) and applied tariffs over time in its LAC destinations markets. MFN tariffs measure the normal non-discriminatory tariffs charged on imports from Colombia. Applied refers to the effectively applied tariff, which is defined as the lowest available tariff. If a preferential tariff exists, it will be used as the effectively applied tariff. Otherwise, the MFN applied tariff will be used. The LAC average reports Colombia’s simple average of MFN and applied import tariff rates across the 17 LAC destination markets, respectively. lations over time, we document the share of all regulations entered before vs. after 2007, the first year of data collection, and the share remained active vs. abolished in 2017, the last year of data collection. This leaves us with four margins of variation, of which all but one – regulations entered prior to 2007 that remained active in 2017 – can be exploited in our empirical strategy. Of the 6,211 regulations, 17 (0.3%) were implemented prior 12 to 2007 and abolished before 2017; 2,367 (38.1%) regulations were implemented before 2007 and we still in place in 2017; 84 (1.4%) regulations were introduced and then abolished between 2007 and 2017; and 3,743 (60.3%) regulations were introduced between 2007 and 2017 and remained active in 2017. Table 6 reports the corresponding figures for each LAC member. Ecuador and Panama stand out in the region as they introduced more than 83% of regulations between 2007 and 2017, far above the region’s average. In sum, this means that the majority of applicable regulations in 2017 were implemented after 2007. This provides us with sufficient variation to estimate the impact of NTMs on Colombian exports using an identification strategy based on the introduction and withdrawal of NTMs. Table 6: NTM turnover Entry before 2007 Entry after 2007 Abolished before 2017 Active in 2017 Abolished before 2017 Active in 2017 ARG 1.3 65.5 3.4 29.8 BOL 0.0 74.4 0.0 25.6 BRA 0.3 39.2 3.0 57.6 CHL 0.2 52.9 0.6 46.2 CRI 0.0 27.3 0.3 72.4 CUB 0.0 66.7 0.0 33.3 ECU 0.1 10.9 2.0 87.0 GTM 0.0 62.9 0.0 37.1 HND 0.0 49.3 0.0 50.7 MEX 1.7 40.5 2.7 55.1 NIC 0.0 39.9 0.0 60.1 PAN 0.0 15.1 1.0 83.9 PER 0.0 38.6 0.0 61.4 PRY 0.0 56.3 0.0 43.7 SLV 0.0 61.6 0.0 38.4 URY 0.0 43.2 0.3 56.5 VEN 0.0 59.7 1.1 39.2 Total 0.3 38.1 1.4 60.3 Source: Author’s calculation based on UNCTAD’s NTM data. Note: This table reports the turnover of NTMs for each of the 17 LAC countries. The row “Total” reports the average of the total NTM turnover in all 17 LAC countries. Next, we study the margin with the greatest importance: regulations introduced between 2007 and 2017 and active in 2017. There the composition of NTMs is heavily skewed toward technical measures: SPS and TBTs make up 40% and 52%, respectively, of all new regulations introduced be- 13 tween 2007 and 2017. Table 7 shows the number of active regulations in 2017 by country and NTM-type and its share implemented after 2007, the first year of NTM data. The majority of NTMs reported in 2017 are implemented after 2007. For the whole sample, this is the case for 53% of SPS measures, 48% of TBTs and 59% of quantity control measures. Costa Rica and Panama stand out in particular as they have implemented more than 86% of their SPS measures and TBTs since 2007. Given the prominence of technical measures, the empirical section will focus on the effect of SPS, TBTs, quantity control and PSIs measures. Table 7: Regulations implemented since 2007 and effective in 2017 SPS measures TBTs Pre-shipment Inspections Quantity control measures Other 2017 Since 2007 2017 Since 2007 2017 Since 2007 2017 Since 2007 2017 Since 2007 ARG 143 41% 265 26% 9 22% 10 40% 18 28% BOL 31 23% 37 19% 1 0% 5 60% 8 50% BRA 199 54% 491 63% 3 67% 36 50% 7 29% CHL 260 45% 194 49% 5 40% 4 50% 13 31% CRI 163 86% 115 56% 0 0% 8 75% 6 33% CUB 24 50% 40 25% 1 0% 6 33% 1 0% ECU 150 73% 802 92% 6 67% 24 83% 6 67% GTM 75 20% 71 56% 0 0% 8 50% 5 0% HND 84 50% 52 50% 0 0% 0 0% 4 75% MEX 134 45% 236 64% 3 33% 18 78% 3 0% NIC 146 61% 258 57% 0 0% 22 91% 5 80% PAN 539 95% 128 47% 11 100% 8 75% 8 0% PER 231 72% 92 39% 3 33% 1 0% 10 30% PRY 27 22% 74 49% 3 33% 15 73% 7 14% SLV 18 78% 221 33% 0 0% 5 100% 6 50% URY 114 60% 174 55% 6 33% 12 58% 8 62% VEN 58 33% 91 40% 3 33% 10 80% 12 42% Total 2,396 53% 3,341 48% 54 27% 192 59% 127 35% Source: Author’s calculation based on UNCTAD’s NTM data. Note: This table reports the number of active regulations in 2017 by country and NTM type. In addition to the number of active regulations and the relative importance of each NTM-type, another way to measure the incidence of NTMs is through their product coverage. Table 8 reports the number and share of products affected by at least one active NTM in 2017. Cuba and Argentina exhibit the highest product coverage: 98% of products that could be produced and imported at the HS 6-digit level are affected by at least one technical measure (SPS, TBT or PSI) while non-technical measures, such as price and quantity restrictions, affect at least 96% of imported products. At the other extreme, Paraguay has the lowest product coverage with 9% of products affected by at least one technical measure and less than 14 6% by non-technical measures. Table 8: Number (share) of products affected by at least one active NTM in 2017 Technical Non-technical ARG 4,479 (81) 5,205 (95) BOL 1,930 (35) 945 (17) BRA 3,960 (72) 1,221 (22) CHL 3,309 (60) 107 (2) COL 2,963 (54) 3,254 (59) CRI 1,777 (32) 322 (6) CUB 5,363 (98) 5,362 (97) ECU 2,937 (53) 2,198 (40) GTM 1,233 (22) 54 (1) HND 1,633 (30) 588 (11) MEX 1,848 (34) 837 (15) NIC 1,898 (35) 752 (14) PAN 1,717 (31) 111 (2) PER 2,045 (37) 84 (2) PRY 1,527 (28) 393 (7) SLV 1,937 (35) 88 (2) URY 2,555 (46) 367 (7) VEN 4,891 (89) 4,803 (87) Source: Author’s calculation based on UNCTAD’s NTM data. Note: The total number of products at the HS 6-digit is 5,400 (HS Combined). 3.4 Colombian Firms and Non-Tariff Measures in LAC We now study the extent to which Colombian exporters face NTMs in their regional product-destination sales markets. The focus is on technical trade regulations since they are the most frequently used type of NTM, as discussed above. To make sense of the high-dimensional data, we define three groups of exporting firms. Fully exposed firms (green bar chart) are defined as those facing at least one NTM in each product-destination market. Partially exposed firms (blue bar chart) enjoy at least one product- destination market without trade regulation. The last group is made up of firms whose exports do not face any trade regulations in product- destination markets. 15 Figure 1: Incidence of technical regulations at the firm-level Source: Author’s calculation based on Colombian customs data and UNCTAD TRAINS NTM data. Note: Zero refers to the share of Colombian firms facing no technical NTMs in any of their product-destination markets. Partial refers to the share of Colombian firms facing at least one technical NTM in one of their product-destination markets. Full refers to the share of Colombian firms facing at least one in each of their product-destination markets. Figure 1 illustrates a growing incidence of technical regulations in the product-destination markets of Colombian exporting firms. While in 2007 only 17% of Colombian exporters faced sales markets replete with trade regulations, the share grew to 30% in 2017. Conversely, the share of Colombian exporters facing zero trade regulations in their sales markets declined from 50% in 2007 to 34% in 2017.6 6The relative importance of these three groups of firms may be biased by the incidence of multi-product, multi-destination firms. However, as demonstrated by the previous analysis, the incidence of multi-product or multi-destination firms remains stable during the period of interest. 16 4 Methodology 4.1 Empirical Framework We use a firm-level structural gravity model to explain firm behavior at the intensive and extensive margins of exports as a function of changes in NTMs and tariffs. Specifically, + ˆ Rljpt ∑ β3,l I M Yf jpt = exp β1ln(1 + τijpt) + ∑ β2,l NTMlijpt l∈L l∈L 1 + β4ln(Imports f jpt ) + ω f jp + ω f pt + ωjt + ε f jpt (1) i ∈ Colombia, ∀ f ∈ { 1, . . . , 17468} , ∀ j∈ { 1, . . . , 17} , ∀ p ∈HS6 { 1, . . . , 4287} , ∀ t ∈ { 2007, . . . , 2017} , l∈ { TBT, SPS, Preshipment Control, Quantity Control, Price Control} where the subscripts f , p, j and t denote, respectively, Colombian firm, HS 6-digit product, LAC destination country, and year. i refers to Colombia, the only exporting country. We consider three dependent variables Yf pjt to capture firm-level export decisions. This requires expanding the initial dataset so that each firm- product-destination has an observation in all sample years, with a 0 export value in a year when exports by the firm-product-destination are not occurring.7 Setting up the dataset in this way for all three margins of exports 7The objective in constructing the expanded dataset is to have observations that are computationally feasible and make economic sense, i.e., that indicate plausible firm choices with the fewest assumptions possible. To build intuition on our fill-in procedure, consider an observation from the initial dataset in which firm f starts to export product p to destination