MASTER THESIS
An Extensive Study of Interaction Effects on the Corporate Social
Responsibility-Customer Loyalty Relationship and the Mediating Effect of
Customer Satisfaction
By
Marjolijn Onrust
Msc Research Master for Economics and Business Marketing Profile
Supervisors: Prof. dr. Peter C. Verhoef
Dr. Jenny van Doorn
July, 2011
University of Groningen Faculty of Economics and Business
Duindoornstraat 11 9741 NM Groningen
Table of Contents
1. Introduction 2
2. Conceptual Framework and Hypotheses 11
3. Methodology 28
4. Analyses and Results 47
5. Discussion 60
6. References 65
Abstract
Although previous research has addressed the effect of corporate social responsibility (CSR) on customer outcomes, no study has yet attempted to explain the variation in these effects through the analysis of a wide range of interaction effects. This study aims to investigate whether CSR has a positive effect on customer loyalty over and above the effect of customer satisfaction, while simultaneously examining various moderating variables at the consumer level, company level and sector level. Additionally, the mediating effect of customer satisfaction on the CSR-customer loyalty relationship is analysed. Results show that CSR positively and significantly affects customer loyalty, a relationship that is moderated by gender, age, and income at the consumer level, by corporate ability at the company level and by competitive intensity and product visibility at the sector level. In contrast with previous research, only marginal support can be provided for a (partial) mediating effect of customer satisfaction. The hierarchical structure of the large sample of consumers, companies and sectors is taken into account by means of a three-level regression analysis.
Key words: Corporate Social Responsibility (CSR), customer loyalty, customer satisfaction,
1. Introduction
In this period of time, it has become standard practice for most multinational corporations to publish corporate social responsibility (CSR) statements, in addition to their regular financial statements. It appears that CSR occupies such a prominent place in today’s business, that not engaging in it has become a competitive disadvantage (Du, Bhattacharya and Sen, 2007). CSR is a broad concept, of which many different corporate examples exist. Two of such examples are the following. First, Shell is located in a sector where CSR is particularly important due to the potential risk carried by operations in damaging the environment. In the 2010 Shell Sustainability Report (2010), the company emphasises the following list of CSR actions: ‘producing more natural gas for power generation; focusing on sustainable biofuels; helping to develop carbon capture and storage technology; and making our own operations more energy efficient’ (p. 1). This example shows an application of CSR from a technological developmental perspective. A second example is a company that has a very different nature of operations, but is also active in CSR: the retail chain H&M. The nature of the retail sector in which it is located leads H&M to emphasise fair dealings with workers and the effect of operations on the environment in their CSR policies. In their 2010 sustainability report, H&M (2010) management therefore primarily focuses on equitable treatment of workers, decent salaries, workers’ rights, health and safety, stable purchasing and sales markets, responsible use of resources and the production of a conscious clothing line from environmentally adapted materials. Additionally, H&M is one of the largest retailers using certified organic cotton in the production of clothing. These two examples of companies engaging in CSR indicate that CSR initiatives are important in a wide range of sectors and are implemented in many different forms.
commitment to ensure societal and stakeholder wellbeing through discretionary business practices and contributions of corporate resources’ (Du, Bhattacharya and Sen, 2010; Kotler and Lee, 2005; Luo and Bhattacharya, 2006). Several papers have suggested that a CSR definition should state that CSR goes beyond the interests of the company (e.g. McWilliams and Siegel, 2001), but from a marketing perspective, it is acknowledged that CSR can have both a positive effect on society and on the performance of corporations. In this marketing study, the main focus therefore lies on the customer stakeholder group.
It has been discovered that CSR positively affects both company-level and customer-level performance outcomes in marketing and other research areas, which are included in Table 1. First, at the company level, CSR has an effect on financial corporate performance in several studies. In 2008, over 225 studies were identified on CSR-corporate financial performance (CFP) relationships by the Socially Responsible Investing Studies Web site (Godfrey, Merrill and Hansen, 2009; Studies of Socially Responsible Investing). In marketing studies on CSR, several scholars have focused on this relationship. Orlitzky, Schmidt and Rynes (2003) perform a meta-analysis of 52 studies and find a positive relationship of CSR on CFP. Luo and Bhattacharya (2006) show that CSR positively affects both Tobin’s q (a stock price-based measure) and stock return, thus increasing the financial or market value of the company. Also focusing on the stock market are Gallego-Álvarez et al. (2010), who show that all CSR practices, and particularly those focusing on enhancing corporate image, have a positive effect on shareholder value creation. Finally, Luo and Bhattacharya (2009) find that corporate social performance (CSP) decreases undesirable firm-idiosyncratic risk, which is the risk carried by firm-specific strategies.
2007; Lichtenstein, Drumwright and Braig, 2004; Marin, Ruiz and Rubio, 2009; Pirsch et al., 2007; Pivato, Misani and Tencati, 2008), customer commitment (Bhattacharya, Korschun and Sen, 2009; Lacey and Kennett-Hensel, 2010) and word of mouth (WOM) or recommendation intentions (Du et al., 2007; Du et al., 2010; Lacey and Kennett-Hensel, 2010; Vlachos et al., 2009). From this list of CSR customer outcomes, loyalty is used as outcome measure in this research.
In addition to many different outcomes on which the effect of CSR is measured, various moderating and mediating mechanisms have been included in CSR research. Table 1 gives an overview of these variables. The mediators most often included are customer-company congruence or identification (Sen and Bhattacharya, 2001), trust (Vlachos et al., 2009) and customer satisfaction (Luo and Bhattacharya, 2006). These variables are found to both partially (Luo and Bhattacharya, 2006) and fully (Vlachos et al., 2009) mediate the relationship between CSR and its outcomes. In addition to mediating effects, a long list of moderators are known to affect CSR relationships. Moderators that have been tested in past research include innovativeness capability (Luo and Bhattacharya, 2006), perceived service quality (Vlachos et al., 2009), R&D (Luo and Bhattacharya, 2009), and advertising (Luo and Bhattacharya, 2009). Generally, these factors strengthen the positive effect of CSR on an outcome, but innovativeness capability is an example where the effect of CSR on market value becomes negative (Luo and Bhattacharya, 2006). It appears that, when innovativeness capability is low, CSR leads to negative market value, probably because in such cases a noncompetitiveness signal is sent to the market. However, a drawback in previous CSR research is that only a low number of moderators have been examined simultaneously. This study therefore investigates the effect of a wide range of moderating effects in one model.
