Seeing the forest, beyond the trees: dimensionality of context specific Organizational Citizenship Behaviour in a Sri Lankan context

This article is published under the Creative Commons CC-BY-ND License (http://creativecommons.org/licenses/ by-nd/4.0/).This license permits use, distribution and reproduction, commercial and non-commercial, provided that the original work is properly cited and is not changed anyway. Abstract: Factor analyses of existing measures of Organizational Citizenship Behaviour (OCB) administered across different contexts and country settings demonstrate cross-setting consistency and support a common dimensionality for OCB irrespective of context. Challenging the use of generic models of OCB are concerns over the local relevance of such generic models and the substantial variation in OCB dimensionality in different country settings found through research using inductive qualitative methods. Even though these findings give conflicting understandings of the context specificity of OCB, a review of literature found no studies that subjected qualitatively derived dimensions to factor analysis. This research uses factor analyses to test a qualitatively derived, context-specific OCB model to explore the context specificity of OCB. University academic staff in Sri Lanka participated in two studies. Findings support a generally accepted two-factor model and not the context specific model. The findings suggest that even when high contextual variations in constructs seem evident, conceptual similarity at a more general level may be present. Past research exploring OCB conceptualisations have used either qualitative or quantitative methods. While qualitative methods indicate complex and contextually specific dimensions, quantitative methods support simpler models with more general applicability. The findings suggest that even when the model and items are localised, factor analyses support a general OCB model.


INTRODUCTION
Organizational Citizenship Behaviours (OCB) are behaviours that support the social and psychological environment in which task performance takes place (Organ, 1997). OCBs have received much attention within human resource management because they are associated with organizational effectiveness, in terms of indicators ranging from higher customer satisfaction to greater profitability (e.g., Podsakoff et al., 2009;Sun et al., 2007). Because they have substantial bearing on employee performance ratings (Podsakoff et al., 2009), OCBs are central to human resource management.
Evidence of the significance of OCB is demonstrated in research conducted in many settings, including in Sri Lanka. Among school teachers in Sri Lanka, perceptions of justice and teacher empowerment were found to be a predictor of organizational citizenship behavior (Anuja, 2016;Rauf, 2014;Rauf, 2015a, Rauf, 2015b. OCB directed toward other employees was found to predict performance and OCB mediates the relationship between perceived organizational support and performance quality among shop floor employees at ISO 9001:2000 certified manufacturing firms in Sri Lanka (Wickramasinghe & Perera, 2014). OCB also predicted work alienation (Rauf, 2015b). Another study June 2019 Sri Lanka Journal of Social Sciences 42 (1) conducted in Sri Lanka found that OCB is positively related to work-family conflict but that the effect was moderated by work-family conflict efficacy (Rauf, 2013). Among a sample of 224 Sri Lankan employees supported that organizational justice is related to organizational citizenship behavior, and employee work alienation is demonstrated to be a mediator in this relationship.
As OCBs are described as highly grounded in the context in which they occur (Organ, 1998;Organ et al., 2006), understanding how they vary across cultural contexts is particularly important. However, the literature on OCB across regions and cultures show varying degrees of consistency in the dimensionality of the construct. In this paper, the researchers explore the context specificity of OCB by subjecting a highly contextually grounded set of OCB items, designed to capture contextually grounded OCB dimensions, to analytical methods that generally demonstrate high consistency in dimensions. By doing so, the extent to which variability in dimensionality is associated with methodology is explored.
Although much of the initial work on OCB was conducted in North America (Podsakoff et al., 2000), a growing body of literature is based in other regions (a Google Scholar search for 'organizational citizenship behaviour' and 'cross cultural' from 2013-2016 resulted in over three times as many hits as it did for the period between 2007-2009). Much of this research simply borrows early conceptualisations of OCB, such as Organ (1998), and uses translated versions of existing scales such as Podsakoff et al. (1990). Other research uses existing scales, but attempts to establish measurement equivalence in the new context generally through factor analysis and other methods (e.g.; Takeuchi et al., 2015;Zhao et al., 2012). Both these sets of studies demonstrate consistency in OCB dimensions across settings. A final set of studies have attempted to ground their research in the new context and to develop conceptualisations endemic to local settings (e.g., Farh et al., 2002;Rauf & Kumar, 2015). These latter studies report a range of OCB dimensions and suggest a large variation across settings. The dimensions developed through this final set of studies seem to, however, have not been subjected to model testing using factor analytical methods.

