OPERATIONALISING CORPORATE SOCIAL RESPONSIBILITY (CSR) AND THE DEVELOPMENT DEBATE

Headnote

ABSTRACT

A considerable amount of attention has been paid to the construct of Corporate Social Responsibility (CSR) and yet research on the precise measurement of CSR has remained limited. Measures have been hampered by a lack of clarity in theoretical frameworks and empirical methods for the CSR construct. Given that the empirical study of CSR measurement is in an undeveloped state, this research describes efforts to justify and prove the relationship between measurement items and the construct. Based on a study among Malaysian stakeholders, this research conceptualises CSR as a formative construct consisting of eight dimensions: process, policy, values, environment, personal, profit, people and politics. The analyses reveal alternative approaches from a conceptual and methodological standpoint that makes clear the danger of misspecifying formative models as reflective, or vice versa. In this regard, it is proposed that the agenda and scope of CSR, as well as the measures used to implement it, are a manifestation of the formative construct that corporations have to operationalise in order to perform CSR better or more efficiently.

Keywords: Corporate Social Responsibility, formative constructs, Malaysia

INTRODUCTION

This research originated from an interest in how Corporate Social Responsibility (CSR) is measured in theory and in practice. There has been a resurgence of interest in the CSR construct among researchers and practitioners (Turker, 2009). From an empirical point of view, research on CSR has often involved a rather incomplete and simplistic methodology (Turker, 2009). For instance, Pederson (2010) identified a lack of consensus on important valid features for CSR research. Effective measurement is still considered to be the greatest hurdle for stakeholders (Dahlsrud, 2008). In addition to the lack of consensus on the dimensions of CSR, recent publications have challenged the common approaches to incorporating complex constructs (Galbreath, 2009) and identify the linkages between CSR and management characteristics (Pederson, 2010). The relevance of this discussion for empirical research needs to be investigated through a more systematic manner of studying CSR. Bollen (2002) notes that all measurement in the social sciences assumes effect indicators, and in structural equation modelling, every construct is assigned a set of indicators. However, in publications, the epistemic relationship between variables and indicators is often not considered. Latent variables may be associated with reflective or formative indicators. Most researchers assume a reflective relationship, meaning that the unobserved latent variable affects the indicators. In this case, all indicators "measure the same thing and should covary at a high level if they are good measures of the underlying variable" (Bagozzi, 1994, p. 331). If the latent construct is of its indicators, such as an index or ranking, it needs to be measured formatively. "Formative indicators give rise to the unobserved theoretical construct. In this case the empirical indicators produce or contribute to the construct" (Fornell & Bookstein, 1982, p. 8). As Hulland (1999) claims, the type of indicator specification used is very important from a conceptual and methodological standpoint. Diamantopoulos and Siguaw (2002, p. 11) emphasise that the "alternative approaches to deriving measures can produce substantially different operationalisation of the same construct".

The dangers of misspecifying formative models as reflective, or vice versa, are thereby made clear. Despite this interest and guidance, research on CSR measures more generally has remained limited due in part to the lack of consensus regarding what CSR really is. Does CSR have a framework or set of dimensions encapsulating its salient characteristics? There is a pressing need for better measurement of CSR, as the current measures appear to be inadequate. The motivation for writing about CSR and its measures in this study is not simply to draw attention to the shortcomings of other methods and concepts of CSR, but rather to provide an appropriate measure and begin to settle the measurement issues in CSR. With that aim, this research first developed dimensions indicating CSR as a multidimensional construct. Then, the study provided a formative measure to capture a multidimensional conceptualisation of CSR.

We structure the remainder of the article as follows. First, we consider background research on CSR measurement and problems of misspecification, developing hypotheses around possible conceptual dimensions of CSR and considering CSR as a formative construct. Second, we outline the methodology for our study, including data collection techniques. Third, we present analysis and findings. Finally, the discussion and conclusion outline our study's contribution and limitations and avenues for further research.

BACKGROUND OF THE STUDY

In this section, past research findings on CSR measurement are first reviewed. Then, the nature of multi-dimensional constructs is discussed.

CSR: How Is It Measured?

Discussion of CSR measurement evokes comment, criticism and argument. Foremost is the diversity of opinions concerning the character of CSR, as currently constructed, reflected in the competing philosophical and moral positions of the various commentators on the CSR debate (Epstein, 1977). CSR can have multidimensional constructs, which may include a wide range of business operational behaviours (internal and external processes) such as pollution control investment and environmental strategies, treatment of women and minorities, quality of products, customer needs and desires and philanthropic programmes within society (Wood, 1991) and industry performance (Waddock & Graves, 1994). In terms of CSR measures, past researchers have focused on forced-choice and Likert scale survey instruments (Aupperle, 1991). Financial reports, including return rate, the Fortune reputational scales and social responsibility indexes (Wolfe & Aupperle, 1991), social disclosures (Ullmann, 1985), pollution control investment (Shane & Spicer, 1983) and scale development (Kim & Kim, 2010) have also been used. Current CSR measures are single or even uni-dimensional and incomplete, and these may not adequately and completely reflect CSR (Brickson, 2007). Thus, these measures are difficult to apply consistently across the range of industries and corporations to be studied. Moreover, research on CSR is largely limited to developing countries and is not well recognised as a global or international concept (Freeman & Hasnaoui, 2010). Similarly, Matten and Moon (2008) show how recent CSR practices are moulded as "Americanisations" (p. 406). While CSR practices are becoming more universal with worldwide adoption, suggesting that a framework has broad applicability, discrepancies between definitions of CSR do occur between countries, and it is therefore important to address these concerns in developing countries as well. While a multidimensional construct offers the ability to increase granularity and features for dissimilar aspects of a construct, the number of measures necessary increases, as does the complexity of analysis. Consequently, it is vital for researchers to comprehend the diverse choices available when investigating and specifying a given construct in a research model. Failing to grasp the normative models and mind sets of business practitioners as regards CSR will make it difficult to understand and predict how firms should respond to societal demands (Pedersen, 2010).

The Problem with Misspecification

There is growing evidence that the misspecification of the construct (CSR) can create bias in the structural model. Evidence has shown that researchers with an understanding of formative constructs may decide to avoid their use in theoretical models. Rather than foregoing the use of formative constructs completely, researchers may also decide to simply model the construct as reflective because formative constructs have in the past been more difficult to employ when analysing data via covariance-based SEM approaches (Chin, 1998). However, this approach is problematic; researchers cannot just convert formative constructs or choose to measure the construct reflectively or vice versa.

Formative versus reflective

First, the formative construct is formed from the individual measurement items that are hypothesised to cause changes in the latent construct-which is usually conceptualised at a higher hierarchical level than the measurement items. The formative approach is constant, with the idea that the items are completely uncorrelated. Therefore, there is no need for uni dimensionality in this formative approach. Indeed, the reason one uses a formative approach is usually that the related construct is seen as comprising different dimensions and that different measurement items are required to tap into them. As a result, a high level of internal consistency for individual measurement items is not what one seeks in this case, and Cronbach's Alpha is not a useful estimate of reliability (Bollen & Lennox, 1991; Diamantopoulos & Winklhofer, 2001; Jarvis, MacKenzie, & Podsakoff, 2003). Thus, it can be noted that changes in the formative measures may cause changes in the construct, the content validity is paramount and internal consistency is irrelevant. Figure 1 illustrates the formative construct. Note in Figure 1 that the directionality of the arrows leading from the X's to the etas is the mark of the construct and is thought to be, and hence modelled as, formative. A way of reading this directionality is to think of the eta as "being caused by the indicators" in the case of the formative construct.