In addition to simultaneously testing moderators that have already been included separately in previous CSR research, I also include various new moderating effects. At the consumer level, consumer confidence and involvement are novel constructs in CSR research. At the company level, market leadership is a performance measure that distinguishes the market leader in a sector from the market followers. Finally, at the sector level, competitive intensity, product type with respect to the distinction between hedonic and utilitarian products and product visibility have not yet been tested in previous studies. This leads to the discovery of interesting new insights in the effect of CSR on customer loyalty.
good practice to also take customer satisfaction into account. By this means, if CSR is known to increase customer loyalty when satisfaction is also accounted for, this provides even more support for the positive CSR-loyalty relationship. However, because a comprehensive overview of the moderating effects on the satisfaction-loyalty relationship is still non-existent, this study also explicitly includes satisfaction as an independent variable in the model and examines its source of variation by means of a list of moderators. Additionally, because satisfaction and similar constructs are included as mediating effects in CSR relationships in several studies (Luo and Bhattacharya, 2006; Swaen and Chumpitaz, 2008) the CSR-loyalty model mediated by satisfaction is also examined at the end of this paper.
With respect to the sectors included in previous CSR studies, Table 1 shows that most papers focus on only one sector, while some include a few, but with a maximum of three. This low number of sectors is a substantial gap of previous CSR research, since the simple example of Shell and H&M already shows vast differences in CSR initiatives between sectors. Because of these disparities, the effect of CSR on customer outcomes is also expected to be different. Therefore, several scholars have proposed the inclusion of a sector dummy as moderating variable (Bhattacharya et al., 2009; Du et al., 2010). However, this is only useful if various companies from many different sectors are included, which is generally not possible because of data limitations. Additionally, including dummy variables for sectors does not account for different levels of variance that exist in a hierarchical data set on customers of companies within sectors, which is another gap of previous CSR research. This study wishes to fill this research gap by collecting an initial sample of 95 companies from 18 sectors and taking the hierarchical data structure into account by means of multilevel regression analyses.
regression analysis. Fourthly, the mediating effect of customer satisfaction is further examined. Finally, several novel moderating effects are included, that have not yet been investigated in previous CSR research.
TABLE 1
Overview of Studies on Direct, Moderated and Mediated Effects in CSR Research
Authors Year Branche/Products Company
Outcomes Consumer Outcomes Identified Mediator Identified Moderator Independent Variables Remarks1
Sen and Bhattacharya 2001 Calculators Computers Printers
Company
evaluation Purchase intention Customer-company congruence (+) CSR support (+); CSR-corporate ability (CA) beliefs (+); CSR domain (+) CSR information (+) Orlitzky, Schmidt and Rynes 2004 CFP (corporate financial performance) Managerial learning (+); Company reputation (+) Measure of CSP (+): accounting based (stronger than) market based CSP (corporate social/ environmental performance) (+) Meta-analysis Lichtenstein, Drumwright and Braig
2004 Food chain Store loyalty;
Emotional attachment; Store interest; Customer donations; % of shopping at store; Year-to-date purchases Customer-Company Identification (+)
Perceived CSR (+) Field survey and
experiment
Klein and Dawar 2004 Energy Purchase intention Brand evaluation
(+) CSR importance (+) CSR associations (+) Luo and
Bhattacharya 2006 Various from Fortune 500 companies
Market value Customer
satisfaction (+) Corporate ability: Product quality (+);
Innovativeness (+, when low)
Authors Year Branche/Products Company Outcomes Consumer Outcomes Identified Mediator Identified Moderator Independent Variables Remarks1
Marin and Ruiz 2007 Retail banking Identity attractiveness
Consumer-company congruence (+)
CSR and CSR support (+) Du, Bhattacharya and
Sen 2007 Yoghurt Customer-company identification;
Customer loyalty; Advocacy (long-term relational outcomes like WOM) CSR positioning (+) CSR beliefs (+)
Pirsch, Gupta and
Grau 2007 No distinction made (random
consumers)
Customer loyalty; Attitude towards the company; Consumer scepticism; Purchase intent CSR programs: institutionalised programs and promotional programs (+) Exploratory online survey
Madrigal and Boush 2008 Clothing Attitude toward the
ad;
Attitude toward the brand;
Attitude toward the product
Willingness to reward for social responsible behavior (+) Social responsibility of brands (+) Experiment
Pivato, Misani and Tencati
2008 Organic
supermarket products
Brand loyalty Customer trust (+) CSP (+)
Luo and Bhattacharya
Authors Year Branche/Products Company Outcomes Consumer Outcomes Identified Mediator Identified Moderator Independent Variables Remarks1 Bhattacharya,
Korschun and Sen2 2009 No distinction made Identification; Customer
commitment; Customer trust; Customer satisfaction; Several behavioural outcomes Identification; Customer commitment; Customer trust; Customer satisfaction; Several behavioural outcomes Stakeholder level factors (demographic profile: e.g. gender & involvement); Organisational level factors (company reputation, sector) Stakeholders perceptions of CSR initiatives Meta-analysis: General conceptual framework on customer responses to CSR
Vlachos et al. 2009 Mobile telecom Repeat patronage
intentions; Recommendation intentions
Customer trust (+) Perceived service quality (+) Consumer perceptions of a company’s motives for engaging in CSR actions (+/- depends on motive) Marin, Ruiz and
Rubio 2009 Banking Customer loyalty; Identity attractiveness Company evaluation (+);
Consumer-company identification (+)
Identity salience
(+) Perceived CSR
Authors Year Branche/Products Company Outcomes Consumer Outcomes Identified Mediator Identified Moderator Independent Variables Remarks1
Du, Bhattacharya and
Sen2 2010 No distinction made External outcomes for
investors: amount of invested capital, loyalty Internal: awareness, attributions, attitudes (identification), customer trust; External outcomes for consumers: purchase, customer loyalty, advocacy Stakeholder characteristics: stakeholder types, issue support, social value orientation; Company characteristics: reputation, industry, marketing strategies CSR
communication Literature exploratory paper
Gallego-Álvarez et
al. 2010 Not specified (120 largest European companies) Corporate reputation; Shareholder value creation CSR as marketing strategy (+) CSR practices (+) Bügel 2010 Banking Telecom Supermarkets Health insurance Automotive Customer
commitment Customer satisfaction (+) CSR support by customers (CSRs) (ns); Company innovativeness (ns satisfaction, + commitment); Sector (+) CSR of the firm (CSRf) (+)
2. Conceptual Framework and Hypotheses
Figure 1a shows the conceptual framework that is tested in this paper, examining the direct effect of CSR on loyalty, with moderating effects indicated by dotted lines. Figure 1b presents the conceptual model that investigates a mediating effect of satisfaction on the CSR-loyalty relationship, which is tested separately from the moderating and direct relationships. Moderating effects are tested on the direct relationships instead of in the mediation model, because there does not yet exist a straightforward manner to test multiple moderated effects in a mediated multilevel regression model. Apart from that a too complex model with too many parameters would result in that case, the technique to test mediation itself in multilevel regression is not yet reliable, because the significance of the multilevel mediated effect cannot simply obtained as proposed by Baron and Kenny (1986). This is caused by that the separate mediation slopes also vary across the second-level and third-level units, which is not accounted for by standard mediation tests (Kenny, Korchmaros and Bolger, 2003).