j in year t. If in the expanded dataset we add an observation with a 0 export value for firm f product p destination j, in year t − 1, this implies that in year t − 1 we are allowing firm f to choose whether to export product p to destination j and the firm chooses not to do so. This seems like a plausible and not overly restrictive assumption. In contrast, expanding the initial dataset into a square matrix where every firm has an observation for every product-destination-year combination possible would retain many 17 and using stringent fixed effects allows us to exploit the panel dimension in the firms’ decision to export or exit a product-destination market as NTMs change over time. Using this expanded dataset, we define our dependent variables Yf pjt as follows: 1. Intensive margin: a continuous outcome variable Exp f pjt capturing exports by Colombian firm f of product p to LAC destination country j in year t and equal to 0 otherwise; 2. Extensive margin I: a firm export participation dummy Participation f pjt equal to 1 in year t if Colombian firm f exports a positive value of products to destination j, and equal to 0 otherwise. Participation reflects both the creation and the continuation of a trade relationship; 3. Extensive margin II: a dummy for firm exit from a product-destination market Exit f pjt equal to 1 if Colombian firm f does not export product p to destination j in year t but did so in year t − 1 and equal to 0 if the firm exported product p to destination j in year t − 1 and continues to do so in year t.8 Our variable of interest is NTMl ijpt , a count variable that indicates the num- ber of NTMs of type l applied on imports from Colombia i of product p by country j at time t. The set L includes the most frequent NTM 1-digit zeros – as most firms tend to export a single product to a single destination – and would impose computational challenges. Similarly, less data-intensive fill-in procedures lack economic sense. Consider a second example to that end: firm f exports products p1 and p2 at some point during the sample period in the initial data set. If in the expanded dataset we add observations with 0 export values for firm f for all other possible products in any year, this implies that in any year we are allowing firm f to choose whether to export any possible product. This is an implausible assumption because other products may be completely different from what the firm’s capabilities in terms of its technology and other inputs allow it to produce. 8Firms that export to a product-destination market in every year and thus have a 0 in the dependent variable in every year will be effectively dropped from the estimating sample given the specific fixed effects (firm-product-destination) included in our specifications. Moreover, if a firm has positive exports to a product-destination market only in the last year of the sample (and no exports to that product-destination market in previous years of the sample) it is not included in the exit analysis. 18 chapters reported in the data: SPS measures, TBTs, PSIs, quantity control measures, and price control measures. We include a variety of control variables in equation 1. First, we isolate the impact of tariffs from NTMs. ln(1 + τijpt) controls for the applied ad valorem tariff rate imposed by destination markets j on imports from Colombia i in HS 6-digit product p in year t. To include zero tariffs, we add 1 before taking the natural log. Second, we control for product-level demand conditions in destination markets, such as time-varying business cycles or import-demand shocks. These are proxied by the natural log of Imports f pjt, which represents imports by destination market j of product p in year t. Note that we subtract firm f ’s exports from these imports to avoid correlation by construction for large firms, hence the subscript f in Imports f pjt. Moreover, we exploit the granularity of the data to control for three types of interacted fixed effects. This strongly reduces concerns about alternative explanations. First, firm-destination-HS 6-digit product fixed effects ω f jp account for unobserved heterogeneity at the panel level and allow us to identify our coefficients of interest based on within firm-destination-product changes in exports as NTMs enter into force over the sample period. Second, firm-product-year fixed effects ω f pt capture product-specific productivity differences of Colombian firms and supply shocks. Third, destination-year fixed effects ωjt account for time-variant demand and macroeconomic shocks in destination markets.9 Taken together, these fixed effects control for multilateral resistance terms in a structural gravity equation (Baier and Bergstrand, 2007; Felbermayr et al., 2020). Our final control variables are the Inverse Mills Ratios of each NTM type l I M ˆ Rljpt, discussed in detail in the next section. Our coefficients of interest β2,l measure the percentage change in export value, probability of participation, and exit following the introduction of an NTM of type l for Colombian exporters to a destination-product market relative to product-destination markets without NTMs. 9Note that we cannot control for HS 6-digit product-destination-year fixed effects as they would absorb all variation in our variable of interest NTMl pjt. 19 We estimate all three margins of exports in equation 1 with the Pseudo Poisson Maximum Likelihood (PPML) estimator. It represents the standard in the trade literature for outcome variables with a high number of zeros and the presence of nonnegative data (Santos Silva and Tenreyro, 2006; Correia et al., 2020). Inference is based on Huber-White robust standard errors, clustered at the HS6 product-destination level to control for autocorrelation and heteroscedasticity. 4.2 Identification Issues Estimating equation 1 faces two econometric challenges. First, our sample may suffer from selection bias since not all Colombian firms export products to destination markets in every year of our sample. Indeed, the imposition of NTMs in destination markets may prevent Colombian firms from exporting products over time, meaning that these observations would not appear in our dataset. Thus, using a logarithmic dependent variable excludes cases with zero exports, generating selection bias in favor of bigger sales markets. To address selection bias at the intensive margin, we use an exponential formula in equation 1, include zeros to the export data along the time dimension and we estimate it with PPML. Second, omitted variable bias may occur since NTMs may correlate with the unobserved component of a Colombian firm’s exports ε f pjt . This concern pertains especially to protectionist trade policy measures that might be imposed to control Colombia’s exports. Specifically, if larger Colombian exports induce authorities in, say, Argentina, to impose more protectionist measures, then our NTM coefficients could be biased because the error term correlates with the NTM regressors. To account for omitted variable bias, we follow Kee and Nicita (2022) and estimate a probit control function for each type l of NTM: NTM ljpt = Φ ∑ γl NTM l jpt + εlpjt ′ ′ (2) l′ ∈L 20 where NTM ljpt indicates a dummy outcome variable equal to 1 if destination j has an NTM of type l on product p in year t and 0 otherwise; Φ the cumulative distribution function of the standard distribution; γl′ evaluates to which extent a country is more likely to implement an NTM of type l if its three closest countries implement an NTM of type l’; and NTM l ′ jpt the average NTM of type l’ of the three closest countries of destination j.10 While the introduction of NTMs in Argentina could be correlated with that of Argentina’s neighbors due to regional and cultural proximity, Colombia’s exports to Argentina of a particular product should not influence the trade policies of Argentina’s closes countries for the same product. Put differently, we argue that Argentina’s political economy motives against Colombian imports will not directly shape trade policies among Argentina’s neighbors for those Colombian products. As a result, these instruments meet the criteria for the exclusion restriction. Table A1 shows positive estimates of the probit control function, suggesting that the neighbors’ imposition of NTMs predicts the country’s own implementation of NTMs. Next, we obtain γˆ l ′ to compute the Inverse Mills Ratio (IMR) for each type of NTM-type l: ) ˆ l ′ NTM l ′ jpt Φ ∑l′ ∈ L γ ˆR IM = (3) ∈ l jpt ) ϕ ∑l′ L ˆ l ′ NTMl′ jpt γ where ϕ represents the standard normal density function. IMR captures the hazard of non-selection: if the IMR for an NTM is higher, the importing country is less likely to have implemented the NTM, considering the NTMs imposed by its neighbors. We include the IMR vector as a control variable in equation 1 to account for the correlation that the importing country enforces an NTM despite a high IMR, our endogeneity concern. This ensures that we compare treated and untreated units that have similar chances of being treated, based on the actions of neighboring countries. Our prefer- 10To define closeness, we rely on bilateral distance between countries, weighted by population of main cities, provided by Mayer and Zignago (2011). 