Department of Political Science

(1)

Department of Political Science

‘Fossilfritt Sverige’

Which Companies Engage with the Climate initiative, and Why?

Justus Vasänge

Independent Research Project in Political Science, 30 credits International Master’s Programme in Political Science Year, Term: 2021, Spring

Supervisor: Naghmeh Nasiritousi

(2)

‘Fossilfritt Sverige’

Which Companies Engage with the Climate initiative, and Why?

Abstract

The climate initiative Fossilfritt Sverige, largely comprised of companies, has been advanced as an integral part for the national climate ambition of Sweden becoming a fossil free welfare state by 2045. This paper seeks to explore and provide insights about the relationship between company characteristics, and motives for, voluntary engagement in the initiative. The study uses a mixed method approach to test hypotheses and expectations regarding such

relationships. In doing so an original dataset of 400 of Swedish companies is analysed using binary regression analyses along with a thematic analysis of 15 previously carried out interviews with company representatives. The findings of the study support a hypothesis of ISO 14001 certified companies being more likely- and further indicates that companies with higher annual net revenues are less likely to engage with the initiative. Moreover, the findings indicate that engagement is especially influenced by competitive motives. It is further

(3)

1. Introduction

In December 2015, the Paris Agreement (PA), an agreement with the ultimate objective of dealing with climate change and its effects, was adopted. The PA among other things states that “Developed country Parties should continue taking the lead by undertaking economy-wide absolute emission reduction targets…” (United Nations 2015, Article 4). Consequently, the European Union (EU) has enshrined in law a goal of net-zero greenhouse gas (GHG) emissions by 2050. Even more ambitiously, the Swedish government in 2015, prior to the finalisation of the PA, announced its overarching goal to become one of the world's first, and later to be the first, fossil free welfare state (Regeringskansliet 2015; 2017). Further

commitment to climate mitigating actions in Sweden continued in the following years with the adoption of a Climate Policy Framework in 2017. The framework contains an underlying goal of reaching net-zero GHG emissions by 2045, environmental laws persistent to political changes, and features an independent expert review body called the Climate Policy Council (Naturvårdsverket 2020).

(7)

2

reduce emissions and are encouraged to with their industry specific expertise jointly generate roadmaps of how various industries may realize a fossil free state. Subsequently, in the Climate Policy Council’s latest report (2020), the initiative Fossilfritt Sverige is advanced as a potentially essential part in bringing about further, and intensifying current, climate

mitigation commitments and engagements. Similar pronouncements, emphasising the significance of the initiative, are conveyed in one of the government's sustainable

development goals report to the United Nations (UN) (SDG n.d.), and moreover in the long-term emission reduction strategy report to the EU (Regeringskansliet 2019).

The increased importance of the private sector and public-private collaborations in emission mitigation and adaptation measures has increasingly been illustrated in the literature (e.g. Hickmann 2017; Bäckstrand 2008). Accordingly, an intensified attention has been directed towards how businesses may improve their environmental performance together with how such performances may be evaluated. However, less academic attention is placed on why and

which companies take a meaningful part in enhancing their environmental commitments.

Expanded insights into which and why companies voluntarily participate in a climate initiative as Fossilfritt Sverige may provide guidance to future national and sub-national policy formulation to further encourage, orchestrate, and scale up emission reduction measures to reach the set-out goal by 2045. Moreover, an increased understanding of this type may provide insights into the association between various organisational characteristics and emission reduction engagements.

1.1. Purpose

(8)

3

stated to declare the necessity and desirability of Sweden to be at the forefront of becoming fossil free (Regeringskansliet 2015). Achieving the goal of the initiative is conveyed to not only be a social responsibility but also competitively beneficial (ibid.). While a participant's failure to fulfil its set out climate objective would not be punished, at least not formally so, the initiative seeks to stimulate its participants, and non-participants, to undertake climate change mitigation activity. Expressed in the initiative’s declaration is similarly to act as a role-model and inspire others to follow their lead (Fossilfritt Sverige n.d.). The intention of this study is to study which type of companies, and why, voluntarily engage in the initiative Fossilfritt Sverige and thus stand behind such statements. Given that companies are a dominating and essential group of actors in the initiative, together with the voluntary nature of engagement, there is a useful possibility to study the characteristics of, and motives for, companies taking part in the initiative. Currently, there exists a limited body of literature studying company characteristics and their relationship with engagement in voluntary initiatives. Furthermore, there is a shortage of studies addressing alike investigations with a quantitative approach together with a qualitative perspective (Balasubramanian et al. 2020).

This study thus contains two components. The first is to quantitatively study the relationship between companies and voluntary emission reduction engagements, conveyed through engagement in the climate initiative Fossilfritt Sverige. This is implemented through binary logistic regression analyses of an original dataset consisting of 400 observations with data of 200 companies that are engaged in the initiative, and 200 that are not. The aim of this

component revolves around studying the association between certain company characteristics and voluntarily choosing to intensify emission reduction ambitions. The second is to

(9)

4

1.2. Research questions

● What is the relationship between certain company characteristics and engagement in the climate initiative Fossilfritt Sverige?

● Why do companies engage with the climate initiative Fossilfritt Sverige?

2. Previous research

This study takes its place amongst a diverse body of research concerning the role and effect of non-state actors in environmental governance. The global climate governance architecture has been found to be particularly complex with participation and following demands and from numerous state- and non-state actors (Widerberg 2016). As a result, various non-state or hybrid initiatives and partnerships, distinct from solely traditional formal treaties between states, across scopes of governance have been established. Although the intensified role of non-state actors in global climate governance, and its importance, is recognized in both academic literature and state-led organisations, it is not illustrated without its concerns. For example, the increased authority of non-state actors possibly being in zero-sum relationship with state authority has been found to be a concern for developing countries (Biermann & Pattberg 2008:282).

Other concerns relate to an increasing incoherence in priorities of objectives. An increasing trend in climate governance is therefore the move from state actors and state-led

(10)

5

interests rather than environmental interests being the main driving force for policy formulation (ibid.:406). Moreover, in Bulkeley et al.’s (2012) evaluation of sixty

transnational climate initiatives and their functions, they found that private-private initiatives are more inclined to create enforcing mechanisms while hybrid initiatives instead are inclined to utilize voluntary and incentive-based mechanisms (ibid.:609), with initiating actors mainly claiming expertise as the source of legitimacy. Although relying on voluntarism is not a feature that necessarily is a drawback, it asks for further inquiries into the motives for voluntary engagements of private actors in PPPs.

