Voluntary programs and emissions revisited: What is the effect of EU trade agreements with environmental provisions?

Multinational enterprises operating global value chains are being increasingly pressured to source from suppliers that adopt green private standards. Likewise, public policymakers are also pressured to reduce national pollution levels to contribute to sustainable development initiatives. In this context, while there is extensive debate on how domestic, country-specific environmental regulations interact with private standards (adopted by firms) in reducing national pollution levels, less is known about the role of international trade policies, which have recently embraced an array of sustainability issues. Our paper seeks to extend our understanding of the extent to which ISO environmental certifications affect a country’s level of emissions of greenhouse gases and air pollutants, and whether the European Union’s environmental protection (EP) standards – as mediated through trade agreements – condition this response. Prior research provides mixed evidence on the impact of the adoption of ISO-14001 on pollution reduction. Based on prior literature and using institutional theory and environmental stewardship perspectives, we expect that membership of trade agreements with EP provisions would complement the effect of ISO-14001 uptakes in reducing national pollution levels. Our arguments and results emphasize the complexity between private and public regulations on pollution reduction.


INTRODUCTION
Environmental concerns are increasingly driving both multinational enterprises (MNEs) and regulators to seek less-polluting and more sustainable methods of production (Christmann, 2004;Delmas & Toffel, 2004;Hartmann & Uhlenbruck, 2015;Ramus, 2002). Reducing emissions is an important part of combating climate change, as denoted during the COP26 climate conference, where 153 countries announced new targets for reducing emissions, covering over 90% of global GDP (COP26, 2021). However, there remain cross-country differences in terms of environmental regulations and their enforcement, largely due to scientific uncertainties about the consequences of pollution and disagreements about the costs vs. the benefits of environmental protection schemes. In this context, the adoption of private regulatory standards has been popular, arising both from corporate self-interest (a desire to exploit market-based advantages) and institutional pressures (Kolk, 2016;Prakash & Potoski, 2006). Nonetheless, the implications on the environment from the adoption of specific environmental standards remain contested.
In global value chains (GVCs), compliance with international private standards is becoming increasingly important, especially for exporting companies in developing countries, where local regulations are weak (Nadvi, 2008;Schmitz, 2004). The International Standards Organization (ISO) 14001 environmental standard has attracted particular attention in the literature (Arimura et al., 2016;Delmas & Montes-Sancho, 2011;Jiang & Bansal, 2003;King et al., 2005). This is because it has diffused most widely among the corporate communities across both industries and economic regions. As a result, it has proved to be a popular governance mechanism in GVCs within the environmental context, so examining its impact is apposite. Additionally, it is a private regulation, so its diffusion speaks to the phenomenon of private regulation, which is distinct from state regulation in the international economy (Büthe, 2010). Private regulatory regimes have gained considerable academic attention over the past three decades, as their existence and development appear to challenge conventional notions of state-centric governance (Boddewyn, 1985;Vogel, 2008).
A central question in the private-public governance debate is the extent to which private regulatory regimes are complementary to public regulation (Coslovsky & Locke, 2013;Heyes & Maxwell, 2004;Potoski & Prakash, 2005a, 2005b. In the context of ISO-14001, on the one hand, some studies suggest that stringent domestic environmental regulations are likely to substitute the utility of ISO-14001 in reducing pollution because stronger domestic regulations will reduce the benefits of adopting the standard in terms of reputational gains (Prakash & Potoski, 2014). On the other hand, there is also evidence of a complementary effect between public and private regimes in the other related contexts of labor standards in GVCs . Prior research on this joint effect does not examine trade agreements, however. The business literature considers private regulation as a mechanism for firms to reduce transaction costs in the context of weak market supporting institutions (El Ghoul et al., 2017). Political economy and international law perspectives are more varied, not the least in acknowledging the potential complementarities between the two. It has been argued that policymakers welcome private regulation where issues are novel or complex (Lawton & McGuire, 2003;Mügge, 2006).
As part of the ongoing debate on the effectiveness of public-private governance in GVCs, our paper examines the impacts of both public and private regulatory actions on greenhouse gas (GHG) and air pollutant emissions. Although the effectiveness of voluntary environmental programs has been questioned, Potoski and Prakash (2013) demonstrated that ISO-14001 certifications reduce SO 2 emissions at the country level, indicating that private standards can be helpful in achieving sustainable development (Kolk & Pinkse, 2008). We aim to revisit their study, but to also extend the scope of the analysis to a broader set of emissions, including GHGs, which have been held significantly responsible for global warming and climate change (Cucchiella et al., 2017;Nordhaus, 1993). Yet, most prior studies on the impact of ISO-14001 have focused on air pollutants (such as SO 2 ), rather than on GHGs, such as carbon dioxide (CO 2 ), methane (CH 4 ) and nitrous oxide (N 2 O), which collectively account for approximately 97% of GHGs. Thus, examining the relationship between ISO-14001 adoption and the emissions of GHGs at the national level is important to achieve a fuller picture of the effectiveness of the ISO-14001 standard.
A second important point is that prior research on the joint effect of private and public regulations on pollution reduction has focused primarily on the stringency of 'domestic' (or national) regulations (Heyes & Maxwell, 2004;Prakash & Potoski, 2014) and surprisingly little on 'international' trade policies, despite private standards (such as the ISO-14001) being the most popular among exporting firms in GVCs (Christmann, 2004;Christmann & Taylor, 2002). Over the past few decades, the number of trade agreements with environmental provisions has risen. For instance, the Trans-Pacific Partnership (TPP) is argued to have approximately 136 environmental norms, whereas the Canada-EU trade agreement has approximately 114 such norms (Morin et al., 2017). Due to these recent developments, we focus on the EU's environmental policies in its trade agreements (as a form of public regulation) to test its moderating effect on ISO-14001 in reducing emissions. The EU has been largely successful in reducing the emissions of both GHG and air pollutants, and is seen as a leader in driving the global climate agenda. Between 1990 and 2012, GHG emissions in Europe decreased by 18%, and the European Council plans to reduce emissions by a further 40% by 2030 (Cucchiella et al., 2017). The EU has also been successful in exporting its regulatory model (Falkner, 2007;Kelemen, 2010): the EU has both a large internal market and a large number of MNEs that lead global value chains and thus benefits from bargaining power to diffuse both private and public regulatory standards globally (Lavenex & Schimmelfennig, 2009).
Our analysis, based on ISO-14001 adoption levels in 147 countries over the 1999-2014 period, shows that a greater uptake of the ISO-14001 standard is associated with lower levels of emissions of greenhouse gases (GHGs), namely carbon dioxide (CO 2 ) and methane (CH 4 ). At the same time, membership of EU trade agreements with environmental protection provisions is directly associated with lower emissions of harmful air pollutants (SO 2 and NO x ) but has no direct effect on emissions of GHGs. Importantly, for carbon dioxide (CO 2 ) we find that the effect of ISO-14001 on pollution reduction is greater for countries that become part of EU trade agreements with environmental provisions, implying a positive interaction effect between the levels of ISO-14001 adoption (at the country level) and the membership of trade agreement with the EU with environmental protection provisions. We thus contribute to the literature discussing the efficacy of private standards in reducing pollution at the national levels, as well as on the complementary effect of trade policy as a public regulation.
The remainder of this paper is organized as follows. The next section provides a review of the literature. Following this, we provide a theoretical background and develop our hypotheses. This is followed by a description of our methods and our data, followed by the presentation of our empirical results. Finally, we discuss our findings and highlight the implications and limitations of our research.