The direct effect of satisfaction is also included in Figure 1a and the moderating effects are tested on this relationship as well, but since much research has already focused on the effect of satisfaction on loyalty and because CSR is the main focus of this study, this chapter focuses on the relationships with CSR in Figure 1a. The satisfaction portion of the model is shortly explained at the end of this chapter. Additionally, Figure 1a indicates that two controls are included in the model: relationship duration and switching costs. Relationship duration is taken into account, because I wish to control for past consumer behaviour, that can lead to customer loyalty merely because of inertia effects (Ganesh, Arnold and Reynolds, 2000; Verhoef, 2003). Switching costs are controlled for because customer loyalty can also be affected by company-induced switching barriers aside from the intrinsic consumer motives to remain loyal to a company (Bell, Auh and Smalley, 2005). The subsequent paragraphs provide an explanation of the variables in the frameworks and several hypotheses are proposed on the relationships between the constructs. An overview of the hypotheses can be found in Table 2 at the end of this chapter.
2.1 CSR
CSR
Customer Loyalty Customer
Satisfaction
about a company (whether real or manipulated) is provided by researchers to participants in the form of a story and the effect of CSR is subsequently measured by consumer reactions to the CSR story (Klein and Dawar, 2004; Sen and Bhattacharya, 2001). However, because CSR
FIGURE 1a
Conceptual Framework Direct and Moderating Effects
FIGURE 1b
Conceptual Framework Mediating Effect of Customer Satisfaction
experiences of actual consumers are not measured in these types of research, the practical conclusions that can be drawn for companies’ CSR policies are limited. Therefore, during the
CSR Customer Satisfaction Customer Loyalty Individual Level - Gender - Age - Income - Consumer confidence - Involvement Sector Level - Competitive intensity/dynamism - Product type (hedonic/utilitarian) - Product visibility to others
past decade, marketing research has primarily focused on the perception of company CSR by consumers (Du et al., 2007; Lichtenstein et al., 2004; Luo and Bhattacharya, 2006; Marin and Ruiz, 2007; Marin et al., 2009). Vlachos et al. (2009) have investigated the effect of the underlying company motive to engage in CSR as perceived by consumers. Perceived CSR is generally measured by means of several questionnaire items, to which consumers attribute a score on a Likert scale.
While CSR as perceived by consumers has been the dominant way of measuring CSR in marketing research, CSR is a broad concept including many types of CSR initiatives. Consequently, papers measuring perceived CSR tend to vary widely on the range of items included in questionnaires. For instance, where Lichtenstein et al. (2004) primarily focus on CSR as the extent to which they are engaged in charitable contributions, Marin and Ruiz (2007) also include items on equitable treatment of women and minorities. As a result of these different interpretations on what initiatives are encompassed by the CSR concept, recent research has focused on more general measures of consumers’ perceptions of the social responsibility of companies and their brands (e.g. Du et al., 2007).
Another measure of CSR that has been used in recent research is the CSR ranking of a list of companies, where companies are ranked from best-CSR score to worst-CSR score (Gallego-Álvarez et al., 2010; Luo and Bhattacharya, 2006; Luo and Bhattacharya, 2009). This approach allows researchers to include a large sample of companies from different sectors, since survey data on customers’ perceived CSR can generally only include one or a few companies, due to sample availability and time constraints. The ranking method generally refers to CSR as corporate social performance (CSP), because a ranking more clearly shows how companies perform differently on CSR, which can subsequently be related to a company’s financial performance (Orlitzky et al., 2003). However, ranking companies based on their CSR means that less variance is included in the independent variable than when consumers provide scores as an indication of companies’ CSR policies. Therefore, this study combines previous research by measuring CSR as perceived by consumers by means of CSR scores (instead of rankings), but for a large sample of companies located in different sectors, which enables the determination of the extent to which CSR differs between these companies and sectors.
2.2 Customer Loyalty
areas (Kumar and Shah, 2004). In order to build a sustainable competitive advantage, building customer loyalty through marketing efforts is an important corporate objective (Dick and Basu, 1994). It has not only been found important in the fast-moving consumer goods sector, but also for ‘industrial goods (vendor loyalty), services (service loyalty), and retail establishments (store loyalty)’ (Dick and Basu, 1994, p. 99). Loyalty is affected by several constructs in marketing, including affective commitment (Johnson, Herrmann and Huber, 2006), which is defined as ‘the degree to which a customer identifies and is personally involved with a company’ (p. 123) and related factors such as customer satisfaction and trust. Affective commitment is related to CSR in that acting socially responsible strengthens consumers’ affective attitude towards the company, which subsequently creates higher loyalty, irrespective of the actual use of the company’s product or service (Salmones, Crespo and del Bosque, 2005). The mechanism underlying this effect is that companies’ positively interpreted CSR policies leads those companies to be connected to consumers’ knowledge systems (i.e. schemata) on social responsibility (Andreassen and Lindestad, 1997), leading to higher affect for the company and subsequently higher loyalty.
consumer and higher identification with the company. CSR as perceived by the consumer is therefore expected to positively influence loyalty. The following hypothesis results:
H1: CSR has a positive effect on customer loyalty.
2.3 Moderating Effects
2.3.1 Consumer-level moderators
At the level of the individual consumer, the relationship between CSR and loyalty is affected by various variables. Several hypotheses are proposed on the interactions with CSR, to provide further explanation under which conditions this direct relationship is more relevant. These hypotheses concern interactions at the same (consumer) level at which CSR and loyalty are measured, and include both demographic and situational variables.
Gender, age and income
H2: The positive effect of CSR on customer loyalty is stronger for women in
comparison to men.
Throughout the years, individuals are expected to follow several stages of moral development (Borkowski and Ugras, 1992; Erikson, 1987; Kohlberg, 1984). As we grow older, we become increasingly aware of moral principles and we learn to care for others. While most people progress from self-focused individuals to community-oriented members of society, the ultimate stage consists of the adoption of universal ethical principles, something which, according to Kohlberg (1984), can only be attained by a number of adults after the age of 20. However, research on the purchase of environmentally friendly products shows that such consumption was more popular among young consumers than older consumers several years ago, while recently the effect seems to be reversed (Laroche et al., 2001). This can be explained by that green products were considered trendy right after their introduction, due to their novelty, and therefore appealed more to young consumers, while this effect has now faded. Furthermore, CSR does not only concern the purchase of green products, but is a much wider concept, as explained in the introduction. Because CSR relates to the moral responsibility of consumers, where older consumers are expected to possess a higher degree of moral responsibility, the effect of CSR is expected to strengthen with age. This results in the following hypothesis:
H3: The positive effect of CSR on customer loyalty is stronger when the age of
consumers increases.
products (Welsch and Kühling, 2009) and a similar effect is expected to apply to the more broadly defined concept of CSR:
H4: The positive effect of CSR on customer loyalty is stronger when income increases.