With this approach, one cannot leave out or eliminate any part of the object to perform a formative approach: 'dropping a measure from a formative -indicator model may omit a unique part of the conceptual domain and change the meaning of the variable, because the construct is a composite of all the indicators' (MacKenzie, Podsakoff, & Jarvis, 2005, p. 712). Turning to the reflective construct, the basic statement is that covariation among the measurement items is caused by variation in one underlying factor (the latent constmct). Consequently, each individual measurement item to be included in a measure (usually referred to as a multi-item scale) should sufficiently reflect the same latent construct, which means that the validity of the measure is not hypothesised to change much if a single item is removed (or added). In other words, it is understood that the indicators are unidimensional (Bollen & Lennox, 1991; Gerbing & Anderson, 1988; Jarvis et al., 2003).

Moreover, given that unidimensionality is confirmed, Cronbach's Alpha can be used to assess reliability in terms of internal consistency. It should be noted that Cronbach's Alpha does not provide information about unidimensionality; other means, normally confirmatory factor analysis, are needed for this consideration (Gerbing & Anderson, 1988). Thus, it can be noted that the changes in the construct create changes in the indicators. Internal consistency (i.e., reliability) is paramount for a reflective construct. Figure 2 illustrates the reflective construct. Note in Figure 2 that the Ys to the etas are the mark of the construct and are thought to be, and hence modelled as, reflective. A way of reading this directionality is to think of eta as 'causing the indicators' in the case of reflective

Reflective indicators are archetypal of classical test theory and factor analysis models; they are invoked in an attempt to account for observed variances or covariances (Jarvis et ah, 2003). Therefore, a reflective indicator is an explanation for observed variances or covariances, and reflective models minimise "the trace of the residual variances in the 'outer' (measurement) equations"(Fornell & Bookstein, 1982, p. 442). Moreover, the direction of causality is from the construction to the indicators, and changes in the underlying construct are hypothesized to cause changes in the indicators (Fornell & Bookstein, 1982; Bollen & Lennox, 1991).

According to Bollen and Lenox (1991), although reliability estimates (e.g., Cronbach's alpha) for the set of indicators will be lower if fewer indicators are included in the measurement model, the construct validity is unchanged when a single indicator is removed because all facets of a unidimensional construct should be adequately represented by the remaining indicators. As shown in the graphic view below, each indicator of a reflective construct is thus represented by its own equation. An example of the appropriate application of the reflective indicator model based on the graphic shown below is attitudes and purchase intention toward healthy products. Typically, attitudes are viewed as predispositions to react in a favourable or unfavourable manner toward an object and are generally measured on multi-item scales such as good-bad, like-dislike, and favourable-unfavourable.

However, purchase intentions are typically measured using subjective estimates of how likely-unlikely, probable-improbable, and/or possible-impossible future purchases are perceived to be (e.g., MacKenzie, Lutz, & Belch, 1986). Thus, the individual attitude toward purchasing healthy products is very subjective. The consumer might have in mind whether the healthy product may help to balance their diet. Therefore, in a reflective model, the latent variable influences the indicators, accounting for their intercorrelations. An important point to note here: reflective indicators of a principal factor latent construct should be internally consistent and because all of the measures are assumed to be equally valid indicators of the underlying construct, any two measures that are equally reliable are interchangeable.

Multi-dimensional constructs

A construct could be measured reflectively or formatively. Moreover, each dimension can be measured using formative or reflective indicators. The dimensions may be formative or reflective depending on the construct. As highlighted in the previous section, the reflective construct should be unidimensional. The measures are tightly centred on a concept. Multidimensional constructs are another concept that relates to formative constructs. Multidimensional constructs contain multiple dimensions and are grouped because there is some theoretical relationship between the various dimensions. These multiple dimensions 'are grouped under the same multidimensional construct because each dimension represents some portion of the overall latent construct' (Law & Wong, 1999, p. 144). Figure 3 illustrates the multidimensional constructs.

Figure 3 illustrates an example of a multidimensional construct that is comprised of three subconstructs (i.e., Y1, Y2 and Y3). In multidimensional constructs, the measurement items are intended to tap into the different subconstructs, and multicollinearity is protected by ensuring that the items do not tap into similar aspects. Thus, formative constructs are an example of multidimensional constructs because one measurement item is used for each dimension or subconstruct. However, Petter et al. (2007) have strongly argued that not all multidimensional constructs are formative. Similarly, MacKenzie et al. (2005) state that there is also potential for the construct to have some subconstructs measured using reflective items, while others are measured using formative items and/ or a combination of both formative and reflective paths between the construct and subconstructs. These authors also added that the choice of whether to model and analyse a construct as reflective, formative or multidimensional depends primarily on the construct under study and 'the generality or specificity of one's theoretical interest' (MacKenzie et ah, 2005, p. 713).

Moreover, it is important to note that if the main topic of study involved is a complex construct, it may be worth modelling the study as a multidimensional construct, which allows more thorough measurement and analysis. Further, developing a multidimensional construct that has a formative relationship between the construct and subconstruct should take place when multiple subconstructs and measurement items are needed to fully capture the entire domain of the construct (Petter et al., 2007). However, it is general practice among researchers to constrain the subconstruct items into a one-dimensional construct when measuring and analysing a multidimensional construct. Importantly, we must also note that evaluating the construct as a first-order unidimensional construct, together with all of the items from each subconstruct as a single reflective construct, produces a construct that is not unidimensional because the items comprising the constructs are, in fact, measuring different aspects of the construct. Given this concern, some researchers specify the construct as a first-order multidimensional construct in that aggregate measures for each of the subconstructs are specified as formative indicators. Again, these approaches to constrain a high-order multidimensional construct into a single construct can compromise validity. Therefore, these approaches may also lead to a measurement problem (Jarvis et al., 2003, Petter et al., 2007).

Given the above discussion, researchers should conduct this practice carefully because it can adversely impact the validity of measures. Researchers need to make a distinction between the orders of the construct, either first or second order, and its dimensionality. While a multidimensional construct offers the ability to increase granularity and features for dissimilar aspects of a construct, the number of measures necessary increases as does the complexity of analysis. Consequently, it is vital for researchers to comprehend the diverse choices available when investigating and specifying a given construct in a research model.

The decomposition of models could lead to serious misspecification problems (Jarvis, MacKenzie, & Podsakoff, 2003). Therefore, the decomposition must be appropriately modelled and the distinctions between measurement models should be carefully recognised beforehand. Consequently, misspecification has potentially had a number of detrimental effects on progress in the research field. Firstly, parsimony is avoided, although parsimonious models can provide abstractions that cause insightful explanations about complex phenomenon; secondly, the decomposed model may provide different theoretical implications when compared to the formative model; and finally, a decomposed model can result in atomistic fallacy (Diez-Roux, 2002). A few past studies, for example Jarvis et al. (2003) (see also the simulation to determine the ramifications of misspecifying formative constructs as reflective by MacKenzie, Podsakoff and Jarvis [2005]), have performed a simulation to resolve the ramifications of misspecifying formative constructs as reflective. Petter, Straub and Rai (2007) have tried to replicate and extend Jarvis et al.'s simulation because these authors wanted to examine the implication of the downward and upward bias in the parameter towards Type I and Type II error. To detect if Type 1 error can occur due to mismodelling, they performed an additional series of simulations on structural models that restricted a non-significant path. The unexpected finding from their series of simulations is that Type I error can occur regardless of whether the formative construct is specified correctly or not. From the results, they have concluded that 'when the formative construct was correctly specified and the path was statistically significant, the practical significance of the parameter estimate was minimal thus suggesting to the researcher that a problem may exist with the parameter estimate' (Petter et ah, 2007, p. 631).

