Does exposure to severe problems lead to more criticism?

(1)

Does exposure to severe problems lead to more

criticism?

Eva Sander

10666818

29/06/2015

Master Thesis

MSc in Business Administration- Strategy Track

Supervisor: Daniel Waeger

(2)

Statement of Originality

This document is written by Student Eva Sander who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document are original and that no

sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of

(3)

Abstract

CSR has become an increasingly important topic during the last 20 years. Due to the changes that came with globalization firms nowadays have to deal with problems that were previously dealt with by governments. Most firms have come to accept these new responsibilities; some, however, do not comply. These firms often face criticism from NGOs and the general public. This study analyzes the effect that being exposed to severe social problems has on receiving criticism. The results show that firms that are more exposed to social problems do indeed receive more criticism than other firms. Moreover, a company‟s good CSR performance may not seem to weaken this effect but strengthen it further. Operating in countries with high press freedom appears to strengthen the effect being exposed to social problems has on receiving criticism for a firm. The study also found that it does not matter whether a firm‟s customers are other firms or end-consumers: if it is exposed to problems, the firm will be criticized just the same. These findings help bridge the gap between the extensive CSR and globalization literature and add to the developing understanding of the codes to which firms should adhere in a global marketplace.

Keywords: CSR, Globalization, NGO, Criticism, Press Freedom, B2C, B2B, Human Development Index, Social Problems

(4)

1. Introduction

Corporate social responsibility (CSR) has become increasingly important in the last 20 years (Scherer & Palazzo, 2007). A reason for this is that companies have realised the importance of a good CSR policy and how severe the consequences can be of not having one. While it was enough in the past to rely on the home country‟s legislation and act according to it, this is very difficult for companies nowadays. With globalization great deals of companies are no longer active in only one country or market. Since each of these countries comes with a different set of legislations that might vary in its strictness, a company has to decide for itself what actions it deems acceptable or not. Moreover, technology develops at such a fast pace that legislation generally cannot keep up (Sandel, 2011). From companies‟ perspectives, a good reputation is now more important than in the past. Cases such as the boycotts on firms like Nike and Shell illustrate the importance of behaving in what is considered to be a

morally acceptable manner. A firm‟s CSR policy nowadays has to be the same across all its geographical operations. Due to technology that was not in place ten or twenty years ago, the firm‟s stakeholders are now more connected. Therefore it is no longer feasible for a firm to only act ethically with regard to its customers: other stakeholders must be considered. Non-governmental organisations keep track of a firm‟s behaviour towards its workers in poorer countries, the environment in producing countries and its general behaviour at home and abroad. The demands that stakeholders have for a firm are rising. It is no longer acceptable to only care about the direct business. Nowadays society demands firms to commit to solving problems even if they are not directly related to their area of business (Matten & Crane, 2005). As Schrempf (2014) notes, companies are socially connected to their environment and stakeholders through their operations and products. However, this also means that firms are exposed to problems they have not encountered before. Due to globalization firms could be

(6)

2 exposed to problems that the respective government is not able to resolve (Palazzo &

Scherer, 2009). Since the mid-1990s firms have come to accepted partial responsibility in taking care of these problems. Consumers and governments have also adjusted their

expectations and will criticize firms that refuse to help (Schrempf, 2014; Maistriau, 2014). In order to achieve the best outcome for everybody the firms that are exposed to the worst problems should get criticized the most. This leads to the following research question:

Does exposure to severe social problems lead to more criticism for a firm? This research is based on 541 companies covered in the Covalence EthicalQuote index, whose analysts strictly monitor the criticism received by these companies. Since the sample consists of large MNEs it is an appropriate research sample because the population of interest is companies which operate across vastly different geographical markets.

This study consists of several chapters. The first chapter will discuss the existing literature on all aspects connected to this topic to gain insight that is complete and recent. Afterwards, the theoretical framework and the hypotheses are introduced. In the following chapter data collection and the way it was analyzed will be discussed. Subsequently the statistical analysis and the results will be presented. The final chapter contains the conclusion, the limitations, and implications for management and suggestions for further research.

(7)

3

2. Literature Review

2.1 The Traditional View of Corporate Responsibility

Even though scholars do not agree on the exact purpose of a firm, many of their explanations revolve around the notions of maximising shareholder value or being efficient while maximizing shareholder value as their top priority (Coase, 1937; Porter, 1990; Sundaram & Inkpen, 2004). The issue of tackling social problems is not seen as a responsibility that should be borne by a firm (Henderson, 2001).

In the traditional view of most scholars, the role of dealing with social and

environmental problems is assumed by the relevant governments. Palazzo and Scherer (2011) mention the Westphalian order as an example for this traditional view. The Westphalian order refers to the order of states as organized in the treaty of Westphalia and it “rests mainly on the steering capacity of the state authorities of sovereign countries with both a monopoly on the use of force on their territory and more or less homogeneous cultures that lead to a

stabilization of social roles and expectations within coherent communities.” (Palazzo & Scherer, 2011). They further point out that with the state being a dominant regulator the firms take dependent roles.

Sundaram and Inkpen (2004) argue furthermore that when a firm is faced with social or environmental problems it should continue to pursue its most important goal of

maximizing shareholder value and let the state deal with these problems, because states are better equipped to deal with social and environmental issues by creating and applying laws, enacting regulations and enforcing contracts.

Moreover, according to Friedman (1962) the state is responsible for the well-being of its citizens. It is therefore up to the government to promote a healthy society by creating rules and laws to which citizens, as well as firms, must adhere. He further argues that the firm‟s

(8)

4 only responsibility is to adhere to these laws. Levitt (1958) further argues that it is the state‟s responsibility to predict problems and conflicts and tackle them in order to ensure well-being for its citizens. Jensen (2001) concurs, stating that a firm‟s highest goal should be value maximization while the state‟s highest goal should be to ensure for the well- being of its citizens. If this is endangered in any way it is up to the state to react by changing the legislation and ensure that all parties involved adhere to it.

2.2 Globalization

Globalization is one of the main reasons why the traditional view just described does not hold up anymore. Palazzo and Scherer (2009) define globalization as “the process of intensification of cross-area and cross-border social relations between actors from very distant locations, and of growing transnational independence of economic and social activities.” The Westphalian order became outdated and was replaced by a “post-national constellation” (Palazzo & Scherer, 2011). This means that the ability of states to regulate business activities is declining. The state is still the lawmaker in its own territory; however, due to businesses‟ new freedoms they can choose under which legal system and set of social, environmental and labor regulations they want to operate (Scherer & Palazzo, 2007). It should be noted that some governments, particularly those that possess comparatively less wealth, might not have as elaborate laws and regulations or the ability to enforce them as most developed countries. New situations can emerge where a firm has more financial power and more abilities to enforce economic and social changes than the respective government under which it operates.