Consumer confidence
Consumer confidence operates through two mechanisms: an economic mechanism and psychological one (de Vries et al., 2011). This research does not test consumer confidence as the economic mechanism persé, since this assumes consumer confidence as an objective measure of ability to buy. In that scenario, necessity products have a downward sloping Engel curve, where the proportion spent on such products decreases as income increases, while luxury products have an upward Engel curve (Vickers and Renand, 2003). Instead, the focus here lies on the psychological mechanism, which assumes that subjective attitudes are held by consumers about their finances and the economy. Katona (1974) underlines that people save more and spend less when their confidence in the economy is low. Furthermore, low consumer confidence often coexists with a decrease in income, which leads to even lower spending. When consumers feel that they have less money to spend now or in the future, they generally cut back on products or product features that are non-essential. Instead, low confidence consumers pay more attention to aspects such as price, while high confidence consumers (continue to) value aspects such as a strong brand (de Vries et al., 2011). This is the reason why luxury products generally experience lower demand when consumer confidence in the economy is low. In addition to luxury products, the growth of green product has been shown to slow down because of the recent economic crisis (Martin, 2008; Nielsen, 2008).
H5: The positive effect of CSR on customer loyalty is stronger when consumer
confidence increases.
Involvement
Oliver (2009) defines involvement as ‘a focused orientation toward specific products and services of a more intense nature, consisting of greater pre-purchase behaviour (e.g. search), greater attention to the act of consumption, and greater processing of consumption outcomes’ (p. 23). When a consumer is more highly involved with consumption, he or she thus exhibits a higher level of interest and effort with respect to the consumption process. Gordon, McKeage and Fox (1998) consider involvement as a motivational state that can result from many sources of value, such as utilitarian, sign and hedonic value. Because CSR policies can function as an addition of value to the consumption of products and services through its signalling of a social identity, Bhattacharya et al. (2009) emphasise customer involvement as an important stakeholder-level contingency to moderate the effect of CSR on customer outcomes. Therefore, when consumers are highly involved in consumption, the social identification effect of companies’ CSR policies, that allows customers to associate themselves with social responsibility, become even stronger. This leads to the following hypothesis:
H6: The positive effect of CSR on customer loyalty is stronger when consumer
involvement increases.
2.3.2 Company-level moderators
Apart from consumer-level moderators, several company-level effects are expected to explain the relationship between CSR and customer loyalty. These so-called ‘cross-level effects’ lead to the formulation of several hypotheses.
Corporate ability
consumer to the company and increased social identification (Bhattacharya et al., 2009; Salmones et al., 2005; Tajfel, 1978). Therefore, one might expect that corporate ability strengthens the effect of CSR on loyalty, something that has been shown in several studies. Indeed, Brown and Dacin (1997) find both a positive direct effect of both CA and CSR on company and product evaluations. Furthermore, Berens, van Riel and van Rekom (2007) show that CA and CSR can complement each other in certain areas, since both concepts result in a positive company image being transferred to consumers. In literature on the moderating effect on CSR relationships, Sen and Bhattacharya (2001) show CSR has a stronger effect on purchase intentions when it is CA-relevant. Finally, Luo and Bhattacharya (2006) find a positive moderating effect of the corporate ability product quality on market value, which indicates that even higher financial returns are reaped as a result of CSR when company products are known for their high quality.
However, literature also indicates that corporate ability can negatively moderate the effect of CSR. Consequently, Luo and Bhattacharya (2006) find that company innovativeness makes the effect of CSR on market value negative in cases of low innovativeness, because low innovativeness sends a signal to the market that the company is not being competitive and has the wrong priorities. Similarly, Luo and Bhattacharya (2009) test R&D as moderating the relationship between CSP and firm-idiosyncratic risk, where the simultaneous pursuit of CSP, advertising and R&D increases firm-idiosyncratic risk, instead of decreasing it. An explanation is that a company which simultaneously focuses on multiple activities that are not directly profit-generating, sends the message to investors that management is distracted from core activities such as production efficiency and quality. In that case, CSR and the other corporate abilities represent trade-offs that have to be made, leading to a weakening of the CSR effect (Sen and Bhattacharya, 2001). According to Luo and Bhattacharya (2009): ‘social responsibility programs not only can be costly but also can compete for a firm’s limited financial resources with other critical marketing instruments’ (p. 198). Therefore, it is expected that when a company focuses on both CSR and corporate abilities that are not central profit-generating activities, it sends the signal that it does not sufficiently invest in central activities. This is the reason why the moderating effect of corporate ability is expected to make the positive CSR-loyalty relationship weaker. The following hypothesis results:
H7: The positive effect of CSR on customer loyalty is weaker when corporate ability
Advertising
Luo and Bhattacharya (2009) emphasise the importance of advertising to CSR, where advertising can make CSR an intangible market-based asset for the company. Higher advertising creates greater consumer awareness of the company and its new and existing products, and thus greater awareness of CSR policies. Therefore, several studies on CSR and advertising have examined the effect of CSR communication (programs) on CSR relationships (Du et al., 2010; Pirsch et al., 2007), or have included advertising as a moderating variable (Luo and Bhattacharya, 2009). Du et al. (2010) emphasise the importance of effectively communicating CSR to customers. Furthermore, Pirsch et al. (2007) find that institutionalised CSR programs (long-term CSR policies that meet the needs of all stakeholders) generally lead to higher customer loyalty, while promotional CSR programs comparatively lead to higher purchase intent. Finally, Luo and Bhattacharya (2009) include advertising as an intangible market-based asset to the firm, to moderate the relationship between CSP and firm-idiosyncratic risk. They find that advertising, in addition to CSP, leads to an additional decrease in firm-idiosyncratic risk. This leads to the following hypothesis:
H8: The positive effect of CSR on customer loyalty is stronger when advertising
increases.
Market leadership
relationships with customer outcomes (Bhattacharya et al., 2009; Du et al., 2010). High performance, by being a market leader instead of a market follower, is therefore expected to increase the positive effect of CSR on loyalty:
H9: The positive effect of CSR on customer loyalty is stronger for market leaders in
comparison to market followers.
2.3.3 Sector-level variables
Several variables at the sector level are hypothesised to affect the CSR-loyalty relationship. Sector is proposed as a moderating variable in various CSR papers (Bhattacharya et al., 2009; Du et al., 2010). However, a multilevel analysis takes differences between sectors into account by design, yielding the inclusion of a sector dummy to account for differences between sectors redundant. The multilevel analysis therefore allows for the inclusion of specific sector-level moderating effects, resulting in cross-level interactions between the individual consumer variables and those varying per sector.