Given this discussion, it is clear that researchers need to be aware of the danger of Type I and II errors that may exist in research studies. In the Type I error, researchers may build new theories and models based on prior research that finds support for a given relationship that does not actually exist. Consequently, this type of error may affect the research impact for both academics and practitioners because the misspecification may direct researchers to create unlike research models and generate different insights and implications than what reality actually implies. However, if a Type II error occurs, it may result in many of the relationships within the model being found to be non-significant. Thus, this type of valuable research may have difficulty getting published in highly ranked journals. In contrast, if no construct has been misspecified and a large number of hypotheses are significant, the chances of publication in atop tier journal is high (Petter et ah, 2007). Generally, determining whether CSR measures should assume reflective or formative measurement depends on four considerations, as follows: (1) the nature of the CSR construct; (2) the relationships among the observed CSR indicators; (3) the direction of causality between the CSR construct and the indicators; and (4) a theoretical judgment on CSR (Coltman, Devinney, Midgley, & Venaik, 2008). Failure to classify formed attributes correctly leads to an inappropriate structure for identifying components and the omission of crucial items (Rossiter, 2002). Thus, theoretical justification is needed to define the nature of the CSR construct, the direction of causality, and the items used to measure the constructs. Following theoretical justification, empirical justification is required, and a number of approaches can be taken: testing for indicator intercorrelation, the relationships of indicators with their antecedents and consequences, and measurement error and collinearity to detect the causal direction between constructs and their indicators, helping to justify the adoption of a formative approach to CSR measurement.

In sum, it is critical for researchers to note that these biases and errors affect the statistical significance of the estimates, thus increasing the danger from the misspecification of constructs in any research. For instance, it is likely that numerous studies have been rejected in the review process because reviewers insisted on high internal consistency, reliabilities and required a principal factor model to fit the data. Consequently, constructs that are truly formative in nature may have received less attention in the literature and/or they may have been more likely to have been modelled as scale scores without taking measurement model relationships into account (Jarvis et ak, 2003). As cited in Jarvis et ak, (2003, p. 216), "an equally large number of studies have been published with severely restricted construct domains due to the same reviewer bias". In addition, the construct domain restriction indisputably contributes to the inconsistency in findings across studies and may partially account for the generally low proportion of variance explained in many criterion variables (Peterson, Albaum, & Beltramini, 1985). Jarvis et ak, (2003) also noted another implication of measurement error: a substantial proportion of the empirical results in the literature may be potentially misleading.

Having outlined the dangers of misspecification of the CSR construct, we now develop our hypotheses.

HYPOTHESIS DEVELOPMENT

This section is divided in two parts. First, a dimension that indicates CSR as a multidimensional construct is considered. Then, a formative measure to capture a multidimensional conceptualisation of CSR is discussed.

CSR Dimensions

CSR is a theory-based formative construct (Gjølberg, 2009; Poolthong & Mandhachitara, 2009), so the issues of construct validity and reliability that typically apply in a reflective construct are not as relevant. Creating a formative measure of CSR suggests that changes in the survey items affect the CSR scales, rather than the other way around. Some authors have also suggested there is no test of reliability for formative measures (Coltman et al., 2008), but the indicators should have the same directional relationship with the latent construct and the indicators should be checked for collinearity. To identify the dimensionality of the scale, this study reports the reliability data for the total set of observations because the number of coding categories can affect the estimate of reliability and, indirectly, can lower the confidence limit (Perreault, Jr. & Leigh, 1989). This study is concerned with the quality of the raw responses as well as with the coding scales. The first step of index construction or scale development requires the specification of the construct domain by providing a conceptual definition of the construct (Diamantopoulos & Winklhofer, 2001). This present study agrees with the existing literature that the scope of CSR is unclear, is subject to diverse interpretation and has no consensus as to its definition (see Amaeshi & Adi, 2007; McWilliams, Siegel, & Wright, 2006). Therefore, regarding the specification of the construct domain, the first step is to provide a clear definition of CSR. The qualitative study serves to understand the construct definition of CSR as well as the existing CSR dimensions identified in the literature and from the practitioner.

Content analysis

The researcher has delineated what is included in the definition of CSR and what is supposed to be excluded. Themes were constructed and developed to apply to the assessment of CSR definitions in the literature and interview transcripts. The primary task in developing the themes was to determine words that might be regarded to be components of a CSR definition. It was imperative that the themes reflected in the CSR definitions be regarded as sufficiently important and relevant to be disclosed in the literature. Consistent with the approach taken in previous literature, reference was first made to the themes employed in prior studies (Carroll, 1979; Romlah, Takiah, & Jusoh, 2003 [as cited in Janggu, Joseph & Madi, 2007]; Zain & Janggu, 2006; Dahlsrud, 2008). The following steps were taken to construct the theme.

To ensure the relevance and applicability of the CSR definition to this study, it was decided that the preliminary list should resemble the CSR definition from its original text. Reference was also made to the studies employed by Carroll (1979) and Dahlsrud (2008), as they represent categorisations of CSR in the literature reviewed at the beginning of this research. It is noted that in all of these studies, the categorisations constructed were based on the authors' assumptions.

1. Reference was then made to the academic personnel to check the theme construct.

2. The list of themes compiled from the literature was then combined with themes discovered from the interviews.

3. Expert opinions were then sought (counter-checking with interviewees and academic personnel)

4. The list was then tested for purity using a statistical test.

The quantitative study serves to confirm the proposed definition developed in this study. The results indicate that CSR is a contested concept and shows multi-dimensional constructs. Therefore, the following sections refer to these parts as first-order dimensions of the second-order CSR construct.

Eight dimensions, derived from prior literature, have been proposed and defined as follows:

1. Process: long-term activities or business between and among stakeholders

2. Policy: compliance to regulations that extend beyond legal and ethical conduct

3. Values: the core beliefs that help a firm to differentiate its reputation and identity and that guide communication efforts

4. Environment: effective management and protection of natural resources while balancing these with stakeholders' activities (i.e., ensuring that these do no harm to the Earth)

5. Personal: individual character as represented by all stakeholders, subject to individual perception and expectation

6. Profit: the monetary gain of the firm while fulfilling their economic obligation and as a return on CSR investment

7. People: the objects of a firm's responsibility and commitment (e.g., shareholders, employees, customers, suppliers, governments, non-governmental organisations and communities)

8. Politics: manipulation by particular organisations or individuals' for their own agenda and interests

This conceptualisation of CSR as a formative second-order construct with eight dimensions provides a basis for the next steps in the CSR operationalisation process. Compounding these types of formative constructs are the links that have been made among the dimensions.

CSR As A Formative Construct

As discussed earlier, when a measurement scale is proposed for a construct, it is necessary to consider whether, from the conceptual point of view, it is a formative model or a reflective one (Diamantopoulos & Winklhofer, 2001). With respect to the direction of causality, it appears clear that dimensions such as philanthropy are not manifestations of CSR but defining characteristics of it. Furthermore, changes in these indicators cause changes in the construct, and not vice-versa. With respect to the formative criterion, the dimension of CSR has its own antecedents and consequences. CSR is therefore a formative construct, but the dimensions of CSR are in turn measured by their own scales. That is to say, the eight dimensions proposed in this study have their own indicators. For this reason, Jarvis et al. (2003) consider CSR to be a second order formative scale: the dimensions of CSR are formative but the indicators of these dimensions are reflective. This aspect is important when proposing and testing the model. Accordingly, the following hypotheses (H1a, H1b, H1c, H1d, H1e, H1f, H1g, and H1h) are advanced.

