Policy Research Working Paper 10323 Pursuing Environmental and Social Objectives through Trade Agreements Joseph Francois Bernard Hoekman Miriam Manchin Filippo Santi Development Research Group & Macroeconomics, Trade and Investment Global Practice February 2023 Policy Research Working Paper 10323 Abstract Using a data set covering more than 120 countries spanning investigates linkages between non-trade provisions and offi- several decades, this paper employs a synthetic difference in cial development assistance. The analysis finds no evidence difference estimator to study whether non-trade provisions that provisions related labour or civil rights improved the on labor standards, environmental protection and civil and associated outcome indicators, while evidence on environ- human rights in trade agreements yield improvements in mental outcomes is mixed. Official development assistance corresponding indicators. The paper distinguishes between is significantly greater with non-binding environmental and binding (enforceable) and non-binding provisions and civil rights provisions, but not with labor standards. This paper is a product of the Development Research Group, Development Economics and the Macroeconomics, Trade and Investment Global Practice. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at joseph.francois@gmail.com. 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 Pursuing Environmental and Social Objectives through Trade Agreements ∗ Joseph Francois, Bernard Hoekman, Miriam Manchin, and Filippo Santi Keywords: Trade agreements, non-trade objectives, binding commitments, development cooperation, environment, labor standards, civil rights JEL codes: F13, F16, F17, F18, F35, O19 ∗ corresponding author: joseph.francois@gmail.com We are grateful to Ryan Abman, Keith Maskus, Michele Ruta, Luca Salvatici, Anirudh Shingal, Alissan- dra Tucci, and participants in the conference on Deep Trade Agreements: Effects Beyond Trade (World Bank), the World Trade Forum 2022, the 2022 Galbino summer school (aka the Galbino Project, Univer- sity of Bern), the 63rd annual conference of the Italian Economic Association and workshops at the EUI, Sapienza University of Rome and the University of Turin for comments on earlier drafts. This paper has benefited from support from the World Bank’s Umbrella Facility for Trade trust fund financed by the governments of the Netherlands, Norway, Sweden, Switzerland and the United Kingdom, and builds on research supported by the European Union Horizon 2020 research and innovation program (RESPECT). Joseph Francois, World Trade Institute, University of Bern, Switzerland and CEPR Bernard Hoekman, Robert Schuman Centre, European University Institute, Florence Italy and CEPR Miriam Manchin, University College London and Politecnico di Milano, Italy Filippo Santi, University of Turin and European University Institute 1 Introduction The number of preferential trade agreements (PTAs) has been increasing steadily since the 1980s, as has the number of signatories and coverage of non-trade policy areas in such agreements (Dür et al., 2014; Lechner, 2018; Mattoo et al., 2020). Non-trade provisions (NTPs) are a prominent feature of the more recent PTAs concluded by the EU and the US, as well as by many other OECD member countries. They pertain to such matters as environmental protection, labor standards and human rights and have become central to political support for PTA negotiations (Van den Putte, 2015; Lechner, 2016). The increasing prevalence of NTPs in PTAs in recent decades is accompanied by extensive heterogeneity in national non-trade outcome indicators. As discussed further in Section 2. The direction of change in some non-trade outcome indicators associated with NTPs has often been negative. Given the prominence of non-trade issues in recent debates on globalization in general and trade agreements in particular, whether the inclusion of NTPs in PTAs supports the realisation of underlying non-trade policy objectives, i.e. they actually work to promote climate and socioeconomic goals, is a fundamental policy issue. This is the primary research question analyzed in this paper – does inclusion of NTPs in PTAs improve associated non-trade outcomes? We address this question by exploiting new data on trade agreement coverage (Mattoo et al., 2020) and the synthetic difference-in-difference (SDID) estimator recently proposed by Arkhangelsky et al. (2021) to address endogeneity concerns. We extend the literature by allowing for the potential effectiveness of NTPs to depend on the nature of the commitments made. Specifically, we distinguish between NTPs depending on whether they take the form of binding, legally enforceable commit- ments or non-binding (non-enforceable) soft law provisions that only provide a framework for policy dialogue on the matters covered.1 In both cases, NTPs provide a potential focal point for technical and financial assistance to support cooperation to improve associated 1 The literature on the effectiveness of binding vs. non-binding provisions in treaties (and the asso- ciated institutional conflict resolution mechanisms) goes well beyond PTAs. For example, McLaugh- lin Mitchell and Hensel (2007) find that peaceful resolution of disputes between states is more likely if pertinent treaties establish binding conflict resolution mechanisms serviced by an international organi- zation. 1 non-trade outcome indicators. By explicitly considering both the enforceability of NTPs and the relationship between NTPs and ODA, we contribute to the policy debate on the appropriate design of NTPs in PTAs. Some proponents of NTPs argue that to be effective, commitments should be binding and enforceable through dispute settlement procedures (Bronckers and Gruni, 2021). Others argue that soft law provisions that are not subject to dispute settlement are more likely to improve outcomes of interest insofar as they involve (support) a process of active engagement and cooperation between governments and concerned stakeholder groups, especially if accompanied by official development assistance (ODA) projects and programs targeting specific non-trade goals.2 Most studies in international economics do not consider non-trade outcomes, instead largely relying on structural gravity models to analyze the trade effects of PTAs in general or specific PTA provisions.3 Robust empirical evidence on the effects of NTPs on associ- ated non-trade outcomes that control for endogeneity is both relatively scarce and limited in scope. The extant empirical evidence is mixed, context- and indicator-specific, often failing to apply methods that permit identification of a causal relationship between PTAs and non-trade outcomes.4 Research that seeks to identify causal relationships between NTPs and non-trade outcomes often focuses on a specific non-trade area. Examples in- clude Baghdadi et al. (2013), Abman and Lundberg (2020), and Abman et al. (2022) on environmental and air quality; Van den Putte (2015), Aissi et al. (2018), and Lundberg et al. (2022) on labor and workers’ rights and Zerk (2019) on civil and human rights. More recent studies focus on even more specific and narrowly defined non-trade outcome 2 See e.g., Yildirim et al. (2021). This debate was a core feature of the recent EU review of Trade and Sustainable Development chapters in EU PTAs (Velut et al., 2022), which centered on the design of co- operation to achieve sustainable development, including in the area of labor standards and environment, in EU PTAs (Marx et al., 2016; Hoekman and Rojas-Romagosa, 2022). One result of the review was a decision to put greater emphasis on compliance and enforcement of NTPs. 3 For example, Breinlich et al. (2022) adopt a machine learning approach to study the effect of WTO- plus PTA provisions, finding positive effects for provisions on areas covered by the WTO, such as subsidies and product standards. There is also some recent evidence on the trade impacts of NTPOs, again with gravity models. See for example Carrère et al. (2022). 4 There is a substantial literature built largely around case studies, focusing on specific countries or specific provisions of PTAs (e.g., Hafner-Burton (2009),Kim (2012), Spilker and Böhmelt (2013), and Postnikov (2014). Hafner-Burton (2009) argues that binding human rights clauses in trade agreements are more likely to induce compliance, but Spilker and Böhmelt (2013) have shown this finding is affected by potential selection bias. 2 or performance measures, such as geo-located tree coverage (Abman et al., 2021).5 We break with the recent literature by adopting a broader perspective. We focus collectively on three non-trade policy domains: environmental protection, labor stan- dards, and civil, social and human rights. We consider both the effects of enforceable vs. non-binding NTPs and the main potential complementary mechanism that may be used to pursue the non-trade policy objectives reflected in NTPs: development assistance projects targeting environmental improvement, labor rights protection, and civil and hu- man rights in developing countries. ODA constitutes a potentially important instrument to support efforts by countries that sign PTAs with NTPs to implement provision-related policy changes. Insofar as NTPs drive changes in associated non-trade outcomes this may be conditional on cooperation, including financial and technical assistance targeting the areas of interest. In the case of soft law NTPs development aid may be an important channel for efforts to improve non-trade outcomes given the absence of binding commit- ments and legal enforcement mechanisms. In cases where NTPs are enforceable ODA can also support implementation, but a binding NTP might be regarded by a donor country as a substitute for ODA, inducing a reduction in assistance. Countries that ac- cept binding NTPs may already have better underlying performance, so that ODA is not needed to attain a particular norm, and sustained implementation or compliance can be assured through dispute settlement procedures. These dimensions of NTP design figure prominently in policy debates but have been largely neglected in empirical research. We seek to identify the causal relationships between NTPs and a range of non-trade outcome indicators associated with these three policy areas of interest. In doing so, we advance the literature by (i) applying a consistent framework across the different policy domains, as opposed to the issue-, country- or PTA-specific focus in much of the research in this area; (ii) differentiating between the effects of enforceable versus non-binding 5 Abman et al. (2021) explore the causal impact of environmental provisions in PTAs on forest coverage loss. Their estimates suggest that PTAs tend to increase pressure on the environment (especially, via deforestation in developing countries), but that this negative impact can be at least partially offset by the inclusion of binding environmental protection obligations. Similarly, Tian et al. (2022) estimate that increased global CO2 emissions accompanying growth in production and trade created by the Regional Comprehensive Economic Partnership (RCEP) will dominate any potential mitigation commitments by signatories. 