Private and Public Regulation in Environmental Protection
Cross-border economic activity raises the prospect of multinational enterprises (MNEs) exploiting national regulatory differences and engaging in regulatory arbitrage. In the context of environmental standards, concerns that MNEs would invest in developing countries to escape more stringent regulations at home gave rise to the 'pollution haven' hypothesis (Cole, 2004;Harrison & Eskeland, 1997). The extent of this downward pressure on environmental standards is unclear, however. Porter's hypothesis alternatively argues that stringent environmental regulations in fact enable firms to engage in greener practices as well as in environmental innovations, which provide firms with competitive advantages (Porter & Van der Linde, 1995).
Broadly, studies show that regulatory arbitrage can be reduced through formal governmental regulations or voluntary, self-regulatory action by firms. Potoski and Prakash argued across a series of papers (see for example, Potoski & Prakash, 2005a, 2005bPrakash & Potoski, 2006, 2007 that the diffusion of ISO environmental standards has challenged the pollution haven hypothesis. Private regulatory arrangements have attracted considerable academic and policymaking interest, as a wide range of company-sponsored regulatory arrangements have proliferated (Mayer & Gereffi, 2010). Why we have seen this proliferation is disputed. For some it reflects a broader triumph of neo-liberal ideas about how to operate a market economy, with a concomitant preference for private activity over public (Mayer & Gereffi, 2010). Alternatively, private regulation arises as a response to 'governance gaps', particularly in developing countries where the capacity and capability to regulate is low. Private regulation fills these institutional voids (Christmann & Taylor, 2001;Silvestre, 2015). Finally, self-regulation by firms, especially for those operating in developing countries, is argued to act as an important signal to consumers and regulators in their target markets about firms' commitment to socially responsible behavior (Fiaschi et al., 2017;Leipziger, 2017;Marano et al., 2017).
Among the more prominent private regimes are those relating to environmental and labor standards, where consumer preferences in developed economies motivate companies to create private standards in host countries (Leonidou et al., 2015). As Bartley (2003) argues, in areas such as apparel production and forestry, this has been done in response to pressures from social groups and consumer interests in key final markets. Private regulatory standards are not, in the strictest sense, a new phenomenon. For instance, Büthe (2010) suggests that international standards for telecommunications equipment have for a long time fallen under the ambit of the International Electrotechnical Commission (IEC), a private organization founded in 1906 that saw the need for common and accepted standards to facilitate the sector's growth. Firms have, in fact, powerful incentives to create common standards that open opportunities to exploit economies of scale. Furthermore, innovation in products or processes poses regulatory challenges to governments, that face demands to regulate (Stewart, 1993). In these circumstances, firms themselves have incentives to support private regulatory innovation and exploit their greater understanding of the relevant product or service markets.
A growing body of literature seeks to understand the interaction of private regulation, public regulation, and firm strategies and responses. International trade agreements increasingly contain provisions relating to a number of non-traditional trade issues, such as labor standards and environmental protection (Lechner, 2016). Making good on these commitments can involve the adoption of international standards, such as the ISO (Heras-Saizarbitoria, 2021). It is seen that the level of integration into the international trading system affects firms' tendency to adopt more socially responsible practices (Dau et al., 2021). However, the extent to which policy initiatives, including trade agreements and regulatory standards, reach local firms -even those nested in global value chains (GVCs) -can be questioned. Much of the literature focuses on the role of lead firms, usually MNEs. The presumption is one of control: GVCs are dominated by MNEs, although this is not to suggest that GVCs are not subject to important external pressures (Buckley, 2021). These firms are assumed to enforce standards for a mix of competitiveness and reputational reasons, as well as stakeholder expectations (Heras-Saizarbitoria, 2021). This overlooks both the complexity and difficulty of enforcing agreements, and the agency of firms farther down the value chain. In fact, enforcement can be difficult. Relying on episodic inspections of facilities may be an inadequate means of verification, with local firms able to game the process (Nadvi, 2008). More positively, simply seeing international regulatory standards to be transmitted down from MNEs overlooks the embeddedness of local firms, which provides them with context-specific knowledge and social capital that can be used to support social goals, albeit in ways not anticipated by the MNE or policymakers. National or local policy environments would thus seem to be important enablers. The literature on institutional voids has been used to account for ways in which firms step in to fill governance gaps in economies in order to facilitate their operations. However, the robustness of the domestic regulatory environment can be a crucial explanatory variable in understanding why local firms adopt policies linked to trade agreements (Dau et al., 2021). In this way, the literature on regulatory standards is enriched by an understanding of how firms embedded in GVCs respond to incentives (Sinkovics et al., 2021b).

Adoption and Outcomes of ISO-14001
The ISO-14001 provides an important voluntary standard to reduce the scope for regulatory arbitrage in environmental protection; although, whether this standard actually works to reduce pollution at both the facility and national levels is disputed in the literature. In addition, ISO-14001 adoption can be seen as having a signaling effect, likewise the adoption of, for example, standards for sustainable forestry. Firms signal their interest in maintaining standards to important stakeholder groups.
Various studies have focused on the specific reasons why firms adopt ISO-14001 (Aragó n-Correa & Sharma, 2003;Darnall et al., 2008;Delmas & Montiel, 2008). These studies mainly suggest that pressure from lead firms in the value chain, international regulators, and other stakeholders, such local non-governmental organizations (NGOs), are important reasons for manufacturing plants to adopt ISO-14001. The benefits to firms from the adoption of ISO-14001 include improved efficiency of energy and resource consumption (Radonjič & Tominc, 2006), lesser risks of environmental accidents, better compliance with environmental regulations, improved corporate image, better motivated employees working in pollution-intensive manufacturing, and greater competitiveness in the market (Morrow & Rondinelli, 2002). Recent studies have also found evidence of other indirect benefits of adopting ISO-14001, such as greater innovative capabilities (He & Shen, 2019).
However, several studies did not find any significant effect of ISO-14001 adoption on the above firm-level aspects (for a full review, see Boiral et al., 2018), thus resulting in overall mixed evidence of the impact of the standard on firm-level benefits. In this respect, contextual factors related to the ways in which the standard is implemented have been argued as being important in achieving success from the ISO-14001. Managerial commitment is an important factor (Chan & Wong, 2006;Curkovic & Sroufe, 2011), while effective training and involvement of employees in the adoption of the standard also plays an important contingent role (Djekic et al., 2014;Ivanova et al., 2014). Firm size and maturity of the certification has also been found to have an effect, such that larger firms with greater resources and capabilities are known to implement the standard more successfully (Yin & Schmeidler, 2009).
A number of studies have also focused on whether the adoption of ISO-14001 has a positive impact on pollution reduction at a national (or regional) level. Boiral et al. (2018) suggest that waste minimization and air emission reduction have been the most important indicators used to assess the impact of ISO-14001 on pollution reduction at such aggregate levels. They suggest that, again, the evidence about the impact of ISO-14001 on both waste management and emission reduction is mixed; however, a large number of studies report a positive impact of the standard. Prakash and Potoski (2014), for instance, found that greater adoption of ISO-14001 at the country level is associated with reduced levels of SO 2 and particulate matter (PM10) emissions. However, whereas most studies have focused on the impact of ISO-14001 on visible air emissions, very few studies have investigated the impact of ISO-14001 on the reduction in water contamination, and these did not find that the ISO-14001 standard contributes significantly to this issue (Potoski & Prakash, 2013).