Another point is that globalization promotes an ongoing deterritorialization of economic, social and political interactions (Scholte, 2005). This deterritorialization creates new transnational risks such as global warming, spread of diseases like the bird flu and social problems like drugs and organized crime (Palazzo & Scherer, 2009). It is impossible for a

(9)

5 government in a developed country to prevent these spreading in its own country even when it has laws and regulations in place to prevent the creation of such problems within its own borders. The situation is even worse for developing countries‟ governments that are exposed to these problems due to globalization but lack the ability to prevent them within their own borders (Scholte, 2005).

2.3 Firms’ New Responsibilities

With this new situation, where governments can no longer single-handedly tackle social and environmental problems, the need thus emerged to find another entity to fulfil this role. Wolf (2008) argues that these new challenges exceed the power of states, therefore non-state actors need to apply their problem-solving potential in order to get the most efficient solution. Palazzo and Scherer (2011) point out that the distinction between economic and political activities are blurring due to pressure from NGOs. They further argue that many firms have adapted to this new concept of responsibility and changed their approach to CSR. The change from liability CSR to social connection CSR illustrates the changes that firms encountered in getting new responsibilities assigned. Previously, firms practiced liability CSR which entails that the companies only felt responsible for issues that directly resulted from their actions or, for example, a malfunctioning of their product (Schrempf, 2014). This approach has been replaced by the social connection CSR. According to Schrempf (2014) this happened due to five main reasons.

Firstly, assigning responsibility should go beyond the causal relationship between what a firm does and the outcome these actions have. The author proposes that it is not enough to only assign responsibility to a problem when it is a caused by a direct action of the firm but rather if the problem can be connected to the firm and its business in any way.

(10)

6 Along the same line of reasoning, Schrempf (2014) gives a second reason why

liability CSR is outdated: the isolation of who gets assigned responsibility and who does not. Under the liability CSR model, one actor bore the full responsibility for a problem to which it was assigned and then had to find a suitable solution. The counter argument, and reason why this is no longer true, is that it is very difficult nowadays to clearly assign responsibility to only one actor. Problems today are too complex and transnational to be solved by only one actor, dealing with them requires the cooperation of different parties from different sides, for example the government, companies, suppliers, consumers and NGOs (Schrempf, 2014; Palazzo & Scherer, 2007). According to the social connection CSR model, if one actor is more powerful, it should also get assigned more responsibility since it will better be able to handle the problems associated therewith (Schrempf, 2014; Palazzo & Scherer, 2011; Matten & Crane, 2005).

Schrempf (2014) further argues that the social connection model should be preferred because it is, unlike liability CSR, not retrospective. In the liability CSR model blame will be assigned once a problem happened. With the new type of CSR an approach is sought that proactively prevents bad things from happening. This can be achieved by having more actors involved that can question the firm and its product.

The fourth reason why the liability CSR is outdated is because it assumes that all circumstantial factors are static (Schrempf, 2014). In modern times this is no longer the case, with a firm being exposed to different legal systems and moral standards. With the liability CSR model it would be nearly impossible, when taking into account the change of these circumstantial factors, to assign blame to an actor and hold that actor responsible for a problem.

(11)

7 The final argument made by Schrempf (2014) in support of social connection CSR is that community is used as a reference point, quoting Jones (1991) who argues that “an

irresponsible act is either illegal or morally inacceptable to a larger community”. This means that as long as a firm adheres to the law and to what most people in its community think is moral it has no further responsibilities. However, laws and morals differ between countries. Despite that, the social connection CSR still does not approve a company applying different moral standards in different countries but asks for a universally accepted CSR, usually meaning that the strictest moral rules of one community are applied to every part of the business (Matten & Crane, 2005).

The transition in favor from the liability CSR to the social connection CSR model illustrates how the expectations that companies face nowadays have changed. It shows how companies get assigned more responsibilities regarding problems to which they are

connected. This is due not only to them causing the problems but, also, merely being exposed to the problems and the companies‟ ability in helping to solve them.

2.4 Firm’s Resistance

As argued above, firms are faced with new responsibilities that were none of their concern before. While some companies seem to accept these, a great deal of firms does not accept these responsibilities voluntarily. They only participate in solving them if they are pressured by the public (Palazzo & Scherer, 2011; Palazzo & Scherer, 2007; Maistriau, 2014).

2.4.1 Different ethical standards

One of the reasons for this is that, as mentioned before, there are different ethical standards and legal systems from which a firm can choose. A company can pick its own preferred legal system for labour, social and environmental regulations without breaking a law (Scherer & Palazzo, 2011). According to Shamir (2004) “MNCs are in a position to

(12)

8 effectively escape local jurisdiction by playing one legal system against the other, by taking advantage of local systems ill-adapted for effective corporate regulation, and by moving production sites and steering financial investments to places where local laws are most hospitable to them.” Adding to that, some governments try to attract companies to their territory by offering, for example, a tax holiday. Firms see these opportunities and decide to exploit them. Especially if the host market is geographically and culturally distant from the home market a company might adopt different CSR for each country. A case that supports this argument is the behaviour of McDonald‟s after the release of Morgan Spurlock‟s movie “Super Size Me”. After McDonald‟s received bad publicity for targeting their unhealthy products to children they stopped these advertising campaigns in Europe and the US but continued to do so in China (Schrempf, 2014), a market unexposed to the film. Legally, the company cannot be blamed: morally, this behaviour was accepted in one market but not the other. Situations like this can be tempting for firms, especially since taking care of and eliminating social and environmental problems is usually connected to spending a great amount of money (Maistriau, 2014). The fact that some companies still exploit these differences by adapting different standards to their own benefit illustrates that they have accepted their new responsibilities to take care of problems, but will only do so if they are publicly pressured into it.

2.4.2 Downsides of CSR

Another reason why a company might be hesitant to take its role in solving problems voluntarily is that there are downsides of CSR for the firm in behaving responsibly. The biggest downside of CSR is the costs connected to it. Behaving in socially responsible manner almost always means having to spend more money (Vogel, 2005). Examples are having to pay workers more money, invest in better factories to decrease pollution or materials and investing more money in quality control measures. Sometimes it even means

(13)

9 not being able to advertise to a whole group of potential consumers as the McDonald‟s

example illustrates. This has a negative effect on the firms‟ total profit. As Crane et al. (2014) suggest, investing more money in CSR activities and behaving socially responsibly is not always honored by the consumers in such a way that more sales could be made or an increase in price to justify the costs could be made. Especially firms with powerful stockholders might be more concerned about their profitability than engaging in CSR and will only change their mind if their profitability is endangered due to public pressure such as boycotts (King, 2008; Carroll & Shabana, 2010).