Non-state actors such as companies have been found to increasingly commit to climate mitigating engagements voluntarily, often through PPPs, which in part is explained in the literature as states and state-led organisations actively seeking to encourage non-state climate mitigation actions (Hale 2016). However, although the business-sector is often affected by the actions and rulemaking of the state, they themselves have a notable role in influencing decisions and setting agendas (Biermann & Pattberg 2008). Businesses have been found to have several strategies, and motives, to take on environmental commitments including joining certification schemes or taking on corporate social responsibility (CSR) practices. One such certification scheme is the environmental management adoption, and subsequent certification, of the voluntary environment management system standard series ISO 14000 developed by The International Organization for Standardization (ISO). Previous studies have found that, besides potential economic gains, legitimising and marketing motives are significant factors influencing adoption (Darnall 2006:374). Thus, participation in PPPs is only one way for businesses to legitimise their activities.

(11)

6

independently (ibid.:7). She finds that private authority materializes from supply and demand and does not necessarily conflict with public authority but might rather synergize with it (ibid.:9). Such claims are based on the benefits for both public and private actors when authority is claimed or accepted, often associated with the legitimising and beneficial expertise the private actor brings.

Voluntary engagements with climate practices by non-state actors may often be regarded as trying to claim entrepreneurial private authority when it is found to grant benefits. Voluntary business-based initiatives have been advanced as a significant and influential component in developing effective national climate policies (Hickmann 2017:102). A significant reasoning for such ideas is the substantial environmental impact of business industries (Etayeb & Zailani 2009:94), and companies increasingly finding voluntary engagements as desirable (Hickmann 2017:95). Chan et al. (2018) further highlights the importance of non-state actors in climate governance. Nonetheless, in his evaluation of climate mitigating actions post-PA he suggests that superior ways of motivating private actors to participate must be found (ibid.:33). In the later work, the importance of coordinating and mobilizing action in a national context is put forward (Chan et al. 2019). The primary obstacle is argued to be to encourage engagement beyond the initial actors making voluntary climate mitigation commitments (ibid.:6). Similarly, Hale (2016:19) asks under what conditions the more reluctant companies decide to engage in climate mitigation. Thus, this seems to indicate that further insights into which, and why, companies choose to engage might prove useful for mobilizing further voluntary engagements.

In sum, previous research has outlined non-state actors as an increasingly involved group of actors in climate governance. While research finds that this change in governance

architecture is not without its concerns, non-state actors are illustrated as crucial components. Thus, types of collaborative governance have escalated, often with a voluntary condition for engagement and commitments. Moreover, with voluntary engagements being put forward as essential for both climate targets and policy formulation there are therefore valuable insights to be gained in terms of which, and why, companies voluntarily engage with climate

initiatives. This study seeks to contribute with insights into such inquiries. Firstly, by

(12)

7

motives of companies voluntarily engaged in the initiative will be a part of study. This study on this account seeks to contribute to the literature on private actors, and their collaborations with public actors, by examining what characteristics are associated with- and motives that lead companies to voluntarily engaging with the climate initiative.

3. Theoretical framework

This section is divided into two separate segments. The initial part will convey theories and findings related to which type of companies, in terms of characteristics, choose to engage with voluntary environmental practices. These are used to classify characteristics, or independent variables, of interest and subsequent hypotheses of how they are expected to influence the likelihood of companies engaging with the initiative. The second part will then form a conceptual framework directed towards why companies choose to voluntarily engage in environmental practices. This second segment will outline a conceptual framework formed from previous theories and findings, which will then be used to analyse the qualitative

material of the study.

3.1. Company characteristics and voluntary

environmental engagements

Literature assessing the effect of company characteristics on issue areas exists across a number of fields of research. While such research does exist in the environmental field, it is a topic relatively unexplored (Balasubramanian et al. 2020). Businesses are in the literature often portrayed as profit-maximizing actors that also must take into consideration the

demands of societal actors (Ervin et al. 2013). Decisions about voluntarily participating in an initiative, or committing to other environmental practices, can thus differ between various businesses depending on demands, costs, and other company circumstances. Although there is no clear consensus on which company characteristics are of interest, how these are

(13)

8

as relevant in this context. Among them are variables concerning company size, business-sector, and environmental engagements separate from the specific environmental procedure or operation of interest (Balasubramanian et al. 2020). The relationship between engagement in the climate initiative Fossilfritt Sverige (the dependent variable) and the company

characteristics (the independent variables) of size, revenue, business-sector, and ISO 14001 certification, is investigated in this study. These independent variables are considered as internal features of the company which, motivated by previous research findings, affect their environmental practices. A few selected characteristics, here considered as more indicating external features, will then act as control variables.

Independent variables

3.1.1. Company size

The size of the company is a commonly studied factor when assessing the relationship between company characteristics and voluntary environmental engagements. Although company size occasionally has been measured through strictly classifications of monetary resources such as annual revenue (Hourneaux et al. 2014), the majority of studies has operationalized company size as the number of employees (e.g. Collins et al. 2007; Min & Galle 2001, Etayeb & Zailani 2009; Jabbour et al. 2016). The hypotheses of how company size affects the engagement in voluntary environmental practices includes their incentives and barriers to internal and external elements. Internal characteristics amongst other things linked to knowledge and expertise (Min & Galle 2001), and financial resources (Kasseeah 2020). Together with external characteristics such as the extent of pressure and number of stakeholders (Collins et al. 2007). Additionally, Swedish legislation requires companies with more than 250 employees to annually submit a sustainability report where the company’s work with environmental and humanitarian concerns is conveyed (Bolagsverket 2019). This is also the case for companies with an annual net-revenue larger than 350 million Swedish Kroner. Previous research has found that publishing environmental reports with following objectives and principles causes companies to increasingly take part in substantial

environmental practices (Kolk 1999). The general hypothesis in the literature is that larger companies, relative to smaller companies, are more likely to commit to voluntary

(14)

9

between company size and engagement in the initiative is therefore that larger companies, in terms of number of employees, are more likely to engage with the initiative.

Hypothesis 1: The larger the company is, the more likely it is to engage with the climate

initiative.