THEORETICAL BACKGROUND AND
HYPOTHESES We develop our hypotheses by combining insights from institutional theory (DiMaggio & Powell, 1983) alongside the literature on environmental stewardship (Guimaraes & Liska, 1995;Potoski & Prakash, 2013). It is well known that institutional forces (coercive, mimetic and normative) pressurize firms to conform to the 'rules of the game' (North, 1996). Institutional pressures for environmental management can be (i) 'coercive' -via strong environmental regulations enforced by countries; (ii) 'normative' -via strong civic cultures and consumers' demands or activist groups promoting green products and processes; and (iii) 'mimetic' -via imitation of competitors and other similar firms' processes and practices (Darnall et al., 2008;Delmas & Toffel, 2004). Conformance to these different pressures has been regarded as important for firms involved in trading internationally and in achieving legitimacy (or 'social license' to operate) among a wide range of stakeholders (DiMaggio & Powell, 1983).
The literature on environmental stewardship complements this view. Under this concept, some firms take on strong environmental management activities to position themselves as environmentally responsible citizens (Guimaraes & Liska, 1995). This is because they believe that being environmentally friendly will not only lead to a better efficiency of their operations, but also allow them to compete on the basis of 'greener' products and services (Iatridis & Kesidou, 2018). Environmental stewards commit a large amount of resources to monitoring, researching, fundraising and developing leadership on specific environmental issues. They also invest in learning in detail about the adverse effects of local air pollution and its effects on the ecosystem by taking field trips with environmentalists and engaging politically with the media and other lobby groups to become an effective voice in the policymaking arena (Lerner & Carr, 1994). This view is also reflected in the 'responsibility matrix' framework (Sinkovics et al., 2021a(Sinkovics et al., , 2021b, which maps firms' responsible actions in terms of their width and depth. Here, firms taking on operational activities at advanced levels are able to tackle the root causes of environmental issues, such as by encouraging their suppliers to use renewable resources, avoiding unnecessary transportation, reusing production materials and reintegrating them into their supply chains. Some firms also embed their business activities at an advanced level of responsibility by engaging in environmental innovations (Horbach, 2008;Veugelers, 2012) and by developing products and services that reduce environmental harm, such as by replacing chemical-based products with plantbased products (Sinkovics et al., 2021a). Overall, environmental stewards spill over their good environmental practices to other firms in the community both nationally and internationally (Rondinelli & Berry, 2000).

ISO-14001 and Pollution Reduction
Adopting private regulations (such as the ISO-14001) has been argued as an important manifestation of environmental stewardship, as the adoption of such voluntary standards signals a firm's commitment to environmental performance beyond legal requirements (Potoski & Prakash, 2013). Obtaining and maintaining the ISO-14001 certification requires a firm to setup an environmental management system, to make substantial investment in training its personnel to maintain the system, and to accredit the system from thirdparty auditors (Berliner & Prakash, 2013). The obligations of ISO-14001 are continual, and firms are expected to analyze their current position regarding pollution and waste management and establish a policy to reduce them using practices in day-to-day implementation (Iatridis & Kesidou, 2018). Several studies have found that the adoption of ISO-14001 leads to an improved compliance with environmental standards, as well as reductions in air pollution at the facility level (Arimura et al., 2008;Potoski & Prakash, 2005a, 2005b. We expect that (as a baseline hypothesis) greater levels of adoption of ISO-14001 should lead to reduced emission levels at the country level, as previously argued (Potoski & Prakash, 2013).
First, the relationship between greater levels of firms' adoption of ISO-14001 and country-level pollution reduction is attributed to the diffusion of the standard by environmental stewards or by firms taking on environmental responsibility at advanced levels (Sinkovics et al., 2021a). Typically, large or visible firms adopt the ISO-14001 standard themselves and require their relevant suppliers to do so (Rondinelli & Berry, 2000). As such firms lead regional or global value chains, they often (coercively) require their suppliers to 'harmonize' various management systems, such as production and accounting systems, with their own systems. In this context, requiring suppliers to adopt ISO-14001 to harmonize their environmental management system with theirs is considered a normal practice (Berliner & Prakash, 2013;Christmann & Taylor, 2001;Iatridis & Kesidou, 2018;Jiang & Bansal, 2003;Sims, 1999). Such coercive pressures cause suppliers in GVCs to enhance their responsible operational actions to maintain their legitimacy in the GVC setting (Sinkovics et al., 2021a). Eventually, as each facility undertaking the ISO-14001 standard is obligated to reduce pollution by achieving quantifiable targets, one could expect that it would contribute to reducing pollution at the national level (Potoski & Prakash, 2013).
Second, and importantly, the effect of ISO-14001 on pollution reduction is also attributed to 'spillover effects' of the standard occurring via mimetic and normative isomorphism mechanisms, as well as through knowledge dissemination by firms adopting the ISO standard. An important aspect of ISO-14001 adoption is to disseminate information about pollution throughout the industry, to rivet attention toward the resource efficiencies and technological improvements gained from minimizing hazardous waste, and to reduce uncertainty that investments in doing so will be valuable (Delmas & Montes-Sancho, 2011). Due to such knowledge dissemination, manufacturing plants operating in the vicinity of an ISO-14001 certified plant are likely to adopt 'ISO-14001-like' practices (Potoski & Prakash, 2013). Other reasons for doing so could also be because ISO-certified firms are often seen to service important foreign customers and to have better relationships with external stakeholders (Jiang & Bansal, 2003). In some cases, firms -typically small and medium-sized enterprises -adopt a partial set of ISO-14001 implementation methods because they lack the resources to adopt the fully 'codified' set of processes and organize an expensive third-party audit. By adopting partial processes, they claim to be 'ISO-ready' to attract consumers. Thus, non-participating firms, i.e., those outside the formal ISO-14001 'club', would also be influenced to adopt greener practices in their own facilities, as ISO-14001 adopters are expected to share their knowledge and skills in reducing pollution to other non-participants within the cluster to enable them to reduce pollution at lower costs. Thus, the overall impact of ISO-14001 in reducing the national pollution levels may reflect not only the efforts of firms adopting the standard, but also the diffusion and spillover of greener practices to other firms.
Overall, similar to the arguments made by Potoski and Prakash (2013), we expect that the adoption of ISO-14001 by firms should diffuse through both participating and non-participating firms in countries and should result in reduced levels of air pollution in the country. Based on this, we propose the following hypothesis: Hypothesis 1: Greater levels of ISO-14001 adoption in countries should lead to reduced levels of air pollution.