Along with the monetary disadvantage of CSR there can also be organizational problems. Implementing a good CSR policy that helps to eliminate problems can be very time-consuming for firms. It could lead to the necessity to track down and control the supplier network, which can be hard and very time consuming. If no agreement can be reached with the suppliers or if they turn out to be causing the problems it could even lead to the need to restructure the company‟s supplier network (Swartz, 2010). Furthermore, in some cases a company could be forced to rethink its marketing strategy (Schrempf, 2014) or even make changes to its product (Maistriau, 2014). Conducting these changes takes a great deal of time and commitment for executives that they would rather spend on aspects of the business that they see as more profitable. For managers, bonuses are often tied to the firm‟s

profitability but seldomly to its CSR achievements. This gives a greater incentive for them to concentrate their time on other parts of the business (Agrawal & Mandelker, 1987; Crocker & Slemrod, 2007).

Combined, the monetary and organizational problems with CSR together with the cultural and legal differences lead to a situation where firms are expected to behave in a responsible manner but where they are hesitant to do so, since they still adhere to the laws and can make a larger profit with lower levels of CSR activity.

(14)

10

2.5 Pressure on Firms

The situation described above changes if pressure is put on a firm. As Castaldo et al. (2009) argue, the only reason firms behave in a socially responsible manner is because they fear penalties if they do not behave according to expectations. This raises the important question of who is putting pressure on firms, and how this is achieved.

With respect to who puts pressure on a company, it is primarily its stakeholders that pressure the organization. Stakeholders can come from many different parts of the firm, but the most important stakeholders of a firm are its suppliers, customers, shareholders, the government and NGOs (Doh & Guay, 2006). It has been explained earlier how governments can pressure firms into behaving responsibly. For suppliers it depends on how important they are to the firm. This can be assessed by considering whether they supply something that falls into Barney‟s (1991) VRIN criteria. This means something that it valuable, rare, inimitable or non-substitutable for the firm. The more VRIN criteria a supplier covers, the more pressure it can put on a firm (Barney, 1991). Suppliers can put pressure on another company by starting to sell to a competitor, raising its prices or refusing to supply to it at all (Provan &

Gassenheimer, 2007).

Customers occupy the other end of a firm‟s activities. There are two different types of customers a firm can have: other firms or end-consumers. The pressure they can put on a firm to behave responsibly again depends on power. We can reverse the VRIN scenario from the supplier example. The more of these criteria a firm meets in providing its product to its customers, the more bargain power it has and the less likely it is that the firm will be criticized by these customers (Barney, 1991). It should be noted that end-consumers, especially, are the target of NGOs who try to convince them into boycotting a company (King, 2007).

(15)

11 The shareholders of a firm are its owners. If they disagree with what the firm does it is easy for the shareholders to put pressure on the firm or, more accurately, its management. The shareholders vote for the management board and it can therefore be expected that if the majority of shareholders disagree with the board‟s actions they will replace it (Karpoff, 2001).

The groups mentioned so far were all primary stakeholders, meaning the company is dependent on them for survival. NGOs belong to the group of secondary stakeholders, meaning that the firm is not dependent on them for survival (King, 2007). This fact makes it harder for NGOs to put pressure on the firm successfully. However, if they manage to become successful they also tend to criticize a firm more strongly since the survival of the firm is not their primary concern (Schrempf, 2014). The first achievement that an NGO has to make is to get collective action. King (2007) argues that collective action “consists of

coordinated behavior of two or more people that, at least in some minimal way, satisfies individual goals and produces a joint experienced outcome”. McAdam et. al. (1996) add that the more people can be mobilized, the greater the outcome. In order to be able to mobilize more people and be able to put more pressure on a company an NGO needs to establish a cause that is of concern to the end-customers. It is important for the NGO to phrase its criticism in such a way that people will be concerned and care for it. Many NGOs use highly vocal tactics to achieve this, meaning that they do not sit down quietly but seek public attention first and then make their demands (Swartz, 2010). Being only reprimanded by an organization will not have the same effect as being reprimanded by an organization that is backed up by the firm‟s customers that would be willing to institute a boycott (Schrempf, 2014). In order to achieve that the NGO needs to communicate the criticism it has for a firm to the broader public. To achieve this, having access to the media is important. This can either be achieved if the NGO itself is well-known with a large number of members or if it

(16)

12 uses tactics that raise public attention. In addition to that, the company that will be targeted is of great importance. It is easier to find public support if a company is widely-known. This effect is compounded if the company, or the industry it is in, has received criticism before, whether from the public or the government (King, 2008). In addition, an NGO will most likely only pick one company in the supply chain and will force it to take responsibility for a particular issue. This is most likely also the company that has the most bargaining power and is financially capable of making changes (King, 2008). Moreover, it is easier to target a company on which public pressure in the form of boycotts, for example, has a direct effect. This means a firm that sells directly to customers will be more likely to be picked as a target (King, 2008).

Above it has been explained how companies started to adopt responsibilities that are not directly related to their business because they have the power -- sometimes even more power -- than individual governments, to solve the problems to which they are exposed. However, not all of these firms adopt these responsibilities voluntarily. They have to be put under pressure by the players described above. This, however, leads to the important question of whether the right type of firms are placed under pressure; whether it is really the

companies that are exposed to the most problems or instead other firms. Thus, the research question of this thesis: Does exposure to severe problems lead to more criticism?

(17)

13

3. Theoretical Framework and Conceptual Model

As discussed earlier in this literature review, recent research has found that companies take on responsibilities which are beyond the usual scope of CSR (Palazzo & Scherer, 2011; Matten & Crane, 2003). Many firms have accepted at least partial responsibility in resolving serious issues (Zadek, 2007). However, for reasons stated above, some firms do not take on these responsibilities voluntarily and will act only if they are publicly criticized (Swartz, 2010). The mechanisms by which a firm is criticized have also been discussed above. It is also important to ensure that criticism is directed at the correct firms, which is interpreted in this thesis to be firms that are exposed to the most severe issues. In order to achieve the best public outcome, firms that are exposed to the worst problems should be pressured to solve them, rather than firms that are exposed to fewer or less severe problems (Doh & Guay, 2006). The first hypothesis to be tested is therefore:

H1: Being exposed to serious problems increases the risk of a firm being criticized.

Figure 1

One way to measure the degree to which a firm voluntarily helps in resolving

problems is by looking at its CSR performance. According to Bleekman (2013) a good CSR performance should cover a firm‟s economic, environmental, social and corporate

governance and include stakeholders on the inside such as suppliers and employees, as well as stakeholders from the outside such as customers. A good way to ensure that all these issues are addressed is by consulting stakeholder organizations to ensure that the CSR performance meets all of these requirements (Swartz, 2009).

(18)

14 As mentioned before, a company that moves abroad will in doing so be exposed to different ethical standards as well as new problems (Palazzo & Scherer, 2009). This means that it also has to adjust its CSR performance to meet the new demands. Maistriau (2014) mentions that the best practice here is to take the strictest requirements and use them for all of the countries. Choosing different CSR standards for different countries will only attract more criticism if it is detected (Maistriau, 2014).