The purpose and scope of the study
The present study explores the factor structure of an OCB measure for Sri Lankan university academic staff derived using inductive methods and designed to capture dimensions that were specific to that context to ascertain whether these dimensions are robust when subjected to factor analysis. The study attempts to address how research generated from two different perspectives may provide different perspectives of a single construct and how these findings may be interpreted. This research is not meant to validate a scale and therefore, does not address scale equivalence and other concerns (Hui & Triandis, 1985) that should be addressed in a validation study. This study focuses on classical conceptions of OCB (e.g., Katz & Kahn, 1966;Organ, 1988), on which later extensions of the construct are derived (e.g., Boiral, 2009;Nohe & Michaelis, 2016).

ETIC AND EMIC APPROACHES TO ORGANIZATIONAL BEHAVIOUR
An on-going debate on the nature of psychological phenomena and the extent to which constructs remain comparable across cultural settings becomes significant as organizations expand geographically. Berry (1969), warns against research carried out assuming that constructs remain largely transferable across cultures, referring to such research as using an imposed etic approach. The extent to which the locality might affect the nature and importance of a construct and how the context may influence the nomenological network associated with it remain unaddressed in research that unquestioningly transplants theories and tools to new contexts.
Even within research that does not simply ignore the possibility of differences between cultural contexts, the understanding of what such differences mean with respect to conceptualisation and methodology vary. Within a cross cultural paradigm, emphasis is placed on understanding and theorising differences between cultures, and, as a result, dimensions have been identified along which cultures differ, such as Hofstede's (1980) value orientations (see also Norenzayan & Heine, 2005, for a review). While this research may acknowledge the role of cultural differences, if measures developed elsewhere are used with little attention to their relevance in the new setting, the research would still be based on an imposed etic approach. To be truly effective, an etic approach, which seeks generalities or universals across cultures, would require establishing equivalence in the constructs used to compare groups.
Research emanating from cultural or indigenous psychological frameworks (Ratner, 2008) emphasises developing contextualised and historicised understandings of phenomena, with comparisons across cultures, at the most, a secondary concern (Ratner, 2008). Much of this research would follow an emic approach and put greater emphasis on locally derived conceptualisations of constructs. However, irrespective Sri Lanka Journal of Social Sciences 42 (1) June 2019 of paradigm (cross-cultural, cultural or indigenous), ultimately this research reflects endeavours to develop comprehensive understandings of phenomena even though they differ in the process of arriving at such understandings. Cross-cultural methods would use crosscultural comparisons to identify differences in phenomena across contexts; thereby ultimately allowing for theories that would be able to explain these differences (Ho & Wu, 2001). Cultural and indigenous paradigms would begin by exploring the distinct nature of the phenomena within a context to later develop a more comprehensive understanding, which is perhaps more nuanced than those developed through cross-cultural methodologies (Chakkarath, 2012).
As these approaches, founded on different epistemological traditions, use different processes to achieve similar objectives, the methods used also differ. Cross-cultural research adopts a positivist epistemological position and tends to use quantitative indicators of difference and similarity. Statistical equivalence is frequently used to establish similarities in conceptualisations and measures. The other approaches, in contrast, are likely to use qualitative methods to identify the nature of phenomena in a particular context. Because their approaches to research differ, it is likely that the nature of the findings from the different approaches may yield results that are different in scope and make the ability to draw on both bodies of literature to arrive at a comprehensive understanding of the world challenging. This paper uses Organizational Citizenship Behaviour (OCB) as a focal construct to explore this proposition.