Competitive intensity/dynamism
Competitive intensity refers to the number of competitors in the market and how fierce they compete, where in a market of high competitive intensity customers have many alternatives to satisfy their needs and preferences (Kumar et al., 2011). Market dynamism refers to the frequency of major changes that are related to the market, which implies that the sector and its consumer preferences are fast-moving. Market dynamism, according to Homburg and Pflesser (2000), requires an above-average degree of market orientation on behalf of the company, due to the high frequency of changes in consumer demands. A high degree of market orientation is also needed in markets of intense competition, because in those sectors, it is imperative for companies to differentiate themselves from competitors. Both competitive intensity and dynamism are aspects that characterise environmental uncertainty and are very much related, which is the reason for discussing the effect of the two jointly (Homburg, Krohmer, and Workman, 1999).
companies to engage in CSR: to merely keep up with competitors, or to take a proactive stance through founding CSR principles and pioneering CSR innovations. In the latter case, CSR has the possibility to create a competitive advantage for the company (Bhattacharya and Sen 2004; Porter and Kramer, 2006). Additionally, Du et al. (2007; 2010) refer to CSR positioning as an important way to increase the positive effect of CSR communication to consumers, where CSR positioning refers to ‘the extent to which a company relies on its CSR activities to position itself, relative to the competition, in the minds of consumers’ (Du et al., 2010, p. 15). Du et al. (2007; 2010) argue that consumers pay more attention to the CSR message, have a higher belief in the authenticity of the company’s CSR initiatives and are subsequently more highly persuaded in favour of the company, when that company is positioned specifically on CSR and differentiates itself from competitors.
Based on the reasoning above, CSR is expected to be higher in sectors where competitive intensity and dynamism are high, because companies are motivated to invest in CSR as a strategy to obtain a sustainable competitive advantage and subsequently a more stable customer base. By this means, CSR allows companies to differentiate themselves from competitors. An example of a study that explicitly discusses the potential moderating effect of competitors is that by Bhattacharya and Sen (2004). They also refer to sector-level effects, since high levels of CSR activity can be found in certain sectors (e.g. oil, gasoline, alcohol and athletic footwear), either due to the nature of the operations in the sector, or because of actions by companies related to social (ir)responsibility in the past. The following hypothesis is formulated:
H10: The positive effect of CSR on customer loyalty is stronger when competitive
intensity increases.
Product type (hedonic/utilitarian)
utilitarian goods are primarily instrumental and functional (microwaves, minivans, personal computers, etc.)’ (p. 60). Therefore, the consumption of hedonic products is generally related to emotive aspects of product usage experience (Hirschman and Holbrook, 1982), while the consumption of utilitarian products relates more to cognitive functional or practical aspects of the usage experience (Dhar and Wertenbroch, 2000).
There are two lines of research that are contrasting with respect to the moderating effect of hedonic versus utilitarian product type. First, the purchase of products and services mainly for hedonic pleasure and fun can induce a feeling of guilt in consumers, which may diminish the pleasure of consumption (Strahilevitz and Myers, 1998). Consequently, consumers can attempt to reduce their feelings of guilt by means of altruistic behaviour, either in the form of donations or by means of purchasing from a company that is known for its strong CSR policy. This line of research suggests that the effect of CSR is stronger for hedonic products in comparison to utilitarian products. However, in contrast, consumers also wish to fully enjoy hedonic products and adding an organic or CSR claim to hedonic products therefore may diminish the pleasure of consumption, while utilitarian products are functional, making consumption already neutral or less enjoyable. If the consumption of utilitarian products becomes socially responsible, this will therefore add a positive element to the consumption experience. Van Doorn and Verhoef (2011) examine willingness to pay (WTP) by consumers for products with an organic claim, differentiating between vice (hedonic) and virtue (utilitarian) products. Market shares are discovered to be lower for vice food categories than for virtue food categories, because vice products with an organic claim are associated with lower quality, while this effect is absent for virtue categories. Therefore, it appears that the type of product in the sense of being hedonic or utilitarian, is associated with the degree of product quality. Product or service quality is included as a moderator in several CSR studies, having a positive moderating effect on CSR relationships (Luo and Bhattacharya, 2006; Swaen and Schumpitaz, 2008; Vlachos et al., 2009). Based on support from these previous CSR studies on the moderating effect of product quality, it is expected that in sectors that are known for their hedonic products, the CSR-loyalty relationship is weaker than in utilitarian product sectors:
H11: The positive effect of CSR on customer loyalty is stronger for utilitarian products
Product visibility to others
Product visibility refers to the extent to which products or services are noticed by others at the moment when these are consumed (Rust, Zeithaml and Lemon, 2000). Certain types of products, such as investment products (Hoffmann and Broekhuizen, 2010), have low visibility, because they are privately consumed. Although such visibility can be increased by means of WOM, according to Hoffmann and Broekhuizen (2010), the direct consumption of these products is not noticed by other consumers. Lemon, Rust and Zeithaml (2001) emphasise the importance of product visibility to other consumers in building brand equity. Particularly in the Western culture, the ownership of certain products is a key indicator of wealth and social differentiation. This is discussed by psychology-based theories, that explain how possessions are used by consumers as symbols of personal and social identity (Dittmar, 2008). More specifically, material possessions serve to maintain identity of the self, to enhance aspects of the self identity and act as symbols of status, wealth and group membership. Particularly the latter is applicable to this research, but the precise effect based on previous research is unpredictable.
H12: The positive effect of CSR on customer loyalty is weaker when product visibility
increases.
2.4 Mediating Effect of Customer Satisfaction
In general, customer satisfaction is found to be an important mediator in relationships with various customer intentions as the outcome (e.g. Cronin, Brady and Hult, 2000; Garbarino and Johnson, 1999). For instance, Szymanski and Henard (2001) prove the important role of customer satisfaction in a meta analysis of seventeen studies in affecting customers’ repurchase intent. In CSR literature, customer satisfaction is included as a mediating variable in a wide range research papers (e.g. Luo and Bhattacharya, 2006), as is the mediating effect of several satisfaction-like variables (brand evaluation by Klein and Dawar, 2004; customer commitment by Lacey and Kennett-Hensel, 2010). According to Bhattacharya et al. (2009), this mediating effect of satisfaction can be explained based on feelings of reciprocity and social identification with the company. First, feelings of reciprocity arise in the consumer because CSR implies that the company gives something back to the community. This increases customer satisfaction and subsequently activates a reciprocity norm in the form of rewarding the company by remaining loyal to it (De Wulf, Odekerken-Schröder and Iacobucci, 2001; Salmones et al., 2005). Second, social identity theory explains the effect of CSR on loyalty and the mediating effect of satisfaction: individuals wish to think of themselves in a positive way and when an organisation communicates an identity that is positive to consumers, they wish to identify themselves with that organisation (Tajfel 1978; Bhattacharya et al., 2009). Therefore, when a company engages in CSR activities, a customer of that company will identify more with it to enhance his/her self-image. The underlying mechanism signifies that CSR first increases satisfaction and, eventually, loyalty. Figure 1b presents satisfaction as mediator in the relationship between CSR and loyalty. The following hypothesis results:
H13: Customer satisfaction positively mediates the relationship between CSR and
customer loyalty.