3.1.2. Revenue

Although the size of a company is occasionally measured through the annual revenue of the company, this study differentiates the two characteristics. Annual net revenue is additionally an often used characteristic when studying voluntary environmental engagements of

companies, although a few studies use annual purchases (Min & Galle 2001), or annual profits (Kasseeah 2020), as classification of monetary resources. Financial costs have been found to be a great barrier to environmental practices, especially for small-medium

enterprises (SMEs) (Collins et al. 2007). Similarly, the findings of Kasseeah (2020) point to that greater monetary resources have a positive relationship with environmental measures. On the contrary, other studies have found that companies with lesser financial capabilities are more prone to voluntarily engage with environmental practices due to decreased costs in production and thus resulting in increased financial performance (e.g. Molina-Azorín et al. 2009). Furthermore, in contrast to the findings of Collins et al. (2007), smaller and thus more resource constrained companies have also been found to be more likely to act upon

(15)

10

engagement in the climate initiative could be called into question. However, the presumption of this study is that the majority of submitted company objectives of climate mitigation, and industry specific roadmaps, require monetary resources- and are intended to be substantially realized. Which leads to the second hypothesis:

Hypothesis 2: The larger the annual revenue of the company is, the more likely it is to

engage with the climate initiative.

3.1.3. Business-sector

As previously suggested, which business-sector or area of business the company has its primary operation of business may also be highly influential on its voluntary environmental engagements. In Sweden, the Swedish Standard Industrial Classification (SNI) is used to categorize businesses into 21 main categories (A-U), as well as with further detailed distinctions, based on their business activities. This system is managed by the Swedish Central Bureau of Statistics (SCB) but is based on the system of industry classifications used in the EU named NACE (SCB n.d.-a). The literature on the relationship of interest frequently scrutinizes companies and voluntary environmental engagements in specific industry

(16)

11

associated with natural resources. This in turn, due to competitiveness, encourages companies to voluntarily regulate their environmental impact (Etzion 2007). The observations of this study will encompass companies with different business activities across all business-sectors. However, a distinction is made between companies manufacturing with- or extracting raw materials and companies operating in the service sector. Furthermore, in line with previous research, the following hypothesis is tested:

Hypothesis 3: Companies engaged in sectors operating with raw materials are more likely to

engage with the initiative.

3.1.4. ISO 14001 certification

The International Organization for Standardization (ISO) is an independent international organisation, a network of national standard bodies from 165 countries, developing voluntary standards of operation across various issue areas (ISO n.d.-a). Among them is the

environment management system standard series ISO 14000. Central to this family of standards is the ISO 14001 that establishes and specifies how an environmental management system should be organised. A standard of environmental management system which is capable of being certified to while other standards in the family work as guidelines (Curkovic et al. 2005). The ISO 14001 certification is applied to business facilities and not companies. However, it is not unusual for companies, even for multinational enterprises, to adopt the management system to all their facilities (Christmann & Taylor 2001:443-445). The standard does not specify environmental performance indicators and is thus applicable to all industries and facilities of all sizes but is expressed as having the design of fulfilling and enhancing environmental performance as well as objectives (ISO n.d.-b). Although the adoption of the ISO series assures that an environmental management system is present, it does however not automatically assure capabilities of substantial environmental performance (Curkovic et al. 2005). Nevertheless, besides a previously mentioned enhanced perception of legitimacy, ISO 14001 facilities have been shown to, relatively to those that are not, have a more desirable environmental performance and recurrence of environmental considerations (Melnyk et al. 2003). Moreover, Arocena et al. (2021) finds that ISO 14001 incorporation not only reduces emissions but also increases financial profits. Several previous studies examine the

(17)

12

relationship between ISO 14001 certification and (further) voluntary environmental practices (Arocena et al. 2021; González-Benito & González-Benito 2008). Although the relationship between ISO 14001 is connected to waste and emission reduction, it has also been found to increase environmental awareness and lead to further voluntary environmental commitments (Boiral et al. 2018; González-Benito & González-Benito 2008). The expected relationship between ISO 14001 certification and engagement in the climate initiative is therefore, in line with previous findings, that companies that are certified are more likely to voluntarily engage with the initiative.

Hypothesis 4: Companies with an ISO 14001 certification are more likely to engage with the

initiative.

In Table 1 below the four variables of interest and the respective hypotheses of their association with the likelihood to engage with the climate initiative Fossilfritt Sverige that will be tested is presented.

Hypothesis 1 The larger the company is, the more likely it is to engage with the climate initiative.

Hypothesis 2 The larger the annual revenue of the company is, the more likely it is to engage with the climate initiative.

Hypothesis 3 Companies engaged in sectors operating with raw materials are more likely to engage with the initiative.

Hypothesis 4 Companies with an ISO 14001 certification are more likely to engage with the initiative.

(18)

13

Control variables

3.1.5. Primary stakeholder

A substantial amount of literature scrutinizes the effect of various stakeholders on voluntary environmental initiatives. However, such literature is mainly considering such elements when examining why companies participate rather than using it as a way of characterising

companies. In such literature, stakeholders and stakeholder pressure has been found to have a substantial effect on voluntary environmental practices beyond compliance to regulatory frameworks (e.g. Henriques & Sadorsky 1999). Although several primary stakeholders can be recognized, with primary internal stakeholders such as employees and primary external stakeholders such as the government (Shubham et al. 2018), one essential group of

stakeholders is the intended customer of the services or products offered. In this regard, one possible distinction is the division of companies as oriented towards either organisations such as companies or directly towards individuals. It has previously been found that companies that have a direct interaction with the end-consumers, meaning those that end up using the product or services, are more likely to engage with climate mitigation practices (Damert & Baumgartner 2018). With this definition of the concept, certain companies are themselves end-consumers of some services and products. However, consumers as individuals have been argued to be increasingly environmentally conscious in their consumption and utilization of services and products (Etzion 2007:647). Swedish consumers have especially been found to politically participate through political and ethical consumption, defined as “consumers’ use of the market as an arena for politics in order to chance institutional or market practices found to be ethically, environmentally, or politically objectionable” (Stolle & Micheletti 2013:39). Furthermore, there have been found to be a relationship between societal pressure, eventual ISO 14001 adoption, and the public image of a company (Arocena et al. 2021). Thus, in line with these findings, whether the company is oriented towards either other companies or directly towards individual consumers will be controlled for.