3 provisions; and (iii) evaluating the relationship between NTPs and official development assistance (ODA) allocated to the respective policy areas of interest. Our findings suggest an absence of consistent and statistically significant causal rela- tionships between NTPs and labor and civil rights, and indicative evidence of ambiguous effects of environment-related NTPs and certain types of environmental outcomes. Aver- age Treatment on the Treated (ATT) estimates indicate that NTPs dealing with labor, civil and human rights have no effect on associated outcome indicators, with the excep- tion of enforceable labor-related NTPs in EU PTAs, where we actually find a significant reduction in a measure of worker protection. We find mixed evidence on whether NTPs improve environmental outcomes. ATT estimates for binding versus non-binding NTPs sometimes have opposite signs, suggesting the type of NTP – enforceable or soft – may matter for different types of activities and thus environmental outcomes. In addition, we find that development assistance increases for countries that join PTAs with non-binding NTPs pertaining to the environment and civil and human rights. No such result obtains for NTPs on labor rights. The ODA results are consistent with the notion that soft law non-trade provisions need to be associated with other instruments if they are to be effective. As with the estimated null effect of labor-related NTPs on associated outcome indicators, agreeing to NTPs in this area also is not associated with greater aid. Labor rights are the only non-trade area where there is no evidence of improved performance or greater effort by donor countries to assist partner countries that sign NTPs to ameliorate outcomes. In the rest of the paper we proceed as follows. In Section 2 we introduce our data on NTPs in PTAs and on indicators that measure different dimensions of the non-trade issues associated with those NTPs. In Section 3 we discuss our methodology and present empirical results. We first analyze non-binding NTPs, comparing signatory countries to other countries that did not sign any PTA with the NTP of interest, and then assess binding NTPs. This is followed by an SDID analysis focusing on the relationship between binding vs. non-binding provisions and development assistance. Section 4 concludes. 4 2 Data We combine information from three sources. Data on NTPs come from the World Bank Deep Trade Agreement Database (Hofmann et al., 2019; Mattoo et al., 2020). This in- cludes information on all non-trade related provisions contained in 279 agreements signed between 1958 and 2015. The database distinguishes 14 “core” provisions that reconfirm existing WTO disciplines or impose additional (WTO-plus) obligations in policy areas that are covered by the WTO, as well as 38 provisions in areas that go beyond extant multilateral commitments (WTO-extra provisions). The dataset documents the growth in the inclusion of provisions on civil rights, environmental protection, and labor rights in trade agreements (Figure 1). In addition to reporting on the existence of provisions on a given subject, the database also provides information on their legal nature, including their enforceability.6 We focus on three types of WTO-extra provisions in PTAs; those related to civil and human rights promotion, labor rights protection, and environmental protection. We reduce the original bilateral dataset to a panel defined at country ∗ year level. For each country, we consider the year it signs a PTA containing the provisions of interest, the partner country (whether the PTA includes the EU, the US, or other countries), and whether the obligations arising from the agreement can be considered as binding (i.e., if they can be enforced through a dispute settlement mechanism). In case a country signs more than one agreement, we consider the first one in which a provision of interest is signed and assume that it stays in force even when additional agreements (possibly with different sets of partners and different sets of obligations) are signed. We allow for changes in legal enforceability over time with subsequent agreements. 6 Information on legal enforceability is reported for 52 selected policy areas in total. An extended version of the dataset provides more detailed indicators for a subset (18) of these policy areas. We work with agreements signed in the period 1995-2010. See https://datatopics.worldbank.org/dta/about-the- project.html. 5 Figure 1: Evolution of provisions on environment, labor, civil and political rights in PTAs Notes: Number of issue specific provisions in PTAs signed by sample countries over time. Source: Hofmann et al. (2019). We merge these data on country level PTA provisions with the NTPOID_v2 dataset (Manchin, 2021) characterizing non-trade related outcomes. This database includes in- formation on a broad range of economic, political, environmental, and social indicators.7 We integrate both datasets with the Environmental Performance Index (EPI) dataset (Wendling et al., 2020) which provides indicators related to climate, environment, and ecosystem vitality. We map relevant outcome indicators on each of the three non-trade policy domains of interest that are available for a long enough time-span for a large num- ber of countries. The conditions of continuity, country coverage and time span needed for estimation reduces the set of suitable indicators for labor, civil and human rights, as many are covered by indices characterized by limited variability, a short time span, or 7 The database was constructed by drawing on multiple sources including the Political Institutions 2017 Codebook (DPI) database, the International Political Economy Data Resource database - Version 3.0 (IPE), the Structural policy indicators database for economic research (SPIDER), the 2018 Quality of Government dataset (QoG)) and World Bank World Development Indicators. 6 periodic breaks. Table A-5 lists the resulting set of indicators and their sources. We retain 14 distinct environmental indicators, covering different types of pollution, habitat preservation, and forest protection. Three come from the World Bank World Development Indicators: the ratio of CO2 emissions in kg to GDP (in PPP $), forest area (% of land area) and SO2 emissions. Eleven additional outcome measures are sourced from the EPI data. These include two measures related to Climate Change mitigation: black carbon intensity and greenhouse gas emissions (measured in terms of average an- nual rate of increase); three indicators related to Environmental Health (PM2.5 exposure, NOx Intensity, and Ozone Exposure, all measured using the number of age-standardized disability-adjusted life-years lost per 100,000 persons due to exposure to ground-level ozone pollution); and six measures of Ecosystem Vitality - a protected areas index (the proportion of biologically scaled environmental diversity included in a country’s terrestrial protected areas, with higher values reflecting better performance), the species habitat in- dex (the proportion of suitable habitat within a country that remains intact for each species in that country relative to a baseline set in the year 2001), the sustainable nitro- gen management index measuring the efficiency in application of nitrogen fertilizer with maximum crop yields (higher values indicating worse performance), a species protection index measuring the species-level ecological representativeness of each country’s protected area network, tree cover loss (five-year moving average of the percentage of forest lost relative to forest cover in 2000), and wetland loss (quantified using a five-year moving average of the percentage of gross losses in wetland areas relative to a 1992 baseline). For EPI indices, a higher value indicates (sometimes counter-intuitively) better performance with respect to the outcome of interest.8 For labor protection we use two outcome variables: the Mosley-Uno labor rights in- dicator, a measure of collective labor rights, and the QOG worker rights indicator, a composite index measuring freedom of association in the workplace, the right to bargain collectively and a prohibition on the use of any form of forced or compulsory labor; a minimum age for the employment of children; and conditions of work with respect to min- 8 See further details on the dataset at https://epi.yale.edu/. 