Policy Research Working Paper 11005 Decarbonization in MENA Energy Transition, Trade, and Labor Markets Raymond Robertson Gladys Lopez Acevedo Poverty Global Department December 2024 Policy Research Working Paper 11005 Abstract This paper examines two key questions about decarboniza- compliance costs, thereby reducing trade flows. Middle tion and its economic implications. First, it analyzes how East and North Africa–specific findings suggest that while environmental provisions in trade agreements affect bilat- regional trade agreements may be less advantageous for eral trade flows, with a specific focus on the Middle East and countries in the region, compared to other regions, envi- North Africa region. By constructing a detailed dataset of ronmental provisions can counterbalance this by improving trade agreements that include environmental provisions and the region’s environmental standards and reputation, ulti- applying an augmented gravity model, the study reveals that mately supporting trade growth. The second focus of the while regional trade agreements generally promote trade paper explores labor market consequences from rising by reducing barriers and fostering economic cooperation, carbon prices and the transition to renewable energy in the inclusion of environmental provisions introduces com- Tunisia. The findings indicate that districts heavily reliant plexity. Environmental provisions can enhance or hinder on fossil fuels experience significant employment declines, trade flows depending on the nature of the provisions and particularly among male workers, as carbon prices rise. The the economic context. Provisions related to general envi- results underscore the importance of targeting policies to ronmental goals and judicial enforcement tend to promote mitigate job losses in carbon-intensive sectors while pro- trade, whereas more stringent regulations often impose moting “green” job creation in renewable energy industries. This paper is a product of the Poverty Global Department. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at gacevedo@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Decarbonization in MENA: Energy Transition, Trade, and Labor Markets 1 Raymond Robertson (Texas A&M) Gladys Lopez Acevedo (World Bank) JEL: D3, F16, J16, O19. Keywords: trade shocks, labor markets, formal employment, wages, green jobs. 1This paper is a product of the Poverty and Equity Practice at the World Bank Group. It was funded by the MENA Chief Economist Office. We thank the following colleagues for their immense support during this process specially Deeksha Kokas, Juan Segnana, Daniel Lederman, Ernest John Sergenti, Chahir Zaki. The views expressed herein are those of the authors and do not necessarily reflect the views of the World Bank. 1. Introduction As the global transition to green energy intensifies, developed countries have begun employing trade policies to promote decarbonization. These policies often involve imposing import taxes and export subsidies based on the carbon content of goods. For example, the European Union’s (EU) carbon border adjustment mechanism imposes tariffs on imports from countries with lower carbon taxes. Such policies aim to incentivize trading partners, especially developing countries, to raise their environmental standards, aligning them with the stricter environmental regulations of wealthier nations. This shift in trade policy is part of a broader trend in which environmental provisions (EPs) are increasingly embedded in trade agreements, requiring countries to comply with national and international environmental regulations. As environmental standards become more stringent, the cost of carbon-emitting production processes is expected to rise. Understanding the implications of rising costs associated with carbon-emitting production processes on trade flows, as well as on labor markets and employment, in developing countries is crucial. For example, the MENA region is highly vulnerable to climate change, with 60 percent of its population living in areas facing severe water stress, a situation expected to worsen. Climate change may reduce rainfed crop yields by 30 percent, threatening food security and increasing dependence on imports. Rising temperatures and sea levels also put coastal cities at risk of displacement and economic loss, while more frequent and intense climate-related disasters, such as droughts and floods, further exacerbate the region's challenges. The global shift to green energy and the pressure to reduce carbon emissions are reshaping both trade dynamics and labor markets in MENA, presenting unique opportunities and challenges. Figure 1 clearly outlines the conceptual framework. The transition to cleaner energy sources will lead to significant shifts in relative prices of energy. These shifts will inevitably disrupt domestic and international product and factor markets, potentially resulting in significant distributional and welfare consequences. According to International Energy Agency (IEA, 2022) forecasts, the price of oil is expected to steadily increase and stabilize over the next three decades, contingent upon the effectiveness of political leaders in curbing fossil fuel demand. In contrast, renewable energy prices are projected to continuously decline due to technology advancements. As transition towards renewable energy accelerates, fossil fuels will experience significant relative price increases. This transition will disrupt numerous labor markets reliant on carbon-intensive production or usage. 2 Figure 1: Conceptual framework Effects mediated by domestic labor market institutions and policies Labor Welfare Trade Policy Carbon Price Trade Flows Market and Outcomes Poverty (1) Green transition is necessary and trade policy may possibly help; (2) Green transition will change carbon prices; (3) Changes in carbon prices will shift trade flows, production and jobs, causing job displacements, especially for poor workers; (4) We need a green transition and policies to mitigate the employment and welfare impact in developing countries Given this background, the first key question this study addresses is: how do trade-related environmental policies impact carbon-intensive industries and trade flows, particularly in developing countries? Early theoretical work, such as Pethig’s (1976) general equilibrium model, demonstrated that environmental regulations can shift comparative advantage between countries. This framework was extended by McGuire (1982), who integrated environmental factors into the Heckscher-Ohlin model, illustrating how countries adjust their production in response to environmental policies. Copeland and Taylor (1994) furthered this analysis by highlighting how countries might specialize in "dirty" or "clean" goods depending on the stringency of their environmental regulations, particularly in North-South trade dynamics. Grossman and Krueger (1991) provided the foundational scale, composition, and technique effects framework to understand how trade affects the environment. Despite these advances in theory, empirical literature on environmental provisions in trade agreements has found mixed repercussions. Antweiler et al. (2001) and Frankel and Rose (2005) suggested that trade could enhance environmental quality, while Cole and Elliott (2003) found the relationship to be more ambiguous, with effects varying by industry and region. More recently, Mattoo et al. (2020) demonstrated that environmental provisions in trade agreements have become more common over the past 30 years, especially in agreements involving high-income countries, where they serve as a bargaining tool to promote environmental goals. However, Dai et al. (2021) found that stringent environmental policies had little effect on trade in environmental goods but did influence the trade of other carbon-intensive products. Similarly, Berger et al. (2020) showed that environmental provisions in preferential trade agreements (PTAs) can moderate the overall trade-boosting effect of PTAs, particularly in developing countries, where compliance costs may be higher. This paper contributes to this literature by making some of the first estimates of how trade- related EPs affect carbon-intensive industries and trade flows in