3.1.6. Internationality

A majority of previous studies have a national scope with attention on companies established in a certain country (e.g. Min & Galle 2001; Jabbour et al. 2016), while a minority

(19)

14

most of the studies is thus to nationally isolate the relationship between company

characteristics and voluntary environmental engagements rather than make cross-country comparisons. Nonetheless, globalization and following company multinational characteristics have been found to influence environmental engagements. Christmann & Taylor (2001) emphasize that activities in more than one country enhances state and non-state

environmental demands and thus self-regulatory voluntary environmental initiatives of companies. Similarly, internationality is linked to an increased number of stakeholders, and thus the potential gain or loss in legitimation dependent on environmental practices of the company (Darnall et al. 2010). Furthermore, company size, and thus availability of resources, is also found to be linked to whether a country is foreign based or has activities in multiple countries (Etayeb & Zailani 2009). While this study examines the relationship between Swedish companies and the voluntary engagement in climate initiative, a number of these companies either originate from foreign countries or alternatively have extended their services and exports of products to further countries. The potential effect of internationality will therefore be controlled for.

3.1.7. National (geographical) scope

In addition to differences in either an international or national scope of company business activity, companies may also geographically differ in their provision of services and products on a national scale. While digitalization has facilitated many businesses to extend their activity across larger geographical areas, others are local or regional in their activities. This is perhaps particularly the case for companies in specific business industries and is

hypothetically also often connected to both company size and revenue. Darnall et al. (2010) control for company market scope using local, regional, national, and global scales. The motive for the inclusion of such a variable is because the scale of business activities is highly related to the number of stakeholders. While companies operating under a local or regional scale generally have less (both internal and external) stakeholders, the approval or

(20)

15

3.2. Company motives for voluntary environmental

engagements

The incentive to voluntarily engage with the climate initiative Fossilfritt Sverige and

contribute to its commitments is expressed, by the initiative itself, as being both virtuous and competitively beneficial. Nevertheless, such incentives originate from the initiative and not its participating companies. A number of studies scrutinize why companies voluntarily choose to engage with environmental practices. Although some distinctions exist in this body of research, a few major themes have been identified. These are distinguished into three categories of motives for voluntary environmental engagements: competitiveness,

legitimation, and social responsibility. Additionally, and closely linked to these motives, the literature puts forward corresponding stakeholders associated with the level of these motives.

3.2.1. Competitiveness, legitimation, and social responsibility

The specific conceptual categorization of motives for companies to engage with voluntary environmental practices into competitiveness, legitimation and social responsibility, derives from the model put forward by Bansal & Roth (2000) but shares its similarities in

conceptualization with other studies (e.g. Berry & Rondinelli 1998; Ervin et al. 2013). In their work, Bansal & Roth (2000) define motives of competitiveness ultimately as motives to enhance financial benefits and resource efficiency. Thus, with an emphasis on competitive advantage and maximizing business profits. Motives of legitimation in turn indicate the company's inclination to comply with governmental policies and regulations, but also to adjust to the society’s norms and values. The emphasis here is expressed as being compliance to minimize costs and risks, but also that such weight on compliance might lead to voluntary environmental engagements ahead of future legislation or expectations. Lastly, social

responsibility originates from the values, and perceived obligations towards society, of the company itself. Thus, the emphasis is on normativity and environmental values (Bansal & Roth 2000:724-728).

In the literature, the extent of these environmental motives is the demands and concerns of various stakeholder groups. For example, Henriques & Sadorsky (1999) finds three

(21)

16

profile, effect on company motives for voluntary environmental engagements. These are regulatory stakeholders, organizational stakeholders, and community stakeholders. Among these stakeholder groups, using Shubham et al. 's (2017) specified classifications are stakeholders such as the government, customers, employees, shareholders, NGOs, and competitors. These stakeholders have been found to substantially affect company motives for voluntary environmental practices of companies in previous literature (e.g. Shubham et al. 2007; Berry & Rondinelli 1998). Not in the least these stakeholders may influence how companies perceive cost-benefit calculations of voluntary environmental engagements (Ervin et al. 2013). Bansal & Roth (2000:726) associate some of these stakeholders and their

pressures with the categorization of motives. Influenced by such associations, and

complemented by corresponding literature, these stakeholders are expected to be linked to the extent of different motives for voluntary environmental engagements. Firstly, motives of competitiveness are expected to be linked to pressures and possible advantages gained from customers, investors, competitors, and shareholders. Secondly, motives of legitimation to complying with government regulations, but also with societal norms pushed by employees, local communities, and NGOs. Lastly, motives of social responsibility to ‘society as a whole’ or explicitly no stakeholder at all. The intention of this study is to examine the extent of these motives being expressed by companies engaged in the initiative, with a following expectation to find linkages between these categories of motives and their associated stakeholders.

4. Method

To investigate the research questions of this study a mixed method deductive approach is carried out. These two separate methods are intended to answer separate research questions, with the quantitative method answering the first and the qualitative answering the second, but also complementary provide with an integrated perspective on the respective research

(22)

17

4.1. Quantitative approach

4.1.1 Binary logistic regression

The four theoretically derived hypotheses presented in the earlier segment of this paper are empirically tested, with the use of the statistical software Stata, through binary logistic regression analyses. A logistic regression analysis is commonly utilized when analysing outcomes that have a binary character where in this study, whether the company is engaged in the climate initiative is the binary dependent variable. In comparison to a linear regression, a logistic regression avoids certain problems of functional form when the dependent variable is binary. Linear regression coefficients with a binary dependent variable are interpreted as the change in probability of one unit change in the independent variable (Pampel 2000:3). Given that we are dealing with probabilities, the interval of values can only have a range of 0-1. There is no such thing as a negative probability and a probability may not be above one, and thus such finds are not meaningful. However, the functional form of a linear regression allows for an indefinite value on the independent variable, giving the possibility of predicted probabilities that are either negative or above the value 1. Furthermore, assuming linearity may not only not fit the data well, it also often does not make theoretical sense (Pampel 2000:6). Rather than a linear function, the logistic function has a s-shaped curve, which is more fitting with a binary dependent variable as it does not assume that the change in probability is constant (ibid.). Logistic regression calculates the effect of the independent variables on the probability of a binary event occurring. It is advantageous as it provides a measure of how appropriate a predictor is, as well as the direction of the association (positive or negative). The logistic regression can be expressed as in equation 1 below.

p̂

=

(1)

(23)

18

straightforward or meaningful. A regular technique is therefore to convert these coefficients into odds ratios (Menard 2010), which will be the approach used to interpret the findings towards the posed hypotheses. Furthermore, the goodness-of-fit of the models will mainly be assessed and compared through the pseudo r-squared and likelihood-ratio tests. Although the pseudo r-squared value is occasionally criticized and expected to be low in logistic models, it is an instructive way of comparing models with each other (Hosmer & Lemeshow 2000:164-167). Additionally, further controls and robustness checks will be employed and elaborated on in the analysis.