21 ence over an instrumental variable approach stems from the binary nature of the endogenous NTM variable and the utilization of a PPML estimator with high-dimensional fixed effects. In addition, we conduct a host of robustness tests to rule out omitted variable bias. First, we lag both tariffs and NTM variables by one year, following Fugazza et al. (2018). Our expectation is that the use of an NTM in a previous year is exogenous to firms’ exports in the current year. Second, following Fernandes et al. (2021), we exclude from the sample the largest exporting firms in Colombia which may influence the imposition of NTMs in destination markets. The largest exporting firms are identified for each destination market and year as those in the top 1% of the distribution of total firm exports. Third, and in the same vein, we exclude from the sample the HS 2-digit sectors with the highest degree of export concentration across firms. Concentrated sectors are defined as those for which the largest 1% of firms are responsible for more than 50% of sector exports in at least one sample year. Fourth, we run placebo tests to show that future NTMs do not have an effect on current firm-level export values, participation or entry. To that end, we evaluate the placebo treatment of leading NTM types and tariffs by one year. In the Appendix B, we conduct additional robustness tests to show that NTM overlap and collinearity are not a concern for our identification strategy. 5 Results 5.1 Summary Statistics Table 9 displays the summary statistics of our variables. Panel A displays the sample for the intensive margin of trade, i.e., for all observations in a firm, product, destination, and year combination. Conversely, Panel B reports the statistics for both extensive margins I and II: probability of export participation and export exits. In our sample, Colombian firms export, on average, around US $151,000 22 Table 9: Summary Statistics Variable N Mean sd Min Max Panel A - Intensive Margin of Exports Exports (in USD) f ,p,j,t 253,212 151,748 1,936,183 0.0 298,207,937 Log (Tariffi,j,p,t) 253,212 1.224 1.27 0.0 4.4 SPSi,j,p,t 253,212 1.910 3.02 0.0 69.0 TBTi,j,p,t 253,212 3.577 3.60 0.0 31.0 Pre-Shipmenti,j,p,t 253,212 0.117 0.33 0.0 3.0 Quantity Controli,j,p,t 253,212 0.493 0.79 0.0 5.0 Price Controli,j,p,t 253,212 0.305 0.49 0.0 5.0 Log (Import Demand f ,p,j,t) 253,212 10.593 5.55 -2.7 20.8 Firm size 1: Exports per product (in USD) f ,t 39,365 679,199 6,211,960 0.0 327,057,215 Firm size 2: Exports per sector (in USD) f ,t 28,786 3,421,507 19,717,287 0.0 399,821,436 Firm size 3: Total exports (in USD) f ,t 24,247 6,850,923 25,315,153 0.1 399,821,877 GVC dummy: firm exports and imports f ,t 253,212 0.244 0.43 0.0 1.0 Panel B - Extensive Margin of Exports Prob (Export Participation f ,p,j,t) 253,212 0.570 0.50 0.0 1.0 Prob (Export Exit f ,p,j,t) 17,221 0.551 0.50 0.0 1.0 Log (Tariffi,j,p,t) 17,221 1.167 1.26 0.0 4.0 SPSi,j,p,t 17,221 1.998 2.95 0.0 29.0 TBTi,j,p,t 17,221 3.951 3.65 0.0 28.0 Pre-Shipmenti,j,p,t 17,221 0.151 0.37 0.0 3.0 Quantity Controli,j,p,t 17,221 0.528 0.79 0.0 5.0 Price Controli,j,p,t 17,221 0.328 0.50 0.0 5.0 Log (Import Demand f ,p,j,t) 17,221 11.171 5.19 -0.4 19.8 Firm size 1: Exports per product (in USD) f ,t 5,239 659,912 2,991,801 0.1 42,844,647 Firm size 2: Exports per sector (in USD) f ,t 3,546 1,080,343 4,747,268 0.1 48,113,669 Firm size 3: Total exports (in USD) f ,t 2,762 1,496,305 5,610,819 0.1 48,113,669 GVC dummy: firm exports and imports f ,t 17,221 0.264 0.44 0.0 1.0 This table presents descriptive statistics for the three dependent variables between 2007 and 2017. The upper panel presents the sample used in the estimations of the intensive margin: export value. The lower panel presents results used in the estimations of the extensive margin I and II: probability of export participation and export exits. The table also includes statistics on the incidence of NTM types and tariffs, control variables and firm size measures. per product and destination market. However, this average masks a great dispersion of firm-level exports as seen by the massive standard deviation.11 Panel B shows that export participations happen in 57% of possible cases. That is, Colombian firms export, on average, into 57% of all product- destination-year combinations in our sample for the extensive margin I. This means export participation is a likely event, on average. Similarly, the 11While the underlying export values are measured in nominal US dollars, note that our regressions have destination-year fixed effects among others, so our dependent variables can be interpreted in real, not nominal, terms. 23 probability of firms leaving product-destination markets occurs in 55.1% of possible cases. The incidence of NTMs varies by type. In our sample, TBTs are ubiquitous and occur, on average, 3.5 times in firm’s export relationships (product- destination-year cells) at the intensive margin and 3.9 times at the extensive margin. SPS and quantity control measures are present, on aver- age, 1.9 and 0.5 times of firms’ export relationships. On average, Colombian firms face 1.2% of applied ad valorem tariffs in regional destination markets in our sample. Our three proxies for firm size – explained in Section 5.5 – indicate that firms that already export (intensive margin) are bigger than those with intermittent exports (extensive margin II). Colombian firms participating in global value chains (GVC), defined as those that both import and export goods, make up 24% in our sample.12 5.2 Intensive Margin of Trade Column (1) of Table 10 evaluates the principal component of all five NTM types. Column (2) provides our baseline results with all NTM types and control variables. The remaining columns provide robustness tests. Specifically, Column (3) lags all NTM types and tariff variables by one year. Column (4) excludes large Colombian exporting firms that may influence the imposition of NTMs in destination markets. To the same effect, Column (5) excludes HS 2-digit sectors with the highest degree of export concentration across Colombian firms. Finally, in Column (6) we run placebo tests to show that future NTMs do not have an effects on current firm-level export values. All columns use the battery of stringent fixed effects α f jp, δ f pt and ηjt that allow us to interpret coefficients akin to a difference-in-differences setup. For expositional clarity, the coefficients of the IMR are not reported. For comparison purposes, we report standardized coefficients in all tables by normalizing the regressors to have mean zero and unit standard deviation. 12This GVC definition follows World Bank (2020). 24 The impact of market access conditions on the intensive margin of exports is, on average, negative. Column (1) suggests that an increase of one standard deviation of the principal component of NTMs is associated, on average, with a decrease of 19% in firm exports.13 The impact of market access conditions varies greatly across NTM types, however. Column (2) shows that quantity controls undermine firm-level exports the most, followed by TBTs. Specifically, an increase of one standard deviation in quantity control in Colombia’s destination markets associated with, on average, a 43% decrease of firm-level exports. In contrast, the introduction of new TBT measures in Colombia’s destination markets translates into a 9% reduction of its firm-level exports. Tariffs, SPS, quantity and price control do not have, on average, an effect on Colombian firms’ exports. As for control variables, their estimated coefficients yield the expected significance and sign; increases in import demand at destination increases Colombian firm-level exports. Our robustness tests confirm the relative importance of quantity controls over TBTs at the intensive margin of exports. Columns (3) show that the results obtained with lagged NTMs and tariffs are consistent with our baseline estimates. Moreover, excluding the largest firms and concentrated sectors in Columns (4) and (5) maintains the consistency of our findings, providing additional proof that reverse causality is not driving our results. Consistent with our baseline results, all three robustness tests highlight the importance of quantity controls, followed by TBTs, to discourage Colombian firm-level exports. Finally, Column (6) provides evidence that future introductions of NTMs are not correlated with existing firm-level exports. Overall, we note the extremely high explanatory power of our model, as evidenced by the adjusted R2 of 0.95. 