developing countries. The key novelty of this study is that we assess how EPs in trade agreements affect trade flows by 3 constructing a database of trade agreements with environmental provisions. This dataset combines measures of: (i) bilateral trade flows; (ii) export and import-specific factors, as well as various trade costs; (iii) membership in RTAs; and (iv) information about the types of environmental clauses specified in the RTAs. While information on (i) and (ii) can often be found in standard databases, we build our database of environmental clauses in RTAs by analyzing each trade agreement individually from the World Trade Organization’s (WTO) RTA database. The second key question this study addresses is: what are the labor market consequences of rising carbon prices and the transition to renewable energy, particularly for developing economies reliant on fossil fuels? Several studies have documented the employment effects of transitioning to renewable energy, particularly in developed countries. Chan and Zhou (2023) found that the expansion of solar and wind energy in the United States (U.S) from 2005 to 2019 increased employment and wages, especially benefiting younger and lower-educated workers, while reducing reliance on government welfare programs. In Germany, Sievers et al. (2019) noted that the energy transition boosted employment, particularly in construction and electricity generation, with regions like northern and eastern Germany benefiting the most from investments in renewable energy. However, the shift away from carbon-intensive industries poses significant challenges for regions heavily dependent on fossil fuels. Hanson (2023) emphasized the negative labor market consequences in such regions, highlighting the need for robust social safety nets and place-based policies to mitigate job losses. Baran et al. (2020) examined decarbonization in coal-dependent countries such as Poland, noting net job losses when cleaner energy sources, such as imported gas, replaced coal. Similarly, Jarvis et al. (2022) highlighted the social and environmental costs of Germany’s nuclear phase-out, which led to increased coal-fired energy production and associated public health risks, underscoring the complexity of the energy transition. In developing countries, how rising carbon prices affect labor markets remain underexplored. For example, Casey et al. (2020) showed that state-level carbon pricing in the U.S. led to a 2.7% reduction in local employment, with neighboring states absorbing displaced workers. However, countries like Tunisia, which rely heavily on imported fossil fuels, face different challenges. Tunisia aims to increase its renewable energy capacity from 8% to 35% by 2030, which will create new job opportunities in green sectors such as solar and wind power. Yet, this shift is also likely to result in job displacement in carbon-intensive industries, such as oil and gas. As Tunisia diversifies its energy mix, understanding the labor market effects of these energy price shifts is crucial for designing policies that mitigate the negative consequences of the transition. Our study addresses this gap by examining the labor market consequences of rising carbon prices in Tunisia. By focusing on employment, we will assess how the green transition affects vulnerable labor markets. The rest of the paper is organized in the following manner. 2 Section 2 on methodology discusses the econometric approaches undertaken to study the dual objectives of this paper: (i) estimate how EPs in trade agreements affect trade flows, and (ii) estimate the local labor 2 Annex B provides some additional results analyzing if a carbon adjustment tariff can be effective. 4 market effects of rising carbon prices. Section 3 discusses all relevant data sources used in this paper. Section 4 provides a detailed description of the results. Section 5 concludes. 2. Methodology We follow a two-step methodology to study the outlined research objectives. Our approach is to combine gravity model estimation in both static and dynamic settings (Boehm, Levchenko and Pandalai-Nayar, 2023) as well as empirical local labor market estimation (for example, Hanson, 2023) to generate some of the first policy-relevant estimates of how these trade policies affect labor markets. Our approach can be applied to both specific groups of commodities (for example, palm oil, soybeans, and carbon-based fuels) and more "macro" variables such total trade flows and local unemployment rates. To estimate the level of impact of EPs in trade agreements, we apply an empirical gravity model to estimate trade elasticity and help anticipate expected outcomes of changes in policy variables. 3 Further, we augment the standard gravity model to assess the elasticity of trade flows to the existence of environmental provisions in RTAs, relative to an RTA without an environmental provision. To derive our empirical specifications, we start from the following specification of the gravity-regression, which we derive from a structural model in Annex A: 4 , = −θ τ � sd + υ, � , + ξ, + δ, + ψ (1) where , is the trade flow between countries s and d in period t; , , , are the destination-year and source-year-fixed effects, respectively; � are source-destination fixed-effects. � , in a way to use variability in RTA provisions to In our following analysis, we will define τ estimate the elasticity of trade to RTAs and to environmental provisions in RTAs. In particular: ̃, = ⋅ , + ⋅ , ⋅ { , ∈ } (2) where { R ∈ } is an indicator function denoting whether the trade agreement is in the set of agreements with environmental provisions . The term −γ captures the average treatment effect of trade agreements over bilateral trade flows. The term − captures the marginal effect of environmental provisions of trade agreements with respect to the average effect. If it is positive, it means that environmental provisions, on average, boost bilateral trade more than the average trade agreement. The total treatment effect of trade agreements that include environmental provisions is −θ(γ + ), making clear that is a shifter around the average treatment effect. 3 In short, a gravity model is a way to decompose trade flows from a given source s to a given destination d into a source-specific component, a destination- specific component, and trade costs: = + − ⋅, where , are destination and source fixed effects, is some measure of trade costs, and is the trade flow from s to d. The goal is then to consistently estimate the trade elasticity . Most modern trade models deliver a gravity-like relationship, which implies that this empirical strategy is grounded in economic theory and can be used for both causal inference and structural estimation. This is true of essentially all modern trade models.3 The gravity model of trade became popular in empirical trade economics, owing to high predictive power in explaining the patterns of international trade flows. Indeed, economic mass and distance explain most of the variance in international trade flows. 4 By “semi-structural specification”, we mean that we derive a cross-sectional version of the equation above from a structural model with the only difference that the empirical version has source-destination fixed effects and residuals. 