H1: Corporate Social Responsibility is a multidimensional formative construct comprised of eight dimensions:

a) process; b) policy; c) values; d) environment; e) personal; f) profit; g) people; h) politics.

METHODOLOGY

This section first introduces the steps that were adopted to develop the measurement instmments. Then, the data gathering process is presented.

Instrument Development

The first step of index construction or scale development requires the specification of the construct domain by providing a conceptual definition of the construct (Diamantopoulos & Winklhofer, 2001). Therefore, regarding the specification of the construct domain, the first step is to provide a clear definition of CSR. Our qualitative study offers an understanding of the construct definition of CSR as well as the existing CSR dimensions identified in the literature and from practitioners. The second step of index constmction, indicator specification, uses the construct definition to identify those indicators that capture distinct facets of the construct. In contrast to reflective constmcts, formative constructs require a census of indicators (Bollen & Lennox, 1991), and the indicators must cover the scope of the construct (Diamantopoulos & Winklhofer, 2001). Table 1 shows the themes across the content analysis. The quantitative research serves to confirm the proposed dimensions identified in this study. In this study, the researcher had to think carefully about whether the CSR measures cause (or define) the construct (i.e., formative relationship) or represent a reflective relationship. The study choice of a formative form of epistemic relationship has been justified clearly and consistently.

Procedure

A coding scheme was developed using the emergent coding technique. The process of emergent coding is described in detail by Fierros, Gulek and Wheelock (1997). The steps as applied in the current study are detailed as follow:

1. A random sample of CSR definitions was selected from each document.

2. An initial list of 65 definitions was used to develop the coding scheme. The researcher independently extracted the dominant themes that emerged from the sample of documents.

3. The researcher also acquired four independent participants1 for this study to check and reach a consensus as to the major themes. From these themes, an explicit coding scheme was developed. Specifically, it was decided that the coding scheme would be dichotomous in nature and would retain ten major themes, each with varying number of thematic elements.

4. The rater agreement (reliability) of the newly devised coding scheme was assessed using inter-rater agreement measures as reported in Table 2.

5. The researcher then proceeded to independently code the remaining 42 CSR definitions (total = 107).

6. Finally, the rater agreement (reliability) of the coded items was assessed again using inter-rater agreement measures (the researcher and colleague as the inter-raters) as also reported in Table 2.

Computing the rater agreement across major themes

The purpose of establishing the reliability of the coding scheme used in this study was to provide a preliminary estimate of inter-judge reliability for diagnostic purposes. Alternatively, the coding scheme provided a summary index to reflect the quality of the final coded data in this study, alongside assessing the extent to which the raters agreed when attempting to apply the themes to various CSR definitions. Indirectly, the reliability is important in the developed coding scheme to exclude the elements of bias on the agreement of major themes. In this study, the rater agreement of the major themes in the developed coding scheme was measured by specifically analysing the:

1. inter-rater agreement amongst the independent participants (n = 2), and

2. inter-rater agreement amongst the researcher and colleague (n = 2).

There are number of ways to quantitatively report the agreement ratings for inter-rater and intra-rater agreement. In attempting to assess the reliability of a coding scheme, the simplest measure of rater agreement would be the overall percent agreement as discussed earlier in the Research Methodology chapter. Cohen's Kappa (Crocker & Algina, 1986) was used to adjust for the inflated coefficient that would result from using a simple percent agreement. Crocker and Algina (1986, p. 201) highlighted that "a kappa value of 0.2 can be interpreted to mean that 20% of the total possible increase over chance consistency was observed for the decisions". In relation to this interpretation, Kvalseth (1991) recommends that a kappa coefficient of 0.61 represents reasonably good overall agreement.

At this stage, the researcher looked to an outside audience (independent raters) to further validate the coding scheme. As mentioned before, independent raters were used to guard against any shared meaning the researcher may have generated amongst others. Krippendorff (1980) suggests that this process would yield a highly reliable or unreliable coding scheme between the researcher and the outside world. The inter-rater agreement amongst the independent participants was conducted using seven randomly selected definitions for each document. Both independent participants were given detailed written instructions and were asked to separately code the CSR using the major themes and elements developed through the emergent coding. When the two independent raters' coding was compared, it was found that the overall inter rater agreement for the major themes was .69 (n = 280)2. Specifically, the rater agreement for the major themes of CSR from books, journals, articles and interview transcripts was .74, .73, .53 and .76, respectively.

Inter-rater agreement between the researcher and a colleague was also evaluated using seven randomly selected CSR definitions for each document. Both also separately coded CSR using the major themes developed through the emergent coding. When the researcher compared the findings, it was found that their overall inter-rater agreement for the major themes was .80 (n = 280). Specifically, the reported agreement (kappa) for the major themes of CSR from books, journals, articles and interview transcripts was .84, .83, .66 and .87. Table 2 illustrates the various agreement ratings reported for the CSR coding scheme.

Data Collection

The current research is conducted in Malaysia. Malaysia is chosen because the awareness of CSR amongst Malaysian stakeholders is increasing (Turker, 2009). This research uses the multi-stage sampling method and a non-probability sampling approach. The research instruments were developed through several methods and stages - a qualitative study and quantitative study. The first stage involves an examination of the literature in the study of construct measurement. This review shows that study of the CSR construct is valuable because it helps to achieve a better understanding of the correct measures for CSR. The literature search and interviews are used to fulfil the exploratory part of the study, to define the construct and also to generate the vocabulary and items/scales to develop the CSR measures. The great challenge this research faced was how to develop themes using content analysis from interviews and literature search data. Specifically, for the quantitative study, an online survey was conducted using different industry settings in Malaysia. In each industry, a contact person was asked to disseminate the survey link among colleagues at all levels of the organisations and in as many different departments as possible. In this study, closed-ended questions together with 5-point Likert scales are adopted through the questionnaire, and PLS graph 3.0 was used to test measurement model. Table 3 illustrates the CSR items.

ANALYSIS AND RESULT

A summary of the correlations of the latent variables is presented in Table 4. Inter-construct correlations were below the common cut-off threshold of 0.9. A content analysis was used to develop the construct and the CSR dimensions of process, policy, values, environment, profit, personal, people and politics.

The analysis used PLS to test for hypotheses H1a, H1b, H1c, H1d, H1e, H1f, H1g and H1h. Testing these hypotheses showed a significant effect for policy (? = 0.188, p < 0.05), values (? = 0.215, p < 0.05), profit (? = 0.279, p < 0.01), and politics (? = 0.178, p < 0.05), but no significant effect for process (t = 0.6711, n.s.), environment (t = 0.1416, n.s.), personal (t = 0.2261, n.s.), and people (t = 1.3016, n.s). These dimensions do not support H1a, H1d, H1e or H1g. Therefore, H1 is partially supported. Given the results for Hypothesis 1, it was necessary to operationalise the developed construct (i.e., CSR). As in any empirical research, the results of the present study cannot confirm the construct without taking into account the external construct validity (Jarvis et ah, 2003). Despite the fact that four of eight indicators were not significantly related to the latent variable (i.e., process, environment, personal and people), this study did not drop these indicators because they contribute conceptually to the CSR construct. Although statistical considerations should be taken into account, conceptual reasoning holds more influence than statistical results when deciding whether to drop formative indicators (Petter et ah, 2007). Other interesting findings show that the personal dimension represented a negative weight of sign (see Table 5). The best option may be to identify this dimension as a reflective indicator instead of a formative indicator when it is conceptually appropriate (Jarvis et ah, 2003). Furthermore, formative indicators of the same construct "can have positive, negative, or no correlation" with one another (Bollen & Lennox, 1991, p. 307). As specified in this model, the CSR multidimensional construct prevails: (a) the eight dimensions are viewed as separate facets that are defining characteristics of the CSR construct, (b) changes in the eight dimensions are expected to cause changes in the CSR multidimensional construct, (c) changes in the CSR multidimensional construct do not cause changes in the eight dimensions, (d) the eight dimensions do not share a common theme, (e) eliminating a dimension (e.g., profit) may alter the conceptual domain of the CSR multidimensional construct, (f) changes in one of the dimensions (e.g., process) is not necessarily expected to be associated with changes in all of the other seven dimensions and (g) the eight dimensions are not expected to have the same antecedents and consequences.