The Moderating Effect of Trade Agreements with Environmental Provisions
As previously suggested, there is limited prior research showing how private governance can interact with international trade policy instruments in terms of pollution reduction. Work on the apparel sector in southern Africa demonstrates how changes in regional value chains are a function of both private decisions by firms and the policy constraints and opportunities created by regional trade agreements (Pasquali et al., 2020). Similarly, research shows that sustainability initiatives at the corporate level can be driven not merely by stakeholder pressure, but also by domestic regulatory change that incentivizes behavior (Lashitew, 2021).
A variety of reasons have led to the introduction of non-trade (i.e., sustainability) issues into trade agreements (Bechtel & Tosun, 2009;Lechner, 2016). A key reason, drawing from the institutional perspective, is policymakers' own preferencesespecially in developed countries -that are also lobbied through NGOs and other environmental activists (Postnikov, 2014). The European Union (EU), for instance, is widely seen as the ''greenest'' of major economic powers. The increasing integration of EU states via the single market has also had an effect of enhancing the coherence of EU policymaking at the supranational level. The EU's experience of regulating a large internal market was essentially 'exportable' (Damro, 2012;Lavenex, 2014). The EU has used its trade policy to export its regulatory model to developing countries. This so-called 'Brussels Effect' is a version of the 'California Effect' articulated by Vogel (1986), whereby the largest and most populous US state was able to unilaterally implement environmental regulations, knowing that firms would comply or risk exclusion from the market. Across a range of issue areas, the EU has acted similarly, exporting its regulatory model in areas aside from environmental compliance to include labor standards (Sinopoli & Purnhargen, 2016).
Environmental provisions in trade agreements are intended to create a 'level playing field' with the view of reducing cross-country differences in environmental protection so that firms cannot arbitrage from these differences (Bechtel & Tosun, 2009). As previously suggested, the economic weight of the country (or the political unit) demanding these provisions also matters. The provisions are implemented through the involvement of civil society actors who provide advisory services and set up forums to facilitate the direct exchange of information between regulators involved in the trade agreement implementation process and other stakeholders, including firms. There is little evidence of 'hard' mechanisms (e.g., sanctions) if partner countries fail to comply with trade policy obligations on sustainability issues, and typically non-compliance is argued to be managed through the dispute-settlement process between trading partners (Kettunen et al., 2020). Due to such relatively 'softer' approaches in trade policy implementation regarding environmental provisions, firms exporting to a region that demands an environmentally friendly trade policy (the EU, in our case) will have greater incentives to effectively utilize voluntary initiatives, such as ISO-14001, as well as to spillover their knowledge to other firms in the vicinity to reduce national pollution levels. This is because reduced national pollution levels will engender greater stability of the trade agreement (e.g., via lesser disputes), allowing exporting firms located in the focal country to gain distinct long-term competitive advantages in terms of gaining preferential market access to the demanding country (or region). At the same time, being a member of both a 'green club' (such as ISO-14001) and being located in a country having such 'green trade agreements', an exporting firm would also be able to potentially reduce its 'liabilities of origin' (Marano et al., 2017) through reputational and legitimacy advantages. As such, we expect that trade agreements with EP provisions would complement the core effect of ISO-14001 adoption by firms in improving the positive impact on the environment through reduced levels of pollution. We thus formulate our moderating hypothesis as follows: Hypothesis 2: The effect of the adoption of ISO-14001 in countries leading to reduced levels of air pollution will be greater among countries having trade agreements with environmental protection provisions.

Sample and Data
We utilize a panel dataset of 147 countries over the 1999-2014 period to analyze how the adoption of ISO-14001 certifications by firms affects countries' emissions of a variety of gases and pollutants. The time period under analysis is dictated by the available data. We extract the number of ISO certifications per country from the ISO-14001 survey, 1 which is available from 1999 to 2017. We matched the ISO data with data on emissions from the 2018 Environmental Performance Index (EPI) report. We use information on all emissions with data over the 1999-2014 period (2014 being the latest available year): carbon dioxide (CO 2 ),  Figure 1 shows the uptake of ISO-14001 certifications over time, cumulated by groups of countries: we distinguish between (1) countries that from 1999 onwards were always members of EU trade agreements containing EP provisions (e.g., EU countries, north-African countries); (2) countries that, over this period, entered EU agreements with EP provisions (e.g., African countries, members of European Partnership Agreements, Canada, Chile, central American countries); and (3) countries that never entered EU agreements with EP provisions (e.g., Australia, India, Russia and the USA, among others), and China, which is by far the country with the largest number of certifications. The growth of ISO-14001 certifications was rapid, but it was uneven across countries, with differences emerging mostly in the post-2009 period, when the overall growth slowed down. Countries in the EU trade agreements with EP provisions kept increasing the number of ISO-14001 certifications, as did China, at an even faster rate. In countries that never entered EU trade agreements with EP provisions, instead the growth of ISO-14001 certifications seems instead to have plateaued post 2009. Joiners of EU trade agreements find themselves in the middle, exhibiting a modest growth of ISO certifications throughout the period under analysis.
In Table 1, we provide further insight into the distribution of ISO-14001 certifications across countries by showing, in terms of total ISO-14001 certifications since 1999, the top ten countries from each of the groups in Fig. 1. China is by far the country with the most certifications, followed by Japan and large European countries.

Measurements
Our dependent variable is the level of emissions of air pollutants and greenhouse gases at the country level. Compared to Potoski and Prakash (2013), who account for levels of SO 2 (for air pollution) and BOD (for water pollution), we account for a wider variety of emissions, i.e., carbon dioxide (CO 2 ), methane (CH 4 ) nitrous oxide (N 2 O) and black carbon (BLC), which are greenhouse gases, and sulfur dioxide (SO 2 ) and nitrogen oxide (NO x ), which are air pollutants. The emission levels are obtained from the EPI report (Wendling et al. 2018). The EPI report has been widely used to measure a country's environmental performance (Atici, 2009;Roy & Goll, 2014;Zhou et al., 2008).
Our independent variable is the aggregate number of ISO-14001 certifications at the country level. These were obtained from the ISO, following Potoski and Prakash (2013).
For our moderating effect, to assess the role of EU trade agreements that feature EP provisions, we combine the data on the content of trade agreements from the 'Design of Trade Agreements' (DESTA) project (Dür et al., 2014), together with data on the degree of legalization of each agreement built by Lechner (2016). DESTA allows us to identify countries that become part of trade agreements, the year in which the agreement entered into force, and the countries that participate in each agreement. Lechner's data are based on the set of agreements in the DESTA project and extend its content identifying agreements with environmental protection (EP) provisions, as well as the degree of EP legalization of each agreement. Lechner (2016) identifies all the EP provisions included in an agreement and classifies them according to three dimensions of legalization -'obligation', 'precision' and 'delegation' -according to a methodology first introduced by Abbott et al. (2000). A score for each dimension is computed based on the total number of areas covered. 3 We then construct two trade policy indicators to (1) identify countries that are part of EU trade agreements and feature EP provisions and (2) identify countries that are part of EU trade agreements with an above median legalization of EP provisions. The rationale behind the construction of two indicators is that EU trade agreements differ widely in their coverage of EP issues, with a marked gap representing the EU Treaty of Lisbon in 2009: it is since then that ''sustainable development'', the umbrella under which EP issues are included in trade agreements, has become one of the key principles of the EU's trade policy (Borchert et al., 2021). EP principles have acquired a significantly larger dimension in all the ''new generation'' trade agreements, starting with the EU-Cariforum agreement, marking a neat distinction between the agreements signed since the late 2000s and the preceding ones.
To isolate the effect of ISO-14001 certifications on emissions, we compile a list of control variables that are likely to impact national pollution levels by following Potoski and Prakash (2013). Since a country's wealth, urbanization and industrialization have been argued to impact pollution levels (Grossman & Krueger, 1995), we extract data To account for the role of international factors, we use the share of exports in GDP (WDI data) and the stock of inward Foreign Direct Investment (FDI) from UNCTAD: a higher level of exports, signaling trade dependence, can lead governments to tolerate higher pollution levels, with a similar argument being made for FDI (Vogel, 2008).
To account for political and regulatory factors, we include in the analysis a country's 'level of democracy' (Polity2 database) (Fredriksson et al., 2005) and, importantly, the effect of 'environmental regulations coming from international treaties'. This latter variable was constructed utilizing the number of each country's environmental treaty commitments in each year from the International Environmental Agreements (IEA) database (Mitchell, 2002(Mitchell, -2019. 4 Other than on the countries part of the agreements, environmental treaties could also have an effect on emissions through exports, as importers whose citizens demand environmental protection might indirectly pressure the exporter to reduce its level of pollution (Vogel, 2008). To capture this aspect, we construct a variable that measures each country's exports as a proportion of its GDP, weighted by the number of environmental treaties in the destination country.
Finally, we control for the share of each country's 'exports directed to the EU'. This variable serves two purposes. On one hand, we account for the effect of EU citizens' demand for 'green' products on exporters' emissions. On the other hand, we reduce concerns of omitted variable bias, which could arise from the likely relation between the presence of EP provisions in trade agreements and the importance of the EU as a trade partner.
This list of control variables accounts rigorously for time-varying country-specific confounding factors, such as countries' size and wealth, industrial activity, international economic involvement (trade and investment), and regulatory policies. Our empirical approach, described in the previous section, also controls for all time-constant factors within each country (country fixed effects) and all time-varying effects common to all countries (time fixed effects). Appendix 1 in the 'Supplementary file' 5 provides a full list of our variables and measures.