5.3 Extensive Margin of Trade I Models with heterogeneous firms predict a negative effect of market access conditions on the extensive margin of trade, as measured by firms’ export 13exp(-0.2179) -1 25 participation. The empirical results shown in Table 11 are in line with theoretical predictions. Column (1) suggests that an increase of one standard deviation of the principal component of NTMs is associated, on average, with a decrease of 13% of firms’ probability of participating in exports. Considering the 57% unconditional probability of export participation, the estimated economic impact of NTMs is thus sizable. Various NTM types influence firms’ probability of export participation. While coefficients of tariffs, quantity controls, TBTs, and SPS are consistently negative and statistically coefficients across all specifications, quantity controls and tariffs undermine firms’ export participation probabilities the most. However, the magnitude of coefficients is low compared to those at the intensive margin. Moreover, our robustness tests in Columns (3)–(5) highlight the relative importance of tariffs and quantity controls in lowering firms’ export participation. In demonstrating consistency with our baseline results, they marshal evidence that reserve causality is not driving our results. Finally, demand conditions at destination positively affect the extensive margin in all specifications. This implies that the likelihood of exporting is larger in bigger destination markets. 5.4 Extensive Margin of Trade II NTMs may also force Colombian firms to stop exporting to a product- destination market. To explore this hypothesis, we evaluate the roles that tariffs and specific NTM types play for firms’ probability of export exits, the second extensive margin of interest. We find that, on average, NTMs increase firms’ probability to stop exporting. Column (1) in Table 12 suggests that a one standard deviation increase of the principal component of NTMs is associated, on average, with a increase of 22% in firms’ exit probability. This implies a big economic impact, considering the 55% unconditional probability of stopping exporting. Our results show that new TBTs in product-destination markets lead to higher exit rates from those markets. Column (2) suggests that the intro- 26 duction of TBT measures in Colombia’s product-destination markets translates into a 22% increase in firms’ likelihood of leaving export markets. Other NTM types and tariffs sustain no effect. Control variables have the expected sign and statistical significance. Intriguingly, however, we find that that new quantity controls in destination-product markets, on average, lower the exit rates of Colombian firms from those markets. The effect is consistently negative and statistically significant across all specifications in Columns (2)–(5). A potential rationale for this finding is that we are picking up average firm-level responses across the entire firm-size distribution. Indeed, we will show in the next section 5.5 that small firms benefit from quantity controls to the detriment of bigger firms, which are more constrained by the imposition of quantity controls in product-destination markets. Our robustness tests in Columns (3)–(5) confirm the importance of TBTs in increasing firms’ exit rates, maintaining similar coefficients as in our baseline specification. Similarly, the robustness tests confirm that new quantity controls reduce exit rates for Colombian firms. This provides additional evidence against reverse causality, i.e., showing that the effect of quantity controls for firms’ exit rates is not the result of pressure by influential domestic firms or sectors to design NTMs to their advantage. Finally, our placebo test in Column (6) also shows that no future NTM-type correlates with firms’ present exit rate. We also find that increases in tariffs and SPS measures in product- destination markets make it less likely that Colombian firms will continue exporting to these markets. This effect is significant and robust across all specifications, revealing that firms are responsive to additional market access conditions at this margin. 5.5 Results by Firm Size Larger firms may be able to more easily overcome the fixed costs needed to comply with a new market access conditions in the importing country. That is why we expect, on average, a smaller export-restricting effect of 27 NTMs for larger firms. Indeed, the largest exporters could gain from new market access conditions as demand is redirected toward them when small exporters are priced out of the market through new market access conditions. The literature, however, provides inconclusive evidence on the heterogeneous impact of NTM types and firm size. Fugazza et al. (2018) shows that new tariffs, TBT and PSI measures in destination markets benefit very large Peruvian exporters, while hurting smaller ones. On the other hand, Fernandes et al. (2021) finds that provisions that harmonize SPS and TBT regulations in PTAs are more beneficial for exports of smaller firms. To explore this hypothesis, we consider the following specification: Yf jpt = exp β1ln(1 + τijpt) + ∑ β2,l NTMlijpt + ˆ Rljpt ∑ β3,l I M l∈L l∈L + β 4 ln ( Imports f jpt) + ∑ β5,l NTMl ijpt × BigFirmf (4) l∈L 1 + β6ln(1 + τijpt) × BigFirmf + ω f jp + ω f pt + ωjt + ε f jpt Our coefficients of interest are β5,l and β6 in equation 4. They measure how the size of a firm creates a varied impact of NTMs and tariffs on the three margins of firm-level exports. We consider two mutually exclusive firm-size categories. As a baseline, we define firm size by the export market share a firm has in a 6-digit product market in its first defined year in the sample.14 Our definition of firm size is guided by how specific NTMs are; we assume that most are applicable to very narrowly defined HS 6-digit products. The underlying reasoning is that the introduction or withdrawal of NTMs could differentially benefit firms that are small from the point of view of that product’s market. 14We are limited to rely on export-based measures of firm characteristics because we do not have information on headcount, turnover or capital of Colombian exporting firms. We define firm size categories separately for each of the three margins of exports. This helps to account for the fact that firms have, on average, greater exports – and are thus bigger – at the intensive margin, as opposed to the extensive margins. 28 We then binarize the continuous firm-size variable based on its median and evaluate the differential impact of NTMs on BigFirmf in equation 4. We conduct two robustness checks on the definition of firm-size categories. First, we define firm size based on the export market share a firm has in an HS 2-digit sector in its first defined year in the sample. Second, we define firm size based on the export market share a firm has in its first defined year in the sample, considering all products and sectors. Thus, our additional firm-size definitions grow broader in scope.15 Our results confirm the heterogeneous nature of market access conditions on firm size. Table 13 attests to that end for the intensive margin of Colombian firm-level exports. The exports of small firms decline significantly when TBT and in particular quantity control measures are introduced in their export markets. In contrast, the effect is muted for big firms. That is why the interaction effects between TBT and Quantity controls and BigFirmf are statistically significant and positive. Moreover, this effect is consistent across our three definitions of firm size in Columns (2)–(4). It confirms previous evidence of the export-promoting effect of TBTs for big Peruvian exporters in Fugazza et al. (2018). Similarly, TBT and quantity control measures exert a regressive effect on the extensive margin of firm-level exports. Table 14 shows that small firms are significantly less likely to export when new TBT and quantity control measures are introduced in sales markets. Conversely, the impact is muted for big firms, irrespective of how we define firm sizes in columns (2)–(4). Firm size also shapes how TBTs and quantity control measures affect the likelihood of firms leaving markets. Table 15 reveals that small firms are more likely to leave sales market when TBT measures are introduced. That is why the baseline TBT coefficient – pertaining to small firms – is statistically significant and positive across all three definitions of firm size in Columns (2)–(4). In contrast, the likelihood of leaving these markets is not elevated for big firms, even though the interaction effect is not statistically 15As in equation 1, we use the same battery of three sets of high-dimensional fixed effects, include zeros along the time dimension and rely on the PPML estimator for all three margins of exports. 29 significant. The only progressive NTM effect comes from new quantity controls. They decrease the likelihood of exits from product-destination markets for small firms, yet increase it for big firms. This finding is consistent across all three definitions of firm size, as shown in the coefficients of Quantity Controli,j,p,t × Big Firm f in Columns (2)–(4). An explanation for this finding is a composition effect: quantity controls are more binding for big firms, increasing their likelihood of leaving markets. Quantity controls, in turn, impose fewer constraints for smaller exporters, making them less likely to leave markets. 