4.1.2. Research design

The quantitative approach of this study has a case-control design. The procedure consists of collecting data for observations with a rare outcome and comparing these with other random observations that do not have this outcome. Thus, the data collection is based on the

dependent variable. Such a design is useful in studies using logistic regression analysis with the purpose of studying the causes to a particular binary outcome (Forgues 2012). The design of this study is further based on what King & Zeng (2001) calls an equal shares sampling design. Meaning that the dependent variable will have an equal number of observations coded as 0 as 1, with the value 1 here indicating engagement in the climate initiative. Optimally, the dataset would consist of all the observations with the rare and limited outcome 1, and more proportionally to the population a greater number of observations without the rare outcome (0). Nevertheless, due to time limitations such a design is not feasible for this study, and thus the design is based on an approach that has also been found to be adequate, being a random selection of observations within the samples of different outcomes (Forgues 2012). The selection of observations was made through simple random sampling, meaning that all

companies had an equal probability of being chosen (Dillman et al. 2014). This is the case for both samples of companies with separate outcomes.

4.1.3. Data collection and delimitations

The data collection of this study ultimately generated 400 observations of Swedish companies, creating an original dataset. Given the design of the study, 200 of these

(24)

19

initiative (based on the list of actors displayed on their website), meaning all these companies had an equal probability of ending up in the sample. This randomisation was used to avoid selection bias in the data collection. The remaining 200 observations are in turn Swedish companies not participating in the climate initiative, thus coded with the value 0 in the

dependent variable. These were randomly selected with the help of the website ‘allabolag.se’. The website contains a database, and some overview financial information, of the limited companies (aktiebolag or AB in Swedish) in Sweden. Thus, a delimitation of the study is that the dataset only consists of such limited companies, which fit with most of the companies engaged in the initiative. The choice of solely including this form of companies was in turn made since they are required to submit annual financial reports, which was a significant source in the collection of data. With the help of coding, these 200 non-participating companies were randomly selected based on the first letter of the company name and their index number in the database of companies, with each company having an equal probability of being chosen.

Further exclusions of randomly selected companies were companies that had filed for bankruptcy, those that were a part of a business entity already randomly selected, those that were found to report ambiguous data, and those companies where sufficient data was not found. The guideline used in the data collection was that companies with missing data on more than two variables were excluded to enhance the robustness and validity of the dataset. However, as realized in the later stages of the data collection, potential limitations of the study occurred due to missing data values, which was addressed and will be further

elaborated on in the analysis. Lastly, the intended sample size was in turn determined from the put forward minimum sample size needed to represent the true population, here being the population of Swedish limited companies, in Dillman et al. (2014). Based on calculations made with Cochran’s formula for calculating sample size, the minimum sample size needed for the 95% confidence interval and a 5% margin of error is 384 observations (Dillman et al. 2014:80), given the true population, just over 600 000, of Swedish limited companies in Sweden (SCB n.d.-b).

4.1.4. Material

A novel dataset containing information on 400 Swedish companies was created using

(25)

20

of data, namely companies’ annual reports, sustainability reports, and websites. On a few observations, the database of allabolag was used to complement data not found in the primary choice of sources. This was mainly used to identify the company industry classification, which is used to operationalise and classify the companies to different business-sectors. The guideline used was to as a rule first use company reports, where annual and sustainability reports are often combined, as a primary source of data. Both because it is feasible to find data for all the independent variables in the reports and because annual reports are

dependable for reliable data, not in the least due to legal requirements. Considering annual reports relate to the past year, and due to this study being conducted during the period of annual report submission, there was a lack of complete availability of company reports for the year 2020. Thus, annual reports from the year 2019 were used for the data collection, meaning that the dataset therefore consists of data from 2019. The data for annual net revenue and number of employees was consistently found in the annual reports, meaning that the values of those variables represent the year 2019. Subsequently, when the company report was not sufficient for the collection of data, company websites were utilized as the source of data. The subsequent data collection was in turn guided by the operationalisation of variables.

4.1.5. Operationalisation of variables

In this section the operationalisation of the variables is presented (see Appendix 1 for a table briefly summarising the operationalisation).

Engagement in Fossilfritt Sverige

Engagement in the climate initiative is the binary dependent variable of this study and is thus operationalised as a nominal dummy variable. The register of participating companies on the Fossilfritt Sverige website was used as the verification of engagement in the initiative. The value is coded with 0 if the company is not engaged with the climate initiative. If the company is engaged with the climate initiative the value is coded with 1.

(26)

21 Company size

Company size is in this study measured as the number of employees in the company. This data is collected from company annual reports where it is required to report the average number of full-time employees during the related year. Thus, the company size is measured as the average number of full-time employees in the company in 2019. Although the data collected initially was unaltered, this data is operationalised into an ordinal variable. Similar operationalisation has commonly been used in previous studies (e.g. Collins et al. 2007; Etayeb & Zailani 2009), and the belief is that changes in company size theoretically has a more substantial effect when moving up intervals of company size rather than individual employees. This operationalisation further the advantage of avoiding any issues of outliers in the data. The classification of categories is based upon the company size classification of SCB but is slightly adjusted to fit the distribution of the data. More specifically, the

adjustment involves expanding the range of value 9 and using value 10 as an end point. The following classification of values is used:

1 = 1-4 employees 2 = 5-9 employees 3 = 10-19 employees 4 = 20-49 employees 5 = 50-99 employees 6 = 100-199 employees 7 = 200-499 employees 8 = 500-999 employees 9 = 1000-1999 employees 10 = 2000 + employees Revenue

(27)