Empirical Model Specification
To estimate how countries' emissions are affected by the adoption of ISO-14001 certifications, together with the impact of EP provisions in EU trade agreements, we follow the empirical approach of Potoski and Prakash (2013) and use a generalized method of moments (GMM) estimator in a dynamic panel setting. We formally estimate the following equation: where Y i;t denotes the level of emissions of country i in year t, Y i;tÀ1 denotes the level of emissions in year t-1, ISO14001 i;tÀ1 denotes the number of ISO-14001 certifications in country i in year t-1, TA EU i;tÀ1 is a binary variable denoting countries part of EU trade agreements with EP provisions in year t-1, X i;t denotes a vector of control variables at the countryyear level, r i denotes country fixed effects, s t denotes time fixed effects, and e i;t denotes an idiosyncratic error term. The main equation is estimated with the system GMM estimator developed by Blundell and Bond (1998). The system GMM jointly estimates the dynamic model both in differences and in levels (of emissions), using lagged levels as instruments for the regression in differences and lagged differences as instruments for the regression in levels. 6 This estimator allows us to account for the strong persistence in the level of emissions by using a lagged dependent variable while simultaneously addressing a series of endogeneity concerns. The use of a lagged dependent variable in a specification with fixed effects at the level of the cross-sectional panel dimension (i.e., countries, in our application) is problematic, as it generates a correlation with the error term (Nickell bias). This requires instrumenting Y i;tÀ1 , and the GMM estimator does so by utilizing further lags of the dependent variable. Additionally, the number of ISO-14001 certifications could spur reverse causality concerns. Utilizing lags of ISO-14001 certifications as instruments in the GMM setting allows us to deal with this source of endogeneity. 7 The main coefficients of interest are c, the effect of ISO-14001 certifications, #, the effect of being part of an EU trade agreement with EP provisions, and q, which picks up the differential effect of ISO-14001 certifications for countries part of such trade agreements. We expect c and # to be negative, implying that a larger number of ISO certifications, as well as the presence of EP provisions in EU trade agreements, determine a reduction in the level of emissions. The q coefficient on the interaction term is also expected to be negative if the effect of EU trade agreements is to reinforce the impact of ISO certifications on emissions.

Descriptive Statistics and Correlations
Our descriptive statistics and correlations are presented in Table 2. The correlation table shows that some of the control variables (such as GDP, GDP per capita and urban population) are highly correlated with one another. We address potential multicollinearity issues in the following ways. First, we estimate our GMM model in first differences to remove any country-specific unobservable timeconstant factors. The pairwise correlations between the variables in first-differences 8 are much lower than those in the actual levels shown in Table 2. Thus, we believe that our estimation approach further mitigates any multicollinearity concerns (Schroeder et al., 1990). Second, we also rerun the regressions by selectively dropping some of the regressors that are more highly correlated. This alters some of the estimates but leaves the results unchanged. 9 However, we believe that the model would be mis-specified if we removed some of these regressors, so we therefore opted for the specification presented in the paper. Potoski and Prakash (2013) We begin by attempting to replicate the main result of Potoski and Prakash (2013) for emissions of sulfur dioxide (SO 2 ). A perfect replication exercise is, unfortunately, impaired by the lack of our data on ISO-14001 certifications over the same period that Potoski and Prakash (2013) use in their analysis. Potoski and Prakash (2013) use data over the 1991-2005 period, whereas the ISO survey we have access to includes data that start in 1999 only. We were able to retrieve data from the ISO archives 10 on certifications from 1995 to 1999, although we were warned that these initial years cannot be considered as informative as data post-1999. 11 We nonetheless attempted to replicate the results of Potoski and Prakash (2013), utilizing data over the 1995-2005 period, being aware of the limitations imposed by the provisional nature of pre-1999 ISO-14001 data, and the lack of data from 1991 to 1995. Table 3 shows these results.

Replication of
In column 1, we adopt the same methodology as Potoski and Prakash (2013), i.e., a difference GMM estimator (Arellano & Bond, 1991). We confirm    that ISO-14001 certifications have a negative impact on SO 2 emissions, but the coefficient that we estimate is smaller and statistically insignificant. The lack of significance can be attributed to the shorter time period at our disposal, and to the difference in size.
To guide the choice of the estimator and reassure the correct specification of the GMM model, in the remaining columns of Table 3, we compare the results by utilizing different estimators and periods of analysis. In contrast with Potoski and Prakash (2013), in this paper we make a choice in favor of the system-GMM estimator. In columns 2 and 3 we compare the results from a fixed-effects (FE) ''within'' estimator and a system-GMM estimator. While biased in a dynamic panel setting, the FE estimator can be taken as a lower bound estimate of the lagged dependent variable coefficient. Bond (2002) suggests that the system-GMM is to be preferred to the difference-GMM in case the GMM estimate of the lagged dependent variable (Ln(SO2) t-1 ) is close to or below the FE estimate: the difference-GMM estimate in column 1 is well below the FE estimate in column 2, suggesting that the system-GMM is the preferable estimator to use in our context.
In columns 4 and 5 we show how the results change when we extend the period under analysis to 2014, the latest available year in our data. In column 4, the effect of ISO-14001 certifications is reduced a great deal with respect to the results up to 2005, and in column 5, where we restrict the data to the (preferred) post-1999 period, the sign We instrument the lagged dependent variable and the number of ISO-14001 certifications with lagged levels in the equation in difference (GMM lags) and lagged differences. The number of lags is adjusted across specifications to obtain estimates with valid instruments according to the AR and Hansen tests. Robust errors in parentheses; 'p \ 0.15, *p \ 0.10, **p \ 0.05, ***p \ 0.01.