5.6 Results for GVC Firms Firms participating in global value chains (GVCs) may face muted effects from NTMs. Since they boast extensive contractual relationships, including from importing, these firms enjoy enhanced access to information about regulatory changes, market conditions, and compliance requirements. As a result, GVC firms may find it easier to meet evolving regulatory requirements emanating from NTMs. To explore this hypothesis, we re-estimate specification 4 by interacting GVC Firm f t with tariffs and NTM types. We define GVC Firm f t as an indicator variable equal to 1 if a firm both imports and exports goods in a given year and 0 otherwise. GVC firms make up around 24% of all observations, as seen in Table 9. Intriguingly, only 20% of big firms are also GVC firms, while 55% of GVC firms are also big firms. Table 16 demonstrates a heterogeneous impact of market access conditions for GVC firms. Column (2) shows results for the intensive margin; Column (4) for the extensive margin I and Column (6) for the extensive margin II. To put these results into perspective, Columns (1), (3) and (5) provide the respective baseline results presented above. Looking at the intensive margin, GVC firms benefit from a dampened effect from new trade regulations in product destination markets. While non- GVC firms see a significant decline of their exports associated with 30 new quantity control, SPS and TBT measures, the positive interaction coefficients for GVC firms in Columns (2) suggest these effects are muted for them. Both effects are statistically significant and of economic relevance. GVC firms are also more likely to export as market access conditions affect them less than non-GVC firms. Column (4) shows that tariffs, pre-shipment and quantity control measures undermine the likelihood of non-GVC firms participating in exports. In contrast, these measures have a muted impact on GVC firms’ likelihood to participate in exports as their positive and significant interaction effect attests to. Intriguingly, GVC firms are more likely to leave markets when new quantity control measures are introduced. This is reflected in the posi- tive and statistically significant interaction effect Quantity Controli,j,p,t × GVC Firm f t in in Column (6), indicating a higher exit probability for GVC firms. Conversely, non-GVC firms are less likely to exit markets when quantity controls are introduced. A potential explanation is that new quantity measures are more binding for GVC firms than non-GVC firms due to their elevated exports in the first place. 6 Conclusion This paper studies how NTMs, the dominant instrument of today’s trade policy, challenge firm-level export decisions to access foreign markets. Our panel analysis of Colombian firms exporting to Latin America reveals that both TBT and quantity control measures decrease their exports on average. We rationalize this trade-deterring effect through an increase in firms’ compliance costs relative to consumer benefits. Other non-tariff measures and tariffs play a minor role. These average effects of trade regulations mask significant heterogeneities. TBT and quantity measures reallocate trade from small to big firms and those participating in global value chains. However, quantity control measures exert progressive effects under specific conditions, making it more likely that big firms will leave export markets, with the opposite effect occurring for small firms. We argue that the quantity controls are 31 more binding for big firms, increasing their likelihood of leaving markets and conversely helping small firms to take over their space. Our results are important for policy. First, they support evidence that trade regulations today are more trade restrictive than tariffs (Nicita and Melo 2018). This highlights the importance of mechanisms to harmonize trade regulations across countries to reduce their costs. Chief among them are international trade agreements, including through the World Trade Organization (Fernandes et al., 2021). That will effectively help firms access international markets, which is critical for lifting demand-side constraints on national development (Goldberg and Reed 2023). Second, our finding that even standard-like measures like TBTs undermine Colombian firm-level export decisions aligns with firms choosing not to embrace a new signal due to its associated costs and benefits. It challenges the trade-promoting effects of TBTs and that consumers receive useful information about product quality and increase trade, relative to the adverse impact on trade caused by any increase in cost (Zavala et al. 2023; Beghin et al. 2015). Moreover, it underscores the need for assistance to help firms comply with these trade regulations. This is especially warranted when market failures prevail, notably consumers take time to adjust their demand after receiving new quality signals (Bai, 2021). Third, our finding that trade regulations tend to favor large firms at the expense of small ones implies a concentration of world markets. It suggests that big firms benefit from protectionism, not globalization. This contrasts with liberalizing trade, which would result in a more equal distribution of export market shares among firms. It would also likely reduce wage inequality within the exporting country since small firms tend to be more less skilled relative to large firms (Cruz et al., 2017). However, with today’s new economic interventionism fueling trade protectionism, this is all but happening. Technological change and the backlash against globalization are making export-oriented industrialization, as seen in East Asia, much more difficult to achieve. 32 Table 10: Intensive margin: Existing exports Dependent Variable: Exports f pjt PCA Baseline Robustness tests Lags w/o large firms w/o concentrated sectors Placebo (1) (2) (3) (4) (5) (6) NTM Principal Componenti,j,p,t -0.2179 (0.044)*** Log (Tariffi,j,p,t) -0.0351 -0.0696 -0.0438 -0.0982 -0.0333 (0.036) (0.054) (0.047) (0.051)* (0.051) SPSi,j,p,t 0.0021 -0.0146 -0.0914 -0.0161 -0.0328 (0.048) (0.042) (0.050)* (0.048) (0.048) TBTi,j,p,t -0.0971 -0.0811 -0.0968 -0.1387 -0.0650 (0.044)** (0.028)*** (0.042)** (0.039)*** (0.044) Pre-Shipmenti,j,p,t 0.0388 -0.0154 0.0094 0.0323 0.0350 3 (0.053) (0.046) (0.049) (0.047) (0.056) Quantity Controli,j,p,t -0.5739 -0.4905 -0.2355 -0.3295 -0.2107 (0.158)*** (0.155)*** (0.100)** (0.115)*** (0.159) Price Controli,j,p,t -0.0483 -0.0410 -0.1115 -0.0793 0.0420 (0.040) (0.036) (0.028)*** (0.035)** (0.046) Other NTMsi,j,p,t 0.1252 0.1040 0.2088 0.0491 0.0289 (0.101) (0.097) (0.100)** (0.097) (0.075) Log (Import Demand f ,p,j,t) 0.0448 0.0324 0.0251 0.0291 0.0285 0.0340 (0.009)*** (0.006)*** (0.006)*** (0.005)*** (0.007)*** (0.007)*** Observations 253212 253212 227964 200738 250213 219733 Adjusted R2 0.93 0.96 0.96 0.94 0.96 0.96 Fixed Effects − p − j, f − p − t , j − t f − p − j, f − p − t, j − t f − p − j, f − p − t, j − t f − p − j, f − p − t, j − t f − p − j, f − p − t, j − t f − p − j, f − p − t, j − t This table estimates equation 1 for the intensive margin of Colombian firm-level exports. Columns (2-6) show coefficients standardized with zero mean and unit standard deviation. All columns use firm-HS6 product-destination, firm-HS6 product-year and destination-year fixed effects. All columns also control for the Inverse Mills Ratios for each NTM type following estimation of probit model of NTM selection based on NTM intensity in neighboring countries. Clustered standard errors at the HS6 product-year level are presented in parenthesis. ∗ p < 0.10,∗∗ p < 0.05,∗∗∗ p < 0.01 Table 11: Extensive margin I: Probability of export participation Dependent Variable: Probability of Export Participation f pjt PCA Baseline Robustness tests Lags w/o large firms w/o concentrated sectors Placebo (1) (2) (3) (4) (5) (6) NTM Principal Componenti,j,p,t -0.1448 (0.008)*** Log (Tariffi,j,p,t) -0.0483 -0.0571 -0.0512 -0.0488 -0.0242 (0.011)*** (0.011)*** (0.013)*** (0.011)*** (0.022) SPSi,j,p,t -0.0280 -0.0283 -0.0363 -0.0307 -0.0219 (0.013)** (0.013)** (0.016)** (0.013)** (0.024) TBTi,j,p,t -0.0343 -0.0424 -0.0496 -0.0348 -0.0098 (0.012)*** (0.012)*** (0.014)*** (0.012)*** (0.013) Pre-Shipmenti,j,p,t 0.0099 -0.0142 0.0169 0.0089 -0.0030 (0.009) (0.009) (0.012) (0.009) (0.013) 3 Quantity Controli,j,p,t -0.0537 -0.0454 -0.0391 -0.0501 -0.0193 (0.021)*** (0.022)** (0.022)* (0.021)** (0.022) Price Controli,j,p,t -0.0043 0.0018 -0.0073 -0.0044 0.0008 (0.014) (0.014) (0.016) (0.014) (0.014) Other NTMsi,j,p,t -0.0091 0.0248 -0.0053 -0.0172 -0.0254 (0.022) (0.020) (0.030) (0.022) (0.019) Log (Import Demand f ,p,j,t) 0.0184 0.0110 0.0105 0.0138 0.0114 0.0107 (0.001)*** (0.001)*** (0.001)*** (0.002)*** (0.001)*** (0.002)*** Observations 253212 253212 227964 200738 250213 219733 Adjusted R2 0.13 0.15 0.15 0.15 0.15 0.15 Fixed Effects f − p − j, f − p − t , j − t f − p − j, f − p − t , j − t f − p − j, f − p − t , j − t f − p − j, f − p − t , j − t f − p − j, f − p − t , j − t f − p − j, f − p − t , j − t This table estimates equation 1 for the extensive margin I of Colombian firm-level exports. Columns (2-6) show coefficients standardized with zero mean and unit standard deviation. All columns use firm-HS6 product-destination, firm-HS6 product-year and destination-year fixed effects. All columns also control for the Inverse Mills Ratios for each NTM type following estimation of probit model of NTM selection based on NTM intensity in neighboring countries. Clustered standard errors at the HS6 product-year level are presented in parenthesis. ∗ p < 0.10,∗∗ p < 0.05,∗∗∗ p < 0.01 Table 12: Extensive margin II: Probability of export exits Dependent Variable: Probability of Export Exit f pjt