7 imum wages, hours of work, and occupational safety and health. This second measure of labor protection captures regulation of labor conditions more broadly. For civil and human rights, we use an electoral democracy index – polyarchy – a measure of freedom of expression, obtained from the Variety of Democracy database, which ranges between zero and one, with higher values indicating better outcomes; a female political participation (empowerment ) index; a political liberties index; and an index for freedom of association which measures the right of citizens to assemble freely and associate in political parties, unions, or cultural organizations. Figure 2 provides a snapshot for the three areas of the average direction of change over time and across countries in the non-trade outcome indicators used in the empirical analysis.9 They reveal that for several developing countries in Africa, Asia and Latin America there were improvements in environmental outcomes, while the trend appears to be less positive in terms of labor standards and civil rights protection. There is significant heterogeneity across indicators and countries during the period considered, with deterioration in non-trade outcomes observed in numerous instances. Data on Official Development Assistance (ODA) flows are sourced from the OECD Creditor Reporting System.10 This includes disaggregated data on aid allocated for en- vironmental, labor protection and civil rights projects as well as total ODA received by a country. To improve the match between treated countries (those having signed NTPs in these areas) and their synthetic counterfactual, in the empirical analysis we include measures of country-level economic openness using trade data from UN COMTRADE (excluding gas and oil) and market and country size (population, GDP, per capita GDP). 9 Ferrari et al. (2021) use principal components analysis to construct outcome indicators for these three areas to analyze the role of NTPs in EU PTAs. 10 https://stats.oecd.org/index.aspx/Index.aspx?DataSetCode=CRS1. 8 Figure 2: Non-trade outcomes across the world, 1990-2015 No Change Improvement Deterioration Civil Rights No Change Improvement Deterioration Environmental Protection No Change Improvement Deterioration Labor Rights Notes: Percentage variation in Civil Rights, Environmental, and Labor Rights Protection between 1990 and 2015. Simple averages of indicators described in the text for the three policy areas considered. 9 The resulting combined data covers more than 120 countries, for which we have an uninterrupted series for both the outcome and the explanatory variables. In order to guarantee a long enough stretch of pre-treatment and post-treatment observations, set to a minimum of 5 years respectively, we focus on all PTAs signed between 1995 and 2010.11 We restrict the partner countries in PTAs that include the EU, US or other OECD countries to a set of low and middle income non-EU countries.12 When focusing on the EU, US, and rest-of-world (ROW) agreements, we further restrict the sample of potentially treated countries by excluding all those with socio- economic conditions comparable to the EU or the US.13 The rationale for this additional restriction is that such countries might already apply high standards in the non-trade outcomes of interest. For instance, with respect to the EU or the US, we exclude high income countries such as Canada, New Zealand and Australia. The countries that are not considered in each exercise are listed in Annex Table A-1. Table 1 summarizes the number of PTAs with binding or non-binding NTPs in our sample, as well as the number of countries making such commitments. While EU PTAs have both binding and non-binding provisions (except in civil and human rights, where provisions are non-binding), the US only has binding NTPs in its PTAs, and does not in- clude provisions on civil and human rights. Most provisions entail environmental commit- ments, with some 80 countries signing binding or non-binding environmental provisions. Annex Table A-2 provides further information on signatories of PTAs with NTPs, in- cluding the number of “overlapping provisions” – instances where a country has accepted NTPs in more than one PTA. 11 In a few cases we partially impute some covariates and outcome indicators to preserve a minimum sample size. 12 We exclude EU accession countries from the sample, given the special nature of their agreements with the EU, which included extensive financial transfers and much deeper cooperation than observed in other trade agreements. 13 ROW refers to all agreements signed with all possible non-EU, non-US partners. This list includes but is not limited to other OECD countries such as Australia, Japan, Switzerland and Korea, and all partners in South-South PTAs involving non-OECD countries. 10 Table 1: Agreements and Number of Signatories by Type of Provision Panel A: Agreements with active provision ALL EU USA ROW Not Binding Not Binding Not Binding Not Binding Binding Binding Binding Binding Environment 57 54 18 16 0 12 39 26 Labor 20 50 4 15 0 12 16 23 CHR 21 2 16 0 0 0 5 2 Panel B: Countries with active provision ALL EU USA ROW Not Binding Not Binding Not Binding Not Binding Binding Binding Binding Binding Environment 78 77 17 26 0 18 73 63 Labor 30 66 2 26 0 18 35 52 CHR 53 8 18 0 0 0 38 8 _Notes: Panel A reports the number of agreements including either an Environmental, Labor, or Civil and Human Rights related provi- sion. Panel B lists instead the number of signatories. The notation “Binding” refers to all provisions that also establish some form of legal enforcement mechanism. 3 Empirical specification and results 3.1 Synthetic difference in difference estimation Evaluating the existence, sign and magnitude of the causal effect of NTPs on correspond- ing non-trade outcomes in partner countries is a major challenge. Countries that commit to a given NTP might differ from those that do not, violating the parallel trends assump- tion that is necessary in most causal inference analysis. This implies that identifying a suitable counterfactual scenario (what would have happened if a country did not sign a PTA with the NTP of interest?) is challenging. Matters are further complicated by the staggered adoption of PTAs by countries over time, which makes it difficult to create control units across time needed for counterfactual analysis (i.e., how to estimate the effect of signing a given NTP, when countries may do so at different points in time and possibly, accept the same type of provision with different partners?). To address both of these issues, we apply the Synthetic Difference-in-Differences (SDID) estimator proposed by Arkhangelsky et al. (2021). The SDID estimator com- bines the desirable features of difference in difference (DID) estimators and the flexibility of the synthetic control methodology. In effect the SDID can be regarded as a “doubly- weighted” two-way fixed effects (TWFE)-DID estimator, where unit- and time-specific 11 weights are computed from the data to (a) align pre-exposure trends in outcomes for treated (in our case, countries signing a PTA with a given type of NTP) and non-treated countries; and (b) balance pre-exposure and post-exposure time periods to reduce the influence of the staggered nature of signing PTAs. These two forms of “weighting” turn the TWFE estimator from being “global” to “local” in constructing a suitable control group, giving greater weight to countries that are more similar to the “treated” ones, and to time periods that are proximate to the treatment. The constructed comparability that derives from the double-weighting procedure al- lows the SDID estimator to potentially compensate for a lack of parallel pre-trends be- tween treated and untreated units in the raw data, an issue that might affect the robust- ness of traditional DID estimators. At the same time, because of the inclusion of two-way fixed effects and a different weighting algorithm, it does not require an exact match of pre-treatment trends of treated and non-treated units, a rarely satisfied requirement of the synthetic control method (Hollingsworth and Wing, 2020; McClelland and Mucciolo, 2022). Equation 1 presents the optimization process implemented by the SDID estimator to identify the average causal effect of the treatment on the treated countries (referred to as τ ) N T (ˆ ˆ, α τ, µ ˆ) = arg min ˆ, β (Yi,t − µ − αi − βt − Wi,t τ )2 ω ˆ sdid ˆ sdid λ (1) t τ,µ,α,β i=1 t=1 The part of equation 1 in parentheses comprises the TWFE component of the SDID estimator, where the term Yi,t refers to the pre-exposure trend in the performance of a PTA signatory country with respect to a given non-trade outcome indicator); αi is the equivalent of the Abadie et al. (2010) term for the effect of the intervention for country i, while βt controls for the difference between exposed and unexposed countries (i.e. between those countries that signed a given PTA provision and those that did not) at time t; and Wi,t τ denotes the exposure to a (binary) treatment, taking value 1 in ˆ sdid ) the post-PTA/NTP signing period. The first term outside of the parentheses (ω is similar to the unit weights in Abadie and Gardeazabal (2003) used to construct the ˆ sdid represents the time weight used to discount the synthetic counterfactual. Finally, λ t 12 distance in time across different treatment periods across countries. The latter term is a new addition compared to estimators used previously in the literature. The two parameters allow weighting the difference between treated and control units (from the TWFE component) by how much the two are comparable.14 We use the SDID estimator to identify the Average Treatment effect on the Treated (hereafter, the ATT) of signing a PTA containing at least one NTP addressing one of the three policy domains of interest: environmental protection, labor market regulation, and civil and human rights promotion. We consider a country as “treated” if it signs an agreement that includes the provision of interest in a given year.15 Following the synthetic control and DID literature, we include a set of additional country characteristics to improve the matching between PTA signatories and control units. Annex Table A-3 A reports the basic summary statistics for the covariates included in the matching algorithm of the SDID. We include three measures of market size and wealth (GDP, population and GDP per capita), a measure of government accountability as a proxy for the likelihood a government will comply with international agreements (from the WDI database), the value of total exports and the share of total exports accounted for by the EU and the US, respectively. All else equal, greater trade dependence is likely to increase implementation of NTPs. Similarly, greater reliance on trade with the EU and the US (the two major proponents of NTPs globally) is also more likely to improve compliance. 