22

is based on revenue classifications of SCB but slightly adjusted, resulting in the following classifications: 1 = 1 - 9 999 tkr 2 = 10 000 - 99 999 tkr 3 = 100 000 - 299 000 tkr 4 = 300 000 - 999 999 tkr 5 = 1 000 000 - 4 999 999 tkr 6 = 5 000 000 + tkr Business-sector

The data collection of company business-sectors initially derived from collecting data of the company industry classifications. As previously stated, companies in Sweden are required to classify their business activities according to SNI standards. The 21 main categories of these standards, classified alphabetically as A-U are used to identify which industry the company is mainly operating in. While some companies have registered business activities in more than one industry, the guideline used is that the industry with the most registered classifications is their main industry. At the occurrence of an equal amount, a more in-depth analysis of their business activities is used to determine their main industry of operation. Further

categorization of companies into business-sectors is then formulated, between companies extracting and manufacturing with the use of raw materials and a service sector. The operationalisation is based on the categorization of the economy into four sectors, the primary, secondary, tertiary, and quaternary sector. Here the primary sector simplified signifies the extraction of natural resources or raw material. The secondary sector represents manufacturing and production of goods with the use of raw material. The main feature of the tertiary sector, also called the service sector, is in turn the production of services rather than the production of goods, and the quaternary sector specifies the output of information

services (Kenessey 1987; Fisher 1939). These sectors were operationalized into four dummy variables indicating whether the company is operating in the sector:

(28)

23

Given some difficulties in determining especially whether a company is a part of the tertiary or the quaternary sector, and particularly because of the interest to investigate the relationship between companies operating with raw materials and engagement in the climate initiative, these categories are further classified into two dummy variables. Consequently, even though the related hypothesis 3 is associated with the occurrences of activities in sectors operating with raw materials, a variable testing the effect of companies operating in the service sector is also included.

Raw material

0 = The company does not operate in sectors extracting or manufacturing with raw

materials.

1 = The company does operate in these sectors

Service

0 = The company is not operating in the service sector 1 = The company is operating in the service sector

ISO 14001 certification

The rationale for measuring ISO 14001 certification rather than for example the EU

counterpart, the European Union Eco-Management and Audit Scheme (EMAS), is based on a few considerations. First, the adoption of the ISO certification is considerably more

(29)

24 0 = Not certified

1 = ISO 14001 certified

Primary stakeholder

This variable is operationalised as a nominal dummy variable. Where the value is coded with 0 if the company does not offer products or services directly to individual consumers, and 1 if they are found to do so. If the company offers products or services to both other businesses or organizations and individual consumers, or if such a specification is not evidently clear, the value is coded as 1.

0 = The company is oriented towards other organizations

1 = The company is oriented towards individual consumers (or both)

Internationality

This variable is operationalised as a nominal dummy variable. Where the value is coded as 0 if no part of the company organisation is found to offer products or services in countries outside Sweden, and 1 if they are found to do so.

0 = No businesses activity internationally 1 = Business activity internationally

Geographical (national) scope

This variable is operationalised as a nominal dummy variable. If the company specify that they are offering products or services locally, here identified as in a single city or county, the value is coded as 0. If they do the opposite, offer services or products digitally, or make no such specification, the value is coded as 1.

0 = The company is active locally

(30)

25

4.2. Qualitative approach

4.2.1. Material

The qualitative component and method used to address the second research question of this study is based on 15 interviews with representatives of companies engaged in the climate initiative Fossilfritt Sverige. The interviews were carried out within the project Action for Climate Transformation in Sweden (ACTS) and the material was accessed through my thesis supervisor who herself carried out a few of the interviews. The ACTS is a collaborative research project set out ‘...to understand how the interplay between state and non-state climate action unfolds in the post-Paris climate policy landscape.’ (ACTS n.d.). The consideration that the interviews are not conducted by the author of this study entails a limitation in composing the interviews to fit the purpose of the study. However, components in the interviews were found suitable to address the research question of scrutinizing

expressed motives of companies to engage with the initiative. Bryman (2016:594) further acknowledges that the analysis of interview transcripts collected by someone else does not necessarily have to be a limitation, but rather that such usages may have ethical concerns. An emphasis has therefore been put on the anonymity of the companies.

(31)

26

emphasis on participation in climate initiatives. Those questions that are of main interest for this study may be summarized as questions about:

● The importance of the climate issue

● The motives or circumstances that led them to join the initiative

● The reasoning for participating, or not participating in another climate initiative ● The effects, advantages, and disadvantages of participating in the initiative

Given that the material exclusively consists of 15 interview transcripts, it does not represent a generalisable view of the motives for engaging with the climate initiative but should rather be taken into consideration as case-specific expressed motives for 15 companies in a diverse and larger group. The company representatives interviewed were all sustainability chiefs or CEOs and therefore provide insights into why they decided to join the initiative. This material is examined with the use of a thematic analysis (Bryman 2016:586), using a conceptual framework of motives and linkages to stakeholders, to identify and interpret the circumstances of different categories of motives.

4.2.2. Thematic analysis

The thematic analysis of this study takes a deductive approach, meaning the analysis is driven by a few predetermined theory-driven themes. To reiterate, these are the themes of

competitiveness, legitimization and social responsibility and their corresponding expected linkages to key stakeholders. Given that the transcribed interviews were relatively few and short, in general 5-10 pages long, a keyword search method was not considered to be necessary, and the transcripts were instead assessed in full. The coding and subsequent analysis is guided by the categories of motives together with the associated stakeholders. Table 2 designates the themes, or categories, and their expected linkages to key stakeholders of interest. Furthermore, examples of indicators for these categories are specified.

Themes Key stakeholders Example of indicator

Competitiveness Customers, competitors,

investors, shareholders

(32)

27

Legitimation Government, employees,

NGOs, local communities

Legislation, compliance, credibility

Social responsibility ‘Society’, ‘none’ Responsibility, normative

values Table 2. Concepts guiding the analysis

Stakeholders have occasionally in previous studies been categorized as a separate or fourth category of motives (Berry & Rondinelli 1998), but the analysis is here guided by examining to what extent they are expressed to bring about and giving rise to these three distinct

categories of motives. The categorization of motives and links between stakeholder groups and these motives is mainly guided by the model of Bansal & Roth (2002), the findings of Ervin et al. (2013), and by the classifications of Shubham et al. (2018). Nevertheless, these classifications of motives are not considered to be exclusive to each other. The occurrence of an expressed competitive motive for engaging with the initiative does not exclude nor

disprove an expressed motive of social responsibility. Furthermore, while linkages between the themes and stakeholders are expected they are not seen as self-evident but are rather, to what extent they are linked, used as a general theoretical basis to be reflected upon. The examples of indicators in Table 2 demonstrate elements that are deemed to be signifying for the separate categories of motives. Thus, for example, weight on cost-benefit considerations suggests competitive motives, emphasis on legislation and compliance in turn denotes motives of legitimation, while stressing responsibility and making normative considerations underlines motives of social responsibility.