PCA Baseline Robustness tests Lags w/o large firms w/o concentrated sectors Placebo (1) (2) (3) (4) (5) (6) NTM Principal Componenti,j,p,t 0.1995 (0.031)*** Log (Tariffi,j,p,t) -0.0545 0.0382 -0.0614 -0.0493 -0.0390 (0.057) (0.055) (0.059) (0.058) (0.070) SPSi,j,p,t 0.1098 0.0390 0.0860 0.1316 0.0898 (0.087) (0.084) (0.094) (0.088) (0.093) TBTi,j,p,t 0.1939 0.2936 0.2148 0.2012 0.0538 (0.072)*** (0.077)*** (0.083)*** (0.074)*** (0.080) Pre-Shipmenti,j,p,t 0.0516 0.1018 -0.0357 0.0777 -0.1040 (0.096) (0.079) (0.101) (0.093) (0.099) 3 Quantity Controli,j,p,t -0.2712 -0.1944 -0.3128 -0.3292 -0.1175 (0.114)** (0.101)* (0.125)** (0.114)*** (0.132) Price Controli,j,p,t -0.0495 -0.1269 0.1028 -0.0229 -0.0480 (0.087) (0.118) (0.108) (0.090) (0.092) Other NTMsi,j,p,t -0.1053 -0.1477 -0.1053 -0.0639 -0.0706 (0.074) (0.060)** (0.077) (0.076) (0.078) Log (Import Demand f ,p,j,t) -0.0282 -0.0211 -0.0210 -0.0310 -0.0218 -0.0322 (0.006)*** (0.008)** (0.008)** (0.010)*** (0.008)*** (0.010)*** Observations 17221 17221 17221 13455 16961 13354 Adjusted R2 0.13 0.18 0.18 0.18 0.18 0.18 Fixed Effects f − p − j, f − p − t , j − t f − p − j, f − p − t , j − t f − p − j, f − p − t , j − t f − p − j, f − p − t , j − t f − p − j, f − p − t , j − t f − p − j, f − p − t , j − t This table estimates equation 1 for the extensive margin II of Colombian firm-level exports. Columns (2-6) show coefficients standardized with zero mean and unit standard deviation. All columns use firm-HS6 product-destination, firm-HS6 product-year and destination-year fixed effects. All columns also control for the Inverse Mills Ratios for each NTM type following estimation of probit model of NTM selection based on NTM intensity in neighboring countries. Clustered standard errors at the HS6 product-year level are presented in parenthesis. ∗ p < 0.10,∗∗ p < 0.05,∗∗∗ p < 0.01 Table 13: Intensive margin: Existing exports Dependent Variable: Exports f pjt Big Firm: exports > median in HS6 product-Year HS2 sector-Year Year (1) (2) (3) (4) Log (Tariffi,j,p,t) -0.0351 -0.0024 -0.0036 -0.2244 (0.036) (0.093) (0.085) (0.107)** Log (Tariffi,j,p,t) x Big Firm f ,t -0.0332 -0.0314 0.1962 (0.090) (0.090) (0.108)* SPSi,j,p,t 0.0021 -0.0840 -0.1102 0.0374 (0.048) (0.097) (0.067) (0.114) SPSi,j,p,t x Big Firm f ,t 0.1022 0.1256 -0.0300 (0.115) (0.079) (0.126) TBTi,j,p,t -0.0971 -0.2802 -0.2091 -0.0552 (0.044)** (0.078)*** (0.099)** (0.160) TBTi,j,p,t x Big Firm f ,t 0.1452 0.1136 0.1566 (0.071)** (0.060)*** (0.073)** Pre-Shipmenti,j,p,t 0.0388 -0.0929 -0.0327 0.1404 (0.053) (0.060) (0.077) (0.062)** Pre-Ship.i,j,p,t x Big Firm f ,t 0.1676 0.0797 -0.1017 3 (0.064)*** (0.087) (0.073) Quantity Controli,j,p,t -0.5739 -0.5769 -0.7281 -0.7879 (0.158)*** (0.134)*** (0.151)*** (0.150)*** Quantity Controli,j,p,t x Big Firm f ,t 0.267 0.1658 0.2286 (0.103)*** (0.082)** (0.100)** Price Controli,j,p,t -0.0483 -0.1686 -0.0706 -0.0782 (0.040) (0.059)*** (0.053) (0.100) Price Controli,j,p,t x Big Firm f ,t 0.1442 0.0173 0.0242 (0.072)** (0.055) (0.097) Log (Import Demand f ,p,j,t) 0.0324 0.0319 0.0322 0.0323 (0.006)*** (0.006)*** (0.006)*** (0.006)*** Observations 253212 253212 253212 253212 Adjusted R2 0.96 0.96 0.96 0.96 Fixed Effects f − p − j, f − p − t, j − t f − p − j, f − p − t, j − t f − p − j, f − p − t, j − t f − p − j, f − p − t, j − t This table estimates equation 4 for the intensive margin of Colombian firm-level exports. Columns (1-4) show coefficients standardized with zero mean and unit standard deviation. All columns use firm-HS6 product-destination, firm-HS6 product-year and destination-year fixed effects. All columns also control for the Inverse Mills Ratios for each NTM type following estimation of probit model of NTM selection based on NTM intensity in neighboring countries. Clustered standard errors at the HS6 product-year level are presented in parenthesis. ∗p < 0.10,∗∗ p < 0.05,∗∗∗ p < 0.01 Table 14: Extensive margin I: Probability of export participation Dependent Variable: Probability of Export Participation f pjt Big Firm: exports > median in HS6 product-Year HS2 sector-Year Year (1) (2) (3) (4) Log (Tariffi,j,p,t) -0.0483 -0.0487 -0.0659 -0.0302 (0.011)*** (0.015)*** (0.020)*** (0.020) Log (Tariffi,j,p,t) x Big Firm f ,t 0.0007 0.0228 -0.0223 (0.017) (0.022) (0.022) SPSi,j,p,t -0.0280 -0.0096 -0.0240 -0.0145 (0.013)** (0.022) (0.023) (0.025) SPSi,j,p,t x Big Firm f ,t -0.0313 -0.0077 -0.0198 (0.025) (0.026) (0.028) TBTi,j,p,t -0.0343 -0.0568 -0.0874 -0.0874 (0.012)*** (0.017)*** (0.022)*** (0.023)*** TBTi,j,p,t x Big Firm f ,t 0.0351 0.0700 0.0671 (0.019)* (0.025)*** (0.026)** Pre-Shipmenti,j,p,t 0.0099 -0.0179 -0.0071 0.0155 (0.009) (0.013) (0.016) (0.019) Pre-Ship.i,j,p,t x Big Firm f ,t 0.0469 0.0226 -0.0062 3 (0.012)*** (0.017) (0.020) Quantity Controli,j,p,t -0.0537 -0.1089 -0.1008 -0.0686 (0.021)*** (0.022)*** (0.027)*** (0.028)** Quantity Controli,j,p,t x Big Firm f ,t 0.0925 0.0574 0.0553 (0.017)*** (0.023)** (0.025)** Price Controli,j,p,t -0.0043 -0.0053 0.0004 -0.0427 (0.014) (0.019) (0.024) (0.025)* Price Controli,j,p,t x Big Firm f ,t 0.0016 -0.0074 0.0460 (0.021) (0.027) (0.027)* Log (Import Demand f ,p,j,t) 0.0110 0.0109 0.0110 0.0110 (0.001)*** (0.001)*** (0.001)*** (0.001)*** Observations 253212 253212 253212 253212 Adjusted R2 0.15 0.15 0.15 0.15 Fixed Effects f − p − j, f − p − t, j − t f − p − j, f − p − t, j − t f − p − j, f − p − t, j − t f − p − j, f − p − t, j − t This table estimates equation 4 for the extensive margin I of Colombian firm-level exports. Columns (1-4) show coefficients standardized with zero mean and unit standard deviation. All columns use firm-HS6 product-destination, firm-HS6 product-year and destination-year fixed effects. All columns also control for the Inverse Mills Ratios for each NTM type following estimation of probit model of NTM selection based on NTM intensity in neighboring countries. Clustered standard errors at the HS6 product-year level are presented in parenthesis. ∗p < 0.10,∗∗ p < 0.05,∗∗∗ p < 0.01 Table 15: Extensive margin II: Probability of export exit Dependent Variable: Probability of Export Exit f pjt Big Firm: exports > median in HS6 product-Year HS2 sector-Year Year (1) (2) (3) (4) Log (Tariffi,j,p,t) -0.0545 -0.0671 -0.0596 -0.0010 (0.057) (0.074) (0.081) (0.080) Log (Tariffi,j,p,t) x Big Firm f ,t 0.0723 0.0272 -0.0659 (0.102) (0.105) (0.100) SPSi,j,p,t 0.1098 -0.0437 0.0408 -0.1746 (0.087) (0.124) (0.128) (0.142) SPSi,j,p,t x Big Firm f ,t 0.3218 0.1290 0.4476 (0.168)* (0.168) (0.177)** TBTi,j,p,t 0.1939 0.2654 0.3064 0.2809 (0.072)*** (0.106)** (0.113)*** (0.116)** TBTi,j,p,t x Big Firm f ,t -0.0746 -0.1811 -0.1256 (0.150) (0.149) (0.143) Pre-Shipmenti,j,p,t 0.0516 0.0292 0.0221 0.0468 (0.096) (0.139) (0.139) (0.143) Pre-Ship.i,j,p,t x Big Firm f ,t 0.0108 0.0461 -0.0358 3 (0.178) (0.164) (0.181) Quantity Controli,j,p,t -0.2712 -0.2904 -0.3151 -0.2507 (0.114)** (0.133)** (0.118)*** (0.120)** Quantity Controli,j,p,t x Big Firm f ,t 0.0857 0.0129 0.0921 (0.095) (0.090) (0.088) Price Controli,j,p,t -0.0495 -0.1460 -0.1276 -0.0254 (0.087) (0.110) (0.106) (0.160) Price Controli,j,p,t x Big Firm f ,t 0.1414 0.1469 -0.0319 (0.118) (0.117) (0.155) Log (Import Demand f ,p,j,t) -0.0211 -0.0208 -0.0217 -0.0215 (0.008)** (0.009)** (0.009)** (0.009)** Observations 17221 16672 17077 17141 Adjusted R2 0.18 0.18 0.18 0.18 Fixed Effects f − p − j, f − p − t, j − t f − p − j, f − p − t, j − t f − p − j, f − p − t, j − t f − p − j, f − p − t, j − t This table estimates equation 4 for the extensive margin II of Colombian firm-level exports. Columns (1-4) show coefficients standardized with zero mean and unit standard deviation. All columns use firm-HS6 product-destination, firm-HS6 product-year and destination-year fixed effects. All columns also control for the Inverse Mills Ratios for each NTM type following estimation of probit model of NTM selection based on NTM intensity in neighboring countries. Clustered standard errors at the HS6 product-year level are presented in parenthesis. ∗p < 0.10,∗∗ p < 0.05,∗∗∗ p < 0.01 Table 16: Heterogeneous effects of market access conditions for GVC firms Dependent Variable: Exports f ,j,p,t Prob (Export Participation f ,j,p,t) Prob (Export Exits f ,j,p,t) (1) (2) (3) (4) (5) (6) Log (Tariffi,j,p,t) -0.0351 0.1146 -0.0483 -0.0713 -0.0545 -0.0781 (0.036) (0.130) (0.011)*** (0.018)*** (0.057) (0.095) Log (Tariffi,j,p,t) x GVC Firm f ,t -0.1509 0.0258 0.0516 (0.118) (0.014)* (0.080) SPSi,j,p,t 0.0021 -0.1688 -0.0280 0.0126 0.1098 -0.1405 (0.048) (0.055)*** (0.013)** (0.022) (0.087) (0.134) SPSi,j,p,t x GVC Firm f ,t 0.1720 -0.0470 0.3144 (0.062)*** (0.020)** (0.129)** TBTi,j,p,t -0.0971 -0.2451 -0.0343 -0.0293 0.1939 0.0967 (0.044)** (0.101)** (0.012)*** (0.027) (0.072)*** (0.182) TBTi,j,p,t x GVC Firm f ,t 0.1463 -0.0046 0.0932 (0.106) (0.024) (0.168) Pre-Shipmenti,j,p,t 0.0388 -0.0033 0.0099 -0.0245 0.0516 0.1074 (0.053) (0.103) (0.009) (0.013)* (0.096) (0.119) 3 Pre-Ship.i,j,p,t x GVC Firm f ,t 0.0506 0.0391 -0.0740 (0.079) (0.010)*** (0.074) Quantity Controli,j,p,t -0.5739 -0.7828 -0.0537 -0.0904 -0.2712 -0.5528 (0.158)*** (0.185)*** (0.021)*** (0.026)*** (0.114)** (0.132)*** Quantity Controli,j,p,t x GVC Firm f ,t 0.2191 0.0364 0.2449 (0.119)* (0.020)* (0.081)*** Price