2.5 Direct and Moderated Effects of Customer Satisfaction
satisfaction or a similar concept would therefore lead to omitted variable bias, which has the potential to inflate the effect of CSR on customer loyalty. Customer satisfaction is measured as ‘the consumer’s general level of satisfaction based on all experiences with the firms’ (Garbarino and Johnson, 1999, p. 71). It is an important construct in marketing research, which is indicated by the development of a popular satisfaction measure, the American Customer Satisfaction Index (ACSI) (Fornell et al., 1996). The two main consequences of customer satisfaction are the degree of customer complaints and customer loyalty, because, when customers are dissatisfied, they have two options: to either complain about this, or leave the company (Fornell et al., 1996; Fornell and Wernerfelt, 1987). In marketing literature, the direct effect of satisfaction on loyalty has received high support (Agustin and Signh, 2005; Fornell et al., 1996; Shenkar et al., 2003). According to Fornell et al. (1996), ‘loyalty is the ultimate dependent variable’ of customer satisfaction (p. 9). Therefore, if it can be shown that CSR has a significant positive effect on loyalty while the effect of satisfaction on loyalty is accounted for, this leads to even higher support for the CSR-loyalty relationship.
TABLE 2
Overview of Hypotheses and Sources for Interactions Satisfaction
Moderating Variables Hypotheses CSR Hypotheses Satisfaction Sources Interactions Satisfaction Research
H1 CSR has a positive effect on
customer loyalty (+)
H2 Gender Stronger for women in
comparison to men (+) comparison to men (+) Stronger for women in Homburg and Giering (2001); Melnyk, van Osselaer and Bijmolt (2009) H3 Age Stronger when the age of
consumers increases (+) Weaker when the age of consumers increases (-) Homburg and Giering (2001) Gilly and Zeithaml (1985);
H4 Income Stronger when income increases (+)
Weaker when income increases (-)
Cooil et al. (2007); Homburg and Giering (2001)
H5 Consumer confidence Stronger when consumer
confidence increases (+) Stronger when consumer confidence increases (+) Katona (1974); de Vries et al. (2011)
H6 Involvement Stronger when consumer
involvement increases (+) involvement increases (+) Stronger when consumer Bloemer and de Ruiter (1999); Goodman et al. (1995); Homburg and Giering (2001);
Oliver (2009) H7 Corporate ability Weaker when corporate ability
increases (-) Stronger when corporate ability increases (+) Luo and Bhattacharya (2006)
H8 Advertising Stronger when advertising
increases (+) Stronger when advertising increases (+) Crosby and Stephens (1987); Dick and Basu (1994
H9 Market leadership Stronger for market leaders in comparison to market followers
(+)
Weaker for market leaders in comparison to market followers
(-)
Anderson, Fornell and Lehmann (1994); Fornell (1992); Griffin and Hauser
(1993) H10 Competitive intensity/
dynamism intensity/dynamism increases (+) Stronger when competitive intensity/dynamism increases (+) Stronger when competitive Kumar et al. (2011)
H11 Product type H/U Stronger for utilitarian products in comparison to hedonic
products (-)
Stronger for hedonic products in comparison to utilitarian
products (+)
Babin and Griffin (1998)
H12 Product visibility Weaker when product visibility
increases (-) Weaker when product visibility increases (-) Fisher and Price (1992)
3. Methodology
In this chapter, the research design and data collection are discussed, followed by an explanation of the measures of each of the constructs. Subsequently, preliminary checks and controls are performed to examine the combination of items to variables, reliability analyses and common-method bias. After this, a description of the data set is provided. At the end of the chapter, the econometric models are specified for the direct and moderating effects in Figure 1a and for the mediating effect outlined in Figure 1b.
3.1 Research Design
Data was obtained from a combination of customer surveys, expert surveys and two measures (advertising and market leadership) were collected through other sources. The majority of consumer-level variables was gathered by means of web-based surveys distributed to Dutch consumers in 2010. Company-level information was obtained from a combination of the same customer surveys and from several databases that supplied 2010 data. Additional information was provided by experts (managers), who, after a pre-test was conducted, were asked to fill out a survey on sector characteristics. This data was obtained in 2011. Ultimately, the fact that variables have been measured by means of multiple data collection instruments decreases the risk of common method bias, and thus diminishes the chance of spurious effects.
3.2 Data Collection
balanced over the five age categories provided, which indicates a good representation of consumers from all ages. With respect to income, most observations are in the 30.000-60.000 category. In general, the sample is found to properly represent the Dutch population.
At the company level, corporate ability was measured in the consumer survey and later aggregated to the company. However, advertising and market leadership have been gathered from outside sources. Advertising data of companies has been obtained from Nielsen, a large market research company, which provided advertising expenditures of companies in million Euros. Market leadership was based on a performance ranking of companies within sectors. Relative performance was determined by means of company revenues and, when revenues were unavailable, comparable measures were used. Various sources were drawn upon to gather this data. First, revenues have been searched for in company statements, the Orbis database and by means of financial documents from the Dutch chamber of commerce. Subsequently, when data was missing, market shares were obtained for several sectors with help from the company Nielsen, or the ranking within sectors was based on another performance measure similar to revenues, such as number of website hits for the online booking sites. Finally, when available, this performance data was rechecked with various sources (such as newspaper articles), to diminish the bias of different reporting standards. A performance ranking was then made of all companies within each sector. Because the ranking could not be completed due to too much information that was still missing, a market leadership dummy variable was created, to indicate whether a company was a market leader vs. a market follower in each sector.
At the sector level, the expert questionnaires were sent to 1440 managers via email. Experts could select only those clusters of sectors they were most knowledgeable of and comfortable to answer questions about. Each cluster consisted of approximately three sectors, where the clusters were identified by a name. For example, the ‘Financial Group’, included questions on the general insurance, health insurance and banking sectors. A total of 244 questionnaires were sent back, yielding a response rate of 16.9%. After scrutinizing the answers to the questionnaire items, a total of 158 usable and complete surveys resulted, with a minimum of 15 responses in the special retail cluster (including electronics, furnishing and DIY sectors) and a maximum of 59 responses in the financial group cluster.
to the exclusion of another company: Blokker. This decision is caused by the fact that Blokker does not clearly belong to the department store sector, nor does it fit in any other sector. Furthermore, additional performance data and other information on Blokker could not be obtained due to a very non-transparent corporate policy. This has ultimately resulted in a sample of 93 companies and 8722 responses to be used for further analyses.