3.2 Results: Non-binding provisions We first investigate the impact of non-binding NTPs on related non-trade outcome indi- cators. We compare outcomes in countries signing non-binding NTPs to those observed in countries that do not sign any agreements including the relevant NTPs (i.e., we exclude those countries that, at some point before or during the sample period sign a PTA with 14 Most of the tests showing the consistency of SDID under different weighting schemes performed by Arkhangelsky et al. (2021) are based on a single treated unit. We exploit the foundation for the staggered treatment scenario in their Appendix A, where they allow for multiple treated units and multiple treatment periods. Pailañir and Clarke (2022) implement both procedures in STATA, allowing for the inclusion of additional covariates to improve the matching in the pre-treatment period. 15 If a country has signed more than one agreement containing the same provision in the period of interest, the earlier one is chosen. 13 a binding NTP). Table 2 reports estimates of the ATT of signing a non-binding NTP, distinguishing between environmental protection, labor market regulation, and civil and human rights. The first two columns report the ATT from non-binding NTPs for all agreements signed during the period of interest, as well as the estimated average percentage changes that the ATT translates into, compared to the sample average. The second two columns do the same for non-binding provisions included in EU PTAs (with no subsequent signa- ture of an agreement with the US.) Finally, the last two columns report results for PTAs with countries other than the EU or the US, again limiting the sample to countries that have not committed to NTPs in PTAs with the EU or the US.16 The results for environmental performance indicators are heterogeneous. When con- sidering all PTAs, only 6 of the 14 indicators considered point to a positive association with environmental NTPs. A statistically significant improvement (at the 10% confidence level) is only observed for greenhouse gases. Conversely, we find that signing a PTA with non-binding environmental provisions leads, on average, to a significant deterioration in four indicators of environmental quality; CO2 emissions, ozone exposure, protected areas (the share of biological environmental diversity included in a country’s terrestrial pro- tected areas) and sustainable nitrogen management in agricultural production.17 Figure 3 plots the standardised coefficients and their statistical significance for the sample of all PTAs. The standardised coefficients highlight that among environmental outcomes, the largest estimated change is the increase in CO2 emissions. Other estimated changes are smaller both in magnitude and in relative terms. 16 We do not consider US agreements separately as the US only has binding NTPs in its PTAs. The full Synthetic DID output tables are reported in Annex B. 17 The (+)/(−)n sign next to each measure in Tables 2 and 3 indicates to the direction of change that is associated with an improvement in the related outcome indicator. A negative sign reflects an improvement in the case of CO2 and SO2 emissions. 14 Table 2: Average Treatment Effect on Signatories - Non-Binding Provisions All agreements EU Rest of the World Variable ATT δ pct. ATT δ pct. ATT δ pct. CO2 (-) 0.141 48.17% -0.007 -2.33% 0.116 39.84% Forests (+) -0.766 -2.35% 0.314 0.96% -0.878 -2.70% SO2 (-) 3.870 0.80% 10.261 -2.13% 3.695 0.77% Black Carbon (+) -2.350 -4.56% -5.995 11.65% -1.083 -2.10% Greenhouse Gases (+) 2.315 3.93% -1.059 1.80% 2.408 4.09% Nox (+) -2.809 -5.98% 6.296 13.42% -3.340 -7.12% Ozone Exposure (+) -4.556 -9.10% 1.467 2.93% -3.807 -7.61% Environment Protected Areas (+) -1.779 -6.64% 0.823 3.07% -1.523 -5.68% PM2.5 (+) -1.105 -2.72% 1.264 3.11% -1.212 -2.98% Species Habitat (+) -0.557 -0.60% 2.334 2.51% 1.813 1.95% Nitrogen Management (+) -2.563 -7.68% 4.698 14.08% -3.393 -10.17% Species Protection (+) 0.487 0.74% -0.583 -0.89% -0.282 -0.43% Tree coverage loss (+) -3.854 -9.68% -4.321 -10.86% -2.336 -5.87% Wetland loss (+) -3.985 -6.96% 5.184 9.06% -7.130 -12.45% 15 Labour Rights (+) -0.770 -3.35% -1.172 -5.10% Labour Market Regulation Workers Protection (+) 0.094 9.86% -0.0668 -6.98% Polyarchy (+) 0.020 4.29% Freedom of Expression (+) 0.001 0.15% Civil and Human rights Women Political Participation (+) 0.000 0.07% Political Liberties (+) -0.004 -0.65% Freedom of Association (+) 0.096 8.55% _Notes: Average Treatment effect on the Treated (ATT) of non-binding provisions in PTAs on signatories. Percentage change refers to the relative change compared to the outcomes’ sample averages. Bold changes signal statistically significant effects. Boldface font indicates a statistically significant effect. The sign in brackets indicates the direction of change denot- ing an improvement in the related outcome. Table A-5 reports the sources and summarises the sign indicating an improvement for each outcome variable considered. Data constraints preclude estimation of the ATT for EU NTPs addressing labor and civil and human rights. US agreements do not include non-binding provisions on Civil and Human Rights. Figure 3: Average Treatment Effect on the Treated - Non-Binding Provisions, All Agree- ments Environmental Protection Labor Mkt. Reg. Civil and Human rights Notes: Standardised ATT effect. Plotted coefficients refer to the column “All Agreements ” in Table 2. 16 In the case of EU PTAs with non-binding NTPs, none of the outcome indicators are affected by NTPs, with the exception of tree coverage, for which there is a significant deterioration, equivalent to an 11% decrease compared to the mean (Table 2). Results for rest of the world PTAs are more similar to those for all PTAs, in that the estimates sug- gest a significant increase in CO2 emissions, deterioration in ozone exposure, protected areas, and nitrogen management, coupled with a smaller improvement in the greenhouse gases emissions (of about 40% and 4.09% respectively). In addition, there is an estimated reduction in forests, equivalent to about 2.7% of the average in the sample and an increase in wetland loss. These overall negative findings are consistent with the literature, and point to the pressure put on the environment by the increase in production and trade associated with the signing of a PTA. We take this to mean that while the core com- mitments of the trade agreements studied led to pressure for environmental degradation, the flanking features of those agreements, in the form of non-binding NTPs, failed to ameliorate those pressures. No significant effects are found for the other two issue areas, labor and worker rights and civil and human rights. 3.3 Results: Enforceable non-trade provisions Table 3 replicates the exercise focusing on enforceable NTPs, distinguishing between all PTAs, EU agreements without US, US agreements without EU, and ROW agreements that do not include the EU or the US. For the sample as a whole, binding NTPs are associated with a small but significant increase in forest coverage, as well as improvement in both PM2.5 emissions and ozone exposure index. Figure 4 plots the standardised co- efficients for the sample of all agreements. The results again mask heterogeneity across provisions. In the case of EU agreements we find a large and significant improvement in SO2 emissions, equivalent to an 85% change relative to the sample average. We also find a significant improvement in NOx emissions and ozone exposure, as well as in species habi- tat protection. In the case of US PTAs, binding provisions lead to a significant increase in CO2 emissions (similar to what was found for non-binding provisions), amounting to a 32% increase compared to countries that did not sign an agreement with the US. Fur- 17 thermore, PM2.5 emissions, species protection and habitat improve significantly. Finally, looking at non-EU, non-US PTAs, the only significant impact found is a 4% improvement in PM2.5 emissions. Unlike in the case of non-binding provisions, there is no other signif- icant change, despite the direction and large magnitude of the non-significant coefficients. We find no significant relationships between NTPs and civil and human rights outcome indicators. Binding labor-related provisions in EU agreements have a small but significant impact in terms of improving the indicator of collective labor rights, but are associated with a large deterioration in the broader measure of labor standards, which includes occupational safety and health, hours of work, and minimum age for employment of children. This index includes more measures associated with actual work conditions. The estimated impact is a 27% decrease in the indicator. 