5 We estimate the models using the definition of trade costs above as Poisson Pseudo Maximum Likelihood (PPML) regressions. We then calculate a distribution of treatment effects by calculating the predicted treatment effect for each source-destination-year.5 To study the localized impacts of changes in carbon prices, we use exogenous changes in energy prices in the past. From the producers’ perspective, these costs are similar to increasing carbon cost through environmental taxes prevalent in green trade policies since both shocks increase production costs, especially for those using energy-intensive processes. Our key explanatory variable (Δ, ) is the exposure of a given region r to carbon prices, and will be defined as the following: ,,−1 ,−1 Δ, ≡ � ⋅ ⋅ Δ (3) ,−1 ∑ ,−1 where ,,−1 is the labor force in region r, sector i, and period t-1; ,−1 ≡ ∑ ,,−1 is the ,,−1 national labor force in sector i, such that is the share of national labor force of sector i ,−1 that is in region r; ,−1 is the energy (or carbon) intensity per value-added of sector i, such that ∑ ,−1 ∈ (0,1) is the relative energy (or carbon) intensity of sector i; and Δ is the ,−1 change in the international price of a barrel of oil. Provided that changes in price of oil (Δ ) were as good as random relative to local labor market characteristics, we would be able to estimate the causal effect of changes in oil prices over local labor markets by regressing some outcome on the shift-share instrument defined above: ΔOr,t = αt + βΔ, + K ′ r,t−1 δ + εr,t (4) for which estimation of is consistent if �Δ ⋅ ε, �,,−1 , ,−1 , K ′ ,−1� = 0 for every s and r pair. 6 In other words, conditional on controls (K ′ ,−1 ), regional employment and energy (or carbon) intensity, changes in international price of oil are uncorrelated with local labor markets characteristics; that is, the price of oil does not differentially affect local labor markets in a country except through its effects through labor market composition and energy (or carbon) intensity of sectors. Outcomes of interest ΔOr,t include employment variables across different demographic dimension. A spike in oil prices has the potential to reveal how employment responds in rising carbon prices because oil is possibly the leading carbon fuel and input. It can serve as a main proxy for the price of carbon to estimate how labor market outcomes respond to changes in the price of carbon. 5 In simulations, Martínez-Zarzoso (2013) compares the PPML estimator with another method: the Gamma Pseudo-Maximum-Likelihood (GPML), a nonlinear least square (NLS) estimator and a feasible, generalized least squares (FGLS). Confirming previous arguments of Santos Silva and Tenreyro (2006), PPML appears to be less affected by heteroscedasticity. Simulations without zero values show that GPML produces the lowest bias and standard errors. Thus, Martinez-Zarzoso (2013) concludes that the selection of the most appropriate estimator requires a number of tests and depends on the characteristics of each dataset (Santos Silva and Tenreyro, 2006). Head and Mayer (2014) also state that, depending on data structure, there exist cases where GPML is preferable to PPML. 6 For a formal treatment of this identification strategy in shift-share designs, see Borusyak, Hull and Jaravel (2022). 6 3. Data To address the first question, our analysis starts by combining two sources of data: the CEPII BACI gravity dataset and the World Bank’s Handbook of Deep Trade Agreements. The CEPII data include two measures of trade for each country pair in each year (1980-2020). The two measures of trade for each country pair reflect the asymmetry in the direction of trade. For each pair, each country is both an importer and an exporter. The CEPII data include several trade measures of pair-wise trade. One draws upon COMTRADE and the other draws upon the International Monetary Fund (IMF). Additionally, CEPII adds a revised version of these variables based on their analysis of positive trade. The analysis of positive trade is important because trade data are missing for many country pairs. It is not always clear if the missing data reflect positive trade not recorded (“true” missing) or no trade in a given year between two countries (“zero trade” values). In this version, we treat missing trade values as zeros and “square” the data so that we have two observations of trade for each county pair. The dataset we use has 241 importers and 241 exporters for a total of 241*241=58,081 country pair observations for each of 41 years from 1980 to 2020 (inclusive). The working dataset therefore has 2,381,321 observations. To conduct analysis for the second question, we combine multiple data sources and assess local labor market outcomes of energy transitions in Tunisia. Labor market indicators come from the Tunisian Government’s National Survey on Population and Employment (Enquˆete Nationale sur la Population et l’Emploi [ENPE]). The survey is a repeated cross-section of households with its main goal producing statistics regarding the social, educational, and economic characteristics of the population that is either working, unemployed, or out of the labor force. Among other information, the ENPE reports individuals’ employment status by industry (if any), gender, and province for intermittent years between 2006 and 2016. It does not provide data for wages or any other variables of interest. ENPE provides individual level weights that aggregate to representative statistics of the Tunisian population. To construct the labor market indicators used in the empirical part of this paper, we excluded individuals under age 15 and aggregated the data at the district-year level using the provided weights. The emission intensity data at the sectoral level comes from IMF climate dashboard, providing data between 1995 and 2018. We then harmonize industries to merge the sectoral codes in ENPE with emission intensity data from IMF climate dashboard. We also merge energy price data for Tunisia obtained from the CEIC database. This captures the price at which energy is sold to industrial customers, such as manufacturing plants, factories, or large commercial enterprises. We were then able to construct a provincial panel for labor market and price exposure. 4. Results 4.1. Effects of trade-related environmental policies on carbon-intensive industries and trade flows 7 Trends in Trade and Regional Agreements It is important to start this section by indicating that our paper is different from Brandi et al. (2020) in that they focus on how environmental provisions affect green versus non-green trade flows. We focus on the labor-market and overall trade impact of the environmental provisions. Second, to address the collinearity issue we group the provisions as well as we examine them individually. Once we get the initial negative effect, it is important to include all of them because we directly compare them with each other. To address multicollinearity, we combine different provisions, and we try to do that by grouping them into similar categories. The results show that the probability of trade and having an RTA increase over time in the data. Figure 2 shows the adjusted coefficients from probit equations, with dependent variables being a dummy, indicating whether the country pair has positive trade in each year or a dummy variable indicating whether a country pair has an RTA (conditional on having positive trade). Figure 2 shows the rise of the recent wave of globalization in the 1990s as both the probability of trade (the “extensive margin” of trade) and RTAs increased. RTAs continue to become more likely as trade along the extensive margin continues to slowly rise. Figure 2: Rise of the Recent Wave of Globalization Notes: Series reflect the adjusted estimated coefficient on the dummy variable for each year in a probit regression with dependent variable equal to one if a country pair has positive trade or has a regional trade agreement (RTA) in that year and controlling for distance (for the positive trade estimation) and distance and having positive trade (for the RTA estimation). The probit estimation has 2,381,321 observations and omits 1980 as the year variable base value. MENA countries followed these global trends. Figure 3 shows that MENA countries closely follow global trends along the extensive margin. Figure 3 contains the simple average of the share of country pairs with positive trade for all country pairs in which a MENA country is an importer or exporter. Note that the COVID crisis in 2020 is evident in both the MENA and 8 global averages. MENA countries also closely follow global trends in the share of pairs with RTAs. Figure 3: Global Trends in MENA and the Rest of the World Share with Positive Trade .6 .4 .2 0 1980 1990 2000 2010 2020 Year Rest of World MENA Notes: Series reflect the share of all possible country pairs in the data that have positive trade in each year. The number of possible country pairs is constant throughout the date range and missing values for trade in the CEPII data are treated as zeros. Environmental Provisions (EPs) in Trade Agreements In this section, we describe the global pattern of EPs in trade agreements. We contrast MENA’s experience against global averages to