A CSR policy that contains all of the points above, and which is subsequently enacted, should therefore lead to behaviour that is harder to criticize by stakeholders and other outsiders. Thus, the second hypothesis to be tested is:

H2: The relationship between a firm’s exposure to severe problems and the amount of criticism it receives is moderated by its CSR performance, meaning a better CSR performance will weaken the effect.

Figure 2

As previously mentioned, multinationals are often criticized for severe problems when operating in different markets because of varying ethical thresholds in these markets. Maistriau (2014) shows that selling to markets with more press freedom leads to higher CSR performance. It is posited that in markets with higher press freedom a firm is more likely to

(19)

15 be criticized for ethically questionable actions than in markets with lower press freedom. Maistriau (2014) describes a scenario where high press freedom can lead to more informed consumers who make choices based on their knowledge, which could lead to them favoring a competitor‟s products or services if they disagree with a firm‟s CSR policy. In extreme cases, high press freedom can even lead to NGOs starting consumer awareness campaigns and consumer boycotts against firms. It should also be noted that such criticism is not automatically generated but is critically reliant on active NGO campaigning and public criticism, which are easier and more likely to occur in countries with higher press freedom (Maistriau, 2014; Scherer & Palazzo, 2007). In particular, media attention is crucial for NGOs to campaign successfully against a firm. It is a broad platform where a great amount of people can be reached and awareness can be raised (Gamson, 2004). Without the media it is very difficult for organizations to educate the people and raise awareness to a broader group of people. More resources, both monetary and human, are required to reach the same amount of people without the media (Maistriau, 2014).

Maistriau (2014) states that firms which move from a home market with high freedom to a new market with lower press freedom keep their CSR policies while companies that go from a home market with low press freedom to a market with higher press freedom have to adjust their CSR policies to be accepted in their new environment.

This leads to the third hypothesis:

H3: Higher press freedom in the countries in which a company operates moderates the effect that exposure to severe problems has on the firm’s criticism in such a way that higher press freedom will increase the effect.

(20)

16 Figure 3

There are two types of firms: those that sell directly to end-consumers and those that sell to other firms. As argued before, the type of customers a firm has can have a severe effect on the criticism it receives (King, 2008). Scholars agree that consumers in developed

countries care about the ethical behaviour of firms from which they buy products and are willing to reprimand these firms if they do not behave ethically (Schrempf, 2013; Maistriau, 2014; King, 2008; Palazzo & Scherer, 2011). When firms choose a supplier, however, they are primarily concerned with quality, cost, flexibility and delivery performance (Cebi & Bayraktar, 2003; Verma & Pullman, 1998). CSR performance only becomes an issue for these firms if unethical behaviour of their suppliers could have an effect on their own reputation. Companies that do not sell to end-consumers are not likely to be affected by unethical behaviour of their suppliers.

In addition, it is easier for an NGO to campaign against a firm that consumers know and that they can punish by means of boycotting than a firm which only sells to other firms (King, 2008; Swartz, 2010). It is therefore less likely that unethical behaviour will be detected as fast as for other firms.

(21)

17 H4: Selling directly to end consumers moderates the effect that the exposure to severe

problems has on the amount of criticism a firm receives in such a way that it will strengthen this effect.

(22)

18

4. Research Methods

4.1 Sample and Data Collection

The sample used for this study consists of 541 companies tracked by the Covalence EthicalQuote index. Covalence bases its research on 50 CSR-criteria developed by the Global Reporting Initiative. It uses them to track online news on topics relevant to corporate social responsibility (EthicalQuote, 2015).They base their selection on 541 companies from 18 different industries according to the Dow Jones Sector Titan Index. This index consists of the most established and largest companies in each sector. All of these companies are

multinationals. The dataset only contains data for the year 2011. Nevertheless, the sample containing only data from 2011 was judged as sufficiently large so as to discern any potentially significant results, although a formal a priori statistical power analysis was not conducted. Furthermore, 2011 was sufficiently recent as for companies in the dataset to be extant at the time of conducting the study. Moreover, the worst effects of the global financial crisis of 2007-08 were diminishing by 2011 and it can be expected that the results will not be irreparably distorted by such effects. Additionally, it is the most recent year for which full data from Covalence EthicalQuote is available. As mentioned, the data is cross-sectional: arguments on the basis of causality can therefore not be made.

For each of the 541 companies, the dataset contains the primary industry in which that company is active. Each of the companies can be thus categorized into one of six different industries. Moreover, the companies vary in size as well as business model: some sell directly to customers while others sell exclusively to other businesses. Finally, they have a variety of different home countries. A point that all companies have in common, however, is that they are all active in more than one country, meaning they are all multinationals. For the sake of the research question it is important for the companies under study to constitute a

heterogeneous group to minimize the effects of any potential systematic biases which are otherwise not controlled for.

(23)

19 The dataset also contains information about how often in 2011 the companies in question were mentioned in the media. For each mention, the source of the article, the place of the incident and the category it falls in, and whether it was criticism or praise for the company is available. All of these points will be important for calculating the dependent variable. This information was gathered from Covalence EthicalQuote

Another source for this study was data provided by the Carbon Disclosure Project (CDP), a group based in the UK that works to disclose major firms‟ greenhouse gas emissions. CDP sends out questionnaires every year to different companies requesting information about where the companies are active and how much carbon emission they produce in each of these countries. This provides a good source to determine in which

countries a company is active in and rank them according to how active the company is there. The companies are not forced to answer but a great deal of them does it voluntarily to exhibit increased awareness of environmental issues. Because it is voluntary, the sample size also had to be reduced since there was not information for every company available. The data is also for the year 2011.

Another major source of data was Datastream, in particular the Worldscope and Asset4 databases, both part of Thomson Reuters. The latter provides information about the environmental, social and governance activities of over 4300 companies. This is achieved by having analysts examine the companies‟ reports and judging the companies‟ performances according to 250 key performance indicators. The Worldscope database is assembled and maintained in the same way as the Asset4 database, the only difference being that its focus is on providing a variety of financial information ranging from historical financial statements to current earnings per share.