3.3 Measures
Based on the relationships outlined in Figure 1a in the previous chapter, several variables were measured as input for the analysis. These measures were developed based on previous academic research and are discussed below. An overview of all constructs, including how they were measured and coded, can be found in Table 8, at the end of this chapter.
3.3.1 Consumer-level measures
Loyalty to the company can be assessed by the number of customers that have recently left the company as a percentage of existing customers (i.e. churn), but this data is generally not collected by companies and therefore unavailable to research. Additionally, if companies do know these numbers, they are generally not willing to provide such sensitive material. Therefore, loyalty intentions are often used as a proxy for loyalty, allowing consumers to specify the proportion of products or services they expect to buy at a specific company in the future (Rust, Lemon and Zeithaml, 2004). Since it is a forward-looking measure, it also helps companies in determining their strategy for the future, and, because the same measure is adopted for consumers of all companies, it is particularly useful to this study in comparing loyalty across companies and sectors (de Vries et al., 2011). Based on Rust et al. (2004), loyalty intentions were measured by first asking respondents to imagine they would again purchase a product within the specific sector, and subsequently to indicate the probability of purchasing this product at the companies included in a sector. A list of companies was provided per sector and respondents could distribute 100 points to the companies to indicate the probability that they would repurchase a product at the respective companies.
Customer satisfaction was derived from two items, based on Fitszimmons (2000):
‘overall, how satisfied are you with this company?’ and ‘my choice for this company turned
out better than expected’. Respondents could indicate their answer on a 7-point Likert scale, for the first item ranging from 1 = very unsatisfied to 7 = very satisfied, while the second item ranged from 1 = totally disagree to 7 = totally agree.
The moderating variables at the individual respondent level were all measured by means of the web-based questionnaires. With respect to demographics, gender was measured by ‘what is your gender?’, Age was measured by ‘what is your age?’, where respondents could choose among six age range categories. Income was measured by ‘could you provide an
indication on your gross family income at a yearly basis?’, where respondents could choose
among four income range categories and one ‘I’d rather not say’ option.
The customer-level moderator consumer confidence was obtained from three items: ‘did the personal financial situation in your household deteriorate, stay the same, or improve over the last 12 months?’, ‘how do you think the personal financial situation in your
household will develop during the upcoming 12 months?’, and ‘how do you think the general
Dutch economy will develop during the upcoming 12 months?’. These items were derived from the Index of Consumer Sentiment (ICS) of the University of Michigan’s Survey Research Center (Curtin, 1982; Huth, Eppright and Taube, 1994). Respondents could indicate their answer on 7-point Likert scales, from 1 = will become worse via 4 = will remain the same to 7 = will become better.
Two questions were used as indicators for the final consumer-level moderator involvement: ‘to what extent are [sector]products important to you?’ and ‘to what extent are you interested in [sector]products?’ (Mittal, 1992; Zaichkowsky, 1985). 7-point Likert scales ranged from 1 = to a very low degree to 7 = to a very high degree.
The control variables, relationship duration and switching costs, were also measured by the customer survey. With respect to relationship duration, one question was asked to respondents: ‘how long have you been a customer of [this company]?’, where respondents could choose among six categories ranging in years (with a minimum of ‘shorter than 1 year’ and a maximum of ‘longer than 10 years’) and one ‘I don’t know’ option. Concerning switching costs, this was measured by: ‘it would take a lot of effort, in terms of both time and
money, to switch to another company [within the same sector]’ (Burnham, Frels and Mahajan,
3.3.2 Company-level measures
At the level of the company, three moderators are proposed in this research.
First, in previous research, corporate ability has been measured by means of several different items. Following the CSR study by Luo and Bhattacharya (2006), this study relates corporate ability to company innovativeness. It is believed that innovativeness most closely reflects the definition of corporate ability as ‘those associations related to the company's expertise in producing and delivering its outputs’ (Brown and Dacin, 1997, p. 68). In addition, the strength of company image is found to be highly correlated with innovativeness, which is the reason for also including a measure on this. Corporate ability was measured by means of two items in the web-based customer questionnaire: ‘[this company] has a strong brand’ and
‘[this company] has an innovative brand’ (Luo and Bhattacharya, 2006). 7-point Likert scales
ranged from 1 = totally disagree to 7 = totally agree. Responses were aggregated to the company level to obtain measures on the strength of the company image and company innovativeness.
The second company-level moderator, advertising, was collected as advertising expenditures in million Euros per company, by way of the procedure as explained in the section above.
Finally, market leadership data was based on a performance ranking in each sector from company revenues and similar measures. A distinction is made between a company as market leader (coded as 1) and the other companies as market followers (coded as 0) per sector.
3.3.3 Sector-level measures
Sector-level moderators have been determined by means of the questionnaires sent to experts. Responses from managers were aggregated to the sector to obtain sector-level measures.
First, competitive intensity was measured by means of two items: ‘could you give an indication of the competitive intensity in the following sectors?’ and ‘in the following sectors, the companies compete with each other for new customers and to keep current customers’ (Slater and Narver, 1994; Kumar et al., 2011). Market dynamism, analysed together with competitive intensity due to their similarity as two measures of environmental uncertainty, was measured by: ‘could you indicate to what degree the market in the following sectors can
be identified as dynamic?’. This was followed by an explanation of dynamism in terms of the
all questions on competitive intensity and market dynamism could be provided on 7-point Likert scales from 1 = to a very low degree to 7 = to a very high degree.
Second, with respect to the hedonic vs. utilitarian product type, respondents were first presented with an explanation on the characteristics of both hedonic and utilitarian products. Subsequently, the following question was asked: ‘based on the characteristics provided above, can you indicate to which category the main product/service offered by [this sector]
belongs?’ (Dhar and Wertenbroch, 2000; Strahilevitz and Myers, 1998). Respondents could
choose one of three response categories: 1 = mainly utilitarian, 2 = mainly hedonic, and 3 = both utilitarian and hedonic. However, instead of simply aggregating these responses to the sector level, the percentage of respondents that chose the specific categories were collected. This has resulted in three variables: hedonic product, utilitarian product and both, which each contain continuous data on the extent to which the different sector products are hedonic, or utilitarian, or both.
Finally, product visibility to others was represented by one item: ‘when the products/services from [this sector] are being used, this is noticed by people in the nearby
environment’ (Fisher and Price, 1992). A 7-point Likert scale included answers ranging from
1 = totally disagree to 7 = totally agree.