OCB DIMENSIONALITY
The literature on OCB has its roots in North America and includes a number of conceptualisations of OCB (see Podsakoff et al., 2000), which are generally implicitly framed as universal even though such generalisations are problematic (see Podsakoff et al., 2000). Two conceptualisations stand out in the literature. The first, developed by Organ (1988), is significant as it is the first to have been introduced, and therefore, the basis on which many later models were built. Williams & Anderson's (1991) two-factor solution is significant for its parsimony and popularity. Organ's (1988) original model, identifies five dimensions of OCB, in which employees engage at their discretion: altruism, behaviour targeted to helping a specific other with an organizationally relevant problem; conscientiousness, behaviour that goes beyond minimum organizational role requirements related to aspects such as attendance, obeying rules and regulations, and taking breaks; sportsmanship, willingness to tolerate less than ideal circumstances without complaining; courtesy: behaviours aimed at preventing work related problems with others from occurring; civic virtue, behaviour that indicates an employee's participation, involvement and concerns about the company. Podsakoff et al. (1990) developed a widely used measure based on the model by generating items that fit definitions of the five dimensions through subject matter experts. They then used a Q-sort method and factor analysis to validate the dimensions.

CONCEPTUALISATIONS OF OCB IN THE CULTURAL AND CROSS-CULTURAL LITERATURE
The literature exploring the dimensionality of OCB across regions addresses the extent to which OCB conceptualisations are similar or different across cultures. Two predominant themes seem to characterise this literature. The first focuses on the extent to which perceptions of OCB and in-role behaviour differ across cultures. Demarcating the scope of one's job does not seem to depend solely on the content of one's job description but varies as a function of other factors such as context (Graen, 1976). Of the latter, the sociocultural setting is important to consider. Generally, this research indicates that Confucian Asian settings (e.g., China, Japan, Singapore), which constitute a substantial proportion of the non-Anglo populations studied, have broader conceptions of their job role than Anglo settings (e.g., United States, Australia, England), meaning that some of what Anglo participants may view as OCB would be viewed as task performance in Confucian Asian settings (Lam et al., 1999).

June 2019
Sri Lanka Journal of Social Sciences 42 (1) The second theme, which is of more relevance to the present research, examines the extent to which OCB dimensions are universal, or etic in conception. Some of these studies, which are designed to establish measurement equivalence across settings, indicate a common structure across populations (e.g., Lam et al., 1999;Takeuchi et al., 2015). Many of these studies, which use factor analytic methods, indicate that conceptualisations of OCB transcend countries and even continents (even though these studies hypothesise somewhat different dimensional structures from each other). Other studies, which are rather limited, are attempts to inductively identify OCB dimensionality (Farh et al., 1997;Farh et al., 2002;Kumar & Bakhshi, 2009;Rauf & Kumar, 2015;Yutaka & Atsuko, 2011).
Studies that attempt to use inductive methods to determine dimensionality, again fall into two categories. Some studies determine dimensionality through factor analysis of items derived inductively, while others use qualitative methods of grouping to do so. To illustrate the former type of research, Farh et al. (1997), in Taiwan, initially developed OCB items through a sample of MBA students and then used exploratory factor analysis to identify five dimensions. Of these dimensions, three were described as etic (universal) as they parallel dimensions identified in Organ (1988): identification with company (similar to civic virtue), altruism and conscientiousness. The additional two, interpersonal harmony; actions aimed at facilitating and preserving harmonious relations, and protecting company resources; actions that save company resources, were described as emic (culture specific). This conceptualisation of OCB and the associated measurement tool has been the most extensively used of those derived from Asian contexts (e.g., Lam et al., 2009;Zhao et al., 2012). However, Zhao et al. (2012) provide evidence through a series of studies, to suggest that the two culture specific dimensions identified by Farh et al. (1997) are simply the positive end of deviant behaviours and should not be classified at OCB. Similarly, in India, Kumar & Bakhshi (2009) generated items through an interviewing process. Principal components analysis was used to identify dimensions. Five dimensions were identified, similar to those that had already been described in the literature: conscientiousness, helping co-workers, group activity participation, sportsmanship and courtesy.
Other studies, in contrast do not use factor analysis at all, and have both generated items inductively and then again use inductive methods to identify dimensions. Farh et al. (2004) examined the dimensions of OCB using a Hong Kong based sample. They used a purely qualitative classification system and arrived at 10 dimensions of OCB, of which five had already been identified [cf. taking initiative (similar to conscientiousness), helping co-workers (similar to altruism or helping), voice (similar to voice in the extant research but broader), participation in group activities (similar to civic virtue) and promoting company image (similar to loyalty)]. They reported the additional dimensions: self-learning, which refers to improving one's own knowledge or working skills; social welfare participation, which refers to employees' participation in activities of public welfare or community service; keeping work place clean, which are discretionary behaviour of employees to keep the workplace clean; interpersonal harmony, which refers to employee actions aimed at facilitating and preserving harmonious relations in the workplace; protecting company resources, or efforts to save company resources, use personal resources to aid the company and protect the company from disasters in the workplace. These latter five dimensions were reported as emic in nature.
A study conducted in Japan (Yutaka & Atsuko, 2011) yielded ten dimensions through a qualitatively sorting of items. Seven of the identified dimensions were aligned to those already found in studies conducted in the Western contexts (altruism, civic virtue, supporting students, courtesy, conscientiousness, sportsmanship and selfdevelopment) and three were identified as specific to Japan (maintaining harmonious relationships, mutual understanding, and maintaining a clean workspace).
In Sri Lanka, Rauf & Kumar (2015) identified eight dimensions through content analytic methods. They are: conscientiousness, going beyond the minimum role requirements of the organization with respect to attendance, obeying rules and regulations, and dedication at work; altruism, willingness to help another person with his or her work or non-work related problem; civic virtue, acts of creativity and innovation to improve the organizational performance; defending, acts that create goodwill towards the organization by defending it against threats; self-training, behaviours designed to improve personal knowledge, skills, and abilities; sportsmanship, willingness to tolerate minor inconveniences, courtesy, actions aimed at preserving harmony and preventing problems with others; conservation of the organizations' property, actions that preserve property (see Rauf & Kumar, 2015).
Thus, it is evident that the findings are confounded by the methods used. The use of techniques to establish equivalence seems to demonstrate consistency across settings. Even factor analyses of inductively generated items seem to yield a limited number of dimensions, which show substantial consistency across settings. In contrast, qualitative coding seems to yield a larger range of dimensions, which are sensitive to a greater extent than the previously described methods, to identify culture specific dimensions.