identify the source of differences in the relationship between EPs and trade between MENA and the rest of the world. The World Bank’s Handbook of Deep Trade Agreements (HDTA) includes an excellent overview of EPs in trade agreements. Table A.1. includes the list of EPs laid out in the HDTA. These provisions are not equally likely to appear in RTAs. Specifically, Figure 4 shows the mean of a dummy variable equal to “one” if the RTA includes any of the provisions in each category for both the world (excluding MENA) and for MENA countries. The main message is that the coverage of most EP categories is higher for agreements in which MENA participates than agreements in the rest of the world. The clear exceptions are the last two categories (labeled MEA (Multilateral Environmental Agreement) compliance and Participation). This could be because the agreements in which MENA participates are more homogeneous (similar provisions appear in more agreements) or that the agreements in which MENA participates cover more countries. Figure 4: Comparing Coverage-Weighted Incidence of Environmental Provision Categories 9 .8 .6 .4 .2 0 World MENA Goals Balance Enforcement Assistance Areas MEA Compliance Participation Notes: Figures represent the weighted average of a dummy variable equal to one if any of the environmental provisions in each category appear in RTAs for non-MENA countries (“World”) and MENA countries. The categories correspond to Table A.1. The weights are the number of country pairs covered by the RTAs. Country pairs without RTAs are excluded. “MEA” stands for “Multilateral Environmental Agreement.” In both cases, however, these results raise the question of which specific provisions are more likely to appear in the agreements that cover MENA countries. To explore the prevalence of specific provisions, Figure 5 compares the country-pair weighted coverage of each of the specific provisions listed in Table A.1. The diagonal line represents equality in coverage of non-MENA agreements (“World”, measured along the horizontal axis) and MENA agreements (measured along the vertical axis). The goal of Figure 4 is to identify which specific provision is more prevalent in MENA RTAs than in world averages. Several components stand out that represent the broad categories shown in Table A.1. Most of the provisions, however, are much less prevalent in MENA agreements. Exceptions include the General Exception to Other Obligations for Environmental Reasons (8), State-to-State Dispute Settlement (17), Water Management (33), Dumping Hazardous and Toxic Wastes (30), and Preventing Pollution by Ships (24). Note that “specify an objective of environmental protection or sustainable development” is also just slightly above the diagonal line of equality. Since these provisions stand out, it will be important to include these specific provisions to see how they might affect overall estimates of how EPs affect trade. 10 Figure 5: Environmental Provisions in RTAs Environmental Provisions in RTAs 1 General Exception to Other Obligations for Environmental Reasons 8 State-to-State Dispute Settlement 17 .8 2 .6 Prevent Pollution by Ships MENA Prohibits Dumping Hazardous and Toxic Wastes .4 Water Management 6 4 32 33 24 30 25 21 .2 14 5429 535 31 193 181011 35 13 16 45 52 12 27 49 48 3655 4739 382837 26 23 46 9 1 7 15 20 22 34 40 42 43 44 50 5141 0 0 .2 .4 .6 .8 World Notes: Scatterplot represents the simple averages of a dummy variable for each environmental provision listed in Table A.1. that are found in RTAs for non-MENA (“World”) countries and in MENA RTAs for agreements in effect in 2019. The solid line represents equal shares, so points above the line indicate the provisions that appear more frequently in MENA agreements than in agreements in the rest of the world weighted by country coverage. Of course, some of these provisions might be highly correlated because they reflect overlapping concerns of RTA negotiators. In fact, simple correlation coefficients between these provisions are quite high (see Table A.2). For example, the correlation coefficient between Provision 24 (Preventing Pollution by Ships) correlation with Provisions 30 and 33 is over 65 percent, and correlation between Provision 30 (Hazardous and Toxic Waste) and Provision 33 (Water Management) is over 76 percent. Table 1 presents the results from the gravity model for the list of countries who signed a trade agreement with an EP between 1980-2020. The provisions are binary indicators of having any provision within the seven main Categories in Table A.1. First, the analysis reveals that RTAs generally increase trade flows, evidenced by the positive and highly significant coefficient for RTAs in both regression tables. In Table 1, the RTA coefficient of 0.235 suggests that RTAs facilitate trade by reducing barriers and fostering economic cooperation among member countries. This finding aligns with the broader literature, which consistently demonstrates the trade-promoting effects of RTAs. However, inclusion of EPs within these agreements introduces a more complex dynamic. The marginal effect of having any environmental provision, as indicated by the negative coefficient of - 0.118 in Table 1, implies that while RTAs boost trade overall, the addition of EPs can reduce trade flows. This reduction is likely due to increased compliance costs and regulatory burdens associated with these provisions. When examining repercussions of specific EP 11 categories, the results vary considerably, indicating that different types of EPs affect trade in distinct ways: • Category I provisions, which have a coefficient of 0.067, correlate with significant increases in trade flows. This suggests that general commitments to the agreement's overall environmental goals—such as specifying objectives for environmental protection and sustainable development, promoting high levels of environmental protection, and encouraging regulatory cooperation—have a beneficial effect on trade. These provisions also ensure the right to regulate environmental issues is preserved. • Category II provisions, on the other hand, with a coefficient of -0.114, are associated with a decrease in trade flows. This category focuses on balancing environmental with trade and investment goals, including provisions for environmental exceptions in trade and investment chapters, prohibiting the dilution of environmental protections to promote trade or investment, and promoting liberalization of trade in environmental goods. • Category III provisions, with a positive coefficient of 0.083, are associated with increased trade flows. This category outlines the requirements for maintaining judicial or administrative proceedings for enforcing environmental regulations and subjecting environmental provisions to state-to-state dispute settlements. • Category IV, with the highest positive coefficient of 0.168, highlights significant positive effects from external assistance provisions, which include technical, financial, and capacity-building support specifically in the environmental area. • Category V provisions, in contrast, have a coefficient of -0.266, indicating substantial negative repercussions on trade flows. This category likely includes general environmental protection areas covering various specific obligations, such as controlling ozone-depleting substances, preventing pollution by ships, fisheries management, conservation of marine species, and promoting renewable energy. • Category VI provisions, with a coefficient of 0.076, appear to have little effect on trade flows, suggesting that compliance with multilateral environmental agreements (MEAs) like CITES, the Montreal Protocol, and the Basel Convention can be significant enough to influence trade dynamics. • Lastly, Category VII provisions, with a coefficient of -0.352, are associates with strong negative effects on trade flows. This category includes establishing intergovernmental committees on the environment, facilitating civil society involvement, and ensuring transparency and private rights in environmental provisions. The mixed results highlight the complexity of integrating EPs into trade agreements, where some provisions support trade by enhancing environmental performance, while others may hinder it due to increased regulatory burdens. 