Voluntary programs and emissions revisited
Mattia Di Ubaldo et al.
switches to positive: in both cases, however, the coefficients lack statistical significance.

Main Empirical Findings
Given the findings in Table 3, for the remainder of the analysis we utilize the system GMM estimator and data from 1999 to 2014. The subsequent tables report the results for emissions of carbon dioxide (CO 2 ), methane (CH 4 ), nitrous oxide (N 2 O), and black carbon (BLC), as well as air pollutants, such as sulfur dioxide (SO 2 ) and nitrogen oxide (NO x ). The coefficient on the lagged dependent variable is positive and strongly significant in all the reported specifications and in all the results tables. This is an expected finding, confirming that emissions have a very high degree of persistence from one year to the next and that our empirical model needs to include this regressor.  Estimates are obtained with a system-GMM estimator. For N 2 O, we instrument the lagged dependent variable and the number of ISO-14001 certifications with lagged levels dated from t-5 to t-8 in the equation in difference and lagged differences dated t-1 in the equation in levels. For BLC, we instrument the lagged dependent variable and the number of ISO-14001 certifications with lagged levels dated from t-2 to t-4 in the equation in difference and lagged differences dated t-1 in the equation in levels. Robust errors in parentheses; 'p \ 0.15, *p \ 0.10, **p \ 0.05, ***p \ 0.01.     Table 4 shows the results for emissions of nitrous oxide (N 2 O) and black carbon (BLC). For nitrous oxide (columns 1-4), neither ISO-14001 certification nor membership in an EU trade agreement with EP provisions appears to have an impact on emissions. For black carbon (columns 5-8), ISO-14001 certification appears to have no impact on these emissions after all the controls are added in the estimation. There is a negative effect emerging (column 6) when inspecting the role of trade agreements with any degree of legalization of EP provisions: surprisingly this is found to be significant only for countries with no ISO certifications. When using the stricter trade policy indicator, no negative effect on emissions is detected, with no significant interaction with the ISO certifications. Taken together, for black carbon there is little evidence that either ISO certification or EU trade agreements influence countries' emissions. Thus, in the context of nitrous oxide (N 2 O) and black   Estimates are obtained with a system-GMM estimator. For SO 2 , we instrument the lagged dependent variable and the number of ISO-14001 certifications with lagged levels dated from t-2 to t-4 in the equation in difference and lagged differences dated t-1 in the equation in levels. For NO x , we instrument the lagged dependent variable and the number of ISO-14001 certifications with lagged levels dated from t-3 to t-9 in the equation in difference and lagged differences dated t-1 in the equation in levels Robust errors in parentheses; 'p \ 0.15, *p \ 0.10, **p \ 0.05, ***p \ 0.01.

Voluntary programs and emissions revisited
Voluntary programs and emissions revisited carbon (BLC) emissions, neither Hypothesis 1 nor Hypothesis 2 is supported. Table 5 shows that the number of ISO-14001 certifications has a negative impact on CO 2 emissions (columns 1-4). Importantly, this negative impact is strongly driven by the subset of countries that are part of an EU trade agreement with EP provisions, with no impact of ISO-14001 certifications for countries not part of such agreements. On the other hand, EU trade agreements are not associated with lower CO 2 emissions: no impact is detected on aggregate (columns 1 and 3), and a positive coefficient is found for countries part of an EU trade agreement with EP provisions but with zero ISO certifications. This latter, apparently odd finding can be explained by comparing columns 2 and 4. Countries with no ISO certifications are those at a lower level of development and tend to have a low degree of legalization of EP provisions in their agreements with the EU. When we restrict the trade agreement indicator to identify only those countries in agreements with a high degree of EP legalization (columns 3 and 4), the positive effect for countries with zero ISO certifications is weakened, while the negative interaction between ISO certifications and trade agreements remains negative and significant. Taken together, these findings suggest that ISO certifications are crucial for lowering CO 2 emissions, but only for countries that are part of EU trade agreements with EP provisions. Simply being part of an EU trade agreement with EP issues, however, especially where EP issues are not strongly legalized, has no impact on reducing CO 2 emissions. Thus, in the context of carbon dioxide (CO 2 ) emissions, both Hypothesis 1 and Hypothesis 2 are supported. Table 5 also shows the results for emissions of CH 4 . Similar to CO 2 , ISO-14001 certifications have a strong and negative impact on methane emissions. The impact is robust to including all controls, but it does not appear to be driven by members of EU trade agreements with EP provisions. Interacting the trade agreement indicators with the number of ISO certifications in columns 6 and 8 cuts the coefficient of ISO certifications by approximately half, but does not yield a statistically significant coefficient on the interaction term. This implies that ISO certifications reduce methane emissions regardless of the countries' membership in EU trade agreements with EP provisions. Finally, there is no statistically significant impact on methane emissions from these trade agreements, per se. Thus, in the context of methane (CH 4 ) emissions, Hypothesis 1 is supported, but Hypothesis 2 is not. Table 6 explores the effect of ISO certification and EP issues in EU trade agreements on air pollution (SO 2 and NO x ). We first examine the impact of emissions of sulfur dioxide (SO 2 ). For this pollutant, ISO-14001 certifications are found to have a positive effect on emissions. On a more positive note, however, being part of EU trade agreements containing EP provisions, in particular agreements with an above median legalization of EP provisions, has a negative impact on SO 2 emissions. The negative effect of trade agreements is larger than the positive effect of ISO certifications, but the interaction term between the two variables is not statistically significant. Finally, nitrogen oxide (NO x ) emissions also appear to be negatively affected by membership in EU trade agreements with EP provisions, but not by ISO certifications. Thus, in the context of sulfur dioxide (SO 2 ) and nitrogen oxide (NO x ) emissions, neither Hypothesis 1 nor Hypothesis 2 is supported.