THE EFFECTS ON METHODOLOGY ON OCB DIMENSIONALITY
The methods used by researchers attempting to demonstrate equivalence are likely to show consistency in conceptualisations. On the other hand, researchers attempting to identify a culture specific conceptualisation are likely to identify these specificities. The authors could not find any studies that attempted to validate qualitatively derived dimensions using factor analysis. Review of the literature indicated that, while scales developed using the former procedures are frequently adopted by others, those that use qualitative classifications are not. Perhaps these latter dimensions do not translate well into the generally established methods of demonstrating equivalence, i.e., through factor analysis. Perhaps factor analysis does not yield the same structures that qualitative coding would, and the former may yield a more parsimonious and generalised conception that glosses over the nuances that qualitative approaches yield. On the other hand, perhaps factor analytically derived dimensions replicate across studies using factor analysis for validation purposes. However, these two perspectives provide valuable but different approaches to determine the content domains of constructs with etic approaches using theoretically derived conceptionalisations and emic approaches using the data itself to derive the content domain of OCB. Thus, both perspectives provide a different 'truth' regarding OCB, which in the present literature is not easily integrated.
Two studies were conducted. Study 1 used Exploratory Factor Analysis (EFA) and Study 2 used both exploratory and Confirmatory Factor Analysis (CFA) to explore the dimensionality of OCB in Sri Lanka among academic staff members. A pool of items generated in Rauf & Kumar (2015) was subjected to this process. The researchers hypothesised that factor analysis would yield a small set of factors that are consistent with existing conceptualisations of OCB. Both studies were based on samples of permanent (tenured) academic staff members from different universities in Sri Lanka but used slightly different procedures for data collection. Academic staff was chosen because of the high role discretion inherent in academic positions (Shazia & Munazza, 2011) should result in a large variation in behaviour (Mischel, 1977). The questionnaires were administered in both Sinhala and Tamil languages.