Controli,j,p,t -0.0483 -0.1134 -0.0043 -0.0138 -0.0495 -0.0685 (0.040) (0.064)* (0.014) (0.021) (0.087) (0.126) Price Controli,j,p,t x GVC Firm f ,t 0.0655 0.0102 0.0216 (0.061) (0.016) (0.094) Log (Import Demand f ,p,j,t) 0.0324 0.0321 0.0110 0.0110 -0.0211 -0.0212 (0.006)*** (0.006)*** (0.001)*** (0.001)*** (0.008)** (0.008)** Observations 253212 253212 253212 253212 17221 17221 Adjusted R2 0.96 0.96 0.15 0.15 0.18 0.18 Fixed Effects f − p − j, f − p − t, j − t f − p − j, f − p − t, j − t f − p − j, f − p − t, j − t f − p − j, f − p − t , j − t f − p − j, f − p − t, j − t f − p − j, f − p − t , j − t This table estimates equation 4 for the intensive margin in columns (1) and (2); extensive margin I in column (3) and (4) and extensive margin II in column (5) and (6). GVC firm is an indicator variable equal to 1 if the firm both imports and exports goods in a given year and 0 otherwise. All columns show coefficients standardized with zero mean and unit standard deviation. All columns use firm-HS6 product-destination, firm-HS6 product-year and destination-year fixed effects. All columns also control for the Inverse Mills Ratios for each NTM type following estimation of probit model of NTM selection based on NTM intensity in neighboring countries. Clustered standard errors at the HS6 product-year level are presented in parenthesis. ∗ p < 0.10,∗∗ p < 0.05,∗∗∗ p < 0.01 Appendix A Control Function Estimates Table A1: Control Function Estimates SPS TBT Pre-Shipment Quantity Control Price Controls Other (1) (2) (3) (4) (5) (6) Neighbor SPSj,p,t 0.2903 -0.0916 0.0394 -0.0825 -0.0651 -0.1370 (0.001)*** (0.001)*** (0.001)*** (0.001)*** (0.001)*** (0.002)*** Neighbor TBTj,p,t 0.0918 0.1835 0.0051 -0.0805 -0.0301 0.0085 (0.002)*** (0.002)*** (0.002)** (0.002)*** (0.002)*** (0.002)*** Neigbor Pre-Shipmentj,p,t -0.3264 1.3487 0.0281 0.6419 0.3193 1.0509 (0.020)*** (0.021)*** (0.019) (0.015)*** (0.017)*** (0.022)*** Neighbor Quantity Controlsj,p,t -0.3772 -0.2079 -0.1883 -0.6891 -0.2958 -0.6978 (0.014)*** (0.014)*** (0.017)*** (0.014)*** (0.015)*** (0.023)*** Neighbor Price Controlsj,p,t -0.1582 0.0111 0.1451 -0.1864 0.5267 0.5255 (0.009)*** (0.008) (0.010)*** (0.009)*** (0.008)*** (0.011)*** Neighbor Otherj,p,t -1.0265 -0.1832 0.6604 -1.8394 0.1387 -1.8006 (0.017)*** (0.016)*** (0.017)*** (0.022)*** (0.015)*** (0.040)*** Observations 468889 468889 468889 468889 468889 468889 This table presents Probit Estimates of Equation 2. Each column uses a different NTM type as dependent variable. ∗p < 0.10,∗∗ p < 0.05,∗∗∗ p < 0.01 B Robustness Tests B.1 NTM overlap Simultaneous changes of multiple NTM types may complicate isolating their respective effects. It occurs when two or more NTM types enter the same product-destination market over time. We refer to this phenomenon as NTM overlap and define it as multiple introductions or withdrawals of distinct NTM types in a product-destination cell. The classical example refers to a situation in which both an import quota and a TBT are applied. A firm may be able to cope with TBT requirements, but because of the quota imposed at the destination, it might not be able to export to that destination. The impact of the TBT is altered by the presence of the quota. Indeed, the effect of one specific NTM may absorb the effect of any other. Table B2, however, reveals that overlap in NTM changes is limited. In fact, policies with multiple NTM changes affect, on average, 22.3% of ex- 40 port relationships at the HS 6-digit product-destination level.16 Given the limited overlap of NTMs, we are thus confident that our empirical framework clearly identifies the impact of NTM types on Colombian export patterns. Overlap may also occur when two or more measures of the same type (e.g., two SPS measures) come into effect for the same product. This, however, is less of a concern, since the scope of our empirical assessment is to identify the average effect of the presence of broad categories of NTMs, rather than the impact of some specific regulation. Table B2: Distribution of simultaneous NTM changes at the product- destination level 1 2 3 4 5+ 2008 0.897 0.103 0.000 0.000 0.000 2009 0.910 0.090 0.000 0.000 0.000 2010 0.948 0.052 0.000 0.000 0.000 2011 0.682 0.056 0.022 0.095 0.145 2012 0.992 0.008 0.000 0.000 0.000 2013 0.303 0.115 0.461 0.121 0.000 2014 0.987 0.013 0.001 0.000 0.000 2015 0.612 0.388 0.000 0.000 0.000 2016 0.997 0.003 0.000 0.000 0.000 2017 0.966 0.034 0.000 0.000 0.000 Pooled 0.777 0.066 0.097 0.038 0.021 Source: Author’s calculation based on UNCTAD’s NTM data. Note: This table reports the distribution of changes in distinct NTMs (measured at the chapter level) at the product-destination level over time. The pooled sample reports the distribution of NTM changes across all years. B.2 Collinearity between NTMs The collinearity between introductions of NTM types may pose a concern to our identification strategy. In particular, it may complicate the interpretation of significance since we evaluate coefficients of NTM types jointly. 161-77.7%. 41 This begs the empirical question of how much do NTMs correlate after controlling for our fixed effects. We put forward four arguments that collinearity between market access conditions is not a concern. First, Table B3 shows correlations between NTM types, conditional on firm-HS6 product- destination, firm-HS6 product-year and destination-year fixed effects. The correlations between NTM types are positive, but not strikingly large.17 Therefore, multi-collinearity seems a less relevant concern for our empirical strategy. Table B3: Correlations of NTMs, After Fixed Effects Variables Tariff SPS TBT Pre- Quantity Price shipment control control Tariff 1.000 SPS -0.008 1.000 TBT -0.010 0.344 1.000 Pre-Shipment -0.013 0.358 0.123 1.000 Quantity Control -0.000 0.420 0.209 0.404 1.000 Price Control -0.002 0.005 0.074 -0.007 0.000 1.000 This table displays bivariate correlation coefficients for tariffs and NTM types after conditioning on firm- HS6 product-destination, firm-HS6 product-year and destination-year fixed effects. Second, we address concerns that NTM types exhibit collinearity by running specification (1) with only two cross sections, using data from the first and last year of data. To that end, Table B4 shows results based on data from 2007 and 2017 only. The results are consistent with our baseline specification in column (2) of Table 10 (intensive margin) and 11 (extensive margin I).18 Third, we conduct a principal component analysis (PCA) of all NTM types. This helps us define a continuous NTM index and evaluate its role to explain Colombian firm-level exports across the three margins considered. All NTM indexes prove statistically significant and with the expected sign. 17Moreover, the correlation between tariffs and NTM types is about 0, allaying concerns over our estimation strategy. 18We can’t estimate the extensive margin II (probability of export exits) in Table B4 because exits are not defined with data from 2007 and 2017 only. 42 Table B4: Results with two cross sectional data only (2007 and 2017) Dependent Variable: Exports f ,j,p,t Probability of Export Participation f ,j,p,t (1) (2) Log (Tariffi,j,p,t) -0.1715 -0.0818 (0.123) (0.037)** SPSi,j,p,t -0.1221 -0.0053 (0.092) (0.031) TBTi,j,p,t -0.1052 -0.0873 (0.049)** (0.031)*** Pre-Shipmenti,j,p,t 0.0409 -0.0614 (0.074) (0.025)** Quantity Controli,j,p,t -1.1977 -0.1079 (0.226)*** (0.057)* Price Controli,j,p,t -0.1056 0.0099 (0.084) (0.035) Other NTMsi,j,p,t 0.1318 -0.0048 (0.141) (0.059) Log (Import Demand f ,p,j,t) 0.0331 0.0157 (0.018)* (0.006)*** Observations 14622 14622 Adjusted R2 0.98 0.12 Fixed Effects f − p − j, f − p − t, j − t f − p − j, f − p − t, j − t This table estimates equation 1 with two cross section of data in 2007 and 2017 only. Column (1) shows results for the intensive margin and column (2) for the extensive margin I of Colombian firm-level exports. All coefficients are standardized with zero mean and unit standard deviation. All columns use firm-HS6 product-destination, firm-HS6 product-year and destination-year fixed effects. All columns also control for the Inverse Mills Ratios for each NTM type following estimation of probit model of NTM selection based on NTM intensity in neighboring countries. Clustered standard errors at the HS6 product-year level are presented in parenthesis. ∗p < 0.10,∗∗ p < 0.05,∗∗∗ p < 0.01 43 References Adarov, Amat and Mahdi Ghodsi (2023). “Heterogeneous effects of nontariff measures on cross-border investments: Bilateral firm-level analysis”. In: Review of International Economics 31.1, pp. 158–179. ISSN: 14679396. DOI: 10.1111/roie.12621. Arkolakis, Costas (2010). “Market penetration costs and the new consumers margin in international trade”. In: Journal of Political Economy 118.6, pp. 1151–1199. ISSN: 00223808. DOI: 10.1086/657949. Asprilla, Alan, Nicolas Berman, Olivier Cadot, and Mélise Jaud (2019). “Trade Policy and Market Power: Firm-Level Evidence”. In: International Economic Review 60.4, pp. 1647–1673. ISSN: 14682354. DOI: 10 .1 111/iere.12409. Bai, Jie (2021). “Melons as lemons: Asymmetric information, consumer learning and seller reputation”. In: CID Faculty Working Paper Series 2021.396, Harvard University, Cambridge. Baier, Scott L. and Jeffrey H. Bergstrand (2007). “Do free trade agreements actually increase members’ international trade?” In: Journal of International Economics 71.1, pp. 72–95. ISSN: 0022-1996. DOI: 10.1016/J.JINTE CO.2006.02.005. Bao, Xiaohua and Larry D. Qiu (2012). “How Do Technical Barriers to Trade Influence Trade?” In: Review of International Economics 20.4, pp. 691–706. ISSN:1467-9396. DOI: 10 . 