3.4 Discussion Overall we find that NTPs do not have a clearly identifiable impact on labor or civil rights, whether binding or non-binding, with the exception of binding labor provisions in EU PTAs, which are associated with a deterioration in the broad measure of labor standards. There is some evidence that NTPs may affect certain environmental indicators, but no consistent impact across the range of environmental outcome indicators. Many estimated changes in outcomes are not significant. Those that are significant are heterogeneous, with a mix of positive and negative estimates for some indicators, and with differences across EU, US and ROW agreements. Some of the statistically significant estimates imply that inclusion of NTPs is associated with a worsening of outcomes. A comparison of the estimates for non-binding and binding NTPs suggests that non-binding NTPs may lead to improved performance in some areas where binding provisions do not do so. Conversely, binding provisions may help improve performance in an area where non-binding NTPs do not. This suggests that insofar as the two types of NTPs are effective, this is issue-specific. 18 Table 3: Average Treatment Effect on signatories - Binding Provisions All Agreements EU USA Rest of the World Variable ATT δ pct. ATT δ pct. ATT δ pct. ATT δ pct. CO2 (-) -0.018 -6.08% -0.018 -6.23% 0.094 32.17% -0.006 -2.18% Forests (+) 0.549 1.69% 0.104 0.32% 0.025 0.08% 0.162 0.50% SO2 (-) -403.147 -83.81% -411.207 -85.49% 96.687 20.10% -787.828 -163.78% Black Carbon (+) -3.929 -7.63% 7.013 13.62% -8.570 -16.65% -6.579 -12.78% Greenhouse Gases (+) -0.395 -0.67% 0.305 0.52% 0.565 0.96% -1.943 -3.30% Nox (+) 2.341 4.99% 14.782 31.49% -0.371 -0.79% -5.692 -12.13% Ozone Exposure (+) 3.147 6.29% 5.081 10.15% 0.064 0.13% 0.999 2.00% Environment Protected Areas (+) 0.179 0.67% 2.119 7.90% 0.107 0.40% -0.064 -0.24% PM2.5 (+) 1.720 4.23% -0.865 -2.13% 4.119 10.13% 1.553 3.82% Species Habitat (+) -2.679 -2.89% 2.978 3.21% 2.164 2.33% -2.664 -2.87% Nitrogen Management (+) -1.092 -3.27% 0.589 1.77% 0.418 1.25% -0.268 -0.80% Species Protection (+) 0.511 0.78% -0.517 -0.79% 1.483 2.26% -1.155 -1.76% Tree coverage loss (+) -0.159 -0.40% -0.593 -1.49% -1.793 -4.51% -0.758 -1.91% Wetland loss (+) -0.077 -0.13% 1.941 3.39% -5.545 -9.68% -5.718 -9.99% 19 Labor Rights (+) 0.489 2.13% 0.212 0.93% -0.125 -0.54% 0.333 1.45% Labor Market Regulation Workers Protection (+) -0.050 -5.23% -0.264 -27.53% 0.152 15.78% -0.125 -13.08% Polyarchy (+) 0.017 3.78% -0.004 -0.81% Freedom of Expression (+) -0.002 -0.31% -0.009 -1.65% Civil and Human rights Women Political Participation (+) -0.013 -2.02% -0.111 -17.77% Political Liberties (+) -0.002 -0.42% 0.003 0.46% Freedom of Association (+) 0.130 11.59% 0.338 30.21% _Notes: Average Treatment effect on the Treated (ATT) of binding provision signatories. Percentage change refers to the relative change compared to the outcomes’ sample averages. Boldface font indicates a statis- tically significant effects. The sign in brackets refers the direction denoting an improvement in the related outcome. Table A-5 in the annex reports data sources and summarises the sign indicating an improvement for each outcome variable considered. Figure 4: Average Treatment Effect on the Treated - Binding Provisions, All Agreements Environmental Protection Labor Mkt. Reg. Civil and Human rights Notes: Standardised ATT effect. Plotted coefficients refer to the column “All Agreements ” in Table 3. 20 The results highlight a need for better understanding the incentive effects and ef- fectiveness of binding NTPs that are accompanied by legal enforcement mechanisms as compared to cooperation motivated by soft law types of NTPs. In the next sub-section we undertake an exploratory analysis of one such instrument that may play a role in influ- encing whether NTPs are implemented: official development assistance. The underlying hypothesis is that the impact of NTPs may be influenced by the use of complementary instruments that seek to improve non-trade outcomes. 3.5 Development assistance and NTPs Development assistance-funded projects and programs targeting issue areas addressed in NTPs may influence outcome performance. While there is an extensive literature investigating the effects of development assistance on trade, including so-called aid for trade projects and programs that seek to enhance trade capacity and facilitate trade,18 little attention has been given to the role of ODA targeting the non-trade issue areas that are the focus of this paper. Aid recipients may or may not have a PTA with the donor, and these may or may not include NTPs, which may be binding or non-binding. Of interest is not only whether ODA varies with (type of) NTPs but whether overall ODA flows are affected by signing an NTP or a change in the type of NTP that a partner country signs.19 Figure 5 plots the average amount of ODA (in USD million) received by countries that have not signed PTAs that include NTPs, those that have agreed to non-binding NTPs and those accepting binding provisions.20 Countries that have accepted binding environmental provisions receive more ODA than signatories of PTAs that only include non-binding provisions. There is little difference between the latter set of countries and those that have not agreed to any NTPs. The opposite is true for aid for civil and human rights-related activities, where countries that have joined PTAs that include non-binding 18 See e.g., Younas (2008); Pettersson and Johansson (2013); Hoekman and Shingal (2020). 19 Hoekman et al. (2023) find that (lagged) bilateral aid is positively and significantly associated with the propensity of developing countries to accept PTAs that include NTPs, including legally-enforceable provisions. 20 Table A-6 reports the t-tests for significance in mean differences. 21 NTPs receive substantially more aid than those that have not done so. The substantial difference between the amount of additional aid allocated for civil and human rights- related projects to countries that sign non-binding NTPs and those that make binding commitments in this area is not a function of the type of NTP but a reflection of the fact that the EU and the US do not include enforceable NTPs dealing with civil and human rights in their PTAs.21 Nonetheless, the difference between the pattern of aid and NTPs for environment and civil rights is noteworthy. Finally, in the case of labor rights, there is little difference in the average amount of ODA received by countries that have agreed to NTPs in this area and those that have not, nor does the type of NTP appear to matter much in terms of average dollar values disbursed. To assess whether and to what extent signing a PTA affects official development assis- tance flows, we again use the SDID estimator (see Eq. 1). We are interested in determin- ing whether signing an agreement affects the amount of development assistance received, and whether there are differences in this respect between signing binding provisions vs. non-binding ones. Table 4 presents the SDID estimates for this second exercise. We report the ATT estimates of signing a PTA that includes a NTP of interest (covering environmental, labor, or civil rights protection) on aid flows targeting the associated policy domain, as well as on total aggregate ODA received. Given the focus on ODA, we restrict our sample to the set of countries that were eligible for receiving financial support during the period 1990-2015. To improve matching, we include per capita GDP, population, total trade, and share of trade with the US and the EU.22 The results for binding (non-binding) provisions are reported in the upper (lower) panel of Table 4. The first three rows of each panel focus on the relationship between the amount of issue specific ODA received and acceptance of a NTP in the corresponding area. The last three rows of each panel report the impact of having accepted an issue 21 There are other donor countries with PTAs that do have binding provisions in this area that are captured in the regressions. 22 We consider all countries that received ODA at least once from at least one donor, and that did not lose the status of potential beneficiaries during the period considered. Country and year fixed effects included in the SDID estimator control for both time- and country-specific invariant factors that could also explain the allocation of ODA. 22 Figure 5: Development Assistance and NTPs Environmental Protection Labor Mkt. Reg. Civil and Human rights Notes: ODA in constant values, received by beneficiaries before and after signing an agreement. The category No Provision includes ODA received by countries before signing an agreement with the provision of interest and aid beneficiaries that never sign a PTA with such a provision. The remaining two box plots report the average amount of ODA received after signing an agreement including a non-binding or a binding provision, respectively. As noted, the EU and US do not have PTAs with binding civil and human rights-related NTPs. 23 specific NTP and total ODA received. Accepting binding environmental and labor NTPs has no statistically significant im- pact on ODA. For those countries that have signed enforceable NTPs addressing civil and human rights we observe a large reduction in average ODA received, both issue-specific aid and overall ODA. While statistically significant, it is important to note that this result is driven by the small number of countries that are members of PTAs that have binding civil rights provisions. As mentioned, these do not include US or EU PTAs. Non-binding NTPs dealing with environment and civil rights are associated with large, statistically significant, increases in the average value of overall ODA allocated to recipient countries. The positive effect on aid for civil and human rights is largely driven by EU agreements as the US does not include such NTPs in its PTAs. The absence of any relationship between ODA and labor-related NTPs suggested by the descriptive analysis in Figure 5 is confirmed by the SDID regression results. These results are consistent with the idea that non-binding NTPs may act as focal points for cooperation on an issue area, and that when assessing their effects account should be taken of complementary policy instruments such as technical cooperation and financial assistance. The estimates indicate that high-income PTA members – the major proponents of NTPs – allocate more aid to countries that agree to non-binding provisions in the area of environment and civil rights. The large additional amount of aid received by signatories of such NTPs suggests an apparent link between committing to soft NTPs and obtaining complementary support for measures to improve outcomes. This said, the absence of such a relationship for labor-related NTPs and aid is striking, and may help to explain the lack of evidence that NTPs in this area drive better labor standards in partner countries. 