Study 1 Methods
To measure each of the eight OCB dimensions reported in Rauf & Kumar (2015), three to six items were identified. The items were based on the responses generated in Rauf & Kumar (2015), where participants had been asked to provide instances of OCB, based on its definition, which were then content analysed and sorted to develop the final eight dimensions. The highly contextualised nature of the original items were retained to maintain specificity at the item level and to ensure consistency across the research.
Using this process, 38 items were created. All items were originally in English and translated into Sinhala and Tamil via a double blind back-translation technique. Each item was rated on a 7-point scale ranging from strongly disagree to strongly agree. In addition, information of the respondents' university and faculty, age, gender and years of service were also gathered.
First, the questionnaire was administered to a sample of 40 academics and minor modifications were made to ten items. The finalised measure was administered to a sample of 210 academicians representing different disciplines from two national universities. A stratified sampling technique, with strata based on university, faculty, gender, age and positions, was used. The sample consisted of 145 men and 65 women. The respondents' age averaged 36 years and ranged from 26 to 58 years. They had an average of 6.8 years on the job. Data were collected through two modalities. One of the authors visited the universities and distributed hard copies of the questionnaire to participants. Later, questionnaires were sent through email for participants to return electronically. A cover letter accompanied each survey stating that participation was voluntary and anonymous in nature. Contact information was provided to participants in case of questions or concerns. The letter also provided a simple definition of OCB and the objectives of the study.

Results
EFA was conducted, using maximum likelihood extraction and an oblique rotation, which resulted in a 2-factor solution that accounted for 71% of the variance, not the eight dimensional model derived by Rauf & Kumar (2015). Six items which had smaller loadings (<.40) were dropped from the model. While the initial model produced a chi-square goodness-of-fit index of 2580.80, df = 523, p<.01, the EFA with the final 32 items resulted in a chi-square of 1997.63, df = 494, p< .05. Table  1 lists the 38 items, their loadings and the theoretical OCB category that each item was initially intended to measure. Items loading greater than .40 were classified as representing a particular factor. As a result, 18 items loaded on Factor 1 and 14 on Factor 2 (see Table 1). The resulting model supported the substantive categories of OCBO and OCBI. Reliability for both factors were high, α OCBI = .96 and α OCB0 = .97.
To conclude, the present study supports a two factor model of OCB that fits with Williams & Anderson's (1991) OCBI-OCBO conceptualisation. The results also indicate that dimensions identified in Rauf & Kumar (2015) fall neatly into the theoretically consistent Williams & Anderson (1991) dimensions.
While the present study used exploratory methods to identify the underlying structure, literature suggests that a more robust method of examining factor structures would be to test a structure that is developed a-priori (Treiblmaier & Filzmoser, 2010). As CFA requires a larger sample size than available from Study 1 (see Bentler & Chou, 1988), the next study, using a larger sample, attempts to replicate the results of Study 1 and uses CFA, in addition to EFA, to test model fit.

Items
Hypothesised

Study 2
The previous study examined the factor structure of the OCB construct and revealed a two factor model, consistent with Williams & Anderson's (1991) dimensions of OCB directed towards the organization (OCBO) and OCB directed towards the individual (OCBI). This study further validates the factor structure of OCB through using EFA and then testing the model using Confirmatory Factor Analysis (CFA) on a new and bigger sample.

Method
The sample was of academic staff members drawn from three Sri Lankan state universities, which represent three groups of universities classified according to when they were established (50, 30 and 20 years ago). A stratified random sampling method, using strata based on gender, position and university, was used to select participants.  (Bentler & Chou, 1988). Due to incomplete data, 36 cases were discarded. A final set of 264 cases was used.