1111 / J . 1467 - 9396 . 2012 . 01047 . X. URL: https://onlinelibrary.wiley.com/doi/full/10.1111/j.1467-9396 .2012.01047.x%20https://onlinelibrary.wiley.com/doi/abs/10.1 111/j.1467-9396.2012.01047.x%20https://onlinelibrary.wiley.c om/doi/10.1111/j.1467-9396.2012.01047.x. Beghin, John Christopher, Anne Célia Disdier, and Stéphan Marette (2015). “Trade restrictiveness indices in the presence of externalities: An application to non-tariff measures”. In: 48.4, pp. 1513–1536. DOI: 10 . 1111 /CAJE.12157. Bernard, Andrew B., J. Bradford Jensen, Stephen J. Redding, and Peter K. Schott (2012). “The empirics of firm heterogeneity and international 44 trade”. In: Annual Review of Economics 4, pp. 283–313. ISSN: 19411383. DOI: 10.1146/annurev-economics-080511-110928. URL: https://www .annualreviews.org/doi/abs/10.1146/annurev-economics-080511- 110928. Bernard, Andrew B., Stephen J. Redding, and Peter K. Schott (2011). “Multi- product Firms and Trade Liberalization”. In: The Quarterly Journal of Economics 126.3, pp. 1271–1318. ISSN: 0033-5533. DOI: 10.1093/QJE/QJR021. URL: https://dx.doi.org/10.1093/qje/qjr021. Bratt, Michael (2017). “Estimating the bilateral impact of nontariff measures on trade”. In: Review of International Economics 25.5, pp. 1105–1129. ISSN: 14679396. DOI: 10.1111/ROIE.12297. Cebeci, Tolga and Ana M. Fernandes (2015). “Microdynamics of Turkey’s Export Boom in the 2000s”. In: World Economy 38.5, pp. 825–855. ISSN: 14679701. DOI: 10.1111/twec.12183. Correia, Sergio, Paulo Guimarães, and Tom Zylkin (2020). “Fast Poisson estimation with high-dimensional fixed effects”. In: The Stata Journal: Promoting communications on statistics and Stata 20.1, pp. 95–115. ISSN: 1536-867X. DOI: 10.1177/1536867X20909691. URL: http://journals.s agepub.com/doi/10.1177/1536867X20909691. Cruz, Marcio, Emmanuel Milet, and Marcelo Olarreaga (2017). “Online exports and the skilled-unskilled wage gap”. In: Policy Research Working Paper Series, The World Bank 8160. ISSN: 19326203. DOI: 10.1371/journa l.pone.0232396. URL: https://journals.plos.org/plosone/article ?id=10.1371/journal.pone.0232396. Curzi, Daniele (2020). “Standards, trade margins and product quality: Firm-level evidence from Peru”. In: Food Policy 91.February, p. 101834. ISSN: 03069192. DOI: 10.1016/j.foodpol.2020.101834. URL: https: //doi.org/10.1016/j.foodpol.2020.101834. Disdier, Anne Célia, Lionel Fontagné, and Mondher Mimouni (2008). “The Impact of Regulations on Agricultural Trade: Evidence from the SPS and TBT Agreements”. In: American Journal of Agricultural Economics 90.2, pp. 336–350. ISSN: 1467-8276. DOI: 10 .1111 /J.1467 - 8276 .2007 45 .01127.X. URL: https://onlinelibrary.wiley.com/doi/full/10.111 46 1/j.1467-8276.2007.01127.x%20https://onlinelibrary.wiley.com /doi/abs/10.1111/j.1467-8276.2007.01127.x%20https://onlineli brary.wiley.com/doi/10.1111/j.1467-8276.2007.01127.x. Disdier, Anne Célia and Stéphan Marette (2010). “The combination of gravity and welfare approaches for evaluating nontariff measures”. In: American Journal of Agricultural Economics 92.3, pp. 713–726. ISSN: 00029092. DOI: 10.1093/ajae/aaq026. URL: https://onlinelibrary.w iley.com/doi/full/10.1093/ajae/aaq026%20https://onlinelibrar y.wiley.com/doi/abs/10.1093/ajae/aaq026%20https://onlinelibr ary.wiley.com/doi/10.1093/ajae/aaq026. Falciola, Justine, Marion Jansen, and Valentina Rollo (2020). “Defining firm competitiveness: A multidimensional framework”. In: World Development 129, p. 104857. ISSN: 0305-750X. DOI: 10.1016/J.WORLDDEV.2019.1 04857. Felbermayr, Gabriel, Aleksandra Kirilakha, Constantinos Syropoulos, Erdal Yalcin, and Yoto V. Yotov (2020). “The global sanctions data base”. In: European Economic Review 129, p. 103561. ISSN: 0014-2921. DOI: 10.1 016/J.EUROECOREV.2020.103561. Fernandes, Ana, Kevin Lefebvre, and Nadia Rocha (2021). “Heterogeneous impacts of SPS and TBT regulations : Firm-Level Evidence from Deep Trade Agreements”. In: Policy Research Working Paper, World Bank 9700. Fernandes, Ana M., Caroline Freund, and Martha Denisse Pierola (2016). “Exporter behavior, country size and stage of development: Evidence from the exporter dynamics database”. In: Journal of Development Economics 119, pp. 121–137. ISSN: 03043878. DOI: 10.1016/j.jdeveco.2015 .10.007. Fontagné, Lionel and Gianluca Orefice (2018). “Let’s try next door: Technical Barriers to Trade and multi-destination firms”. In: European Economic Review 101, pp. 643–663. ISSN: 00142921. DOI: 10.1016/j.euroecorev.2 017.11.002. Fontagné, Lionel, Gianluca Orefice, Roberta Piermartini, and Nadia Rocha (2015). “Product standards and margins of trade: Firm-level evidence”. 47 In: Journal of International Economics 97.1, pp. 29–44. ISSN: 0022-1996. DOI: 10.1016/J.JINTECO.2015.04.008. Fugazza, Marco and Alessandro Nicita (2013). “The direct and relative effects of preferential market access”. In: Journal of International Economics 89.2, pp. 357–368. DOI: 10.1016/j.jinteco.2012.09.001. Fugazza, Marco, Marcelo Olarreaga, and Cristian Ugarte (2018). “On the heterogeneous effects of market-access barriers: Evidence from small and large Peruvian exporters”. In: LACEA. ISSN: 1098-6596. arXiv: arXi v:1011.1669v3. Goldberg, Pinelopi and Tristan Reed (2023). “Demand-Side Constraints in Development : The Role of Market Size, Trade, and (In) Equality”. In: Econometrica 91.6, pp. 1915–1950. ISSN: 1468-0262. DOI: 10.3982/ECTA20 787. Kee, Hiau Looi and Alessandro Nicita (2022). “Trade fraud and non-tariff measures”. In: Journal of International Economics 139, p. 103682. ISSN: 18730353. DOI: 10 .1016 /j.jinteco.2022 .103682. URL: https:/ / do i.org/10.1016/j.jinteco.2022.103682. Li, Yuan and John C. Beghin (2012). “A meta-analysis of estimates of the impact of technical barriers to trade”. In: Journal of Policy Modeling 34.3, pp. 497–511. ISSN: 0161-8938. DOI: 10.1016/J.JPOLMOD.2011.11.001. Macedoni, Luca and Ariel Weinberger (2022). “Quality heterogeneity and misallocation: The welfare benefits of raising your standards”. In: Journal of International Economics 134, p. 103544. ISSN: 18730353. DOI: 10.10 16/j.jinteco.2021.103544. URL: https://doi.org/10.1016/j.jinte co.2021.103544. Mayer, Thierry and Soledad Zignago (2011). “Notes on CEPII’s Distances Measures: The GeoDist Database”. In: CEPII Working Paper No. 2011-25. DOI: 10.2139/SSRN.1994531. URL: https://papers.ssrn.com/abstrac t=1994531. Melitz, M (2003). “The impact of trade on intra-industry reallocations and aggregate industry productivity”. In: Econometrica 71.6, pp. 1695–1725. DOI: 10.1111/1468-0262.00467. 48 Melitz, Marc J. and Stephen J. Redding (2015). “New Trade Models, New Welfare Implications”. In: American Economic Review 105.3, pp. 1105–46. ISSN: 0002-8282. DOI: 10.1257/AER.20130351. Nicita, Alessandro and Jaime de Melo (2018). “NTM : economic assessment and policy options for development”. In: Non-Tariff Measures : Economic Assessment and Policy Options for Development UNCTAD/DITC/TAB/2017/2, pp. 14–79. Ronen, Eyal (2017). “Quantifying the trade effects of NTMs: A review of the empirical literature”. In: Journal of Economics and Political Economy 4.3, pp. 263–274. ISSN: 2148-8347. DOI: 10.1453/JEPE.V4I3.1360. URL: http://kspjournals.org/index.php/JEPE/article/view/1360. Santos Silva, J. M.C. and Silvana Tenreyro (2006). “The log of gravity”. In: Review of Economics and Statistics 88.4, pp. 641–658. ISSN: 00346535. DOI: 10.1162/rest.88.4.641. URL: https://www.mitpressjournals.org/d oix/abs/10.1162/rest.88.4.641. UNCTAD (2019). International classification of non-tariff measures - 2019 version. Tech. rep. Geneva: UNCTAD/DITC/TAB/2019/5, UNCTAD/DITC/TAB/2018/2. URL: https:/ / unctad .org / webflyer/ int ernational-classification-non-tariff-measures-2019- version. UNCTAD and World Bank (2019). The unseen impact of non-tariff measures: Insights from a new database. Vol. UNCTAD/DITC/TAB/2018/2. UN. URL: https://books.google.com.pk/books?id=HwdszgEACAAJ. World Bank (2020). World Development Report: Trade for Development in the ages of GVCs. The World Bank Washington, DC. ISBN: 978-1-4648-0667- 4. DOI: http://hdl.handle.net/10986/32437. URL: http://www.doing business.org/$%5Csim$/media/GIAWB/Doing%20Business/Documents /Annual-Reports/English/DB16-Full-Report.pdf. Yousefi, Ayoub and Mengyi Liu (Dec. 2013). “The Impact of Technical Barriers to Trade: The Cases of Trade Between China, Japan, Korea, and the US”. In: Innovation in the High-Tech Economy. Ed. by Pang Chuan, Vasil Khachidze, Ivan K.W. Lai, Yide Liu, Sohail Siddiqui, and Tim Wang. 127th ed. Contributions to Economics. Springer, pp. 23–34. DOI: 10.100 49 7/978-3-642-41585-2. URL: https://ideas.repec.org/h/spr/conch p/978-3-642-41585-2_3.html. Zavala, Lucas, Ana Fernandes, Ryan Haygood, Tristan Reed, and Jose- Daniel Reyes (2023). Quality Regulation Creates and Reallocates Trade. November, pp. 1–6. DOI: 10.1596/1813-9450-10601. 50