(24)

20

4.2 Dependent Variable

The dependent variable in this study is corporate criticism. This refers to negative publicity for a firm in any sort of media. The sources of criticism and praise are not of the same origin and include reporters, the government, NGOs and other sources. Furthermore, the criticism can originate from a company‟s home country, another country the firm is active in, or a country with which it has no ties at all. The variable is constructed from data provided by Covalence EthicalQuote. For the 233 companies that remained after eliminating missing cases there are 16,313 events which are classified as either criticism or praise. After

eliminating the positive events 4,566 remained. The motivation behind the dependent variable of this study is to capture the extent to which a company is criticised. This could be achieved by taking the dependent variable to be the total number of criticisms a company has in the dataset or, alternatively, normalize this value by counting the fraction of all mentions per company that are critical. It was decided to adopt the former approach of counting the total number of criticisms. One benefit of this approach is that the results will not be distorted by how active a company is in the media since the list includes publications from the

companies themselves. It might appear at first that this approach of counting the total number of criticisms without any kind of normalization would be unfair to larger companies, since it seems reasonable to expect them to receive a larger number of criticisms purely by virtue of their larger size. This problem is eliminated by inclusion of firm size as a control variable.

4.3 Independent Variable

The independent variable is the degree of corporate exposure to severe problems. This is a composite measure consisting of two distinct parts: the degree to which each company is active in each country, and the degree of social problems faced by each country in which the company is active. A proxy metric for the degree of social problems in each country can be calculated by using a weighted average (reflecting the degree to which a company is active in

(25)

21 that country) to give an overall measure of the company‟s exposure to severe problems. Since the “degree to which a company is active in a country” can be interpreted in various ways and, because it is not practicable to measure the degree of severe “problems” in a country directly, we require proxy measures for both. These measures are discussed in turn below.

To quantify the degree to which a firm is active across the suite of countries in which it operates, we calculate a weighting based on data provided by the Carbon Disclosure Project. For a given company, the weighting of a country‟s “problem” score is calculated as the carbon footprint of the firm in that country divided by the firm‟s total carbon footprint. Other studies like Maistriau (2014) use a similar weighting based on sales rather than carbon footprint. This latter approach, however, has the disadvantage that the poorer countries where a company is active might be the source of most of the firm‟s direct activity (such as

producing its products), but do not contribute significantly to the firm‟s sales, do not get weighed fairly with respect to how exposed the firm is to that country‟s problems.

To quantify the extent of social problems in each country, the Human Development Index (HDI) is used. The HDI is a metric developed by the United Nations to rank and measure countries on social and economic criteria. It bases its measurements on four dimensions: mean years of schooling, life expectancy at birth, expected years of schooling and gross national income per capita (Human Development Report, 2011). The scores are between 0 and 1 and increase with the level of development. The UN publishes a report every year with the new scores and explanations of new developments. For this study we use HDI as an inverse proxy for the number of “problems” a country faces: lower HDI scores

correspond to greater problems. The HDI scores used in this study are all for the year 2011 and can be found in the Human Development Report of 2011.

In order to get a final measure of each firm‟s exposure to problems across the countries in which it operates, the HDI score of each country is weighted according to its

(26)

22 fraction of the firm‟s total carbon footprint to provide the independent variable for this study. It should be noted that higher values of the independent variable correspond to a firm facing lower levels of problems.

4.4 Moderating Variables

The second hypothesis states that CSR activities have a negative moderating effect on the criticism a firm receives. That means the more actively a firm engages in CSR activities, the less likely it is that the firm is criticized despite its being exposed to severe problems. The aforementioned Asset4 database was used to construct the moderator CSR activity.

As mentioned before, the Asset4 database operates with 250 key performance indicators. These can be grouped in four pillars: economic performance, environmental performance, social performance and corporate governance performance (Thomson Reuters, 2012). These numbers are normalized and the values lie within 0 and 100. The CSR variable was constructed by taking the overall score, adding all pillars together and dividing by the number of pillars. Aggregating pillars to compute CSR performance is an approach commonly used in the literature (Cheng et. al., 2014).

The third hypothesis states that press freedom positively moderates the effect exposure to severe problems has on firm criticism. This means that the higher the press freedom in a country, the more likely firms that are exposed to problems are criticized.

The moderating variable press freedom is constructed by means of the Press Freedom Index. This index is compiled and published annually by the organisation Reporters Without Borders. It incorporates censorship by the government and other authorities and the degree to which the journalists and news organisation are free to publish without restraint (Reporters Without Borders, 2012). Since these reports reflect the previous year, the data was collected from the World Press Freedom Index of 2012 (Reporters Without Borders, 2012) in order to obtain data for 2011. The variable was calculated by means of a weighted average of all the

(27)

23 countries a company is active in. For this, the data from the Carbon Disclosure Project was used. As argued before, using this data has the advantage of ensuring that poorer countries where fewer goods are sold will still be represented.

The last hypothesis states that selling to end consumers positively moderates the effect that being exposed to problems has on receiving criticism. This means that consumers are more concerned with products to which they are directly exposed, meaning that they consume directly and firms producing these products are therefore potentially more likely to be criticized. The firms in the dataset were therefore assessed on whether they sell directly to consumers (B2C) or not. We employ an indicator (or “dummy”) variable that contains the value “1” or “0”, meaning respectively that the firm sells directly to consumers, or does not. No distinction was made between companies that only sell to end consumers and firms that sell to both end consumers and other companies. The data for these variables were retrieved from the companies‟ annual reports, in most cases from the initial section where an overview of the company‟s business segments is given. For most companies this was sufficient to ascertain whether or not they sell to end customers. For the other firms descriptions about their customers could be found in the annual reports.

4.5 Control Variables

The research question of this study is whether exposure to severe problems leads to more criticism for a company. In order to be able to test the hypothesis without a bias there are some other factors that have to be controlled for since they might have an influence on the dependent variable.

The first control variable is firm size. It is measured by a firm‟s total assets. Since it can be expected that this variable can span multiple orders of magnitude, its natural logarithm will be used. The rationale behind controlling for firm size is that larger firms usually have more spending power and it could therefore be assumed that they are also more capable of

(28)

24 dealing with social problems to which they are exposed. Alternatively, because of their larger size, they could experience more criticism (Schrempf, 2013). This is potentially because NGOs consider it more effective to target a large firm than multiple smaller players. One consideration is that larger firms might have more bargaining power in the supply chain and will be therefore able to force suppliers to abide by ethical behaviour (Swartz, 2010). Alternatively, it might be easier to raise awareness and get media attention for a large firm than for a smaller firm (King, 2008).

The second control variable is the industry in which each firm operates. The dataset containing the criticism data for each firm also groups them by industry. We control for this because, as mentioned by Schrempf (2013), sometimes bad behaviour of one firm in a

particular industry can lead to further inspection of other firms in the same industry. Once the public is concerned about a topic, other firms will have to ensure that they are able to deal with it, otherwise NGOs might campaign against them (King, 2008). This can lead to particular industries being a greater focus of NGO, public and media attention than other industries and thus more likely to be criticized.

The final control variable is firm profitability. Profitability is quantified in the form of return on assets. The Worldscope Database provides information on return on assets for the companies in the sample. The motivation for controlling for profitability is that a firm which makes a larger profit may be seen as more capable of resolving problems and therefore receive more criticism (Schrempf, 2014). The public will also be more likely to judge a firm that makes a sizeable profit while not helping to resolve the problems to which it is exposed (King, 2008).