Results
First EFA revealed very similar loading from Study 2 as with Study 1 (see Table 1). Next, CFA was used to examine the fit of the two-factor model derived from Study 1. In confirmatory factor analysis, the covariance matrix is used and fit indices are calculated. Initially, the two factor model was specified using the 32 items and maximum likelihood techniques were used to estimate the model. For TLI, GFI and CFI, values greater than .95 constitute good fit and values greater than .90 acceptable fit (Medsker et al., 1994). For the RMSEA, values less than .05 constitute good fit, values between .05 and .08 are considered acceptable fit, values between .08 to .10 reflect marginal fit and values greater than .10 reflect poor fit (Browne & Cudeck, 1992). The initial analyses yielded low values of fit with a Chi-square of 2614.05 (df = 344), p < .001 and GFI=.72, CFI=.80, TLI=.87, RMSEA = .14, suggesting poor fit. Therefore, modification indices and residuals were examined and, as a result, four items were dropped and four sets of error variances were allowed to co-vary. This resulted in a significant improvement in model fit. The two factor model, measured by 28 items, resulted in a Chi-square of 1702.36 (df = 293), p < .001, and GFI=.91, CFI=.93, TLI=.91, RMSEA=.07 (see Figure 1 for final parameter estimates), suggesting acceptable fit.

DISCUSSION AND CONCLUSIONS
The present studies support the Williams & Anderson's (1991)  This paper follows up on Rauf & Kumar (2015) to investigate the dimensionality of OCB in university teachers in the Sri Lankan context. For this purpose, two studies were conducted. Study 1 used Exploratory Factor Analysis (EFA) and Study 2 used both exploratory (EFA) and Confirmatory Factor Analysis (CFA). Although the studies supported Williams & Anderson's (1991) two factor conceptualisation of OCB, and not Rauf & Kumar's (2015) eight dimensional model, the items designed to test the Rauf & Kumar (2015) dimensions fell neatly into OCBO or OCBI in a theoretically consistent manner. It is possible that factor analysis is not sensitive to subtle differences between dimensions that are captured through item sorting methods and that factor analysis captures a broader bandwidth (Ones et al., 1993). Alternatively, perhaps sorting methods are highly contextual and not easily captured through follow up research attempting to replicate these dimensions. In other words, the OCB structure may depend on the method of investigation, with nuances specific to the context captured through highly contextualised qualitative methods and broader dimensions captured through factor analysis.

With
respect to the question of whether conceptualisations are easily transferrable from location to location, the present study supports a body of literature on personality that suggests the affirmative (McCrae et al., 2005). The present research found that a simpler model of OCB was captured even in an instance where the items were highly contextualised. However, the study also suggests that the original context specific conceptualisation is meaningful as the items designed to measure those dimensions were incorporated into broader dimensions in a theoretically meaningful manner. Therefore, the question of which level of specificity is important in a given context becomes important. At times, greater specificity may be important so that the model is able to address the nuances of the context. At other times, such as when cross national comparisons are called for, broader dimensions seem appropriate (Kashima, 2016).
The theoretical physicist and Nobel Laureate, Werner Heisenberg, in highlighting how the methodology and questions adopted in research affect findings, stated "What we observe is not nature itself, but nature exposed to our method of questioning" (Heisenberg, 2000(Heisenberg, /1958. Similarly, regarding social phenomena, the findings suggest that perhaps the two approaches, emic and etic, have a propensity to elicit results that are consistent with the paradigm the research adopts. This researchers' results provide a possible means through which findings from emic and etic approaches to understanding the conceptualisation of constructs can be synthesised to arrive at the final goal of such research, which is to develop a more comprehensive understanding of a phenomenon that is able to integrate generalisations and specificities and is able to build on each other. Which approach is relevant would depend on the lens through which OCB is viewed. In these two studies, the authors used the term context to refer to Sri Lankan university teachers and as such 'context' may be interpreted to refer to the fact that the study was conducted in Sri Lanka, at Sri Lankan universities, or with academic staff at Sri Lankan universities. In the cross-cultural literature, it is not uncommon to study a specific subset of individuals to derive an understanding of the 'cultural' context, which is then frequently used as a proxy for those represented in the country. Hofstede's (1980) work is a classic example of such research. This study did not make such a generalisation as it was deemed unwarranted as the study simply explores a measure derived from an explicitly specific sample located within specific work and national contexts. The researchers did not approach this study assuming that the dimensionality found in this study would generalise to other broader contexts outside of academic staff at Sri Lankan universities. However, the fact that these findings suggest that a model found elsewhere fits this very specific context even when the items were designed to assess other dimensions, provides support to the robustness of Williams & Anderson's (1991) model.