Robustness Tests
To test the robustness of our findings, we conducted a series of further analyses. First, we re-ran our analyses by excluding EU member countries from our sample. This is because of Hypothesis 2, where we argued that trade agreements with the EU having environmental protection (EP) provisions moderate the effect of ISO-14001 on emissions. Although EP provisions in these trade agreements apply to both EU and non-EU members, one may argue that the EU is 'demanding' the inclusion of these provisions in trade agreements. It is therefore of interest to inspect whether EP provisions also have an effect on emissions of the subsample of countries on which these provisions are 'imposed'. Our results are presented in the online 'Supplementary File' in Tables A1.1, A1.2, and A1.3. All of our main results are confirmed.
Second, since the ISO survey was at its preliminary stages at the beginning of our observation period (1999), it may be possible that the effect of the ISO-14001 certifications on emissions might have taken some time to materialize. Additionally, the work of Potoski and Prakash (2013), which we replicate and substantially extend, utilized data over the 1995-2005 period. It is of interest, therefore, to explore whether restricting our analysis to more recent years shows results that differ significantly from those obtained with the full data. This Voluntary programs and emissions revisited allows us to exclude particular effects that might have been at work during the early years over which firms could obtain the ISO-14001 certification, and to conduct an analysis over a set of years that do not overlap with those examined by Potoski and Prakash (2013). In this test, we restrict the time dimension of our panel by dropping one year of data at the time to observe whether the impact of ISO-14001 certifications and EU trade agreements with EP provisions is affected by the exclusion of the initial years of our data. Our results are presented in Tables A2.1-A2.6 of the online 'Supplementary File', where we re-estimate our main empirical models over different time periods. The empirical results are broadly confirmed.
Third, because institutional pressures on environmental protection will increase with a greater population (Aravind & Christmann, 2011), we investigate whether our results hold if we used the per-capita emissions of the countries rather than the overall aggregate volume of emissions (as included in our main analysis) or by controlling for the population of the countries. We re-ran our analysis by doing so. The results are presented in the online 'Supplementary File' in Tables A3.1.1-A3.3.3. Our results, largely, remain intact.
Finally, we re-estimated our main empirical models by adding the (log) value of export of mineral fuels as a control variable. This variable includes exports of coal and peat, petroleum and derived products, and gas, both natural and manufactured. The data for this test were obtained from the UN Comtrade database, from which we used exports of Section 3 of the Standard International Trade Classification (SITC) for all the countries in our sample. Controlling for exports of fuels could be of relevance if this variable were correlated with our main variable of interest, i.e., the number of ISO-14001 certifications or the presence of a trade agreement with environmental provisions, thereby causing an omitted variable bias. The results are presented in our online 'Supplementary File' in Tables A4.1, A4.2, and A4.3. All our results remain intact.

DISCUSSION AND CONCLUSION
The great challenges of global warming and climate change certainly require coordinated efforts by a variety of stakeholders, including businesses and governments, on a global scale. Emissions of air pollutants and greenhouse gases significantly contribute to these issues and will increasingly have an adverse impact on human health and sustainable development (Kolk & Pinkse, 2008). The spread of private regulatory standards for environmental management, importantly the ISO-14001, has been argued to prevent MNEs from locating their polluting subsidiaries in countries with lax environmental regulations (Christmann & Taylor, 2001), although the impact of this standard on the environment is contested (Boiral, Guillaumie, Heras-Saizarbitoria, & Tayo Tene, 2018). There is extensive literature on the reasons for the adoption of ISO-14001, which includes the role of the expectations of companies higher up the supply chain on lower tier suppliers (Boiral et al., 2018). Our paper seeks to build on and extend the existing literature on the impact of both the private ISO 14001 standards and that of international governance mechanisms on the emissions of air pollutants and greenhouse gases (Delmas & Montiel, 2008;Jiang & Bansal, 2003;Potoski & Prakash, 2013).

Theoretical Implications
Our paper looked at (1) whether ISO-14001 levels are associated with reduced levels of greenhouse gases (CO 2 , CH 4 , N 2 O and BLC) and air pollutants (SO 2 and NO x ) and (2) how ISO-14001 certification interacts with broader political and regulatory expectations by considering EU trade agreements and the environmental protections contained therein. In Hypothesis 1, we expected that greater levels of ISO-14001 adoption would reduce national emissions of both greenhouse gases and air pollutants, following the same theoretical logics as Potoski and Prakash (2013), and complementing with environmental stewardship and MNE responsibility frameworks (Dau et al., 2021;Guimaraes & Liska, 1995;Kolk, 2016;Kolk & Van Tulder, 2010;Sinkovics et al., 2021aSinkovics et al., , 2021b. The results paint a somewhat more complicated picture than might be expected: ISO-14001 certifications are associated with lower emissions of the greenhouse gases CO 2 and CH 4 , but not in the case of N 2 O and BLC. An explanation of this could be that CO 2 and CH 4 comprise 91% of greenhouse gases and are thus 'highly visible'. CO 2 , in fact, comprises approximately 81% of the greenhouse gases and is thus a leading contributor to climate change. It is thus plausible that there is greater focus on reducing CO 2 and CH 4 emissions within the ISO-14001 standard; however, this warrants further research. This finding also implies that corporations focus on sustainability matters that are prominent or salient to the wider public because of potential reputational effects, as distinct from a desire to achieve an environmental goal (Lashitew, 2021). It also implies that most firms, especially those engaged in GVCs, apply sustainability standards at operational levels to maintain their legitimacy, and only a few firms may be able to undertake deeper levels of responsible actions (Sinkovics et al., 2021a).
We also find that ISO-14001 certifications are not associated with lower levels of emissions of air pollutants SO 2 and NO x. Other than unexpected, this finding is at odds with the results of Potoski and Prakash (2013), who found a negative impact of ISO-14001 on SO 2 emissions. A possible explanation of this could be that due to tighter public regulations in reducing these 'harmful' industrial pollutants in many emerging economies (Ni et al., 2019;Pei et al., 2020), the utility of ISO-14001 in reducing these emissions might have been reduced. This can also be reflected in our finding that membership of an EU trade agreement with EP provisions is (directly) associated with lower emissions of SO 2 and NO x .
In relation to Hypothesis 2, the moderating effect of EU trade agreements with EP provisions on the association between ISO-14001 and reduced emissions is only supported for the case of CO 2 . As such, the joint effect is seen only in the case of the highly visible greenhouse gas, but not in the case of other pollutants. Again, one could argue that this is due to the larger focus on CO 2 in climate change discussions, thus engendering collective action at a global level (Ostrom, 2010). This is also largely due to countries' commitment to reducing CO 2 emissions under the Kyoto Agreement and the greater focus on reducing GHGs in the UN's Sustainable Development Goals (Goal 13). Our findings in this regard speak to the complex interactions of international trade agreements, national governance, and firm responses. Dau et al. (2021) suggested that national regulatory quality was an important element in the transfer of international agreements to local firms. Gereffi et al. (2021) examined how trade agreements affect firm strategies, arguing that trade restrictions present opportunities for firms to upgrade their operations or reconfigure them. There is no reason to think that it is only due to traditional trade restrictions; voluntary export restraints, for example, are also implicated here. Our findings suggest that firms adapt to environmental provisions in much the same manner.