(29)

25

4.6 Data Analysis

As mentioned previously, the initial sample consisted of 524 companies. However, data for each variable could not be found for all companies, so the final sample size is 233 firms.

The method used to detect possible outliers is the outlier labelling rule (Banerjee & Iglewicz, 2007). With this method an upper and a lower boundary are calculated based on each variable‟s quartiles. The formula for the upper boundary is Q3+(2.2*(Q3-Q1)), the formula for the lower boundary is Q1-(2.2*(Q3-Q1)). As can be seen, the multiplier to set the boundary for the outliers is 2.2. This value was chosen since it ensures that the tails of a normal distribution will not be cut off. With this method a total of 20 outliers could be determined (approximately 8.6%). According to Cousineau and Chartier (2010), winsorizing the outliers or replacing them with the mean is not a favourable solution since it can lead to a more leptokurtic distribution and a higher chance for a type-I error. Because deleting the cases would decrease the size of the dataset by a great amount it was also decided against deleting the outliers. Chen et al. (2014) point out that keeping outliers is an acceptable approach when using a Poisson or negative binominal regression since these are more robust to outliers. Afterwards, the descriptive statistics for the dependent, independent, moderators and control variables are calculated. The results can be found in Table 1. Following that, the variables have to be tested on multicollinearity. This was done by performing a bivariate correlation analysis using the Pearson Correlation Index, the results of which can also be found in Table 1. The results showed some values that were above the threshold of 0.8 and could therefore be a sign of multicollinearity. In order to ensure that there is no

multicollinearity, the variance inflation factor (VIF) was calculated for each of the variables. The VIF indicates the degree to which the variance of an estimated regression coefficient is inflated due to collinearity. If the VIF is above 5 then variables are likely to be correlated.

(30)

26 Since all of the VIFs for this set of variables were under 2 it can be assumed that

multicollinearity is not an issue.

The dependent variable Total Criticism has a mean of 18.62 and a standard deviation of 37.72. This variable is a count variable and the values it contains range from 0 to 321. With the dependent variable being a count variable, a Poisson regression would be appropriate or, if the model is overdispersed, a negative binominal regression could be performed (Hilbe, 2007). Both types of regressions are robust to outliers and do not require normally distributed variables. The differentiating factor will be whether or not the dependent variable is overdispersed. One way to measure overdispersion is to compare the mean with the variance. If the variance is higher than the mean, the dependent variable is overdispersed or, if it is lower, the dependent variable is underdispersed. In this case the variance is higher than the mean and it can therefore be assumed that the variable is overdispersed. However, this is a measure of apparent overdispersion and not real overdispersion. Apparent

overdispersion can occur due to some features in the dataset that might not harm the

regression. Real overdispersion on the other hand affects the reliability of the model and its fit in general (Hardin et. al., 2007). Since this model exhibits apparent overdispersion, conducting a Poisson regression would lead to flawed results and therefore the negative binomial regression will be used to test the hypotheses. Various forms of negative binomial regression exist; the version used here, NB-2, models the variance as a second-order function of the mean, specifically V = μ(1 + αμ), where α is a so-called overdispersion parameter.

When α = 0 this reduces to a Poisson regression. To verify that real overdispersion is indeed present, a negative binomial log-linear regression was performed with the overdispersion parameter allowed to vary. The 95% confidence interval for α was found to be [1.692, 2.475]. This range does not contain 0, so we reject the hypothesis that the dependent variable is equidispersed, and proceed with negative binomial regression with the default overdispersion

(31)

27 parameter of 1. This leads to a so-called geometric regression, which is a special case of a negative binomial regression (Hilbe, 2007).

Finally it is worth noting that 21.9 percent of the values of the dependent variable are exactly zero. It therefore needs to be considered whether the zeros are being generated by a separate process than the rest of the values. If this would be the case, a zero-inflated negative binomial regression would have to be performed instead of a regular negative binomial regression. In that case the excess zeros in the dependent variable are modelled with a logistic regression or a probit model, and the rest of the values of the dependent variable modelled with a negative binomial distribution. According to Allison (2012) this is only the case if it was impossible for one part of the data to obtain a count value higher than 0. With this

dataset and the nature of the dependent variable it does not seem plausible to hypothesize that criticism counts for some firms could be artificially constrained to zero. Performing a zero-inflated regression would therefore distort the model and lead to incorrect conclusions (Ridout et. al. 1998).

The independent variable only contains numbers between 0 and 1. Its mean is 0.81 and its standard deviation is 0.13. The variable is extremely negatively skewed and highly leptokurtic. Since the negative binominal regression does not assume normality this variable does not have to be transformed to make it follow a normal distribution. The moderator exposure to press freedom shows a mean of 15.24 and a standard deviation of 14.13 which means the sample is very diverse with regard to press freedom. It is substantially positively skewed and slightly leptokurtic. The next moderator CSR performance shows a mean of 87.1 and a standard deviation of 12. This implies that the CSR performances of the observed companies are not as diverse. This variable is again extremely negatively skewed and moderately leptokurtic. 66.5% of companies in the sample can be categorized as B2C companies. The companies are not spread evenly over the six different industry

(32)

28 classifications with 76.4% being regarded as industrial, 9.4% belonging to the utilities

industry, 1.3% belonging to the transportation industry, 7.3% belonging to the bank/savings and loan industry, 3.9% belonging to the insurance industry and 1.7% belonging to other financial industries. The variable Total Assets has a mean of 932,458,607.06 and a standard deviation of 3,768,257,997.57 and is extremely positively skewed as well as extremely leptokurtic. After it was log-transformed it shows only slight signs of positive skewness and leptokurtosis. The variable Profitability has a mean of 6.78 and a standard deviation of 5.78. It is moderately positively skewed and slightly leptokurtic.

In order to assess the effects of each covariate on the model, a number of regressions were performed, using a total of seven models. The first model contains only the control variables, with the second model containing control variables and the independent variable. The third, fourth and fifth models contain the independent variable, control variables and, respectively, each of the three moderators along with its interaction term with the

independent variable. The penultimate model contains all pure variables but no interaction terms, with the final model adding in the interaction terms for the three moderator variables.

As just mentioned, models three, four, five and seven contain interaction terms. A variable, by definition, only has a moderating effect if the interaction term has a statistically significant effect on the model. Furthermore, when a moderator does have a significant moderating effect, not only the magnitude but also the direction of the relationship between independent and dependent variable can change depending on the value of the moderator variable.