DISCUSSION AND CONCLUSION

This section begins with a discussion of the study findings along with their managerial implications. The primary limitations of the analysis and directions for future research are then proposed.

This study defines the contested CSR construct. The research indicates that CSR is a second-order construct that consists of eight dimensions: process, policy, values, environment, profit, personal, people and politics. In addition, the indicators of each of these dimensions encompass CSR items that comprise the relevant dimension.

The first dimension, process, pertains to measuring long-term activities or business between and among stakeholders. However, the weakness of the current CSR dimension is that there is a greater focus of the criteria on goal-oriented outcomes and the 'macro' level of all large corporations (Tuzzolino & Armandi, 1981). In light of this weakness, this present research views CSR as a 'process'. When managers are more conscious of the social consequences of their decisions, CSR changes from being a goal-oriented to an institutionalised process. Moreover, current management practices, particularly in the field of CSR, are based on outputs rather than processes, which creates difficulties in understanding the concept. One of the primary reasons the CSR framework can be ineffective in practice is that it does not take into account a coherent portfolio of CSR business practices covering all stakeholders (Lamberti & Lettieri, 2009). In fact, top managers have become increasingly aware that CSR is not only an alternative means to increase profitability in the short term but is a pillar representing the company's system of values and its mission (van Marrewijk, 2004; Willard, 2002). This study holds that firms that consider this dimension for their CSR initiatives are more likely to develop trust in CSR as a means to increase business profitability and stability.

The second dimension, which relates to policy, relies on compliance to regulation that extends beyond legal and ethical conduct. The policy dimension is observed as being a key issue in determining CSR dimensionality. This dimension provokes some unexpected transformations in terms of business values and processes because the common response of some corporations to such allegations is either the development of a CSR policy or a reference to and the potential reform of such a policy. For example, it has recently been argued by some ethical campaigners that the CSR policies of corporations such as Gap and Nike are not as prominent as they ought to be and, as a result, allegations of negligence have been made against them (Rosselson, 2009). These allegations could potentially elicit a number of different reactions (Rosselson, 2009). Some might argue that it is the role of the state to regulate corporations more strictly, whereas some would say that corporations are best left to regulate themselves and to be relied upon to make ethical decisions and comply with policy (Whitehouse, 2006). This paper has found that CSR policy affects people's opportunities in life, regardless of whether CSR comes from efforts made by the state or by corporations. The current circumstances demand a different type of response and a response can be related to the personal and political ideals of the dimension.

The third dimension, values, relates to determining the core beliefs that help a corporation to differentiate its reputation and identity and that guide communication efforts. This dimension is considered to be 'invisible' and begins with issues of image and reputation. In other words, stakeholders appear to see corporate reputation as the key driver to promoting and embedding CSR internally in organisations, and they view corporate image and reputation as leverage to force corporate change towards implementing CSR. Moreover, the values dimension is involved if the goods, services, or activities satisfy a need or provide benefits that contribute positively to the quality of life, knowledge, or safety of firms' stakeholders (Haksever, Chajanti, & Cook, 2004). Thus, this study proposes that the values dimension is a key driver in framing and embedding CSR in corporate strategy and elicits social values that will enhance prominent and favourable impacts on competitiveness in a core business. Moreover, how a company values its corporate social responsibility department and projects will reflect its world view and corporate culture. By making CSR policies part of 'corporate value', the notion of what it means to be ethical is made part of the commercial value of a product. Therefore, immeasurable resources are devoted to the publication of glossy brochures and advanced websites for the CSR division of many corporations. Hence, CSR is something that can be bought and sold like any other product (Nan & Heo, 2007).

The fourth dimension, environment, relates to the effective management and protection of natural resources while balancing this with stakeholders' activities and interests. Nevertheless, there are corporations that misuse natural resources. The main culprits in this respect are oil, mining, logging and mineral exploitation corporations, with conflict arising over hydroelectric dams and bio-fuel plantations as well as coal, copper, gold and bauxite mines (Vidal, 2009). The central arguments of this sort of allegation are that corporations are involved in activities that will lead to the eventual annihilation of local peoples as well environmental deprivation and loss of biodiversity. An example in this regard includes the Trafigura scandal in the UK, in which the corporation tried to conceal its responsibility for the illegal dumping of waste and the consequent ill health of those who came into contact with it in the Ivory Coast as well as trying to suppress reporting of this by the press (The Guardian, 2009). In this respect, the government's hydro-energy project in Malaysia, the Bakun dam project, has environmentalists up in arms, questioning the need for the dams and the planned development of the Sarawak State (The Star, 23 July 2008). Although government officials claim the dams will be necessary to meet energy demands, the project has been highly controversial after environmentalists suggested that Sarawak's national park may be threatened. In many such cases, it is alleged that the state is also complicit in these abuses, supporting the activities of the corporation rather than those of its citizens (Vidal, 2009). From these current circumstances, there is evidence that the stakeholders (e.g., environmentalists) will contribute positive or negative feedback on environmental issues, regardless of whether the project is handled by a corporation or by government entities.

The fifth dimension, which relates to profit, refers to firms making an investment in CSR and consequently seeking monetary gain while fulfilling their economic obligations. This dimension is considered somewhat clichéd, an opinion that is consistent with many other CSR studies (e.g., Carroll & Buchholtz, 2008; Vilano va, Lozano, & Arenas, 2009). In this regard, an important implication of this study of CSR from the stakeholders' perspective is the understanding that CSR is often associated with monetary gain or profit for the initiator. From the firm's perspective, however, CSR is often a defensive strategy or is believed to be conducted at the expense of profit (e.g., Milton Friedman). Hence, there has been a long, sustained search for the "business case for CSR". Given this historical perspective, it is important for a study briefly to address the question of CSR as a monetary expense as well (particularly when the study identified some of the stakeholders as shareholders). Profit is an evaluative condition that requires some clarification.

Accordingly, from the standpoint of Malaysian stakeholders regarding the assessment of the subject, some of the stakeholders perceived CSR as taking place at the expense of profit, but many have agreed that CSR is conducted merely to make more profit. Many firms claim that they really 'practise what they preach'; however, some of them have also profited from their CSR initiatives. Consequently, the criticism regarding profit in relation to CSR will never end, as it is ambiguous in nature. Perhaps the profit debate is a good illustration of how a developing country adopting CSR (in the legally mandated accountability sense) can be contrasted with developed countries' defensive claims of CSR practice (in the voluntary sector and 'beyond the law'). In addition, it should be possible to assert that CSR is used aggressively in Malaysia. This usage also reflects the complexity inherent in a contested concept of CSR due to the relationship between business and society existing in different social contexts.