12 Table 1: PPML HDFE Full Sample Estimates (1) (2) VARIABLES Trade Trade RTA 0.198*** 0.235*** (0.022) (0.023) I. Environmental Goals/Objectives 0.067** (0.032) II. Balance between Environmental and Trade/Investment Goals -0.114*** (0.040) III. Enforcement Mechanism 0.083* (0.042) IV. External Assistance 0.168*** (0.022) V. General Environmental Protection Areas -0.266*** (0.026) VI. MEA Compliance 0.076** (0.036) VII. Participation in Promoting Environmental Objectives -0.352*** (0.039) Any Environmental Provision = 1 -0.118*** (0.022) Constant 16.220*** 16.200*** (0.009) (0.009) Observations 1,724,223 1,724,223 Notes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Each Roman numeral category represents a dummy variable equal to one if the regional trade agreement (RTA) includes any of the provisions listed in that category in Table 1. “PPML HDFE” stands for Poisson Pseudo Maximum Likelihood High Dimensional Fixed Effects” following Correia et al. (2019a, 2019b). Analysis of the HDFE PPML regression results reveals significant insights into the interplay between RTAs and EPs, with a particular focus on the MENA region (Table 2 and Table 3). RTAs generally promote trade by reducing barriers and fostering cooperation between member countries, as evidenced by the positive coefficients in both regressions (0.329 and 0.295). However, the negative coefficients for the MENA RTA variable (-0.182 and -0.1) indicate that MENA countries do not experience the same level of benefits from these agreements as other regions do. This discrepancy suggests that unique economic, political, or infrastructural challenges in the MENA region might impede full RTA potential. Contrastingly, EPs—typically associated with a reduction in trade flows due to increased compliance costs and barriers—are distinctly positive in terms of increasing MENA countries’ trade flows. The negative coefficients for the general EPs variable (-0.252 and -0.227) highlight the trade-reducing effect of these provisions overall. Yet, the positive coefficients for the MENA EP interaction term (0.492 and 0.951) suggest that MENA countries benefit from these provisions. This counterintuitive finding could be attributed to enhancement of MENA countries' reputations and alignment with global environmental standards, making their products more attractive in international markets. 13 The combined effects underscore a critical insight: while RTAs might not be as advantageous for MENA countries, the inclusion of EPs within trade agreements can mitigate these shortcomings and even promote trade for the region. This highlights the potential for tailored policy interventions that incorporate environmental standards to boost trade competitiveness for MENA countries. Policy makers should explore specific RTAs involving MENA countries and the unique terms within these agreements to understand and address the barriers preventing full benefits. Moreover, a detailed examination of the EPs that drive increases in MENA trade is crucial for formulating strategies to leverage these provisions and enhance market access and trade flows. Table 2: PPML HDFE Estimates of MENA and Environmental Provisions in RTAs (1) (2) (3) VARIABLES Trade Trade Trade RTA 0.226*** 0.207*** 0.261*** (0.027) (0.025) (0.026) MENA x RTA -0.057 -0.012 0.112*** (0.043) (0.042) (0.039) I. Environmental Goals/Objectives -0.153*** 0.046 (0.027) (0.035) II. Balance between Environmental and Trade/Investment Goals -0.151*** (0.048) III. Enforcement Mechanism 0.092** (0.045) IV. External Assistance 0.150*** (0.023) V. General Environmental Protection Areas -0.302*** (0.025) VI. MEA Compliance 0.127*** (0.039) VII. Participation in Promoting Environmental Objectives -0.353*** (0.041) MENA x PI 0.511*** -0.064 (0.043) (0.074) MENA x PII 0.307*** (0.059) MENA x PIII -0.536*** (0.077) MENA x PIV 0.150** (0.073) MENA x PV 0.651*** (0.065) MENA x PVI 0.175** (0.080) MENA x VII -0.126 (0.085) I. Environmental Goals/Objectives MENA x PI Does the agreement require states to prevent pollution by ships? -0.278*** 14 (0.037) Any Environmental Provision = 1 -0.164*** (0.028) MENA x Any Environmental Provision 0.290*** (0.037) Constant 16.211*** 16.214*** 16.188*** (0.009) (0.009) (0.009) Observations 1,724,223 1,724,223 1,724,223 Notes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. . Each Roman numeral category represents a dummy variable equal to one if the regional trade agreement (RTA) includes any of the provisions listed in that category in Table 1. “PPML HDFE” stands for Poisson Pseudo Maximum Likelihood High Dimensional Fixed Effects” following Correia et al. (2019a, 2019b). Since multi-collinearity can affect coefficient estimates, Table 3 shows various specifications that remove some of the highly correlated provisions. Two main results emerge from Table 3. The first is that the results are somewhat sensitive to specification, as expected when multicollinearity is present. MENA’s RTA experience is generally positive, as shown in columns 1 and 5. Moving from left to right, column 1 contains the results with RTA, the MENA x RTA interaction, and specific environmental provisions highlighted in Figure 4. Estimates of the relationship between pair-wise trade and Provisions 8 and 24 are consistently significantly negative for the full sample. Other estimates (for example, provision 30, which has correlation coefficients among the highest in Table A.2.) change sign, magnitude, and significance. When we restrict the sample to just country pairs that include at least one MENA country (either as exporter or importer), these results change significantly. In particular, the individual estimates become positive and only provision 8 is statistically significant. This implies that the implementation of these provisions for MENA countries do not discourage trade in the same way that they do for the rest of the world. These results also help explain the result in Table 2 showing that environmental provisions are associated with more trade, rather than less, for MENA countries. Table 3: PPML HDFE Estimates of Specific Provisions and Trade Flows (1) (2) (3) (4) (5) Controllin MENA VARIABLES Trade Trade Trade g for MC only RTA 0.164*** 0.175*** 0.159*** 0.162*** 0.065 (0.023) (0.022) (0.022) (0.021) (0.040) MENA x RTA 0.141*** (0.037) 8. Does the agreement -0.116*** -0.121*** -0.103*** -0.079*** 0.261*** provide for a general exception to other (0.037) (0.038) (0.037) (0.022) (0.033) obligations for environmental reasons? 17. Does the agreement 0.027 0.030 0.035 subject environmental provisions to general state (0.041) (0.041) (0.042) to state dispute settlement? 24. Does the agreement -0.324*** -0.322*** -0.299*** -0.302*** 0.037 require states to prevent (0.037) (0.037) (0.040) (0.039) (0.092) pollution by ships? 15 30. Does the agreement -0.017 -0.021 0.129*** 0.135*** 0.085 prohibit of dumping hazardous and toxic (0.031) (0.031) (0.031) (0.030) (0.092) wastes? 33. Does the agreement 0.296*** 0.304*** require states to implement water management? (0.028) (0.028) 14.617** Constant 16.228*** 16.228*** 16.232*** 16.231*** * (0.009) (0.009) (0.009) (0.008) (0.013) Observations 1,724,223 1,724,223 1,724,223 1,724,223 336,700 Notes: Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. RTA stands for regional trade agreement. Numbers in the list of variables correspond to Table 1. 