Policy Implications
The findings from our study offer a number of policy implications for both public policymakers and firms. Since reducing pollution requires a global and collective effort, the United Nations (UN) has proposed the National Sustainable Development Strategy (NSDS) since 1992 to call upon countries to integrate economic objectives with social and environmental initiatives. The UN also encourages governments to incorporate the 17 Sustainable Development Goals (SDGs) into their national strategies to achieve the 2030 agenda. In this context, the European Union (EU) is widely seen as having a strong commitment to sustainability, as manifested in the non-trade elements of its international relationships (Lashitew, 2021). As such, for public policymakers, our findings based on the EU's trade agreements with sustainability provisions show that such trade policy arrangements with environmental provisions positively interact with private or voluntary systems, such as ISO-14001, in reducing emissions of GHGs (especially CO 2 ) at the national level. We also find that environmental policy provisions in trade agreements have a direct effect on reducing harmful SO 2 and NO x pollutants. In short, for policymakers we present a positive case of trade policy arrangements with environmental provisions, as these tend to be effective (at least partially) in achieving climate change and other environmental objectives (SDG-13, for example).
Our empirical findings also indirectly suggest that developing countries (with lower levels of GDP per capita) are more likely to pollute. Developing countries (e.g., the global South) are important locations for MNEs to conduct their manufacturing activity and to export intermediate and finished products to global markets. Policymakers in these countries are often pressured, on the one hand, to lower their environmental regulations in order to attract pollution-intensive manufacturing activity and maintain trade competitiveness, leading many developing countries to be ''stuck at the bottom'' (Porter, 1999). On the other hand, they are also pressured to engage in sustainable development, despite internal institutional voids and governance gaps in enforcing strong environmental regulations (Blackman, 2008;Cave & Blomquist, 2008). According to our findings, developing countries can approach this dilemma by negotiating trade agreements (e.g., lower trade barriers) that include sustainability provisions with their major export Voluntary programs and emissions revisited markets (such as the EU, in our case), since environmental provisions provide a 'soft law' approach and complement voluntary programs mandated by lead firms in GVCs. Brandi et al. (2019) find that non-trade provisions in trade agreements represent credible commitments, and translate into regulations that impose higher costs on the operations of foreign multinationals. Further, Brandi et al (2020) find that environmental provisions in trade agreements reduce dirty exports and increase green exports from developing countries. The ''greening'' of trade agreements also enjoy strong public support in developed markets (Esty, 2001), potentially reducing the liabilities of origin (Marano & Kostova, 2016) of imports from developing countries. Overall, our work relates to other observations about the interaction of policy -particularly EU policies -on voluntary standards and corporate responses to societal pressures (Lashitew, 2021).Our work has also sought to engage with Van Assche's (2018) call for international business (IB) scholarship to gain a seat at the policymaking table alongside international economics. The relative absence of an account of the interaction of policy and corporate activity has placed IB scholarship in a peripheral position (Gereffi, 2019). We also engage with calls for IB scholarship to respond to the great challenges of our time, in this case climate change and pollution (Buckley et al., 2017).
For managers, although our analysis does not account for micro-firm-level factors, we do present a positive case for the adoption of voluntary standards, as well as for engaging in deeper levels of responsible behavior in environmental management (Sinkovics et al., 2021a). Managers have mixed perceptions about whether the adoption of voluntary standards and conformance to public policies makes a real difference (Cacioppe et al., 2008;Murmura et al., 2017). They are also constrained by firm-level resources and capabilities in adopting responsible actions and thus engage in such actions at different levels in terms of their width and depth (Sinkovics et al., 2021a). However, at the same time firms are increasingly required to report their responsible actions and link them to wider sustainability frameworks such as the UN's SDGs and the UN Global Compact's principles. Our findings show that for CO 2 (carbon dioxide) and CH 4 (methane), which comprise approximately 91% of the GHGs, the adoption of ISO-14001 significantly reduces their emissions. Thus, the adoption of voluntary environmental standards (especially ISO-14001, which we focus on) is associated with lowering GHG emissions; goal 13 under UN SDGs and principles 7, 8, and 9 under the UN Global Compact. As there has been mixed evidence (and belief) on whether ISO-14001 indeed helps reduce emissions, our positive finding on the impact of this standard can be encouraging for firms to adopt this standard.

Limitations
Despite our contributions, and like all academic studies, we acknowledge that our study has limitations, and we suggest that these provide worthwhile avenues for further research. First, our study is based on cross-country data on both ISO-14001 adoptions and emission levels and lacks a firm-level analysis. Analysis at the firm or MNE level would have allowed us to explicate the deeper nuances of the competitive effects of firm and industry characteristics and would have significantly improved our contributions and implications. Secondary data on firm-level adoption of ISO-14001 certifications are, however, very rare and would have required us to conduct a questionnaire survey, which itself results in data limitations. Our limitation of using country-level data also limits us from testing the effect of firm-level factors (in our case, ISO-14001 adoption) on country-level outcomes (i.e., reduction of emissions) in a more robust way. One may argue that the 'aggregation' logic of firm-level decisions on country-level outcomes would require a certain threshold of firms in adopting ISO-14001 to have a normative force on other firms to adopt environmentally friendly practices to subsequently impact country-level emissions. This could be possible if we had the data at a firm level or at subnational levels and by using multilevel techniques of analysis (Hitt et al., 2007). We suggest future research to explore this. Second, despite ISO-14001 being a globally and widely adopted environmental certification standard, we also acknowledge that we do not account for other standards, such as the EU's Eco-management and Audit Scheme (EMAS) and Responsible Care. Both of these standards are, however, region specific -for example, the EMAS is diffused more widely in the EU, whereas Responsible Care is diffused more widely in the United States' chemical industry. Including a number of standards would again have reduced our sample significantly and would have undermined the generalizability of our findings regarding ISO-14001. We suggest that future studies can consider the aforementioned possibilities.

Voluntary programs and emissions revisited ACKNOWLEDGEMENTS
We would like to thank the European Union for funding our research under the Horizon 2020 scheme. NOTES 1 https://www.iso.org/the-iso-survey.html.

2
The EPI data also includes other emissions, such as nitrogen oxide (NOx) and black carbon, but these are only available up to 2010. This reduced time period results in issues in the empirical analysis, as we detail below. 3 We refer the reader to Lecher's 2016 paper for more details on the calculation of the legalization score.  Estimating the model in terms of both differences and levels addresses the weak instrument problem arising from using lagged-levels of persistent explanatory variables as instruments for the regression in differences (Blundell and Bond, 1998). 7 As we interact the number of ISO-14001 certifications with the trade agreement indicators, we instrument these latter interaction terms with lags of themselves as well. 8 The table with the correlations of the variables in first-differences can be made available upon request. 9 We re-ran the regressions by dropping the highly correlated regressors and these results are also available upon request. 10 We thank Laurent Charlet from the ISO for help with this.
11 The ISO-14001 survey was at its very preliminary stages in 1995, and for this reason a decision was made by ISO to only make the 1999 onwards data available in the online version of the ISO survey. Additionally, another limitation is that the sample of countries included in the 1995-1999 data is only half of that included in the 1999-onwards data.

ABOUT THE AUTHORS
Mattia Di Ubaldo is a Research Fellow in Economics at the University of Sussex Business School, and a Fellow of the UK Trade Policy Observatory. His current research spans several areas: trade policy effects on trade, FDI, and non-trade policy objectives; deep trade agreements; Brexit and its impact on firms and workers. He holds a PhD in Economics from the University of Sussex.

Steven McGuire is Professor of Business and Public
Policy and Head of the University of Sussex Business School. He has held previous academic appointments at Aberystwyth University, Audencia Nantes School of Management, the College of Europe, and the University of Bath. His research interests are in the areas of international political economy, international business and corporate political activity.
Vikrant Shirodkar is a Senior Lecturer in International Business at the University of Sussex Business School. His primary research interest is on the nonmarket (political and social) strategies of multinational enterprises (MNEs). He is also interested in the strategies of MNEs based in the context of emerging markets.