The sixth dimension, personal, pertains to measuring an individual character, subject to individual perception and expectation. Particularly in recent times, there has been a perception that large global corporations are more socially responsible than small companies. Employees from global corporations are expected to represent top wage earners and average wage earners. One study indicates that, in the United States at least, in the period from 2002 to 2006, the top 1 of wage earners gained 75% of all income growth, a trend that is attributed to an explosion in top wages and salaries (Saez, 2008, p. 2-3). However, are these employees happy to receive huge salaries if their work life is not balanced with their personal well-being? Another common perception of corporations is that they engage in CSR for the sake of their shareholders' benefit, growth, and wealth. Moreover, Korten (2001) argues that corporations are responsible for causing much of the world's poverty and inequality. However, there is an argument that CSR has a considerable effect on people's lives in developing countries such as Malaysia. Recently, a popular sentiment has emerged that CSR also has had a direct impact upon people's lives in the industrialised, developed countries of Europe and North America in terms of income and wealth from shares, pensions and savings, as well as in terms of job security and inevitable redundancy in difficult economic circumstances. The recent financial crises have emphasised the idea that in many Western societies, corporations wield a great deal of power over people's lives; this power is increasingly coming under widespread scrutiny. From the above discussion, the salient fact is that everyone sees and responds to CSR in a different way. Therefore, stakeholders may perceive CSR from the standpoint of their own particular interests. For example, as mentioned earlier, firms may think that their employees will be satisfied and that employees' welfare has been taken care if the firm offers them a high salary. In contrast, the employee may well feel differently. Employees might not be at all satisfied if their working lives and personal lives are not balanced, even if they are receiving a very high salary.

The seventh dimension, people, refers to the objects of a firm's responsibility and commitment (e.g., shareholders, employees, customers, suppliers, governments, non-governmental organisations and communities). Other studies have also recognised that CSR can be decomposed into people components (Johnson & Greening, 1999) and stakeholder management issues (Hillman & Keim, 2001). In relation to this observation, the Ministry of Women, Family and Community Development of Malaysia, aims to recognise companies that have made a difference to the communities in which they operate through their CSR programmes (New Straits Times, 15 December 2010). In 2010, the Sunway Group received the award in the field of Education, together with Malakoff Corporation Berhad (Environment), CIMB Group (Culture and Heritage, and Community and Social Welfare), Saito College Sdn Bhd (Small Company CSR), L'Oreal Malaysia Sdn Bhd (Empowerment of Women) and Telekom Malaysia (Workplace Practices), while Media Prima Berhad received the Media Coverage Special Award for their CSR initiatives (New Straits Times, 15 December 2010). These awards do not just benefit the companies involved but also the community. This recognition serves both as an example and an inspiration for other companies to follow in these company's footsteps and help transform the nation's lives.

However, the focus on CSR may vary a great deal depending on the company's character. For example, the Body Shop Malaysia represents itself as a beauty and skincare company with a conscience and has hoped, through a campaign, to create widespread awareness of violence against women and communicate the idea of friendship as a means of assisting victims of domestic violence. Since 2000, it has campaigned actively with the Women's Aid Organisation (WAO), a non-profit-making organisation, and their campaign, which they called Break the Silence on Domestic Violence: Talk to a Friend, has focused on the often hushed-up issue of domestic violence. For a contrasting example, Sime Darby and Kuala Lumpur Kepong Berhad (KLK) employs 2,933 women, most of whom come from disadvantaged backgrounds and hence are not highly educated. As part of its general philosophy of caring, the company provides semi-detached houses with free water and electricity to all its employees. Housing benefits for employees are the same, irrespective of gender. Women employees are also paid the same salary as men, and enjoy more benefits, especially with regard to their health and general well-being. KLK provides regular health check-ups for its women employees, including pap smears and breast examinations. In addition, lactating women are exempted from certain duties, such as spraying chemicals. Nevertheless, a common allegation made against corporations in regards to this dimension is that of the abuse of human rights. Other common allegations that have been made in this regard concern the use of child labour, sexual abuse, and the denial of freedom of speech to corporate employees in many places (Human Rights Watch, 2007). In summary, if corporations want to be profitable in the long run, they should consider the social welfare of internal and external stakeholders, for instance the employees and surrounding community, in their plans.

The eighth dimension, politics, relates to determining the existence of manipulation by certain organisations or individuals' for their own agenda and interests. An element of this dimension is that corporations are able to potentially take advantage of current situations; for instance, the phenomenon of globalisation. As such, for example, corporations can exert their rights to extract resources and demand people's land by claiming it for the use of infrastructure development (e.g., building roads or railways), for which they pay low compensation to local people. As a result of these actions, corporations are making a great deal of money. Furthermore, it is also alleged that many corporations are able to take advantage of globalisation by paying employees in developing countries far less than they would be paid in developed countries. However, there are also corporations that have used a CSR agenda for their own benefit and yet at the same time still consider the welfare of the community. In this regard, Malaysian corporations like the Sunway Group, Malakoff Corporation Berhad, CIMB Group, Saito College Sdn Bhd, L'Oreal Malaysia Sdn Bhd, Telekom Malaysia and Media Prima Berhad have incorporated their CSR efforts into their company strategies (New Straits Times, 1 December 2010). Further, corporations can gain advantages from particular technical strategies such as exploiting tax loopholes, transfer pricing and tax havens for their commercial gain. In relation to these types of activities, going beyond Carroll (1979), this paper shows that the political dimension necessarily enhances social welfare despite its pros and cons because it all depends on stakeholders' perceptions. Thus, this situation reflects back to the personal dimension.

As this paper has emphasised, current CSR research using a formative construction needs to be addressed as a technical specification question to improve specification and usage of the CSR construct. The results have several implications for managers, including guidelines for implementing CSR; for example, the importance of CSR dimensions in terms of CSR success. However, such actions may be poorly implemented simply because firms do not consider the distinction between CSR dimensions. Generally, the findings suggest that when assessing CSR, firms should not employ general measures of CSR, but should ensure that all aspects of CSR's specific dimensions are evaluated. Indeed, there may be a need to use all eight CSR dimensions, as proposed in this study, because the study results could influence managerial decisions in at least two areas:

a) Operationalising formative measures of CSR

A simple, direct measure is inadequate; managers who employ direct CSR measures are unlikely to capture the conceptual richness of this multidimensional construct. The proposed higher-order, formative operationalisation of CSR implies that such a measure must contain several CSR dimensions, as their omission prevents a comprehensive understanding of the construct. The failure to specify a measurement model properly can bias estimates of the structural relationships between constructs, and misspecifications can lead to poor or incorrect decision making. For example, a well-intentioned manager of food and beverages might decide to invest in expensive modern technology and infrastructure to improve CSR, thus perceiving an effect on the quality of the product (e.g., healthy food). On the basis of the strength of the stmctural relationships, the manager makes certain assumptions about the expected impact of shareholders' value perceptions, shareholders' future behaviour, and, ultimately, the firm's return on investment. However, such a decision may be inappropriate if the manager's measurement instmment over-emphasises the relative contribution of CSR (i.e., expensive modern technology and infrastructure) to enhance the quality and safety of a product by ignoring other aspects of CSR (e.g., profit and policy, as the budget to improve the quality of the range of healthy foods should not neglect the company's income and policy). The allocation of resources may also vary according to the conceptualisations of CSR. A formative model enables a manager to determine which CSR dimensions are the most influential in forming stakeholders' relationships. In the preceding example, the food and beverages manager could decide to allocate budgets to smooth the process of producing healthy food. The manager could train workers in favourable environmental practices rather than investing in expensive infrastructure to more cost-effectively influence consumers' product equity value perceptions. That is, the proposed model indicates the relative importance of each of the formative measures, such that managers do not jeopardise their CSR efforts by focusing on relatively unimportant dimensions of CSR. In the end, the use of poor CSR measures can result in poor and unbalanced managerial investment decisions.

b) Performance of the company

Stakeholders' perceptions of CSR depend significantly on CSR efforts and implementation; therefore, the eight proposed dimensions must be an integral part of any CSR strategy. In particular, the formative measurement model decomposes process, environment, people, profit, policy, personal, values and politics dimensions into CSR elemental parts. It is argued that by improving on one or more of these factors, managers can affect CSR performance. The model also demonstrates that across contexts, formative CSR measures provide the strongest drivers of CSR value, supporting the previous literature that highlights these dimensions as essential pillars of the CSR value creation process (Green & Peloza, 2011; Jo & Haqoto, 2011). Therefore, all interaction with stakeholders should emphasise these eight formative dimensions. Managers must focus on delivering reliability and consistency to achieve perceptions of high CSR value. Again taking the example of the food and beverages company, the reliability and consistency in providing a healthy and safe product are critical to the consumers' perceptions of value. The application of a standard policy developed by the International Organization for Standardization addressing food safety, for example ISO 22000, can have a significant impact on consumers' value perceptions. Managers therefore should invest in building positive standard policies in food management systems, understand what their products represent to consumers, and consistently reinforce this image at every contact point with stakeholders.