As an example, consider a simple OLS regression with independent variable x, moderator z and dependent variable y. The regression equation including interactions is y = B0 + B1x + B2z + B3(x*z). Clearly the marginal effect of x on y is (B1 + B3z), thus

(33)

29 + B3z) can change sign, indicating a reversal in the direction of the effect of x on y under certain values of moderator z. For this reason interaction plots for each moderator whose interaction term attained statistical significance were constructed. These display the

relationship between independent and dependent variable for three values of each moderator: at its mean, and at a value one standard deviation on either side of its mean, under the

assumption that all other variables remained at their mean value. Two of the three proposed moderators were found to be significant: CSR activity and degree of press freedom. The respective interaction plots are shown in Figures 5 and 6. Due to pre-packaged routines for construction of interaction plots being unavailable for negative binomial regression, these plots were created manually. Note that since negative binomial regression is a log-linear model, the ordinate of these plots is the logarithm of the dependent variable.

4.7 Results

The results of the regression for each of the seven models described above will now be presented in turn.

The first model with only the control variables showed a statistically significant model with p < 0.01 significance and a likelihood ratio chi-square of 62.522. The first control variable, the logarithm of total assets, was significant at p < 0.01. However, the third control variable, return on assets, was not significant. Of the six industries only the insurance industry was significant at p <0.01.

The second model tests the first hypothesis which states that exposure to severe problems has a positive effect on receiving criticism, by including the independent variable exposure to HDI (whose composition was previously described). The model was significant at p < 0.01 and with likelihood ratio chi-square of 69.013. The regression shows that the independent variable has a negative effect on the dependent variable, and is significant at p = 0.016. It has to be kept in mind however that with HDI a higher score indicates better

(34)

30 conditions and hence fewer problems. A negative relationship between this variable therefore confirms the hypothesis that exposure to more social problems will lead to more criticism. Therefore, hypothesis 1 is not rejected. There was no change from the first model in the significance of the control variables.

The third model tests the second hypothesis which states that CSR activity negatively moderates the effect that the independent variable (exposure to severe problems) has on the dependent variable (receiving criticism). The model is significant at p < 0.001 with a likelihood ratio chi-square of 105.192. The independent variable is significant at p < 0.01. The variable CSR activity was found by itself to have a significant effect with p < 0.019. The interaction term showed significance at p = 0.001. However, it should be noted that both the CSR variable and the interaction term have a positive effect, the latter being relevant for this hypothesis. Figure 5 shows the interaction effect for this model. At the mean value of CSR (solid line in the figure), the effect of HDI on total criticism is negative. However, a high HDI score implies fewer problems, so total criticism is positively correlated with amount of

problems. At higher values of CSR activity the effect of HDI on criticism are weakened (and even changes sign when the moderator value is one standard deviation from its mean), so the effect of the amount of problems on criticism is therefore enhanced. However, hypothesis 2 states that higher CSR activity would weaken the effect of the amount of problems on

criticism. This leads to the rejection of hypothesis 2. There was no change in the significance of the control variables compared to the previous two models.

(35)

31 Figure 5

The fourth model tests Hypothesis 3 which states that press freedom strengthens the effect exposure to severe problems has on the criticism a company receives. The model itself is significant at p < 0.01 compared to an intercept-only model. In this case, the variable exposure to press freedom and its interaction term both show significance at p = 0.002. Figure 6 is used to interpret the interaction effect. The coefficient for the variable itself is negative; however, the interaction term has a positive coefficient (again, only the latter is relevant for this hypothesis concerning the moderating effect). At the mean value of press freedom, the effect of HDI on total criticism is slightly negative, i.e. higher exposure to HDI leads to lower levels of criticism. This means that more exposure to severe problems leads to higher criticism. As can be seen from the top line (one standard deviation above mean), higher press freedom leads to significantly more criticism compared to the mean level of press freedom at higher HDI levels than at lower HDI levels. The same applies to the lower line when compared to the mean, i.e. at high HDI values the difference in the amount of criticism between the mean level of press freedom and low level of press freedom is much larger than the difference at low HDI values. Thus it can be said that press freedom positively moderates the effect of exposure to HDI on total criticism. It should also be noted that at high

(36)

32 levels of press freedom the direction of the effect in fact reverses, and in this case higher exposure to HDI leads to higher levels of criticism. Exposure to HDI and exposure to problems are negatively correlated. However, it does not change the fact that press freedom positively moderates the effect of exposure to problems on total criticisms. Hypothesis 3 is therefore not rejected. The significance and effect of the other variables did not change compared to the previous models.

Figure 6

The fifth model tests the final hypothesis which states that being a B2C firm, i.e. directly selling to end customers, strengthens the effect that exposure to severe problems has on being criticized. However, neither the independent variable, the moderator nor the

interaction term were found to have a significant effect. The fourth hypothesis therefore has to be rejected. There was no change in the significance of the control variables.

The sixth model, i.e. with all variables but without the interaction terms, showed a statistically significant model with significance < 0.001 and a likelihood ratio chi-square of 106.266 when compared to an intercept-only model (i.e. a model with no predictors). All of

(37)

33 the variables individually had a significant effect on the dependent variable except for

exposure to press freedom, the control variable return on assets and the industry variables with the exception of bank/savings&loan and insurance. These two industries were both significant at p < 0.05. All effects were as predicted except for the variable exposure to press freedom that had a negative effect on criticism and the moderator CSR which had a positive effect on criticism. Note that since interaction terms were omitted in this sixth model, the effects just mentioned refer to the variables functioning as control variables rather than moderators.

For the final model, all the interaction terms are included, even the interaction term for B2C which had been found to be insignificant in the model where only B2C was considered as a moderator (the fifth model). By including the interaction terms the whole model improved to a likelihood ratio chi-square of 126.177, again with significance p < 0.001 against an intercept-only model. All of the variables were significant except for the control variable return on assets, the moderator CSR and the interaction term for B2C and all the industry variables with the exception of bank/savings&loan and insurance.

Bank/savings&loan is significant with p < 0.05 and insurance is significant at p < 0.01. The moderator exposure to press freedom still has a negative effect, the interaction term however is positive. The moderator CSR is negative as predicted but not significant while its

interaction term is significant but positive. The moderator B2C is positive and significant as predicted, the interaction term however is not significant. All regression coefficients can be found in Table 2.