24 Table 4: Average Treatment Effect on the Treated (ODA recipients) ATT Std. Error δ Pct. Binding Provisions ODA: Environment Protection 2.35 11.10 70.28% ODA: Labor Rights Protection 2.41 2.20 271.61% ODA: Civil and Human Rights -23.06 11.59 -256.68% Total ODA - Environment Provision 119.51 119.71 63.66% Total ODA - Labor Mkt. Provision -13.97 99.95 -7.44% Total ODA - Civil, Human Rights Provision -229.28 58.18 -122.13% Non-Binding provisions ODA: Environment Protection 0.96 3.50 28.48% ODA: Labor Rights Protection -0.28 1.12 -31.70% ODA: Civil and Human Rights 26.81 21.49 298.44% Total ODA - Environment Provision 192.53 82.49 102.56% Total ODA - Labor Mkt. Provision 10.02 180.76 5.34% Total ODA - Civil, Human Rights Provision 236.10 133.22 125.76% _Notes: Average Treatment effect on the Treated (ATT) on ODA received by signatories of binding provision. The sample is limited to countries that are eligible to receive ODA from the DAC donors. The first 3 rows (ODA) refer to sector specific disbursements. Percentage change refers to the relative change compared to the received ODA sample averages. Statistically significant effects are reported in bold. 4 Conclusion Non-trade policy objectives are increasingly incorporated in developed countries’ trade policies. The Treaty on the Functioning of the European Union for example requires that external trade policy supports and promotes EU values relating to human rights, environment, and sustainable development.23 An implication is a need to balance the pursuit of material interests through trade with measures to ensure minimum standards are met regarding the conditions under which goods are produced. Conditioning pref- erential access to markets on non-trade provisions that relate to values and processes used to produce traded goods is a prominent instrument through which this balance is operationalized in the trade policy of high-income nations. The inclusion of NTPs in PTAs is not simply a reflection of a desire to promote improvement in labor and environmental standards and civil rights in partner countries. Commercial considerations may also play a role, including concerns of domestic firms to ensure a “level playing field” when confronting competition from imports. From the perspective of firms in partner countries, implementation of NTPs by their governments may increase production costs and reduce their competitiveness. Whatever the underlying 23 http://data.europa.eu/eli/treaty/tfeu_2012/oj 25 political economy and trade effects of NTPs (and PTAs more broadly) our focus in this paper is on the question whether the inclusion of NTPs leads to better non-trade outcomes as reflected in a range of performance indicators for which sufficient data are available. Whether or not NTPs are motivated primarily by a desire to safeguard and promote values, they should at the very least be associated with (gradual) improvement in outcomes. If this is not the case, not only is the values argument undercut, but so are concerns that NTPs are in fact (at least partly) driven by commercial concerns. For NTPs to have effects on competitiveness (e.g., by raising production costs in developing partner countries) they must be accompanied by attainment of higher standards in the areas covered. The results of our empirical analysis suggest that inclusion of NTPs in trade agree- ments does not have consistent, clear (i.e. significant) effects on non-trade outcomes in partner countries in the area of labor and civil rights, with the exception of EU agree- ments where we actually find a significant deterioration in the case of binding provisions. We find some evidence that NTPs are associated with specific environmental outcomes, though many of the estimates are not statistically significant, while those which are signif- icant are heterogeneous, with some indicators improving and others deteriorating. There is no generally consistent pattern. Our results also suggest non-binding NTPs may do more to improve performance in some areas than binding provisions, and vice versa. An implication is that the efficacy of these two types of NTPs is issue-specific, with one type of NTP potentially being “better” than the other. We also find evidence that development assistance increases for recipient countries that are members of PTAs that include non-binding NTPs, whereas this is not the case for countries that have accepted binding NTPs. This is consistent with the idea that that non-binding NTPs can act as a focal point for cooperation between PTA partners in the policy areas covered, utilizing non-trade policy instruments to pursue shared non-trade policy objectives. 26 References Abadie, A., Diamond, A. and Hainmueller, J. (2010). ‘Synthetic control methods for comparative case studies: Estimating the effect of california tobacco control program’, Journal of the American Statistical Association, vol. 105(490), pp. 493–505. Abadie, A. and Gardeazabal, J. (2003). ‘The economic costs of conflict: A case study of the Basque country’, American Economic Review, vol. 93(1), pp. 113–132. Abman, R. and Lundberg, C. (2020). ‘Does free trade increase deforestation? the effects of regional trade agreements’, Journal of the Association of Environmental and Resource Economists, vol. 7(1), pp. 35–72. Abman, R., Lundberg, C. and Ruta, M. (2021). ‘The effectiveness of environmental pro- visions in regional trade agreements’, World Bank, Washington, DC. Abman, R., Lundberg, C. and Szmurlo, D. (2022). ‘Trade, emissions, and environmental spillovers: Issue linkages in regional trade agreements’, The World Bank, forthcoming. Aissi, J., Peels, R. and Samaan, D. (2018). ‘Evaluating the effectiveness of labour provi- sions in trade agreements: An analytical and methodological framework’, International Labour Review, vol. 157(4), pp. 671–698. Arkhangelsky, D., Athey, S., Hirshberg, D.A., Imbens, G.W. and Wager, S. (2021). ‘Syn- thetic difference-in-differences’, American Economic Review, vol. 111(12), pp. 4088– 4118. Baghdadi, L., Martinez-Zarzoso, I. and Zitouna, H. (2013). ‘Are RTA agreements with environmental provisions reducing emissions?’, Journal of International Economics, vol. 90(2), pp. 378–390. Breinlich, H., Corradi, V., Rocha, N., Ruta, M., Santos Silva, J. and Zylkin, T. (2022). ‘Machine learning in international trade research-evaluating the impact of trade agree- ments’, CEPR Discussion Paper No. DP17325. Bronckers, M. and Gruni, G. (2021). ‘Retooling the sustainability standards in EU free trade agreements’, Journal of International Economic Law, vol. 24(1), pp. 25–51. Carrère, C., Olarreaga, M. and Raess, D. (2022). ‘Labor clauses in trade agreements: Hidden protectionism?’, The Review of International Organizations, vol. 17(1), pp. 453–483. Dür, A., Baccini, L. and Elsig, M. (2014). ‘The design of international trade agreements: Introducing a new dataset’, The Review of International Organizations, vol. 9(3), pp. 353–375. Ferrari, A., Fiorini, M., Francois, J., Hoekman, B., Lechner, L.M., Manchin, M. and Santi, F. (2021). ‘EU trade agreements and non-trade policy objectives’, The European University Institute, Robert Schuman Centre for Advanced Studies Research Paper No. RSC 2021/48. Hafner-Burton, E.M. (2009). Forced to be good : why trade agreements boost human rights, Ithaca: Cornell University Press. 27 Hoekman, B. and Rojas-Romagosa, H. (2022). ‘EU trade sustainability impact assess- ments: Revisiting the consultation process’, Journal of International Economic Law, vol. 25(1), pp. 45–60. Hoekman, B., Santi, F. and Shingal, A. (2023). ‘Trade effects of non-economic provisions in trade agreements’, European University Institute. Hoekman, B. and Shingal, A. (2020). ‘Aid for trade and international transactions in goods and services’, Review of International Economics, vol. 28(2), pp. 320–340. Hofmann, C., Osnago, A. and Ruta, M. (2019). ‘The content of preferential trade agree- ments’, World Trade Review, vol. 18(3), p. 365–398. Hollingsworth, A. and Wing, C. (2020). ‘Tactics for design and inference in synthetic control studies: An applied example using high-dimensional data’, SSRN. Kim, M. (2012). ‘Ex Ante Due Diligence: Formation of PTAs and Protection of Labor Rights’, International Studies Quarterly, vol. 56(4), pp. 704–719, ISSN 0020-8833. Lechner, L. (2016). ‘The domestic battle over the design of non-trade issues in preferential trade agreements’, Review of International Political Economy, vol. 23, pp. 840 – 71. Lechner, L.M. (2018). ‘The European Union’s inclusion of Non-Trade Issues in Preferen- tial Trade Agreements’, RESPECT mimeo. Lundberg, C., McLaren, J., Abman, R. and Ruta, M. (2022). ‘Child labor standards in regional trade agreements: theory and evidence’, The World Bank. Manchin, M. (2021). ‘Description of version 2 of the panel dataset on non-trade policy outcome indicators’, EUI RSCAS Working Paper. Marx, A., Lein, B. and Brando, N. (2016). ‘The protection of labour rights in EU bilateral trade agreements. a case study of the EU-Colombia agreement’, EU Trade Policy at the Crossroads: between Economic Liberalism and Democratic Challenges. Mattoo, A., Rocha, N. and Ruta, M. (2020). Handbook of deep trade agreements, World Bank Publications. McClelland, R. and Mucciolo, L. (2022). ‘An update on the synthetic control method as a tool to understand state policy’, Tax Policy Center. McLaughlin Mitchell, S. and Hensel, P.R. (2007). ‘International Institutions and Compli- ance with Agreements’, American Journal of Political Science, vol. 51(4), pp. 721–737. Pailañir, D. and Clarke, D. (2022). ‘Sdid: Stata module to perform synthetic difference- in-differences estimation, inference, and visualization’, Boston College Department of Economics. Pettersson, J. and Johansson, L. (2013). ‘Aid, aid for trade, and bilateral trade: An em- pirical study’, The Journal of International Trade & Economic Development, vol. 22(6), pp. 866–894. 