However, managers should also take note of the importance of company profit by incorporating other dimensions of CSR into their measures of company performance and CSR image. The literature reveals the importance of profit-making to a company's survival, but the findings of this study also emphasise the relevance of other dimensions in the context of CSR formative measures. For example, the political dimension is particularly salient in aspects that are generally 'invisible', such as contest participation. Although these efforts show some evidence of the hidden agenda of the programme, they are important alternatives for indirectly boosting sales. Managers need to be aware of the varying effects of the antecedent and cause constructs on CSR. Therefore, managers must recognise that stakeholders attach varying levels of importance to what they perceive to be CSR and therefore understand what balance of dimensions to emphasize, how formative measures affect one another, and where to place them for the firm to have superior performance. In particular, the findings strongly suggest that the mere broad implementation of CSR will not necessarily lead to the desired effect; it may even have a negative effect.

The fact that the existing measurements of corporate social responsibility were developed most likely represents the first and foremost limitation of the study. The researcher had to select the journals beforehand using a key word search. These caveats call for systematic qualitative inquiries examining how stakeholders define CSR. Future research could further examine the extent to which this study acknowledged dimensions of CSR. In particular, more inquiry is needed to validate whether CSR is a formative construct with other related variables, and the CSR scale should be further tested by examining the stability of the eight developed CSR dimensions. Given that future research should consider validating this CSR scale, a number of variables must be tested in structural models. Hopefully, in future, researchers will take more care in making sure that the measurement models used match their conceptualisation. In conclusion, this study led to the development of a practitioner-based model of a CSR multidimensional formative construct that in some aspects differs from the existing measurement model of CSR. Delineating the correct measurement in detail indicates some compliance and a greater understanding of social responsibility towards stakeholders. The research also perhaps indicates a gradual evolution away from focusing solely on Carroll's (1991) seminal inclusion of economic, legal, ethical, and philanthropic components. With respect to normative and instrumental stances toward the stakeholder, critically, the findings recognise that an operationalisation of CSR should be formed as a multidimensional formative construct.

Footnote

NOTES

1. Four independent participants volunteered to be the raters. They are also the participants for the interviews.

2. Note that the η = 280 refers to the number of categories coded. Twenty-eight definitions were coded, each of which contained 10 potential themes: 10*28 = 280.

References

REFERENCES

Amaeshi, K. M. & Adi, B. (2007). Reconstructing the corporate social responsibility construct in Utlish. Business Ethics: A European Review, 16(1), 3-18.

Aupperle, K. E. (1991). The use of forced-choice survey procedures in assessing corporate social orientation In. J. E. Post (ed.), Research incorporate social performance and policy (Vol. 12, pp. 269-279). Greenwich, CT: JAI Press.

Bagozzi, R. P. (1994). Structural equation models in marketing research: Basic principles. In R. Bagozzi (ed.), Principles of marketing research (pp. 317-385). Oxford: Blackwell Publishing.

Bollen, K. (2002). Latent variables in psychology and the social sciences. Annual Review Psychology, 53, 605-634.

Bollen, K., & Lennox, R. (1991). Conventional wisdom on measurement: A structural equation perspective. Psychological Bulletin, 110(2), 305-314.

Brickson, S. L. (2007). Organizational identity orientation: The genesis of the role of the firm and distinct forms of social value. Academy of Management Review, 32(3), 864-888.

Carroll, A. B. & Buchholtz, A. K. (2008). Business & society: Ethics and stakeholder management (7th Ed.). Mason, OH: South-Western Cengage Learning.

Carroll, A. B. (1979). A three-dimensional conceptual model of corporate performance. Academy of Management Review, 4, 497-505.

Carroll, A. B. (1991). The pyramid of CSR: Toward the moral management of organizational stakeholders. Business Horizons, 34, 39-48.

Chin, W. W. (1998). Issues and opinion on structural equation modeling. MIS Quarterly, 22(1), 7-16.

Coltman, T., Devinney, T., Midgley, D., & Venaik, S. (2008). Formative versus reflective measurement models: Two applications of formative measurement. Journal of Business Research, 61(12), 1250-1262.

Crocker, L., & Algina, J. (1986). Introduction to classical and modern test theory. Orlando, FL: Harcourt Brace Jovanovich.

Dahlsrud, A. (2008). How corporate social responsibility is defined: An analysis of 37 definitions. Corporate Social Responsibility and Environmental Management, 15, 1-13.

Diamantopoulos, A. & Siguaw, J. A. (2002). Formative vs. reflective indicators in measure development: Does the choice of indicators matter? (The Center For Hospitality Research Working Papers), Ithaca.

Diamantopoulos, A., & Winklhofer, H. M. (2001). Index construction with formative indicators: An alternative to scale development. Journal of Marketing Research, 38(2), 269-277.

Diez-Roux, A. V. (2002). A glossary for multilevel analysis. Journal of Epidemiology and Community Health, 56, 588-594.

Epstein, E. M. (1977). The social role of business enterprise in Britain: An American perspective: Part IF. Journal of Management Studies, 14(3), 281-316.

Fierros, E., Gulek, C., & Wheelock, A. (1997). Using multiple methods of assessment to promote district level reflection about instructional improvement. Paper presented at the meeting of the Annual Symposium of the American Educational Research Association, March, Chicago, IL.

Fomell, C., & Bookstein, F. L. (1982). Two structural equation models: LISREL and PLS applied to consumer exit-voice theory. Journal of Marketing Research, 19(4), 440M52.

Freeman, I. & Hasnaoui, A. (2010). The meaning of Corporate Social Responsibility: The vision of four nations. Journal of Business Ethics. doi: 10.1007/s10551-010-0688-6.

Galbreath, J. (2009). Building corporate social responsibility into strategy. European Business Review, 21(2), 109-127.

Gerbing, D. W. & Anderson, J. C. (1988). An updated paradigm for scale development incorporating unidimensionality and its assessment. Journal of Marketing Research, 25 (May), 186-192.

Gjølberg. (2009). The origin of corporate social responsibility: Global forces or national legacies?. Socio Economic Review, 7(4), 605-637.

Green, T., & Peloza, J. (2011). How does corporate social responsibility create value for consumers?. Journal of Consumer Marketing, 28(1), 48-56.

Haksever, C., Chajanti R. & Cook, R. G. (2004). A model of value creation: Strategic view. Journal of Business Ethics, 49, 291-305.