4.2. Local labor market impacts of carbon prices To address this question, we specifically focus on Tunisia as it heavily relies on imported fossil fuels for its energy needs, and a move towards decarbonization, diversification, and adaptation to climate change can have several effects. Owing to data constraints, we have added relevant and possible descriptives. Decarbonization efforts can promote development of renewable energy sources such as solar and wind power. Through March 2022, Tunisia had about 472 MW of installed renewable energy capacity, of which 244 MW was wind power, 166 MW solar power, and 62 MW hydroelectric power, representing a combined 8 percent of national energy production capacity. The country aims to raise the usage of renewable energy resources to 35 percent of total power capacity by 2030. This shift can lead to a more diversified and resilient energy sector, reducing dependency on fossil fuel imports and promoting energy security. The renewable energy sector can create new job opportunities, ranging from manufacturing and installation of renewable energy infrastructure to research and development of clean technologies. These green jobs can contribute to economic growth, reduce unemployment, and enhance skills and expertise in the workforce. However, predicting the overall employment and welfare consequences of this energy shift in Tunisia is challenging. Tunisia's economy has traditionally been reliant on sectors such as agriculture, textiles, and tourism. Tunisia’s regions vary in in terms of exposure to energy intensity, which suggests that changes in energy prices will affect employment across the country unevenly. For example, Table 4 shows that urban and industrial hubs like Tunis (1,125.23), Sfax (914.27), and Nabeul (871.18) exhibit the highest energy exposure in the country measured in terms of CO2 emissions per unit of output, weighted by the share of each district in total industry employment. These areas are highly industrialized, particularly in manufacturing, which contributes heavily to overall emissions. In contrast, districts with moderate CO2 emissions like Gafsa (246.97), Gabes (275.05), and Beja (265.67) likely rely on sectors such as agriculture, mining, and light manufacturing, which generate fewer emissions. Low-emission districts, including Jendouba (70.50), Tozeur (95.65), and Tataouine (103.51), tend to be less industrialized or depend on lower- 16 impact industries such as agriculture. The variation in emissions intensity suggests that districts with higher shares of industrial activity, like Tunis and Sfax, are more vulnerable to employment disruptions from rising energy prices, while less developed regions with smaller industrial bases may be less affected or may benefit from energy-efficient practices. The context described above serve as a strong motivation to pursue a study on the labor and distributional effects of transitioning from a carbon-intensive production system to a non-carbon intensive one. This situation in Tunisia offers the opportunity to expand the scope of the literature by studying localized employment impacts of decarbonization. Table 4: Energy intensity exposure, by district 2016 district energy intensity exposure Jendouba 70.50 Tozeur 95.65 Tataouine 103.51 Kebili 121.98 Zaghouan 168.55 Siliana 180.05 Le Kef 230.74 Gafsa 246.97 Beja 265.67 Gabes 275.05 Sidi Bouzide 294.35 Kasserine 305.91 Manouba 373.73 Mahdia 377.60 Mednine 397.77 Kairouan 443.54 Bizerte 490.39 Monastir 583.54 Ben Arous 637.82 Ariana 639.42 Sousse 641.47 Nabeul 871.18 Sfax 914.27 Tunis 1,125.23 Note: This represents regional energy exposure district wise in 2016. The index is defined as CO2 Emissions per $1M Output, Weighted by Industry Employment. The regression results following the methodology outlined in the earlier section demonstrate a significant relationship between changes in energy exposure due to carbon prices and employment outcomes, revealing how increased carbon costs influence labor market dynamics across various demographic and educational groups (see Table 5). 17 Column 1 reports estimates for the full sample, while Columns 2 and 3 for male and female workers, respectively. The findings indicate that overall employment declines in regions with higher exposure to energy price increases, with the price exposure index showing a significant negative effect of −0.00503. Male employment is particularly affected, with a significant negative coefficient of −0.00442, indicating that male workers experience more substantial employment reductions due to rising energy prices. Female employment also experiences a negative impact (−0.00443), although the decline in employment is less significant. This suggests that while both genders are adversely affected, employment losses are more pronounced for male workers. Table 5: Results (1) (2) (3) Full Sample Male Female Price exposure --- - - index 0.00503*** 0.00442*** 0.00443* (0.00149) (0.00154) (0.00225) Observations 120 120 120 R-squared 0.154 0.150 0.164 Year FE Yes Yes Yes District FE Yes Yes Yes Individual Controls Yes Yes Yes District cluster robust standard errors in parenthesis. * p < 0.1; ** p < 0.05; *** p < 0.01 5. Conclusion Achieving sustainability in the energy sector is expected to lead to a net increase in jobs worldwide by 2030 compared to a business-as-usual path. This employment growth is largely due to the greater labor intensity of renewable energy production relative to fossil fuel-based power generation. However, this positive trend comes with substantial regional disparities. While regions such as the Americas, Asia, and Europe are expected to experience net job creation, the Middle East and Africa may face net job losses, estimated at -0.48% and -0.04%, respectively, if their economic structures remain unchanged (ILO-World Employment and Social Outlook series: Greening with jobs). Policy interventions are thus essential to mitigate these potential job losses and address the negative impacts. For MENA countries, adapting policies to encourage green transition is not just beneficial but necessary to ensure sustainable employment growth and reduce reliance on fossil fuels. 18 This paper contributes to a growing body of literature on the intersection of environmental policy, trade, and labor markets. The paper offers new insights into how environmental provisions (EPs) in trade agreements and rising carbon prices affect trade flows and labor outcomes, particularly in the MENA region and Tunisia. On the first front—environmental policy and trade—the study highlights the dual role of EPs within regional trade agreements (RTAs). While RTAs generally boost trade by reducing barriers and fostering cooperation, the inclusion of EPs introduces significant complexity. Our analysis reveals that effects on trade flows vary significantly depending on the exact kind of EP: provisions that support broad environmental goals and enforcement mechanisms can enhance trade by aligning countries with global sustainability standards; but more stringent provisions—especially those that impose costly environmental regulations—tend to reduce trade flows, particularly for carbon-intensive industries. These mixed effects underscore the need for careful design and implementation of EPs within RTAs. The findings suggest that policy makers must strike a balance between advancing environmental protection and ensuring that these policies do not disproportionately hinder trade, particularly for developing countries in the MENA region. Importantly, the study shows that MENA countries can leverage EPs to enhance their reputation and align with global environmental standards, turning potential trade barriers into opportunities for sustainable trade growth. The second key contribution of this study is its focus on the labor market consequences of carbon pricing and energy transition, using Tunisia as a case study. Rising carbon prices, driven by both market forces and environmental regulations, have led to employment losses in fossil fuel-dependent regions. Our findings show that these shifts can disproportionately lead to job losses for men. This highlights the pressing need for targeted social safety nets, ensuring they are not left behind as economies decarbonize. Moreover, while the shift to renewable energy presents opportunities for new job creation in “green” sectors, policies are needed to facilitate worker retraining and reskilling for those displaced by the decline of carbon- intensive industries. Without such measures, the transition to a “greener” economy may exacerbate inequality and lead to further labor market disruptions. In conclusion, this study underscores the complex interplay between trade, environmental policy, and labor market outcomes. For policy makers, the challenge lies in designing EPs and carbon pricing mechanisms that balance environmental sustainability with economic and social objectives. References Antweiler, W., Copeland, B. R., and Taylor, M. S. (2001). Is free trade good for the environment? Am. Econ. Rev. 91, 877–908. doi: 10.1257/aer.91.4.877 Berger, A., Brandi, C., Morin, J. F., & Schwab, J. (2020). The trade effects of environmental provisions in preferential trade agreements. International Trade, Investment, and the Sustainable Development Goals, 111-139. 