3.4 Preliminary Checks and Controls
To determine whether items may be combined to measure the variables in Figure 1a, correlation matrices are provided of the independent and moderating items at each level (see Tables 3, 5 and 6) and a factor analysis is performed at the consumer level (see Table 4). The correlation tables provide information on convergent validity and discriminant validity, although the latter will again be observed at the variable level, after the items of have been combined to variables. After the discussion on the combination of items, reliability analyses are performed and common method bias tests are discussed.
3.4.1 Combination of items at the consumer level
Demographic variables have been excluded from Table 3, because each of these is measured by only one item, which are clearly separate from the rest of the variables.
because correlations between the unrelated items are relatively low. Only the second measure of satisfaction correlates somewhat higher with both CSR items, but these correlations do not reach above .4, which is sufficiently low to not signify any multicollinearity problems. A rotated factor analysis of the items can be found in Table 4, which shows four clearly separate factors, supporting both high convergent and discriminant validity. Based on the correlations and factor analysis, all items are combined to form the respective variables.
TABLE 3
Correlation Matrix of Consumer-Level Items
1 CC refers to consumer confidence. ** p < .01.
* p < .05.
3.4.2 Combination of items at the company level
Table 5 includes the correlation coefficients of the items measured at the company level. The market leader variable is not included, because it is dummy coded. Corporate ability is measured by multiple items, while advertising has only one measure, which is also the reason why a factor analysis is not conducted for the variables measured at the company level. Table 5 shows that the two items on corporate ability correlate highly with each other, indicating high convergent validity, while both correlations with advertising are sufficiently low, which also signifies high discriminant validity.
TABLE 4
Factor Analysis Results at Consumer Level
Factors Items 1 2 3 4 CC_21 .848 -.021 .020 .030 CC_11 .748 .037 .031 -.010 CC_31 .731 .004 -.043 -.013 CSR_2 .006 .923 .117 .149 CSR_1 .018 .921 .108 .165 Involvement_2 .023 .096 .891 .071 Involvement_1 -.020 .113 .887 .074 Satisfaction_1 .005 .056 .049 .891 Satisfaction_2 -.002 .269 .102 .801
1 CC refers to consumer confidence.
TABLE 5
Correlation Matrix of Company-Level Items
** p < .01.
3.4.3 Combination of items at the sector level
Table 6 shows the correlations between items at the sector level. The two items of competitive intensity correlate sufficiently to combine them into one measure and, as expected, market dynamism is also highly related to competitive intensity. Market dynamism even correlates higher with the first measure of competitive intensity than the second measure of competitive intensity does. Because there is such high overlap between the two items of competitive intensity and the market dynamism, the three are combined in one variable. Furthermore, the three product type measures created to identify the degree to which a sector is characterised by utilitarian products, hedonic products or both, correlate very highly with each other and with product visibility. Because of these multicollinearity problems, the three product type measures are recoded into dummy variables. The utilitarian product category is used as reference category, resulting in a dummy for hedonic products (1 = hedonic product, 0 = other) and both utilitarian and hedonic products (1 = both, 0 = other).
1 2 3
1. Corporate ability_1 1.00
2. Corporate ability_2 .770** 1.00
TABLE 6
Correlation Matrix of Sector-Level Items
1 U&H product refers to the both utilitarian and hedonic product type. ** p < .01.
* p < .05.
3.4.4 Reliability analyses
To examine whether the items can be combined into variables based on the correlations and factor analysis results, reliability analyses are performed. At the consumer level, all proposed combinations have a Cronbach α higher than .6. Excluding the third measure of consumer confidence, because it correlates somewhat lower with the other two consumer confidence items, reduces Cronbach α to .647, while including it results in a Cronbach α of .668. Therefore, based on this and the fact that the three measures were part of one clearly discriminant factor, all three items are combined in one consumer confidence variable. At the company level, the two items of corporate ability have a Cronbach α of .870. Finally, the two sector items on competitive intensity are combined with market dynamism, because the Cronbach α is higher when the three items are combined compared to the combination of the two competitive intensity measures alone. This Cronbach α of competitive intensity/dynamism is above .6 as well.
3.4.5 Common-method bias
Several of the variables have been measured by means of a method other than questionnaires. However, because much information, including that on the dependent variable, has been
obtained from survey research, a common-method bias test is performed to examine whether significant (spurious) effects may exist only due to the use of a common data collection method. First, the Harman’s one-factor test (Podsakoff and Organ, 1986) poses that common-method bias arises when one general factor emerges after unrotated factor analysis, or one big factor results that accounts for the vast majority of covariance. This is not the case in this research, because multiple factors arise, with a balanced number of variables per factor. Second, partial correlation with a marker variable (Lindell and Whitney, 2001) could not be performed as common-method bias test because a marker variable is not present in the data set, but no different patterns for correlations among survey-measured variables and among/with those obtained from other sources are found (see Table 7), which provides sufficient evidence that common-method bias is not present in the data.
3.5 Description of Data
With respect to missing data, only CSR has eight missings and customer satisfaction has six. These represent random missing data and are therefore included in the analysis without recoding. However, the switching costs variable containts 480 missing values, all located in the entire gasoline sector. This is an industry that is characterised by low awareness on where consumers purchase, because consumers generally just buy their gasoline at the nearest gas station when they run out of gas. Because the analysis program uses listwise deletion, keeping these 480 values as missing would eliminate all responses in the gasoline industry. Therefore, it is decided to replace the missing values of switching costs in the gasoline industry by the mean value of switching costs.
Additionally, two variables contain categories that are recoded, due to interpretation difficulties. First, income contains the option where respondents can indicate ‘I’d rather not say’. Where the other four options represent increasing levels of income, this category cannot be meaningfully interpreted in an analysis. Also, because 30.6% of respondents chose this answer, it would lead to a great loss of data if the values would be coded as missings, since listwise deletion for missing values means that these cases are excluded. It is therefore decided to impute the mean value of income for ‘I’d rather not say’ . The same procedure is followed for the control variable relationship duration, which contains six options that are ascending in years, and one ‘I do not know’ category. 8.8% of respondents chose this answer, which is replaced by the mean value of relationship duration.
both utilitarian and hedonic product type variables are not included, because these are dummy coded. Table 7 indicates that, in general, discriminant validity is high, since all correlations do not succeed .4 and most are even much lower than this threshold. It is common to depict the correlations of the variables measured at the different levels, although this correlation table does not correctly take into account the variances at the specific levels in the data set (e.g. Zohar and Luria, 2005). The sample sizes for the company-level measures and those at the sector level are also different than those at the consumer level, something which is not considered here. Multilevel regressions in the next chapter will therefore properly include the hierarchical structure of the data.
3.6 Analysis and Model Specifications