Hillman, A. & Keim, G. (2001). Shareholder value, stakeholder management, and social issues: What's the bottom line? Strategic Management Journal, 22, 125. Retrieved 15 November 2010, from http://www.anugerahcsrmalaysia.org/2010/ 12/08/firms-urged-to-heed-staff-community-well-being/

Hulland, J. (1999). Use of partial least squares (PLS) in strategic management research: A review of four recent studies. Strategic Management Journal, 20(2), 195-204.

Human Rights Watch (2007). Human Rights Watch - excellent reports on human rights situations in different countries and also on various issue areas. Retrieved 10 November 2010, from http://www.hrw.org/

Janggu, T., Joseph, C., & Madi, N. (2007). The current state of Corporate Social Responsibility among industrial companies in Malaysia. Responsibility Journal, 3(3), 9-18.

Jarvis, C. B., MacKenzie, S. B., & Podsakoff, P. M. (2003). A critical review of construct indicators and measurement model misspecification in marketing and consumer research. Journal of Consumer Research, 30(2), 199-218.

Jo, H., & Haqoto, M. A. (2011). Corporate governance and firm value: The impact of Corporate Social Responsibility. Journal of Business Ethics, 103, 351-383. doi: 10.1007/s10551-011-0869-y

Johnson, R. A., & Greening, D. W. (1999). The effects of corporate governance and institutional ownership types on corporate social performance. Academy of Management Journal, 42, 564-576.

Kim, Y., & Kim, S. (2010). The influence of cultural values on perceptions of corporate social responsibility: Application of Hofstede's dimensions to Korean public relations practitioners. Journal of Business Ethics, 91(4), 485-500.

Korten, D. (2001). When corporations rule the world (2nd Ed.). San Francisco, CA: Berrett-Koehler Publishers.

Krippendorff, K. (1980). Content analysis: An introduction to its methodology. Newbury Park, CA: Sage.

Kvalseth, D. (1991). The relative useful information measure: Come comments. Information Sciences, 56, 35-38.

Lamberti, L. & Lettieri, E. (2009). CSR practices and corporate strategy: Evidence from a longitudinal case study. Journal of Business Ethics, 87, 153-168.

Law, K. & Wong, C. (1999). Multidimensional constructs in structural equation analysis: An illustration using the job perception and job satisfaction constmcts. Journal of Management, 25(2), 143-160.

MacKenzie, S. B., Lutz, R. J. & Belch, G. E. (1986). The role of attitude toward the ad as a mediator of advertising effectiveness: A test of competing explanations. Journal of Marketing Research, 23(May), 130-143.

MacKenzie, S. B. & Podsakoff, P. M. & Jarvis, C. B. (2005). The problem of measurement model specification in behavioural and organizational research and some recommended solutions. Journal Applied Psychology, 90, 710-730.

Matten, D. & Moon, J. (2008). Implicit and explicit CSR: A conceptual framework for a comparative understanding of corporate social responsibility. Academy of Management Review, 33(2), 404-424.

McWilliams, A., Siegel, D. S., & Wright, P. M. (2006). Corporate social responsibility: Strategic implications. Journal of Management Studies, 43(1), 1-18.

Nan, X. & Heo, K. (2007). Consumer responses to corporate social responsibility (CSR) initiatives: Examining the role of brand-cause fit in cause-related marketing. Journal of Advertising, 36(2), 63-74.

New Straits Times (1 December 2010). Malaysia's per capita income to more than double by 2020. Retrieved 15 December 2010, from http://www.nst.com.my/nst/ articles/Malaysia_spercapitaincometomorethandoubleby2020/Article/

New Straits Times (15 December 2010). Firms urged to heed staff, community wellbeing. Retrieved 15 December 2010, from http://www.anugerahcsrmalaysia.org/ 2010/12/08/firms-urged-to-heed-staff-community-well-being/

Pedersen, E. R. (2010). Modelling CSR: How managers understand the responsibilities of business towards society. Journal of Business Ethics, 9E 155-166.

Peneault, W. D" Jr., & Leigh, L. E. (1989). Reliability of nominal data based on qualitative judgments. Journal of Marketing Science, 2<5(May), 135-148.

Peterson, R. A., Albaum, G., & Beltramini, R. F. (1985). A meta analysis of effect sizes in consumer behavior experiments. Journal of Consumer Research, 12(June), 97-103.

Petter, S., Straub, D. & Rai, A. (2007). Specifying formative constructs in information systems research. MIS Quarterly, 31(4), 623-656.

Poolthong, Y., & Mandhachitara, R. (2009). Customer expectations of CSR, perceived service quality and brand effect in Thai retail banking. The International Journal of Bank Marketing, 27(6), 408-427.

Romlah, J., Takiah, M. L, & Jusoh, M. (2003). An investigation of environmental disclosure in Malaysia. International Journal of Business and Society, 3(2), 55- 68.

Rosselson, R. (2009). Inertia means we're failing to boycott the least ethical clothing companies, The Guardian. 12 March 2009. Retrieved 18 November 2010, from http ://www. guardian, co .uk/enviromnent/ethicalhvingblog/2009/mar/12/boycott-unethical-fashion

Rossiter, J. R. (2002). The C-OAR-SE procedure for scale development in marketing. International Journal of Research in Marketing, 19(4), 305.

Saez, E. (2008). Striking it richer: The evolution of top incomes in the United States (Update using 2006 preliminary estimates). Retrieved 8 November 2010, from http://topforeignstocks.com/wp/wp-content/uploads/2009/05/saez-ustopincomes-2006prel.pdf

Shane, P. & Spicer, B. (1983). Market response to environmental information produced outside the firm. The Accounting Review, July, 521-538.

The Guardian (2009). Trafigura and The Probo Koala. Retrieved 1 November 2010, from http://www.guardian.co.uk/world/trafigura-probo-koala

The Star (23 July 2008). Dam Worry. Retrieved 1 November 2010, from http://whatrainforest.com/2008/07/23/dam-worry-the-star-23-july-2008/

Turker, D. (2009). Measuring corporate social responsibility: A scale development study. Journal of Business Ethics, 85(4), 411-427.

Tuzzolino, F., & Armandi, B. R. (1981), A need hierarchy framework for assessing corporate social responsibility. The Academy of Management Review 16(1), 21- 28.

Ullmann, A. A. (1985). Data in search of a theory: A critical examination of the relationships among social performance, social disclosure, and economic performance ofU.S. firms. Academy of Management Review, 10, 540-557.

Van Marrewijk, M. (2004). Concepts and definitions of CSR and corporate sustainability: Between agency and communion. Journal of Business Ethics, 44, 95-105.

Vidal, J. (2009). Food land grab puts world's poor at risk. The Guardian Weekly. 10 July 2009, p. 12.

Vilanova, M., Lozano, J. M., & Arenas, D. (2009). Exploring the nature of the relationship between CSR and competitiveness. Journal of Business Ethics, 87, 57-69.

Waddock, S. A., & Graves, S. B. (1994). Industry performance and investment in R&D and capital goods. Journal of High-Technology Management Research, 5, 1-17.

Whitehouse, L. (2006). Corporate social responsibility: Views from the frontline. Journal of Business Ethics, 63(3), 279-296.

Willard, B. (2002). The sustainability advantage: Seven business case benefits of a triple bottom line. British Columbia, Canada: New Society Publishers.

Wolfe, R., & Aupperle, K. E. (1991). Introduction. In J. E. Post (ed.). Research incorporate social performance and policy (Vol. 12, pp. 265-268). Greenwich, CT: JAI Press.

Wood, D. J. (1991). Corporate social performance revisited. Academy of Management Review, 16, 691-718.

Zain, M. M. & Janggu, T. (2006). Corporate social disclosure (CSD) of construction companies in Malaysia. Malaysian Accounting Review, 5(1), 85-114.