19 Can, M., Ahmed, Z., Mercan, M., and Kalugina, O. A. (2021). The role of trading environment- friendly goods in environmental sustainability: does green openness matter for OECD countries? J. Environ. Manage. 295:113038. doi: 10.1016/j.jenvman.2021.113038 Cole, M. A., and Elliott, R. J. (2003). Determining the trade-environment composition effect: the role of capital, labor and environmental regulations. J. Environ. Econ. Manage. 46, 363– 383. doi: 10.1016/S0095-0696(03)00021-4 Cole, M. A., Elliott, R. J., and Zhang, J. (2011). Growth, foreign direct investment, and the environment: evidence from Chinese cities. J. Region. Sci. 51, 121–138. doi: 10.1111/j.1467- 9787.2010.00674. Copeland, B. R., and Taylor, M. S. (1994). North-South trade and the environment. Q. J. Econ. 109, 755–787. doi: 10.2307/2118421 Correia, S., Guimarães, P., Zylkin, T. (2019). "fpmlhdfe: Fast Poisson Estimation with High- Dimensional Fixed Effects", 2019; arXiv:1903.01690. Correia, S., Guimarães, P., Zylkin, T. (2019). "Verifying the existence of maximum likelihood estimates for generalized linear models". arXiv:1903.01633. Dai, Z., Zhang, Y., & Zhang, R. (2021). The impact of environmental regulations on trade flows: A focus on environmental goods listed in APEC and OECD. Frontiers in Psychology, 12, 773749. De Melo, J., and Solleder, J.-M. (2020). Barriers to trade in environmental goods: how important they are and what should developing countries expect from their removal. World Dev. 130:104910. doi: 10.1016/j.worlddev.2020.104910 Frankel, J. A., and Rose, A. K. (2005). Is trade good or bad for the environment? Sorting out the causality. Rev. Econ. Stat. 87, 85–91. doi: 10.1162/0034653053327577 Grossman, G. M, and Krueger, A. B. (1991). Environmental Impacts of a North American Free Trade Agreement. NBER working paper 3914. doi: 10.3386/w3914 Mattoo, Aaditya, Nadia Rocha, and Michele Ruta, eds. 2020. Handbook of Deep Trade Agreements. Washington, DC: World Bank. doi:10.1596/978-1-4648-1539-3. License: Creative Commons Attribution CC BY 3.0 IGO. McGuire, M. C. (1982). Regulation, factor rewards, and international trade. J. Public Econ. 17, 335–354. doi: 10.1016/0047-2727(82)90069 Jug, J., and Mirza, D. (2005). Environmental regulations in gravity equations: evidence from Europe. World Econ. 28, 1591–1615. doi: 10.1111/j.1467-9701.2005.00748. Pethig, R. (1976). Pollution, welfare, and environmental policy in the theory of comparative advantage. J. Environ. Econ. Manage. 2, 160–169. doi: 10.1016/0095-0696(76)90031-0 20 Poncet, S., Hering, L., and Sousa, J. (2015). Has Trade Openness Reduced Pollution in China? Working Papers 2015-11, CEPII Research Center. Tamini, L. D., and Sorgho, Z. (2018). Trade in environmental goods: evidences from an analysis using elasticities of trade costs. Environ. Resour. Econ. 70, 53–75. doi: 10.1007/s10640-017- 0110-2 Tobey, J. A. (1990). The effects of domestic environmental policies on patterns of world trade: an empirical test. Kyklos 43, 191–209. doi: 10.1111/j.1467-6435.1990.tb00207. Van Beers, C., and Van Den Bergh, J. C. (1997). An empirical multi-country analysis of the impact of environmental regulations on foreign trade flows. Kyklos 50, 29–46. doi: 10.1111/1467-6435.00002 Xu, X. (2000). International trade and environmental regulation: time series evidence and cross section test. Environ. Resour. Econ. 17, 233–257. doi: 10.1023/A:1026428806818 Zugravu-Soilita, N. (2019). Trade in environmental goods and air pollution: a mediation analysis to estimate total, direct and indirect effects. Environ. Resour. Econ. 74, 1–38. doi: 21 Annex A Table A.1: List of Environmental Provisions Code Description 1 I. Environmental Goals/Objectives Does the agreement specify an objective of environmental protection or sustainable 2 development? 3 Does the agreement specify an objective of high levels of environmental protection? 4 General obligation of environmental cooperation Does the agreement call for regulatory cooperation or harmonization in environmental 5 regulation? 6 Does the agreement preserve the right to regulate in the environment? 7 II. Balance between Environmental and Trade/Investment Goals Does the agreement provide for a general exception to other obligations for environmental 8 reasons? 9 Does the investment chapter provide for an environmental exception? 10 Does the agreement prohibit dilution of environmental protection to promote trade? 11 Does the agreement prohibit dilution of environmental protection to promote investment? Does the agreement provide for differential and greater liberalization of trade in 12 environmental goods? Does the agreement require states to take science into account when preparing and 13 implementing environmental regulation? Does the agreement require states to prepare an environmental impact assessment of the 14 PTA? 15 III. Enforcement Mechanism Does the agreement require states to maintain judicial or administrative proceedings for 16 enforcement of environmental regulation? Does the agreement subject environmental provisions to general state to state dispute 17 settlement? 18 Does the agreement provide special environmental state to state dispute settlement? Does the agreement provide international remedies of compensation or retaliation for 19 violation of environmental provisions? 20 IV. External Assistance Does the agreement provide for technical assistance/financial assistance/capacity building 21 specifically in the environmental area? 22 V. General Environmental Protection Areas 23 Does the agreement require states to control ozone-depleting substances? 24 Does the agreement require states to prevent pollution by ships? 25 Does the agreement require states to implement fisheries management? Does the agreement require states to take measures for conservation of specified marine 26 species? 27 Does the agreement provide for differential restriction of fishing subsidies? Does the agreement require states to protect wild fauna and flora at risk and/or prevent 28 illegal trade in these species? Does the agreement require measures to prevent deforestation and/or require sustainable 29 trade practices in forest products? 30 Does the agreement prohibit of dumping hazardous and toxic wastes? Does the agreement require states to promote and protect biodiversity (including 31 indigenous-traditional knowledge)? 32 Promotion of renewable energy and improving energy efficiency 33 Does the agreement require states to implement water management? 34 VI. MEA Compliance 22 35 Does the agreement require states to comply with MEAs generally? 36 Does the agreement specify supremacy of MEA obligations over PTA obligations? 37 Does the agreement require states to comply with CITES? Does the agreement require states to comply with the Montreal Protocol on Ozone- 38 Depleting Substances? Does the agreement require states to comply with the Basel Convention on the Control of 39 Transboundary Movements of Hazardous Wastes and Their Disposal? Does the agreement require states to comply with the Inter-American Tropical Tuna 40 Convention 41 Does the agreement require states to comply with MARPOL? Does the agreement require states to comply with the International Convention on the 42 Regulation of Whaling? 43 Does the agreement require states to comply with the Ramsar Convention on Wetlands? Does the agreement require states to comply with Convention on Conservation of 44 Antarctic Marine Living Resources? Does the agreement require states to comply with the UN Fish Stocks Agreement, the FAO Code of Conduct for Responsible Fisheries, the 1993 FAO Agreement to Promote 45 Compliance with International Conservation and Management Measures by Fishing Vessels on the High Seas (Compliance Agreement) and the 2001 IUU Fishing Plan of Action/IUU measures in general? Does the agreement require states to comply with the 2005 Rome Declaration on IUU Fishing, the Agreement on Port State Measures to Prevent, Deter and Eliminate Illegal, 46 Unreported and Unregulated Fishing, 2009, as well as instruments establishing and adopted by Regional Fisheries Management Organisations? Does the agreement require states to comply with the UN Convention on the Law of the 47 Seas (1982)? Does the agreement require states to comply with the UN Conference on Environment and 48 Development (UNCED)? 