Thematic analysis is sometimes named and used interchangeably as a content analysis,

(33)

28

questions of the study, the qualitative analysis is to a small extent guided by the quantitative approach. Given the depiction in previous literature of companies as actors taking, in addition to their company features, the demands of societal actors into consideration in their business activities (e.g. Ervin et al. 2013), the predetermined conceptual framework was regarded as significant. In turn, a limitation in taking a deductive approach is, given the predetermined themes guiding the coding and subsequent analysis, that fruitful insights might not be captured from the material. Therefore, while a deductive approach is put forward as

providing a more detailed examination of the chosen aspects, representation of the material overall is less comprehensive (Vaismoradi et al. 2013:401). Furthermore, a deductive

approach can increase possibility of confirmation bias, interpreting and fitting the material to demonstrate the themes. Thus, particular emphasis in the analysis is placed on the

conversational context and interpretation of the underlying intention of the expressed motives.

5. Analysis

This section will present the result of the analyses in two segments. Initially, the result of the quantitative analysis is presented. Secondly, the findings of the qualitative analysis is put forward.

5.1. Quantitative analysis

As previously presented, the data collection resulted in 400 observations. Table 3 illustrates the descriptive character of the data.

Variable Obs Mean Std. Dev. Min Max

(34)

29

international 398 .563 .497 0 1

stakeholder 400 .505 .501 0 1

nationalscope 400 .838 .369 0 1

Table 3. Descriptive statistics

In Table 3 a few elementary summary statistics of the variables and their values are

presented. The mean of the dependent variable engagement is 0.5, indicating that 50% of the observations are engaged with the climate initiative as per the design of this study. Similarly, the mean of the ISO14001 certification variable is 0.618, designating that approximately 62% of the 338 observations have adopted the certification. Obs here indicates the number of observations collected for each variable. For instance, the variable iso, which measures whether the company is ISO14001 certified, limited only has 338 observations.

Thus, for some of the companies included in the analysis data is lacking in certain variables. The restriction causes the models to ignore observations with missing data on one or more of the variables. This is a potential limitation of the study as it reduces the number of

observations, not fulfilling the set-out aspiration of 384 observations and increases the risk of estimation bias. As displayed in Table 3, the variable of iso has missing data namely in 62 observations. To solve the issues with missing data each missing data in the variable iso, with a binary operationalisation, is imputed with the value 0. Only the missing data of the variable signifying the presence of an ISO 14001 certification is imputed as it was regarded as

sufficient to increase the number of observations. The imputation assignment is guided by the previously stated relationships between companies and ISO 14001 adoption. For example, the advantageous environmental performance, but not in the least the pressures for- and gains in public image obtained from stakeholders at adoption. It was assessed that if companies were to be ISO 14001 certified they would likely clearly convey such a company quality given it found benefits. Which was the premise, together with the fact that it was relatively

unchallenging to find evidence of certifications and commitments for most of the remaining companies, used to substantiate an imputation of the value 0 to all the missing data for the variable in the dataset.

5.1.2. Preliminary tests

(35)

30

in the relationship of interest. However, in the preceding correlation analysis the variables sizeordinal and revenueordinal, which measured a company’s size and revenue respectively, proved to be highly correlated. A correlation matrix displaying the correlation between all independent variables is presented in Table A in Appendix 2. Consequently, in the correlation matrix the correlation coefficient between sizeordinal and revenueordinal, 0.820, indicates a strong positive relationship between the variables, meaning that they parallelly change in direction together. This strong correlation may be problematic as multicollinearity could risk biasing the estimates. Supplementary, a test of the fit of the model was made using a

likelihood-ratio test. The likelihood-ratio test compares the goodness of fit of two models with a different amount, but close to the same, independent variables and evaluates which one better fits the data (Hosmer & Lemeshow 2000). The null hypothesis tested with the

likelihood-ratio test is that the restricted model, the model with less variables, better fits the data. Thus, a significant result denotes that we can reject the null hypothesis and conclude that the model with more predictors has a better fit. Given that the probability value (0.1597) in Figure A (see Appendix 2) does not indicate statistical significance we cannot reject the null hypothesis. The model with solely revenue therefore has a better fit compared to a model which includes both predictors. The strong correlation between the variables together with the results from the likelihood-ratio test suggests that the two variables should not be a part of the same model. Instead, two separate binary logistic regressions models are estimated

(36)

31 5.1.3. Results

Table 4 below presents the results from the logistic regression analyses estimating the effect of the independent variables on engagement in the initiative. Column (1) presents the results of model 1 including the variable of company size, while column (2) displays the model 2 containing the variable of revenue.

(1) (2) Engagement Engagement Company size .951 (.044) Annual revenue .782*** (.059) ISO certification 3.161*** 3.211*** (.772) (.765) Raw material .9 1.015 (.237) (.258) Service 3.243 2.521 (2.415) (1.709) Internationality .602** .686 (.153) (.171) Stakeholder .893 1.016 (.2) (.224) National scope 2.191** 1.906** (.735) (.617) Constant .162** .384 (.131) (.287) Observations 382 397 Pseudo R2 _.067 _.068 Standard errors are in parentheses

*** p< .01, ** p< .05, * p< .1

Table 4. Main logistic regression models

(37)

32

initiative. Important to emphasize is that these values do not indicate causation but rather correlation, meaning that a value above 1 indicates that a unit increase in the independent variable increases the likelihood of engaging in the initiative but not that it necessarily causes engagement. The estimated coefficient of company size in model 1 has an odds ratio of 0.951, signifying that the larger the company is, in terms of number of employees, the less likely it is to engage with the climate initiative, which is opposite to the hypothesised relationship between the variables, although indicating a weak effect. However, in the same way as with the variables indicating business-sector, and the stakeholder control variable, the relationship is not found to be statistically significant, meaning a null hypothesis of no

relationship cannot be rejected (see Table B and C in Appendix 2 for more detailed indicators for statistical significance).