28 Postnikov, E. (2014). ‘The design of social standards in EU and US preferential trade agreements’, in (D. Deese, ed.), Handbook of the International Political Economy of Trade, pp. 531–549, chap. 22, Edward Elgar Publishing. Spilker, G. and Böhmelt, T. (2013). ‘The impact of preferential trade agreements on gov- ernmental repression revisited’, The Review of International Organizations, vol. 8(3), pp. 343–361. Tian, K., Zhang, Y., Li, Y., Ming, X., Jiang, S., Duan, H., Yang, C. and Wang, S. (2022). ‘Regional trade agreement burdens global carbon emissions mitigation’, Nature Communications, vol. 13(1), pp. 1–12. Van den Putte, L. (2015). ‘EU bilateral trade agreements and the surprising rise of labour provisions’, International Journal of Comparative Labour Law and Industrial Relations, vol. 31(3). Velut, J., Baeza-Breinbauer, D., De Bruijne, M., Garnizova, E., Jones, M., Kolben, K., Oules, L., Rouas, V., Tigere Pittet, F. and Zamparutti, T. (2022). ‘Comparative analysis of trade and sustainable development provisions in free trade agreements’, London School of Economics. Wendling, Z.A., Emerson, J.W., de Sherbinin, A., Esty, D.C., Hoving, K., Ospina, C., Murray, J., Gunn, L., Ferrato, M., Schreck, M. et al. (2020). ‘Environmental perfor- mance index’, New Haven, CT: Yale Center for Environmental Law And Policy. epi. yale. edu. Yildirim, A., Basedow, R., Fiorini, M. and Hoekman, B. (2021). ‘EU trade and non- trade objectives: New survey evidence on policy design and effectiveness’, Journal of Common Market Studies, vol. 59(3), pp. 556–568. Younas, J. (2008). ‘Motivation for bilateral aid allocation: Altruism or trade benefits’, European Journal of Political Economy, vol. 24(3), pp. 661–674. Zerk, J. (2019). ‘Human rights impact assessment of trade agreements’, London, England: Royal Institute of International Affairs, Chatham House. 29 Annex A Additional data description Table A-1: Countries Considered Countries Partners in PTA Countries Excluded BDI,BEN,BFA,BLR,BRB,CHL,CHN,CIV,COL, CRI,DOM,DZA,EGY,GNB,GTM,GUY,HND,IDN, Other EU countries (Including accession) IND,JAM,JOR,KAZ,KEN,KGZ,KWT,LAO,LBN, USA, JPN,NZL,NOR,LIE,KOR,ISL,CHE,CAN, MEX,MMR,MYS,MLI,NER,NIC,OMN,PAN,PER, AUS PHL,QAT,RUS,RWA,SAU,SGP,SEN,SLV,SUR, THA,TJK,TTO,VNM,TGO,TUN,TZA,UGA,ZAF Notes: The countries considered as potentially treated, reported in the first columns, signed an agreement including at least one of the non- trade provisions of interest with either the EU, the US, or either other OECD or high income countries as a partner. No country belonging to the latter group (listed in the column Countries Excluded ) is considered in the “treated” or in the donor pool. Potentially treated coun- tries are included as potential control units in the donor pool, which also include countries that never signed a PTA (overall, with the EU, or with the US). 30 Table A-2: Countries with active agreement with both EU and USA Country Environmental Protection Labor Market Regulation Human and Civil Rights EU USA EU USA EU USA Not Bind. Not Bind. Not Bind. Not Bind. Not Bind. Not Bind. Bind. Bind. Bind. Bind. Bind. Bind. Chile 1 0 0 1 0 0 0 1 1 0 0 0 Colombia 1 0 0 1 1 0 0 1 1 0 0 0 Costa Rica 0 1 0 1 0 1 0 1 1 0 0 0 Guatemala 0 1 0 1 0 1 0 1 1 0 0 0 Honduras 0 1 0 1 0 1 0 1 1 0 0 0 Israel 1 0 0 0 0 0 0 0 1 0 0 0 Jordan 1 0 0 1 0 0 0 1 1 0 0 0 Korea, Rep. 0 1 0 1 0 1 0 1 0 0 0 0 Morocco 1 0 0 1 0 0 0 1 1 0 0 0 Mexico 1 0 0 1 0 0 0 1 1 0 0 0 31 Nicaragua 0 1 0 1 0 1 0 1 1 0 0 0 Peru 1 0 0 1 1 0 0 1 1 0 0 0 Salvador 0 1 0 1 0 1 0 1 1 0 0 0 _Notes: The 13 countries reported here signed an agreement with both the EU, EU member states, and the US. Table A-3: Matching covariates used in Synthetic Diff-in-Diff Variable Mean Std. Deviation Min Max Obs Export Share to EU 0.239152 0.176254 0.005185 0.871214 3422 Export Share to USA 0.076059 0.095722 0 0.671639 3422 Environmental Law Provision L.E. 0.463808 0.829501 0 2 3422 Labor Mkt. Provision L.E. 0.440085 0.824812 0 2 3422 CHR Provision L.E. 0.029805 0.241105 0 2 3422 GDP current (Log) 24.02366 2.292423 18.42872 30.49477 3422 Governance 0.905488 1.409029 -2.46429 4.270286 3422 GDPcap 9320.000 15000.000 22.700 103000.000 3422 Population 44.607 150.674 0.071 1397.029 3422 GDPc (Log) 21.70597 1.661391 16.93909 25.35765 3422 Population (Log) 2.317693 1.691523 -2.65039 7.237138 3422 Total ODA (US$ mil.) 618.000 1280.000 0.000 21700.000 3179 Total Trade (US$ mil.) 48600.000 24600.000 17800.000 88900.000 3422 ODA: Environment Protection (US$ mil.) 11.600 47.900 0.000 853.000 3179 ODA: Labor Market Regulation (US$ mil.) 2.978 18.600 -5.051 507.000 3179 32 ODA: Civil and Human Rights (US$ mil.) 28.8000 96.400 0.017 2050.000 3179 ODA to GDP: Total ODA 0.0411 0.075 0.000 1.126 3179 ODA to GDP: Environment Protection 0.0006 0.001 0.000 0.038 3179 ODA to GDP: Labor Market Regulation 0.0001 0.000 0.000 0.008 3179 ODA to GDP: Civil and Human Rights 0.0016944 0.0051576 -0.00005 0.0988238 3179 Table A-4: Summary statistics: Non-trade outcome indicators Environmental Protection Variable Mean SD Min Max N CO2 0.292338 0.19624 -1.35997 2.638186 7774 Forests 32.53986 18.44064 0 98.98526 7774 SO2 505.4478 2132.555 -2040.19 29989.1 7774 Black Carbon 51.47063 21.04038 0 100 7774 Greenhouse Gases 58.84966 26.44156 0 129.1304 7774 NOx 46.93604 21.01648 0 100 7774 Ozone Exposure 50.04097 15.96843 0 100 7774 Protected Areas 26.80994 14.3315 0 100 7774 PM2.5 40.65553 13.92697 0 100 7774 SO2 Trend 58.32562 21.90723 0 100 7774 Species Habitat 92.82481 8.956446 0 100 7774 Nitrogen Management 33.35488 14.12421 0 99.47662 7774 Species Protection 65.63925 16.71921 0 100 7774 Tree Coverage loss 39.79158 11.6473 0 100 7774 Wetland loss 57.25232 19.01902 0 100 7774 Labor-standards indicators Labor Rights 22.96748 6.242013 0 36.81034 7774 Worker Protection 0.95846 0.586932 0 2 7774 Civil and Human Rights Promotion NTPOs Polyarchy 0.462098 0.268917 0.013789 0.94937 7774 Freedom of Expression 0.570917 0.29947 0.014093 0.988696 7774 Women Political Participation 0.627106 0.228782 0.047552 0.999952 7774 Political Liberties 0.588804 0.302271 0.01185 0.993807 7774 Freedom of Association 1.146401 0.665681 0 2.040891 7774 33 Table A-5: Non-Trade Outcomes Environmental Protection Law Outcome Source Sign for Improvement CO2 World Development Indicators − Forests World Development Indicators + SO2 World Development Indicators − Black Carbon Environmental Protection Index + Greenhouse Gases Environmental Protection Index + Nox Environmental Protection Index + Ozone Exposure Environmental Protection Index + Protected Areas Environmental Protection Index + PM2.5 Environmental Protection Index + Species Habitat Environmental Protection Index + Nitrogen Management Environmental Protection Index + Species Protection Environmental Protection Index + Tree coverage loss Environmental Protection Index + Wetland loss Environmental Protection Index + Labor Market Regulation Outcome Source Sign for Improvement Labor Rights QOG Institute + Workers Protection QOG Institute + Human Rights Protection Outcome Source Sign for Improvement Polyarchy International Political Economy Data Resource V3 + Freedom of Expression International Political Economy Data Resource V3 + Women Political Participation International Political Economy Data Resource V3 + Political Liberties International Political Economy Data Resource V3 + Freedom of Association QOG Institute + Table A-6: Differences in Mean - T-test No Provision No Provision Non Binding vs Non Binding vs Binding vs Binding t-crit p-val t-crit p-val t-crit P-val Environment -4.34 0.00 -6.43 0.00 -4.15 0.00 Labor Market -1.09 0.28 1.62 0.11 2.06 0.04 CSHR -13.97 0.00 21.06 0.00 21.26 0.00 _Notes: T-test on ODA averages by type of provisions. 34 B Additional Results All Agreements Table B-1: ATT from Non-Binding Provisions Environment ATT Std. Error t-crit P-Value Lower CI Upper CI CO2 0.141 0.062 2.29 0.02 0.02 0.26 Forests -0.766 0.488 1.57 0.12 -1.72 0.19 SO2 3.870 33.670 0.11 0.92 -62.12 69.86 Black Carbon -2.350 3.806 0.62 0.55 -9.81 5.11 Greenhouse Gases 2.315 1.300 1.78 0.08 -0.23 4.86 Nox -2.809 3.399 0.83 0.41 -9.47 3.85 Ozone Exposure -4.556 0.691 6.59 0.00 -5.91 -3.20 Protected Areas -1.779 0.792 2.25 0.02 -3.33 -0.23 PM2.5 -1.105 0.945 1.17 0.25 -2.96 0.75 SO2 trend -5.649 3.740 1.51 0.13 -12.98 1.68 Species Habitat -0.557 1.418 0.39 0.71 -3.34 2.22 Nitrogen Management -2.563 1.204 2.13 0.03 -4.92 -0.20 Species Protection 0.487 1.653 0.29 0.78 -2.75 3.73 Tree coverage loss -3.854 3.456 1.12 0.27 -10.63 2.92 Wetland loss -3.985 4.361 0.91 0.37 -12.53 4.56 Labor Market ATT Std. Error t-crit P-Value Lower CI Upper CI Labor Rights -0.76991 0.802418 0.96 0.342 -2.34264 0.802834 Workers Protection 0.094497 0.125805 0.75 0.462 -0.15208 0.341076 Civil and Human Rights ATT Std. Error t-crit P-Value Lower CI Upper CI Polyarchy 0.019821 0.020212 0.98 0.332 -0.01979 0.059436 Freedom of Expression 0.000869 0.032032 0.03 0.978 -0.06191 0.063652 Women Political Participation 0.000443 0.020196 0.02 0.986 -0.03914 0.040027 Political Liberties -0.00381 0.031723 0.12 0.912 -0.06599 0.058365 Freedom of Association 0.095609 0.081169 1.18 0.24 -0.06348 0.2547 _Notes: Average Treatment effect on the Treated (ATT) in response to signing a non-binding non-trade provision. Bold- face font indicates statistically significant effects. 