(38)

34 Table 1: Correlation Matrix for all variables

* Correlation is significant at the 0.05 level (2-tailed) ** Correlation is significant at the 0.01 level (2-tailed)

Variable Mean Standard Deviation 1 2 3 4 5 6 7 8 9 10 11 12 13

1. Total Criticism 18.61 37.715 1

2. Exposure to HDI 0.8102 0.1314 -0.061 1

3. CSR 87.0904 11.9934 0.097 -0.061 1

4. Exposure to Press Freedom 15.2396 14.1295 0.034 -0.179** 0.107 1

5. B2C 0.67 0.473 0.149* -0.033 0.032 0.136* 1

6. Industrials Industry 0.7639 0.4256 -0.029 -0.032 0.084 0.076 -0.287** 1

7. Utility Industry 0.0944 0.293 0.126 -0.014 -0.001 0.058 0.105 -0.581** 1

8. Transportation Industry 0.0129 0.113 -0.026 0.05 0.01 -0.045 0.081 -0.205** -0.037 1

9. Bank/Saving & Loan 0.073 0.2606 -0.03 0.01 0.069 -0.07 0.199** -0.505** -0.091 -0.032 1

10. Insurance Industry 0.0386 0.1931 -0.082 -0.003 -0.142* -0.103 0.142* -0.361** -0.065 -0.023 -0.056 1

11. Other Financials Industry 0.0172 0.1301 0.017 0.08 -0.208** -0.046 0.024 -0.238** -0.043 -0.015 -0.037 -0.026 1

12. LogTotal Assets 18.2322 1.9318 0.131* -0.001 -0.227** -0.048 0.167* -0.347** 0.039 -0.069 0.386** 0.115 0.164* 1

(39)

35 Table 2: Regression coefficients by model

Dependent Variable: Total Criticism Model 1 Model 2 Hypothesis 1 Model 3 Hypothesis 2 Model 4 Hypothesis 3 Model 5 Hypothesis 4 Model 6 Model 7

Constant -4.202 (0.003) -2.692 (0.078) 5.113 (0.181) -0.503 (0.775) -3.576 (0.041) -5.555 (0.000) 2.725 (0.510)

Independent Variables

Exposure to HDI -1.577 (0.016) -15.903 (0.000) -4.197 (0.000) -0.299 (0.794) -1.760 (0.008) -13.246 (0.006)

CSR -0.094 (0.019) 0.033 (0.000) -0.065 (0.118)

Exposure to Press Freedom -0.159 (0.002) -0.008 (0.244) -0.151 (0.003)

B2C 1.742 (0.125) 0.581 (0.000) 2.683 (0.000)

CSR x Exposure to HDI 0.162 (0.001) 0.129 (0.014)

Exposure to Press Freedom x Exposure to HDI 0.210 (0.002) 0.191 (0.005)

B2C x Exposure to HDI -1.418 (0.307) -2.662 (0.082)

Control Variables: Firm characteristics

Total Assets 0.358 (0.000) 0.352 (0.000) 0.410 (0.000) 0.343 (0.000) 0.318 (0.000) 0.360 (0.000) 0.391 (0.000)

Return on Assets 0.011 (0.486) 0.010 (0.504) -0.004 (0.807) 0.003 (0.825) 0.006 (0.699) -0.001 (0.946) -0.012 (0.424)

Control Variables: Industry

Industrial 0.543 (0.323) 0.396 (0.473) -0.136 (0.826) 0.409 (0.460) 0.504 (0.364) 0.152 (0.796) 0.006 (0.992)

Utility 0.863 (0.127) 0.764 (0.178) 0.308 (0.624) 0.899 (0.116) 0.741 (0.192) 0.414 (0.490) 0.453 (0.477)

Transportation 0.382 (0.642) 0.370 (0.653) 0.258 (0.765) 0.601 (0.467) 0.221 (0.788) 0.077 (0.927) 0.358 (0.684)

Bank/Savings & Loan -0.713 (0.212) -0.858 (0.135) -1.638 (0.013) -8.50 (0.139) -0.876 (0.128) -1.413 (0.024) -1.679 (0.011)

Insurance -1.679 (0.009) -1.83 (0.005) -2.436 (0.001) -1.804 (0.005) -1.899 (0.003) -1.623 (0.022) -2.405 (0.002)

Other Financial 0ᵃ 0ᵃ 0ᵃ 0ᵃ 0ᵃ 0ᵃ 0ᵃ

Model Fit

N 233 233 233 233 233 233 233

Likelihood Ratio Chi-Square 62.544 69.013 105.192 78.296 83.617 106.266 126.177

P-value 0.000 0.000 0.000 0.000 0.000 0.000 0.000

ᵃ Set to zero because this parameter is redundant Significance level shown in parenthesis

(40)

36

5. Discussion

This study analyses the relationship between a firm receiving criticism and being exposed to severe problems. Most importantly, the results of hypothesis 1 show that a firm that is exposed to more severe problems does indeed face more criticism than a firm that is not.

Moreover, having a stronger CSR policy does not dilute the effect that being exposed to problems has on receiving criticism. It was found that there was an effect, however it was in the opposite direction: that a stronger CSR policy tends to strengthen the relationship between exposure to problems and criticism received. This raises the question as to whether firms that are criticized a lot are forced into stricter CSR policies. Another explanation could be that a firm has a strict CSR behaviour in one country but does not set the same standards for all countries in which it is active. Especially if such a case is detected by an NGO the firm will receive a great deal of criticism.

Moreover, firms that are active in countries with high press freedom tend to

experience more criticism when they are exposed to severe problems. This finding is in line with previous research and confirms the theories other scholars have put forward. This can be explained by the crucial role that the media plays in criticizing companies nowadays. Without unhindered access to the media it is harder for NGOs to raise criticism and inform the general public, i.e. end consumers.

Lastly it was found that selling to end consumers does not have a significant effect on receiving more criticism when being exposed to social problems. This means that NGOs (a major source of criticism) do not focus exclusively on consumer companies but address all companies in a more equitable way, perhaps in accordance with the perceived severity of these firms‟ specific practices. In turn, end consumers do not only seem to care about

(41)

37 companies with which they are in daily contact: if the problem is severe enough, they will criticize the company irrespective of its business model.

5.1 Academic Relevance

This research adds new insights to the literature. CSR and how firms implement it has been a popular topic in the academic literature for the past 20 years. Companies moving abroad and being exposed to severe problems has also been discussed. This study works to bridging the gap between CSR policy and the problems of firms‟ geographic expansion. Even though there is extensive literature on both of the topics mentioned above, relatively few scholars address whether it is the „right‟ firms that are receiving criticism, meaning the firms that are exposed to the greatest burden of problems (whether by number or severity thereof) and whose change in behaviour would therefore have the most powerful effect on solving the problems.

5.2 Managerial Implications

The managerial implications of this study are that managers can see it as a guideline when making decisions to go abroad. This study proves that it is crucial to assess not only the business aspect but also the social circumstances of a new business opportunity. Managers must assess whether it is worth saving money on labour costs, for example, while receiving more criticism, or alternatively if moving somewhere with slightly higher labour costs but less severe problems is a better solution.

Furthermore, this study shows not to underestimate the importance of the media (and the degree of freedom thereof) in a country and how media attention can be both favourable and critical.

Moreover, the results indicate that good CSR performance does not protect a company from being criticized. Managers must be aware of this and pay close attention to what they put in their CSR policy and how they implement it.

Does exposure to severe problems lead to more criticism?