The output in model 1 displays that the odds ratio for the ISO certification variable is

(38)

33

Hypothesis Null

hypothesis

Support for hypothesis H1 The larger the company is, the more likely it is to

engage with the climate initiative.

Accepted Not

supported H2 The larger the annual revenue of the company is, the

more likely it is to engage with the climate initiative.

Rejected*** Not supported H3 Companies engaged in sectors operating with raw

materials are more likely to engage with the initiative.

Accepted Not

supported H4 Companies with an ISO 14001 certification are more

likely to engage with the initiative.

Rejected*** Supported

Table 5. Result of hypothesis testing

In Table 5 hypothesis 1-4 are reiterated. The rejection of a null hypothesis of there being no relationship, and therefore acceptance of an alternative hypothesis of there being a

relationship between the variables, is then summarised. Finally, whether the findings corroborate the hypothesis is stated. As displayed, no statistical significance is found to support hypothesis 1 and 3, meaning no significant relationship is found between the variables of company size as well as business-sector and engagement in Fossilfritt Sverige. Furthermore, while the relationship between company annual net revenue and engagement in the climate initiative is found to be statistically significant, the nature of the relationship is found to be in the opposite direction and thus not supporting the expectations establishing hypothesis 2. Lastly, hypothesis 4 of there being a positive association between companies being ISO 14001 certified and engagement in the initiative is corroborated by the result.

5.1.4. Robustness of the results

(39)

34

it made better theoretical sense than to attribute values to the variable based on the

distribution of values in the observed data. To control for the effect of the imputation method, an alternative way of solving issues with missing data, multiple imputation, was conducted. The basic idea of the multiple imputation procedure is to estimate likely values for the

missing data based on the distribution of values for the non-missing data with the inclusion of random elements to represent variability (White et al. 2011:377). Thus, models using this method to estimate values of the missing data for the ISO certification variable were carried out using this mechanism (see Table D and E in Appendix 2 for the models using this

procedure). The results from these models are similar in terms of direction and significance to those in the baseline models presented in Table 4 above. Furthermore, Table F in Appendix 2 displays the results from a logistic regression including the same variables as in the main model but without any imputation of the iso variable. These regressions therefore have a lower number of observations, however the significant results in the baseline model remain. Finally, omitted variable bias due to running separate models is in this case unlikely as the condition of the omitted variable being correlated with the dependent variable is not met. However, Table G in Appendix 2 includes both the correlated variables signifying size and revenue together to see how the variables behave. In the model the significant effects remain, but while size remains insignificant it changes direction, becoming positive, which is a common effect from multicollinearity. Moreover, although there is no clear academic

consensus of the potential effect of heteroscedasticity on logistic regressions, the variables of iso and revenue furthermore remain at their level of significance in models with the inclusion of robust standard errors (see Table H in Appendix 2), while the control variables slightly reduce in level of significance.

The significant relationship between the iso variable and the dependent variable thus does not change across models but rather persists through all model specifications. Furthermore, the imputation method does not seem to skew the relationships, but rather gaining in the number of observations seems to strengthen the significance of other relationships. This is for

(40)

35

The pseudo-r squared values and likelihood-ratio tests have in turn mainly been used in this study to compare competing models and assisted in assessing which model has the most advantageous model fit, rather than being finalized measures of fit. The choice of not using measures such as AIC and BIC for model comparisons is partially derived from practical difficulties of obtaining the criterions for all models, but also that such indicators are highly influenced by sample size. Nevertheless, although the increase in pseudo r-squared value in the main models relative to other models appears to be marginal, the pseudo r-squared displayed in the models is specifically McFadden's pseudo r-square with the interpretation of values indicating a good fit differing from the r-squared values of linear models. McFadden’s pseudo r-square can essentially be said to measure how well the model predicts the outcome compared to a model without independent variables. McFadden (1977) states that values between 0.2 to 0.4 indicate a very good model fit. Therefore, while the pseudo r-squared values of model 1 and 2 (0.0.67 & 0.068) indicate a better fit relative to other specified models of this study using similar operationalisation, the models indicate that a more

appropriate model exists. Moreover, to further explore the unexpected insignificant result of business-sector, an additional robustness check was to assess the relationship between business-sectors and engagement in the initiative through using an operationalisation of specific industry classifications rather than classifying them into two categories as guided by hypothesis 3 (see Table I in Appendix 2). These models provide some support to there being a relationship between companies involved in specific business activities and engagement in the initiative, and an improved model fit, which will further be elaborated on in the

(41)

36

5.2. Qualitative analysis

As previously defined, the thematic analysis assessed the expressed motives for participating in Fossilfritt Sverige of interviews with 15 separate company representatives. Specifically, the analysis was guided by the three established categories of motives and their expected linkages to certain stakeholder groups. The findings of this analysis will thus be presented in three separate segments, demonstrating what was found to be indicative for, and

distinguishing within, the categories of motives. I will in this segment refer to the company representatives, the interviewees, as company 1, 2 and so forth. Although the interviewees are not in every case a part of the top management of the company, jointly for the representatives are their positions as parts of the company’s sustainability management. Thus, the viewpoints are assessed to portray the company’s environmental viewpoint. Furthermore, while some of the expressed motives are regarding the climate issue and climate mitigation in commitments in general, the overall subject of the interviews is about participation in initiatives. Thus, some of the motives are implicitly linked to motives for engagement in the initiative.

5.2.1. Competitiveness

Motives related to competitiveness are found to be the most recurring expressed motive for engagement in Fossilfritt Sverige, with all but two companies expressing some sort of indication of either the competitive cost-benefit or a competitive pressure for engagement. These motives are thus characterized by a rationale of enhancing economic opportunities. The motives of competitiveness are further found to have a close linkage to the expected group of stakeholders, such as competitors, investors, shareholders, and customers. In particular, customer demands are distinguished as a primary source of motivation.

“...I feel that there is a greater demand from our customers as well, and if we can't offer the

solutions that are okay then someone else will take that part. So, I feel that we have sharper customer inquiries...” (Interview with company 7).

Department of Political Science