49 Does the agreement require states to comply with the UN Environment Program (UNEP)? 50 Does the agreement require states to comply with the International Energy Program? 51 VII. Participation in Promoting Environmental Objectives 52 Does the agreement establish an intergovernmental committee on environment? Does the agreement require states to facilitate civil society involvement and/or establish a 53 forum on trade and environment? 54 Does the agreement include special obligations of transparency in the environmental field? Does the agreement include private rights to make submissions regarding environmental 55 provisions? Table A.2: Correlation Coefficients of Specific Provisions Provision 8 Provision 17 Provision 24 Provision 30 Provision 33 8. Does the agreement provide for a general exception to other obligations for environmental reasons? 1 17. Does the agreement subject environmental provisions to general state to state dispute settlement? 0.6974 1 24. Does the agreement require states to prevent pollution by ships? 0.3125 0.406 1 23 30. Does the agreement prohibit of dumping hazardous and toxic wastes? 0.3685 0.4756 0.7344 1 33. Does the agreement require states to implement water management? 0.4516 0.58 0.6505 0.7614 1 24 Annex B: Can a carbon adjustment tariff be effective? Policies that induce a change in the price of carbon will play a decisive role in the energy transition. Out of this menu, one possibility is a carbon-adjustment tariff or a global tariff on carbon. While different in their minutia, these policies share the fact that they induce changes in trade flows as a response to trade costs. One can measure the responsiveness of trade flows to these policies through the elasticity of trade of carbon intensive products. This object measures the marginal response of trade flows between two countries to a change in the cost of trading between these countries, induced by a policy instrument such as a tariff. A higher (lower) trade elasticity implies that trade flows will change more (less) in response to a change in tariffs. A carbon adjustment tariff would only be effective at moderate costs if consumers are sufficiently responsive to them, adjusting their sourcing decisions away from carbon- intensive sources after the implementation of the policy. By the same token, if the trade elasticity is too low, a decrease in carbon intensive goods can only be achieved by imposing higher welfare costs.7 In this box, we take a first step towards estimating the trade elasticity of carbon-intensive goods. In doing so, our goal is twofold. First, we want to inform policy. 8 Second, we extend the scientific literature by estimating heterogeneous trade elasticities at a more disaggregated level than available in existing studies (Boehm, Levchenko and Pandalai- Nayar, 2023). We will apply the methodology recently developed by BLP to estimate the trade elasticity at different horizons. In short, the identification strategy exploits the plausibly exogenous time-variation of changes in Most Favored Nations (MFN) tariffs, using countries benefiting from preferential tariffs as the control group. The estimation strategy builds up from standard gravity regressions but extends it to a dynamic framework. The details of the methodology can be found in Annex C. We compare the average elasticity across all products with the average elasticity of a sample of products included in the Carbon Border Adjustment Mechanism (CBAM) of the European Union. The EU’s CBAM includes a list of “carbon intensive and at most significant risk of carbon leakage: cement, iron and steel, aluminium, fertilisers, electricity and hydrogen.” We isolated these products using their HS Codes and estimated the trade elasticity for a full sample of products (replicating BLP) as well as for a subset of carbon intensive products. Our point estimates suggest that the short-run trade elasticity of carbon-intensive products (-0.08) is close to the overall trade elasticity (-0.27) but its long-run trade elasticity (-5.84) is much higher than the overall long-run trade elasticity (-1.30). If these point estimates reflect the true parameter an implication would be that carbon-intensive products are very 7 In a large class of structural trade models, if one has access to those two summary statistics, then, for a given country d, one 1 can denote changes in welfare after a 1% change in trade costs as: � � is the change in the share of sales � ) where = ( that are spent domestically and < is the trade elasticity. Note that changes in welfare are decreasing in changes in domestic trade share and that this effect is controlled by the trade elasticity. For an equal change in the domestic trade share, if products have equal initial budget allocations, products with lower trade elasticities will contribute more to welfare changes. 8 The “trade elasticity”, which can be understood as the responsiveness of trade flows to trade barriers, is a central object in trade policy. It is also relevant where trade and climate change intersect. For instance, potential policies like carbon tariffs are only feasible if the trade elasticity on carbon-intensive goods is high enough. Furthermore, the trade elasticity is typically expected to be lower in the short run than in the long run, since firms are usually more capable of substituting away from inputs over a longer time horizon. To use the same policy example, this would imply that to induce a constant level of decarbonization, carbon tariffs should be higher over the short run. 25 responsive to tariffs over the long-run and policies like a carbon-adjustment tariff can be effective over the medium run at moderate welfare costs. However, while the long-run elasticity is statistically significant, confidence intervals of the CBAM sample estimates are very wide, covering between -11.5 to -0.16 at a 5% significance level. Therefore, we interpret this as initial suggestive evidence that a carbon-adjustment tariff can be effective but emphasize that further research is necessary with more statistical power is necessary to have a conclusive answer. 5 Trade elasticity 3 1 -1 -3 -5 -7 95% confidence interval -9 -11 CBAM Sample -13 Full Sample -15 0 1 2 3 4 5 6 Horizon 26 Annex C: Let ,, denote exports from s to d of product p in period t, while let ,, denote ad-valorem tariffs, using the same notation. At any horizon h, the elasticity of trade is defined as: ,,ℎ − ,,−1 ℎ ≡ (1) ,,ℎ − ,,−1 Estimation consists of a two-stage least square strategy, where the second stage uses all variation in tariffs: ,,ℎ − ,,−1 = βℎ � ,,ℎ − ,,−1 � + δ,,ℎ + δ,,ℎ + ,,ℎ (2) ℎ ∈ {, + 1, … } while the first stage uses one instrument for changes in tariff rates, which can be endogenous: � ,,ℎ − ,,−1 � = βℎ τ � ,, − ,,−1 � + δτ ,,ℎ + δτ ,,ℎ + τ ,,ℎ (3) Following BLP, we will construct the instrument in the following way: � ,, − ,,−1 � ≡ �,, = ,, � × �,,−1 = ,,−1 � × � ,, − ,,−1 � (4) together with the restriction that observations are dropped if: �,, = ,,, � × �,,−1 = ,,−1 � = 1 and ( − 1 ) + ( − 1 ) + ( ) + ( ) > 0 The intuition behind the construction of this instrument is the following. When changing its MFN tariffs, a country might be targeting one of its major trading partners. However, exactly because of that, for all third parties affected by the MFN tariff the timing of the policy change is as good as random and can be used as a treatment to identify the trade elasticity compared to countries unaffected by MFN policy changes. Our data entails UNCOMTRADE bilateral trade data as well as TRAINS tariff data (both tariff and preferential). After constructing a dataset on bilateral tariffs and identifying which countries are beneficiaries of preferential tariffs, we need to superimpose the restrictions mentioned above to construct the instrument described in (6). Finally, once the instrument is constructed, we can use UNCOMTRADE trade data to estimate a sequence of dynamic local projections impulse response functions (Jordà, 2005) using the 2SLS strategy described by equations (X3) and (X4). 27