35 Table B-2: ATT from Binding Provisions Environment ATT Std. Error t-crit P-Value Lower CI Upper CI CO2 -0.01776 0.0206022 0.86 0.397 -0.05814 0.022621 Forests 0.549063 0.2512702 2.19 0.028 0.056574 1.041553 SO2 -403.147 390.4396 1.03 0.307 -1168.41 362.1147 Black Carbon -3.92905 4.336667 0.91 0.369 -12.4289 4.570819 Greenhouse Gases -0.39533 0.6239059 0.63 0.54 -1.61819 0.827526 Nox 2.34134 3.491818 0.67 0.513 -4.50262 9.185303 Ozone Exposure 3.147339 0.9559039 3.29 0.001 1.273767 5.020911 Protected Areas 0.17913 0.7075864 0.25 0.814 -1.20774 1.565999 PM2.5 1.719509 0.5980004 2.88 0.004 0.547428 2.89159 SO2 trend 8.15026 4.315405 1.89 0.058 -0.30793 16.60845 Species Habitat -2.67891 2.344744 1.14 0.257 -7.27461 1.916786 Nitrogen Management -1.0918 1.154168 0.95 0.348 -3.35396 1.170374 Species Protection 0.510788 1.093949 0.47 0.651 -1.63335 2.654928 Tree coverage loss -0.15936 1.286582 0.12 0.912 -2.68106 2.362341 Wetland loss -0.07688 2.586414 0.03 0.978 -5.14625 4.992489 Labor Market ATT Std. Error t-crit P-Value Lower CI Upper CI Labor Rights 0.489115 0.4133784 1.18 0.24 -0.32111 1.299337 Workers Protection -0.05014 0.094709 0.53 0.608 -0.23577 0.135492 Civil and Human Rights ATT Std. Error t-crit P-Value Lower CI Upper CI Polyarchy 0.017481 0.0128477 1.36 0.175 -0.0077 0.042663 Freedom of Expression -0.00177 0.0183558 0.1 0.927 -0.03775 0.034203 Women Political Participation -0.01264 0.0237998 0.53 0.608 -0.05929 0.034006 Political Liberties -0.00249 0.0168085 0.15 0.89 -0.03544 0.030452 Freedom of Association 0.129613 0.0932787 1.39 0.165 -0.05321 0.312439 _Notes: Average Treatment effect on the Treated (ATT) in response to signing a binding provision. Boldface indicates a statistically significant effect. 36 Agreements With the EU as signing partner Table B-3: ATT from Non-Binding Provisions - PTAs with EU as a signatory Environment ATT Std. Error t-crit P-Value Lower CI Upper CI CO2 -0.007 0.045 0.15 0.89 -0.095 0.081 Forests 0.314 0.366 0.86 0.40 -0.403 1.031 SO2 -10.261 99.813 0.1 0.93 -205.894 185.371 Black Carbon -5.995 7.984 0.75 0.46 -21.644 9.655 Greenhouse Gases -1.059 0.862 1.23 0.22 -2.748 0.630 Nox 6.296 5.907 1.07 0.29 -5.281 17.874 Ozone Exposure 1.467 1.151 1.27 0.21 -0.789 3.723 Protected Areas 0.823 0.892 0.92 0.36 -0.926 2.572 PM2.5 1.264 1.009 1.25 0.21 -0.713 3.241 SO2 trend 5.532 7.050 0.78 0.44 -8.287 19.351 Species Habitat 2.334 2.382 0.98 0.33 -2.335 7.003 Nitrogen Management 4.698 3.613 1.3 0.20 -2.384 11.779 Species Protection -0.583 0.848 0.69 0.50 -2.245 1.079 Tree coverage loss -4.321 2.311 1.87 0.06 -8.850 0.207 Wetland loss 5.184 7.762 0.67 0.51 -10.030 20.398 Civil and Human Rights ATT Std. Error t-crit P-Value Lower CI Upper CI Polyarchy 0.020 0.031 0.66 0.52 -0.040 0.080 Freedom of Expression -0.001 0.062 0.02 0.99 -0.123 0.121 Women Political Participation -0.003 0.039 0.09 0.93 -0.080 0.073 Political Liberties -0.002 0.064 0.03 0.98 -0.127 0.124 Freedom of Association 0.129 0.100 1.29 0.20 -0.067 0.326 _Notes: Average Treatment effect on the Treated (ATT) in response to signing a non-binding provision. Boldface indicates statistically significant effects. It was not possible to estimate the ATT for non-binding Labor Market related provisions. 37 Table B-4: ATT from Binding Provisions - EU Environment ATT Std. Error t-crit P-Value Lower CI Upper CI CO2 -0.014 0.017 0.81 0.426 -0.047 0.019 Forests 0.114 0.156 0.73 0.475 -0.191 0.419 SO2 -322.757 155.133 2.08 0.037 -626.819 -18.696 Black Carbon 6.214 8.367 0.74 0.468 -10.185 22.614 Greenhouse Gases -0.017 1.607 0.01 0.993 -3.168 3.133 Nox 12.445 5.105 2.44 0.015 2.438 22.451 Ozone Exposure 4.344 1.020 4.26 0 2.345 6.343 Protected Areas 1.582 1.387 1.14 0.257 -1.136 4.300 PM2.5 -0.803 0.524 1.53 0.126 -1.830 0.225 SO2 trend 9.131 10.633 0.86 0.397 -11.710 29.972 Species Habitat 2.676 1.529 1.75 0.08 -0.321 5.674 Nitrogen Management 0.283 1.599 0.18 0.867 -2.852 3.417 Species Protection -0.444 0.434 1.02 0.312 -1.295 0.406 Tree coverage loss -0.195 1.527 0.13 0.905 -3.187 2.798 Wetland loss 1.942 4.504 0.43 0.68 -6.887 10.771 Labor Market ATT Std. Error t-crit P-Value Lower CI Upper CI Labor Rights 0.163 0.068 2.4 0.016 0.030 0.297 Workers Protection -0.212 0.100 2.11 0.035 -0.408 -0.015 _Notes: Average Treatment effect on the Treated (ATT) in response to signing a binding provision. Boldface indi- cates statistically significant effects. It was not possible to estimate the ATT for binding, Civil and Human rights protection-related provisions. 38 Agreements With the US as signing partner Table B-5: ATT from Binding Provisions - USA Environment ATT Std. Error t-crit P-Value Lower CI Upper CI CO2 0.028399 0.036708 0.77 0.45 -0.04355 0.100346 Forests 0.24378 0.68086 0.36 0.732 -1.09071 1.578265 SO2 275.5559 236.2318 1.17 0.245 -187.458 738.5702 Black Carbon -1.85771 5.702112 0.33 0.754 -13.0338 9.318433 Greenhouse Gases 0.941502 1.01159 0.93 0.358 -1.04121 2.924219 Nox 1.602137 5.402121 0.3 0.777 -8.98602 12.19029 Ozone Exposure 3.324542 1.183079 2.81 0.005 1.005707 5.643377 Protected Areas 0.003143 0.913218 0 1 -1.78676 1.79305 PM2.5 2.5335 1.142401 2.22 0.026 0.294394 4.772606 SO2 trend 12.45017 3.300849 3.77 0 5.980506 18.91983 Species Habitat 0.568695 2.591109 0.22 0.837 -4.50988 5.647268 Nitrogen Management -1.39872 1.350941 1.04 0.303 -4.04657 1.249123 Species Protection 3.640571 2.473755 1.47 0.142 -1.20799 8.489131 Tree coverage loss -1.272 2.894777 0.44 0.673 -6.94576 4.401768 Wetland loss 1.996312 3.001094 0.67 0.513 -3.88583 7.878456 Labor Market ATT Std. Error t-crit P-Value Lower CI Upper CI Labor Rights -0.2938 0.829378 0.35 0.739 -1.91938 1.331779 Workers Protection 0.060735 0.140436 0.43 0.68 -0.21452 0.33599 _Notes: Average Treatment effect on the Treated (ATT) in response to signing a binding provision. Boldface indi- cates statistically significant effects. It was not possible to estimate the ATT for binding, Civil and Human rights protection-related provisions. 39 Agreements With all countries excluded the EU, the US as signing partners Table B-6: ATT from Non Binding Provisions - PTAs among non-EU, non-US countries Environment ATT Std. Error t-crit P-Value Lower CI Upper CI CO2 0.129245 0.041899 3.08 0.002 0.047123 0.211368 Forests -0.93871 0.488261 1.92 0.054 -1.8957 0.018286 SO2 7.3654 24.42961 0.3 0.777 -40.5166 55.24744 Black Carbon -4.99189 2.774301 1.8 0.071 -10.4295 0.445738 Greenhouse Gases 2.681199 1.132112 2.37 0.018 0.462259 4.900139 Nox -3.1096 2.784895 1.12 0.266 -8.56799 2.348798 Ozone Exposure -4.17789 0.54844 7.62 0 -5.25283 -3.10294 Protected Areas -1.68004 0.749064 2.24 0.025 -3.14821 -0.21187 PM2.5 -1.02904 0.99164 1.04 0.303 -2.97265 0.914577 SO2 trend -5.70608 2.923449 1.95 0.051 -11.436 0.023877 Species Habitat -0.2494 1.44293 0.17 0.875 -3.07754 2.578747 Nitrogen Management -3.03963 1.014901 2.99 0.003 -5.02883 -1.05042 Species Protection 0.780171 1.554832 0.5 0.63 -2.2673 3.827642 Tree coverage loss -2.6246 2.780202 0.94 0.353 -8.07379 2.8246 Wetland loss -7.14907 4.996949 1.43 0.153 -16.9431 2.644955 Labor Market ATT Std. Error t-crit P-Value Lower CI Upper CI Labor Rights -0.78825 0.804666 0.98 0.332 -2.36539 0.788899 Workers Protection -0.09021 0.124893 0.72 0.481 -0.335 0.154585 Civil and Human Rights ATT Std. Error t-crit P-Value Lower CI Upper CI Polyarchy 0.016847 0.01657 1.02 0.312 -0.01563 0.049323 Freedom of Expression 0.008372 0.021726 0.39 0.71 -0.03421 0.050955 Women Political Participation -0.00038 0.017686 0.02 0.986 -0.03505 0.034281 Political Liberties 0.003367 0.017361 0.19 0.86 -0.03066 0.037394 Freedom of Association 0.051205 0.11223 0.46 0.658 -0.16877 0.271176 _Notes: Average Treatment effect on the Treated (ATT) in response to signing a non-binding provision. Boldface ATT indi- cates statistically significant effects. 40 Table B-7: ATT from Binding Provisions - Agreements among non-EU, non-US countries Environment ATT Std. Error t-crit P-Value Lower CI Upper CI CO2 -0.01389 0.017878 0.78 0.444 -0.04893 0.021146 Forests 0.571814 0.318321 1.8 0.071 -0.0521 1.195723 SO2 -465.032 477.7811 0.97 0.337 -1401.48 471.4189 Black Carbon -4.39428 4.813709 0.91 0.369 -13.8291 5.040594 Greenhouse Gases -0.73606 0.752863 0.98 0.332 -2.21167 0.739552 Nox -4.36918 3.594744 1.22 0.224 -11.4149 2.676514 Ozone Exposure 2.279647 1.069882 2.13 0.033 0.182678 4.376616 Protected Areas -0.6553 0.771309 0.85 0.403 -2.16706 0.856467 PM2.5 1.477123 0.560081 2.64 0.008 0.379363 2.574883 SO2 trend 5.895188 4.333213 1.36 0.175 -2.59791 14.38829 Species Habitat -3.95925 2.81741 1.41 0.159 -9.48137 1.562875 Nitrogen Management -1.45454 1.647489 0.88 0.386 -4.68362 1.774541 Species Protection 0.594122 1.326157 0.45 0.666 -2.00515 3.19339 Tree coverage loss -1.20712 1.060469 1.14 0.257 -3.28564 0.871395 Wetland loss -2.8613 3.082695 0.93 0.358 -8.90338 3.180787 Labor Market ATT Std. Error t-crit P-Value Lower CI Upper CI Labor Rights 0.322519 0.389044 0.83 0.414 -0.44001 1.085045 Workers Protection -0.07003 0.095573 0.73 0.475 -0.25735 0.117292 Civil and Human Rights ATT Std. Error t-crit P-Value Lower CI Upper CI Polyarchy -0.00373 0.02158 0.17 0.875 -0.04603 0.038566 Freedom of Expression -0.00945 0.014175 0.67 0.513 -0.03723 0.018337 Women Political Participation -0.11143 0.089539 1.24 0.217 -0.28693 0.064061 Political Liberties 0.002686 0.012681 0.21 0.845 -0.02217 0.027541 Freedom of Association 0.337724 0.303473 1.11 0.27 -0.25708 0.93253 _Notes: Average Treatment effect on the Treated (ATT) in response to signing a non-binding provision. Boldface indicates refer to statistically significant effects. 41